43176319

CAPITAL MARKET THEORIES AND PRICING MODELS:
EVALUATION AND CONSOLIDATION OF THE

AVAILABLE BODY OF KNOWLEDGE
by
EUGENE RUDOLPH LAUBSCHER
submitted in fulfilment of the requirements

for the degree of
\
MASTER OF COMMERCE
in the subject
j
ACCOUNTING
at the
UNIVERSITY OF SOUTH AFRICA
SUPERVISOR: PROF E SAENGER

JOINT SUPERVISOR: PROF BL STEYN /
MAY 2001
ACKNOWLEDGEMENTS
Without the power of the Lord this study could not have been commenced, much less
completed. I would like to thank the following persons for their contribution to this study:
• Prof E Saenger for her guidance, patience, encouragement and interest shown.
It has been a privilege to be your student.
• Prof BL Steyn for his contribution to the study.
• Me YB van Stuyvenberg, librarian, for her assistance in accessing and obtaining

the necessary information.
• Me EM van Deventer for her assistance with the technical wordprocessing

aspects related to this study.
• Me ME Joubert for her editorial assiatance.
• My family and colleagues for their encouragement.
• My wife, Mariana, whose endless love, understanding and patience supported

me throughout this study.
SUMMARY
CAPITAL MARKET THEORIES AND PRICING MODELS:

EVALUATION AND CONSOLIDATION OF THE
AVAILABLE BODY OF KNOWLEDGE
by
Eugene Rudolph Laubscher
The study investigates whether the main capital market theories and pricing models pro-
vide a reasonably accurate description of the working and efficiency of capital markets,
of the pricing of shares and options and the effect the risk/return relationship has on in-
vestor behaviour. The capital market theories and pricing models included in the study
are Portfolio Theory, the Efficient Market Hypothesis (EMH), the Capital Asset Pricing
Model (CAPM), the Arbitrage Pricing Theory (APT), Options Theory and the Black-
Scholes (8-S) Option Pricing Model.
The main conclusion of the study is that the main capital market theories and pricing
models, as reviewed in the study, do provide a reasonably accurate description of
reality, but a number of anomalies and controversial issues still need to be resolved.
The main recommendation of the study is that research into these theories and models
should continue unabated, while the specific recommendations in a South African con-
text are the following: ( 1) the benefits of global diversification for South African investors
should continue to be investigated; (2) the level and degree of efficiency of the JSE Se-
curities Exchange SA (JSE) should continue to be monitored, and it should be esta-
blished whether alternative theories to the EMH provide complementary or better de-
scriptions of the efficiency of the South African market; (3) both the CAPM and the APT
should continue to be tested, both individually and jointly, in order to better understand
the pricing mechanism of, and risk/return relationship on the JSE; (4) much South
African research still needs to be conducted on the efficiency of the relatively new
options market and the application of the B-S Option Pricing Model under South African
conditions.
KEY TERMS:
Efficient Market Hypothesis (EMH); Portfolio Theory; Arbitrage Pricing Theory (APT);
Capital Asset Pricing Model (CAPM); Black-Scholes; Options; Diversification; Risk;
Return; Accounting Theory.
CONTENTS
Page
TABLE OF FORMULAE iv
TABLE OF FIGURES v
CHAPTER 1
THESCOPEANDPURPOSEOFTHESTUDY
1.1 Introduction 1
1.2 Background to the study 2
1.3 Definition of the problem 4
1.4 Purpose of the study 6
1.5 Motivation for and importance of the study 6
1.6 The research approach 7
1.7 Organisation of the study 9
1.8 Summary 11
CHAPTER 2
CAPiTAL MARKET THEORIES AND PRICING MODELS WITHIN THE CONTEXT
OF ACCOUNTING THEORY AND INVESTMENT DECISION-MAKING
2.1 Introduction 12
2.2 The concepts "capital market theories", "pricing models",
"accounting theory" and "investment decision-making" 14
11
2.3 The historical background to the development of capital

market theories, pricing models and investment decision-
making 17
2.4 The role of capital markets and pricing models in the formu-
lation of accounting theory 21
2.5 The need for and importance of capital market theories and
pricing models in investment decision-making 39
2.6 Summary 40
CHAPTER 3
THE DEVELOPMENT OF CAPITAL MARKET THEORIES
3.1 Introduction 43
3.2 The role of capital market theories in investment decision-
making 44
3.3 The development of Portfolio Theory 55
3.4 The development of the Efficient Market Hypothesis 75
3.5 Summary 116
CHAPTER4
THE DEVELOPMENT OF PRICING MODELS
4.1 Introduction 118

4.2 The development of the Capital Asset Pricing Model 120
4.3 The development of the Arbitrage Pricing Theory 184
4.4 The development of options theory and the Black-Scholes
Option Pricing Model 214
4.5 Summary 270
lll
CHAPTER 5
RECENT DEVELOPMENTS IN THE VARIOUS CAPITAL MARKET
THEORIES AND PRICING MODELS

5.2 Recent developments in Portfolio Theory 277
5.3 Recent developments in market efficiency and the Efficient
Market Hypothesis 281
5.4 Recent developments in the Capital Asset Pricing Model 343
5.5 Recent developments in the Arbitrage Pricing Theory 361
5.6 Recent developments in Options Theory and the Black-Scholes
Option Pricing Model 375
CHAPTER 6
SUMMARY, CONCLUSIONS AND RECOMMENDATIONS

6.2 Summary of the literature study 397
6.3 Conclusions 406
6.4 Recommendations 410
6.5 Summary 411
BIBLIOGRAPHY 412
lV
TABLE OF FORMULAE
Formula 3.1 Standard deviation of an investment 48
Formula 3.2 Actual return on an investment 49
Formula 3.3 Expected return on an investment 49
Formula 3.4 Standard deviation of a portfolio 63
Formula 3.5 Expected return on a portfolio 64
Formula 3.6 The lower confidence limit 68
Formula 3.7 Maximum utility of wealth 70
Formula 4.1 The Capital Asset Pricing Model 135
Formula 4.2 Alpha 136
Formula 4.3 The Arbitrage Pricing Theory factor-model 189
Formula 4.4 Put-call parity 224
Formula 4.5 The Black-Scholes Option Pricing Model 233

v
TABLE OF FIGURES
Figure 3.1 Standard deviation as a measure of risk 47
Figure 3.2 The relationship between risk and return 50
Figure 3.3 The relationship between income (wealth) and utility 54
Figure 3.4 Risk reduction through diversification 60
Figure 3.5 Efficient frontier 65
Figure 3.6 Risk-averse investors' utility indifference curves 69
Figure 3.7 Selection of the optimal portfolio 71
Figure 3.8 Reactions of capital markets to new information 79
Figure 4.1 The capital market line 131
Figure 4.2 The security market line 134
Figure 4.3 Boundaries of call option values 221

1
CHAPTER 1
THE SCOPE AND PURPOSE OF THE STUDY
1.1 Introduction 1
1.2 Background to the study 2
1.3 Definition of the problem 4
1.4 Purpose of the study 6
1.5 Motivation for and importance of the study 6
1.6 The research approach 7
1. 7 Organisation of the study 9
1.8 Summary 11
1.1 INTRODUCTION
The aim of this chapter is to discuss the scope and purpose of the study. The chapter
commences with an outline of the background to the study, whereafter the problem is
defined. The third aspect to be covered is the establishment of the main purpose and
supplementary purposes of the study, and this is followed by the motivation for and
importance of the study. The method of study is then outlined and the chapter ends with
a description of the organisation of the study.
2
1.2 BACKGROUND TO THE STUDY
The development of modern finance theory started in the 1950s and was refined during
the following decades resulting in a unified framework of finance theory during the
1980s. The rapid development of finance theory, especially the formation of theories
defin_ing the nature and working of capital markets, resulted in the establishment of
flexible pricing models which are widely applied in capital markets. The establishment
of options markets had a dramatic impact on finance theory. Research in the field of the
pricing of options provided additional important advances in the development of pricing
models.
The major advances in modern finance theory were the development of Portfolio
Theory, the Efficient Market Hypothesis, the Capital Asset Pricing Model, the Arbitrage
Pricing Theory and the Black-Scholes Option Pricing Model. With reference to Harring-
ton (1987:1-4), Van Horne (1992:3-5) and Milne (1995:4-8), the development of the
major capital market theories and pricing models are summarised below.
The development of Portfolio Theory, based on the work of Markowitz (1952, 1959),
revolutionised finance theory and was the first step in the development of pricing
models. The basic idea behind Portfolio Theory is that the risk of an investment should
be considered on the basis of its contribution to the overall risk of a portfolio of invest-
ments and not according to deviations in the individual investment's expected return.
Therefore the efficient diversification of investment portfolios will result in a reduction
of the level of risk associated with a given expected return.
The development of the Efficient Market Hypothesis (EMH), based primarily on the
work and findings of Fama (1970), had a major impact on how the workings of the capi-
tal market were viewed. The basic suggestion of the EMH is that, since the capital mar-
ket is efficient, the changes in the prices of securities are uncorrelated and the security
prices fully reflect the price implications of all publicly available information in an un-
biased manner.
3
A further development was the extension of Markowitz's (1952, 1959) work by Sharpe
(1964), assisted by the efforts of Treynor (1961 ), Lintner (1965) and Mossin (1966), re-
sulting in the now well-known Capital Asset Pricing Model (CAPM). The main principle
of the CAPM is that some of the risk underlying an investment can be diversified away
by holding a portfolio of investments. Further, the expected return on an investment is
equal to the sum of the risk-free rate of return and a risk premium derived from the
investment's sensitivity to the risk of the market as a whole. The development of the
CAPM had a major impact on the selection and the measurement of the performance
of individual investments and portfolios of investments.
A further important development was Black and Scholes's (1973) contribution of an

option pricing model. Known as the Black-Scholes (B-5) Option Pricing Model, this
model is based on the same basic principles as the CAPM. The development of the B-S
model expanded the use of pricing models since it can also be applied with other
financial instruments such as bonds, foreign currency, etcetera.
The development of the Arbitrage Pricing Theory by Ross ( 1976) as a rival/alternative

to the CAPM, further contributed to the development of finance theory. The APT differs
from the CAPM in that it does not only take the market risk factor into account, but also
relies on a number of, yet undetermined, risk factors. The development of the APT has
created a great deal of interest, and even controversy, in the field of asset pricing,
insofar as to determine all the risk factors to be taken into account in the pricing of
assets, and whether or not it is more flexible and useful than the CAPM.
Finance theory is continuing to develop through new ideas, techniques and research,
and all the major capital market theories and pricing models are examined by acade-
mics, researchers and investment practitioners on an ongoing basis. The controversy
and debate surrounding these theories and models, especially with regard to the EMH,
CAPM, APT and the B-S Option Pricing Model, continue unabated.
4
1.3 DEFINITION OF THE PROBLEM
As indicated in the preceding section, the main capital market theories and pricing
models have been developed and refined during the past fifty years. Since their
development, the research and debate over these theories and pricing models have
been ongoing with the result that today there is an enormous volume of research and
academic literature on these subjects. The volume of research conducted to date and
the extensive literature available, show that there is a definite need to attempt to
consolidate the available body of knowledge before this task becomes too voluminous.
The consolidation of the available knowledge and a review of the extensive body of
research will enhance knowledge of the subject in South Africa and will serve as a basis
for future South African research on the issues concerned. It should, however, be noted
that the usefulness of such a consolidation of knowledge is not only limited to South
Africa, but may also be useful on an international scale. Most of the research under-
taken to date and most of the debate surrounding the various subjects, in the broad
context of capital market theory, have taken place on an international scale outside the
borders of South Africa.
This gives rise to the definition of the problem of this study:
Does the available body of knowledge and evidence from em-

pirical research on the main capital market theories and pricing
models show that these theories and models provide a reasonably
accurate description of reality?
5
The importance of such a consolidation of the available body of knowledge and a review
of research is confirmed by Griffen (1987:4,5) who is of the opinion that such a merging
of research can serve a broad educational objective in that
it filters and translates a large body of work written mostly by aca-

demics to satisfy scholarly endeavours. Doubtless, that scholarly focus
has merit, but for the non-academics many points of broader interest
are lost because the reader has neither the means nor the inclination
to decipher the researcher's technical language. Periodically research
needs to be summarized and coherently organized to enhance and
broaden people's understanding, particularly those who have an en-
during interest in financial reporting. An organized and coherent trans-
lation helps researchers and readers alike.
The sub-problems of the study can be summarised as follows:
• How do the existing capital market theories and pricing models describe and
explain the nature and working of the capital market?
• Has research shown that these theories and models are valid and that they have
descriptive and predictive ability?
• Have recent studies and research led to new developments of importance in the
broad context of capital market theory?
• Can the recent research and developments be summarised and consolidated to

provide a framework for future research?
• Where should future research, especially under South African conditions, be

directed in this field of study?
This study will attempt to consolidate the available subject-related knowledge and pro-
vide a comprehensive review of the subject-related research. Within this framework, the
study will also attempt to provide answers to the sub-problems.
6
1.4 PURPOSE OF THE STUDY
With the problem areas as specified in the preceding section in mind, the purpose of
this study is to review the development of the major capital market theories and pricing
models within the context of accounting theory and investment decision-making, investi-
gate and analyse the research done and identify new developments in these theories
and pricing models. This would then be used as a basis to determine whether these
theories and pricing models provide an accurate description of reality and to identify
areas of possible future research.
1.5 MOTIVATION FOR AND IMPORTANCE OF THE STUDY
The large volume of literature and research on the various capital market theories and
pricing models, identified in sections 1.2 and 1.3 above, revolutionised the area of fi-
nance theory. When reviewing the literature and research, the importance of these
theories and pricing models is emphasised. Evidence of this can, for example, be found
in the works of Myers (1973:49), Clark et al. (1979:181 ), Seneque (1987:28) and Viljoen
(1996:40). This point of view is also clearly illustrated by the following references:
Two of the most exciting developments in microfinance and capital

markets have been the efficient market hypothesis and the capital
asset pricing model (Bicksler 1977:xiii).
The pricing of capital assets has been the subject of intensive research
by scholars and practitioners over the last thirty years (Gourley et al.
1987:219).
Arbitrage pricing theory and Black and Scholes (1973) option pricing
theory are two of the most recent possibilities suggested for use in
describing investor behaviour (Harrington 1987:94).
Notwithstanding the fact that these theories and models were developed a number of
decades ago, much debate and even controversy still surround them and the research
continues unabated. Evidence of this can, for example, be found in the work of
7
Casabona (1986:iii), Blume (1993:8), Karnosky (1993:56) and Ross (1993: 11 ).
The following references elucidate this point of view:
Development of portfolio theory ... an empirically testable capital asset

pricing model (CAPM), along with the efficient market hypothesis, has
generated an enormous volume of accounting research (Spoede
1982:1 ).
The now generally recognized controversy over the efficient market

hypothesis (EMH) ... must be seen as being existent in tandem with the
controversy over the validity of the CAPM (Seneque 1987:30).
Arbitrage pricing theory (APT) was first developed in 1976 ... However,
as with the CAPM, there are major difficulties with testing its validity
which have not yet been overcome (Allison 1991: 121 ).
The motivation for the review and consolidation of the available knowledge is that it pro-
vides a proper understanding of the subject of capital market theory and pricing models
and serves as background information for the analysis of recent developments in the
latter part of the study.
This study differs from past research in that it does not concentrate on one or two areas
of the subject, but attempts to incorporate all the related aspects of the subject into one
logical unit. Such an up-to-date consolidation of information and analysis of recent
developments have, to the researcher's knowledge, not as yet been done The Human
Sciences Research Council (HSRC) has confirmed that to date this has not been done
in South Africa.
1.6 THE RESEARCH APPROACH
The study will consist mainly of a literature study, that is, a study of the theoretical
aspects and published empirical tests of the major capital market theories and pricing
models, within the context of accounting theory. The topic of the study and its related
aspects are covered in a wide range of literature sources, including books, professional
8
journals, theses, dissertations and technical reports.
The topic of the study will be investigated as follows:
• A literature study to review the development and historical research of accoun-

ting theory, capital market theories and pricing models.
• A literature study to review and analyse the recent developments.
• A summary of the current knowledge of all the major capital market theories and
pricing models investigated, in an attempt to identify areas of possible future
research, especially in a South African context.
The main reason for making use of a literature study only is that in an attempt to
establish current knowledge and to define possible areas for future research, it is
necessary to review the development and testing of the capital market theories and
pricing models and to evaluate recent research and developments in order to determine
the future direction of the theoretical developments and empirical research in this area.
The literature study will be conducted as follows:
• Intensive consultation and assimilation of the existing literature and empirical

evidence.
• Critical evaluation of the literature and empirical evidence.
• A summary of the literature and empirical evidence leading to conclusions and

recommendations that contribute to the field of study.
9
Saenger (1991 :6,7), however, notes a number of limitations in such a literature study
and review of empirical research. These limitations can be summarised as follows:
• There may be publication bias, in that certain issues, events, circumstances,

etcetera receive greater recognition and more attention than others.
• Important research may have never been published, may have gone out of print
or may have been edited in such a manner that it detracts from its value.
• The retrieval of information may be in accordance with preconceived ideas.
• The researcher may show preference for those authors who supports his/her
point of view.
• The researcher's communication skills may be limited.
In addition to these limitations, the availability of literature sources, the language barrier
and the researcher's judgement of the importance and relevance of the research per-
formed contribute to the limitations of a literature study and review of empirical research
1.7 ORGANISATION OF THE STUDY
The study consists of six chapters:
• Chapter 1 describes the scope and purpose of the study. The chapter starts with
the background to study, followed by the definition of the problem, the purpose
of the study and the motivation for and importance of the study. Thereafter the
research approach is discussed.
• Chapter 2 focuses on the role of the capital market theories and pricing models
in the formulation of an accounting theory (the market-based approach to the
development of an accounting theory) and their role in investment decision-
making. The chapter starts with an explanation of the concepts "capital market
theories", "accounting theory" and "investment decision-making". This is followed
by a review of the historical background to the development of the capital market
10
theories, pricing models and investment decision-making, the role of the theories
and models in the formulation of an accounting theory and the importance
thereof for investment decision-making.
• Chapter 3 concentrates on the development of the main capital market theories

and commences with an investigation of the role of the capital market theories
in investment decision-making. Of special importance in this section is the rela-
tionship between risk and return and their effect on investment decisions. There-
after the development of the two major capital market theories is investigated.
Firstly, the nature of Portfolio Theory and its implications for the risk/return rela-
tionship are examined. An important aspect of the chapter then follows, that is,
the review of the empirical research related to Portfolio Theory up to the early
1980s. From this review the conclusions about Portfolio Theory are summarised.
Secondly, the development of the Efficient Market Hypothesis and the forms and
degrees of market efficiency are examined. This is followed by an examination
of the tests of market efficiency, the efficient market anomalies, the implications
of the Efficient Market Hypothesis and the misconceptions about market effi-
ciency. The final section of the examination of the Efficient Market Hypothesis
reviews the empirical research, and the conclusions drawn from it, related to
capital market efficiency up to the early 1980s.
• Chapter 4 consists of an examination of the development of the three major

pricing models, that is, the Capital Asset Pricing Model, the Arbitrage Pricing
Theory and the Black-Scholes Option Pricing Model and an investigation of
Options Theory. The chapter starts with the Capital Asset Pricing Model, that is,
its nature, assumptions, parameters, how it describes risk and its application in
investment decision-making. The final sections on the Capital Asset Pricing
Model deal with the testing of the validity of the model, a review of the empirical
research related to it, up to the early 1980s, and the conclusions drawn from it.
Next the Arbitrage Pricing Theory is examined, including its nature, how it
describes risk, its application in investment decision-making and how it compares
11
with the Capital Asset Pricing Model. The final part of this section considers the
tests of the validity of the model, reviews the empirical research related to the
model up to the early 1980s and draws conclusions from it. Finally, options
markets, the nature, types and value of options and the application of the Black-
Scholes model in the pricing of options are examined. This is followed by a
review of the tests on and results of empirical research related to option markets
and the Black-Scholes model up to the early 1980s, ending with the conclusions
that can be drawn from it.
• Chapter 5 reviews recent developments in and recent research related to the

capital market theories and pricing models examined in chapters 3 and 4. Inclu-
ded in this is some alternative theories to the Efficient Market Hypothesis and the
possible role of options in the stock market crash of October 1987. The. chapter
also draws conclusions from the recent empirical research conducted on each
of the capital market theories and pricing models.
• Chapter 6 provides a summary of the literature study, that is, within the context
of accounting theory, the nature and implications of investment decision-making
and the principles of Portfolio Theory, the Efficient Market Hypothesis, the Capi-
tal Asset Pricing Model, the Arbitrage Pricing Theory, Options Theory and the
Black-Scholes Option Pricing Model. This is followed by the main conclusions
that can be drawn from the empirical research on the above. The chapter ends
with recommendations for future research in a South African context.
1.8 SUMMARY
The aim of this chapter was to provide information on the scope and purpose of the
study. The chapter started with the background to the study, followed by a definition of
the problem of the study and a discussion of the purpose, motivation for and importance
of the study. The last two sections of the chapter dealt with the research approach that
is going to be adopted and the organisation of the study.
12
CHAPTER2
CAPITAL MARKET THEORIES AND PRICING MODELS WITHIN THE

CONTEXT OF ACCOUNTING THEORY AND INVESTMENT DECISION-
MAKING
2.1 Introduction 12
2.2 The concepts "capital market theories", "pricing models", "accounting

theory" and "investment decision-making" 14
2.3 The historical background to the development of capital market theories,

pricing models and investment decision-making 17
2.4 The role of capital markets and pricing models in the formulation of
accounting theory 21
2.5 The need for and importance of capital market theories and pricing
models in investment decision-making 39
2.6 Summary 40
2.1 INTRODUCTION
According to Weston and Copeland ( 1992: 193) capital markets, to be able to efficiently
allocate scarce capital resources, should price securities only on economic grounds
based on publicly available information.
13
Accounting information, especially as contained in interim and annual financial reports,

is an important source of information available to capital markets since it provides infor-
mation about the financial performance and condition of business entities. As such, and
to be useful for capital market investment decision-making, accounting information
should be both relevant and reliable.
The importance of the usefulness of accounting information is emphasised by the

American Institute of Certified Public Accountants (Belkaoui 1992:23) in its definition
of accounting:
Accounting is a service activity. Its function is to provide quantitative

information, primarily financial in its nature about economic entities that
is intended to be useful in making economic decisions, in making re-
solved choices among alternative courses of action.
Thus, by implication, the value of accounting information can be established by mea-

suring the impact of new publicly available information on security prices on capital
markets. Capital market theories, combined with the application of pricing models,
provide a basis according to which various accounting policies, disclosure practices and
accounting methods can be evaluated through measuring the impact of alternative
accounting policies, practices and methods on security prices (Firth 1977: 140).
The main focus of this chapter is, therefore, on the role of the market-based approach
to the formulation of accounting theory, since accounting information is an important aid
to capital markets in its function of efficient allocation of capital resources and it is an
essential investment tool which enables investors in capital markets to hold optimal
investment portfolios which reflect their risk-return preferences. This chapter commen-
ces with descriptions of the concepts "capital market theories", "pricing models",
"accounting theory" and "investment decision-making" to provide a frame of reference
for the ensuing sections. Following on this is a review of the background to the develop-
ment of those capital market theories and pricing models which, apart from being essen-
tial elements in the market-based approach to the formulation of accounting theory, are
also the main focus of this study.
14
It should be noted as this stage that the market-based approach to the formulation of
accounting theory is but one of many such approaches, hence the various approaches
to the formulation of accounting theory are summarised in the next section with specific
emphasis on the market-based approach.
Jn addition, and of particular importance to this study, the need for and importance of
capital market theories and pricing models in investment decision-making are dis-
cussed, with specific reference to the role of accounting information.
2.2 THE CONCEPTS "CAPITAL MARKETS THEORIES", "PRICING

MODELS", "ACCOUNTING THEORY" AND "INVESTMENT DE-
CISION-MAKING"
According to Kam (1990:45,46,485,486) theories, hypotheses and models can be

viewed as synonyms in their role of explaining and predicting what happens in reality.
The American Accounting Association (AAA) (Wolk et al. 1989: 155) defines theory as
" ... a cohesive set of hypothetical, conceptual and pragmatic principles forming a
general frame of reference for a field of study".
Rees (1995:38) describes the role of theory as follows:
In understanding the world of accounting and finance, there is a role for

descriptions of current practice. There is also a need for informed pre-
scriptions as to how to improve that practice.
Theoretical development requires a postulated model which either

describes why things work as they do rather than simply how, or what
reaction can be expected from a given set of circumstances. Put
simply, it either explains or predicts.
15
The main purpose of theory is seen by Hendriksen and Van Breda ( 1992: 16) as:
To provide a framework for the development of new ideas and new

.procedures and to help in making choices among alternative proce-
dures. If these objectives are met, it is not necessary that theory be
based completely on practical concepts or that it be restricted to the
development of procedures that are completely workable and practical
in terms of current technology.
Although no definitive definitions exist for the concepts "capital market theories", "pri-
cing models", "accounting theory" and "investment decision-making" the following
sections attempt to provide reasonable descriptions of these concepts.
2.2.1 The concept "capital market theories"
Capital market theories are concerned with explaining and predicting the relationship
between expected return and risk on investments in capital markets, the effect of
investor's efficiently diversified portfolios on the market pricing mechanism and whether
the market is able to ensure that security prices fully and correctly reflect all available
information (Fama 1976:136,258,320).
2.2.2 The concept "pricing models"
The purpose of capital market pricing models are to accurately predict the fair value of
securities based on estimates of the expected returns on risky securities in market
equilibrium. These models therefore attempt to explain the relationship between risk,
return and value based on the behaviour of efficient capital markets (Weston & Cope-
land 1992:401,436; Rees 1995: 153).
16
2.2.3 The concept "accounting theory"
Using the various definitions of accounting theory (Wolk et al. 1989:3; Kam 1990:490;
Belkaoui 1992: 178; Hendriksen & Van Breda 1992:21 }, accounting theory can be de-
scribed in the following manner.
Accounting theory constitutes a coherent set of concepts, models and hypotheses on

which the establishment of accounting techniques are based. These serve as a frame
of reference for the establishment of accounting concepts and principles in the stan-
dard-setting process, with the ultimate aim of providing logical, useful and verifiable
financial information.
At this point in time no comprehensive theory of accounting exists. Different theories

have been and are continually proposed depending on the approach followed (Belkaoui
1992:56).
2.2.4 The concept "investment decision-making"
With specific reference to capital markets, investment decision-making refers to the pro-
cess of using information to assess the performance and prospects of firms with the aim
of investing in or divesting of its securities. Information is essential in the decision-ma-
king process to ensure that investment in securities provides an adequate return to
compensate for the risk involved. The essential element of investment decision-making
is to ensure that the expected return on the investment justifies the level of risk involved
(Dobbins et al. 1994:5; Rees 1995:ix, 157; Scott 1997:4).
With these descriptions of relevant concepts as point of departure, the historical de-
velopment of capital market theories, pricing models and investment decision-making
are now reviewed.
17
2.3 THE HISTORICAL BACKGROUND TO THE DEVELOPMENT OF

CAPITAL MARKET THEORIES, PRICING MODELS AND INVEST-
MENT DECISION-MAKING
Prior to the 1950s the traditional methods for capital market investment decisions and
investment portfolio management were rule-of-thumb, the intrinsic value of securities
(fundamental analysis) and trends or patterns in security prices (technical analysis)
(Anderson 1978:6, 10). Investors and investment managers concentrated on the return
on investments and the role of risk was not properly understood, if considered at all
(Harrington 1987:5; Dobbins et al. 1994: 1,2).
Investors used the rule-of-thumb method as a basis to manage their portfolios, while
they used the intrinsic values of securities and trends or patterns in security prices as
a basis for selection of securities to buy, hold or sell. The two basic methods used for
security selection were fundamental analysis and technical analysis (Anderson 1978: 7).
In using fundamental analysis, investors studied the firm, industry and economy to
invest in securities when their market prices were below intrinsic value and to sell
securities when market prices rose above their intrinsic values. In contrast to this,
technical analysis does not concentrate on the factors influencing individual securities
and the factors reflected in market prices, but focuses purely on price trends or patterns
(Anderson 1978:9).
Investors, using technical analysis, based their investment decisions on observable

price patterns that repeated themselves over time. These patterns were used to deter-
mine whether capital markets underpriced or overpriced securities. Investors, therefore,
maintained or increased the value of their investment portfolios by charting the past
prices of securities in an effort to identify security price trends or patterns that might be
used as signals to buy, hold or sell securities (Anderson 1978: 10).
18
Since the 1950s academics, theorists and researchers started questioning and testing
the traditional assumptions concerning the pricing of securities on capital markets. This
lead to an unification of theory concerning capital markets, business finance and invest-
ment decision-making. New developments, such as Portfolio Theory (PT), the Efficient
Market Hypothesis (EMH), the Capital Asset Pricing Model (CAPM), option pricing and
the Arbitrage Pricing Theory (APT) concentrated not only on return on investment but
rather on the use of information by investors to form expectations about the risk and
return on securities (Anderson 1978:6, 12, 13).
Modern portfolio theory, based on the work of Markowitz (1952, 1959), concentrates as
much on investment risk as on investment return. Further, investment in securities is
evaluated in terms of the security's effect on the investor's entire portfolio of investments
(Harrington 1987:5,6; Dobbins et al. 1994:2).
Fundamental to modern portfolio theory is the postulate that, under conditions of uncer-
tainty, investment in a security should not by judged purely on the basis of its rela-
tionship between risk and return, but rather in relation to its contribution on the overall
risk of the portfolio of investments (Van Horne 1992:4 ). Thus, depending on the in-
vestor's desired level of risk, investors construct portfolios of investments that render
the highest return for a specific level of risk (Harrington 1987:9-11; Linley 1992:3).
In contrast with modern portfolio theory, the traditional investment theory evaluated
investment in securities under the assumption of perfect certainty without any clear or
defined approach in assessing the effects of risk. Based on the assumption of perfect
certainty about future returns, investors were urged to hold one-security portfolios which
would render the highest return for a specific investment period (Anderson 1978:33;
Linley 1992:8).
Following on Markowitz's work on the relationship between risk and return and the
efficient diversification of investments, Sharpe (1964), Lintner (1965) and Mossin (1966)
developed the Capital Asset Pricing Model (CAPM). The CAPM describes the rela-
19
tionship between risk and return in efficient capital markets based on the postulate that
the prices of securities in capital markets will adjust until securities with the same level
of risk offer the same expected return (Anderson 1978:70; Sharpe 1985:xix).
Weston and Copeland (1992:403) summarise the significance of the CAPM as follows:
It provides a measure of the risk of an individual security which is con-

sistent with portfolio theory. It enables us to estimate the undiversifiable
risk of a single asset and compare it with the undiversifiable risk of a
well-diversified portfolio.
After the development of the CAPM, academics, theorists and researchers shifted their
focus to the idea of efficient capital markets and the performance of the pricing mecha-
nism of capital markets (Firth 1977:14).
At first the focus was on whether successive price changes in capital markets were un-
correlated. Later on, after the evidence supported a random walk for price changes, the
focus shifted to how the market priced securities to achieve a random walk (Anderson
1978:13,14).
This led Fama (1970), building on the earlier work of Samuelson ( 1965), to introduce
the Efficient Market Hypothesis (EMH) (Milne 1995:8). The EMH asserts that security
price changes are uncorrelated as capital markets are efficient and rapidly incorporates
all relevant new information into security prices in an unbiased manner (Anderson
1978: 14; Hendriksen & Van Breda 1992: 169).
The EMH is the prevailing description of the effect of publicly available information on
security prices and argues that no abnormal returns can be made on capital markets by
using publicly available information. Abnormal returns can only be obtained through
change or by using insider information (Milne 1995:8).
20
With the opening of the first, and still the largest, options exchange in Chicago in 1973
came another important development in modern investment theory, namely option
pricing (Anderson 1978:7). Two different but equivalent pricing models for determining
fair market values of options are the Black-Scholes Model, developed by Black and
Scholes in 1973, and the binomial model developed by Cox, Ross and Rubinstein in
1979 (Weston & Copeland 1992:446). The binomial model, however, falls outside the
scope of this study.
The Black-Scholes Model, which is quite similar to the CAPM, is used to price call
options in efficient markets, but can also be used to determine put option prices, inclu-
ding the more difficult American put option prices (Weston & Copeland 1992:446; Milne
1995:7).
Finally, due to dissatisfaction with some theoretical and empirical weaknesses of the
CAPM, the APT was developed by Ross in 1976. Unlike the CAPM, which deems the
capital market as the only source of risk (Linley 1992: 19), the APT uses a number of
factors, for example unexpected changes in interest rates, to explain returns on securi-
ties (Weston & Copeland 1992:422).
Although Ross did not specify the various factors which may affect security returns nor
provided instructions how these factors should be measured (Linley 1992:3), it is poten-
tially a more flexible and meaningful alternative to the CAPM, with less empirical testing
limitations (Milne 1995:6).
Due to the recognition of the relationship between risk and return and the explanation
of the effect of individual securities on investors' portfolios, these modern capital market
theories and pricing models have proved popular since they provide a coherent frame-
work for capital market investment and investment decision-making (Harrington 1987:5;
Milne 1995: 11 ).
21
Since the 1980s the focus of investment theory has shifted to the effect of efficient
market anomalies and imperfections in the pricing of securities as well as the effect of
the globalization of capital markets on investment decision-making (Van Horne 1992:5).
Following on the historical background to the development of capital market theories,

pricing models and investment decision-making, the formulation of accounting theory
and the role of capital markets and pricing models therein can now be considered.
2.4 THE ROLE OF CAPITAL MARKETS AND PRICING MODELS IN

THE FORMULATION OF ACCOUNTING THEORY
The various approaches to the formulation of accounting theory can be labelled as

either descriptive or prescriptive. According to Belkaoui (1992:58, 109) the various
approaches can be classified in the following manner.
• Traditional approaches
• Non-theoretical approaches
... Pragmatic (informal) approach
... Authoritarian approach
• Theoretical approaches
... Deductive approach
... Inductive approach
... Ethical approach
... Sociological approach
... Economic approach
• Modern approaches
• Regulatory approach
• Events approach
• Behavioural approach
• Human information processing approach
• Positive approach
22
• Predictive approach
... Prediction of an economic event approach
... Prediction of market reaction: market-based approach
In the following sections the methodologies used in these approaches and the various
approaches themselves are summarised, with particular emphasis on the market-based
approach in its role in the formulation of accounting theory and its relevance to the
subject of the study.
2.4.1 Methodologies for the formulation of accounting theory
Accounting theories can be classified as either being a descriptive theory of accounting

or a prescriptive theory of accounting, depending on the methodology used. Hendriksen
and Van Breda (1992:17) summarises these two methodologies as follows.
a. Descriptive theories
Descriptive (positive) accounting theories attempt to describe the way things are at
present, that is, they attempt to explain what financial information is made available to
users of accounting data and how this information is communicated.
b. Prescriptive theories
Prescriptive (normative) accounting theories attempt to prescribe the way things should
be rather than what is, that is, they attempt to explain what financial information ought
to be made available and how this information ought to be communicated.
Belkaoui (1992:57) confirms the need for both methodologies in the formulation of
accounting theory. Descriptive theory is needed to justify existing accounting practices,
while prescriptive theory is needed to justify the improvement of existing accounting
practices or the adoption of new accounting practices.
23
Following from this, the various approaches to the formulation of accounting theory are
examined.
2.4.2 Approaches to the formulation of accounting theory
This section provides a brief discussion of the various approaches to the formulation of
accounting theory with specific reference to the role of the market-based approach.
2.4.2.1 Non-theoretical approaches
a. Pragmatic (informal) approach

The pragmatic approach to the formulation of accounting theory examines accounting
techniques and principles from the viewpoint of their. practical usefulness. Accordingly,
the usefulness to users of accounting information and the relevance to decision-making
is the main criteria for the selection of accounting techniques and principles (Belkaoui
1992:58).
Advocates of the usefulness criterion and therefore the pragmatic approach include
Fremgen (1967), Beams (1968), Mueller (1967) and Cowan (1968). Critics of this ap-
proach, including Skinner (1972), attack the usefulness criterion since this leads to con-
flicting accounting practices and principles due to the different information needs of the
various decision-makers and users of accounting data (Saenger 1991 :34).
b. Authoritarian approach
The authoritarian approach also seeks to provide practical solutions to accounting pro-
blems and is employed mostly by professional accounting bodies through their attempts
to regulate accounting practices. This approach emphasises the viewpoint that accoun-
ting theory must be practical to be useful, and that this is the basis for accounting
having any function at all (Saenger 1991 :35; Belkaoui 1992:59).
24
Neither the pragmatic nor the authoritarian approach has resulted in the formulation of
a single unified accounting theory. This does not mean that non-theoretical approaches
can be ignored because practical considerations will always be a part of the field of
accounting (Saenger 1991 :35; Belkaoui 1992:59).
2.4.2.2 Theoretical approaches
a. Deductive approach
The deductive approach to the formulation of accounting theory emphasises the
importance of objectives in the development of accounting practices and principles,
since different objectives may require different principles (Hendriksen & Van Breda
1992: 16). The deductive approach, therefore, starts with the basic objectives, postulates
and premises of accounting and through logical inference formulates accounting rules
and practices. Through empirical application of these logically derived conclusions the
theory is then tested to ensure it meets the demands of practice (Belkaoui 1992:60).
Critics of the deductive approach claim that it is too far removed from reality to provide
the basic accounting rules and practices, and it has the disadvantage that if any of the
basic objectives, postulates and premises of accounting are false then the resulting
theory may also be false (Hendriksen & Van Breda 1992: 16).
The deductive approach is usually prescriptive (normative) and has a number of advo-
cates, inciuding Paton (1922), Sweeney (1936), Alexander (1950), Edwards and Bell
(1961), and Sprouse and Moonitz (1962) (Saenger 1991:36; Belkaoui 1992:60).
b. Inductive approach
The inductive approach is primarily descriptive (positive) and is based on drawing gene-
ralised conclusions from specific observations and measurements. This approach con-
structs new accounting practices and principles based on observations of real financial
data of firms when there is sufficient recurrence of relationships between the observed
phenomena (Belkaoui 1992:61; Hendriksen & Van Breda 1992:16).
25
Unlike the deductive approach, the validity of inductive theories rely on empirical ve-
rification (Belkaoui 1992:61 ). The process of induction starts with the formulation of an
hypothesis, based on a possible relationship between observed phenomena, empirical
testing of the hypothesis leading to conclusions regarding confirmation of the hypo-
thesis as being true, or abandonment of the hypothesis as having failed. A confirmed
hypothesis constitutes a theory (Henderson et al. 1992:5).
The inductive approach has a distinct advantage in that researchers are not restricted
by existing practices and principles, but are allowed to make and test any observations
they may consider relevant (Hendriksen & Van Breda 1992: 17).
Proponents of the inductive approach include Hatfield (1927), Gilman (1939), Littleton
(1953) and ljiri (1975) (Saenger 1991 :37).
The main disadvantage of the inductive process is that researchers may be influenced
by preconceived ideas regarding the data to be observed and the relationships to per-
ceive. Further, observations of accounting data are likely to differ from firm to firm, and
this results in difficulties in determining relationships that can lead to useful generali-
sations (Hendriksen & Van Breda 1992: 17).
It is generally considered that the deductive and inductive approaches are not competi-
tive approaches, but can be used together in a complementary manner. Although some
theorists hold the viewpoint that one approach may dominate the other, for example
Kaplan ( 1972) favours induction and Hakansson ( 1969) favours deduction, it can be
concluded that almost all theories will be based on both deductive and inductive rea-
soning (Wolk et al. 1989:35-38; Saenger 1991 :38).
c. Ethical approach
The concepts justice, fairness, truth and equity have been identified by Scott ( 1941) as
the main criteria in the ethical approach for the formulation of an accounting theory.
Other proponents of the ethical approach have rejected some of the concepts, for
26
example Yu (1976) perceived only fairness and justice as ethical norms, while Patillo
(1965) and Spacek (1962) consider only fairness to be a basic standard against which
all accounting principles and practices should be measured (Saenger 1991 :39; Belkaoui
1992:62).
Jn conclusion, the ethical approach should not be seen in isolation since, according to
Belkaoui (1992:63), fairness is a desirable objective in any approach to the formulation
of an accounting theory. The ethical approach, as such, cannot produce sound accoun-
ting practices and principles, but, due to the subjective nature of what can be deemed
as fair and what not, it does mean that other approaches should not ignore the basic
ethical principles (Saenger 1991 :39).
d. Sociological approach
The sociologi_cal approach to the formulation of accounting theory emphasises the wel-
fare of society as a whole, and that this should form the basis for evaluating accounting
principles and practices (Saenger 1991 :39,40). Socioeconomic issues, such as the
costs of environmental pollution and unhealthy working conditions, need to be taken
into account when accounting reporting principles and techniques are evaluated and
established (Hendriksen & Van Breda 1992: 10).
Among the proponents of the social role of accounting are Littleton and Zimmerman
(1962) and Ladd (1963), and they stressed that accounting should serve the public
interest (Belkaoui 1992:63).
Due to increased interest in the impact of the activities of firms on the social environ-
ment, it can be expected that the sociological approach will play a major role in the
future formulation of accounting theory (Belkaoui 1992:63). Devine (1960) and Rappa-
port (1964), in particular, consider the establishment of a statement of socially orien-
tated accounting objectives as fundamental to the development of a general accounting
theory (Saenger 1991 :40). ·
27
e. Economic approach
According to the economic approach, the choice between alternative accounting tech-
niques depends on the particular economic situation and their impact on the national
economy (Belkaoui 1992:64). The economic approach can be divided between two
different approaches, namely the macro-economic and the micro-economic approach
(Hendriksen & Van Breda 1992:9).
The macro-economic approach concentrates on explaining the effect of various re-

porting alternatives on economic activities and economic measurement at an industry
or national level. Sweden is the most typical example of a country that bases its
accounting policies and practices on macro-economic goals. The micro-economic
approach is concerned with explaining the effect of various reporting alternatives on
economic activities and economic measurement at the level of the firm. Advocates of
this approach argue that, since financial information has inevitable economic conse-
quences, it does not matter how financial information is disclosed as long as it is dis-
closed in full.
The formulation of accounting theory and the development of accounting practices and
principles have in general been in the form of a combination of approaches rather than
a single approach. This has resulted from the combined and sometimes competitive
attempts by accounting theorists, professional accounting bodies and governmental
organisations to establish the concepts, principles and practices of accounting (Bel-
kaoui 1992:64,65).
This has given rise to new (modern) approaches to the formulation of accounting theory
based on both empirical and conceptional reasoning. These new approaches are more
reliant on empirical verification than the traditional approaches as method of resolving
accounting issues and problems. Each of these new approaches are innovative and
unique in their way of looking at accounting problems, each with their own metho-
dologies and interests, " ... causing accounting to become a multi paradigmatic science
in a constant state of crisis" (Belkaoui 1992:109, 162, 163).
28
2.4.2.3 Modern approaches
a. Regulatory approach
The regulatory approach is concerned with the establishment and enforcement of
accounting standards for financial reporting (Belkaoui 1992:7_4). Although these
standards are not accounting theories, there have been attempts in recent years to
provide the standards with a formalised theoretical underpinning - known as a con-
ceptual framework. This theoretical foundation is important since it is mainly the quality,
and improvement in quality, of reported information that is being regulated (Saenger
1991 :43).
Although there is debate whether, or to what extent, accounting should be regulated,

the setting of standards is a reality in the field of accounting. The process of setting
standards can be divided into three approaches (Belkaoui 1992:74,85).
i. The free-market approach. The market of supply and demand for information is
considered to be the best mechanism for determining the types of information to be
disclosed, to whom it should be disclosed and the accounting standards which the
information should conform to. Leftwich ( 1980) and Kripke ( 1980) support this approach
and conclude that without regulation market pressures would be sufficient to ensure
adequate accounting disclosure (Belkaoui 1992:85-87).
ii. Private-sector regulation. This approach considers that the public interest in
accounting will be best served if standards are set by the private sector, for example the
Financial Accounting Standards Board (FASS) in the United States. Brown (1982)
deems the participation by the FASS as essential, while Kaplan (1980) and others
oppose the involvement of the FASS. Arguments against private-sector regulation
include the lack of statutory authority and enforcement power and possible lack of
independence from large accounting firms and companies (Belkaoui 1992:87-89).
29
iii. Public-sector regulation. Public-sector regulation has to some extent become a

reality in most countries. Advocates of this approach include Burton (1980) and Chet-
kovich (1980), who conclude that a degree of government intervention is necessary to
protect public interest through leadership and statutory control. Opponents include
Mosso (1980) and Watts (1980), and their criticisms include the possible politicisation
of standard-setting and high corporate costs associated with compliance with govern-
ment regulation of information (Belkaoui 1992:89-92).
Due to the conflicting interests of the parties involved in standard-setting, and the li-
mitations and benefits of each approach, the process of standard-setting will probably
remain a combination of all three approaches. However, a distinct problem that has to
be guarded against and solved is the problem of standards overload, that is, too many
standards and standards being too detailed and rigid (Belkaoui 1992:98).
b. Events approach
The events approach is concerned with providing the information about economic
events that is needed for decision-making. This approach is not concerned with what
information the user wants, but to provide information about significant economic events
to users, to enable users to translate the information into accounting information suit-
able for the particular decisions involved (Wolk et al. 1989:41; Belkaoui 1992: 110). The
result of the events approach is that information in accounting reports are represen-
tative of the real world and not distorted through the possible manipulative use of
alternative accounting techniques (Belkaoui 1992: 111 ).
Utilisation of the events approach could result in a large increase in the information
content of financial reports due to the number of possible events, which may have
monetary and/or non-monetary characteristics, relevant to decision-making (Saenger
1991 :46; Belkaoui 1992: 111 ).
The events approach has the advantage that more information is made available about
economic events that may be relevant to a wide range of users, but it may result in
30
information overload. Supporters of the event approach include Sorter (1969) and
Johnson (1970) (Belkaoui 1992: 110, 113).
c. Behavioural approach
The behavioural approach to the formulation of accounting theory is concerned with
what information users require for decision-making and how this information influences
the behaviour of the individual or groups (Hendriksen & Van Breda 1992: 11, 12). The
behavioural approach has generated much enthusiasm, with the choice of accounting
practices and techniques being evaluated against their economic and sociological
impact. Most studies in behavioural accounting have not attempted to establish a
theoretical framework for accounting, but have in stead focused on the behavioural
effects of accounting information and users' problems in processing the information
(Belkaoui 1992:114).
Dyckman, Gibbins and Swieringa ( 1978) classified the research on the behavioural
effects of accounting information into five areas (Belkaoui 1992: 115, 116):
i. The adequacy of disclosure. These studies include those conducted by Horngren

(1956), Cerf (1961 }, Buzby (1974) and Belkaoui and Kahl (1978). The results indicated
that financial statements are, in general, understandable and comprehensible and thus
disclosure is adequate.
ii. The usefulness of information included in financial statements. The results of

the studies conducted by Chandra (1974), Libby (1975) and Belkaoui (1979,1980),
show that users of financial statements do not, in general, rely only on financial state-
ments for their informational needs, but rather that the relatively important items are
disclosed in the financial statements.
iii. The attributes about corporate reporting practices. The results of the studies by,
Brenner and Shuey (1972) and Godurn (1975) highlighted the differences in attitudes
between various professional accounting groups about the issue of reporting.
31
iv. Materiality judgements. The results of studies conducted by, for example, Woolsey
(1973), Dyer (1973) and Dickhaut and Eggleton (1975) show that individuals have
different judgements regarding materiality and that a number offactors affect materiality.
v. The decision effects of alternative accounting procedures. The results of these

studies generally show that the effect on decision-making of alternative accounting
procedures depends on the nature of the task, environment and users of the informa-
tion. Studies in this field include those of Jensen ( 1966), Barrett ( 1971 ) and Belkaoui
and Cousineau (1977). A new direction of research in behavioural accounting is con-
cerned with agency theory, that is, the costs of monitoring and enforcing relationships
between different parties. These parties include management, owners, government and
creditors, and examples of agency relationships are bond covenants, management com-
pensation contracts and income taxes (Wolk et al. 1989:42,43).
d. Human information processing approach

This approach, which is closely related to the behavioural approach, is concerned with
the quality of the accounting information presented to users. their ability to make use
of the information and the cost of producing the information ( Wolk et al. 1989:43;
Belkaoui 1992: 118).
Four different approaches to researching human information processing have been

identified, and can be summarised as follows (Belkaoui 1992: 119-124 ).
i. The lens model approach. This model was developed by Brunswick (1952) and has
been used in research to examine various accounting decision problems, including (1)
materiality judgements, reasonableness of forecasts and uncertainty disclosures, (2)
bankruptcy and price predictions, and (3) the impact of accounting changes and
feedback methods. Relevant studies include those of Rockness and Nikolai (1977),
Zimmer (1980) and Brown (1981).
32
ii. The probabilistic judgement approach. This approach, also known as the
Bayesian approach, is concerned with the way decision-makers process information
and the choice of techniques which they apply to estimate subjective probabilities. The
studies of Ward (1976), Snowball and Brown (1979) and Biddle and Joyce (1981) are
examples of research of this approach.
iii. Predecisional behaviour, that is, the process of defining the problem, identifying
the information required and searching for the information. Few accounting studies have
been conducted using this approach, and some of the areas researched include mo-
delling expert financial analysis and the strategies used by managers in evaluating
performance reports. Studies in this field include those of Biggs (1979), Weber (1980)
and Libby and Lewis (1982).
iv. The cognitive style approach. This approach is concerned with the factors that
influence the quality of decision-makers' judgements. Accounting studies using this
approach include those ofVasarhelyi (1977), Weber (1980) and Belkaoui (1981, 1982),
and have focused on classifying users of accounting information according to their
cognitive style, and then designing information systems to suite the cognitive style of
these users.
e. Positive approach
The positive approach to the formulation of accounting theory is based on the initial
studies of Beaver (1966) and Ball and Brown (1968), but developed largely through the
influential research of Watts and Zimmerman (1978, 1986, 1990). The positive approach
attempts to make good predictions of real-world events (Henderson et al. 1992:326,
327,360; Scott 1997:221,222).
The positive approach is thus concerned with explaining why certain accounting prac-
tices and procedures are chosen by managers, standard setters and others, based on
a comparison of the costs and benefits of the alternative procedures and practices.
33
Positive accounting theory does not attempt to prescribe to individuals or constituencies

what they should do, but it ultimately attempts to understand and predict accounting
policy choices (Scott 1997:219-221 ).
The predictions made according to positive accounting theory can largely be organised
around three opportunistic hypotheses, as formulated by Watts and Zimmerman (1986)
(Scott 1997:221-222).
i. The bonus plan hypothesis. According to this hypothesis managers of firms with
bonus plans will tend to select accounting procedures that increase current reported
earnings if their bonus is related, at least in part, to reported net income.
ii. The debt covenant hypothesis. According to this hypothesis managers of firms are
likely to shift reported earnings from future periods to the current when the firm is close
to violating accounting-based debt covenants. Therefore, to prevent or postpone such
violations, management may adopt accounting policies to raise current earnings.
iii. The political cost hypothesis. According to this hypothesis managers are likely
to choose accounting policies which, when political costs face the firm, defer reported
earnings from current to future periods. Examples of these political costs are, (1)
attracting excessive media and consumer attention, (2) special regulations or taxes, and
(3) foreign competition.
Positive accounting theory has generated a large amount of empirical research. Studies
include those of Lev (1979), which supports the positive approach in explaining why
different firms may choose different accounting policies; Healy (1985), who found
evidence in favour of the bonus plan hypothesis; Sweeney (1994) reported results
confirming the debt covenant hypothesis; and Jones (1991 ), who found evidence in
favour of the political cost hypothesis (Scott 1997:223-225).
34
Other studies argue that these three hypotheses do not only exist in opportunistic form,
but also in efficiency form. The studies of Christie and Zimmerman (1994) and Dechow
( 1994) supports the viewpoint that accounting policies are not only chosen for
management's own advantage (opportunistic), but also to improve a firm's flexibility
(efficiency) (Scott 1997:227 ,228). Positive accounting theory does not only have
supporters but also critics, including Christenson (1983), Holthausen and Leftwich
(1983), and more recently Sterling (1990) (Henderson et al. 1992:360).
f. Predictive approach
The predictive approach to the formulation of accounting theory, which is of great im-
portance to this study, is concerned with the usefulness of accounting information for
the prediction of future events.
In the search for criteria to use as a basis for choosing between different methods of
accounting measurement, the predictive ability of accounting information came to the
foreground through the study of Beaver, Kennelly and Voss (1968). This approach uses
the criterion of predictive ability to choose between various accounting methods based
on the ability of a particular method to predict events which may be useful to decision-
makers. It focuses on the evaluation of the methods of accounting measurement to
determine which method would be most decision-useful and, therefore, enable decision-
makers to make better decisions (Saenger 1991 :51; Belkaoui 1992: 139).
The predictive approach has attracted much attention from empirical studies and the
areas these studies investigated can be classified into two main categories (Belkaoui
1992:139; Henderson et al. 1992:287,288).
• The behaviour of accounting variables over time; with the aim of establishing the
ability of accounting information to explain and predict economic events.
• The relationship between capital markets and accounting information; with the
aim of establishing the ability of accounting information to explain and predict
capital market reaction to the disclosure of accounting information.
35
1. The prediction of an economic event approach
The predictive approach has been tested and applied in accounting research in the
following ways.
i. Time-series analysis. These studies include examinations of the behaviour of re-

ported earnings and tests of the predictive ability of past reported annual and quarterly
earnings. Studies include those of Brown and Rozeff (1979), Dopuch and Watts (1972),
Griffin (1977) and Collins and Hepwood (1980) (Belkaoui 1992:140; Henderson et al.
1992:289,290).
ii. Relevance of earnings forecasts. The research is centred on the predictive accu-
racy· of earnings forecasts by analysts, management and statistical models. If these
forecasts are found to be inaccurate, they would be of little use to investors and capital
markets, and at this point in time the results of research are inconclusive. Studies on
the relevance of earnings forecasts include those of Cragg and Malkiel (1968), Elton
and Gruber (1972) and Brown, Hughes, Rozeff and Vanderweide (1980) (Belkaoui
1992: 141; Henderson et al. 1992:290).
iii. Distress prediction. Prediction of financial distress is a very important application

of the predictive approach. Two well-known studies are those of Beaver (1966) and
Altman (1973). Despite their limitations, Beaver's univariate study based on a single set
of accounting ratios and Altman's multivariate study based on five accounting ratios
have shown that financial distress can be reasonably accurately predicted (Belkaoui
1992: 142; Rees 1995:298-305).
iv. Information decomposition measures. The investigations of the predictive ability

of information decomposition measures have centred on the areas of corporate bank-
ruptcy and corporate takeover. Information decomposition measures express the sta-
bility over time of financial statement decomposition and the results of studies by Lev
( 1969), Belkaoui ( 1976) and others have pointed to the usefulness of this approach
(Belkaoui 1992:144).
36
v. Explaining corporate restructuring behaviour. Research in this are focussed on

the characteristics of acquired and non-acquired firms and the results of these studies,
for example Tzoannos and Samuels (1972), Rege (1984) and Palepu (1986), have
shown, despite limitations similar to those experienced with distress prediction, the
usefulness of various accounting ratios in predicting takeovers (Belkaoui 1992: 144;
Rees 1995:280,281 ).
vi. Credit and bank-lending decisions. The predictive approach has been used to
predict credit evaluations and research in this field includes the studies of Clarkson
(1962), Ewert (1980) and Dietrich and Kaplan (1982). This approach has also been
extended by Sinkey (1975) and Pettaway and Sinkey (1980) to estimate and predict
financial distress for commercial banks (Belkaoui 1992: 145).
2. The prediction of market reaction: market-based approach

The market-based approach is concerned with reactions of security prices on capital
markets to accounting information, especially with the aim of establishing the relevance
of accounting information and evaluating accounting measurement procedures. This
view is emphasised by Gonedes (1972) and Beaver and Dukes (1972), who contend
that capital market reactions to accounting information may be used as a basis for se-
lecting accounting policies and for the evaluation of alternative accounting policies
according to their information content (Wolk et al. 1989:5,211; Belkaoui 1992: 145, 146).
Belkaoui (p.146) justifies the use of the predictive approach to the formulation of
accounting theory through the role of accounting information in capital markets.
Hendriksen and Van Breda (1992: 168, 169) describe the role and objectives of accoun-
ting information as to ensure optimal allocation of capital resources and, secondly, to
ensure that investors are able to hold optimal portfolios of securities according to their
risk-return preferences.
The research into the predictive approach to the formulation of accounting theory
evaluates and tests accounting policies and techniques on the basis of capital market
37
reactions, using the theory and methodologies provided by Portfolio Theory, the Effi-
cient Market Hypothesis, the Capital Asset Pricing Model, Arbitrage Pricing Theory and
option pricing (Belkaoui 1992: 149; Hendriksen & Van Breda 1992: 168).
Since this study is concerned with an in-depth analysis of research conducted on capital
market theories and pricing models, which will be detailed in chapters 3, 4 and 5, no
review of the research conducted on these theories and models is provided at this
stage. It is, however, useful to briefly list the results of the research regarding the role
of accounting information in capital markets.
According to Henderson et al. (1992:291,295) most of this new area of accounting

research can be labelled as information content studies. These studies investigate the
role of new accounting information in security price movements on capital markets.
They (p.295,296) further identified three areas which have to be simultaneously in-
vestigated to establish the impact of accounting information announcements of the
behaviour of security prices. Models used for such investigations need to be employed
in the following areas:
i. The degree of capital market efficiency. The research is concerned with whether
security prices react when new accounting information is announced and whether this
reaction is of the correct "size".
ii. The expectation of accounting information. The research is concerned with the
application of the models to establish the component of the accounting information an-
nouncements which has already been expected by capital markets to be able to deter-
mine the new (unexpected) component of the accounting information announcement.
This concerns the fundamental assumption that capital markets are efficient and only
reacts to new (unexpected) information not yet reflected in security prices.
38
iii. The measurement of normal returns on investments. In this case, research is

concerned with the application of models to establish what the normal investment
returns, as compensation for the risk of investing, are. It is important to establish
whether any abnormal returns have been gained due to the announcement of new
accounting information.
It can, therefore, be said that the research about the information content of accounting
information announcements assumes that capital markets are efficient in their pro-
cessing of information, and that this assumption can be used to measure the impact of
new (unexpected) accounting information in security prices.
Wolk et al. (1989:5,211) summarises the evidence from capital market research as
follows:
• New information concerning accounting earnings affect security prices, and this
information is rapidly incorporated into security prices in an unbiased manner.
• Changes in accounting policies and methods, which merely have a cosmetic
effect on earnings, do not seem to affect security prices.
• Changes in accounting policies and methods, which have either a direct or in-
direct influence on cash flows, do affect security prices.
• There may be some merit in choosing certain accounting methods and policies
in preference to others, where alternatives exist, due to the effect of indirect cash
flow consequences on security prices.
• Accounting information is useful for assessing risk ,since accounting-based risk
measures are closely linked to capital market risk measures, for example,
variability in market returns is highly correlated with expectations of accounting
earnings and announcements of unexpected accounting earnings.
To conclude this chapter the application of capital market theories and pricing models
in the investment decision-making process is now examined.
39
2.5 THE NEED FOR AND IMPORTANCE OF CAPITAL MARKET

THEORIES AND PRICING MODELS IN INVESTMENT DECISION-
MAKING
Capital market investment decision-making relates to uncertain future events, and

accounting information is an important means for investors to formulate own expec-
tations about the amount, timing and uncertainty of future cash flows. These cash flows
can be either in the form of dividends or the proceeds from the disposal of investments.
Further, for the efficient allocation of capital resources, both the users and the suppliers
of capital require relevant and reliable information to establish whether, and at what
price, to invest or lend (Kam 1990:53-55).
Capital market theories and pricing models provide investors with the means to fully and
properly make use of all information available for investment decision-making. These
theories and models facilitate the proper understanding of the relationship between
investment risk and return in uncertain capital markets. These theories and pricing
models also explain the effect of new information on expectations regarding a security's
risk and return. Lastly, models can be used to determine investment and portfolio values
and these theories and models significantly improve investors' understanding of the be-
haviour and pricing mechanism of capital markets (Hendriksen & Van Breda 1992: 178,
182, 184,243).
Henderson et al. (1992:291) also make the point that if investors have a proper under-
standing of the role of new information in capital markets, they might be able to make
extremely profitable investment decisions if they have access to information before it
is made public.
According to Rees ( 1995: 153, 154) the importance of capital market theories and pricing
models for investment decision-making can be summarised as follows:
40
• Pricing models and capital market theories provide investors with improved
insight into the pricing process on capital markets.
• The methodology and basic techniques of pricing models and capital market
theories enable investors to make better use of investment opportunities.
• It enables investors to choose between a passive or active !nvestment strategy,
depending on their informed perception of the efficiency of capital markets.
In conclusion it can be stated that capital market theories and pricing models provide
investors with the means to improve their investment decision-making and portfolio
selection. Improved investment decision-making also leads to more optimal allocation
of resources in a free economy.
2.6 SUMMARY
In this chapter capital market theories and pricing models were examined within the
context of accounting theory and investment on capital markets.
The first part of the chapter described the concepts "capital market theories", "pricing
models", "accounting theory" and "investment decision-making" to provide a frame of
reference for the remainder of the chapter. In the second instance, the background to
the development of the most important capital market theories and pricing models was
reviewed. This established the role of these theories and pricing models, the role of
information and the risk/return relationship in investing on capital markets. The dis-
cussion then focused on the various approaches to the formulation to an accounting
theory. This discussion culminated in a review of the market-based approach to the
formulation of accounting theory wherein the importance and role of accounting infor-
mation in capital markets were established. This chapter concluded with an examination
of the role and importance of the capital market theories and pricing models in
investment decision-making.
41
The conclusion reached is that investors find real benefit through the application of
these theories and models, as these theories and models facilitate improved investment
and portfolio selection through improved usage of the available information.
Viewed as a whole, this chapter attempted to establish, through an investigation of the

various approaches to the formulation of an accounting theory and the available capital
market theories and pricing models, the importance of information, theories and models
in investment decision-making, and in particular, for investment on capital markets.
In an attempt to provide a deeper understanding of the role of these concepts in invest-

ment decision-making, the next chapter focuses on the relationship between investment
risk and return, Portfolio Theory and the Efficient Market Hypothesis.
42
CHAPTER3
THE DEVELOPMENT OF CAPITAL MARKET THEORIES
3.1 Introduction 43
3.2 The role of capital market theories in investment decision-making 44
3.2.1 The concepts "capital market", "investment", "risk", "uncertainty"

and "return" 46
3.2.2 The relationship between risk and return in investments 50
3.2.3 The effect of risk, uncertainty and return on investment decisions 51
3.2.4 Utility theory and the relationship between risk and return 52
3.2.5 Capital market theories and investment decision-making 54
3.3 The development of Portfolio Theory 55
3.3.1 The concept and nature of Portfolio Theory 56

3.3.2 Portfolio Theory and the concepts risk, return and uncertainty 59
3.3.3 The implications of Portfolio Theory for portfolio selection 66
3.3.4 Optimal portfolios and utility theory 69
3.3.5 Research related to Portfolio Theory 72
3.3.6 Criticism of Portfolio Theory 74
3.4 The development of the Efficient Market Hypothesis (EMH) 75
3.4.1 The concepts "perfect markets", "efficient markets" and "inefficient

markets" 76
3.4.2 Forms and degrees of market efficiency 80
43
3.4.3 Tests of market efficiency 82

3.4.4 Efficient market anomalies 85
3.4.5 Implications of the EMH 87
3.4.6 Misconceptions regarding market efficiency 89
3.4. 7 Research related to market efficiency 90
3.5 Summary 116
3.1 INTRODUCTION
The main focus of this chapter is on three important aspects of investment in capital
markets, namely:
• Investors dislike risk and the more risky the investment the greater the return
expected by investors (Dobbins et al. 1994:6).
• Investors should not view investment in individual securities in isolation, but
rather as a component of the total investment in a portfolio of securities (Sharpe
1985:120).
• Investors require financial information to evaluate investment opportunities (Rees
1995:153).
The first section of this chapter is concerned with the first aspect. namely the relation-
ship between risk and return of investment in capital markets. According to Dobbins et
al. (1994:6) there exists a trade-off between risk and return which results in investors
expecting a higher expected return for more risky investment opportunities.
The second aspect, the implications of Portfolio Theory and diversification for invest-
ment decision-making, is examined in section two of this chapter. Markowitz (1952)
showed that rational investors wishing to maximise their utility of wealth can, when
investing in efficient markets, reduce the risk of their investments by holding well-
44
diversified portfolios of securities and accordingly reduce the level of risk associated
with a given expected level of return.
The third aspect, and final section of this chapter, deals with a major assumption of
modern investment management - that of the efficiency of capital markets. The Efficient
Market Hypothesis (EMH) suggests that capital markets incorporate new information
immediately and correctly in security prices. It is therefore implied that the prices of
securities in efficient capital markets are the best estimates of the securities' true value.
Further, market efficiency implies that investors cannot systematically gain abnormal
profits, nor incur abnormal losses, if they act on the basis of publicly available infor-
mation (Weston & Copeland 1992:93, 94).
A further important part of this chapter is the analysis and classification of research
conducted on.Portfolio Theory and the EMH. The research on these theories, together
with those on the pricing models, are the main areas of emphasis in this study.
As a starting point, the role of the capital market theories in investment decision-making
is examined.
3.2 THE ROLE OF CAPITAL MARKET THEORIES IN INVESTMENT

DECISION-MAKING
According to Weston and Copeland (1992:92) the main role of capital markets can be
described as follows:
• Facilitating the investment process by providing a continuous marketplace where

security prices are tested on an ongoing basis and where buying and selling
transactions can occur efficiently at a relatively low cost.
• Providing a relatively stable marketplace for securities whereby, through its con-
tinuous operations and the resultant frequent and relatively small changes in
45
security prices, investors are protected against volatile security prices.

• To facilitate the allocation of investment capital, capital markets need to value
securities correctly. If securities are mispriced and capital markets are not run
efficiently, it may result in the markets becoming speculative and this may result
in many investors not investing their funds in capital markets (Firth 1977: 1,2).
Portfolio Theory, the first major capital market theory, provides a guideline for investors
in their capital market investment decision-making. According to Portfolio Theory, in-
vestors should not invest all their funds in one security. All securities should be viewed
in the context of being part of a portfolio and, thus, it is not the risk and return of the
security that is important, but rather its contribution to the overall risk and return of the
portfolio. Decisions to buy or sell securities should also only be taken after the diversifi-
cation effect has been taken into account, that is, what the effect of the decision will be
on the overall risk and return of the portfolio (Sharpe 1985: 120; Linley 1992:8).
On the other hand, the second major capital market theory, the EMH, provides investors
with a guideline about how efficient capital markets price securities. Keane (1983:2,3)
provides the following summary of the role and importance of the EMH for investment
decision-making. Capital markets create wealth through the optimal allocation of
investment capital and therefore the prices of securities provide investors with useful
signals of how to construct their portfolios of investments. If the market is inefficient in
its pricing of securities, investor attention would be drawn away from efficient diversifi-
cation towards looking for opportunities of possible exploitation of the perceived ineffi-
cient pricing of securities. The importance of the EMH to investors is that the reliable
pricing of securities enables investors to concentrate on the construction of optimal
portfolios in line with their risk/return preferences, rather than attempting to outperform
the market.
To fully understand the implications of the main capital market theories, it is necessary
to first examine the concepts "investment" and "capital market" and how the concepts
"risk", "return" and "uncertainty" affect capital market investment decision-making.
46
3.2.1 The concepts "capital market", "investment", "risk", uncertainty" and

"return"
3.2.1.1 The concept "capital market"

The capital market, also known as the securities market, is the market where the sup-
pliers of capital (investors) and the demanders of capital (companies and the govern-
ment) meet to trade in financial assets (securities), that is, shares, government bonds,
options, currencies, etcetera. The capital market, therefore, enables investors to trade
in securities and it enables companies and the government to raise capital (Hendriksen
& Van Breda 1992:168; Rees 1995:154; O'Brien & Srivastava 1995:3).
3.2.1.2 The concept "investment"

Sharpe (1985:2) defines investment as:
The sacrifice of something now for the prospect of something later ...
Investment is the sacrifice of certain present value for (possibly uncer-
tain) future value.
The most important aspect of investments is that the investor should be compensated
for the risk taken, that is, the investment should provide an adequate return in terms of
income and capital growth (Rees 1995:157).
3.2.1.3 The concept "risk"

According to Linley (1992:9) and Van Horne (1992: 12) risk is the possibility that the
actual return on an investment will be different from the expected return on that invest-
ment, and that it is possible to attach probabilities to these expected outcomes.
Rees ( 1995: 158) further states that:
The variability in returns is taken to represent the risk of the investment

to investors since the variability reflects the uncertainty attached to
returns.
47
Risk can be measured, either by the average (or mean) variance or the standard de-
viation of returns from their expected values. Standard deviation indicates how far the
actual returns have deviated from the expected returns and is the measure which is
preferred for investment analysis. Standard deviation is an estimate of the likely diver-
gence of actual returns from the expected returns and the larger the spread of diver-
gence, the more risky the investment (Sharpe 1985:122, 123; Broadbent 1992:8). This
is illustrated in figure 3.1 and its calculation is shown by formula 3.1.
Figure 3.1 Standard deviation as a measure of risk
Increasing
probability
of outcome ........................................................... c
where:
A = The average expected return, which is the same for both investments
B Investment with more risk, due to its greater distribution of expected returns (larger
standard deviation)
C = Investment with less risk as there is less variability in its expected returns (smaller
standard deviation)
Source: Harrington (1987:7) and Broadbent (1992:8).

48
Formula 3.1 Standard deviation of an investment
where:
cr 2 = Standard deviation
n = Number of possible outcomes
pi = Probability of i th return
E(R) = Average expected return
Ri Probable ith expected return
Source: Harrington ( 1987 : 20).
3.2.1.4 The concept "uncertainty"

Uncertainty arises when there are various possible expected returns on an investment,
but little evidence exists to predict the possible expected returns, therefore, the alterna-
tive possible outcomes cannot be identified and the investor is unable to estimate the
probability distribution of the expected returns (Clark et al. 1979: 117; Drury 1992:320).
3.2.1.5 The concept "return"

The actual return on an investment, during a certain period, is the income received
during the period together with any change in the value of the investment (Henderson
et al. 1992:292,293). This is expressed in formula 3.2:
49
Formula 3. 2 Actual return on an investment
where:
Rt = Actual return for period t

Dt = Dividends (or interest) received during period t
pt = Market price at the end of period t
pt-! = Market price at the beginning of period t
P1-P1-1 = Capital gain or loss during period t
Source: Harrington (1987:6), Henderson et al. (1992:293) and Dobbins et al. (1994:5,6).
The expected return on an investment is merely the weighted average of probable ex-
pected returns (Sharpe 1985:126), which can be expressed as:
Formula 3.3 Expected return on an investment
11
E(R)= L(RJ~)
i=l
where:
E(R) = Expected return

n = Number of possible outcomes
R; = Return associated with the ith outcome
P; Probability of the occurrence of the ith outcome
Source: Clark et al. (1979: 119), Harrington (1987:20) and Weston and Copeland ( 1992:363).
50
3.2.2 The relationship between risk and return
The risk/return relationship is illustrated in figure 3.2:
Figure 3.2 The relationship between risk and return
Expected
return
···················································-·························· ~ ........................ A
Risk
where:
A Risk-free rate of return

B = As the risk of an investment increases, the greater the return expected by the
investor
Source: Broadbent (1992:8) and Drury (1992:357).
The relationship between risk and return is a major topic in finance theory and the
nature of this relationship is that the higher the risk associated with an investment, the
greater the return expected by the investor to compensate for the level of risk involved
(Broadbent 1992:8; Drury 1992:357).
51
Now that the risk/return relationship has been established, it is important to consider
how investment decisions are influenced by risk, return and uncertainty.
3.2.3 The effect of risk, uncertainty and return on investment decisions
The decision to invest depends largely on whether the investment involved would
render an adequate return to compensate for the risk involved. Any rational investor
would expect a greater return on an investment which bears a higher degree of risk
than an alternative investment opportunity (Broadbent 1992:8).
O'Brien and Srivastava (1995:5,9) states that the following factors and problems affect
investment decision-making:
• The investor's current level of wealth.

• Estimating the expected return on an investment and determining whether the
expected return is sufficient to compensate for the level of risk involved.
• Identifying the sources of risk and measuring the level of risk involved.
• Making the investment decision based on the risk and return involved, the pre-
vailing interest rates and the various other investment opportunities available.
An important aspect in investment decision-making is estimating the value of an invest-

ment, and this value, apart from the underlying intrinsic value, is based to a large extent
on the size and timing of the expected returns combined with the risk involved in obtai-
ning the returns. Of these elements, risk is the most difficult to determine and the har-
dest part of decision-making is deciding on how much return is required to compensate
for the measurable risk involved (Harrington 1987: 1). Three investor attributes towards
risk have been identified (Drury 1992:326; Weston & Copeland 1992:358) and these
attributes can be described as:
• Risk aversion: The investor seeks to avoid risk and, when faced with a choice
between alternative investments with the same expected return, would opt for
52
the investment with the lowest degree of risk.

• Risk seeking: The investor is indifferent to the level of risk involved (gambler)
and, when faced with different investment opportunities with different expected
returns, selects investments which yield the highest expected return.
• Risk indifference: When choosing between investment alternatives that render
the same expected return, the investor does not distinguish between different
risk levels when making an investment decision.
Undoubtedly some investors are either risk seekers or risk indifferent, but empirical
evidence, as shown by the study of Friend and Blume (1975), suggests that investors
avoid risk where possible and will only opt for investments with a higher degree of risk
when the related increase in the expected return exceeds the increase in the risk
involved (Drury 1992:326,357).
The principle of choosing the less risky investment when faced with the choice between
investments rendering the same expected return can also be described using the basic
premises of utility theory (Weston & Copeland 1992:358).
3.2.4 Utility theory and the relationship between risk and return
Under conditions of uncertainty, both expected return and risk need to be considered
when investment alternatives are evaluated. Under conditions of uncertainty, the out-
comes of investment alternatives are only known in a probabilistic form and modern
utility theory (based on the work of Von Neumann and Morgenstern (1947)) was de-
veloped to deal with decision-making under conditions of uncertainty. Utility theory
attempts to provide a basis for rational decision-making. Utility values are attached to
the various alternatives available, which include all aspects relevant to the decision.
When utility theory is applied to investment decision-making, in particular in its appli-
cation to the trade-off between risk and return, the investor must consider the following
factors within his framework of personal preferences (Clark et al. 1979:122; Dobbins
et al. 1994:23):
53
• The opportunity set of all relevant available decision alternatives.

• The determination of a hierarchy of alternatives, together with a trade-off of alter-
natives within this hierarchy.
• The investor's perception of risk and the investor's preference with regards to
the risk/return relationship.
• The investor's attitude towards the choice between current or future consumption
of wealth.
It is reasonable to assume that rational investors will act in such a manner as to maxi-
mise utility and therefore select those investment alternatives which will provide the
greatest level of satisfaction. Fundamental to utility theory is the notion that investors
have a diminishing marginal utility for wealth, and that this affects their attitude towards
risk. Diminishing marginal utility for wealth leads to an investment attitude of risk aver-
sion.
The principle of diminishing marginal utility for wealth can best be described using the
following examples:
• An investor with diminishing marginal utility for wealth will, when faced with a
choice between the two alternatives, prefer not to lose a Rand rather than gain
a Rand.
• An investor with diminishing marginal utility for wealth gains less satisfaction
from the second R 1 000 of wealth earned than from the first R 1 000 and this loss
of satisfaction increases with each additional R 1 000 of wealth earned (Weston
& Copeland 1992:358-361 ).
The relationship between wealth and its utility depends on the investor's attitude to-
wards risk and is illustrated in figure 3.3.
54
Figure 3.3 The relationship between income (wealth) and utility
Utility
Income (Wealth)
where:
A = Investor is a risk taker; the investor has an increasing marginal utility of wealth
B Investor is indifferent to risk; the investor's marginal utility of wealth is constant
C = Investor is risk averse; the investor has a diminishing marginal utility of wealth
Source: Clark et al. (1979: 124) and Weston and Copeland (1992:359).
In conclusion, the manner in which the various capital market theories describe the in-
vestment decision-making process, is considered.
3.2.5 Capital market theories and investment decision-making
As discussed in the preceding sections, obtaining an adequate return to compensate

for the risk involved is the primary motivation of investors. Some investors may actually
have a secondary goal of outperforming the capital market, that is, earning excessive
returns, if they perceive capital markets to be inefficient. In practice, it can be assumed
55
that most investors are not motivated solely by either one of these goals, but rather a
mixture of the two. The problem for investors is that these are two conflicting goals and
the more the investor diversifies his portfolio of investments in order to reduce the level
of risk, the less likely he will be able to beat the market (Keane 1983:99).
Two capital market theories deal specifically with these two investment goals, namely
Portfolio Theory and the EMH. Portfolio Theory is concerned with the relationship
between risk and return and the effect of diversification on this relationship. Portfolio
Theory proposes that it is necessary for investors to hold diversified portfolios, in line
with their risk preferences, which render the maximum rate of return for a specific level
of risk (Hendriksen & Van Breda 1992: 168).
The EMH proposes that capital markets efficiently and correctly price securities and
therefore it is difficult, if at all possible, to beat the market on any consistent basis. The
efficiency aspect of capital markets is also of particular interest to accounting re-
searchers since observations of market reactions to the release of accounting informa-
tion provide both a means of testing the validity of the EMH as well as the usefulness
of accounting practices and standards (Hendriksen & Van Breda 1992: 168).
The next two sections of this chapter are concerned with a discussion of, firstly, Port-
folio Theory and thereafter the EMH.
3.3 THE DEVELOPMENT OF PORTFOLIO THEORY
A combination of securities or investments held by an investor is known as a portfolio

and Portfolio Theory is concerned with the selection of efficient combinations of se-
curities or investments (optimal portfolios) depending on the investor's desired level of
risk (Linley 1992:3; Drury 1992:333).
Optimal portfolios are those that provide the highest return for a given level of risk, or
the lowest level of risk for a given rate of return (Weston & Copeland 1992:366).
56
Portfolio Theory is concerned with the principle that a portfolio of securities will always
be a better investment strategy than holding a single security, and the concept behind
Portfolio Theory is that any rational investor will always opt for the portfolio of securities
which renders the highest possible return for a specified level of risk (Viljoen 1989 :20;
Correia et al. 1993: 113).
Markowitz (1952, 1959) pioneered the development of modern Portfolio Theory, in parti-
cular with regards to risk in a portfolio sense. According to Markowitz rational investors
will select portfolios which render the highest possible return for a specified level of risk,
or will select portfolios with the minimum level of risk for a specified rate of return
(Dobbins et al. 1994: 12).
Diversification, as suggested by Markowitz, improves on naive diversification in the

· sense that investment in individual securities in a portfolio need not be equally
weighted. Before Markowitz, naive diversification stipulated that investment in individual
securities in a portfolio should be equally weighted. Markowitz provided the principle
that risk in a portfolio can be reduced if the portfolio weights are not equal, provided
that the weights are properly managed. Through his contributions, Markowitz provided
a formal theory for portfolio selection (O'Brien & Srivastava 1995:48).
3.3.1 The concept and nature of Portfolio Theory
Portfolio Theory is based on a number of assumptions with regard to the way capital
markets behave and investors function in conditions of uncertainty. Furthermore, Port-
folio Theory has significant implications for investment decision-making and selection.
3.3.1.1 Assumptions of Portfolio Theory

Linley (1992:3,4), Hendriksen and Van Breda (1992:179) and O'Brien and Srivastava
(1995:7,8) summarise the assumptions underlying modern Portfolio Theory as follows:
• Investors view the return on an investment as an adequate description of the

57
outcome of the investment and that investment returns are normally distributed.
• Investors view the risk on a portfolio as being proportional to the variability of the
expected return for the portfolio.
• All investors have fixed single period investment horizons and base their in-
vestment-decisions purely on the parameters of perceived risk and expected
return.
• Investors are rational and risk-averse, that is, investors prefer less risk for a
given expected return and more expected return for a given level of risk.
• All investors have full and equal access to the available information, there are
no transaction costs or taxes, capital markets are perfectly competitive and se-
curities are completely divisible.
• Investors are considered small in relation to the market and hence their indivi-
dual actions will have no significant influence on market prices.
• Proceeds from investments occur at the end of the period and, when these are
of equal amounts, investors are indifferent to the choice between income and
capital gains.
Under these assumptions, portfolios are deemed to be efficient when no other portfolio
renders a higher expected return for a given level of risk or no portfolio has a lower
level of risk for a specified expected return (Linley 1992:4).
3.3.1.2 Stages in portfolio selection

Markowitz (1991 :3) describes a good portfolio in the following manner:
A good portfolio is more than a long list of good stocks and bonds.
It is a balanced whole, providing the investor with protections and
opportunities with respect to a wide range of contingencies. The
investor should build towards an integrated portfolio which best
suits his needs.
Markowitz (1991 :6) also identifies two objectives of portfolio selection which are
common to most investors:
58
• They want return to be high; they prefer more of it to less of it.

• They want this return to be dependable, stable and not subject to uncertainty;
that is, they are not speculators.
Three stages in portfolio selection have been identified:
a. Investment analysis
Assessing the expected return and risk of potential investments, and also the correla-
tion between the various expected returns (Viljoen 1989: 19).
b. Portfolio analysis
Using the information obtained from investment analysis, different sets of investment
opportunities that meet the investor's objectives can be identified (Viljoen 1989:20; Mar-
kowitz 1991 :3).
c. Investment strategy
The final step involves the selection of a suitable strategy for the selection of invest-
ments rendering an acceptable return at an acceptable level of risk together with the
identification of the method of financing the investments (Viljoen 1989:20).
3.3.1.3 Implications of Portfolio Theory

The most important implication of Portfolio Theory is that investors can reduce the level
of risk through diversification and, therefore, earn the same expected return but at a
lower level of risk. This implies that it is not the return and risk of individual securities
that are important, but rather the contribution of an investment to the overall return and
risk of the portfolio (Hendriksen & Van Breda 1992: 179, 180; Drury 1992:336).
Following on the above the effect of diversification on portfolio return and risk can be
considered.
59
3.3.2 Portfolio Theory and the concepts of risk, return and uncertainty
Due to the fact that the effect of economic conditions cannot be predicted without doubt,
uncertainty forms a permanent part of investment in securities. This uncertainty requires
that the potential and weaknesses of all investments in the portfolio need to be carefully
analysed. A further permanent aspect of investment in securities is that returns on
securities are correlated, that is, returns on securities tend to move together, either up
or down. As this correlation is not perfect, especially for securities across various indus-
tries and market segments, risk can be reduced through investment in a well diversified
portfolio (Markowitz 1991 :4,5).
3.3.2.1 Portfolio Theory and the concept of risk

The total risk involved in investment in securities co~sists of two parts:
a. Specific risk
This element of risk can be reduced or even eliminated through efficient diversification.
This risk is specific to the individual security, and results fromeconomic, political and
other factors which are unique to the corporation involved. Specific risk is also known
as unsystematic risk, non-market risk, avoidable risk, unique risk or diversifiable risk.
b. Market risk
This element of risk cannot be diversified away. This risk affects all securities in the
market to a greater or lesser degree through changes in factors such as interest rates,
inflation, foreign exchange rates, taxation, oil prices, world recessions, government
spending and changes in the money supply. Market risk is also known as systematic
risk, non-specific risk, unavoidable risk or non-diversifiable risk (Linley 1992:8;
Broadbent 1992:8; Correia et al. 1993:129; Dobbins et al. 1994:8,9).
60
3.3.2.2 Risk reduction through diversification

Portfolio Theory argues that the risk involved in investment in securities can be reduced
through diversification. The market will not compensate the investor for the specific risk
associated with a poorly diversified portfolio, but a well -diversified portfolio should only
suffer the risk known as market risk. Various studies have shown that when more and
more randomly selected securities are added to a portfolio, the closer the level of risk
resembles the level of market risk, and that 15 to 20 randomly selected securities are
sufficient to remove most of the specific risk of a portfolio (Sharpe 1985: 172; Broadbent
1992:8; Van Horne 1992:69).
This principle of reduction of portfolio risk through diversification is illustrated in figure

3.4, and the following can be seen from figure 3.4:
• The risk of a portfolio is rapidly reduced by increasing the number of securities

in the portfolio until a stage is reached where the addition of more securities
achieves little or no risk reduction.
• There is always a degree of risk involved, even at the point where the portfolio
reflects the composition of the market as a whole (Correia et al. 1993: 129).
Figure 3.4 Risk reduction through diversification
Total
portfolio
•!
i
i
risk
!
.. c
Number of securities
in the portfolio
61
where:
A = Specific risk
B = Market risk
c = Diversified portfolio
Source: Van Home (1992:70), Correia et al. (1993: 128) and Dobbins et al. (1994: 7).
Figure 3.4 illustrates two very important aspects of Portfolio Theory, namely:
• Investors should not expect to be compensated (receive additional return)

for taking avoidable risk.
• The expected return on a portfolio is directly related to the level of market
risk associated with that portfolio (Dobbins et al. 1994:8,9).
3.3.2.3 Portfolio risk

Fundamental to Portfolio Theory is the principle that the risk of a portfolio is not simply
the weighted average of the risk of the individual investments in the portfolio. The risk
of a portfolio does n'"'t only depend on the riskiness of the investments in the portfolio,
but also depends on the relationship between the returns of the various investments
(Sharpe 1985:127; Van Horne 1992:53; Correia et al. 1993:115).
The risk of a portfolio will be at its highest when the returns on the investments are per-
fectly correlated. The purpose of diversification is to include investments in the portfolio
which correlate as imperfectly as possible and thereby reduce the level of risk as far
as possible (Hendriksen & Van Breda 1992: 180). The correlation coefficient is used as
a measure to show the degree to which the returns of the individual investments move
together (correlate}, and in general, returns tend to be positively correlated but are un-
likely to be either perfectly positively nor perfectly negatively correlated (Correia et al.
1993:115).
62
According to Clark et al. (1979:150), Drury (1992:334) and Correia et al. (1993:115) the
correlation between returns can be classified as follows:
a. Perfect positive correlation (+1)

This occurs when the returns of different investments exceed or fall below their expec-
ted values by the same margin. Risk cannot be reduced when perfectly positive invest-
ments are combined.
b. Perfect negative correlation (-1)

This occurs when the returns of some investments exceed their expected values, while
the returns on other investments fall below their expected values by the same margin.
When perfectly negative investments are combined, risk is reduced significantly to the
market risk level.
c. Zero correlation (0)

This occurs when the movement in the returns on investments are independent of one
another. If the returns of some investments exceed their expected values, then the re-
turns on other investments are likely to either exceed of fall below their expected values
by the same margin. A combination of zero correlated investments reduces the level of
risk.
In practice the process of calculation of the correlation coefficients is complex, since

the correlation coefficient of each new investment needs to be calculated in relation to
all the existing investments in the portfolio (Drury 1992:335). The formula for the calcu-
lation of portfolio standard deviation (risk) can be expressed as follows:
63
Formula 3. 4 Standard deviation of a port/olio
The standard deviation of a portfolio is a weighted sum of the standard deviations of

the individual securities adjusted for the covariance among the securities.
n n-1 n
aP = L:x;a;
i=l
+22:: L X;XiaU
j=l i=j+l
where:
Standard deviation of the portfolio
= The proportion of the portfolio invested in securities i and j

respectively
a; = The standard deviation of security i
aij = The covariance between the returns on securities i and j
n = The total number of securities in the portfolio
and the covariance can be calculated as:
where:
= Covariance between the returns on securities i and j
= Correlation coefficient between the returns on securities i and j
Standard deviations of securities i and j
Source: Van Home (1992:55), Correia et al. (1993:126) and Dobbins et al. (1994: 25).
64
3.3.2.4 Portfolio return

When different combinations of securities are evaluated, both the standard deviation
(risk) and expected return of the various combinations need to be considered. The cal-
culation of the expected return of a portfolio is simple, since it is purely the weighted
average of the expected returns on the investments based on the proportional value of
the portfolio (Clark et al. 1979: 150; Sharpe 1985: 127; Ross et al. 1990:257; Weston &
Copeland 1992:366).
The formula for the calculation of the expected return on a portfolio can be expressed
as:
Formu la 3. 5 Expected return on a port/olio
n
E(RP)= Ix;E(R;)
i=l
where:
E(R) = Expected return on the portfolio

X; Proportion of the portfolio invested in security i
E(RJ = Expected return on security i
n = Number of securities in the portfolio
Source: Van Home (1992:53), Correia et al. (1993:115) and Dobbins et al. (1994:24).
Rational investors will choose efficient portfolios and the available efficient portfolios,
which depends on the investor's trade-off between risk and return, together form the
efficient frontier (Harrington 1987: 11 ). The opportunity set and efficient frontier is illus-
trated in figure 3.5.
65
Figure3.5 The efficient frontier
Expected
return
Risk (Standard deviation)
where:
A = Efficient frontier; all efficient portfolios which dominate all portfolios to the right of
it, as portfolio c dominates portfolio d.
B Opportunity set; all attainable portfolios which can be achieved from the available
securities.
a = Portfolio with the least possible risk
b = Portfolio with the highest possible rate of return
Source: Sharpe (1985: 140), Hendriksen and Van Breda (1992:180, 181)
66
3.3.3 The implications of Portfolio Theory for portfolio selection
O'Brien and Srivastava (1995:55) define the objective of portfolio selection in the fol-
lowing manner:
The objective of the portfolio selection problem is to minimise port-

folio risk, subject to:
1) attaining a target expected return, and
2) the sum of portfolio weights equalling one.
As a result, O'Brien and Srivastava (p.55) identify two constraints in the portfolio selec-
tion problem. Firstly, that for a certain expected return a portfolio is not dominated by
any other portfolio with a lower level of risk. Secondly, if all portfolio weights are to be
positive and the sum of these weights must equal one, short-selling is not allowed.
The rest of this section will only concentrate on the first constraint, since it is one of the
underlying principles of the basic capital market theories and pricing models that short-
selling is not allowed.
3.3.3.1 The efficient frontier and portfolio selection

Each investor has an opportunity set (a set of attainable portfolios) which depends on
the investor's personal preferences regarding the trade-off between risk and return. Out
of all these possible portfolios, only some can be considered to be efficient (undomi-
nated). To be considered efficient, a portfolio must satisfy the following criteria (Clark
et al. 1979: 158; Markowitz 1991 :6; Rees 1995: 163; O'Brien & Srivastava 1995:51,52):
• No other portfolio has a lower level of risk for a given expected return.
• No other portfolio has a higher expected return for a given level of risk.
• No other portfolio has a higher expected return combined with a lower level of risk.
67
3.3.3.2 Portfolio selection and global diversification

The return that an investor receives on investments in securities in the domestic capital
market consists of:
• The capital growth of the investment over the period considered.

• The value of income (dividends) received during the period.
A third element must, however, be taken into consideration with return on investments
in securities in international capital markets, namely:
• The change in the relationship between the investor's domestic currency and the
foreign currency of the country of investment.
When there is a difference between the set of efficient portfolios attainable in the do-
mestic and international markets, some international efficient portfolios will dominate
domestic efficient portfolios. Therefore, global diversification should be beneficial in the
sense that investors can decrease the level of risk for a given expected return or in-
crease the expected return for a given level of risk (Dobbins et al. 1994:37,38).
3.3.3.3 Portfolio selection and capital budgeting

Portfolio selection for capital projects are more difficult than for investment in securities.
This is due to the fact that most capital projects are indivisible and cannot be divided
into homogenous units, unlike investment in securities which are divisible units with the
same expected return and risk for each ordinary security of a specific corporation.
Unlike investment in securities, it is most improbable that an investor can acquire a
certain percentage of a capital project and thereby acquire a certain percentage of the
return and the risk (Clark et al. 1979: 161 ) .
When considering a large number of indivisible capital projects, the efficient set of
options involves a large number of combinations and integer programming is usually
used to solve this problem. The lower confidence limit model assumes that the investor
68
has a minimum acceptable return (the lower confidence limit) rather than a maximum
level of risk for a specific level of return (Clark et al. 1979: 161 ). The following equation
is used to establish the lower confidence limit:
Formula 3. 6 The lower confidence limit
where:
L Lower confidence limit

E(R) Expected return on the portfolio
k = Constant which refers to the number of standard deviations in the
normal distribution
= Standard deviation of the portfolio
Source: Clark et al. ( 1979: 161 ).
By using this model, investors can establish a level beneath which the return on the
project must not drop. Investors achieve this by putting a value on k, either 1, 1.5, 2 ... ,
and thereby establishing the minimum acceptable return together with the amount of
acceptable downside risk. The higher the acceptable return (l), the fewer the number
of efficient combinations available (Clark et al. 1979: 161, 162).
3.3.3.4 The selection of porfolios

In the previous part of this section it has been established how to identify efficient port-
folios, but not how to select a specific portfolio. This choice will depend on investors'
attitudes towards risk. Thus, if investors are risk-takers, they will select portfolios with
a higher risk level and with higher expected returns, or if investors are risk averse they
will select portfolios with less risk and lower expected returns (Harrington 1987: 11 ).
69
3.3.4 Optimal portfolios and utility theory
Portfolio Theory is concerned with the selection of optimal portfolios, that is, the alloca-
tion of investors' wealth among securities that will provide the highest return for a given
level of risk or the lowest degree of risk for a specified return. Utility theory can be
applied to select the optimal portfolio from the set of efficient portfolios, in other words,
selecting that portfolio which best represents the risk and return preferences of the
investor (Clark et al. 1979:159; Viljoen 1989:19; Dobbins et al. 1994:22). The process
of selecting the optimal portfolio starts with specifying the objectives of the investor,
which are expressed in terms of the investor's utility function. Following on this, the
expected utility is then calculated to determine the investor's investment choices under
conditions of risk and uncertainty (Van Horne 1992:60; Dobbins et al. 1994:22).
In general, rational investors are considered to have diminishing marginal utility of

wealth, meaning they are risk-averse investors who associate risk with any variance
from the value of expected return (Clark et al. 1979: 159; Van Horne 1992:60,61;
Dobbins et al. 1994:23). The utility indifference curves, which express the different
combinations of risk and return of risk-averse investors, are depicted in figure 3.6.
Figure 3.6 Risk-averse investors' utility indifference curves
Expected
return

70
where:
= Indifference curves; the further the curves move to the left the greater the
utility (more return for less risk)
Source: Clark et al. (1979: 159) and Van Horne (1992:61).
Indifference curves represent a trade-off between risk and expected return where ex-
pected utility is constant and investors are indifferent to any combination of expected
return and risk on a particular indifference curve. The more risk-averse the investor, the
steeper the slope of the indifference curve (Van Horne1992:60; Dobbins et al. 1994:28).
Rational investors want to maximise their expected utility of wealth and will select port-
folios which satisfy this requirement. Expected utility of wealth is calculated as follows:
Formula 3. 7 Maximum utility of wealth
E[u(w)J = LP;u(w;)
i
where:
E[u(w)J = Expected utility of wealth
w, = Wealth from outcome i

P; Probability that outcome i will occur
Source: Dobbins et al. (1994:44).
Investors want to select portfolios which are on the highest indifference curve, and opti-
mal portfolios can be found where the highest indifference curve is at a tangent to the
efficient frontier. The point of tangent represents the highest level of satisfaction inves-
tors can achieve and each investor has a unique point of tangency according to his
preferred combination of securities (Van Horne 1992:61; Dobbins et al. 1991 :29).The
point of tangency that represents investors' optimal portfolios, is depicted in figure 3.7.
71
Figure 3. 7 Selection of the optimal portfolio
Expected
return
where:
A = Efficient frontier
B = Opportunity set
= Utility curves
a = Optimal portfolio; the tangent between the efficient frontier and the utility
curves.
Source: Clark et al. (1979: 159) and Dobbins et al. (1994:29).

72
3.3.5 Research related to Portfolio Theory
Research related to Portfolio Theory has in the main focussed on the effectiveness of
diversification and on the benefits of global diversification.
3.3.5.1 Studies on the effectiveness of diversification

Diversification has the effect of lowering the risk of a portfolio through the reduction of
the variability of the portfolio returns. The results of research studies confirm this and
has in general shown that a portfolio of securities need only consist of a relatively small
number of securities to achieve significant benefits of risk reduction (Correia et al.
1993: 128, 129).
Evans and Archer (1968)

Due to the fact that risk can be quantified, it is possible to identify both the analytical
and empirical properties of risk. Evans and Archer studied the impact of portfolio size
on portfolio risk and concluded, from randomly selected equally weighted portfolios of
varying size, that the benefits from diversification are significantly reduced when the
share portfolio consists of more than 15 to 20 shares (O'Brien & Srivastava 1995:40).
Wagner and Lau (1971)

Diversification has the effect that the larger the number of shares the lower the risk of
the portfolio, but the reduction of risk is relatively small after the first 10 shares. Further,
there is also a level of risk (market risk) that cannot be diversified away. The data from
the study of Wagner and Lau confirms that diversification reduces the level of risk, but
even well-diversified portfolios have some level of risk that cannot be reduced through
diversification (Weston & Copeland 1992:384,385).
Solnik (1974)
Most specific risk can be diversified away by randomly increasing the number of shares
in a portfolio to about 30 and nearly only market risk remains when the portfolio is
increased to resemble the composition of the market as a whole. Solnik's data, from
73
both well-diversified portfolios in domestic and international markets, confirms that an

increase in the number of shares causes the level of risk to become closer to the level
of market risk (Sharpe 1985:172).
3.3.5.2 Studies on the effectiveness of global diversification

Diversification across different capital markets can further reduce risk levels due to
exchange rate movements and due to the fact that different countries have different
economic cycles. As a result, global diversification can lead to lower risk for the same
return or more return for the same risk (Van Horne 1992:63,64).
Grubel (1968)
Grubel applied the principles of portfolio theory to internationally diversified portfolios,
based on two additional assumptions. Firstly, that the only additional risk to consider
is exchange rate movements and, secondly, that there is no increase in transaction
costs. Grubel's study focussed on whether US investors can achieve any benefits by
diversifying internationally. The rnsults of the study confirmed that US investors can
construct international portfolios that render higher returns or bear lower risk in compa-
rison with domestic portfolios (Dobbins et al. 1994:38,39).
Levy and Sarnat (1970)

Levy and Sarnat also examined whether international diversification has any benefits
for US investors. Their study, which ignored dividends and only used exchange rate
movements and capital growth, showed that well-diversified international portfolios do-
minate well-diversified domestic portfolios. Further, the data showed that if international
diversification is not limited only to Canada and Western Europe but also includes
Japan, South Africa and developing countries, then the return achieved can be signi-
ficantly improved for the same level of risk (Dobbins et al. 1994:39-41 ).
Dimson, Hodges and Marsh (1980)

The study of Dimson et al. on the advantages of international diversification for UK in-
vestors showed similar benefits from global diversification. Their data showed that the
74
biggest risk reduction can be achieved when portfolios include investments from Japan
and South Africa, and not only those of Europe and North America. Thus, portfolios
which are diversified globally are likely to be superior to those which are only diversified
across domestic capital markets (Dobbins et al. 1994:41 ).
3.3.6 Criticism of Portfolio Theory
The main objection to Portfolio Theory, as a model to select efficient portfolios, is the
numerous and complex calculations involved. For a portfolio of 100 securities, in the
order of about 5 000 correlation coefficients need to be calculated (Clark et al.
1979:166; Viljoen 1989:22; Linley 1992:3; Dobbins et al. 1994:36). A further objection
is that the assumptions underlying Portfolio Theory, such as that there are no
transaction costs or taxes, securities are completely divisible, investors have equal
access to information, information is freely available and investors have the same time-
horizons, are a major simplification. It allows investment decision-making to be
examined under idealised conditions, but does not consider the practical complexities
of investing in well-diversified portfolios (Linley 1992:3; O'Brien & Srivastava 1995:7,8).
Thirdly, Portfolio Theory states that higher return should be expected on riskier secu-
rities, but does not specify how this risk premium should be calculated. Portfolio Theory
merely states that the risk of securities depends on the relationship between the various
securities' returns (Linley 1992:3,4). A fourth objection to Portfolio Theory is that it is
impractical for most investors to consider investing in proportion to all securities
available on the market. Only investment in some unit trusts or pension funds might
resemble an investment in the market portfolio (Correia et al. 1993: 132).
The problems with estimating probabilities of cash flows in capital budgeting is a fifth
objection to Portfolio Theory, insofar as these complicate the application of Portfolio
Theory to capital budgeting (Clark et al. 1979:166).
75
3.4 THE DEVELOPMENT OF THE EFFICIENT MARKET HYPOTHESIS
Keane (1983:v) states that there are differing views on the efficiency of capital markets.
The behaviour of the vast majority of market participants,

professional and lay, informed and ill-informed, appears to be
premised entirely on the assumption that market prices
'incorrectly' reflect underlying values with sufficient margin and
frequency to justify a policy of actively striving to outperform the
market.
Side-by-side with this activity is an extensive body of research

accumulated over a period of years which, for the main part,
conflicts with the practitioner's basic premise and which suggests
that the conventional investor in search of mispriced securities is
largely a victim of self-deception.
The efficiency of capital markets refers to how successful capital markets are in correct-
ly establishing security prices. Efficient pricing consist of two equally important parts;
the speed of the adjustment of prices and the size and direction of the adjustments. If
the markets are efficient, investors will not be able to systematically earn returns above
the market average (Keane 1983:9).
The EMH suggests that security prices are the best estimate of their investment values
since capital markets adjust prices immediately and fully to reflect new information that
becomes available. This ensures that changes in security prices are unpredictable and
that prices change in a random manner (Sharpe 1985:68).
To fully understand the implications of the EMH and the results of research done there-
on, the concepts of "market efficiency" and "market inefficiency" together with the va-
rious forms, tests and anomalies of market efficiency need to be examined first.
76
3.4.1 The concepts "perfect markets", "efficient markets" and "inefficient mar-
kets"
3.4.1.1 The concept "perfect markets"

According to Keane ( 1993:26) and Rees ( 1995: 176), capital markets need to display the
following attributes to be considered as being perfect markets:
• All investors are rational and base their investment decisions on the principle of
seeking higher returns for lower risk.
• All investors have full and immediate access to all relevant information and the
information is available free of charge.
• There are many buyers and sellers of securities, there are no transaction costs and
taxation has no effect on investment decisions.
Firth (1977:2,3) further states that the requirements for perfect markets are:
• Homogeneity of goods.
• Active trading in all securities.
• Freedom of entry to and exit from the market.
• An unlimited supply of securities.
• Availability of all information relevant to each security.
• Accurate pricing of all securities in accordance with the available information.
In practice, perfect capital markets should correctly price all securities so that each se-
curity's price correctly represents its investment value at all times. No existing capital
market has the attributes nor meet the requirements to be described as a perfect capital
market (Firth 1997:3; Sharpe 1985:67).
3.4.1.2 The concept "efficient markets"

Market efficiency implies that the market price of a security represents all the partici-
pants in the market's consensus value of that security (Van Horne 1992:51 ). To be con-
77
sidered efficient a capital market must instantaneously and unbiasedly incorporate all
available information into the prices of securities. This implies that investors are unable
to achieve abnormal returns on a consistent basis as the capital market does not under-
or overprice securities (Firth 1977: 105; Keane 1983:28; Henderson et al. 1992:292).
The efficiency of capital markets has been the subject of much research, based on the
early work and findings of Fama (1970) who found that changes in security prices oc-
curred randomly due to the correct adjustment to new information. This description of
capital markets has become known as the EMH (Hendriksen & Van Breda 1992: 169;
Van Horne 1992:51; Henderson et al. 1992:291 ).
According to Harrington (1987:26-35) the following assumptions underlie the EMH:
• The objective of investors is to maximise their marginal utility of wealth.

• Investors base their investment decision on the risk/return relationship.
• All investors have the same expectations regarding risk and expected return.
• All investors have a specific (fixed) time-period of investment.
• All relevant information is available to all investors at the same time and at no cost
to them.
The following conditions are deemed sufficient for capital markets to be considered
efficient:
• There are no costs involved in the trading of securities.

• All relevant information is costless and equally available to all investors.
• All investors agree on the implications of the available information for security
prices (Fama 1970:46; Bicksler 1977: 1; Hendriksen & Van Breda 1992: 169).
It can be concluded that if capital markets are efficient then security prices provide the
best possible estimate of the underlying value of the securities (Rees 1995: 176).
78
3.4.1.3 The concept "market inefficiency"

Opposite to the EMH is the possibility that capital markets are inefficient. Market ineffi-
ciency occurs when many investors do not have access to all information or if many in-
vestors incorrectly interpret new information. Further, if investors can predict security
price movements and are able to consistently earn returns in excess of what is con-
sidered normal for the level of risk involved, then capital markets can be considered to
be inefficient (Kam 1990:56; Henderson et al. 1992:292).
'
;'
For ibapital markets to be considered inefficient, the inefficiencies must also be ex-
pl~,ilable, that is, investors must be able to profit from them. Keane ( 1983:24) identifies
Y''~ following criteria to be satisfied for an inefficiency to be exploitable:
/
-·'
• The inefficiency must be identifiable. Even if it can be shown that some investors
or investment strategies can consistently earn abnormal returns, if the inefficiencies
are not identified it will be of no use to the investor.
• Inefficiencies must be material enough to be able to profit from them, to such an
extent that the costs and risks involved in exploiting them are exceeded.
• An inefficiency that has been identified in the past must be shown to continue in the
future. If not, the market may have learned from the past and there may be no
opportunity in future to profit from it.
• All inefficiencies must be shown to be authentic through proper statistical research.
Keane (1983:25,26) adds that if the market inefficiencies can only be perceived by in-
vestors with special insight, the rest of the investors will still consider the market to be
efficient and act accordingly. If the investors with special insight make their knowledge
publicly available, the capital market may act on this information and thus remove the
inefficiency. This means that inefficiencies can only be exploited by the investors who
perceive them and do not communicate their information, otherwise the capital market
becomes efficient and removes the opportunity for achieving abnormal returns. The
effect of the release of new information on security prices in efficient and inefficient
capital markets is illustrated in figure 3.8.
79
Figure 3.8 Reactions of capital markets to new information
Price of the
security
_,,.......-·~·.,·~.::··········· ....................................A
,.. ....
,/ ....:::: ~ ......................................B
.....
.. ......
__________... ......
/::;1·........................................................... c
a Days relative to the day

of announcement
where:
a The day of the release of the new information
A = Inefficient market The market overreacts to the new information and it

takes a number of days for the price to settle at the
true price
B Efficient market The market reacts instantaneously and correctly to the

new information and there is no subsequent movement
in the price
c Inefficient market: The market delays its reaction to the new information
and it takes a number of days for the new information
to be correctly reflected in the price
Source: Ross et al. {1990:340).

80
As can be seen from figure 3.8, investors with special insight have the opportunity to
earn abnormal returns in inefficient capital markets, since security prices take a certain
amount of time to reach their correct levels.
According to the EMH there are varying forms and degrees of capital market efficiency.
These aspects of the EMH is the focus of the next section of this chapter.
3.4.2 Forms and degrees of market efficiency
According to Keane ( 1983: 11) the research of Bache lier ( 1900) was the first reported
tests of market efficiency. It was only in 1959 that Roberts first classified market effi-
ciency according to three levels and these were formalised later by Fama (1970) into
three forms of market efficiency.
Each form of market efficiency relates to a specific set of available information; firstly,
information on past security price movements (weak form), secondly, all publicly avail-
able information (semi-strong form) and, thirdly, all known information, whether publicly
available or not (strong form) (Keane 1983:10; Rees 1995:175).
3.4.2.1 Weak form efficiency

The weak form of the EMH states that current security prices reflect all the information
contained in past security prices. Security prices follow a random walk and a study of
past price movements are of little use in predicting future price movements. Therefore,
only new information will cause security prices to change.
3.4.2.2 Semi-strong form efficiency

According to the semi-strong form of the EMH, security prices reflect all publicly avail-
able information. Security prices reflect new information instantaneously and in an un-
biased manner. Therefore, analysis of all publicly available information will not provide
the investor with the opportunity to earn abnormal returns.
81
3.4.2.3 Strong form efficiency

The strong form of the EMH contends that security prices reflect all information, both
publicly available and privately held. Therefore, this suggests that even the use of
insider information cannot be used to earn abnormal returns (Keane 1983: 10; Ross et
al. 1990:343; Weston & Copeland 1992:94,95; Correia et al. 1993:140; Rees 1995:78).
Keane ( 1983: 10, 11) states further that the three forms are interdependent of one
another. For capital markets to be efficient in the strong form it must also be efficient in
the other two forms. Likewise, for capital markets to be efficient in the semi-strong form
it must also be efficient in the weak form. Most of the evidence for weak form efficiency
seems conclusive, but due to some anomalies there is doubt about semi-strong effi-
ciency. Strong form efficiency, however, is difficult to prove due to lack of opportunity
in studying the effect of insider information. Further, it is quite implausible that security
prices can fully reflect information which is not publicly available (Hendriksen & Van
Breda 1992: 173, 176; Rees 1995: 78).
Keane (1983:26) contends that when market efficiency is considered it is not only the
form of efficiency that is important, but also the d~gree of efficiency.
3.4.2.4 Degrees of market efficiency

Market efficiency can be classified according to the following degrees of efficiency
(Keane 1983:26,27):
• Perfect efficiency: Security prices are so close to their true values that no investors
are able to achieve abnormal returns.
• Near efficiency: Security prices are close enough to their true values that only
investment experts are able to earn abnormal returns. However, these excess re-
turns are only sufficient to compensate them for the transaction costs incurred.
82
• Inefficiency: Securities are so mispriced that even non-expert investors can earn
abnormal returns.
Capital markets can be considered to be perfectly efficient in the weak form, near effi-
cient in the semi-strong forms and inefficient in the strong form. For most practical
purposes capital markets need only be near efficient in the weak and semi-strong form
(Keane 1983:27).
For market efficiency to hold it needs to be empirically verified. The tests used by re-
searchers are classified according to the form of efficiency being tested (Correia et al.
1993:141).
3.4.3 Tests of market efficiency
As future information is not available it is impossible to test the efficiency of capital mar-
kets directly. Therefore, tests must be based on the available information and available
statistical techniques (Dobbins et al. 1994: 16, 17).
Rees ( 1995: 176) identifies two approaches to testing market efficiency:
• The first approach is examining the implications of a specific piece of information

on a security's price. A trading rule is then established and the returns obtained
from trading is then investigated for any abnormal returns.
• Secondly, the performance of a certain group of market participants is investigated
for the possible achievement of abnormal returns.
According to Hendriksen and Van Breda ( 1992: 170, 171) there are also two parts to effi-
ciency that need to be tested:
• Firstly, the speed with which the markets react to new information.
• Secondly, the correctness of the markets' response to the new information.
83
The results of research show that markets do react very quickly to new information, but
it is not clear whether they react correctly. The problem with this is that it is very difficult
to define what a proper response should be. Hendriksen and Van Breda (1992:171)
also identifies two other problems with testing market efficiency.
• The first problem is that most of the tests of market efficiency requires the use of
theoretical pricing models. Therefore, the tests jointly test the efficiency of capital
markets and the efficiency of the pricing models.
• Secondly, a true price against which the market price of securities can be measured
does not exist. Therefore, market efficiency is hard, if not impossible, to prove.
Firth (1977: 118, 119), further, identifies the following factors that need to be considered
when the EMH is tested:
• The tests provide evidence for or against the hypothesis, but cannot be considered
proof of market efficiency or inefficiency.
• The argument is not that the market is perfectly efficient and, therefore, a few in-
efficiencies are expected. The problem is, however, how many inefficiencies are
acceptable.
• The identification of inefficiencies in published research constitute publicly available
information and may cause these inefficiencies to be removed from the market and
prevent their recurrence in the future.
• The results of the tests are a reflection of the efficiency of the market at a certain
point in time and the research needs to be ongoing to establish the overall validity
of the EMH.
3.4.3.1 Weak form tests

Weak form tests of the EMH are concerned with whether past security price movements
can be used to predict future security price movements. Most evidence from studies is
consistent with weak form efficiency, in that no abnormal returns can be earned through
the use of the historic series of prices (Bicksler 1977: 1,2; Keane 1983:33-35; Ross et
84
al. 1990:347; Dobbins et al. 1994: 17). Of the tests that are used are:
• Serial correlation tests: Using correlation coefficients, the relationship between

successive price changes are investigated for possible trends. If the market is weak
form efficient the correlation coefficients would be approximately zero, meaning there
is no consistent correlation between successive price changes.
• Mechanical investment strategies: Strategies such as filter rules and buying and
selling of securities when their prices move beyond their moving average are used
to establish whether higher returns can be earned than from a buy-and-hold strategy
(Correia eta/. 1993:141,142).
3.4.3.2 Semi-strong form tests

These tests are concerned with whether security prices react correctly and swiftly to
new publicly available information. Most of the results of the tests suggest that capital
markets are semi-strong efficient, but some anomalies still cause controversy (Bicksler
1977:2; Firth 1977: 118; Keane 1983: 36,37; Weston & Copeland 1992:95; Dobbins et
al. 1994: 17).
Some information categories that have been tested are news about mergers, capitali-
zation issues, dividend announcements, changes in accounting policies that effect re-
ported earnings and earnings announcements. The tests attempt to establish whether
the price movements can be attributed to the release of the information (Ross et al.
1990:350; Correia et al. 1993: 142).
3.4.3.3 Strong form tests

Tests of the strong form attempt to establish whether investors with access to inside in-
formation can earn abnormal returns and whether certain investors have superior know-
ledge which consistently enable them to earn excess returns. Although there is the
difficulty of obtaining the information about the results of the trading on inside infor-
mation, most evidence suggests that the strong form of the EMH does not hold (Bicksler
85
1977:2; Firth 1977: 118; Ross et al. 1990:354; Weston & Copeland 1992:96; Dobbins
et al. 1994: 17).
Most evidence from tests done on the efficiency of capital markets indicate that inside
information is not impounded in security prices, but that security prices do provide a
quick and correct reflection of publicly available information. This suggests that capital
markets are efficient at the semi-strong level, although anomalies that have been iden-
tified still cast doubt on this conclusion (Bicksler 1977:2; Kam 1990:58; Van Horne
1992:51,52).
3.4.4 Efficient market anomalies
During the 1980s an increasing number of anomalies have been reported and these
anomalies cast some doubt on market efficiency as defined by the EMH (Dobbins et al.
1994: 11 O; Rees 1995:79, 178).
Some of the anomalies that have been recorded include:
• The small firm effect: It appears that in many markets the shares of small firms out-
perform those of larger firms. When the relative risk involved is taken into account,
the expected returns of small firms are much higher than that for large firms. There
also seem to be some link between this size effect and certain calendar effects
(Sharpe 1985:402; Dobbins eta/. 1994:110; Rees 1995:179).
• The PIE effect: Firms with low P/E ratios tend to render abnormal returns when com-
pared with firms with high P/E ratios. These returns seem excessive when compared
with market prices and when the returns are linked to accounting earnings. Evidence
suggests that much of the P/E effect may be explained by the small firm effect
(Dobbins et al. 1994: 11 O; Rees 1995: 180).
86
• The weekend effect: Capital markets show a general tendency to fall over week-
ends and to a greater degree than any other day of the week, except Wednesdays
(Sharpe 1985:408; Rees 1995: 179).
• The January effect: This anomaly is linked to the small firm effect since the shares
of small firms tend to significantly and consistently outperform those of larger firms
during January (Sharpe 1985:405,407).
• Stock market crashes: The crash of October 1987 occurred when no significant
· news was released during the preceding days. This significant drop of between 20%
and 25% seems irrational and not consistent with the concept of market efficiency.
Since the reason for the crash of 1929 is still unknown it does not seem that these
crashes can be explained other than by the theory of the bubble effect of speculative
markets. Share prices must fall back to their true values after a period of speculation
(Ross et al. 1990:353; Dobbins et al. 1994: 121 ).
Although the evidence suggests that there are persistent irregularities which cannot be
explained by the EMH, inefficiencies are expected to be exploitable until they are arbi-
traged away, and this does not seem to be the case. Rees ( 1995: 181) identifies pos-
sible reasons for this:
• The returns from pursuing these irregularities may not be sufficient to compensate
for the costs involved.
• Unexplained returns may be due, firstly, to the fact that what may be considered as
normal returns is difficult to establish and, secondly, to some shortcomings in asset
pricing models.
Correia et al. ( 1993: 140, 141) state that for capital markets to be efficient some investors
need to believe that it is inefficient. Their activities of attempting to exploit inefficiencies
will ensure that capital markets become more efficient. Following on this, it is now ne-
cessary to consider the implications of market efficiency as defined by the EMH.
87
3.4.5 Implications of the EMH
If capital markets are efficient it holds certain implications for investors and investment
advisors. These implications can be summarised in the following manner:
a. Market efficiency implies that abnormal profits can only be achieved by chance
and the best policy for investors is to hold well-diversified portfolios and to mini-
mise transaction costs. This will reduce the risk of their investments underper-
forming (Keane 1983:116).
b. It would be a waste of time for investors to attempt to identify mispriced securities

since the market price of the securities represents the best estimate of the se-
curities' worth (Rees 1995:40).
c. When capital markets are efficient with respect to a specific set of information it
is not worthwhile to attempt to earn abnormal returns by trading on data from that
set of information (Hendriksen & Van Breda 1992:170). This implication can,
further, be sub-divided into three separate implications according to the three
forms of market efficiency.
• Weak form efficient. Past price patterns are already incorporated in cur-
rent market prices and this implies that any form of technical analysis
used to predict future prices will be of no value (Keane 1983: 1O; Ross et
al. 1990:342; Van Horne 1992:51 ).
• Semi-strong form efficient. All publicly available information is already im-

pounded into security prices and, thus, all forms of fundamental analysis
used for trading decisions will not result in the investor being able to earn
abnormal returns (Keane 1983: 1O; Ross et al. 1990: 343; 174; Van Horne
1992:51,52).
88
• Strong form efficient. Security prices reflect all public and private infor-
mation and no-one is able to consistently achieve superior investment
results (Keane 1983: 1O; Ross et al. 1990:343; Correia et al. 1993: 141 ).
d. Due to the concept of market efficiency, the role of investment advisors is to as-
sist their clients in achieving optimum investment performance from their avail-
able resources, and not to attempt to beat the market (Keane 1983: 116).
In addition, the concept of market efficiency also holds some implications for accoun-
ting:
e. Through its effect on share prices the value of any item of accounting information
can be evaluated.
f. The value of alternative accounting disclosure policies and methods can be as-
certained through their impact on share prices.
g. The accounting disclosure policies and methods to be chosen are those which
cause the least long-term variability in share prices (Firth 1977: 140, 141 ).
h. Since capital markets are not fooled by accounting gimmicks or tricks, capital
markets are able to decipher information regardless of its form and only incor-
porates information that affect the expectations regarding the share's risk and
return (Hendriksen & Van Breda 1992: 175, 176).
It is clear that in efficient markets "by chance" and "through luck", not ingenuity, are the
only ways to earn excessive profits on capital markets. Therefore, to prove market in-
efficiency it is not sufficient to show that excessive profits have been earned, but that
it has been earned through skill and on a consistent rather than one-off basis (Keane
1983:14,15).
89
From this, the misconceptions about the EMH need to be considered, as this is an
aspect which is closely related to the implications of the EMH.
3.4.6 Misconceptions regarding market efficiency
Keane (1983:28) states that:
The basic error is the belief that an efficient market is one that should
be able to predict the future. Investors are subject to an illusion of in-
efficiency as a result of a common misconception of how an efficient
market should behave.
Some of the misconceptions about market efficiency that have been identified are listed
below:
a. Frequent changes in security prices are not an indication of market inefficiency,

but rather efficiency as prices are rapidly adjusted to new information becoming
available. A lack of price movements on the other hand might suggest an ineffi-
ciency (Keane 1983:28; Ross et al. 1990:346).
b. Secondly, differences between actual returns and expected returns on securities

do not signify market inefficiencies. Expected return represents a probability
distribution of what is likely to occur, but if subsequent events turn out to be more
favourable or unfavourable than specified by the weighted average of probabi-
lities, actual returns will differ from those expected.
c. Thirdly, since most investors do not hold portfolios of investments which repre-
sent the market as a whole, their investments may outperform the market return.
This does not necessarily imply superior knowledge by the investor, but rather
that economic events do not influence all securities equally and non-represen-
tative portfolios will be affected differently. This is also consistent with market
efficiency (Keane 1983:28,29).
90
d. A fourth misconception is that capital markets are clairvoyant. Market prices are
only a reflection of the knowledge available. With hindsight prices can be shown
to have been incorrect, but that does not imply inefficiency as the EMH simply
states that market prices are set at the correct level according to the information
available.
e. Another misconception is that capital markets would be inefficient if no financial

statements and other accounting information were available. Again, the semi-
strong form of the EMH only asserts that all publicly available information is im-
pounded in security prices. Without financia! statements and accounting infor-
mation the variability and levels of prices may be different, but this too does not
imply inefficiency (Hendriksen & Van Breda 1992: 176, 177).
As it is through empirical testing only that the validity of the EMH can be established,
it is now important to consider the results of some of the research which had been
undertaken.
3.4. 7 Research related to market efficiency
The tests of market efficiency can be divided into three categories (Fama 1976: 136;
Firth 1977: 106), namely:
• Determining whether security prices fully reflect all the information contained in
past prices and price patterns. Also, whether future security prices can be pre-
dicted by using the historic price information.
• Establishing whether security prices react fully and correctly to new publicly
available information. The market's reaction, both in speed and correctness, to
specific items of information is measured.
• Examining whether certain investors have superior knowledge, or access to in-
formation, which allow them to consistently earn above-average returns.
91
From this, results of some of the international research (on the US market, except where
otherwise indicated) conducted until the mid-1980s are reviewed, according to their
area of investigation.
3.4.7.1 Tests of the weak form of market efficiency

Weak form tests of market efficiency are concerned with whether historic prices and
price movements can be used to predict future security prices.
a. Early random walk studies
The earliest research study was that of Bachelier ( 1900) in Franee, who concluded that
commodity prices follow a random walk and that current prices are the best estimate of
future prices. Working (1934) and Cowles and Jones (1937) confirmed the finding that
commodity prices and security prices follow a random walk. Kendall (1952), in the UK,
suggested that security and commodity prices follow a random walk since he failed to
find cycles in these prices, but found that price changes are independent of previous
price changes. Both Roberts (1959) and Osborne (1959) also concluded that security
prices appear to follow a random walk. Cootner (1962) agreed that security prices
follow a random wall<, but argued that it is not valid for all investors since some pro-
fessional investors are able to earn abnormal returns. Finally, Samuelson (1965)
confirmed that security prices change in a random manner and that capital markets are
at least weak form efficient (Firth 1977: 120, 121; Dobbins et al. 1994:70-73).
b. Serial correlation analysis
The correlation between a security's current return and its future return has been the
subject of a large body of research.
Moore ( 1964)
Moore examined the serial correlation between successive price changes of individual
shares and found an average serial correlation coefficient which is not statistically
92
different from zero. Moore concluded that there is little or no dependence between
successive price changes and that any returns obtained from such dependence would
be insufficient to compensate for the trading costs involved. Therefore, historic price
changes cannot be exploited to achieve above-normal returns (Firth 1977: 123; Ander-
son 1978:19; Dobbins et al. 1994:73).
Granger and Morgenstern (1963) and Godfrey, Granger and Morgenstern (1964)
Spectral analysis, a complex serial correlation technique, was used in an attempt to find
cycles in share prices. No significant relationship between successive price changes
was found since the serial correlation was found to be close to zero (Fama 1970:54;
Firth 1977: 125; Dobbins et al. 1994:72).
Fama (1965)
In a major study, Fama found an average serial correlation coefficient of 0.03. This is
statistically not different from zero and confirmed that share price changes follow a
random walk and that information on the history of the prices cannot be used to earn
abnormal returns (Firth 1977:124; Dobbins et al. 1994:74).
Brearly (1970)
In a study in the UK, Brearly found some degree of serial dependence in share returns.
The regularity found, however, is not sufficient to be exploited after transaction costs
are taken into account and as such does not provide evidence against the random walk
hypothesis and weak form market efficiency (Rees 1995: 177)
Solnik (1973)
Solnik tested seven European stock markets and found greater serial dependence than
that found in the US market. This dependence too proves to be insufficient to provide
abnormal returns when transaction costs are taken into account (Keane 1983:48; Ross
eta/. 1990:347; Rees 1995:176,177).
93
Hagerman and Richmond (1973)

Hagerman and Richmond conducted a similar serial correlation analysis, but concen-
trated on smaller and thus less well analysed shares. They also found little serial corre-
lation and concluded that capital markets are weak form efficient for even the smal I type
firms (Firth 1977: 124).
Various studies
Various other studies obtained similar findings in that share prices largely follow a
random walk and that little or no correlation can be found between past prices and
future prices of shares. Some of the evidence, though, suggest that infrequently traded
shares' prices do not follow a random walk. Some of these studies are those of Praetz
(1972) in Australia, Niarchos (1972) in Greece, Dryden (1970), Cunningham (1973)
and Guy (1975) in the UK, Conrad and Juttner (1973) in Germany, Jennergrean and
Korsvold (1975) in Norway and Sweden, Ang and Pohlman (1978) who found the
Japanese market to be highly efficient in the weak form, Errunza (1979) in Brazil and
D'Ambrosia (1980) in Singapore (Frith 1977:124; Gay 1982:189; Keane 1983:43).
According to Keane (1983:48) smaller capital markets, such as Nigeria, Singapore and
South Africa, have shown significant levels of weak form inefficiency. This arises from
the size of these markets, the thinness of trading and weaknesses in disclosure of infor-
mation. However, no evidence has as yet been presented that these inefficiencies are
exploitable and thus the evidence presented regarding smaller markets is inconclusive.
Keane (1983:49) concludes that it is doubtful that the inefficiencies in smaller markets
would present investors with abnormal profit opportunities. The major markets (US,
Europe and Japan) are substantially efficient and, through the flow of capital across
international borders, exploitable inefficiencies would be eliminated.
94
c. Mechanical investment strategies
Another area of research regarding weak form efficiency has been whether mechanical
investment strategies, which provide signals for buying and selling of securities, enable
investors to achieve abnormal returns.
Alexander (1961)
Alexander studied a filter rule which provides buy and sell signals when share prices
move a certain percentage from a previous low and high point. He established that
there are trends in share price movements, but also found that any abnormal returns
disappear when transaction costs are taken into account (Firth 1977: 126; Keane
1983:35; Dobbins et al. 1994:72).
Fama and Blume (1966)

Fama and Blume found similar evidence in that small abnormal returns can be earned
with frequent trading. These excess returns are not sufficient to cover transaction costs
and, in fact, a simple buy-and-hold strategy at the same level of risk can outperform tra-
ding based on the filter technique (Firth 1977: 126; Weston & Copeland 1992:95).
Cootner (1962), Van Home and Parker (1967) and James (1968)
All three studies investigated whether it is possible to earn abnormal returns when
shares are bought or sold when their prices moved away from their moving average.
The evidence showed that this is not a profitable investment rule when transaction costs
are taken into account (Firth 1977: 126, 127).
Latam~ and Young (1969) and Evans (1970)

Both studies investigated the investment strategy of adjusting portfolios at the end of
specific investment periods so that the portfolio weights are the same as at the be-
ginning of the period, that is, the same proportion of the portfolio is invested in the same
shares. The results of the studies showed that this investment strategy does not outper-
form a simple buy-and-hold strategy (Firth 1977:127; Keane 1983:35).
95
Levy (1967) and Jensen and Bennington (1970)

These researchers studied the strategy of ranking shares in terms of their price per-
formance over a specific investment period. Shares from a certain percentage at the
bottom of the ranking are sold and the rest are retained. Despite claims to the contrary,
the results of the studies showed that this does not outperform a si_mple passive strate-
gy (Firth 1977:127; Keane 1983:35).
Black and Scholes (1972)

In an investigation of the options market, Black and Scholes concluded that options are
significantly mispriced. They also found that this cannot be profitably exploited by a tra-
ding rule due to the high transaction costs that would be incurred (Firth 1977: 124 ).
Various studies
Studies by Praetz (1969) in Australia and Dryden (1970) in the UK also found that no
abnormal returns can be earned by using filter techniques (Firth 1977: 126).
Firth ( 1977: 127, 128) adds that many other mechanical investment strategies have been
tested of which none can consistently outperform the market. Yet it can be accepted that
investors who have profitable trading rules would keep it a secret and, hence, it is only
those strategies which have been published that can be examined through empirical
testing.
It can be reasonably concluded from the results of research done that:
• Security prices follow a random walk.

• The future prices of securities cannot be predicted by using information about the
past prices and price changes of the securities.
• Mechanical investment strategies do not provide returns that exceed those of a
passive buy-and-hold strategy.
• Therefore, capital markets are for all intents and purposes efficient in the weak
form.
96
3.4.7.2 Tests of the semi-strong form of market efficiency

Tests of semi-strong efficiency are concerned with the reaction of capital markets to
new information. These tests attempt to establish whether security prices react quickly
and correctly to the new information and, if they do, it implies that investors would not
have the opportunity to earn abnormal returns through trading on the basis of the infor-
mation. Most of the research studied the impact of specific economic events and speci-
fic types of information on security prices to establish whether capital markets are semi-
strong efficient.
a. Annual earnings announcements
Ball and Brown ( 1968)

Ball and Brown studied the impact of annual earnings announcements on share prices.
They found that the prices move gradually during the year and that by the time of the
announcements, approximately 90% of the content of announcements are already
incorporated in the prices. This provides support for the semi-strong form of the EMH
since the market anticipates the information contained in annual reports and it is not
possible for investors to use annual earnings announcements for profitable trading
(Fama 1970:70; Keane 1983:37,39; Weston & Copeland 1992:95,96; Henderson eta/.
1992:296,298; Dobbins et al. 1994:74, 75).
Beaver, Clarke and Wright (1979)

In this study on annual earnings announcements, Beaver et al. not only studied whether
the changes in share prices are in the right direction, but also whether it is of the correct
magnitude. It was found that the market continuously adjust prices and that the market
is quite sensitive to the magnitude of changes in earnings (Keane 1983:39,40).
Beaver, Lambert and Morse (1980)

The Beaver et al. study confirmed that capital markets are able to anticipate earnings
announcements and that its effect is to a large degree already incorporated in share
prices (Keane 1983:40).
97
Various studies
In an Australian study, Brown (1970) found similar evidence to those of US studies.
Later Australian studies by Brown and Hancock (1977} and Brown, Finn and Han-
cock (1977), and a New Zealand study by Emanuel (1984) confirmed the results
obtained in previous studies (Henderson et al. 1992:298).
b. Earnings announcements by similar type firms
Firth (1977)
In this UK study, Firth (1977:133) examined the impact on share prices when similar
type firms release their earnings results. He found that price reactions are in the
direction expected, but was unable to conclude whether the magnitude of changes are
correct. This evidence supports the semi-strong form of market efficiency since capital
markets use all relevant publicly available information to establish share prices.
Foster (1981)
Foster investigated the behaviour of share prices when firms of similar type make an-
nouncements about earnings. He found that the share prices move more than normal
when firms in the same industry announce their earnings (Henderson et al. 1992:303).
c. Dividend announcements
Pettit (1972)
Pettit examined the market's reaction to dividend announcements. The results showed
that the market reacts quickly and that the price adjustments are over immediately after
the announcements. Pettit concluded that this confirmed the hypothesis about strong-
form market efficiency (Firth 1977:131; Dobbins et al. 1994:77).
Watts (1973)
Although Watts disagreed with Pettit (1972) about the importance of dividend changes,
his results confirmed those obtained by Pettit (1972) (Firth 1977: 131 ).
98
d. Share splits
Fama, Fisher, Jensen and Roll (1969)

Fama et al. found results consistent with semi-strong market efficiency when they in-
vestigated the response of share prices to share splits. Share splits constitute an in-
crease in the number of shares per shareholders and as such has no new information
content. However, it was found that share splits are usually followed by an increase in
dividends and that the capital market reacts to the splits in anticipation of this increased
dividends. They found that share prices rise before the announcement of splits, but no
abnormal returns could have been earned at the time of the announcement since the
prices show almost no movement after the announcement, meaning that the price
adjustments are over at that time (Fama 1970:67,69, 70; Anderson 1978:25,26; Keane
1983:40,41; Rosseta/.1990:350; Dobbinseta/.1994:75).
Various studies
The studies of Hausman, West and Largay (1971) and Reilly and Drzycimski (1981)
provided further confirmation of semi-strong efficiency in the case of share splits.
Charest (1978), however, differed from the above conclusions as he found large po-
sitive abnormal returns during the period surrounding the announcement of share splits
(Van Rhijn 1994:28).
e. Capitalisation issues
Firth(1973, 1974, 1977)

Firth examined the impact of capitalisation issues on share prices in the UK, and found
that this has no impact on share prices. This confirmed that markets react rationally and
supports semi-strong efficiency (Firth 1977:134).
Various studies
Studies by Johnson (1966) and in Canada by Finn (1974) also found that no abnormal
returns can be earned based on the announcements of capitalisation issues. Their
99
results showed that the market reacts quickly and accurately to the information re-
garding capitalisation issues (Firth 1977: 131 ).
f. Large block trades
S.choles (1972)
Scholes investigated the impact of secondary offerings (large underwritten sales of
existing shares) on share prices. The findings showed that these trades are associated
with small declines in prices. He found that the declines do not depend on the size of
the trade, but on the seller of the shares. The largest declines are associated with
corporations and corporate officers and this implies that other investors suspect that
insider information is involved. This supports semi-strong form market efficiency as the
market is more concerned with the information content of the sale than with the amount
of the sale (Fama 1970:71; Anderson 1978:27; Dobbins et al. 1994:76).
Kraus and Stoll (1972)

Kraus and Stoll also found an effect on share prices when they investigated the impact
of all block trades. Share prices decline after the trade but recover quickly thereafter,
and they concluded that it is not possible to earn abnormal returns by acting on publicly
available information regarding block trades (Dobbins et al. 1994:76, 77).
Dann, Mayers and Raab (1977)

This study examined the impact of large block trades on the assumption that the sellers
or buyers have access to special information. They found that block trades have an
impact on share prices, but investors must act within five minutes of the trade to earn
returns sufficient to cover transaction costs (Keane 1983:41 ).
g. New share issues
Scholes (1969)
In a study similar to those on share splits, Scholes found evidence that supported the
100
semi-strong form of the EMH. As with share splits, new issues suggest favourable eco-
nomic conditions for firms and prices tend to rise in the period preceding the new issue.
Prices behave in a random manner after the issue which suggests that the market in-
corporates all the information into the prices (Fama 1970:71 ).
h. Exchange listings
Furst (1970)
The study of Furst examined whether listing has any effect on share prices. In general
no impact was found, suggesting that listing is not deemed by investors as representing
a change in the value of the firm. As a result no abnormal returns can be earned by
investing at the time of the listing (Van Rhijn 1994:29).
Van Horne (1970)

Van Horne also studied new exchange listings in order to establish whether it is pos-
sible to profit from them. His results showed that prices move significantly upwards for
a period before the listing, but, when transaction costs are taken into account, no abnor-
mal returns can be earned after listing (Van Rhijn 1994:30).
Ying, Lewellen, Schlarbaum and Lease (1977)

Contrary evidence arose from the study of Ying et al. who found that abnormal returns
can be earned during the month preceding the listing as well as the month thereafter.
The abnormal returns persist after providing for reasonable transaction costs and re-
present evidence against semi-strong form market efficiency (Van Rhijn 1994:30).
McConnell and Sanger (1981)

The results of a subsequent study by McConnell and Sanger confirmed the findings of
Ying et al. (1977).
From the foregoing, it is evident that the results of studies regarding new listings and
the semi-strong form of market efficiency are mixed, since the more recent research
101
provide little support for the semi-strong hypothesis (Van Rhijn 1994:31 ).
1. Mergers and takeovers
Mandelker (1974)
Mandelkerfound that capital markets generally anticipate mergers and that prices begin
to adjust approximately eight months before the actual event (Dobbins et al. 1994:77).
Various studies
The UK study of Firth ( 1977: 133) found that capital markets react quickly and correctly
to the announcement of takeovers. Hong, Kaplan and Mandelker ( 1978) found similar
results which support the EMH in the semi-strong form (Keane 1983:41 ).
j. Accounting for depreciation
Archibald (1972)
The study of Archibald examined the behaviour of share prices when firms change their
methods of providing for depreciation. These changes are for accounting purposes only
and have no tax effect and, therefore, no effect on cash flows. The results of the study
showed that the changes have no effect on share prices, which is consistent with semi-
strong market efficiency (Wolk et al. 1989:208; Hendriksen & Van Breda 1992: 170; Van
Rhijn 1994:33,34).
Kaplan and Roll (1972)

Kaplan and Roll also investigated changes in accounting for depreciation with no tax
effect. Their results showed that the market is not fooled by the resulting increase in
accounting earnings and these charges have no significant effect on prices (Dobbins
et al. 1994:76).
102
Beaver and Dukes (1973)

Beaver and Dukes investigated whether using different methods for the provision for
depreciation have any effect on prices. They adjusted the earnings for different firms to
the same depreciation basis and found that these adjustments have no significant effect
on the price/earnings relationship between the various firms (Wolk et al. 1989:207,208).
Cassidy (1976)
Cassidy also studied changes in depreciation with no tax effect. The results were con-
sistent with semi-strong market efficiency as Cassidy found no effect on prices and no
abnormal share returns were observed (Henderson et al. 1992:313).
k. LIFO inventory valuation
Sunder(1973, 1975)
Sunder studied whether a change to LIFO inventory valuation, with the resulting cash
flow and tax effects, has any effect on share prices. Consistent with semi-strong market
efficiency, prices increase in anticipation of the change, while no abnormal returns can
be earned after the change (Dobbins et al. 1994:77).
Various studies
Findings similar to those of Sunder (1973, 1975) were found by Ball (1972) and Biddle
and Lindahl (1982). The Abdel-Khalik and McKeown (1978) and Brown (1980)
studies contradicted these findings since they found either no response or a negative
response to the LIFO change. This suggests that previous studies had not correctly
isolated the effect of the LIFO change (Wolk et al. 1989:208,209).
I. Changes in the Federel Reserve discount rate
Waud (1970)
The study of Waud provides additional support for the EMH in the semi-strong form.
Waud found significant price changes on the day of the announcement and that the
103
market even anticipates the announcement by a few days. The price changes, however,
do not to persist after the day of announcement (Anderson 1978:28).
m. The effect of extraordinary items on earnings
Eskew and Wright (1976) and Gonedes (1978)

Eskew and Wright investigated whether changes in share prices are associated with the
announcement of unexpected extraordinary items. The results showed such an asso-
ciation and support semi-strong form market efficiency. Gonedes, though, found no
significant association (Henderson et al. 1992:302).
n. The announcement of large investment holdings
Firth (1975)
Firth's UK study examined the impact on share prices of the announcement of large
investment holdings being built up in firms. The behaviour of prices after the announce-
ment supports semi-strong efficiency, but insider trading was suspected as prices re-
acted irrationally in the lead up to the announcement (Firth 1977: 132, 133).
o. Significant world events
Reilly and Drzycimski (1973)

Reilly and Drzycimski examined the announcement of seven unexpected world events
in an attempt to find market inefficiency. The results support semi-strong market effi-
ciency as the market reacts swiftly and appropriately to the announcements. It was also
found that investors are not able to earn abnormal returns on the trading day following
such an announcement (Van Rhijn 1994:31,32).
104
p. Tipsters
Firth (1972)
In this UK study, Firth found that share prices adjust swiftly to new information contained
in the press comments of tipsters. This response is quick enough to prevent profits for
any investors who act on the advice, although the tipsters themselves may be able to
profit before announcing their advice (Keane 1983:43).
Most of the event and information studies have shown that the capital market can in ge-
neral be considered to be semi-strong efficient. Although mixed results were obtained
in some areas, it can be concluded that in general capital markets:
• React swiftly to new information.

• Adjust security prices correctly.
• Respond to new publicly available information in such a manner that investors
are not able to consistently outperform the market average.
3.4.7.3 Tests of the strong form of market efficiency

Tests of the strong form version of the EMH are designed to examine two issues:
• Whether certain investors have superior knowledge and, hence, have superior
success in trading on capital markets.
• Whether access to privileged or inside information can be used by investors to
obtain abnormal returns.
a. Mutual funds and unit trusts
Friend, Brown, Herman and Vickers (1962)

In the pioneering study on the performance of mutual funds, Friend et al. found that
mutual funds on average do not outperform a passive buy-and-hold strategy. This im-
plies that mutual fund managers in general do not have superior knowledge and they
105
concluded that this evidence supports the EMH in the strong form (Anderson 1978:30;
Dobbins et al. 1994:78).
Sharpe (1966)
The study of Sharpe also provides support for the efficiency of capital markets and
found that mutual fund managers on average cannot outperform the market. Sharpe
concluded that the fund managers should rather concentrate on evaluating risk and on
efficient diversification (Anderson 1978:30; Dobbins et al. 1994:78).
Jensen (1968)
Jensen found no evidence of mutual funds having superior performance and when
management expenses are taken into account they actually perform worse than
randomly selected portfolios. Further, successful performance in the past should not be
taken as an indicator of future success (Fama 1970:75, 76; Anderson 1978:30; Keane
1983:43; Ross et al. 1990:351,352; Dobbins et al. 1994:78, 79).
Friend, Blume and Crockett (1970)

Further confirmation that fund managers are unable to outperform unweighted random
portfolios were provided by Friend et al. (Anderson 1978:31; Dobbins et al. 1994:79).
Firth (1977)
Firth studied the performance of unit trusts in the UK. His findings support those of US
research, in that many unit trusts actually have significantly inferior performance and,
therefore, that the managers of unit trust do not have superior knowledge or access to
non-publicly available information (Firth 1977: 137; Dobbins et al. 1994:79,80).
Various studies
Studies by Treynor(1965) and Williamson (1972) and UK studies by Ward and Saun-
ders (1976) and Cranshaw (1977) provide further evidence in support of market
efficiency. They concluded that these funds do not perform consistently and in general
perform worse than the market average (Keane 1983:43; Van Rhijn 1994:50,51 ).
106
b. Forecasting ability
Fitzgerald (1974)
In this UK study, Fitzgerald found that portfolios constructed on the advice from certain
advisers render returns that vary abnormally above and below the market average. He
concluded that the UK market seems less efficient than the US market, but added that
further research was required to confirm his findings (Firth 1977: 134, 135).
Ambachtsheer {1974) ,
Ambachtsheer apparently also found evidence that some investors have a special fore-
casting ability. He did qualify his findings, by adding that forecasting ability is hard to
find since these investors may have had a lucky run, meaning that it may be due to
chance and not forecasting ability (Dobbins et al. 1994:81 ).
Dimson and Marsh (1984)

The UK study of Dimson and Marsh supported Ambachtsheer's 1974 findings, in that
they found that some analysts have a special forecasting ability. They showed that
these analysts' forecasts can be used to earn abnormal returns, although the investors
who receive their recommendations have to act quickly to profit. Dimson and Marsh
qualified their findings by adding that this is not conclusive evidence against market
efficiency since the analysts may have access to inside information (Dobbins et al.
1994:68; Rees 1995:79, 178).
Penman {1980)
Penman investigated the association between earnings forecasts by management and
share returns. The results of the study indicate a strong effect on the day before the
announcement and that abnormal returns are noticeable on the day of the announce-
ment (Henderson et al. 1992:302).
107
Imhoff and Lobo (1984)

The Imhoff and Lobo study found that abnormal returns are associated with analysts'
forecasts, especially when management's own forecasts are not publicly available
(Henderson et al. 1992:302).
Sinclair, Fatseas and Trotman (1986)

The Australian study of Sinclair et al. on forecasts by chairmen of companies found si-
milar results to those of Penman (1980). It was found that abnormal returns are asso-
ciated with announcements of both good and bad news (Henderson et al. 1992:302).
Various studies
Several studies examined the performance of individual investment advisers. The
studies of Craig and Malkiel (1968), Elton and Gruber (1972) and Firth (1972), in the
UK, found that few advisers can claim abnormal performance with their forecasts (Firth
1977:137).
c. Inside information
Niederhoffer and Osborne (1966)

The study of Niederhoffer and Osborne found that stock exchange specialists can earn
abnormal profits when they use their access to information about buying and selling
orders which at that point in time have not yet been filled (Firth 1977: 135; Van Rhijn
1994:49, Dobbins et al. 1994:80).
Reilly and Drzycimski (1975)

The Reilly and Drzycimski study found that stock exchange specialists can trade pro-
fitably on selling and buying directories received after the announcement of significant
unexpected world events (Van Rhijn 1994:49).
108
Various studies
Various studies on the use of insider information found that this information can be used
to earn significant abnormal returns. Although it is difficult to obtain information about
insider trading, studies by De Vere (1968), Jaffe (1974), Collins (1975), Finnerty
(1976) and Seyhun (1986) found sufficient evidence to conclude that the strong form
of market efficiency should be rejected. Share prices do not reflect non-publicly avail-
able information and, hence, such information can be used to trade profitably (Firth
1977:135,136; Keane 1983:36; Ross eta/. 1990:354).
From the available evidence, the following conclusions can be drawn regarding the
strong-form of the EMH:
• Mutual fund and unit trust managers do not have special knowledge or.access
to special information and therefore cannot earn superior returns.
• Although the evidence regarding forecasting ability is mixed, it still casts doubt
on the efficiency of capital markets.
• Corporate insiders and stock exchange specialists do have access to information
that can be used to earn above-average returns.
• Capital markets are therefore not efficient in the strong form, but the evidence
about mutual funds and unit trusts supports semi-strong form efficiency.
3.4.7.4 Tests on efficient market anomalies

A number of anomalies have appeared that cast some doubt on the efficiency of capital
markets. The tests on these anomalies have attempted to determine whether they
persist or are one-off in nature and, whether the anomalies are exploitable.
a. Quarterly earnings announcements
Jones and Litzenberger (1970)

The Jones and Litzenberger study showed that when a mechanical investment strategy
is used on quarterly earnings announcements, the market can be outperformed and,
109
hence, excess returns can be earned (Firth 1977: 132).
Jones (1970)
Jones extended his 1970 study with Litzenberger by using filter rules and again found
that abnormal returns can be earned. The implication is that capital markets do not
correctly adjust prices to the information contained in quarterly earnings announce-
ments (Firth 1977: 132).
Brown and Kennelly ( 1972)

The study of Brown and Kennelly examined the association between share price
movements and quarterly earnings reports. The results of their study showed that ab-
normal returns can be earned by using the information contained in such reports
(Henderson et al. 1992:298,299).
Joy, Litzenberger and McKnally (1977)

The Joy et al. study showed that favourable information contained in quarterly earnings
announcements is not instantaneously reflected in share prices. They concluded that
price changes after the announcements are influenced by the size of the unexpected
announcements (Van Rhijn 1994:35,36).
Foster (1977), Brown and Hancock (1977) and Brown, Finn and Hancock (1977)
Foster's study showed significant movements in prices on the day before the announce-
ments as well as on the day of the announcement. Similar results to those of Foster
( 1977) were obtained in the 1977 Australian studies by Brown and Hancock, and Brown,
Finn and Hancock (Henderson et al. 1992:299).
Ball (1978)
The results of Ball's study indicate that abnormal returns can be earned through trading
based on quarterly earnings reports, but he concluded that this is due to problems with
the asset pricing model rather than an indication of market inefficiency (Van Rhijn
1994:36,37).
110
Watts (1978)
Watts reviewed Ball's 1978 study and stated that the abnormal returns were in fact due
to market inefficiency and not misspecification of the asset pricing model (Van Rhijn
1994:37). The study of Watts covered the period 1962 to 1968 and he found that ab-
normal returns could only have been earned in the 1962 to 1965 period. Thereafter no
significant abnormalities were observed, which suggests that the market had adjusted
to the inefficiency and that it is reasonable to assume that it will not recur after 1965.
Further, the abnormal returns during the 1962 to 1965 period were insufficient to cover
transaction costs and, thus, Watts concluded that it is not reasonable for investors to
expect to earn abnormal returns based on quarterly earnings announcements (Keane
1983:54,55).
b. The firms size effect
Banz (1981) and Reinganum (1981)

Both studies showed that small firms consistently provide investors with larger returns
than do large firms. Neither study attempted to provide a reason for this phenomenon,
although both contended that it is not due to market inefficiency, but rather due to
misspecification of the CAPM. Reinganum also contended that the size effect provides
an explanation for the price/earnings effect (Dobbins et al. 1994: 111, 113; Van Rhijn
1994:38,39).
Dimson (1979) and Roll (1981)

Dimson and Roll also noted the small firm effect, but both contended that this pheno-
menon is due to the incorrect measuring of small firms' risk. Standard risk measures do
not take into account the infrequent trading of small firms' shares. leading to the under-
statement of their risk (Keane 1983:55,56; Van Rhijn 1994:39).
Reinganum (1982)
This study by Reinganum supports the results of the Dimson (1979) and Roll (1981)
studies, but he found that the underestimation of risk does not completely explain the
111
small firm effect (Keane 1983:56; Van Rhijn 1994:40).
Various studies
Several studies also found evidence supporting the size effect, whereby shares of small
firms outperform those of large firms. These studies include those of Ibbotson and
Sinquefield (1982), Blume and Stambaugh (1983), and Braun, Kleiden and Marsh
(1983) in Australian markets, Berges, McConnell and Schlarbaum (1984) in Canadian
markets and in Japan by Nakamura and Terada (1984) (Sharpe 1985:403,404;
Dobbins et a/.1994: 111, 112).
c. The price/earnings (P/E) effect
Basu (1977)
The results of Basu's study showed that firms with low P/E ratios tend to yield higher
returns than expected. This would imply that there is a relationship between the
historical P/E ratio and future market performance. Obviously, this is in conflict with
market efficiency since this provides investors with the opportunity to earn abnormal
returns (Dobbins et al. 1994:113; Van Rhijn 1994:37,38).
Various studies
The studies of Nicholson (1968) and Oppenheimer and Schlarbaum (1981) confirm
the P/E effect. Reinganum (1981) clearly showed that the PIE and size effects are
closely related and that it is difficult to separate the impact of these two effects (Keane
1983:55; Dobbins et al. 1994: 113).
d. The weekend effect
Cross (1973)
An early study that showed that returns on capital markets are not random was that of
Cross, who found significant lower returns for the period between the close of the mar-
ket on Fridays to the close of the market on Mondays (Rees 1995: 179).
112
French (1980)
The study of French also found evidence of the weekend effect and the results showed
that returns on shares are abnormally high on Fridays and, opposed to this, returns are
negative on Mondays (Ross et al. 1990:353).
Rogalski (1984)
Rogalski's study showed that markets on average fall over the weekend. He found that
the weekend effect is too small too exploit by selling shares just before the close of the
market on Friday and then buying them at a lower price on Monday. Rogalski did, how-
ever, find that it would be wise not to buy shares on Friday afternoons or to sell shares
early on Monday mornings (Sharpe 1985:408).
e. The January effect
Rozeff and Kinney (1976)

The research of Rozeff and Kinney showed that share prices tend to fall towards the
end of the year and start to rise in January (Dobbins et al. 1994:114).
Keim (1983)
Keim found a significant relationship between the size effect and the January effect.
The study found that small firms show abnormally high returns in January, especially
during the first five days of January. Further, it was shown that approximately 50% of
the size effect occurs during January (Ross et al. 1990:353; Dobbins et al. 1994:114).
Roll (1983)
Roll's study confirmed the existence of a January effect. In addition to this, the results
showed that firm size has an additional effect in January (Van Rhijn 1994:45).
Grossman and Sharpe (1984)

Grossman and Sharpe found a significant effect for small firms during January. The
results seem to indicate that the entire small firm effect occurs during January, in
113
particular on the last day of December and the first four days of January (Sharpe
1985:405).
DeBondt and Thaler (1985)

The study of DeBondt and Thaler confirmed the existence of the J.anuary effect and, in
addition, identified another aspect, other than the size effect, that occurs in conjunction
with the January effect. They found that portfolios consisting entirely of shares that had
performed poorly during the preceding year show excess returns during January, while
portfolios of shares that had performed well earn less than the market average during
January (Dobbins et al. 1994: 116).
f. Other calendar effects
Various studies
Other calendar anomalies that have been identified are small and cannot be profitably
exploited by investors due to the transaction costs involved. However, they still provide
evidence against market efficiency and include irregularities related to the time of day,
the study of Harris (1986), and the day of the week, the studies of Gibbons and Hess
(1981), Lakonishok and Levi (1982), Keim and Stambaugh (1984) and Jaffe and
Westerfield (1985) (Dobbins et al. 1994: 115, 116).
g. Price to book value/Book value to market value (BV/MV) ratio
Rosenberg, Reid and Lanstein (1985)
Rosenberg et al. identified another irregularity which is closely related to the small firm
and price/earnings effects. They found that investors can earn abnormal returns if they
invest in firms which have a low share price to book value of shares ratio (Dobbins et
al. 1994: 113).
114
h. Market volatility
Various studies
Schiller (1981) investigated the reaction of the market to changes in dividends. In effi-
cient markets, share prices should change with the release of new information, but
Schiller found that the market does not act correctly and that this leads to excess
volatility in share prices (Keane 1983:64; Ross et al. 1990:352,353; Dobbins et al.
1994:126). Le Roy and Porter (1981) found evidence that supported Schiller's conclu-
sions regarding excess volatility. Studies by Flavin (1983), Marsh and Merton (1986)
and Kleidon (1986), however, criticised the results obtained by Schiller (1981) as being
a result of limitations in the model applied (Dobbins et al. 1994: 126).
French and Roll (1986)

The study of French and Roll examined market volatility on the New York Stock Ex-
change (NYSE) during a period in 1968 when the market was closed on Wednesdays.
They found lower volatility on Wednesdays when the market was closed, which sug-
gests that market volatility is due to trading rather than to the release of new infor-
mation. This finding is not consistent with market efficiency (Dobbins et al. 1994: 116).
Although some of the anomalies identified cannot be profitably exploited by investors,

the existence of anomalies is puzzling and casts doubt on market efficiency, especially
on the semi-strong version of the EMH.
Dobbins et al. (1994:121) conclude that the stock market crash of October 1987
suggests that capital markets are irrational and cannot be considered to be consistently
efficient. Markets around the world fell by nearly one-third, although no major release
of new information (bad news) took place during the days preceding the crash.
A limited number of South African studies have also addressed the issue of the effi-
ciency of the JSE, that is, the JSE Securities Exchange SA [previously known as the
Johannesburg Stock Exchange (JSE)), and to date no definite conclusions can be
115
reached from these studies as regards the level of efficiency of the South African
market. The following is a summary of the results of some of the studies.
Affleck-Graves and Money (1975)

These researchers studied industrial shares and concluded that, even though the price
changes are not totally independent, the JSE is at least weak form efficient (Van Rhijn
1994:7).
Gilbertson and Le Roux (1977)

Gilbertson and Le Roux studied both the performance of trading rules and mutual funds.
They found that the trading rules do not outperform a buy-and-sell strategy, nor can the
mutual funds consistently outperform the market. They concluded that the evidence
from their study support the efficiency of the JSE at both the weak and semi-strong
levels (Linley 1992: 14; Van Rhijn 1994:22).
Gilbert and Vermaak (1982)

The Gilbert and Vermaak study also investigated the performance of mutual funds and
found that the JSE is not efficient in the strong form (Linley 1992:14).
Knight and Affleck-Graves (1983; 1985)

The 1983 study on market efficiency and the change to LIFO found the JSE to be ineffi-
cient at the semi-strong level since the market did not react correctly to the change in
accounting earnings. In their 1985 study, however, Knight and Affleck-Graves found
evidence supporting market efficiency and they concluded that the JSE is approaching
efficiency at the semi-strong level (Linley 1992:14; Correia et al. 1993:143; Van Rhijn
1994:7,32,33).
De Villiers, Lowings, Pettit and Affleck-Graves (1986)

The De Villiers et al. study supported the findings on the size effect anomaly and con-
cluded that the size effect also occurs on the JSE (Van Rhijn 1994:41 ).
116
3.5 SUMMARY
This chapter examined the main capital market theories - Portfolio Theory and the Effi-
cient Market Hypothesis - and their impact on investment decision-making.
In the first section of this chapter investment on capital markets, the effect of risk, return
and uncertainty, and the impact of utility theory on investment decision-making were
examined. Through this it was established that risk is the major factor to consider when
investing in capital markets and that it can be concluded that most investors prefer less
risk to more and have diminishing marginal utility for wealth.
Portfolio Theory and the effect of diversification on investment risk and return were exa-
mined in the second part of the chapter. The results showed that diversification can
reduce the level of risk of a portfolio and that the application of utility theory enables
investors to construct optimal portfolios. Through the review of the research related to
Portfolio Theory it was established that diversification does have definite investment
value, even more so when done on an international scale.
The final section of this chapter examined the EMH and the implications of market effi-
ciency for investment decision-making. The various degrees and forms of market effi-
ciency, its implications and misconceptions together with certain anomalies were exa-
mined. In the last section, the extensive body of research related to market efficiency
was reviewed. The results of the tests and research done showed that, for the most
part, capital markets are efficient at a weak and semi-strong level but not at a strong
level. Certain anomalies identified still require further investigation.
The next chapter focuses on the main pricing models that can be used to quantify the
risk and other factors related to specific investments. These models are used to derive
expected returns on investments and hence to provide a decision framework for invest-
ment decisions.
117
CHAPTER4
THE DEVELOPMENT OF PRICING MODELS
4.2 The development of the Capital Asset Pricing Model (CAPM) 120
4.2.1 The nature of the CAPM 121

4.2.2 Assumptions of the CAPM 123
4.2.3 Risk and the CAPM 126
4.2.4 Parameters of the CAPM 127
4.2.5 The concept "capital market line" 130
4.2.6 The concept "security market line" 133
4.2.7 Tests and validity of the CAPM 137
4.2.8 Application of the CAPM in investment decision-making 146
4.2.9 Research related to the CAPM 149
4.2.10 Conclusion 183
4.3 The development of the Arbitrage Pricing Theory (APT) 184
4.3.1 The nature of the APT 184

4.3.2 Assumptions of the APT 187
4.3.3 Risk and the APT 188
4.3.4 Identification of the factors to be included in the APT 191
4.3.5 Tests and validity of the APT 192
4.3.6 The APT versus the CAPM 194
4.3.7 Application of the APT in investment decision-making 197
4.3.8 Research related to the APT 198
118
4.4 Options theory and the Black-Scholes (B-S) Option Pricing Model 214
4.4.1 The importance of options and the 8-S Option Pricing Model 214
4.4.2 The nature, types and value of options 217
4.4.3 The options market 226
4.4.4 The B-S Option Pricing Model 231
4.4.5 The application of the B-S Option Pricing Model in the pricing
of options on the options market 238
4.4.6 Tests on options theory, options markets and the B-S Option
Pricing Model 240
4.4.7 Research related to options markets and the B-S Option
Pricing Model 245
4.5 Summary 270
4.1 INTRODUCTION
Sharpe ( 1985:xix) notes that, although the field of investments continues to be revolu-
tionised, before the 1950s there was a lack of a theory of the formulation of prices and
capital markets.
Since then, development of Portfolio Theory and the Efficient Market Hypothesis, as
examined in Chapter 3, have contributed to our understanding of the risk/return relation-
ship, how capital markets behave and how prices on capital markets are established.
Three other important aspects which have contributed to our understanding of invest-
ment on capital markets and the formulation of prices will be examined in this chapter.
These are the Capital Asset Pricing Model (CAPM), the Arbitrage Pricing Theory (APT)
and options theory, with its related Black-Scholes (B-S) Option Pricing Model.
119
The first section of this chapter is concerned with the CAPM which, according to
Harrington (1993: 1), has made a fundamental contribution to financial theory and, toge-
ther with Portfolio Theory, changed the way investors, analysts, fund managers and
academics thought about investment management.
The CAPM attempts to explain the trade-off between risk and return in efficient markets
and provides a model of the equilibrium risk/return relationship (Sharpe 1985:xix;
Harrington 1993: 1). It uses beta as a measure of risk and as a basis for establishing
expected returns. The CAPM has been the subject of a tremendous amount of empirical
research and it was through empirical difficulties experienced and empirical anomalies
identified that an alternative theory of asset pricing was developed (Blume 1993:6,8).
The second section deals with this alternative asset pricing theory, the APT, which
suggests that returns are a function of various macro-economic risk factors and not only
one, beta, as suggested by the CAPM (Arnott 1993:16).
No clarity currently exists about which model is superior, the CAPM or the APT, but as
Sharpe (1985:xx) notes, it is not necessary to choose one above the other, they may
both hold at the same time. Both have strengths and weaknesses and only time and
empirical work will determine whether either, or both of them are validated or discarded.
Hodges (1988:4) identifies another aspect of the continuing revolution in the field of in-
vestments and that is the development of markets for options. The third section, there-
fore, concentrates on options markets and the option pricing model developed by Black
and Scholes. Their model, the B-S model, was the breakthrough that was required to
improve understanding of share-options as investment alternatives and as a hedge
against risk.
An important element of this chapter's examination and review will be the role of empi-
rical research in the development of these theories and models. Both Sharpe ( 1985:xx)
and Hodges (1988:4) note that research facilitated much of the recent developments
120
and that research is a key element of investment management.
4.2 THE DEVELOPMENT OF THE CAPITAL ASSET PRICING MODEL
The CAPM has been developed to measure the relevant risk of securities and to de-
scribe the relationship between the risk and expected return associated with invest-
ments. Although the CAPM has a number of simplifying assumptions, the importance
of the model does not lie in the realism of its assumptions, but rather in how well it pre-
dicts and describes reality. If empirical evidence confirm that the CAPM is successful
in explaining the returns on risky assets, then it can be concluded that the model is
useful and the realism of its assumptions is thus of less importance (Jones 1998:226).
Two important relationships are associated with the CAPM, namely the capital market
line (CML) and the security market line (SML). The CML depicts the relationship be-
tween total risk and expected return, while the SML depicts the relationship between
systematic risk and expected return. Both the CML and SML apply to individual shares
as well as portfolios. The SML is the graphical depiction of the parameters of the CAPM
and can be used to identify under- and overvalued securities (Jones 1998:230,247).
The CAPMs main conclusions are that there exists a positive relationship between risk
and return and that the relevant risk for a share can be measured by its effect on port-
folio risk. To assess the validity of these conclusions, and of the theory as a whole,
requires that empirical testing be performed and there exists an extensive literature as
regards to testing of the CAPM (Jones 1998:241 ).
In the following sections on the development of the CAPM all of these aspects are exa-
mined, that is, the nature and assumptions of the CAPM, how the CAPM assesses risk,
the parameters of the CAPM, the CML and SML, the tests and validity of the CAPM,
how the CAPM can be employed in investment decision-making, and the evidence from
empirical research done on the CAPM.
121
4.2.1 The nature of the CAPM
The development of the CAPM is generally ascribed to the work of Sharpe (1964),
although Treynor (1961 ), Lintner (1965) and Mossin (1966) made important, though
independent, contributions. Using the implications of Markowitz's (1952) work in the
development of a theory of portfolio selection, they established an equilibrium model for
the pricing of assets and this model, the CAPM, is mostly used to price the shares
traded on capital markets (Casabona 1979:7; Bradfield et al. 1988: 11 ).
The CAPM provides a simple, but powerful, description of the relationship between risk
and return in efficient markets. It is based on several simplifying assumptions and with
its focus being on the relationship between risk and return, it assumes firstly that all
investors view the risk and expected return characteristics of individual shares in the
same manner. Other assumptions include that transaction costs can be ignored; in-
vestors have homogeneous beliefs; investors have access to the same information and
analyse it in the same way; shares are divisible and liquid; investors can borrow and
lend at the risk-free rate of interest; taxes have no real effect on investment in shares;
and investors are only concerned about two aspects of investments, namely risk and
return (Sharpe 1985: 148; Blume 1993:6).
According to the CAPM the best portfolio is one that is fully diversified and in which
each share is held in proportion to its value in the market. The expected risk premium
(the expected return on the market minus the risk-free interest rate) on each share is
proportional to the expected risk premium on the market and the constant of propor-
tionality (beta) is related to the covariance of the returns of the individual shares with
the returns on the market portfolio (Blume 1993:6).
If the CAPM is an accurate description of capital markets it is easy to establish the

relationship between risk and return for efficient investment strategies. This relationship
can be demonstrated graphically and is known as the capital market line (CML). The
CML provides a relationship between expected return and risk, as measured by the
122
standard deviation of returns, for efficient portfolios and the slope of the CML can be
seen as representing the reward for every unit of risk borne (Sharpe 1985: 152, 153).
Since the CAPM considers beta as the accurate measure of risk, every share and port-
folio are expected to plot on an upward-sloping straight line and this relationship
between risk and return is known as the security market line (SML). The SML is a key
implication of the CAPM and represents an equilibrium theory of the expected returns
on shares (Sharpe 1985:160-162).
Since the CAPM is an expectational theory (it makes assertions about the expected
returns on shares and the future value of risk (beta)), it is difficult if not impossible to
test whether it describes reality. Historical data are usually used as surrogates for future
expectations in testing the CAPM. Only if the expected returns and the value of beta
remain constant over a long period of time, and if actual returns conform to these pre-
dictions, can the historical data be used with success in establishing the validity of an
equilibrium theory such as the CAPM (Sharpe 1985:393).
Such tests are joint tests of the CAPM and the assumptions about the stability of the
predictions, together with the conformance of the actual results with the predictions.
Sharpe (1985:394) concludes that:
If such a test fails, it may indicate that the equilibrium theory is in error.
Alternatively, one or more of the other assumptions may be inappro-
priate. As a practical matter, such procedures cannot reject an equi-
librium theory. They can, however, provide suggestive evidence.
The original CAPM makes strong assumptions and has significant implications for
investment decision-making. Since its development, various extended CAPMs have
been proposed and these extensions posit that expected returns are not only related
to beta, but also to additional risk factors (Sharpe 1985: 173, 199).
123
A theoretical examination of these extended models, and the research done thereon,
largely falls outside the scope of this study. This is not a study of the CAPM itself, but
rather of the various capital market theories and pricing models and how research
thereon has contributed to the development of accounting theory.
Broadbent (1992:10) concludes that empirical research suggests that the CAPM is a
useful tool for investment decision-making, as long as investors hold well-diversified
portfolios. However, before the research on the CAPM can be reviewed and conclusions
drawn from it, it is necessary to examine all aspects of the CAPM and their implications.
The discussion commences with a review of the assumptions underlying the model.
4.2.2 Assumptions of the CAPM
The CAPM relies on a number of simplifying assumptions, some of which are unique to
the CAPM, while others also underlie the EMH and Portfolio Theory. In order to derive
the CAPM, the following assumptions are required:
• All investors are rational, risk-averse and have the objective of maximising the
expected utility of their wealth at the end of their investment period. This
assumption is required in order to describe investor behaviour. Investors' utility
of wealth depends on their specific risk and return preferences (Harrington
1987:26-28; Seneque 1987:29).
• Investors base their investment decisions purely on the criteria of expected re-
turn and risk. This assumption is necessary in order to describe how investors
select investments so as to maximise their utility of wealth. Thus, investors are
assumed to make their portfolio choices on the basis of the expected values of
returns and standard deviations (or beta) of the returns (Harrington 1987:26,28;
Elton & Gruber 1995:295).
124
• All investors have identical expectations with respect to risk and expected return.
Without the assumption that investors agree on the market price of risk, a condi-
tion of capital market equilibrium, whereby all investments are efficiently priced
with regard to the level of risk involved, will not exist (Harrington 1987:26,31;
Elton & Gruber 1995:295; Jones 1998:227).
• Investors have a common single-period time horizon for decisions about invest-
ments and are assumed to have the exact same definition of this time-period.
This facilitates comparisons over a common interval, since a single-period model
like the CAPM requires that investors construct their investment portfolios at the
same point in time and sell them at the same unidentified future point in time
(Anderson 1978:69; Harrington 1987:26,33; Elton & Gruber 1995:295).
• Information relevant to the investment decision is costless and simultaneously

available to all investors. Without the assumption that capital markets are effi-
cient and that most investors are in agreement about the future prospects of
shares, the CAPM cannot be applied effectively (Harrington 1987:26,36; Laing
1988:8; Viljoen 1989:23).
• A risk-free asset exists and investors can borrow or lend unlimited amounts at
the risk-free rate of interest. This is a fundamental, and perhaps the most crucial,
assumption underlying to the CAPM. This implies that investors are not con-
cerned about the risk characteristics of individual shares, but rather how the risk
of the portfolio is affected when a portion of the risk-free asset is added to the
portfolio, or funds to invest in the portfolio are borrowed at the risk-free rate
(Harrington 1987:26,36; Keogh 1994: 17).
• There are no market imperfections, that is, no transaction costs, no rules placing
restrictions on short sales and dividend income and capital gains are not subject
to different rates of taxation (Harrington 1987:26,42; Seneque 1987:29). Accor-
ding to Keogh (1994:17) this allows investors to be treated equally as no-one
125
would be able to take advantage of these imperfections. Elton and Gruber

(1995:295) state that transaction costs are probably not important and will only
make the model more complex, and that the major results of the model will hold
if taxes on income and capital gains are not the same.
• There are a fixed amount of shares, meaning that new issues of shares can
probably be ignored (Harrington 1987:25,45). Secondly, shares are infinitely
divisible and thus investors can construct any portfolio, regardless of their wealth
(Elton & Gruber 1995:295). Thirdly, the shares are marketable, that is, they are
liquid and can be bought and sold at the ruling price (Oosthuizen 1992: 11 ).
• There are many investors and no single investor is able to affect prices through
individual buying or selling decisions. Also, since investors are price-takers, they
act as if prices are not affected by their own buying or selling decisions (Elton &
Gruber 1995:295; Jones 1998:227).
Both Elton and Gruber (1995:295) and Pike and Neale (1996:89) state that several, if
not most, of these assumptions are unrealistic and do not hold in the real world. It
should, however, be noted that the CAPM is an expectational model and should not be
judged on the realism of its assumptions, but rather on how well it explains reality and
predicts expected behaviour (Harrington 1987:27; Seneque 1987:29). Pike and Neale
(1996:279) add, further, that the CAPM has stood up well to the relaxation of many of
these assumptions and that the apparently unrealistic assumptions have no significant
negative effect on its implications. Elton and Gruber (1995:295) conclude that it is not
how realistic the assumptions of the CAPM are, but rather how well it describes capital
markets and the performance thereof.
Before any significant conclusions regarding the CAPM can be reached, it is first
necessary to examine its parameters and how it measures risk. These will be examined
in the following two sections of this chapter, whereafter its implications, namely the
capital market line and the security market line, will be discussed. This leads to the final
126
and most important sections of this examination of the CAPM, namely, its implications,
tests of the model, its validity and the research related to the model.
4.2.3 Risk and the CAPM
The CAPM defines risk as the extent to which the share's return covary with the return
on the market, that is, beta (as the designated risk measure) measures the volatility of
the share's return relative to the return on the market portfolio (O'Brien & Srivastava
1995:15).
From Portfolio Theory and the effect of diversification, it is known that the systematic
risk is the only element of risk that is priced and rewarded, and beta is a measure of the
systematic risk that cannot be diversified away. Beta, therefore, is a relative measure
of risk - it is an estimate of the risk of a share relative to the risk of the market portfolio.
That is, beta is a measure of a share's volatility (fluctuations in price) and estimates how
the share or portfolio's expected return will move relative to the market portfolio's return
(O'Brien & Srivastava 1995:16; Jones 1998:234).
Portfolio beta, as with expected return, is simply the weighted average of the betas of
the individual shares contained in the portfolio. The market portfolio has a beta of 1,
while more risky investments have betas higher than 1 and less risky investments have
betas lower than 1. Thus, for shares or portfolios with more or less risk than the market
portfolio, the returns would be proportionally higher or lower than the market return
(Harrington 1987:17; Jones 1998:235).
It can be concluded from Jones (1998:236) that the CAPM provides an elegant, yet
simple, risk/return relationship. This relationship between beta and return states that
investors expect the return on shares to be equal to the risk-free return plus a risk
premium, and the greater the risk the greater will be the risk premium. The next section
examines these individual components (parameters) of the CAPM.
127
4.2.4 Parameters of the CAPM
Implementation of the CAPM requires estimates of the following variables:

• The risk-free rate of return.
• The expected return on the market.
• The beta coefficients for individual shares.
4.2.4.1 The risk-free rate of return

The risk-free rate of return should have no variance and no covariance with the return
on the market, that is, it should be the return on an asset with no risk. Such assets are
difficult to find and there is some doubt that such assets actually exist, meaning that
various proxies are used. These proxies, for example, government bonds and treasury
bills, have been found to be empirically inadequate and theoretically suspect since they
are subject to uncertainty and inflation, and as such do have a degree of variance and
covariance with the market. These proxies have a further problem, namely, the historical
period to use as the period that would be most representative of the future (Firer
1993:27, Rees 1995:170).
4.2.4.2 Return on the market

The market return should be the return on a market portfolio that includes all risky as-
sets and, again, there is doubt that such a market portfolio is ever likely to exist. Usually
a market index is used as a proxy for the market portfolio, but the problem still remains -
which index and historical period to use as proxy for the future (Firer 1993:31 ).
Rees (1995: 170, 171) identifies two approaches that are used to estimate market re-
turns. The first, and most widely used, approach is to select some market index as a
proxy for the market portfolio and then to use long periods of historical data to eliminate
any short-term movements. This approach provides estimates of the market proxy's
return that are generally stable and reliable. The second approach is more expecta-
tional and uses forecasts from market analysts of individual share returns to estimate
market returns. This approach requires a considerable investment in data collection as
128
input for a model to forecast the market return, and is for this reason rarely used.
4.2.4.3 Beta
The third parameter, beta, is crucial to the CAPM and it brings together investors' ex-
pectations of returns and those of the market (Jones 1998:238). There are a number of
factors to take into account when estimating the betas for individual shares.
• A stock market is usually used as a proxy for the market as a whole when esti-
mating beta and this may not fully reflect the market portfolio as required by the
CAPM (Harrington 1987: 103; Jones 1998:240).
• The use of a stock market index as a proxy for the whole market, which in turn
is a proxy for the whole market portfolio of all risky assets, is a major compromise
of the specifications of the CAPM (Harrington 1987: 103; Jones 1998:240).
• A third aspect relates to the stability of beta over time. Historical betas are used
to make estimates of future betas, and whether historical betas are good es-
timates to use in an expectational model, remains a mute point. Empirical studies
have generally shown that portfolio betas derived from historical data are useful
estimates of future betas (Harrington 1987:99; Firer 1993:25; Rees 1995: 169).
According to Sharpe (1985:170) using a share's historical beta without change,

as an estimation of its future beta, is an extreme assumption. Historical betas are
therefore, in practice, adjusted or used in conjunction with other information
when estimates of future betas are made.
• Estimates of beta using historical data vary substantially, depending on the num-
ber of observations used in the calculation, the length of the time periods used
and the presence of outliers. There is no correct time period or number of ob-
servations and estimates can vary from weekly to monthly rates of return over 30
years to periods that are five years and even less. Using a large number of ob-
129
servations result in less variation in betas, but there is also the likelihood that
betas will not remain constant over long periods of time (Rees 1995: 167; Jones
1998:240).
• The estimates of future betas are only estimates and may not be equal or close
to the actual true betas (Jones 1998: 240).
• Beta will not remain perfectly stable over time. Beta will change as the funda-
mental variables such as cash flows and earnings of companies change, and
thus beta will always have a degree of instability (Jones 1998:240).
• The procedure in obtaining reliable estimates of beta is statistically taxing and

investors are well advised to make use of professional beta services. Such ser-
vices are available both in South Africa and internationally and are preferable to
ad hoc estimates of beta (Firer 1993:25).
Firer (1993:25) notes that, due to the estimation problems, it is not possible to apply the
CAPM in its pure form and provides the following summary of the estimation problems.
In order to obtain an estimate for beta, the stock's returns should theo-
retically be regressed against an index representing all risky assets in
the world. No such index exists and so a "true" beta cannot feasiblely
be obtained. Even the existence of a truly riskless asset is question-
able. Thus surrogates must be sought for the CAPM parameters - a
segment of the "world index" for which an index can be found, and a
financial instrument closest to the risk-free ideal.
Before the tests of, and research on the parameters of the CAPM can be reviewed, it
is necessary to consider how the CAPM defines the risk/return relationship. In the next
two sections the concepts "capital market line" and "security market line", which graphi-
cally describe this relationship, will be examined.
130
4.2.5 The concept "capital market line"
When the assumption regarding the existence of a riskless asset is added to the effi-
cient frontier, as defined by Portfolio Theory, the number of investment opportunities
available to investors are increased and a new and more efficient frontier is created. By
being able to borrow or lend at the riskless rate, and combining these with investment
in the market portfolio, investors are able to implement investment strategies that
provide more return for the same level of risk, or less risk for the same level of return,
than with the Portfolio Theory efficient frontier (Firth 1977:89; Harrington 1987: 14).
Firth (1977:89,90) notes that the capital market line (CML) provides investors with the
following investment strategies:
• The investor can lend (invest) all his funds available for investment at the risk-
less rate.
• The investor can lend (invest) a part of the funds in the riskless asset and invest
the balance in the market portfolio.
• The investor can invest all his funds in the market portfolio.
• The investor can borrow funds at the riskless rate and use these funds, in con-
junction with his own funds, to invest in the market portfolio.
The capital market line (CML) and its resultant investment strategies are illustrated in
figure 4.1.
131
Figure 4.1 The capital market line
Expected
return
where:
A Capital market line (CML)

B = Efficient frontier
M = Market portfolio
R1 = Risk-free rate
Source: Viscione and Roberts (1987: 195) and Ross et al. (1990:276).
The CML shows the risk/return relationship from combining lending or borrowing at the
risk-free rate with investing in the market portfolio. The returns that can be achieved by
combining investing in the market portfolio and investing in the risk-free asset or bor-
rowing at the risk-free rate, represent the most efficient portfolios and describes the risk/
return relationship for all shares traded in the market (Drury 1992:400).
The investment strategy an investor will select depends on his risk/return preferences.
An investor who is risk averse will combine investment in the risk-free asset with invest-
ment in the market portfolio - as indicated by point 1 on the CML. Against that, an
132
investor who is prepared to accept more risk will combine investing in the market port-
folio with borrowing at the risk-free rate - as indicated by point 2 on the CML. Point 2
renders a higher expected return, but is subject to a higher level of risk (Firth 1977:90).
Jones (1998:230-232) gives an extensive analysis of the CML and the main aspects to
note are the following:
• The slope of the CML indicates the additional amount of return the market de-
mands for each percentage increase in the risk of a portfolio and as such repre-
sents the equilibrium price of risk in the market.
• Only efficient portfolios that consist of both the risk-free asset and the market
portfolio lie on the CML, that is, all the combinations of the risk-free asset and
the market portfolio that lie on the CML are efficient portfolios.
• Since the CML describes conditions under equilibrium, it must always be upward
sloping, because the greater the risk the greater the return, that is, the market
price of risk is always positive.
• However, on an ex post (after the fact) basis the CML can for a period be down-
ward sloping where the return on the risk-free asset is greater than the return on
the market portfolio. This does not make the CML invalid, but merely shows that
actual returns differ from those expected. Thus, although the ex ante (before the
fact) CML must always be upward-sloping, expectations are not always realized
and it can sometimes be downward sloping on an ex post basis.
• The CML shows the expected return for each level of risk and can as such be
used to measure the optimal expected return for various portfolio risk levels.
According to Firth (1977:91,92) the CML holds for efficient portfolios, but does not
describe the relationship between expected returns on individual shares or inefficient
133
portfolios and their risk, as measured by beta. The relationship between beta and
expected return is called the security market line (SML) and will be considered in
section 4.2.6.
4.2.6 The concept "security market line"
The security market line (SML) differs from the CML in the following ways:
• The CML depicts the risk/return trade-off for efficient portfolios only.
• The SML depicts the risk/return trade-off for all assets, including individual
shares, efficient portfolios and inefficient portfolios.
• Standard deviation appears on the horizontal axis of the CML, while it is beta
that appears on the horizontal axis of the SML.
• The SML provides an unique relationship between systematic risk (as measured
by beta) and expected return on investments (Ross et al. 1990:287; Weston &
Copeland 1992:403; Jones 1998:232,233).
According to Pike and Neale (1996:274) the CAPM postulates that all shares are
correctly priced and that the SML gives the relationship between risk and return under
these conditions of market equilibrium. This relationship between beta and expected
return is depicted in figure 4.2.
134
Figure 4.2 The security market line
Expected
return
Beta
where:
A = Shares/portfolios that are less risky than the market portfolio (beta< 1,0)
B = Shares/portfolios that are more risky than the market portfolio (beta> 1,0)
M = Market portfolio
R1 = Risk-free rate
Source: Sharpe (1985: 161) and Jones (1998:234).
4.2.6.1 Characteristics of the SML

Ross et al. ( 1990:285,286) and Keogh ( 1994:27 ,28) identify the following characteris-
tics that can be associated with the SML:
• The expected return on a share with a beta of 0 is equal to the risk-free rate.
• The expected return on a share with a beta of 1 is equal to the return on the
market portfolio. The beta of the market portfolio is 1 since the market portfolio
represents the average beta of all shares, weighted according to the proportion
of each share's market value to the value of the total market portfolio.
135
• The SML is an upward sloping straight line depicting a linear relationship between
beta and expected return. Beta is the appropriate measure of risk and high-beta
shares will have a greater expected return than low-beta shares.
• Shares not lying on the SML are mispriced. Shares above the line are underpriced
and those under the line are overpriced, and under conditions of equilibrium it is
expected that these prices will adjust so that the SML becomes straight, that is,
depicting a linear relationship between risk and return.
• The SML is a graphical depiction of the CAPM. The equation for the CAPM is
represented by formula 4.1.
• The SML, and thus the Capital Asset Pricing Model, holds for both portfolios and
individual shares.
The CAPM is a model of equilibrium prices for shares and portfolios in efficient
markets, and the expected return on these shares and portfolios is equal to the return
on the riskless asset, plus the beta of the security or portfolio times the market risk
premium (Hendriksen & Van Breda 1992: 183).
Formula 4.1 The capital asset pricing model
where:
R = Expected return on the share/portfolio
Rf = Risk-free rate
~ = Beta (volatility of the share/portfolio relative to the market portfolio)
Rm = Expected return on the market portfolio
Rm - Rf = Market risk premium
Source: Harrington (1987:17), Ross et al. (1990:286) and Jones (1998:235,236).

136
Gay (1982:14) summarises the major implications of the CAPM as being:

• The risk/return relationship is positive and linear.
• That systematic risk, as measured by beta, is the only relevant risk that affects a
share or portfolio's return.
4.2.6.2 Alpha (under- and overvalued shares)

In equilibrium all shares are correctly priced and under conditions of equilibrium prices
of shares or portfolios should lie on the SML. If a share or portfolio does not lie on the
SML, a situation of disequilibrium exists, it means that the share or portfolio is mispriced
and the extent to which the share or portfolio is mispriced, is measured by its alpha
value (Sharpe 1985: 163).
A share or portfolio's alpha value represents the difference between its expected return
and the appropriate (equilibrium) expected return. A positive alpha value indicates that
the share or portfolio is underpriced, that is, its expected return is greater than the ap-
propriate expected return for investments with similar attributes. Against that, a negative
alpha value indicates that the share or portfolio is overpriced, while an alpha value of
zero indicates that the share or portfolio is correctly priced (Sharpe 1985: 163, 164 ).
The equation for the calculation of alpha is shown by formula 4.2.
Formula4.2 Alpha
where:
a = Alpha value
E = Expected return on the share/portfolio
Ee = Appropriate (equilibrium) expected return on the share/portfolio
Source: Sharpe (1985: 164).

137
Alpha implies that investors and investment managers who try to outperform the market
hope to construct portfolios with positive alpha values, that is, construct portfolios that
are underpriced. Those who do not attempt to outperform the market will try to construct
portfolfos with alpha values of zero, while those who make incorrect judgements
construct portfolios with negative alphas, that is, construct portfolios that are overpriced
(Sharpe 1985: 165).
The previous sections of this chapter examined the theory behind the CAPM. In the
next section its validity will be investigated, including tests and areas to be tested to
establish its validity, that is, whether it is realistic and whether it provides a good de-
scription of the behaviour of capital markets.
4.2.7 Tests and validity of the CAPM
Several criticisms have been directed at the CAPM. One area of criticism is noted by
Keogh (1994:36), namely, that the assumptions underlying the CAPM are unrealistic
and, hence, that it cannot predict capital markets in a practical (real world) manner.
Harrington (1987:52) and Rees (1995:173) identify two other areas of criticism. Firstly,
the CAPM may be misspecified, that is, the model is wrong and the actual risk/return
relationship is not linear. Secondly, the CAPM is inadequate, that is, the model does
not include all the factors relevant to the pricing of assets.
4.2.7.1 Validity of the assumptions of the CAPM

According to Harrington (1987:35,47) the assumptions underlying the CAPM are clearly
unrealistic and most, if not all, are violated in the real world. Without examining each
assumption individually it can be stated, according to Anderson (1978:80), that if the
assumptions were to be relaxed, the theoretical market equilibrium conditions cannot
be reached and as a consequence there will not be a true CML and efficient frontier for
all investors.
138
Asset prices will only adjust to a level where proportional returns approximate the total
risk associated with the asset, resulting in investors having many alternative combi-
nations of efficient portfolios. The result of this is that there will be a number of tangen-
cies between the CML and the efficient opportunity set (Anderson 1978:80).
Milton Friedman (1953), however, warned against judging a theory purely on the basis
of the realism of its assumptions, unless the theory aims to provide an accurate de-
scription and explanation of behaviour. But, if the theory aims to provide predictions
which can be tested against reality, the realism of the underlying assumptions is less
important, especially in circumstances where the predictions can be shown to accord
reasonably closely with reality (Pike & Neale 1996:285, 286).
It is therefore inappropriate to judge the CAPM on the basis of the realism of its
assumptions. A more appropriate approach is to conduct empirical studies on the cri-
teria of a good model, that is, does it have explanatory power and/or predictive ability?
If it has either or both, then the model is useful and can be applied to improve decision-
making (Firth 1977:96; Harrington 1987:35).
Before examining the tests of the CAPM, it is necessary to first look at difficulties in
testing the model in order to establish the limitations of the tests and, hence, the
limitations of the conclusions that can be drawn from the tests.
4.2. 7.2 Problems with testing the CAPM

Three inherent problems with the tests, apart from the difficulties experienced with
faulty or insensitive methodology, are generally emphasised:
• The CAPM is an expectational model, but most tests use realized (ex post)
returns, since expected (ex ante) returns are not readily observable. This creates
difficulties in that the model may actually hold for expected returns, although
actual returns may cast some doubt on it (Viscione & Roberts 1987:204). Rees
(1995: 173) deems attempts to model investor's expectations using historical
139
information (mostly as regards to estimating beta), or testing expectations against

realizations, not to be direct tests of the CAPM.
• The CAPM relies on the existence of a risk-free asset and there is doubt that such
an asset actually exists. The tests, therefore, depend on the specification of some
measure to be used as a proxy for the risk-free rate of return (Viscione & Roberts
1987:204; Pike & Neale 1996:287).
• The CAPM requires that share returns be analysed against the market portfolio.
Because such a portfolio does not exist, a proxy must be used and this could
result in different SMLs, depending on the proxy used. Further, the index used as
a proxy may be inefficient and thus distort the results of the tests. Hence, any ex-
post test of the CAPM is as much a test of the validity of the proxy as it is a test
of the model (Viscione & Roberts 1987:204; Rees 1995: 173; Pike & Neale
1996:287).
The role of each of these aspects in testing the CAPM is now examined in more detail.
a. Estimating beta
The stability of beta over time has become an important issue in the tests of the CAPM.
Historical betas are used as predictors of future betas and if beta remains relatively
stable (unchanged) over time, these historical betas will be useful in estimating future
betas. If, however, the historical betas are unstable and change substantially over time,
then they will have little predictive value. Instability of beta may have a further effect,
in that it can cause difficulty in interpreting the results of the tests and may even cause
the results to be invalid (Firth 1977:98, 100; Harrington 1987: 118).
The problem of stability of beta can be divided into three elements, namely the choice
of historical measurement period, the choice of length of interval within the chosen
measurement period and the choice of market proxy to be used as the benchmark
(Harrington 1987:109-115).
140
• The measurement period. The measurement period must be long enough to

achieve a statistically significant result, but must not be too long so as to capture
data that will not be relevant to forecasting the future. Unfortunately the CAPM
provides no guidelines for choosing the measurement period, but studies.have
shown that the length of the period is important and that, in general, the longer the
period the better the estimation of beta (Harrington 1987: 109, 110).
• The choice of interval. The length of the intervals within the chosen measure-
ment period can also affect the estimates of beta. Again, the CAPM provides no
guidelines and researchers have a choice between daily, weekly, monthly and
quarterly intervals, among others. Studies have shown the interval period is
important in order to obtain an adequate set of data, without including irrelevant
data which may bias the beta estimates (Harrington 1987:110, 114).
• The choice of market proxy. If the market index used as a proxy is not fully
diversified and, hence, not a good reflection of the market portfolio, an incomplete
distinction will be made between systematic and unsystematic risk. Due to the
misspecification of systematic risk, this can result in incorrect and even useless
estimations of beta values (Harrington 1987:114, 115).
b. Estimating the risk-free rate

The risk-free rate is of equal importance to the other two parameters of the CAPM and
is used twice in the model. Firstly, as the minimum expected return and, secondly, as
one of the two components of the market risk premium. The choice of proxy should be
carefully considered, since error in the estimation of the risk-free rate can cause a sig-
nificant error in the estimation of the expected return on a share or portfolio (Harrington
1987:149).
c. Estimating the return on the market

Harrington ( 1987: 167) identifies four problems with estimating the return on the market
proxy, namely which index to use as the proxy; how the return should be calculated;
141
whether the index should be value or equally weighted; and the period over which the
return should be calculated.
• The choice of index as market proxy. All share indexes are only fragments of
the total market and usually consist of different kinds of shares. Hence, it is diffi-
cult, and probably impossible, to establish whether a particular index is an ade-
quate proxy for the market portfolio (Harrington 1987: 174,175).
• Calculating the return. Two methods are generally used to calculate returns,
namely, the simple (arithmetic) average method and the compound (geometric)
average method. Again, it is not known which method is best to use when looking
at past performance to forecast a future return (Harrington 1987: 167).
• Value or equal weighting of the index. There exists a choice between weighing
each return in the index according to the market value of the share, or by simply
averaging the returns in the index in order to calculate the return on the market.
Again, there is some disagreement and it is not known which method is the more
correct (Harrington 1987: 168, 170).
• The time period. Another dilemma is the choice of time period to use as a proxy
for the future return on the market. Returns vary substantially between periods of
bull markets and periods of bear markets. This also creates the problem of
choosing the beginning and ending periods of the estimation period. Results can
vary substantially when the period starts with a bear market low and ends with a
bull market high, in comparison with starting high and ending low. It is therefore
important to select a time period that, according to the investor or researcher's
judgement, will best resemble the period expected to realize over the investment
horizon. The time period chosen should, in general, exclude periods of high vo-
latility and should be long enough to construct meaningful frequency distributions
(Harrington 1987:170-173).
142
Despite these difficulties the CAPM has been the subject of a considerable amount of
empirical testing (Rees 1995: 173), and the following aspects have been the focus of
the testing in order to establish whether the model has possible problems with mis-
specification or inadequacy.
4.2.7.3 Tests of possible misspecification

Harrington ( 1987: 56-77), Seneque (1987:30) and Linley ( 1992: 15) describe five condi-
tions which indicate whether the CAPM is valid, that is, not misspecified. These condi-
tions are reviewed below.
There are two basic tests which can be performed to determine whether the CAPM is
reliable:
• Ex post data is examined to establish whether the relationships are the same as
those predicted by the model.
• Historical data is used to forecast the values of beta, the risk-free rate and the
market return and these forecasts are then tested against more recent history to
establish how well the CAPM predicts future behaviour (Harrington 1987:55;
Linley 1992: 15).
The results, constrained by the difficulties of testing the model, of the numerous tests
conducted into the five conditions are summarised below.
a. Condition 1 - Does the CAPM describe reality?

Many studies have examined the stability of beta and compared historical data with
data generated from simulated portfolios. Although the results from the tests on the
stability of beta are not conclusive, it has been found that portfolios have more stable
betas than individual shares. Tests using ex post data have also indicated that the
risk/return relationship is not the same as that predicted by the CAPM (Harrington
1987:56,62; Seneque 1987:30,31; Linley 1992:16).
143
b. Condition 2 - Is the relationship between risk and return positive and linear?
Although the results of research have shown some nonlinearity, in general most of the
evidence support a positive linear relationship between beta and returns for portfolios
of shares. Where the tests of portfolios cover a long period of time (in excess of 10
years) the evidence tends to support the requirement of the CAPM that the beta/return
relationship be positive and linear (Harrington 1987:62: Seneque 1987:31; Linley
1992:17).
c. Condition 3 - Does beta measure risk?

The beta coefficient should be equal to the excess return on the market and if the re-
sults of the tests do not accord with this requirement of the CAPM, beta may be an
incorrect or insufficient measure of risk. Since it had been found that excess portfolio
returns are less than predicted, the results of these studies generally cast doubt on this
requirement (Seneque 1987:31; Linley 1992:17). According to Harrington (1987:64,65),
if investors act as if though diversification removes all unsystematic risk and beta
measures systematic risk, the CAPM may be ratified.
d. Condition 4 - Is alpha zero or at least close to zero?

The results of studies on this condition cast some doubt on the validity of the CAPM.
Low-beta shares have been found to earn significantly higher returns than predicted,
while high-beta shares earn significantly less than predicted (Seneque 1987:31,32;
Linley 1992: 18).
e. Condition 5 - What does the choice of market proxy show?

The use of an index that is not a suitable proxy of the market can provide incorrect
information. If the market indexes do not correlate highly with the true market portfolio,
betas will be affected and the slope and position of the SML will be different than
specified. Although this means that all tests of the CAPM should be treated with
caution, it does also cast doubt on the usability of the model in evaluating portfolio
performance (Harrington 1987:77; Seneque 1987:32; Linley 1992: 18).
144
4.2.7.4 Tests of possible inadequacy

An aspect that has received considerable attention is whether the CAPM is inadequate
in its depiction of the pricing of assets, that is, whether there are other important factors
that the model omits (Harrington 1987:79).
The CAPM is tested for inadequacy by adding other factors deemed relevant to the
model in order to establish whether these additional factors provide better explanatory
power, more stable parameters, or better diversification with fewer assets (Harrington
1987:79).
Listed below are some of the additional factors that have been added to the CAPM as
a means of testing its adequacy.
a. The effect of dividends and taxes

The effect of dividends has always been considered as a factor that affect share prices,
due to possible tax differentials between dividends and capital gains. Although some
evidence of a dividend yield effect have been uncovered, it has been shown that this
effect cannot be attributed to taxes, but rather to some other factors such as firm size
(Harrington 1987:79; Bradfield eta/. 1988:12, 13). Hence, Harrington (1987:79,81) con-
cludes that the effect of both dividends and taxes are as yet not clear and require
further information.
b. The effect affirm size

The size effect is another factor that has received considerable attention and it has
been shown that there is a significant relationship between the market value of shares
and the returns on shares, that is, that small firms show long and persistent abnormal
returns. However, it is not yet clear whether firm size is responsible for the effect, or
whether it is only a proxy for some other factor affecting returns (Harrington 1987:83;
Bradfield et al. 1988: 13).
145
c. The effect of liquidity

The liquidity effect is related to the size effect, in that it can be expected that the shares
of smaller firms will be less liquid than those of large firms. Although little evidence re-
garding the liquidity effect exists, it can be assumed that investors should require a
higher expected return for less liquid shares as compensation for its limited marketabi-
lity and higher cost of trading (Bradfield et al. 1988: 13).
d. The effect of price/earnings (PIE) ratios

Studies on the effect of PIE ratios have shown that low PIE portfolios outperform and
earn higher returns than high P/E portfolios (Linley 1992:16).
4.2.7.5 Summary
The basic results of the tests performed on the CAPM and their implications concerning
the model's validity are summarised below.
• Beta appears to be related to past returns, has better explanatory power than the
standard deviation of returns and the risk/return relationship appears to be linear
and positive (Gay 1982:179; Linley 1992:19; Rees 1995:172).
• Low-beta shares tend to earn more than that predicted by the CAPM and high-
beta shares tend to earn less (Gay 1982: 178, 179).
• The risk-free rate of return appears to be higher than expected, although this
could be as a result of incorrect estimation rather than error with the CAPM, and
the slope of the risk/return relationship is insufficiently steep (Rees 1995: 172).
• Although systematic risk is found to be significant in explaining returns, other

factors have also been found to have explanatory power and their inclusion in the
model may improve its fit (Rees 1995: 172).
146
It can, in general, be concluded that the CAPM is not misspecified and that it is there-
fore valid from this point of view, but investors should be cautious when applying it to
evaluate investment performance. From a validity point of view, it appears as if the
CAPM may be inadequate and that other factors and/or pricing models may provide a
better description of the risk/return relationship (Seneque 1987:32; Linley 1992: 18, 19).
Following from this, the next section briefly examines how the CAPM can be applied
in investment decision-making.
4.2.8 Application of the CAPM in investment decision-making
As noted in section 4.2.7, serious criticisms can be levelled at the CAPM. According to
Harrington (1987:209) the following three criticisms severely restrict the practical appli-
cation of the model in investment analysis.
• The assumptions underlying the model are unrealistic and, thus, the basic CAPM
is flawed.
• Tests of the CAPM have shown that the model does not provide an accurate
description of reality, and is as such flawed.
• Forecasts of the risk-free rate, the market return and beta are hampered by
serious uncertainties and this may cause the CAPM to be of limited use in
practice.
Despite these criticisms and practical problems associated with the tests of the CAPM,
the model does have practical applications. Although application of the CAPM should
be performed with care, corporate management, regulation of public utilities, portfolio
management and share selection have been found to be areas where the CAPM can
be applied with real practical use (Harrington 1987:210).
147
4.2.8.1 Corporate management

Harrington (1987:210) identifies three interrelated fields where the CAPM has, and
continues to be, used by corporate management.
• Determining the hurdle rate for corporate investment.

• Estimating the required rates of return for divisions, business units and lines of
business.
• Evaluating the performance of divisions, business units and lines of business.
a. Hurdle rates - Cost of capital

Management often uses the weighted average cost of capital (WACC) as the yardstick
against which the expected returns of capital investment proposals are evaluated. The
cost of equity is an element of the WACC and the CAPM is often used to calculate the
firm's required return on equity (Harrington 1987:210).
b. Estimating the required rate of return

The CAPM can also play an important strategic planning role in assisting corporate
managers in the implementation of a consistent and systematic method of risk analysis.
The distinction between systematic and unsystematic risk, and the measurement of
systematic risk using beta, are quite useful to corporate managers when applying risk
analysis .in corporate strategic planning (Harrington 1987:211 ).
c. Performance evaluation
The CAPM is also used to evaluate past performance and to establish whether divi-
sions, business units and lines of business have created value and, at least, earned
the cost of equity (Harrington 1987:211 ).
4.2.8.2 Regulation of public utilities

The CAPM can also been used to establish the cost of equity of public utilities. Utility
rates are set so as to recover all costs, including the cost of equity, and thus the CAPM
is a useful tool for managers and regulators of public utilities (Harrington 1987:213).
148
4.2.8.3 Portfolio management and share selection

Investment managers have also been able to apply the CAPM with some success in the
fields of share selection, portfolio construction and portfolio performance evaluation.
a. Share selection
Many practitioners attempt to forecast returns and beta in order to identify undervalued
shares, that is, shares with positive alpha values. The CAPM can be used to forecast
share returns and practitioners have used this process to identify underpriced shares
in order to earn abnormal returns (Harrington 1987:214-216).
b. Portfolio construction
The CAPM can also be used to control the level of portfolio risk. Beta is used, as a risk-
level constraint in linear programming, to manage the level of portfolio risk (Harrington
1987:216-218).
c. Evaluating portfolio performance

Using historical returns and beta, the performance of portfolios can be evaluated
against the SML. In this manner, portfolios that have underperformed, that is, having
negative alpha values, and portfolios that have achieved abnormal risk-adjusted re-
turns, that is, having positive alpha values, can be identified (Harrington 1987:218).
As it is only through empirical testing that the validity, application and descriptive ability
of the CAPM can be established, it is now important to review the results of some of the
research conducted on the CAPM.
149
4.2.9 Research related to the CAPM
Although the CAPM suggests that shares with high forecasted beta values should
render high expected returns, it does not imply that these shares will necessarily render
high actual returns. If there is a substantial increase in the market (the market has gone
up), one would expect the actual returns of high-beta shares to be higher than that of
low-beta shares. If, however, the market shows a substantial decrease (the market has
gone down), one would expect the actual returns of high-beta shares to go down more
than those of low-beta shares. These are, however, only expectations and over time
actual returns may bear little resemblance to expected returns and, hence, to estimated
beta values (Sharpe 1985:395).
This clearly illustrates why many writers have observed that the CAPM is basically
untestable. It is based on investor's expectations about future returns and expectations
are difficult, if not impossible, to measure. Thus, researchers have generally had no
choice but to test the CAPM using past data as proxies for future expected returns
(Pike & Neale 1996:286).
Sharpe (1985:395) identifies a further problem, namely, the measurement of beta. The
CAPM requires that beta should measure a share's sensitivity to a market portfolio
consisting of all risky assets. However, such a market portfolio has as yet not been
identified and researchers are forced to use some stock market portfolio or index as a
proxy for the market portfolio. This causes the problem that the same share can have
different beta values, depending on whether the sensitivity of its returns have been
measured against a stock market index, a country's full stock market portfolio or even
an international stock market portfolio.
Despite these problems, the CAPM has been the subject of extensive research. These
studies have examined the CAPM in many ways, but the essence and basic procedure
of these studies are (levy & Sarnat 1994:336,338; Pike & Neale 1996:286):
150
• The application of time-series analysis to a large sample of historical (ex post)

data of shares in order to estimate beta and expected returns.
• Using regression analysis to apply the estimates of beta and expected return to
form a SML in order to establish whether the projected SML corresponds with the
slope of the SML suggested by the CAPM.
From this, the results of some of the international research (on the US market, unless
otherwise indicated) conducted until the mid-1980s are reviewed, according to the area
of investigation.
4.2.9.1 Research on beta as a measure of risk
Sharpe (1966) and Friend, Blume and Crocker (1970)

These studies represent evidence in support of the CAPM. The performance of mutual
funds was examined and it was found that higher returns are associated with increased
systematic risk (Firth 1977:98).
Arditti (1967,1971)
Arditti investigated the importance of a number of factors, including skewness, and
found that the return on shares is directly related to variance, but inversely related to
skewness. He showed that, although increases in return are correlated to variance, the
increases in return are less than proportional to the increases in risk and concluded
that skewness is an important factor in explaining returns (Harrington 1987:72).
Arditti also found evidence in support of investors' preference for positive skewness
(Elton & Guber 1995:247). He found that actual returns on the market and those for the
sample of mutual funds investigated, showed positive skewness. The mutual funds
exhibit a greater amount of skewness than the market and it was concluded that this
represents evidence that managers are willing to sacrifice some expected return, or
accept more variability in returns, for the chance of earning large returns (Firth
1977:96).
151
Jensen (1968)
Jensen also examined mutual fund performance and found evidence consistent with
the implications of the CAPM. He found a linear relationship between systematic risk
(beta) and returns and also found that portfolio betas are quite stable over time. This
implies that estimates of beta based on historical data are useful in making predictions
of future portfolio betas (Van Rhijn 1994: 103).
Douglas (1969)
The study of Douglas covered 616 shares for the years 1926 to 1960. The covariance
of each share's returns was calculated against an index consisting of the 616 shares
examined. Douglas found evidence in conflict with the CAPM, in that a much stronger
relationship was found between variance and returns than between beta and returns.
This implies that, for the vast majority of the years studied, total risk explained returns
better than systematic risk (beta) could (Firth 1977:96; Van Rhijn 1994: 104, 105; Reilly
& Brown 1997:312).
Douglas further found that the minimal rate of return (the intercept) is not equal to the
actual risk-free rate, that is, the intercept is higher (Harrington 1987:57; Reilly & Brown
1997:312).
Lintner (1969)
Lintner found results similar to those of Douglas(1969) (Firth 1977:96; Harrington
1987:57; Van Rhijn 1994:106). Harrington (1987:57) concludes that the results of the
Douglas ( 1969) and Lintner ( 1969) studies could either be because of faulty test proce-
dures or because the CAPM is wrong.
Francis and Archer (1971) and Jacob (1971)

Jacob used individual shares to study the CAPM, while Francis and Archer used port-
folios, but they found similar results. Again the evidence was similar to that of the 1969
studies of Douglas and Lintner, and in conflict with the CAPM (Anderson 1978:83-85).
152
Black, Jensen and Scholes (1972)

These researchers examined a sample of all the shares traded on the NYSE over the
period 1926 to 1966 (Van Rhijn 1994: 114). They used 10 different portfolios to investi-
gate the relationship between portfolio returns and betas over the 40 year period of
their study (O'Brien & Srivastava 1995: 117). The major conclusions drawn from their
study were that:
• Over time, there exists a close relationship between beta and return, and this
relationship is positive and linear (Casabona 1979: 12; Seneque 1987:31; Pike &
Neale 1996:287).
• The intercept is significantly greater than the risk-free rate, and that the slope of
the SML is flatter than predicted by the CAPM (Casabona 1979:12; Harrington
1987:59; Pike & Neale 1996:287).
• The value of alpha is not equal to zero, since high-beta shares earn returns that
are lower than expected and low-beta shares earn higher returns than expected.
This implies that there are additional factors which explain returns (Firth 1977:97;
Gay 1982:181; Senequa 1987:32).
• Returns are better explained by a two-factor model, which they developed and
called the zero-beta model. Risk-free borrowing is not used as an assumption in
this model (Firth 1977:97; Gay 1982:181; Harrington 1987:61).
Miller and Scholes (1972)

Miller and Scholes found results similar to those of Black et al. ( 1972) in their inves-
tigation of a sample of 631 shares for the period 1954 to 1963 (Firth 1977:97; Van Rhijn
1994: 107). They proposed that the combined effects of skewness of returns and beta
measurement errors could have caused the pre-1972 results, and concluded that the
results of these studies are inconclusive as they concentrated on variance/standard
deviation and did not properly divide total risk into its systematic and unsystematic
parts (Firth 1977:96; Anderson 1978:84,85; Van Rhijn 1994:108).
153
Sharpe and Cooper (1972)

Sharpe and Cooper examined the long-term risk/return relationship using monthly re-
turns and each share's beta relative to an index of the returns on all shares listed on
the NYSE for the period 1931 to1967. They found results for short-term (one year) and
medium-term (up to five years) periods that are consistent with a positive relationship
between beta values and returns. The results for the overall 37-year study period are
less clear, but it shows an overall result of an apparent positive relationship between
beta and long-term returns (Sharpe 1985:395-398).
Elton and Gruber (1995:344) conclude that the work of Sharpe and Cooper represents
strong evidence in favour of the existence of a positive linear relationship between beta
and return.
Friend and Blume (1970) and Blume and Friend (1973)

Friend and Blume investigated the risk/return relationship in order to establish whether
beta is the only and complete measure of risk. They tested 200 randomly selected
portfolios consisting of shares listed during the period January 1960 to June 1968 to
determine whether a linear relationship exists between risk-adjusted returns and two
measures of risk, namely beta and standard deviation. They found that neither beta nor
standard deviation are complete measures of risk and that high-beta portfolios earn
lower returns than do low-risk portfolios. These findings are contrary to the implications
of the CAPM and they concluded that a less rigid expression of the risk/return rela-
tionship would be preferable (Harrington 1987:65,66; Van Rhijn 1994: 108, 109).
In a follow-up study in 1973, Blume and Friend again examined the risk/return relation-
ship for shares listed in the period 1955 to1968. Although they found a linear relation-
ship between realized returns and systematic risk, they concluded that their tests failed
to find evidence in support of the CAPM as the estimated risk-free rate of return and
the estimated risk premium on the market portfolio are significantly different from the
average actual risk-free rate and the return on the market portfolio over the period of
the study (Van Rhijn 1994:109,110).
154
Fama and MacBeth (1973)

The Fama and MacBeth study found evidence in support of the CAP M's positive linear
relationship between portfolio beta and expected returns (Seneque 1987:31; Levy &
Sarnat 1994:338). Fama and MacBeth tested 20 portfolios of NYSE listed shares over
the period 1935 to 1968 and found overall evidence that is consistent with the basic
CAPM. The main conclusions from their study were:
• The relationship between expected returns and risk is linear in efficient portfolios.
• Beta provides a relatively complete measure of risk and no other measure of risk
provides a better explanation of expected returns of efficient portfolios.
• The risk/return relationship is positive since higher risk is associated with higher
return (Firth 1977:97,98; Casabona 1979:13,14; Van Rhijn 1994:114).
Drzycimski and Yudelson (1969), Joyce and Vogel (1970), Klemkosky and Petty
(1973) and Ben-Zion and Shalit (1975)
All these studies identified some problems with using historical beta to measure risk
and recommended that more than one measure of risk should be used (Anderson
1978:96).
Levy (1974)
Levy also concluded that the evidence shows that it would be possible to find a greater
correlation between historical risk and subsequent return if risk is measured by a
variable other than beta (Anderson 1978:96).
McDonald (1974)
This study of mutual fund performance, over a 10-year period, found that higher returns
are associated with increased systematic risk. Although evidence was found that high-
beta portfolios earn higher returns than expected, this was not found to be statistically
significant and the overall results of the study provide evidence in support of the CAPM
(Firth 1977:97,98).
155
Cooley, Roenfeldt and Modani (1977)

The study of Cooley et al. also examined the completeness of beta as a measure of risk
by comparing it with a number of the more traditional measures of risk. They showed
that beta provides additional information to the traditional measures, but there are other
measures with which it overlaps. As with the 1967 study of Arditti, it was found that
skewness provides non-beta related and useful information, and that share returns
which are not normally distributed, but skewed to the right, have less downside risk and
more upside potential, resulting in less risk for the investor (Harrington 1987:71 ).
Firth (1977)
Firth's study of the performance of unit trusts in the UK also found evidence in support
of the CAPM in that higher returns are associated with higher levels of systematic risk
(Firth 1977:98).
Friend, Westerfield and Granito (1978) and Friend and Westerfield (1981)
Friend, Westerfield and Granite added unsystematic risk to the CAPM and found that
this provides a better explanation of share returns. They concluded that systematic risk,
as measured by beta, is not the sole variable explaining returns and that residual stan-
dard deviation is at least as important in explaining returns (Gay 1982: 182; Van Rhijn
1994:131 ).
Gay (1982: 182) notes, however, that it is unclear whether these findings are a true re-
flection of an ex ante phenomenon or whether it is only a statistical artifact. Gay further
notes that the study is tainted due to a limited number of observations. Friend and
Westerfield extended the initial study over a longer time-period and found results which
confirm the initial findings.
Cheng and Grauer (1980)

This study examined the monthly share values for firms listed on the NYSE during the
period 1926 to 1977. In order to ensure that their test results are compatible with the
results of previous studies, they based their portfolio selection technique on that used
156
by Fama and MacBeth in their 1973 study. Cheng and Grauer formulated five testable
hypotheses according to which the CAPM could be judged and concluded from their
results that these hypotheses cannot be accepted as they do not hold jointly and, as
such, that their findings provide evidence against the CAPM (Van Rhijn 1994:131 ).
Lakonishok and Shapiro (1984)

Lakonishok and Shapiro examined portfolios of shares for a 22 year period and found
that risk is rewarded, but that it is variance that is rewarded and not beta. Similarly to
Black, Jensen and Scholes (1972), they found further evidence against the CAPM in
that high-beta shares perform better in up markets and worse in down markets than
low-beta shares, and in general, that high-beta shares earn less and low-beta shares
more than predicted by the model (Harrington 1987:66,68; Seneque 1987:32).
Other researchers, in addition to the studies of Arditti ( 1967), Black et al. ( 1972) and
Cooley et al. ( 1977) also investigated the relationship between beta and return, and the
effect of skewness on expected returns. They found that when returns are normally
distributed there exists a balance between positive and negative observations, that is,
the distribution is symmetrical. Skewness, however, is a measure of the asymmetry of
the distribution of returns and positive skewness refers to a situation where there is an
abnormal number of large positive price changes. Investors should prefer positive
skewness as there exists a strong probability of large payoffs and many researchers
have considered skewness as the reason for the CAPM phenomenon that low-beta
shares earn higher returns and high-beta shares lower returns than expected, that is,
low-beta shares are underpriced and high-beta shares are overpriced (Elton & Gruber
1995:247; Reilly & Brown 1997:314).
McEnally (1974)
The results of the McEnally study confirmed these perceptions about skewness, but
also found that high-beta shares have high positive skewness. A possible explanation
for this finding is that investors prefer shares with high risk and high positive skewness
as this provides the opportunity for very high returns (Reilly & Brown 1997:314).
157
Kraus and Litzenberger (1976)

Kraus and Litzenberger suggested that it is not skewness in itself, but rather the syste-
matic and unsystematic skewness that is important and criticized Arditti ( 1967) for not
differentiating between the diversifiable and non-diversifiable elements of skewness
(Harrington 1987:72; Elton & Gruber 1995:247).
They added a skewness term to the CAPM and tested this new model on 20 portfolios
over a 32-year period. They found that systematic skewness increases the explanatory
power of the CAPM in that it corrects for the apparent mispricing of high-beta and low-
beta shares. They concluded that expected return is a function of both systematic and
unsystematic skewness (Firth 1977:97; Reilly & Brown 1997:314).
Friend and Westerfield (1980)

In a subsequent test of the Kraus and Litzenberger (1976) model, Friend and Wester-
field found mixed results (Reilly & Brown 1997:314), and Gay (1982: 182) concludes
that while skewness was found to be a significant factor, it does not add much to the
explanatory power of the CAPM.
Simkowitz and Beedles (1978)

This study examined the effect of diversification on portfolio skewness and found that
skewness is reduced through diversification, since 92% of skewness is eliminated when
portfolios reach five shares. All positive skewness was found to have disappeared after
portfolios reach six shares and Simkowitz and Beedles concluded that investors who
prefer skewness would not diversify (Harrington 1987:72, 73).
Reinganum (1981)
In this examination of the risk/return relationship, Reinganum used at least 100 days
of daily data to estimate betas and found that, for the period 1926 to1979, the higher
the portfolio beta the higher the returns. These results are however not statistically
significant. He also showed that portfolios ranked according to beta have significantly
skewed results, even when the abnormal returns are excluded (Harrington 1987:62, 72).
158
It can be concluded from the above studies that when historical data is used, the
risk/return relationship is not quite as that predicted by the CAPM (Harrington 1987:62).
Although the evidence is not strong enough to reject the CAPM, it does indicate that
the basic CAPM is flawed. It has been shown that the CAPM provides incorred de-
scriptions of the past and that it does not describe expectations well. The intercept
appears to be higher than predicted and the slope of the SML seems to be less steep
than predicted (Harrington 1987:75, 79).
4.2.9.2 Research on the stability of beta
Blume (1971,1975)
The 1971 study of Blume examined the stability of beta for both portfolios of shares and
individual shares over consecutive seven-year periods (42 years in total). Two main
conclusions were drawn from this study:
• Betas for individual shares are unstable, but those of portfolios tend to be fairly
stable over time, and historical portfolio betas are fairly good predictors of future
betas.
• Betas of firms tend, over the 42 years of the study, to move towards the beta of
the market as a whole and when beta values depart from their average values,
their next value tends to be such that it is brought back towards the average value
(Firth 1977:99; Clark et al. 1979:183).
The evidence from the 1975 study confirmed the 1971 results, in that it was found that:
• Betas for individual shares are not good predictors of future betas, while betas of
portfolios are stable and contains much information about future betas.
• Betas tend to regress towards the mean over time and the shares with extreme
risk characteristics tend to be less extreme over time (Blume 1975:21,31; Jones
1998:240).
159
Fisher (1971) and Jacob (1971)

Fisher found only a slight difference between the beta values provided by various base
time periods, while the study of Jacob suggested that beta values are relatively un-
stable (Firth 1977:98,99).
Levy (1971,1974)
Levy examined, similarly to Blume (1971, 1975), the stability of beta over time and with
similar results and conclusions. The 1971 study, the results of which were confirmed
by the later study, involved an examination of over 500 shares listed on the NYSE over
a 10-year period (Anderson 1978:92; Reilly & Brown 1997:310). Clark et al. (1979:183)
provide the following summary of the conclusions that can be drawn from Levy's
studies:
• Betas for large portfolios are quite stable, more so than for small portfolios and
even more than those of individual shares.
• The longer the beta forecast period, the more reliable the forecast.
• Large portfolios with more than 25 shares and a forecast period of longer than 26
weeks provide historical betas that are good and stable predictors of future betas.
Levitz (1972)
The study of Levitz supports Levy's findings. Although he found poor correlation be-
tween individual share's historical and actual betas. he found significant correlation for
portfolios 'of only 10 shares. (Clark et al. 1979: 183).
Sharpe and Cooper (1972) and Sharpe and Sossin (1976)

Other than finding a positive relationship between risk and return, although not com-
pletely linear, both these studies also examined the stability of beta for individual
shares and found a high level of stability, even for individual shares (Firth 1977:99).
160
Cunningham (1973)
In a study on UK shares, Cunningham examined 950 shares for two different three-year
periods. He found beta to be relatively stable and historical betas to be relatively good
predictors of future betas (Firth 1977:99).
Meyers (1973)
Meyers examined the hypothesis that beta's instability is caused by the fact that the
amount of variance explained by the market varies from one period to the next. The
periods 1952 to 1960 and 1961 to 1967 were investigated and the portfolio beta of 94
shares were found to be reasonably stable, while the variance was virtually unchanged
during both periods. For individual shares, however, the degree of return behaviour
explained by beta varied significantly from period to period (Harrington 1987: 123).
Aber (1972), Brennan (1973) and LeRoy (1973)

Aber also found betas to be relatively unstable. Both Brennan and LeRoy developed
theoretical models which imply that beta can and will not be stable over time (Firth
1977:99).
Various studies (1970-1976)

A number of studies have also confirmed the findings that
• betas for individual shares are relatively unstable,
• betas for portfolios consisting of a large number of shares are quite stable and
have a high degree of predictive ability, and
• beta stability is directly related to the degree of diversification.
These studies include those of Beaver, Kettler and Scholes (1970), Rosenberg and
McKibben (1973), Schwartz and Altman (1973), Vasicek (1973), Schlarbaum and
Racette (1974) and Rosenberg and Guy (1976) (Anderson 1978:92,93).
The studies of Baesel (1974), Klemkosky and Martin (1975) and Porter and Ezzell
(1975) confirmed the earlier findings that the betas for portfolios are more stable than
161
those of individual shares, while Baesel (1974) and the French study of Altman,
Jacquillat and Levasseur (1974) found that the betas of individual shares become
more stable with increases in the estimation period (Harrington 1987: 125; Reilly &
Brown 1997:311 ).
Fielitz (1974) and Porter and Ezzell (1975)

Fielitz found that a portfolio consisting of only eight shares has a beta with a con-
siderable degree of stability, but to have a beta with a high degree of stability the size
of the portfolio needs to be increased substantially. The findings of Porter and Ezzell's
study contradict those of Fielitz, since they found no evidence that an increase in port-
folio size improves the stability of beta (Seneque 1987:30; Reilly & Brown 1997:310).
Elton, Gruber and Urich (1978)

The study of Elton et al. showed that two elements of beta estimates contribute to its
instability. These elements are the correlation between the share and the market and
the volatility of the returns of the share. It seems that, in general, it can be concluded
from this study that more accurate estimates of betas can be obtained if it is based on
the average correlation for all the shares in the sample, rather than estimates of beta
for specific shares (Rees 1995: 169).
Roenfeldt, Griepentrag and Pflamm (1978)

The study of Roenfeldt et al. examined the relationship between the length of time of
the base period and the test period in the estimation of beta. Betas derived from 48
months of data were compared with subsequent betas for 12, 24, 36 and 48 months.
The findings of their study showed that beta estimates using 48 months of data are not
good predictors for subsequent 12-month betas, but do predict 24-, 36- and 48-month
betas well (Reilly & Brown 1997:311 ).
162
Alexander and Chervany (1980)

This study confirmed Meyers' 1973 conclusion that betas of well-diversified portfolios
are relatively stable. They also found that the benefits of diversification on the stability
of beta, to a large degree, disappear after the addition of the tenth share to the portfolio
(Harrington 1987:124)
Carpenter and Upton (1981)

Carpenter and Upton examined the impact of trading volume on the stability of beta and
found an effect related to the small-firm effect. They showed that beta estimates can
be improved when volume-adjusted betas are used (Seneque 1987:30: Reilly & Brown
1997:311 ).
Theobald (1981)
The findings of Theobald explain the earlier findings of Baese I ( 197 4) and partially
explain those of Roenfeldt et al. (1978). Theobald showed that the stability of beta im-
proves with an increase in the calendar period examined and concluded that the op-
timal period could be over 10 years if beta had not shifted during that period (Seneque
1987:30; Reilly & Brown 1997:311 ).
Tole (1981)
Tole agreed with Fielitz's 1974 findings and contradicted those of Porter and Ezzell in
197 5. Tole concluded that 10 to 25 shares in the portfolio still improves the estimates
of beta and that these benefits can be attained when the portfolio grows beyond 100
shares (Seneque 1987:30; Reilly & Brown 1997:310).
Bey (1983)
Using a sophisticated technique, Bey examined the stability of beta for public and utility
shares for the period 1960 to 1979. He found betas of individual shares to change sub-
stantially from five-year period to five-year period and concluded that betas of indivi-
dual shares are not stable (Harrington 1987: 118).
163

In a study on the UK stock exchange, Dimson and Marsh confirmed the earlier findings
of Blume (1971, 1975) (Rees 1995:169).
Grossman and Sharpe ( 1984)

This study examined the beta values on the NYSE during the period 1928 to 1982 and
found that beta values are positive when excess returns exceed the riskless rate and
negative when the reverse occurs, that is, shares with historic high beta values are
more volatile than those with low beta values. They concluded that this is a relationship
that is statistically significant and that historic betas are useful indicators of future
betas, especially when combined with the size effect (Sharpe 1985:399,400).
Kryzanowski and To (1984)

Kryzanowski and To argued that much of what was cited as the cause of the instability
of beta do not in actual fact create real instability. Estimates of beta were generated
using time-series analysis of historic data on past returns, whereas betas are expec-
tational, that is, a function of expected returns. They concluded that
betas estimated using ex post return data can be expected to exhibit

non-stationarity, even when the underlying ex ante security returns are
serially independent and obey a stationary distribution over time
(Harrington 1987: 130).
From these studies it can generally be concluded that betas of portfolios are relatively
stable, while those of individual shares are much less stable (Firth 1977:99). Hence,
the CAPM is more useful in structuring investment portfolios than in estimating returns
on individual shares (Van Rhijn 1994:250). Two other conclusions that can be drawn
from the studies into the stability of beta is that at least 36 months of data should be
used for its estimation and that cognisance should be taken of the share's trading vo-
lume (Reilly & Brown 1997:311 ).
164
4.2.9.3 Research into the association between accounting measures of risk and
beta
Ball and Brown ( 1967)

Ball and Brown found an association between beta, systematic risk and the covariance
of individual firm's returns to an average level of corporate earnings (Firth 1977: 100).
Beaver, Kettler and Scholes (1970)

This twofold study examined the relationship between beta and various accounting
measures of risk, together with whether these accounting measures are good pre-
dictors of beta. The accounting measures investigated were dividend payout, growth,
financial leverage, liquidity, size, variability of earnings and a measure of cyclacility
(earnings beta). The investigation entailed the examination of the accounting data and
beta of 307 firms for the period 1947 to 1965, with the first 10-year period used as a
predictor of beta values for the remaining years. The results of the study showed that
leverage and earnings beta have strong correlations, earnings variability and dividend
payout have significant correlation, while the rest only have weak correlation (Beaver
et al. 1970:287-289; Myers 1973:51-53; Firth 1977: 101 ). The conclusions drawn from
the study were that:
• Accounting data is incorporated in market prices and accounting measures of risk

are good predictors of beta.
• Accounting data provides better predictions of beta than does naive estimates of
beta (Beaver et al. 1970:287-289; Firth 1977: 101 ).
Hamada (1972)
In this study, Hamada examined beta levels and found that a firm's beta is influenced
by its leverage (Firth 1977: 103).
165
Logue and Merville (1972)

Logue and Marville concluded from their research that there is a significant correlation
between a firm's beta value and its liquidity ratios, leverage, dividend payout and pro-
fitability (Clark et al. 1979: 183).
Pettit and Westerfield (1972)

This study, on the other hand, found that accounting measures of risk do no better than
other financial factors in explaining beta values (Firth 1977: 103).
Rosenberg and McKibben (1973)

In their study, Rosenberg and McKibben examined whether accounting data and the
historical distribution of share returns can be used to predict the unsystematic risk and
current levels of beta. They concluded that accounting measures of risk have a certain
degree of explanatory power (Firth 1977: 102).
Lev and Kunitzky ( 1974)

Lev and Kunitzky found that smoothing of accounting data, that is, reducing its varia-
bility, reduces the level of a firm's beta (Firth 1977:103).
Various studies
Various other researchers have examined the association between different accounting
measures of risk and systematic risk. Most of these studies have found a fair, but
variable degree of association and concluded that although the degree of predictive
ability was not ascertained, accounting measures of risk have major explanatory power.
Examples of such studies include those of Breen and Lerner (1973), Gonedes
(1973,1975), Beaver and Manegold (1975) and Bildersee (1975) (Firth 1977:101).
The results of these studies have generally shown an association between market
measures of systematic risk and accounting measures of risk. It is, however, unclear
whether accounting data provides better forecasts of future betas than other financial
data (Firth 1977: 103).
166
4.2.9.4 Research on whether beta is the only factor that explains returns
King (1966)
The study of King was an investigation of whether industry-specific indices are a better
reflection of the returns-generating process. King found that around 50% of the volatility
of share returns are explained by the market index, while an additional 10% can be
attributed to an industry classification and, thus, that industry-specific indices do add
value. King's study has however been rejected by critics on the basis that his classifica-
tion scheme is not acceptable and that any industry factor identified is unsystematic
and, hence, a diversifiable risk (Harrington 1987:86,87).
Cohen and Pogue ( 1967)

Cohen and Pogue found some evidence in support of conclusions of King (1966), but
the usefulness of their study is hampered by severe statistical problems (Harrington
1987:87).
Merton (1973, 1980)

Merton postulated that share returns are determined by a risk measure based on stock
market volatility and that investors' portfolios are a linear combination of three different
portfolios, namely the market portfolio, the riskless portfolio and a portfolio with perfect
negative correlation with the riskless asset (Bicksler 1973:84; Oosthuizen 1992:10).
Merton developed an intertemporal CAPM that allows for changes in the investment
opportunity set. He argued that this model provides a conceptual explanation of the
observed phenomena that high (low) beta portfolios earn less (more) than expected
(Bicksler 1973:84).
Farrell (1974, 1975, 1976)

Farrell postulated that the industry effect is not unsystematic and diversifiable and that
there will be positive and negative correlation between various indices. Farrell deve-
loped a multi-index model which he showed to outperform a single-index model, in that
portfolios created through the use of a multi-index model reduce risk faster or require
167
less diversification to reach the same level of risk of portfolios created using the single-
index CAPM (Harrington 1987:87).
Fouse (1976)
Fouse investigated whether liquidity has an effect on the pricing of shares. Ex ante data
generated by investment analysts was used to project the returns for individual shares
and liquidity was added to this risk/return relationship. Fouse showed that liquidity has
a significant effect on expected returns and that a liquidity factor will have a significant
effect on share selection (Harrington 1987:81 ).
Bassu (1977) and Nicholson (1968)

Both these studies examined the effect of P/E ratios and Bassu confirmed the views of
Nicholson, that is, low PIE portfolios yield higher returns than do high PIE portfolios
(Seneque 1987:31; Linley 1992:16).
Levy(1978)andMayshar(1979)
Both Levy and Mayshar have contended that expected and realized returns are affec-
ted not only by beta, but also by variance. In conditions where diversification is con-
strained by, for example, the number of shares in the portfolio and/or transaction costs,
both these types of risk need to be included in the model as return-generating factors
(Harrington 1987:30).
Arnott (1980)
Arnott also examined whether more than one factor affects expected and realized
returns and found that the market faetor explains only about 30% of the variance of
returns. He concluded that other factors, such as an industry effect, cannot be ignored
when the risk characteristics of a portfolio are evaluated and that the CAPM ignores
such an industry risk factor (Harrington 1987:88,89).
168
Banz (1981)
In an investigation of the relationship between the size effect and share returns, Banz
added a firm size factor to the CAPM. Using data for the 1936 to1975 period, he found
that shares of firms with large market values have, on average, lower returns than do
those of small firms (Harrington 1987:85; Bradfield et al. 1988:13).
Reinganum (1981)
Using a different technique, Reinganum, like Sanz (1981 }, also found that portfolios
consisting of shares of small firms consistently earn higher risk-adjusted returns than
do portfolios of shares of large firms (Harrington 1987:83,85; Bradfield et al. 1988: 13).
Reinganum concluded that the basic CAPM is misspecified, as it seems if the per-
sistence of abnormal returns is caused by risk factors which are omitted from the CAPM
(Seneque 1987:31 ).
Blume and Stambaugh (1983)

Blume and Stambaugh showed that the technique used by Reinganum in 1981 caused
an upward bias in the estimates of the returns on the small firm portfolios. This upward
bias was caused by a bid-ask spread bias which is inversely related to firm size, and
the bias caused the premium of the size effect to be doubled (Bradfield et al. 1988: 13).
Lakonishok and Shapiro (1984)

The beta coefficient should be equal to the excess return on the market, but
Lakonishok and Shapiro found it to be much less than predicted by the CAPM. They
also found that returns on individual shares do not appear to have a specific rela-
tionship with systematic risk (Seneque 1987:31; Linley 1992: 17). They added the size
effect to a regression analysis consisting of beta and variance and found that the size
factor is the only significant variable (Harrington 1987:85).
Chi-Cheng, Reilly and Wong (1985)

The Chi-Cheng et al. study showed that the size effect is related to the liquidity effect.
They suggested that, due to the close relation between liquidity and size, a liquidity
169
factor can be one of a number of factors missing from the CAPM (Bradfield et al.
1988:13).
Amihud and Mendelson (1986)

In their study, conducted over the period 1961 to1980, Amihud and Mendelson showed
that liyuidity, as measured by the bid-ask spread, has a significant and positive rela-
tionship with return. It was found that illiquid (high bid-ask spread) shares earn higher
returns than liquid (low bid-ask spread) shares (Bradfield et al. 1988: 13, 18).
From the evidence available it does seem as if beta does not provide a full description
of risk and that other risk measures are also important in explaining share returns.
These conclusions, together with the findings on the problem of skewness of returns,
seem to indicate that the CAPM may possibly be misspecified (Harrington 1987:79).
4.2.9.5 Research on the estimation of beta
Breen and Lerner (1972)

Breen and Lerner examined the effect of having different choices of holding periods
when estimating beta. They calculated the betas for a number of shares, using linear
regression with monthly intervals,_ and found that, as the holding period lengthens,
significant changes in individual beta values occur (Harrington 1987: 109).
Phillips and Segal (1975) and Levhari and Levy (1977)

Both these studies showed that estimated beta values are different when different
estimation intervals are chosen. Levhari and Levy calculated the betas for a number
of shares, for the period 1948 to1968, using different monthly intervals and found that
most beta values change significantly with a lengthening in the interval period (Harring-
ton 1987: 110).
170
Frankfurter (1976)
Frankfurter investigated the choice of index on beta values. Three different indices
were used to estimate the betas for the same shares and it was found that some
shares' beta values are quite similar while others are not (Harrington 1987:115)".
Peseau ( 1977)
Peseau examined beta values for two overlapping periods, 1971 to 1974 and 1971 to
1975. Although no dramatic changes in beta values were expected, as only a year was
added on in the latter period, it was shown that the betas had changed significantly be-
tween the periods (Harrington 1987: 109).
Alexander and Chervany (1980)

Using data from 1950 to 1967, the aim of this study was to estimate the optimal interval
period which should be used to calculate beta. They found that a six-year horizon pro-
vides the least absolute errors, although this is only insignificantly better than a four-
year period (Harrington 1987: 110).
Hawawini (1983)
Hawawini estimated betas on a daily, weekly, two-weekly, three-weekly and monthly
interval over the period 1970 to 1973 and found the values to differ substantially. He
speculated that this occurs because firms with large market values lead the market and
their betas will increase as the interval shortened, while small firms lag the market and
their betas will decline with a shortening of the interval (Harrington 1987: 111 ).
Various studies have examined the problem of how to correctly estimate beta using
historical data and the conclusion at this stage seems to be that it is not known how
beta should be measured. The trial-and-error experimentation should thus continue
171
4.2.9.6 Research on the estimation of the market return and the validity of the
market proxy
Roll (1977, 1980)

Roll (1977) provided a strong critique on the empirical testing of the CAPM, based on
the argument that the CAPM is basically untestable. Keogh ( 1994:41-44) and Van Rhijn
(1994:126-128) summarise the reasons for Roll's critique as the following:
• The CAPM is an expectational model and can only realistically be tested if the
composition of the true mean-variance efficient market, containing all assets, is
known and used in the tests.
• All other implications of the CAPM stem from the efficiency of the market portfolio
and cannot be tested independently.
• The use of a proxy for the market portfolio creates two basic problems. Firstly, the
proxy may be mean-variance efficient even though the true market portfolio may
not be. Secondly, the various market proxies will be correlated with each other
and with the market portfolio, whether they are mean-variance efficient or not.
• Misspecification of the market portfolio will create bias and non-stationarity in the
risk/return relationship.
These factors imply that empirical tests of the CAPM have not been tests of expecta-
tions, but of what has actually occurred (Harrington 1987:75). All the tests have there-
fore been incorrect , their results should be treated with caution and it would seem that
it is highly unlikely that a correct and unambiguous test of the theory will be accom-
plished in the near future (Dobbins et al. 1994:65; Van Rhijn 1994: 126).
In a follow-up critique, Roll (1980) also criticised the use of the CAPM in the evaluation
of the performance of portfolio managers. This procedure calls for a comparison of the
performance of the managed portfolio with that of an unmanaged portfolio bearing the
same level of risk, and Roll argued that the use of a proxy for the market portfolio as
a benchmark invalidates the evaluation process. This is caused by the impossibility to
172
establish whether the risk-adjusted returns are attributable to a manager's ability or to

the inefficiency of the market proxy. Roll showed that when the performance of different
managed portfolios are compared through the use of different market proxies, the
rankings of the evaluation can be reversed (Dobbins et al. 1994:65).
Gay (1982:190) concludes that although Roll's critique is strictly correct, it is rather
over-exaggerated. All that can really be concluded is that it is unlikely that it will ever
be known whether the CAPM is true, as tests of it will almost certainly deliver imperfect
information.
Solnik (1977)
Solnik extended Roi l's critique of the CAPM to include international tests of the CAPM
and concluded that, as in domestic tests, the problem with the market proxy persists
and that the true mean-variance efficient market portfolio had not been used in these
tests (Berges-Lobera 1982:73,74).
Friend, Westerfield and Granito (1978)

Friend et al. examined the choice of index as market proxy and found that the use of
different indices renders different regression coefficients. They concluded that the
choice of index will have a significant effect on the analysis (Harrington 1987:75-78).
Mayers and Rice (1978)

Mayers and Rice partially refuted Roll's 1977 and 1980 critique of the CAPM. They
showed that the CAPM can provide meaningful conclusions with portfolio performance
tests. As long as the chosen market index consists of a high proportion of the total
market value of shares, it can be an acceptable benchmark for performance mea-
surement and beta estimation (Dobbins et al. 1994:67).
Stambaugh (1982)
The market proxy issue was also examined by Stambaugh. Through the inclusion of
bonds and real estate, he constructed broader market indices and found that the tests
173
are not really sensitive to the choice of market proxy (Bradfield et al. 1988: 12).
Carleton and Lakonishok (1985)

Carleton and Lakonishok investigated two aspects of estimating the market return,
firstly whether simple or compound rates should be used and, secondly, whether the
index used should be value or equally weighted. They examined market returns using
different holding periods of between five to 25 years and found differences between
geometric and arithmetic average share returns. They further found that the differences
between the two methods do not remain constant. Carleton and Lakonishok also found
that an equally weighted index provides higher returns over all the holding periods and
suggested that these results are due to differences in risk, as smaller shares have a
heavier weighting in an equally weighted index.
Carleton and Lakonishok reached two basic conclusions from their study. Firstly, they
suggested that the geometric method should be used to measure changes in returns
for a buy-and-hold strategy over more than one period, while the arithmetic approach
would be preferable when measuring performance over a single historical period.
Secondly, they concluded that the value-weighted index provides a better reflection of
the market and of investors' experiences. Other researchers have, however, disagreed
with this conclusion (Harrington 1987:167, 168, 170).
From Roll's criticisms it can basically be concluded that the use of indices as proxies
for the market portfolio has serious implications for tests of the CAPM. Roll showed it
is possible that the relationship between beta and returns can be shown to be linear
when the market proxy is efficient. Thrs is unfortunately not a true test of the CAPM as
one is not working with the true SML, but only with an estimated SML. It would, there-
fore, seem unlikely that the CAPM could be truly tested in the foreseeable future (Reilly
& Brown 1997:317,318).
174
It should be noted that the market proxy issue does not invalidate the CAPM, it is only
an indication of the measurement problems associated with testing the model as well
as with using the model to evaluate portfolio performance (Reilly & Brown 1997:321 ).
On the issue of estimating the market return, no definite conclusions can be reached
at this stage. Various issues, such as the choice between arithmetic and geometric
methods, the choice of an appropriate historical period and the choice of a suitable
index remain unresolved (Harrington 1987: 185).
4.2.9.7 Research on the estimation of the risk-free rate of return
Grey (1974)
According to Grey, among others, the rate of a long-term government bond or high
quality industrial bond should be used as proxy for the risk-free rate. Although this
proxy is only an approximation and not based on theory, its use is advocated because
empirical studies have shown the intercept of the SML to be higher than Treasury bill
rates and the activities of central banks preclude a free market rate (Firer 1993:27).
Ibbotson and Sinquefield (1979)

Ibbotson and Sinquefield examined the use of US Treasury bills as a proxy for the risk-
free rate over a 54-year period, 1926 to 1979. They found significant negative cova-
riance between share returns and the Treasury bill rate as well as strong serial cor-
relation in the Treasury bill return series. This suggested a pattern to the returns over
time and that the beta for Treasury bills will not be equal to zero, that is, it is not risk-
free (Harrington 1987: 151; Firer 1993:27).
Casabona and Vora (1982)

This study also examined the use of US Treasury bills and found that, even after con-
verting the Treasury bill rate to a perpetuity, there is significant correlation with the mar-
ket return (Harrington 1987: 151 ).
175
Carleton and Lakonishok (1985)

Carleton and Lakonishok found that average returns on US Treasury bills are routinely
used in the applications of the CAPM. They, however, advocated the use of long-term
bonds as a proxy for the risk-free rate since estimates using Treasury bills are also
used to analyse long-term investments, which could include investments in the Trea-
sury bills (Firer 1993:26).
Although doubts have been expressed over the use of various instruments as proxies
for the estimation of the risk-free rate, it can in general be concluded that beta is a valid
and quite complete measure of risk (Mclaney 1997: 177).
4.2.9.8 Research on the assumption of no taxes and no dividends
Miller and Modigliani (1961) and Modigliani and Miller (1963)

These researchers showed that capital gains will have a greater after-tax value in
conditions where dividends are taxed at a higher rate than capital gains. Where no tax
differentials exist, given an investment policy, dividends and capital gains will have
equal attraction (Harrington 1987:43).
Brennan ( 1970, 1971)

Brennan also examined the effect of taxes and suggested that investors in different tax
brackets are interested in different shares and this will cause multiple CMLs if taxes are
important enough to affect share prices. Brennan also showed that, if dividends are
certain, the risk/return relationship will still be linear even if dividends and capital gains
are taxed at different rates (Harrington 1987:43).

Black and Scholes examined the effect of dividends on share returns by adding a divi-
dend payout variable to the CAPM. They found this variable to be not significantly
different from zero, over both the entire period and every subperiod of the study. They
concluded that this implies that there is no significant difference between the expected
176
returns on high and low dividend-yield shares and, therefore, there is no significant
relationship between dividends and share returns, both before and after taxes (Harring-
ton 1987:80; Bradfield et al. 1988:12).
Bar Yosef and Kolodny (1976)

The results of this study contradicted the findings of Black and Scholes (1974), as well
as those of the 1982 study of Miller and Scholes (Harrington 1987:43).
Litzenberger and Ramaswamy (1979)

Litzenberger and Ramaswamy found that investors with different characteristics prefer
different dividend yields and the higher the yield the lower the aversion for dividends.
They concluded that the expected before-tax return on a share has a linear relationship
with its dividend yield and systematic risk (Harrington 1987:80).
Miller and Scholes ( 1978, 1982)

Miller and Scholes (1978) argued that investors are able to efficiently transform divi-
dends into capital gains through efficient leverage, and if investors use tax shelters
there may be no difference between the before-tax return on shares paying dividends
and those that do not (Bradfield et al. 1988: 12).
In 1982 they criticized the findings of the 1979 study of Litzenberger and Ramaswamy
and suggested that their conclusions were based on a dividend announcement effect
and not on a systematic tax effect. After correcting for the information effect of dividend
announcements, they concluded that they could not find any significant relationship
between share returns and dividend yields. They did not state that there was no such
effect, only that it could not be found using short-term measures (Harrington 1987:80).
Vandell and Stevens (1982)

This study investigated the effect of taxes on share prices by examining the differences
between the basic CAPM and a CAPM that includes a personal tax variable. Using
analysts' forecasts, 90-day US Treasury bills and the market return, they examined the
177
ex post CML and a plane derived from a combination of risk and yield. They found
negative correlation between beta and yield, that is, high-yield shares tend to have low
betas, and they concluded that this means that beta does not capture the effect of taxes
(Harrington 1987:80).
Various studies
Several other studies have also examined the tax and dividend effect. Most have identi-
fied the existence of such effects, although the evidence is not unanimous. These
studies include those of Long (1978), Rosenberg and Marathe (1979), Stone and
Bartter (1979), Blume (1980), Gordon and Bradford (1980), Elton, Gruber and
Rentzler (1983) and Keim (1985) (Bradfield et al. 1988:13; Reilly & Brown 1997:309).
4.2.9.9 Research on the CAPM assumptions in general
Brennan (1969)
Brennan examined the assumption that investors can borrow and lend unlimited
amounts at the risk-free rate and showed that differential borrowing and lending rates
means that two different lines are going to the efficient frontier. This has the implication
that investors can either borrow or lend, but borrowing portfolios (subject to higher
interest rates) are less profitable than lending portfolios (Reilly & Brown 1997:306).
Mayo (1971) and Levy (1978)

Both these studies showed that the CAPM may still be usable, even if several of the
assumptions are invalid. They showed that if markets are not perfect, meaning that
there are taxes, transaction costs, less than perfect competition, information inefficiency
and investors are unable to maximise their expected utility of terminal wealth, the
CAPM can still be used (Harrington 1987:46).
Mayers (1972) and Merton (1973)

These studies have developed extensions of the CAPM under conditions where the as-
sumptions were relaxed. In the case of Mayers, a model was constructed which incor-
178
porated the existence of a non-marketable asset such as human capital. Merton, on the
other hand, derived a version of the CAPM under the assumption that trading took
place continuously over time and, secondly, that returns on assets are lognormally
distributed (Bradfield et al. 1988: 12).
Black (1972), Gibbons (1982), Stambaugh (1982) and Shanken (1985)

Black developed a zero-beta CAPM that does not require the existence of a risk-free
asset, and several studies have tested this model. Both the Gibbons and Shanken
studies rejected the model, while Stambaugh found evidence in support of the model
(Reilly & Brown 1997:307,308).
Both Levy and Sarnat (1994:336) and Reilly and Brown (1997:310) note that the CAPM
should not be judged on the basis of its assumptions, and although many of the
assumptions clearly do not hold, in reality it cannot be rejected on these grounds, but
rather on how well it describes and explains the risk/return relationship that exists in
the real world. It is on the basis of this proposition that no conclusions on the validity
of the CAPM are provided in sections 4.2.9.8 and 4.2.9.9.
4.2.9.10 General research on major non-US markets
Modigliani, Pogue, Scholes and Solnik (1972)

In the first major non-US study, Modigliani et al. found, over the five-year period stu-
died, that the CAPM does provide a good description for the French, UK and Italian
markets, while the CAPM is not supported for the German market (Gay 1982: 183).
Altman, Jacquillat and Levasseur (1974)

Altman et al. examined the French market over a seven-year period using 430 shares.
They found that the CAPM explains the variability of share returns just as well in France
as it does in the US. The fact that they did not report the estimated alpha coefficients,
however, limits the conclusions that could be drawn from this study (Gay 1982:185).
179
Lau, Quay and Ramsey (1974)

The study of Lau et al. examined 585 shares listed on the Tokyo Stock Exchange for
a five-year period and found strong support for the CAPM. The study also found no
evidence that the CAPM misprices low and high-beta shares, but it should be noted
that the study purposely omitted shares which were infrequently traded and made no
conclusions regarding the explanatory power of the CAPM (Gay 1982: 185, 186).
Pogue and Solnik (1974) and Friend and Losq (1979)

In a study that expanded the earlier 1972 study of European markets by Modigliani et
al., Pogue and Solnik added the Swiss, Belgium and Netherlands markets to the origi-
nal four markets studied. They concluded that their findings suggest that the CAPM is
better suited for Western European markets than for the US markets. Friend and Losq
examined the 1974 study of Pogue and Solnik and concluded that the study had some
problems with sampling bias. The Pogue and Solnik comparison between European
results and NYSE shares consisted of only a small random sample of NYSE shares,
of which the largest companies were Western European (Gay 1982: 183, 185).
Palacios (1975)
Palacios performed the first study on a non-industrialised nation's capital market,
namely Spain. The study examined 150 shares over a 13-year period. No evidence was
found to support the CAPM and the risk/return relationship was found to be negative.
In addition Palacios found that high-beta shares have negative alphas, while low-beta
shares have positive alphas. Due to these extreme findings, Palacios divided the study
into two subperiods, one of eight years (a pre-mutual fund period) and one of five years
(a post-mutual fund period). By making this distinction, Palacios found some evidence
in support of the CAPM in the post-mutual fund period. However, these findings are not
sufficient to provide any strong support for the CAPM (Gay 1982: 186).
Guy ( 1976, 1977)

In the 1976 study, Guy examined 100 shares in two separate five-year periods on the
UK market and found little support for the CAPM. The findings seem to indicate that the
180
CAPM is only really applicable to UK companies with large market capitalisation. It

should be noted, however, that this study suffers from a number of problems, namely
the small sample examined, the relatively short time period used and measurement
errors. Guy re-examined the German market in 1977. The study covered a 10-year
period and concluded that the earlier study of Modigliani et al. in 1972 suffers from a
time period problem and that the CAPM does in fact suit the German market (Gay
1982: 183, 187).
Morrin (1976)
Morrin investigated the Toronto Stock Exchange in Canada and concluded that the
CAPM seems to be only partially correct and that alpha is significantly different from
zero (Gay 1982:185).
Errunza (1979)
This study of the Brazil market examined 64 shares over a period of approximately four
years. Errunza found that only four shares had insignificant betas and concluded that
the CAPM has good explanatory power. However, the conclusions that can be drawn
from this study are hampered by the fact that the alpha coefficients were not reported
and also by the short time period of the study (Gay 1982: 187).
Levy(1980)
Levy examined the Israeli capital market using 104 shares over an 11-year period. Both
quarterly and annual data on returns were examined and it was found that the CAPM
holds for annual data and does not misprice shares, but it that it fails for quarterly data.
Levy argued that the reason for the CAPM's failure with quarterly returns is because
it does not take full account of investors' individual time horizons. However, Levy's
findings are subject to potential errors in that the work was not adjusted for measure-
ment errors and it is unclear whether any adjustments were made for any estimation
bias caused by infrequently traded shares (Gay 1982: 187, 188).
181
Ritsonis (1980)
In a study of the Greek capital market, Ritsonis found that both systematic and unsys-
tematic risk are significant for Greece (Gay 1982: 188).
Ibbotson, Carr and Robinson (1982)

The study of Ibbotson et al. investigated equilibrium in international capital markets.
The basic postulate underlying their study was that if there is complete integration in
international capital markets, then the relationship between expected return and the
relevant attributes of shares may be described by a type of International Capital Asset
Pricing Model (ICAPM). It is expected that the return on each share will be proportional
to its beta value relative to the world market portfolio. However, a different CAPM may
hold for each country if capital markets are totally segmented. The relationship between
beta and return may be different in each country and within each country the expected
return on a share may be proportional to its beta value relative to the market portfolio
of that country. It is, however, expected that there will be some relationship between
expected return and the total risk of a market portfolio across countries (Sharpe
1985:719).
The study of Ibbotson et al. provided some evidence on the average return and stan-
dard deviation of return, in US dollars, for the share indices of 16 countries over the
period 1960-1980. The countries involved were Australia, Austria, Belgium, Canada,
Denmark, France, Germany, Italy, Japan, Netherlands, Norway, Spain, Sweden,
Switzerland, the UK and the US. They measured, in US dollars, the beta value of the
excess return on each index relative to the world share index, together with the ex post
values of alpha for each country's index relative to the world index. It should be noted
that all the indices used were only proxies for the various countries' market portfolios,
since only shares were included (Sharpe 1985:719).
Ibbotson et al. found that the alpha values varied substantially, but none were statisti-
cally different from zero and concluded that this does not confirm complete international
capital market integration. They also found the relationship between average return and
182
standard deviation across countries to be nearly equally as strong as the relationship

between beta and average return. They concluded that this confirmed that the inter-
national capital markets are not completely segmented (Sharpe 1985:719,720).
It can be concluded that the results from the various international studies have gene-
rally confirmed what is known about the CAPM from the US studies and, thus, no new
or differing conclusions can be drawn from these studies. On the issue of equilibrium
in international capital markets, the evidence is as yet not powerful enough to reject or
accept the hypothesis, that is, that the market is completely integrated or totally seg-
mented, or to suggest where it is lying between these two extremes (Sharpe 1985:720).
4.2.9.11 South African research
Booth (1979)
The objectives of Booth's study were to determine the strengths and weaknesses of the
CAPM in the South African context and to establish whether it can be applied or not.
He concluded that the CAPM cannot successfully separate systematic and unsyste-
matic risk and that it only performs well when the systematic portion is high (Van Rhijn
1994: 142, 143).
Gilbertson (1979)
This study investigated the estimation of the risk-free rate in South Africa and proposed
that short-term, rather than long-term, Government bond rates should be used. This will
ensure that investors' short share-investment horizons are reflected (Firer 1993:28).
Boshoff (1986)
Boshoff, on the other hand, suggested the use of the prevailing trading rates of the
shortest-term gilts as proxy for the risk-free rate. These rates will reflect the highest
return available to investors without any risk being borne (Firer 1993:28).
183
Due to the limited number of studies which had been done on the CAPM in the South
African context by the early 1980s, no definite conclusions can at this stage be reached
on its applicability, validity and explanatory power under South African conditions.
4.2.10 Conclusion
Reilly and Brown (1997:322,323) provide the following summary of the status of the
CAPM up to the mid-1980s:
The tests of the CAPM indicated that the beta coefficient for individual
shares were not stable, but the portfolio betas generally were stable
assuming long enough sample periods and adequate trading volume.
There was mixed support for a positive linear relationship between
rates of return and systematic risk for portfolios of shares, with some
recent evidence indicating the need to consider additional risk
variables or a need for different risk proxies.
In addition, several papers by Roll criticised the tests of the model and
the usefulness of the model in portfolio evaluation because of its
dependence on a market portfolio of risky assets that is not currently
available.
Many studies seem to indicate that the basic CAPM is inaccurate. They have identified
a factor or factors that appear to be omitted from the CAPM and these factor models,
particularly the APT model, were developed in an attempt to capture all aspects influen-
cing the behaviour of share returns. The APT requires only limited assumptions, but it
is important to note at this time that the factors and their impact on share returns can
change quite rapidly. Also, the factor models are descriptive by nature and may or may
not be behavioural, that is, explain what should be (Harrington 1987:93).
In the next section of this chapter the APT, which is the predominant alternative to the
CAPM, will be examined.
184
4.3 THE DEVELOPMENT OF THE ARBITRAGE PRICING THEORY
The most appealing aspect of the APT is that it has less restrictive assumptions and is
more general than the CAPM. The APT is also concerned with the relationship between
risk and return, but is based on the view that several sources of risk affect expected and
realized returns. These APT risk factors are not company-specific by nature and repre-
sent broad economic forces (Jones 1998:242,245).
The APT has important implications for investment management, in that if investors are
able to identify the main risk factors which affect most securities, they would then be
able to improve the design and performance of their investment portfolios. This, how-
ever, is also the APT's biggest drawback, since it offers no clues as to the identity of the
risk factors that affect share returns. It is only through empirical research that the iden-
tity and importance of the various risk factors can be established. Some researchers
have identified and measured a number of risk factors and some have also concluded
that the theory of the APT is superior to that of the CAPM (Jones 1998:245).
In the following sections on the development of the APT, all these aspects are exa-
mined, that is, the nature and assumptions of the APT, risk and the APT risk factors,
tests and validity of the APT, the APT compared with the CAPM, how the APT can be
employed in investment decision-making, and empirical research related to the APT.
4.3.1 The nature of the APT
It is clear from the preceding section of this chapter that, although the CAPM is a simple
and enlightening model, it is beset by a number of problems. The main problems, iden-
tified by Rees (1995: 172, 173), are:
• It is dependent on a number of simplifying assumptions.

• The market is described as the only source of risk.
• Empirical evidence is mixed and the tests may not be genuine tests of the model.
185
• The model is unable to provide an explanation of some persistent anomalies in

market returns.
• It is difficult to make reliable and stable estimates of its parameters.
• There exists doubt that returns are only dependent on systematic risk (beta).
From these and other criticisms, Ross (1976) developed an alternative pricing theory
known as the APT in an attempt to provide a more meaningful and realistic model for
measuring risk. The APT describes the relationship between risk and return in terms of
several factors rather than the single market factor of the CAPM. Although the APT
does not specify what the factors might be, it does provide a testable multi-factor model
which is an interesting and appealing alternative to the CAPM (Harrington 1987:93;
Linley 1992:19).
Linley (1992:19) notes that the APT should not be seen as a perfect model. It has
failings, but it is an important alternative to the CAPM which may in time replace the
CAPM as the predominant asset pricing model.
The logic of the APT is founded on the basic arbitrage principle, namely the law of one
price, which states that two equivalent assets trading in the same competitive market
cannot sell at different prices. Thus, two shares in the same company must sell at the
same price as they are affected by the same factors and will have equal expected
returns. If they did not sell at the same price, investors would simultaneously sell the
higher priced share and buy the cheaper share and thus achieve arbitrage profits at no
risk to them (Seneque 1987:32; Jones 1998:242).
Jones (1998:242) adds further that the APT states that market prices will adjust to
eliminate arbitrage opportunities to achieve conditions of market equilibrium and when
a situation of market disequilibrium did arise, it will only require the actions of relatively
few investors to restore market equilibrium.
186
It is from these arbitrage arguments that the APT derives its share return relationships,
namely an arbitrage portfolio requires no new wealth, has no risk and offers a return
equal to the risk-free rate. If it did offer a return in excess of the risk-free rate, investors
would borrow at the riskless rate to buy the portfolio and thus earn arbitrage profits
without incurring any risk (O'Brien & Srivastava 1995:144; Pike & Neale 1996:288).
According to Harrington (1987: 191, 192) the principle of arbitrage seems intuitively logi-
cal and a reasonable description of investor behaviour. Investors are profit seekers and
believe that opportunities to make profit exist, but in efficient markets the active share
trading activities of these investors eliminate the arbitrage opportunities and over time
no investor can, on average, find arbitrage profit opportunities. However appealing the
APT might be, certain assumptions still apply to it. The theory has less restrictive
assumptions than the CAPM and it assumes that share returns depend partly on macro-
economic factors and partly on company-specific factors. The company-specific factors
can be eliminated through diversification, leaving only the macro-economic risk
variables as determinants of share prices (Pike & Neale 1996:288).
Both the APT and the CAPM expect the relationship between expected return and risk
to be positive, but the APT relates expected return to the risk-free asset and a range of
other common factors that systematically affect returns. It, therefore, has a more general
view of risk than the CAPM, which views it purely on the basis of beta related to the
market portfolio (Harrington 1987: 189; Ross et a/.1990:286).
The APT states that different shares will be more sensitive to certain factors than
others, but it does not postulate what these explanatory factors might be. It does, how-
ever, eliminate a serious problem associated with the CAPM, the need to identify the
market portfolio. The two major areas of uncertainty associated with the APT, which are
also the two areas wherein most of the research and empirical testing of the model have
been conducted, are, firstly, the identity of the macro-economic variables that affect
returns and, secondly, the risk premium for each variable and the measuring of the sen-
sitivity of share returns to these factors (Pike & Neale 1996:288).
187
Before the research into and empirical testing of the APT can be reviewed and conclu-
sions drawn from it, it is necessary to examine all aspects of the theory, its implications
and compare it with those of the CAPM. The next section therefore reviews the assump-
tions of the APT.
4.3.2 Assumptions of the APT
The APT is a simpler theory than the CAPM and requires fewer and less rigorous
assumptions (Seneque 1987:32). However, there are a number of assumptions that it
shares with the CAPM:
• Investors require return as compensation for risk, they are risk averse and wish
to maximise their terminal wealth (Harrington 1987:193; Laing 1988:17; Linley
1992:20). Harrington (1987: 193) adds that it is important to note that no assump-
tion is made about the distribution of returns, nor does it assume that investors
base their decisions on information about mean returns and the variance of
returns.
• Investors are able to both borrow and lend at the risk-free rate of interest
(Harrington 1987: 193; Linley 1992:20). Harrington (1987: 193) states that this is
a logical part of any pricing model but, as with the CAPM, nothing is said about
the borrowing and lending rates.
• Capital markets are perfectly competitive and there are no market imperfections
such as transaction costs, taxes and restrictions on short-selling and all relevant
information is simultaneously and costlessly available to all investors (Harrington
1987:193; Laing 1988:17; Linley 1992:20). Harrington (1987:193) also notes
that, as with the CAPM, these imperfections cause difficulties in pricing, and
hence they are assumed away.
188
In addition to these assumptions, the APT also requires some unique assumptions of
its own. These are:
• Investors are in agreement about the number and nature of the factors which are
important in the systematic pricing of assets (Harrington 1987: 193; Laing
1988:17; Linley 1992:20). This unique assumption of the theory suggests that
returns are generated by a number of factors, not only the market, and that
investors believe these to be all the factors which are important in the pricing of
assets (Harrington 1987: 194; Linley 1992:20,21 ).
• No arbitrage profit opportunities are available. This assumption describes in-

vestor behaviour. Investors are active in their pursuit of opportunities for riskless
profits, and through their activities these opportunities are eliminated (Harrington
1987: 193, 194; Linley 1992:20,21 ).
Following from these assumptions, the next section examines a fundamental aspect in
asset pricing, namely risk, and how the APT views risk and describes the risk/return
relationship.
4.3.3 Risk and the APT
The underlying rationale behind the APT is that if a particular risk factor only affects a
certain sub-set of shares, then this element of risk will be priced in those shares in order
to ensure that share returns compensate for this additional risk factor. If markets are not
efficient and this does not occur, then arbitrageurs will be able to take advantage of
arbitrage opportunities and so eliminate risk in mder to earn abnormal returns (Rees
1995:173).
While it is possible for investors to eliminate unsystematic risk through diversification,

the APT asserts that investors cannot eliminate a share's exposure to various macro-
economic factors. The beta coefficients will, however, vary from share to share, de-
189
pending on the exposure of each share to each macro-economic variable (Dobbins et

al. 1994: 131 ).
As stated above, different shares will have different sensitivities to these systematic risk
factors and it follows that investors' risk preferences will, to a degree, be influenced by
and reflect these risk variables. Each investor will have a different attitude and will, thus,
construct their portfolios based on the exposure of the shares to each of these risk
factors. It follows that it will be possible, if the market price of these risk factors and the
sensitivity of each share's returns to changes in the factors are known, to estimate the
expected return for each share (Jones 1998:245).
The APT states that the risk premium for each share will depend on each factor's
premium in combination with the sensitivity of each share to these factors. It relates the
expected return of a share to the risk-free rate of return and to the various factors that
systematically affect returns (Linley 1992:21; Jones 1998:243).
The factor formula for the APT can be formally stated as per formula 4.3
Formula 4.3 The arbitrage pricing theory factor-model
for:
j 1 ton
n = Number of assets
where:
= The rate of return on asset j during a specified time period

190
The expected rate of return on asset j
b1 = The sensitivity of asset j's returns to a factor
f = A common factor with zero mean that influences the returns on all
assets under consideration
= A random error term, unique to asset}, that, by assumption, is comple-
tely diversifiable in large portfolios and has a mean of zero
Source: Weston and Copeland (1992:422), Reilly and Brown (1997:298,323) and Jones
(1998:243,244).
Reilly and Brown (1997:298,323) provide some further elaboration on two elements of
equation 4.3:
• The term f refers to the various systematic macro-economic factors that are
expected to have an impact on share returns and according to the theory there
are many such factors.
• The term b determines how each share reacts to the common systematic macro-
economic factors.
The APT equation asserts that there is a linear relationship between expected returns
and the various sensitivities to the major factors in share returns (Sharpe 1985: 197).
The equation states that the expected return of a share is related to its factor betas and
these factors represent risk that cannot be diversified away. It follows that the higher
a share's sensitivity to a specific factor, the higher its risk and the equation states that
the expected return on a share is a combination of the risk-free rate plus compensation
for each type of risk borne (Ross et al. 1990:310).
191
According to Jones (1998:243), a factor model like the APT is based on the principle
that there are various risk factors that affect expected and realized returns. These
factors are broad economic factors, are not company-specific and reflect unexpected
changes of the factors from their expected values.
Reilly and Brown (1997:299,323), however, note that the APT does not identify these
factors. The number and nature of these factors will be investigated when the empirical
studies are reviewed. In the next section the nature and characteristics of the APT
factors, and the principles of factor identification, will be examined.
4.3.4 Identification of the factors to be included in the APT
As previously stated, Ross ( 1976), in the development of the theory, made no statement
about the identity of the factors that affect returns. He did, however, state that the whole
theoretical and empirical structure should be explored in order to obtain a better under-
standing of the economic forces/factors that systematically affect actual returns (Van
Rhijn 1994: 153).
Harrington ( 1987: 190) restates the problem associated with the identification of the
factors that affect share returns, by pointing out that the APT makes no statement and
provides no guidance about the nature of the factors, their magnitude or their signs. Nor
does it provide any guidance on how these aspects should or could be determined.
According to Brevis (1998:88) a factor that is priced could be described as an uncertain

economic variable that will affect the expected returns on shares or portfolios of shares.
Fogler (1982) identifies the following three criteria that can be used to identify these
priced factors (Brevis 1998:89):
• The factor should be significant and stable, meaning that the covariance between
the identifiable factors should be the same under all economic scenarios.
192
• The factors should be representative of all systematic movements in returns.

• The identifiable factors should be independent from one another. If the correlation
between the factors is equal to zero, returns could be explained by only adding the
effect of one factor to the effect of the previous factor.
Radcliffe (1990) classifies the factors that influence returns as follows (Brevis 1998:88):
• Factors that influence all the shares.

• Factors that only affect a sub-set of all the shares.
• Factors that only affect a single share.
Factors that only affect a single share can be reduced through diversification and are
not priced factors. Those that affect all the shares should be considered as priced
factors, but it is unclear whether those that only affect some of the shares should be
included. According to Radcliffe (1990) this will depend on whether the uncertainty
related to these factors could be eliminated through rational diversification. Those that
cannot be eliminated through rational diversification should be considered as priced
factors (Brevis 1998:89).
To conclude, the APT was developed without the factors affecting returns being identi-
fied. These factors are, however, likely to differ between industries and even between
shares within industries (Laing 1988:22). It is only through empirical investigation that
these factors and their characteristics can be established. Before these empirical
findings are reviewed, it is first necessary to consider the theory's validity, to compare
it with the CAPM and to consider its application in investment decision-making.
4.3.5 Tests and validity of the APT
The APT has relatively few assumptions, but it provides little help with the identification
of the priced factors and the relationship between these systematic factors and expec-
ted returns. Herein lies the main challenge for researchers in establishing the validity
193
of the theory. Three questions, that need to be answered through empirical testing,
have been identified to establish the validity of the APT (Harrington 1987: 194 ):
• Does more than one factor explain returns?

• Are the APT factors the only risk factors which are priced?
• What are these APT factors?
4.3.5.1 The number of factors determining returns

Only a review of the empirical evidence can provide a definite answer to this question,
but it can generally be concluded that the research on the APT has found there to be
more than one priced factor. Harrington (1987:195,200), however, notes that the re-
search has also found the market to be an important factor in explaining returns and
that the controversy surrounding this issue is ongoing and far from over.
4.3.5.2 The APT factors are the only risk factors which are priced
To answer this question, empirical findings must show that the factors identified are not
correlated to each other and, also, that no other important factors exist. If the empirical
results confirm these conditions, this would suggest that the APT provides a good de-
scription of the pricing mechanism (Harrington 1987:201 ).
Harrington (1987 : 202) also notes that the methods for factor estimation and testing
are not as yet entirely resolved. The question of which method to use and how to inter-
pret the results must still be solved through ongoing research and, thus, it is still un-
clear whether the APT is superior to the single-factor CAPM.
4.3.5.3 The factors of the APT

Again, the results of empirical testing are inconclusive as to the number and identity of
the factors that affect returns. It is only through a review of the empirical tests that the
current knowledge regarding the APT factors can be ascertained, but it is worth noting
that different studies have found different factors for different periods and datasets.
Although these studies suggest a great deal of instability in these factors, the research
194
continues unabated and a comprehensive and stable list of factors cannot be provided
at this stage (Rees 1995: 173).
It is also useful to note that, up to the mid-1980s, the macro-economic variables Iden-
tified were:
• Industrial production (or the market index).

• Changes in a default risk premium.
• Twists in the yield curve.
• Unanticipated inflation.
• Changes in the real rate.
(Chen, Roll & Ross (1986), in Weston & Copeland 1992:423.)
The only conclusion that can be reached at this stage is that the factor issue sur-
rounding the APT remains unresolved and, hence, that the controversy about the APT
versus the CAPM also remains unresolved. In the next section the comparison and
competition between these two models are examined in more detail.
4.3.6 The APT versus the CAPM
Ross et al. ( 1990:309) state that the APT and the CAPM are alternative models of risk
and return, that there are various differences between them and that it is worthwhile to
consider these differences. The following are some of the differences noted:
a. Seneque (1987:34), Laing (1988:31) and Van Rhijn (1994:154,155) list six diffe-
rent reasons why the APT can be considered more robust than the CAPM:
• The APT requires no assumptions about the distribution of returns.

• The APT requires no strong assumptions about investors' utility curves,
other than that they are assumed to be motivated by risk aversion and greed.
195
• The APT states that equilibrium returns can be and are dependent on many
factors, not only beta.
• The APT does not require that the entire universe of assets (the total market
portfolio) be measured for empirical testing. It can be used to establish the
relative pricing of any sub-set of assets.
• The CAPM requires that the market portfolio be known and be efficient, while
the APT places no special emphasis on the market portfolio.
• The APT can easily be extended to a multi-period investment framework.
b. Assuming a one-factor state, in the APT the beta of a share measures its sensi-
tivity to the factor, while in the CAPM the beta measures the share's sensitivity to
movements in the market portfolio (Ross et al. 1990:308).
c. A major disadvantage of the APT is that it does not define the nature and identity
of the factors that affect share returns, but requires that these be established
through empirical testing (Seneque 1987:34).
This provides the CAPM with an important advantage in that it states that only
one, well-defined factor is priced. Although this has had its problems with empiri-
cal testing, it may account for its continued wide-spread use (Seneque 1987:34).
d. Ross et al. (1990:310) identifies another important advantage the CAPM has over
APT. The CAPM ensures the discussion of efficient sets, which 1s of great intuitive
value, but this is not easily achieved through the APT.
e. Ross et al. (1990:310) also identify an off-setting advantage of the APT. The APT
equation adds factors until each share's unsystematic risk is uncorrelated with
every other share's unsystematic risk. Through this process it is not only shown
that unsystematic risk decreases with diversification. but also that systematic risk
does not. This result is also achieved with the CAPM, but to a lesser extent as
unsystematic risk could be correlated across shares.
196
f. Ross et al. ( 1990: 310) state further that the APT has a significant advantage in
that it can handle multiple factors, while the CAPM cannot. The advantage stems
from the fact that multi-factors are probably a better reflection of reality, as it is
probably more realistic that a model should abstract from both industry-wide and
market-wide factors in order to achieve a situation where the unsystematic risk of
a share becomes uncorrelated with the unsystematic risk of all other shares.
g. According to Sharpe (1985:199), the CAPM supplements the APT and, hence,
these two models should not be compared from a replacement point of view. The
two models combined make stronger predictions than they could on their own and
as a combination they improve investment decision-making.
h. Tests of both models have proved to be controversial, but have in general showed
the APT to have greater explanatory power than the CAPM. On the other hand,
investment analysts have criticised the APT in that computer analysis programs
determine the factors and these factors then have no meaning on their own. This
makes it difficult to develop forecasts based on the APT (Viscione & Roberts
1987:206). It is from this experience of investment analysts, that Viscione and
Roberts (1987:206) conclude that investment consultants continue to use a multi-
factor CAPM, which include factors such as dividend-yield, firm size, trading
volume and the market factor, rather than the APT.
Following on the comparison between the two models, it is now appropriate to examine
the application of the APT in investment decision-making. Obviously its application
overlaps those areas already examined with the CAPM, but there are specific appli-
cations for which it is particularly well suited.
197
4.3.7 Application of the APT in investment decision-making
The introduction of the APT has led many to believe that it would solve the theoretical,
empirical and practical problems associated with the CAPM (Harrington 1987:218). It
is argued that it offers an opportunity for and approach to strategic portfolio manage-
ment. Through the identification of the factors which affect share returns, investors
should be able to structure their portfolios in such a way as to improve its design and
performance (Jones 1998:245).
Harrington (1987:219) notes that many portfolio managers have for a long time con-
sidered basic changes in the structure of the economy when making share selection
choices and establishing their portfolio management practices. With the development
of the APT, these models of economic sensitivity could be formalised. Computer models
which can estimate the sensitivity of shares and portfolios to economic factors have
been developed, and these assist portfolio managers to select shares and structure
portfolios to meet certain factor-sensitivity criteria.
Jones ( 1998:246) also emphasises the application of the APT in portfolio management.
It provides managers with the opportunity to design investment strategies whereby their
portfolios would be exposed to one or more risk factors or to structure portfolios in such
a way that its exposure to unexpected changes in a factor or factors mirror that of the
market as a whole.
Through active investment management, managers who believe they could forecast
factor realization would be able to construct portfolios which emphasise or de-empha-
sise a specific factor or factors. This would enable these managers to then select
shares which have exposure to the remaining risk factors that were exactly proportional
to the market. In this way managers would be able to outperform the market for the
forecast period, on the condition that they could accurately forecast the unexpected
changes in the risk factor or factors (Jones 1998:246).
198
Hence, application of the APT provides opportunities for active investment management
and improved share selection and portfolio construction, but this is constrained by
investment managers' ability to measure these factors accurately and to accurately
predict unexpected changes therein. Only time and empirical research will tell how suc-
cessful such APT-based share selection and portfolio management models are. This
now leads to the final part of this discussion on the APT, the review of the empirical
research thereon up to the mid-1980s.
4.3.8 Research related to the APT
Sharpe ( 1985:394) states the following about the APT and the testing thereof:
Since the APT assumes a factor model of security returns, any test
of its predictions must incorporate such a factor model and be, in
effect, a joint test of the equilibrium theory and the appropriateness
of the selected factor model.
Moreover, the APT makes relatively weak predictions. It does imply

that when all pervasive factors are taken into account the remaining
portion of return on a typical share should be expected to equal the
riskless interest rate. Testing this implication is possible iri principle,
but difficult in practice. It may never be possible to find a set of
shares so diverse that consistency of the zero factor with riskless
interest rates can be used as a test of the APT.
A more promising test concerns the prediction that share expected

returns will be related only to sensitivities to "pervasive factors". In
particular, there should be no relationship between expected returns
and shares' nonfactor risks.
Two basic ways to test the APT exist:
• The first is known as factor tests, whereby factor analysis is employed in order to
establish the number of priced factors and their magnitudes (Dobbins et al.
1994:131; Levy & Sarnat 1994:345).
199
• The second is specified factor tests, whereby hypotheses are formulated about the
factors which are relevant to the return-generating process, for example, inflation,
interest rates, changes in production, etcetera. These factors are then used in a
regression analysis to establish whether they explain share returns. It should,
however, be noted that this can be an ad-hoc method as there exists no theo-
retical foundation why these factors are more relevant than others in explaining
returns. Economic theory can, however, be used as guidance (Dobbins et al.
1994:131; Levy & Sarnat 1994:345).
A subset of these tests are tests whereby the APT is compared with the CAPM
with the aim of establishing whether it can explain anomalies, such as the size
effect and January effect, better than the CAPM (Reilly & Brown 1997:327).
Brevis ( 1998: 97) states that the first objective of empirical testing is to establish whether
share returns are affected by multiple risk factors and, secondly, to determine the
nature of these factors.
Following from this the results of the international research (studies on the US market
unless otherwise indicated) conducted up to the mid-1980s will be reviewed, whereafter
the South African studies will be examined.
4.3.8.1 Research on the number of APT factors
Roll and Ross (1980)

In the first published test of the APT, Roll and Ross used factor analysis and daily
share returns for the 1962 to 1972 period to estimate the number and importance of the
factors for 42 portfolios of 30 shares each (Harrington 1987:195; Seneque 1987:34).
Reilly & Brown (1997:325) note that they used the following two steps in their test:
• The expected returns and factor coefficients were estimated using time-series
analysis of individual share returns.
200
• Using these estimates, the cross-sectional pricing conclusions of the APT were
tested to establish whether the expected returns of the shares were consistent
with the factors derived in the first step.
Their initial findings showed that there are at least three important pricing factors, but
also that it is unlikely that there are more than four significant factors. They then
allowed the model to estimate the risk-free rate and found that the original finding of
three factors was probably an overestimate as only two factors were now found to be
consistently significant (Reilly & Brown 1997:325).
In a further test, they tested whether the APT is valid. To achieve this, they examined
whether the shares' standard deviations are significant in affecting expected returns.
This should not be the case because the unsystematic component of risk should be
eliminated through diversification, while the systematic risk should be explained by the
factor loadings. They initially found that a share's own standard deviation is significant,
but this disappears after adjusting for skewness (Reilly & Brown 1997:325).
The final test was an investigation of whether the same factors affect different portfolios
and they found no significant evidence that this is not so (Reilly & Brown 1997:326).
Roll and Ross concluded that, although these initial tests were admittedly weak, their
results provide support for the APT and that there are more than one factor which affect
returns (Seneque 1987:34; Reilly & Brown 1997:326).
Fogler (1982)
In an attempt to show that different shares are affected differently by the priced factors,
Fogler used principal component analysis on 100 shares for the 1941 to 1969 period.
Four common, but unidentified, factors were used and different factor sensitivities were
found. (Harrington 1987: 195).
201
Brown and Weinstein (1983)

Using a bilinear paradigm and the 1980 data of Roll and Ross, Brown and Weinstein
found results in conflict with the APT. After adjusting their test for a large sample size
they found results consistent with a three-factor APT model, but not for five-factor and
seven-factor models. They concluded that there are few rather than many APT factors
(Van Rhijn 1994: 165; Reilly & Brown 1997:329).
Chen (1983) and Reinganum (1981)

Their tests also found that at least three to four factors are important in explaining share
returns (Weston & Copeland 1992:426).
Hedge (1983)
In a review of some empirical tests, using both principal component analysis and factor
analysis, it was shown that the findings are generally in agreement with those of Roll
and Ross ( 1980). However, it was found that the second and third factors do not make
a significant contribution to variance and, although they are statistically significant, may
not be important in asset pricing (Hedge 1983:2, 117, 118).
Cho, Elton and Gruber (1984)

This analysis foliowed the same procedure to that of Roll and Ross ( 1980), but with
different data from a later period (Linley 1992:24; Elton & Gruber 1995:377). They exa-
mined different sets of data in an attempt to establish whether the number of significant
factors remain between three and five. They found that even when a two-factor model
is used to generate returns, it still requires two or three factors to explain returns. Cho
et al. concluded that their findings support the APT model as the model allow these
additional factors to be taken into account (Reilly & Brown 1997:326).
Dhrymes (1984) and Dhrymes, Friend and Gultekin (1984)

Both these studies criticized the findings of earlier studies, especially Roll and Ross's
1980 study, and even questioned the validity of empirical studies of the APT. Dhrymes
et al. confirmed the findings of Dhrymes that the number of significant factors are an
202
increasing function of the size of the sample analysed (Harrington 1987: 196, 197; Van
Rhijn 1994: 158; Elton & Gruber 1995:378).
Seneque (1987:35), Van Rhijn (1994:158,159) and Reilly and Brown (1997:326) note
that the criticisms are centred around two main issues:
• Firstly, Roll and Ross (1980) divided their sample into numerous portfolios of 30
shares each and, thus, the results differ from those of a large portfolio. They also
found no relationship between the factor loadings for groups of 30 shares and
those of 240 shares.
• Secondly, the number of factors change with changes in portfolio size and they
were not able to identify the actual number of factors involved in the return-ge-
nerating process. They found, for example, that portfolios of 15 shares have two
factors, portfolios of 45 shares have four factors and for portfolios of 90 shares
there are as much as nine factors. This also creates the problem of establishing
which of these factors are significant in explaining returns.
Cho (1984)
Cho disagreed with the Dhrymes ( 1984) and Dhrymes et al. ( 1984) studies. Using a dif-
ferent factor analysis technique, five to six factors were found and it was concluded that
the number of factors do not depend on the size of the sample (Harrington 1987: 197).
Roll and Ross (1984) and Ross (1984)

These articles were written in reply to the findings and criticisms of the Dhrymes, Friend
and Gultekin (1984) study. Roll and Ross disagreed with the findings of Dhrymes et al.
and took exception to the conclusion that the number of factors increases with an
increase in sample size. While they agreed that the number on non-priced factors will
increase with sample size, they stated that these are not important and can be diver-
sified away. They concluded that the issue of the number of factors is secondary and
it is more important how many factors are significant (priced) in a well-diversified port-
203
folio and how well the model explains the return-generating process compared to other
models (Harrington 1987:197; Linley 1992:25; Reilly & Brown 1997:326).
In his comments on the Dhrymes et al. ( 1984) tests, Ross stated that some tests can
be classified in three basic categories:
• Tests that are wrong.

• Tests that are misdirected and beside the point.
• Tests that are mystifying and, hence, cannot be classified into either of the first
two categories (Harrington 1987:200).
Dhrymes, Friend, Gultekin and Gultekin ( 1985)

This study repeated the previous tests of Dhrymes et al. ( 1984), using a larger group
of shares. They found evidence which confirmed the earlier results, in that as the
number of shares in each group increases so do the number of factors and number of
priced factors. However, most of the factors are not priced. They found, further, that
total standard deviation is at least as important in explaining returns as is the factor
loadings, that the number in the time-series affects the number of factors and, also, that
the size of the group affects the model's intercept (Reilly & Brown 1997:327).
Reilly and Brown (1997:327) conclude that these findings are evidence against the
validity of the APT, since it indicates dramatic instability in the risk/return relationship
and that the size of the group and number of observations affect the number of factors
identified. The fact that total standard deviation was found to be significant provides
further unfavourable evidence.
Most of the early research was concerned with establishing the number of significant
factors influencing share returns and from these flowed the studies which attempted to
identify these factors. However, by the mid-1980s there was no consensus about the
number of factors and much less about their identity. Much research was still needed
to establish the number and nature of the factors to ensure the APT has practical value.
204
4.3.8.2 Evidence on the testability/usefulness of the APT
Litzenberger and Ramaswamy (1979) and Sharpe (1982)

Both these studies, although not direct tests of the APT, showed that if factors such as
dividend-yield, firm-size and the beta of long-term bonds are ad_ded to the model, it
shows that a multi-factor model has better explanatory power than a single-factor model
(Brevis 1998:103).
Shanken (1982)
In a critique similar to those of Roll (1977, 1980) on the CAPM, Shanken questioned and
challenged the usefulness and empirical testability of the APT. Linley (1992:29) de-
scribes the basis of Shanken's criticism as being concern about its apparent lack of an
exact linear return relationship. Further, that the factor model can be manipulated to
produce any number of variables as factors and, hence, factor analysis is an inade-
quate method for the identification of the random variables that are important in asset
pricing.
A further source of criticism is added by Harrington (1987:200), who points out that
Shanken argued that equivalent sets of shares may be subject to different factor struc-
tures and the APT may, therefore, provide inconsistent implications. Shanken also con-
tended that if returns are not explained by the factor model, it is not considered evi-
dence against the model, but if it does explain returns then it is considered as support
for the model (Van Rhijn 1994:163; Reilly & Brown 1997:329).
Shanken concluded that as the APT provides no guidance on the nature and magnitude
of the factors, it suffers from a similar problem to the impossibility of identifying the true
market portfolio experienced with the CAPM (Linley 1992:29; Van Rhijn 1994: 164;
Reilly & Brown 1997:329). The enthusiasm about the APT as a testable alternative to
the CAPM may thus not be appropriate (Harrington 1987:200).
205
Dybvig and Ross (1985)

Dybvig and Ross countered Shanken's (1982) criticisms and concluded that the APT
is indeed testable and that the criticisms are not really relevant for actual empirical
testing (Harrington 1987:200; Linley 1992:29).
According to Van Rhijn (1994: 166), Dybvig and Ross's conclusions on the testability
of the APT can be summarised as the following:
• The APT assumptions ensure that portfolios with identical factor loadings are
close substitutes and the APT should be empirically testable as an equality.
• Both the CAPM and the APT are closely linked to separation theory, two-fund
separation and k-fund separation respectively.
• If returns are generated by a factor model and no particular asset is in large

aggregate supply, then the CAPM implies the APT, but the APT does not imply the
CAPM.
Shanken (1985)
Shanken disagreed with Dybvig and Ross's (1985) conclusions and argued that a set
of testable equilibrium APT pricing models have been developed, but that the original
APT is not testable as specified (Reilly & Brown 1997:329). He suggested further that
the equilibrium factor models as discussed by Dybvig and Ross (1985) are generali-
sations of the basic CAPM in a multi-beta format (Harrington 1987:200).
Dhrymes, Friend and Gultekin (1984) and Roll and Ross (1984)
Both these studies, the comments thereon and the resultant follow-up findings were
examined earlier in section 4.3.8.1. Laing (1988:38) considers these to be evidence on
the testability of the APT, which is to be considered in conjunction with the studies
already listed.
206
The situation regarding the testability and usefulness of the APT was unclear by the
mid-1980s. Although there were controversial arguments, it does appear as if there was
a growing body of research that considered the APT to be an usable alternative to the
CAPM, but its testability required further investigation.
4.3.8.3 Research on the nature of the APT factors
Sprinkel (1972)
In this study, performed outside the framework of the APT, Sprinkel showed that
changes in money supply and changes in market indices are leader indicators of the
business cycle. He also showed that changes in money supply can be used to predict
changes in market indices (Brevis 1998:121). Brevis (1998:121) adds that money
supply should be investigated as a possible APT factor.
Chen, Roll and Ross (1983,1986) and Roll and Ross (1983,1984)
In their 1983 study, only published in 1986 and reported on by Roll and Ross ( 1983),
Chen, Roll and Ross analysed data on the NYSE for a 27-year period. Although they
did not attempt to predict market returns, they correlated various macro-economic
variables with returns on five portfolios which mimicked the underlying factors, and
found the following five macro-economic variables to be significant in explaining share
returns:
• Growth, both anticipated and unanticipated, in industrial production.

• Unexpected changes in the default risk premium.
• Unexpected changes in the term structure of interest rates.
• Measures of unanticipated inflation and unexpected changes in inflation.
• A market factor, that is, changes in the stock market index (Harrington 1987:91,
202; Seneque 1987:34; Reilly & Brown 1997:472).
The market index is included as a proxy for factors that may have been omitted,
because if the market factor is found to be significant in pricing, then it means that
207
either the factors have not been correctly measured or that there are one or more fac-
tors missing (Harrington 1987:204).
In 1984 Roll and Ross acknowledged that there could be many other factors which can
affect returns, but those identified are the most significant since many of the others only
affect returns through their impact on the four main factors identified (Linley 1992:26).
Elton, Gruber and Rentzler (1983)

This study examined the impact of inflation on returns and, after deriving an equilibrium
model of real returns, the researchers employed the APT assuming inflation is not a
priced factor. They contended that it is important that APT models should be used with
factors which are not only statistically identifiable, but which have some real economic
meaning, such as inflation and growth in real GNP (Van Rhijn 1994: 163; Reilly & Brown
1997:328,329).
Gultekin (1983)
In a study outside the APT framework, Gultekin examined the relationship between
share returns and inflation for 26 countries. These countries excluded the US and UK,
but included South Africa. No reliable positive relationship could be found and it was
also found that the inflation/return relationship is unstable and differed from country to
country (Brevis 1998: 113).
Solnik (1983) and Geske and Roll (1983)

Solnik examined the relationship between share returns and expectations about in-
flation for nine countries during the 1971 to 1980 period and found results similar to
those of Gultekin in 1983. Similar results were obtained by Geske and Roll, who found
a negative relationship between share prices and expected inflation. Both these studies
were also not conducted within an APT framework (Brevis 1998: 113).
208
Various studies
Various other studies, performed outside the APT framework, also found that expected
inflation, unanticipated inflation and changes in expected inflation are negatively related
to share returns. Such studies include those of Lintner (1975), Jaffe and Mandelker
(1976), Nelson (1976) and Fama and Schwert (1977) (Brevis 1998:113).
The identification of the relevant priced factors is still in its inception. Considerable
controversy and argument were associated with the initial testing of the APT, and much
research and debate are still required to identify these factors, to understand the APT
and to determine its strengths and validity (Harrington 1987:204). However, it will not
be easy to determine the appropriate factors and those that have so far been identified
were included for reasons of common sense and convenience, and not derived from a
theoretical basis (Ross et al. 1990:311 ).
4.3.8.4 Research on the APT versus the CAPM
Brown and Warner (1980)

Brown and Warner compared the basic single-factor CAPM with various multi-factor
based CAPMs in an attempt to establish which model best identify the abnormal returns
included in the data. They concluded that there is no evidence that the more complex
models outperform the simple models (Laing 1988:vii).
Reinganum (1981)
Reinganum investigated whether the APT can explain the small firm effect, an anomaly
which the CAPM can not explain, whereby small firms (based on market value) con-
sistently earn higher risk-adjusted returns than large firms. The study examined ten
portfolios consisting of shares with similar risk characteristics and the results do not
support the APT. He found that small-firm portfolios earn statistically significant positive
excess returns, while portfolio for the largest firms earn statistically significant negative
excess returns. Reinganum also used different factor models, a three-factor, four-factor
and five-factor model, but still found that low-market value portfolios outperform high-
209
market value portfolios. Reinganum concluded that, since the results do not support the
APT, it cannot be considered a superior theory or empirical replacement for the CAPM.
He did, however, acknowledge that his analysis was a joint test of several hypotheses
implicit in the theory and, hence, it was impossible to determine the exact origin of the
error (Laing 1988:vi, 36; Van Rhijn 1994:159,161; Reilly & Brown 1997:327,328).
Berges-Lobera (1982)
This study entailed a joint examination of the size effect in both the US and Canadian
stock markets. The APT is not supported as it was found that there exists a significant
negative relationship between firm size and returns, in excess of that predicted by the
APT (Berges-Lobera 1982:127).
Brown and Weinstein (1983)

In their portfolio performance study, Brown and Weinstein found no improvement when
the APT is used as a replacement for the CAPM (Weston & Copeland 1992:426).
Chen (1983)
The study of Chen found evidence in support of the APT against the CAPM as well as
evidence contrary to the 1981 findings of Reinganum. He concluded that the APT is
able to capture missing price information, while the CAPM cannot fully explain returns
and is misspecified (Seneque 1987:35; Linley 1992:28; Reilly & Brown 1997:328).
Chen's analysis consisted of 180 shares and five factors and the cross-sectional
analysis showed the first factor to be highly correlated with beta. Chen found that the
APT is able to provide factors for the residual information related to returns, which the
CAPM is unable to capture. A further test was conducted in order to establish whether
there are some other variables with explanatory power not included in the APT factor
loadings. No such variables were found and Chen concluded that the size effect has
no explanatory power after adjustment for risk through the factor loading. Chen also
contended that the problems associated with sample size and multiple factors are
related to testing of the theory and not the theory itself (Reilly & Brown 1997:328).
210
Roll and Ross (1983) and Bower, Bower and Logue (1984)
Both these studies found that the APT provides better estimates for the cost of equity
of electric utilities than the CAPM (Weston & Copeland 1992:426).
Bower, Bower and Logue (1984)

Using monthly return data for 942 shares and four treasury security portfolios for a
period of nine years, Bower et al. performed various other tests on the APT and CAPM.
They divided the 942 shares into 30 portfolios, used four factors and found that the APT
explains more of the variations in returns between the portfolios than does the CAPM.
They also compared both models' predictive ability. Using ex post analysis of 127
shares over a 10-year period, they compared both models' estimated expected returns
with the actual returns of that period. The evidence strongly supports the APT, with the
standard deviations of the APT estimates being up to 50% lower than those of the
CAPM estimates (Linley 1992:31-33).
Chan, Chen and Hsieh (1985)

Chan et al. investigated whether the APT better explains expected returns and whether
the size effect disappears when the APT is used to measure expected returns on 20
size-based portfolios. The researches found that the APT is a much more appropriate
model than the CAPM, in that the difference in APT expected returns between the
largest and smallest size-based portfolios was found to be only 1,5% per annum, while
the CAPM rendered a difference of 11,5% per annum. It was also shown that the addi-
tional factor in their model, which explains the size effect's variation in returns, is the
difference in return between government bonds and high risk corporate bonds (Elton
& Gruber 1995:425).
Chen, Copeland and Mayers (1983,1987)

In their comparison between the CAPM and APT, Chen et al. found little difference in
the performance between these models, except that the APT better explains the impact
of dividing the sample into portfolios of similar size (Laing 1988:vii; Weston & Copeland
1992:426).
211
Though there is some conflicting evidence, it does seem as if the body of research
renders greater support for the APT, both from a comparison with the CAPM point of
view as well as from an explanatory ability point of view. However, further research is
required to get a more comprehensive set of findings from which stronger conclusions
could be drawn.
4.3.8.5 Research on APT equilibrium in international capital markets
Sharpe (1985:719) notes that tests on an International Arbitrage Pricing Theory (IAPT)
would require the identification of an international factor model, and as no such model
had been developed by the mid-1980s it is impossible to say whether the arbitrage
pricing relationships hold internationally.
4.3.8.6 South African APT research
Taylor (1977), Visser (1983) and Carter (1984)

These studies showed that there are more than one factor affecting returns on the JSE
and, hence, the APT will be able to better explain changes in portfolios returns than
does the CAPM (Brevis 1998:6, 101 ).
Gilbertson and Goldberg (1981)

In their study, which was not directly related to the APT, Gilbertson and Goldberg found
evidence of more than one factor on the JSE and this provided impetus to the APT
studies that followed (Van Ransburg 1998:17). They examined the postulate that the
JSE is a small market and thus heavily influenced by its two main sectors, the mining
and industrial sectors. They used the Actuaries All Mining Index and the Actuaries
Industrial Index in their two-factor model and their results led them to conclude that
there are at least two factors, a mining and an industrial factor, that affect share per-
formance on the JSE (Laing 1988:vi,37; Page 1989:78; Brevis 1998:101).
212
Myers (1981)
In another study outside the APT framework, Myers investigated the influence of the
gold price on all gold, mining financing, metals and minerals, financial and industrial
indices on the JSE. The findings from the study showed that the gold price has a long
and medium-term effect on all the sectors. In the short-term it has the strongest effect
on the gold and mining financing indices, a lesser effect on the metals and minerals
index and a delayed effect on the financial and industrial indices (Brevis 1998: 116).
Page (1986)
The first published South African study on the APT consisted of two main tests for the
JSE:
• Regression analysis to compare the explanatory powers of the APT and CAPM.
• Factor analysis to establish the number of priced factors affecting returns.
The main conclusions drawn from these tests were that:
• On an ex post basis the APT is substantially better than the CAPM in explaining
the variability of returns.
• Security returns are explained by at least two factors and there is a possibility of
more factors (Linley 1992:34; Van Rhijn 1994: 171; Brevis 1998: 101 ).
In addition, Page also compared his findings with those of Gilbertson and Goldberg
( 1981 ). He disagreed with their findings and concluded that it is not the mining and
industrial sectors that influence returns, but rather macro-economic variables that
influence these two sectors to a greater extent. He also added that a considerable
amount of research needed to be done to identify these factors (Linley 1992:34; Van
Rhijn 1994:171).
213
Linley (1992:34,35) notes a definite problem with testing both the APT and the CAPM
and comparing these two models in a South African context, namely the limited trade-
ability of the JSE. This may result in many shares not reflecting all the known informa-
tion relevant to particular shares and, hence, the effect of the various priced factors
may not be reflected in the less traded shares, that is, there may be a degree of market
inefficiency.
Despite this deficiency of the JSE, research has proved to be less controversial than
the international, mostly US, research. It does seem, in general, as if there are more
than one priced factor on the JSE and that the APT may have an advantage over the
CAPM in a South African context. Again, more research is required before more com-
prehensive conclusions can be drawn.
4.3.9 Conclusion
The APT is a relatively new alternative to the CAPM and it does have the advantage
of having less restrictive assumptions than the CAPM. By the mid-1980s it had been
the subject of a number of empirical studies, but these were hindered by a particular
APT difficulty - the theory does not specify the characteristics of the variables that
should affect returns. This has proved to be a major point of discussion.
Although there is some evidence that expected returns are affected by more than one
factor and that the APT may have greater explanatory power than the CAPM, these
results were only tentative. Conclusions on whether the APT provides an accurate de-
scription of reality can at this stage only be deemed as preliminary and inconclusive.
It is only through a review of additional and more recent research that a clearer and
more stable picture of the status of the APT will be obtained.
214
4.4 THE DEVELOPMENT OF OPTIONS THEORY AND THE BLACK-

SCHOLES OPTION PRICING MODEL
The CAPM and the APT deal with the expected risk and return on shares. There are,
however, many other kinds of securities in the market, other than shares. A type of
security that has been widely studied in recent years is options, which are the basic
ingredient in many of the hybrid securities ( Hendriksen & Van Breda 1992: 187).
Investors can purchase securities that present a claim (an option) on a particular share
or group of shares, rather than trade directly in the shares themselves. This option gives
the holder the right to receive or deliver shares under predetermined conditions.
Options are securities that derive most, if not all, their face value from the equity of a
company, but options are often not worth exercising and investors are then able to buy
or sell these equity-derivative securities (Jones 1998:526).
It is important to note that the emphasis in this study will be on share options. Other op-
tions such as interest rate options, foreign exchange options, etcetera fall outside the
scope of the study. The study will examine the importance of options, their nature, types
and values. Further, the emphasis will be on put and call options, how these options
work, how they are valued, their role in investment management, the efficiency of op-
tions markets and the role of the B-S Option Pricing Model in the pricing of options. The
final section entails a review of the empirical research on option markets and the B-S
Option Pricing Model.
4.4.1 The importance of options and the B-5 Option Pricing Model
Although options have been traded for centuries, they remained rather obscure and it
was only with the establishment of the Chicago Board Options Exchange (CBOE) in
1973, and the introduction of exchange-traded options that options became attractive
financial and investment instruments (Cox & Rubinstein 1982:3; Weston & Copeland
1992:436). The listing of options created a more orderly and active market for these
215
securities and option trading experienced gigantic growth - being among the fastest
growing markets for financial assets in the US. Subsequently, options markets were
also established in several major financial markets around the world and the growth in
option trading continued worldwide (Dobbins et al. 1994: 151; Elton & Gruber 1995:570).
Prior to the establishment of options markets, traditional (conventional) options could

only be bought and sold on an ad hoc basis through options dealers. There were a
number of reasons why traditional options remained obscure financial instruments:
• High transaction costs and lack of uniformity among option contracts, since the
date of expiration, the exercise price and the option had to be negotiated for
each contract.
• Difficulties of guaranteeing option contracts against default.
• Options could only be exercised or allowed to expire since no significant secon-
dary market for unexpired options existed (Dobbins et al. 1994: 151 ).
Traded options, which are option contracts with standardised terms, are bought and
sold through a central clearing system and are subject to relatively low transaction costs
(Dobbins et al. 1994: 151, 152). This led to the growth in option trading and as a conse-
quence stirred interest among academics and practitioners regarding the valuing of
option contracts (Elton & Gruber 1995:570). Black and Scholes (1973) started the deve-
lopment of option pricing models when they published their model - the 8-S Option
Pricing Model. This model has become the industry standard and is used as a bench-
mark in academic research, is programmed on hand-held calculators and relied on for
valuations by professional investors (Jarrow & Rudd 1983:ix).
Although largely falling outside the scope of this study, it is worth noting that the 8-S
model has been adapted and used in a number of other applications. Weston and
Copeland (1992:436) note that Merton (1974) used it to analyse how the value of cor-
porate debt is affected by risk; Galai and Masulis (1976) used it to examine how the
value of debt and equity are affected by mergers, acquisitions, expansions and spin-
216
offs; Ingersoll (1976) applied it to value the shares of dual-purpose funds; and Black
( 1976) used it to value commodity options, forward contracts and future contracts.
Hawinkels (1987:3,4) states that options are basically used in two ways by investors in
South Africa:
• Firstly, based on a fundamental or technical outlook, they construct investment

portfolios using options.
• Secondly, using an option pricing model, they value options in an attempt to
identify under- and overpriced options, that is, they apply the principle of arbi-
trage, ensuring that the options market remain liquid and efficient.
Hawinkels (1987:4,5) adds that the B-S model is the most widely used in the South
African market, but notes two limitations of the model:
• It relies on the accurate estimation of the price/yield volatility of the underlying

share.
• It assumes that price relatives are normally distributed.
According to Levy and Sarnat (1994:609) options can be purchased for speculation
rather than investing, but their role in corporate decision-making is increasing. Put
options have an important role in portfolio insurance and shares can be viewed as call
options and can be evaluated using the B-S model.
To summarise, share options trading and the development of the B-S Option Pricing
Model provide significant insights into other financial assets. It is therefore useful to
develop an understanding thereof and this study will examine the nature and charac-
teristics of options and options markets, the development of option pricing models - in
particular the B-S model, the application of options in investment decision-making and
empirical tests on the efficiency of options markets and the accuracy and validity of the
B-S model.
217
4.4.2 The nature, types and value of options
Two parties are involved in an options contract, the buyer (option holder) and a seller
(option writer) and the buyer has the right, but not the obligation, to exercise the option
and if the buyer exercises the option the seller must deliver his part of the contract, that
is, deliver the share or shares. If, however, the share price does not move as expected,
the buyer can let the option lapse as it has become worthless. Alternatively, if the share
price does move as expected, options create the opportunity for substantial gains with
very little outlay (Pike & Neale 1996:321 ).
These, and a number of additional aspects of options and options theory will be exa-
mined in the following sections.
4.4.2.1 The nature of options

To understand options, it is important to note that options are not issued by the compa-
nies whose shares are traded, thus options have no effect on these companies except
for their informational content (Correia et al. 1993:726).
Generally stated, share options can be seen as securities which grant its owner the
right, but not the obligation, to trade in a fixed number of shares at a predetermined
price at any time on or before a given date (Cox & Rubinstein 1982:3; Ross et al.
1990:561 ). Some options have an additional right in that the option need not only be
exercised on the date of expiration, but also on any date before then (Gemmill 1993: 1).
Hence, two kinds of options can be identified:
• American options. Options that may be exercised anytime up to and including

the expiration date.
• European options. Options that can only be exercised on the expiration date
(Jarrow & Rudd 1983:5; Van Horne 1992:105).
218
Gemmill (1993:6) notes that since an American option provides greater flexibility, its
value should be at least equal to, or more, than that of an European option on the same
underlying share.
Other important elements of options that require some further elucidation are:
• Exercising the option. The act of buying or selling the underlying share via the
option contract (Ross et al. 1990:561 ).
• The exercise (strike) price. The fixed per-share price at which the option holder
can buy or sell the underlying asset (Ross et al. 1990:562). Jones (1998:528)
notes that most shares in options markets have options available at different
exercise prices, thus providing investors with a choice, and as the price of the
underlying shares rise, so new options with new exercise prices are added.
Further, traded options are protected against share-dividends and share-splits,
that is, if either of these occur during the life of an option, both the number of
shares in the contract and the exercise price will be adjusted where necessary.
• The expiration (maturity) date. This is the last day on which the option can be
exercised and after this date the option is dead, that 1s, worthless (Ross et al.
1990:562; Jones 1998:528).
• The option premium. The initial price that the buyer of the option pays to the
writer of the option, both for options to buy or to sell shares (Jones 1998:528).
• LEAPs. Long-term options which can have maturities up to two years and
beyond (Jones 1998:46,528).
• Underlying shares. The shares which are involved in option contracts (Levy &
Sarnat 1994:609).
219
The two basic and most widely traded types of options are call options and put options.
Most other options can either be valued as combinations of call and put options or
according to the call or put option valuation methodology (Elton & Gruber 1995:570).
In the next section these two types of options are examined.
4.4.2.2 Types of options

As noted, basically two types of options are available in financial markets, depending
on whether the buyer wants to buy or sell the underlying share, namely call options and
put options (Jarrow & Rudd 1983:4).
• Call options. These are the most common types of options and provide the
holder with the right to buy the underlying share at a fixed price during a parti-
cular period. The value of a call depends on the value of the share on the
exercise date and if the value of the share exceeds the exercise price, then the
call is worth the difference. However, the value of the share may be less than the
exercise price and then the call is worthless and the holder will not exercise the
option (Ross et al. 1990:562,563).
• Put options. Put options are the opposite of call options. They provide the
holder with the right to sell the underlying share at a fixed price during a parti-
cular period. The value of puts are also opposite to those of calls. If the value of
the underlying share is less than the exercise price, the holder will buy the
underlying share and use the put to sell it. However, if the value of the underlying
share exceeds the exercise price, the put becomes worthless and the holder
would be foolish to exercise it (Ross et al. 1990:564).
Sharpe (1985:471) and Elton and Gruber (1995:570,572) note that both calls and puts
can either be American or European by nature, that is, can be exercised any time up to
expiration or only on the expiration date.
220
It is also useful to look at three other option terms. These are expressed from a call
option point of view, but are also applicable to its opposite, the put option.
• ln•the-money options. The current share price exceeds the exercise price of the
call and hence it will be worthwhile to exercise the option (Viscione & Roberts
1987:599; Pike & Neale 1996:321 ).
• Out-of the-money options. The exercise price of the call is above the current
share price and the value of the call equals zero - it cannot be negative (Viscione
& Roberts 1987:599; Pike & Neale 1996:321 ).
• At-the-money options. The exercise price of the option is exactly equal to the
current share price (Viscione & Roberts 1987:599).
To summarise, when an option expires in-the-money the holder gains the difference be-
tween the share price and the exercise price minus the original purchase price, but if
it finishes out-of-the money the option is worthless and the holder loses the original pur-
chase price (Weston & Copeland 1992:437; Levy & Sarnat 1994:620).
The following discussion on option value will focus on European call options with a fixed
exercise date. As puts are mirror images of calls, they are similarly valued. Although
American options have flexible exercise time-frames, the principles of European option
valuation also apply to American options (Correia et al. 1993:726).
4.4.2.3 Option values

The maximum value of a call is the price of the underlying share. No matter how low the
exercise price, if the price of the option is higher than the share price, investors would
ignore the option and buy the share directly. Against this, the minimum value of the call
is the difference between the share price and the exercise price, and the call has no
value if the exercise price exceeds the share price (Correia et al. 1993:726). These
boundaries of option value are illustrated in figure 4.3.
221
Figure 4.3 Boundaries of call option values
Value
of call
option
............................................................. A
..
.. .
~·
..,,····
•..··
../
........................................ B
..···/
....··
,.....
///
................
....~··
................................
Share price
where:
A Upper bound, that is, maximum value of a call option
B = Option price curve
C = Lower bound, that is, minimum value of a call option
D Exercise price of the call option
Source: Correia et al. (1993:727) and Jones (1998:545).
The call option value can vary in the range between the upper boundary and lower
boundary. Other things being equal, the call value moves towards the upper boundary
with an increase in the exercise price and time to expiration. At worst the call value can
drop to zero and it can never exceed the share price (Levy & Sarnat 1994:625).
222
a. Bounds of option values

Gemmill (1993:37,40-42) provides the following discussion on the bounds (limits) of
option values:
• The first bound is that an option, both put and call, can not have a negative value.
The right to being able to do something is either worth zero or a positive amount.
• The second bound is that the value of a call can not exceed the share price. The
right to buy the share can not be worth more than the share itself, since exercising
the right only gives the share and nothing more. Applied to put options, this bound
is that the value of the put can not exceed the exercise price, since that is what
would be paid on expiration.
• The third bound only applies to American options, which can be exercised imme-
diately. The minimum value of an American call is the share price less the exer-
cise price. If this was not so, the purchase of the call and the immediate exercise
thereof would provide a riskless profit. For American puts, this bound ensures that
the intrinsic value of the put equals the exercise price minus the share price.
• The fourth bound is that if the share price is zero, the value of the call is also zero.
An option on a worthless asset must itself be worthless.
• The fifth bound is that the call value approaches its lower bound when the share
price is much higher than the exercise price, since the chance that the call will be
exercised becomes 100% and all uncertainty value is removed.
b. Basic relationships of option values

Elton and Gruber (1995:577-579) describe how characteristics of options should affect
their values in rational markets, and describe a number of basic relationships that any
option valuation model should comply with:
223
• Relative values of call options with different characteristics. An American call

provides the added benefit of early exercise, hence, it cannot be worth less than
the European call. The first relationship is that an European call with the same
exercise price and expiration date as the American call can not be more valuable
than the American call. The second relationship is that if two calls have the same
expiration date, but differing exercise prices, the one with the higher exercise price
cannot be worth more than the one with the lower exercise price. This is logical,
as both holders will have the same share, but the latter will have some cash left
over.
• Minimum value of an European call option. The third relationship is that the
value of an European call is at least the greatest of zero and the difference be-
tween the share price and the present value of the exercise price.
• Early exercise of an American call option. According to the fourth relationship

it is not worthwhile to exercise an American call before expiration if the share does
not pay dividends or if the exercise price is protected against dividends. This
relationship does not hold if the call is unprotected and the share does pay
dividends. The logic behind this relationship is that the call is worth more alive
than dead, since, due to its uncertainty value, it would be better for the investor
to sell the option rather than exercise it.
• Put-call parity. The fifth relationship is put-call parity whereby a call and the
underlying share can be combined so as to render the same payoff pattern as the
put, or a put can be combined with the underlying share so as to provide the same
payoff pattern as the call.
Weston and Copeland (1992:440,441) note that there exists a fixed relationship,
derived by Stoll ( 1969), between European call and put options written on the same
shares, with the same exercise price and having the same maturity date. As a
consequence, no separate pricing models are required to value call and put options.
224
According to Elton and Gruber (1995:581) the put-call parity relationship, although it
may not hold perfectly for American options, has been shown to be a close approxi-
mation of the actual market relationship.
The put-call parity relationship is formally expressed as an equation by formula 4.4.
Formula 4. 4 Put-call parity
P =C - S + EP !( e
11
)
where:
P = Value of a put option
C = Value of a call option
S = Current market price of the underlying share
EP I ( ~) = Present value of exercise price
EP = Exercise price
e = The base of natural logarithms, approximately 2. 71828
r = The riskless interest rate continuously compounded
t The time remaining until expiration, expressed as fraction of a year
Source: Galai (1982:69,70) and Jones (1998:549).

225
c. Determinants of option values

There are seven basic factors which have an influence on the value of options:
• The underlying share price. The higher the share price for a call and the lower
for a put, the greater the value of the option since there exists large anticipated
returns at expiration (Jarrow & Rudd 1983:16).
• The exercise price. The greater the exercise price of a put relative to the share
price, the more valuable the put. Similarly, the lower the exercise price of a call
relative to the share price, the more valuable the call (Jarrow & Rudd 1983:17).
• The time to expiration. The longer the time to expiration the more the present
value of the exercise price is decreased. Since a call holder has to pay this
amount the value of the call increases, but as the holder of the put has to receive
this amount the value of the put decreases. However, the uncertainty associated
with the longer time to expiration increases the possibility of favourable outcomes
for both puts and calls. In conclusion, call values increase with increases in time
to expiration, but with puts the answer is less clear since two opposing factors
influence their values (Jarrow & Rudd 1983: 17).
• The risk-free interest rate. An increase in the interest rate decreases the present
value of the exercise price, hence, call values will increase and put values will
decrease when the interest rate rises (Jarrow & Rudd 1983: 17, 18).
• The volatility of the underlying share. The more variable the underlying share
price the more valuable the call since there is a bigger chance that the exercise
price will be lower than the share price. Volatility also increases the value of a put.
A put that is deep-in-the-money is more valuable than one only slightly-in-the-
money, but a put that is deep-out-of-the-money has the same value (zero) as one
only slightly-out-of the-money. In both cases, volatility decreases the downside
risk relative to upside risk and thus increases value (Ross et al. 1990:573, 576).
226
• Dividends. The payment of dividends on unprotected European options will de-

crease call values and increase those of puts. The payment of dividends generally
causes share prices to drop significantly as they go ex-dividend, increasing the
possibility of calls expiring out-of-the-money and puts expiring in-the-money
(Jarrow & Rudd 1983:17).
• Strong market trends. Strong market trends, either being bull markets (moving
upwards) or bear markets (moving downwards) generally have an effect on option
values since the scope for making profits is increased, depending on the type of
option (Pike & Neale 1996:328).
Van Horne ( 1992: 124) and Weston and Copeland ( 1992:439) note that a perfectly risk-
less (hedged) position could be achieved using options. This could be obtained by buy-
ing the share, buying a put option on the share and selling a call option on the share.
Van Horne ( 1992: 124 ), further, notes that if the markets were efficient, the return on a
perfectly hedged position would be the risk-free rate and this principle can be used to
determine the value of the option at the beginning of the period. In the next section
options markets and their efficiency are examined.
4.4.3 The options market
As can be seen from the preceding section, options, by their nature, can be valuable
securities. This led to options markets being formalised whereby options, especially
share options, can be traded in the same way as shares are traded on stock exchanges
around the world.
4.4.3.1 Background to the development of options markets

Options are not new and options contracts on commodities, such as wheat, have been
used since the middle ages. These initial option contracts were sometimes simple fixed-
price forward contracts, but at other times merchants could also have the option to back
out. These commodity options were not without their problems as in 1636 the Amster-
227
dam options market for tulip bulbs collapsed when the sellers of options refused to buy
the bulbs in terms of their option contracts - the price of tulips had fallen drastically and
they were not prepared to still pay the high prices specified in the contracts (Gemmill
1993:6,7; Pike & Neale 1996:319).
Before the establishment of organised options exchanges, traditional put and call
options (mostly of an European nature) could be bought and sold in the over-the-
counter (OTC) market. Option brokers or dealers brought the buyers and sellers
together and the terms of each contract had to be negotiated. This was a cumbersome
process with high transaction costs and, since the option contracts were not stan-
dardised, no effective secondary market for trading in options existed (Jarrow & Rudd
1983:6; Hodges 1988:3; Jones 1998:528).
These problems were overcome with the establishment of the first traded options
market, the CBOE, in 1973. The CBOE was an immediate success and options trading
had a rapid growth. With this growth other exchanges also started trading in options
(Hodges 1990:3). These included the American Stock Exchange, the Philadelphia
Stock Exchange, the Pacific Stock Exchange and the New York Stock Exchange in the
US. In Europe, both the European Options Exchange (EOE) in Amsterdam and the
London Traded Options Market (LTOM) were established in 1978. LTOM later became
part of the London International Futures and Options Exchange (LIFFE). In the mid to
late 1980s options exchanges were established in many European countries, including
France (MONEP), Germany (OTB), Sweden (OM) and Switzerland (SOFFEX) and in
the far-east time zone options were traded by exchanges in Osaka, Tokyo, Singapore
and Sydney (Gemmill 1993:7; Pike & Neale 1996:321 ).
Hodges ( 1990:4) notes that by 1990 options were traded in at least 14 countries world-
wide and that trading volume continued to grow. Gemmill ( 1993:6) adds that most of the
options traded on exchanges are American style.
228
In South Africa the development of the options market started in the early 1980s, an
OTC market was established and the options were switched from European to
American style. However, the OTC market was thinly traded and inefficient, which led
to the establishment of a traded options market in 1987. This options market on the
JSE has also shown rapid growth and options have become established financial
instruments in the South African market (Mynhardt 1996:3,4).
Sharpe (1985:475) notes that not all the options written are on shares. In recent years
many new and exotic options have been listed and new exchanges have opened that
specialize in such instruments.
4.4.3.2 General characteristics of exchange-traded options

Although various variations of the basic exchange-traded options have been
developed, including various new and exotic options, these largely fall outside the
scope of this study. The discussion that follows only provides the general basic
characteristics of exchange-traded options. This serves as elucidation of the aspects
discussed with the efficiency of options markets and the later review of empirical
research on options and options markets.
Jarrow and Rudd (1983:6) describe four basic characteristics of standardised

exchange-traded option contracts:
• Fixing maturity dates so that options expire at set dates.
• Specifying the exercise prices of option contracts.
• Specifying the number of shares per option contract.
• Standardising the procedures to deal with dividends, share-splits, etcetera.
a. Expiration dates
Options generally have a maturity of six to nine months with different expiration dates
approximately every three months. These expiration dates belong to an expiration
cycle, for example January, April, July and so forth (Jarrow & Rudd 1983:6, 7; Gemmill
1993:7).
229
b. Exercise prices
Exercise prices are rounded-off in bands and as the price of the underlying share
moves, new options are generally introduced, but the earlier options remain valid until
expiration. For example, exercise prices can be R39, R40 and R41 if the current share
price is R40 (Jarrow & Rudd 1983:7; Gemmill 1993:7).
c. Number of shares
The initial number of shares in the contract is standardised so that each option repre-
sents 100 shares of the underlying asset. The number of shares and the exercise price
can, however, change after share-splits or dividend payments. The options exchange
controls this adjustment so that neither party involved in the contract obtains an unfair
advantage (Jarrow & Rudd 1983:7).
d. Dividends and share-splits

Listed options are generally not adjusted for cash dividends, but are adjusted for share-
dividends and share-splits. The listed options are adjusted on the beginning of the
same day as the share's market price is adjusted for the share-dividend or share-split.
Since cash dividends do not form part of the general characteristics of traded options,
they can have a significant impact on option values (Jarrow & Rudd 1983:7,8).
4.4.3.3 The efficiency and tests of efficiency of options markets

The discussion in this section is based on Galai's (1982:47-50) review of empirical
tests performed on option pricing models.
Since the values of options are generated by the values of the underlying shares, mea-
surement of the efficiency of options markets relies on the synchronous price of the
shares. It is important to note that synchronous markets, in this sense, are the stock
market and the options market where trading in related assets take place simulta-
neously and the listed prices reflect this trading simultaneously. Therefore, it is neces-
sary that these markets are synchronized and that both are efficient to obtain good pre-
dictions of option prices.
230
A major problem with options market efficiency is that any tests are joint tests of the va-
lidity of the model, market efficiency, market synchronization and data accuracy. Due
to this problem, it can not be concluded that an option pricing model is invalid if it fails
to accurately predict future prices. It may be due to nonsynchronous or inefficient mar-
kets, rather than an incorrect model. If, however, a model consistently generates ab-
normal risk-adjusted profits, which the others can not, this can be construed as evi-
dence in support of the model.
If it is assumed that the data is accurate and the input parameters of the model is
unbiased, then the following statements can be made about the efficiency and synchro-
nousness of options markets:
• Given the share price, one would expect the model to predict future option prices
accurately, ifthe markets are efficient and synchronous and the model is accurate.
• If the markets are efficient and synchronous, no model should be able to offer ab-
normal profits.
• If the markets are nonsynchronous and the model is correct, it should be possible
to generate abnormal profits using the model.
Galai (1982:49,50) adds some other potential problems that may be experienced with
tests of market efficiency:
• Trading practices may complicate the conclusions drawn from the empirical
studies. With thinly traded or low-priced securities. the discreteness of trading with
minimum price changes may affect results.
• Another problem is that models are unable to handle the valuation of American
options where uncertain future dividends may be paid on the underlying share.
• The existence of transaction costs and taxes may cause distortions in the results
obtained from the model.
231
It is clear from the above that option pricing models are an integral part of testing
options market efficiency and, from the preceding section, in determining the value of
options. In the next section the predominant and most widely used model, the B-S
Option Pricing Model, is examined.
4.4.4 The B-5 Option Pricing Model
Given the insights into the characteristics and determinants of call option values, Black
and Scholes (1973) developed a formal model for the equilibrium pricing of European
call options paying no cash dividends. Due to the existence of a put-call parity rela-
tionship, this model is equally efficient in valuing put options of a similar nature
(Hendriksen & Van Breda 1992: 188; Jones 1998:546).
The B-S model, although having a mathematical and seemingly complex formula, is
widely accepted and used in the investment community and is widely available on
computers and calculators (Jones 1998:546).
Hendriksen and Van Breda (1992:188) note that the B-S model has a number of
assumptions which are similar to those of the CAPM and these assumptions are
examined in the next section.
4.4.4.1 Assumptions of the B-5 Option Pricing Model

The assumptions listed below are the sufficient conditions needed for the B-S model
to be correct. Though, even when these conditions do not hold in reality, variations of
the model often works. The B-S equation can be fine-tuned to account for dividends
and empirical evidence suggests that the model and its variations do value call options
accurately (Ross et al. 1990:582). The assumptions on which the model is based are:
• Only European options are considered, that is, that options are only exercisable
at expiration.
• There are no transaction costs and no taxes.
232
• There are no imperfections when options are written, there are no restrictions on
short-selling and short sellers receive the full proceeds from their transactions.
• The short-term interest rate is known and constant throughout the life of the option
and investors can both borrow and lend at this rate.
• The underlying share pays no dividends.
• The market operates continuously and the variance of return is constant and
known to market participants, that is, share trading is continuous and there are no
jumps in the movement of share prices.
• The probability distribution of share prices is lognormally distributed (Ross et al.
1990:582; Van Horne 1992: 114; Levy & Sarnat 1994:635).
Simister (1988: 11, 12) notes that most of the weaknesses of the 8-S model are caused
by market imperfections and these weaknesses are:
• Transaction costs are not zero.

• Price movements are not continuous.
• The distribution of prices is neither normal nor lognormal.
• The markets are not infinitely deep.
As noted, despite these weaknesses, the 8-S model provides accurate option values
and the next section examines the model's parameters and pricing equation.
4.4.4.2 Parameters of the 8-S Option Pricing Model

The 8-S model provides an exact formula for determining the value of an option and
the model is based on the idea that investors can maintain a reasonably hedged
portfolio over time. Arbitrage then causes returns to be equal to the risk-free rate and,
hence, the price of the option will have an exact relationship with the price of the
underlying share (Van Horne 1992: 124, 125). The equation of the 8-S model is shown
by formula 4.5.
233
Formula 4. j The Black-Scholes option pricing model
where:
C = Current value of the option
S = Current price of the underlying share
EP Exercise price of the call option
e = The base of natural logarithms, approximately 2. 71828
r = The riskless interest rate continuously compounded
t = The time remaining until expiration, expressed as a fraction of a year
N ( d) = The value of the cumulative normal density function
In (SIE P) + ( r + 0, 5 a 2 ) t
a.Ji
d = In(S!EP)+(r-0,5a 2 )t
2
a.Ji
In = The natural logarithm
cr = The standard deviation of the continuously compounded annual rate of return

on the underlying share
Source: Van Home (1992: 116), Elton and Gruber (1995:588) and Jones (1998:547).
234
Ross et al. (1990:582), Upsher (1993:13) and Jones (1998:546) identify the five va-
riables used by the formula as being:
• The current price of the underlying share.

• The exercise price of the option.
• The time remaining until expiration of the option.
• The interest rate.
• The volatility of the underlying share.
Ross et al. (1990:582) add that the 8-S formula is one of the most important contri-
butions in finance, it allows anyone to calculate the value of an option and has the ad-
vantage that four of the parameters are directly observable. It is only the volatility of the
underlying share that needs to be estimated.
Before reviewing the estimation of the share's volatility, it should be noted that the
attractiveness of the 8-S model is increased by the parameters which are not required.
These are:
• The investor's risk attitude does not affect the value of the option. Anyone can use
the formula, whether they are risk averse or not.
• The expected return on the share does not enter the formula. Investors may have
different views on the share's expected returns, but this will not affect the value
of the option. Although the value of the option depends on the underlying share's
price, that price already reflects the divergent views of all investors (Ross et al.
1990:582).
a. Estimating the underlying share's volatility

Volatility, as used in the model, is defined as the standard deviation of the continuously
compounded annual return on the underlying share and is the only parameter that is
not directly observable (Mulder 1998:84). Volatility is an important variable, since the
235
greater the volatility the greater the chance of the share price going up and, thus, the
higher the value of the call option. Hence, there exists a positive relationship between
option values and the volatility of share returns (Jones 1998:547).
Jones (1998:547,548) and Mulder (1998:84,85) provide the following descriptions of

the methods of estimating volatility:
Historical volatility estimates. Historical data on share returns are used to estimate
the standard deviation, and this estimate assumes that past volatility will continue to
hold in the future. It has, however, been shown that the variability of share returns
changes over time and different users will thus have different estimates, resulting in
differing option values. Empirical studies have also shown that the variance obtained
from sources other than historical data is more valuable than historical estimates.
Implied volatility estimates. The implied volatility is an estimate that makes the value
obtained from the 8-S model equal to the current market price of the option. As the
current market price of the option is observable, this figure can be inserted into the
equation to derive the implied volatility. Thus, the market's opinion of the share's
volatility is used as an input to the model.
Gemmill (1993:93) notes that if everybody only uses implied volatility, no forecasting
would be done and the consensus forecast derived from the implied volatility becomes
useless. The following approach for estimating volatility is recommended:
• Estimate the historic volatility.

• Adjust the forecast for abnormal events.
• Adjust more distant volatilities towards the long-term mean.
• Adjust the volatilities towards the consensus forecast provided by the implied vo-
latilities (Gemmill 1993:93).
236
b. Adjusting the Black-Scholes model for dividends

A weakness of the basic 8-S model discussed so far, is that it assumes that the under-
lying share pays no dividends. However, traded options are not protected against cash
dividends and these can have a significant effect on option values. When a cash
dividend is paid, the share price declines and any reduction in share price reduces the
value of a call and increases the value of a put (Jones 1998:546).
Sharpe (1985:505,506) provides a description of the procedures which can be applied

to deal with dividends. These procedures are based on the assumption that dividends
can be predicted with certainty and that the option will not be exercised before expi-
ration.
• Assume that all relevant dividends have already been announced and the share
price has gone ex-dividend for all these dividends. The share price is then re-
duced by the present value of these expected dividends.
• To consider the position where the share price has not yet gone ex-div at the
expiration date, the following procedure can be followed. Assume that the option
is held until just before the ex-dividend date and subtract the present value of all
the dividends, except the last payment, from the current share price.
The B-S formula can be applied using these dividend-adjusted prices. The current
value of the option is then assumed to be equal to the highest of these two values.
Sharpe (1985:506) adds that although these procedures are not exact they are pro-
bably sufficient for most listed options.
237
4.4.4.3 Information required for estimation of the parameters of the B-S Option
Pricing Model
Jarrow and Rudd ( 1983: 132-138) provide a comprehensive discussion of the informa-
tion requirements and information sources for each of the parameters and for any divi-
dend adjustments.
a. The share price

The B-S formula provides a link between two simultaneous prices, and the best sources
of data on share prices are:
• Those services that provide the ability to download closing share prices onto per-
sonal computers.
• Real-time systems that provide the latest information on share trading directly
from stock exchanges (Jarrow & Rudd 1983:133).
b. The exercise price

This information is easy to access, since it forms part of the specifics of the option
contract. However, the effect of any share-splits and share-dividends must also be
taken into account (Jarrow & Rudd 1983:133).
c. The interest rate

The interest rate is assumed to be constant over the life of the option and equal to the
risk-free rate of return. The rate is usually obtained from financial instruments, such as
Treasury bills, which have a maturity date close to the option's expiration date. It should
be noted that the B-S formula is not particularly sensitive to changes in the interest rate
(Jarrow & Rudd 1983: 133, 134).
d. Time to maturity
Again, the expiration date is part of the specifics of the option contract and the time to
maturity is simply determined by counting the number of days between the current date
and the expiration date. A rule of thumb also exists whereby each non-trading day is
238
considered to be equal to a third of a trading day. This is done as a measure of the

varying risk levels which are likely to be encountered before expiration (Jarrow & Rudd
1983:134).
e. Volatility of the underlying share

The estimation of volatility, using either historical volatility or implied volatility, has
already been examined in section 4.4.4.2.a.
f. Adjusting for dividends

Most companies tend to maintain stable dividend growth rates and are relatively con-
sistent in their dividend payment dates, so it is generally not very difficult to make
forecasts of the cash dividends and ex-dividend dates over the life of the option (Jarrow
& Rudd 1983: 138).
It is now appropriate to consider how the principles of options theory can be practically
applied to investment decision-making and corporate finance.
4.4.5 The application of options theory in investment decision-making and

corporate finance
Jones (1998:527) identifies four major reasons why, rather than investing in the shares
themselves, share options are useful investment alternatives.
• The availability of options has expanded the investment opportunity set. Investors,
therefore, have access to risk/return combinations which would otherwise not have
been available and this can improve the risk/return characteristics of their
investment portfolios.
• Investing in call options allows investors to control, for a short period, a claim on
the underlying share at a much lower cost than investing in the share itself.
• With investment in put options, investors are able to duplicate short sales without
needing a margin account and at a much lower cost than the price of the shares.
239
• Options provide investors with the benefit of knowing the maximum amount of their
loss in advance. The maximum that can be lost when an option expires worthless
is the cost price of the option itself.
• Options have a significant leverage effect, whereby percentage gains relative to
cost price is magnified in comparison with investment in the shares.
• With investment in options on stock market indices, investors are able to parti-
cipate in market movements with only a single trading decision.
Hawinkels (1987:5) also identifies a number of investment opportunities which are

available through the use of options:
• Options provide a significant expansion in the range of portfolio returns that are
available to investors.
• Options are useful instruments in hedging against risk.
• Options provide opportunities for speculation, that is, using options for their
leverage effect, since the downside risk is the cost of the option while the upside
potential is nearly unlimited.
• Arbitrage opportunities are available if options are mispriced.
• Capital market transactions can be transformed into money market transactions
in order to take advantage of the slope of the yield curve.
Pike and Neale (1996:331) identify other applications of options theory in corporate
finance and investment decision-making. They state that the implications of options
theory go far beyond just trading in, and valuing of options. Options theory also provide
powerful knowledge for the understanding of various other contractual agreements in
finance, including:
• Share warrants which give holders the right to buy shares directly from the com-
pany at a fixed price over a specified period of time.
240
• Convertible debentures which give holders a combination of an option and a loan.

When the option is exercised the holder exchanges the loan for a fixed number
of shares in the company.
• Debentures to which a call option can be attached, thus providing the company
with the right to repurchase the debentures before maturity.
• Share option schemes which provide company executives with an incentive to
pursue the goals of the shareholders.
• Insurance and loan guarantees are both forms of put options. An insurance claim
is a put option, while a government loan guarantee is a type of insurance. The
underwriting of a share issue is a similar type of option.
• Foreign exchange and interest rate options are also ways of hedging against risk,
or of speculating on movements in these markets.
Before reviewing the research done on the efficiency of options markets and the
reliability of the B-S Option Pricing Model, it is first necessary to examine the various
approaches to the testing of these aspects.
4.4.6 Tests on options market efficiency and the B-S Option Pricing Model
4.4.6.1 Testing options market efficiency

The testing for the efficiency of options markets can be divided into two sections:
• Firstly, testing the actual efficiency of options markets.

• Secondly, testing the impact of the existence of options markets on the prices of
the underlying shares, in order to establish whether it is benign, harmful or irrele-
vant (Gemmill 1993:256).
Both these aspects are important (Gemmill 1993:256). If options markets were found
to be inefficient, then it could either be because the market participants are ill-informed
or because competition is not free and fair. More importantly, if options markets were
241
found to have a destabilising effect on the underlying shares, and thus affect the effi-
ciency of stock markets, then constraining their use should be considered.
a. Direct tests of options market efficiency

According to Gemmill (1993:256), options markets will be efficient if they allocate re-
sources correctly, that is, in a manner that ensures that all marginal costs and returns
are equated. In order to achieve this, option prices must reflect all available information
and transaction costs must be minimised. Thus, we can refer to two elements of effi-
ciency, namely pricing efficiency and transaction efficiency.
Pricing efficiency
Pricing efficiency in options markets are the same as those discussed earlier with the
EMH. The various forms of efficiency will, thus, not be examined again.
The existence of profitable opportunities which are not exploited is evidence against
pricing efficiency. It is, however, difficult to assess whether the unexploited profits are
sufficient to cover transaction costs and risk and, hence, it is difficult to test for pricing
efficiency (Gemmill 1993:257).
Transaction efficiency
These tests are concerned with the size of bid-ask spreads and commissions. Market
makers earn at least part of the bid-ask spread as their remuneration, while brokers
earn commission. Thus, the question of transaction efficiency is whether these interme-
diaries are able to earn excess profits relative to the risk they bear (Gemmill 1993:257).
Tests of these two aspects of options market efficiency include the following:
i. Arbitrage tests
These are the simplest kind of pricing efficiency tests. The tests for simple arbitrage
opportunities are relatively easy as they do not require the use of option pricing models
for its calculations, nor does risk need to be considered since arbitrage is riskless. The
242
main arbitrage opportunities relate to the minimum bounds on call and put prices as
well as put-call parity. The tests usually concern the observation of a series of
transaction prices and the application of lower-bound tests and parity tests on the
series (Gemmill 1993:257).
ii. Model-based efficiency tests

Option pricing models are used to test for more complicated profit opportunities. How-
ever, model-based efficiency tests do have the problem that they are joint tests of both
the model and of market efficiency. In principle, models are used in order to establish
whether options are mispriced, that is, whether the option can be bought and sold and
hedged with the underlying share or another option to achieve a riskless profit. These
tests are concerned with whether these procedures are able to deliver profits (Gemmill
1993:258).
Gemmill (1993:258) notes, however, that if these tests deliver a profit of zero, this does
not necessarily mean market efficiency. Due to the joint nature of these tests, it could
either indicate market inefficiency or model inaccuracy. Positive profits only indicate
market inefficiency if:
• The model provides the correct riskless hedge ratio.
• All transaction costs are taken into account.
• All option values used are a true reflection of market prices.
iii. Implied volatility tests

Rather than concentrate on profit opportunities, these tests concentrate on whether in-
formation is fully reflected in option prices. The aim of these tests is to establish
whether implied volatilities provide good predictions of future volatilities (Gemmill
1993:259).
b. Tests on the impact of options on the underlying shares

Gemmill (1993:259,260) considers this to be the important issue and states that if it
was found that options destabilised the underlying shares, then banning options might
243
be an appropriate action. Gemmill (1993:260) distinguishes between two potential

impacts:
• The general stabilising or destabilising effect of options.
• The destabilising effect on the expiration of options.
Tests on the general impact of options

The potential impact of options depends on the particular assumptions made. In a 8-S
model world no market imperfections exists, so any option can be replicated almost
perfectly. However, the continued existence of options must indicate that market imper-
fections are rewarded and may reduce the cost of speculation. Increased speculation,
both well-informed and ill-informed, may cause an increase in volatility and, hence,
destabilisation. However, these theoretical arguments do not solve the problem and it
is only through empirical testing that the impact of options can be assessed (Gemmill
1993:260).
Tests on the impact of options on expiration

When options expire, all arbitrage positions are unwound and, according to Gemmill
(1993:262), the underlying stock market may suffer a bout of 'indigestion' since, with
the resultant trading in a large amount of shares, share prices may rise or fall signifi-
cantly.
4.4.6.2 Testing the validity of the 8-S Option Pricing Model

These tests are not only relevant to the basic 8-S model, but have been used to test
extensions thereof and for the testing of various other models. Many of these tests have
also been comparative tests, whereby the performance and accuracy of various models
have been compared.
Galai (1982:52,53) identifies four different approaches which have been used to test
the 8-S model's validity.
244
a. Simulations
The first approach uses simulations and quasi-simulations of deviations from the
model's assumptions to test the model's sensitivity to these empirical deviations. These
tests are not direct tests of the model's validity, but are used to examine the robustness
of the model to varying conditions (Galai 1982:52).
b. Direct comparison of model values with actual prices

The second approach uses the actual prices of the underlying shares and the esti-
mated parameters of the model to generate expected option prices. These expected
prices are then compared with actual market prices. These tests attempt to establish
whether the model provides unbiased estimates of actual prices or whether there are
consistent deviations. These deviations can either be due to incorrect estimation or
identification of profit opportunities (Galai 1982:52).
c. Hedging
In the third approach neutral hedge positions are created and the behaviour of the
investment returns are examined. If the model is correct, these returns should be equal
to the riskless rate. These tests have the advantage that the investment returns need
not be adjusted for risk (Galai 1982:52).
d. Implied volatility tests

This fourth approach uses standard deviations derived from actual option prices. As all
the parameters of the model, except volatility, are observable, the standard deviation
can be imputed by equating the actual option price to the model's value. These test
then examine the behaviour of the implied standard deviation (volatility) to establish the
validity of the model (Galai 1982:52,53).
According to Galai (1982:53) the last two approaches have been used in the majority
of recent research. Now that the tests of market efficiency and model validity have been
established, the next section reviews the results obtained from empirical tests. From
245
these tests it will be possible to make conclusions regarding the validity of the B-S
Option Pricing Model and on the efficiency of options markets.
4.4.7 Research related to options markets and the 8-S Option Pricing Model
This section reviews research studies done on options market efficiency and those
done on the B-S Option Pricing Model. Testing of these two areas are interrelated as
most tests are generally joint tests, that is, the model is used to establish the level of
market efficiency and the validity of the model is examined assuming market efficiency.
The efficiency of options markets are generally concerned with whether options
markets accurately price options and whether any price adjustments occur fast enough
to prevent investors and members of the exchange from having the opportunity to earn
abnormal profits. Another area of concern is whether options markets, which are
directly connected to the underlying stock markets, have any destabilising effect on
these stock markets.
As mentioned before, any tests of the B-S model are directly related to tests of options
market efficiency. Without the options markets being efficient, it is almost impossible
to make reliable conclusions about the validity, accuracy and performance of the
model.
There are basically four different approaches to testing the B-S model's validity. In the
first case the model's assumptions are examined to establish its robustness. Secondly,
using the model to determine option values and comparing the values calculated with
market prices to establish the model's accuracy and predictive ability. Thirdly, the
model's predictions are examined to establish whether it provides an accurate descrip-
tion of the market pricing mechanism and, finally, its parameters are tested against the
market to establish its validity.
246
This section starts with an examination of the efficiency of options markets, whereafter
the B-S model's tests will be reviewed. The initial studies reviewed will all be from inter-
national research (on the US market unless otherwise indicated), whereafter studies
in a South African context will be examined.
4.4.6.1 Research related to market efficiency
a. Simple arbitrage tests
i. Minimum bound tests
Galai (1978)
Galai tested the efficiency and synchronization of the CBOE, using daily closing prices
in the first six months of its operations, April 1973 to October 1973. A significant
number of call option lower-bound arbitrages were found (Galai 1982:51; Gemmill
1993:257). Galai (1982 51) summarises the major conclusions from his 1978 study as
being:
• The greatest number of violations were found in cases where options approach
expiration and the frequency is an increasing function of the share price, given the
exercise price.
• The hypothesis of sufficient synchronization between stock exchanges and op-
tions markets is rejected.
• Profitable arbitrage opportunities can be exploited, but these profits will be small
relative to the dispersion of yields.
Bhattacharya (1983)
A similar test by Bhattacharya, using the Berkeley options database, also found that
violations occur, but these are not profitable since transaction costs cannot be covered
(Gemmill 1993:257).
247
Halpern and Turnbull (1985)

In their study on the Canadian market, the researchers also found that violations occur,
but their study did not take full account of transaction costs (Gemmill 1993:257).
ii. Put-call parity studies
Stoll (1969)
Stoll examined put-call parity for European options on OTC markets in the US, using
weekly submissions by the Put and Call Brokers and Dealers Association for ten com-
panies in 1967 (Galai 1982:70,71; Gemmill 1993:257). Stoll used regression analysis
and found that there were deviations from expectations, in that the intercept was higher
and the slope lower than expected (Galai 1982:71 ).
Merton (1973)
Merton showed that the put-call parity relationship does not hold for American put
options. As it may be profitable to exercise the American put before expiration, the
American put will have a higher value than an European put with similar terms (Galai
1982:70).
Gould and Galai (1974)

Their study involved the inclusion of tax and transaction costs in the put-call parity rela-
tionship for American options. It was found that transaction costs increase the upper
bound and decrease the lower bound of the difference between calls and puts. Taxes,
however, have no significant effect.
Gould and Galai re-examined Stall's 1969 results and included data for 1968 and 1969.
They suggested that Stall's use of regression analysis was incorrect and found that
transaction costs to a large degree explain the observed premiums on puts and calls.
However, some degree of market inefficiency still remains (Galai 1982:71 ).
248
Klemkosky and Resnick (1979, 1980)

The 1979 study of Klemkosky and Resnick examined the transactions data of one day
per month for the period July 1977 to June 1978. They used 15 shares with their
options, and were able to construct a nearly simultaneous position for the underlying
share, the put and the call. They concluded that the simultaneous prices observed for
the puts and calls are consistent with put-call parity (Galai 1982:72).
In 1980 they re-examined their 1979 study and concluded that the price corrections on
options exchanges are fast enough to eliminate most profit opportunities for members
of the exchange (Galai 1982:72).
Various studies
Brenner and Galai (1984) also concluded that the put-call parity relationship is valid
and no significant profits exist after transaction costs. Similar results were obtained by
Bodurtha and Courtadon (1986) for currency options and Ball and Torous (1986) for
gold and silver options (Gemmill 1993:258).
It can be concluded from these studies that the put-call parity relationship generally
holds and that options markets are relatively efficient in preventing abnormal profits
being available after accounting for transaction costs. However, market synchronization
is in doubt and possibly invalid.
b. Model-based efficiency tests
Various studies have used option pricing models in an attempt to identify overvalued
and undervalued options and then to test for market efficiency. If 1t was found that
excess returns, after transaction costs, can be earned on a consistent basis then
market inefficiency is implied (Blomeyer & Klemkosky 1982:112, 113).
The results of these studies have produced contradictory findings. Black and Scholes
(1972) concluded that the hypothesis of efficient markets can not be rejected, while
249
Galai (1977) found a degree of inefficiency. Chiras and Manaster (1978) also indi-
cated market inefficiency, but Phillips and Smith (1980) suggested that both Galai
(1977) and Chiras and Manaster (1978) had underestimated transaction costs and that
with the correct transaction costs, both market synchronization and market efficiency
are indicated (Blomeyer & Klemkosky 1982: 112; Gemmill 1993:258).
Later studies by Bhattacharya (1983) showed inefficiency in that a specific spread

strategy was found to be profitable and Evnine and Rudd (1985) found market
inefficiency in the pricing of index options. Chance (1986) refuted Evnine and Rudd's
1985 findings and found index options to be efficiently priced if allowance is made for
the bid-ask spread. Whaley (1986) concluded that options on Standard and Poor 500
futures are incorrectly priced - indicating inefficiency (Gemmill 1993:258).
Again, these tests generally show that options markets are efficient in their pricing me-
chanism, but these tests rely on the accuracy of the model used in the tests and,
hence, no firm conclusions can be drawn without establishing the model's accuracy and
validity.
c. Implied volatility efficiency tests
Latam! and Rendleman (1976) examined whether better forecasts of standard devi-
ations can be obtained when the option pricing model is used to estimate implied
volatilities, rather than using historical data for forecasts of standard deviation. Other
studies which have examined whether volatilities implied by options are good predictors
of future volatilities include those of Schmalensee and Trippi (1978), MacBeth and
Merville (1979), Beckers (1981), Brenner and Galai (1984) and Gemmill's (1986)
study on London data (Gemmill 1993:259).
Most of these studies used a regression of the implied standard deviation from the
option, the historic standard deviation and an error term in order to determine whether
implied volatilities capture all the information in historical volatilities. In most cases it
250
was shown that the implied standard deviation is significantly different from zero, while
the historic standard deviation is not. This implies that historic standard deviation
makes no information contribution and all its information content is captured in implied
volatility (Gemmill 1993:259).
Gemmill (1993:259) concludes that implied standard deviation performs better than
forecasts of volatility based on historical data. He notes, further, that neither implied
volatility nor historic volatility provides good predictions of future volatility.
Chiras and Manaster (1978) suggested that the implied standard deviation should be
weighted with the price elasticity of an option in relation to its implied standard de-
viation. They concluded that weighted implied standard deviation gives improved pre-
dictions of future volatility and has the benefit of not requiring information on historical
share returns (Blomeyer & Klemkosky 1982:105).
These are, again, joint tests of market efficiency and the model used. Market efficiency
is generally indicated, since the market-based volatility estimates have been shown to
be better than historic data-based estimates.
d. Impact of options trading on market efficiency
A number of studies have examined the impact of the introduction of share options on
the underlying shares. Hayes and Tennenbaum (1979) found that the introduction of
options had in general increased the volume of share trading. Both the studies of Tren-
nepohl and Dukes (1979) and Klemkosky and Maness (1980) found no significant
effect on the volatility of shares, while Whiteside, Dukes and Dunne (1983) found no
significant effect on either share trading or share volatility (Gemmill 1993:260,261 ).
251
No significant evidence has been found that options have a destabilising effect on the
stock market, but there is some evidence that stock market liquidity has improved since
the introduction of options and that there has been an increase in the trading volume
of shares due to option trading.
4.4.7.2 Research related to the 8-5 Option Pricing Model
a. Simulation tests on the model's assumptions
i. Distribution/volatility of returns
Various studies
According to Gemmill (1993:112) it follows from the work of Fama (1965) and Taylor
(1986) that the B-S model tends to overvalue both put and call options, since it under-
states the possibility of large positive and large negative returns. Instead of returns
being normally distributed, all sorts of assets have shown fat-tailed return distributions.
The empirical evidence from option returns seems to support the fat-tailed volatility, as
shown by Rubinstein (1985) and by Gemmill (1986). Gemmill ( 1993: 113) notes that
it is difficult to adjust the B-S model for fat-tailed volatility, and that in most cases the
volatility of options are simply up-rated.
Beckers (1980) and Christie (1982) found evidence that the return distributions for
shares are skewed and that the variance rises as the price drops. According to Gemmill
( 1993: 115), the higher variance at low prices is seen as a fat-tail on the left side of the
return distribution. Further, the implication of such skewness would be that the B-S
model will overvalue high-exercise-price options and undervalue those with low-exer-
cise prices. As a consequence, implied volatilities will differ across exercise prices and
this is known as the exercise-price-bias.
252
Various studies have confirmed the existence of such a bias. Black (1975) found that
low-exercise-price call options have low implied volatilities. MacBeth and Marville
(1979) found the opposite, in that high-exercise-price call options have low implied
volatilities. Rubinstein (1985) confirmed that the bias is unstable. In the first period of
his study it was found that low-exercise-price options have low volatilities, while in the
second period of the study the high-exercise-price options have low implied volatilities
(Gemmill 1993:115,116).
Merton (1976)
Merton showed that returns are not lognormally distributed as assumed by the B-S
model. Merton assumed that return distributions are a combination of a jump and diffu-
sion process, but showed that the B-S model can be adjusted for this as long as the
jumps are diversifiable in a large portfolio. Merton concluded that the effect of the mis-
specification of the return distribution on option values will generally be small, except
in the cases of options close to maturity, options on shares with low variance and deep-
out-of-the-money options (Galai 1982:55; Gemmill 1993:113,114).
Boyle and Ananthanarayanan (1977)

This study examined the effect of using estimates of variance in option pricing models.
It was in general found that the bias is relatively small, but that the dispersion of the
distribution of option values may be significant (Galai 1982:55).
Bhattacharya (1980)
In this study the robustness of the B-S model was also examined. The findings from the
simulation tests were that near-in-the-money and near-out-of-the-money options with
five days to expiration are often statistically significantly mispriced and that at-the-
money options with one day to expiration provide operationally and statistically signifi-
cant excess returns. The conclusion was that the B-S model does not misprice options,
except for at-the-money options with one day to expiration, and that the significance of
the mispricing decreases with an increase in time to expiration (Galai 1982:55,56).
253
Bookstaber (1981)
Bookstaber performed simulation tests to examine the effect of nonsimultaneity of share
and option price quotations on option values. He found that nonsimultaneity can have
a mispricing effect where there are less than 20 option trades during a day. The ·effect
is not as significant for at-the-money and out-of-the-money options (Galai 1982:56).
ii. Transaction costs
Leland (1985)
Leland investigated the effect of transaction costs on option values. According to the
8-S model:
• It will not be worth selling an overvalued call and buying the replicating put if the
expected profit does not exceed the transaction costs.
• An option hedge needs to be rebalanced with movements in the underlying share
price and the more frequent the rebalancing, the higher the transaction costs, but
the smaller the hedging errors.
Thus, it would seem to be possible to estimate the expected transaction costs and to
add them to the value of the option in order to establish whether profit opportunities are
actually profitable (Gemmill 1993: 118, 119).
Leland, however, suggested that this approach has three problems:

• It is difficult to estimate transaction costs when prices are volatile and more reba-
lancing will be required.
• Frequent rebalancing will cause transaction costs to be arbitrarily high.
• Transaction costs will be correlated with share and market returns, hence their
risk will not be diversifiable and the attempted hedge will become risky.
Leland suggested that the solution would be to raise the variance by some factor, which
will cause the expected return on the hedge to be zero as required and the risk will not
be correlated with share returns (Gemmill 1993: 121, 123).
254
iii. Dividends
Merton (1973)
Merton developed an adjusted-8-S model that can accommodate a continuous dividend
yield. This model found application on valuing currencies and bonds, but its per-
formance for share options depends on the frequency of dividend payments. It was
found that the model works well with frequent dividend payments, but relatively poorly
when dividends are infrequent and large, for example half-yearly dividend payments.
The model was also found to overvalue call options in some periods and undervalue
it ~n other periods. In general, it seems that it would be better to make an adjustment
to the 8-S model which takes into account the lumpiness of dividends (Gemmill
1993:95).
Black (1975)
Black advocated the use of a pseudo-American adjustment for dividends, whereby the
value of an option exercised just before the ex-dividend date 1s compared with the
value of an option which is allowed to run to expiration. The correct value of the call will
be the higher of the two values (Gemmill 1993:95).
Gemmill (1993:96,97) outlines the principles of the pseudo-American approach:

• It is assumed that there is only one dividend payment during the life of the option.
• The holder of an American call is considered, in effect, to hold an European
option.
• The first call expires immediately after payment of the dividend and the share
price is reduced by the present value of the dividend, but this is offset by the
reduction in the exercise price for the payment of the dividend.
• The second call expires later and pays no dividend. The share price is again re-
duced by the present value of the dividend, but there is no dividend to reduce the
exercise price with.
• Two 8-S valuations are performed and the larger value is used as the correct
value of the option.
255
Gemmill (1993:97) notes that the pseudo-American approach can be extended to

situations where more than one dividend is paid over the life of the option. Further, this
approach is widely used and reasonably accurate, although the call is generally slightly
undervalued because it is assumed that the holder has to decide now when the call will
be exercised.
Galai (1977)
Galai also examined the effect of dividend payments on CBOE options and found that
the B-S model's ability to identify underpriced and overpriced options declines with
increases in dividend yield. He concluded that the model performs better within its
basic assumptions, such as no dividend payments (Galai 1982:62).
Roll (1977), Geske (1979) and Whaley (1981)

Roll developed a model that solved the B-S model's problem with the valuation of
options on shares which are expected to make dividend payments before expiration.
Geske continued with the development of this model and Whaley modified it slightly
(Jarrow & Rudd 1983:143; Van Horne 1992:122; Weston & Copeland 1992:454).
The model is essentially a modification of the B-S model that incorporates the dividend,
the time left to going ex-dividend and the decline in the price of the share on the ex-
dividend date as a proportion of the dividend (Van Horne 1992: 122).
Whaley (1982)
The results of Whaley's test showed that the Roll and Geske model performs well and
that it eliminates all but one bias in the dividend-adjusted B-S model. He did, however,
find that the model still undervalues options on low-risk shares and overvalues options
on high-risk shares. Whaley concluded that these differences in value are small and
that both models perform well when trading imperfections are taken into account
(Jarrow & Rudd 1983:148).
256
Gultekin, Rogalski and Tinic (1982)

Gultekin et al. concluded in their study that the Roll and Geske model tends to over-
value options on high-risk shares and undervalues those on low-risk shares. In-the-
money options are also overvalued and out-of-the-money options are undervalued (Van
Horne 1992: 122).
Sterk ( 1983)
Sterk found that the Roll and Geske model performs well, especially when dividends
are large. The prediction errors are lower than with the dividend-adjusted 8-S model
(Van Horne 1992: 122).
1v. Early exercise of American put options
Merton, Scholes and Gladstein (1982)

They showed in their simulation study that the possibility of early exercise of American
puts has a considerable effect on option prices. Their study covered the period 1963
to 1976 and it was shown that the possibility of early exercise increases the price of
American puts by between 4% and 14% above those of European puts. The biggest dif-
ference was found in periods of high interest rates. They also showed that around 44%
of all at-the-money puts will be exercised before expiration (Galai 1982:70).
Unfortunately, the basic 8-S model is inadequate in valuing American puts where it may
be profitable to exercise the puts early, but there is no other equivalent to the 8-S
equation which allows for the principle that put options may be exercised any time
between now and expiration (Gemmill 1993:99).
Gemmill (1993:99) adds that one approach to this problem is to use the binomial model
instead of the 8-S model, but the binomial model is a very slow time-consuming
process. Another approach is to use some approximation of the 8-S model which may
be reasonably accurate and much faster. The approximation would be to use the higher
of the 8-S value or the exercise price of the put. This can be refined to add some other
257
factor to the B-S model, that is, the difference between European and American put
values. Two approaches to calculating this factor have been identified:
• Using an equation based on the same principles as the B-S equation.
• Using the control-variate approach.
Various studies
A number of studies have examined the use of an equation to calculate the factor. The
study of Johnson (1983) provided a fast but complicated equation which does not
perform well for long-term options. Geske and Johnson (1984) showed that if early
exercise is considered at three or four intervals before expiration, reasonably accurate
forecasts of a limit price can be made. McMillan (1986), published by Barone-Adesi
and Whaley (1987), provided a simple but untidy equation for the calculation of the
factor. The intuition behind this equation is that the factor is itself the value of the
option, which can be analysed by the B-S equation (Gemmill 1993: 100).
Hull and White (1988) published the control-variate approach, which had been widely
used in practice before then. This approach uses two binomial estimates of the put
value, one European and one American, with the factor being the difference between
the two. Although the binomial method is slow (SO iterations are required for con-
vergence of the binomial model to the B-S model), they suggested that only 25
iterations would be necessary to achieve a reasonable estimate of the factor (Gemmill
1993:99). Gemmill (1993: 99) suggested that his experiments showed that only 10
iterations are required for options with a few weeks left to expiration, while 15 iterations
will be sufficient for longer options.
These simulation studies on the assumptions of the 8-S model have generally indicated
that the model is extremely robust. It can be adapted to some real-life realities and still
provide reasonably accurate predictions, it is only with early exercise dividend-unpro-
tected put options that the model has some serious inadequacies.
258
b. Tests consisting of direct comparisons between actual prices and model

values

Black and Scholes found that their model undervalues options on low-variance shares
and overvalues those on high-variance shares (Weston & Copeland 1992:463).
Black (1975)
Black reported that the B-S model undervalues out-of-the-money options and over-
values in-the-money options (Jarrow & Rudd 1983: 140; Weston & Copeland 1992:463).
Merton (1976)
The results of Merton's study showed that the B-S mo~el undervalues both deep:-in-the-
money and deep-out-of-the-money call options (Jarrow & Rudd 1983: 140).
Trippi (1977)
Trippi performed a simple test for the period 30 August197 4 to 14 March 1975, whereby
all options with a B-S value of 15% more than market price were bought and those with
a value of 15% less were sold short. Based on the average weekly returns, Trippi con-
cluded that the C80E is inefficient and that the 8-S model provides accurate results.
It should, however, be noted that the returns were not risk-adjusted and, since option
trading is quite risky, this represents a serious drawback (Galai 1982:56).
Mac Beth and Merville ( 1979, 1980)

In their 1979 study, Mac8eth and Merville examined the difference between 8-S values
and actual call prices. They used the1976 daily closing prices of six shares as data for
their test. It was assumed that the market is efficient and that any difference between
actual prices and model values can be attributed to weaknesses of the model, espe-
cially its assumption of a constant variance of share returns. Mac8eth and Marville con-
cluded that the 8-S model correctly values at-the-money options with at least 90 days
to expiration, but overvalues deep-out-of-the-money options and undervalues deep-in-
259
the-money options. These differences decrease with a shortening in the time to expira-
tion and they suggested that their results may have been affected by the early exercise
of dividend-unprotected calls (Galai 1982:56, 76).
In their 1980 study, they compared the B-S model with the Cox model of constant
elasticity of variance. They assumed that the Cox model will reduce the degree of
mispricing observed in their 1979 study. They repeated their 1979 tests and concluded
that the share return-generating process is better described as a constant elasticity of
variance process (Galai 1982:57; Blomeyer & Klemkosky 1982:112).
Manaster (1980)
Manaster discussed the 1980 study of MacBeth and Marville and concluded that, since
the B-S model is included as a special case in the Cox model, their results were not
surprising. The Cox model should explain option prices at least as well as the B-S
model. They also cautioned against concluding that one model is superior to another,
as these were joint tests of both the accuracy of the model and market efficiency (Galai
1982:57).
Thorp and Gelbaum (1980)

Similarly to MacBeth and Marville (1980), Thorp and Gelbaum also believed Cox's
constant elasticity of variance model to be better than the B-S model. From their expe-
rience in trading on the CBOE they have found that the B-S model undervalues out-of-
the-money options. However, using 1979 data, they found only small deviations be-
tween actual prices and the B-S values. A number of reasons were suggested for the
differences between the two studies' results, including changes in tax rates in Sep-
tember 1976, changes in the volatility of the market, better adjustments of results for
potential early exercise of call options and more accurate estimation of the variance in
share returns (Galai 1982:58).
260
MacBeth (1981)
Using 1978 data, Mac8eth repeated the earlier work done with Marville in 1979 and
1980. The study confirmed earlier findings that the parameters of either or both the Cox
and the 8-S model are nonstationary over time. The results from this study showed that
the 8-S model provides accurate predictions of market prices in some periods, while
in other periods it misprices options in a manner opposite to that observed in their ear-
lier studies (Galai 1982:58).
Rubinstein (1981)
Results similar to those of Mac8eth and Marville (1979) were found. It was further
shown that the bias for in-the-money and out-of-the-money options reversed itself
around 1977 (Weston & Copeland 1992:464).
Blomeyer and Klemkosky (1982)

This study compared the 8-S model with the model developed by Roll. Since Roll's
model is more general than the 8-S model, in that it can value dividend-unprotected
calls, it is expected to give more accurate predictions of market prices. 81omeyer and
Klemkosky examined C80E options for 18 NYSE listed shares for the period July 1977
to June 1978, with all the options having at least three weeks remaining until expiration
and having one ex-dividend date prior to expiration. Their results contradicted those
of Mac8eth and Marville (1979), but are consistent with those found by Thorp and Gel-
baum in 1980. They found that both models have nearly identical pricing-bias characte-
ristics, since both undervalue out-of-the-money call options and value at-the-money
and in-the-money call options fairly accurately (Galai 1982:58: Blomeyer & Klemkosky
1982:104, 111 ).
81omeyer and Klemkosky ( 1982: 112) concluded that the pricing-bias characteristics
found with the 8-S model, the Roll model and earlier with the Cox model suggest that
in-the-money and out-of-the-money pricing bias is not due to the possible early exer-
cise of the call options, but rather to a share return-variance related problem.
261
Blomeyer and Resnick (1982)

This study used a subsample of the 1982 study of Blomeyer and Klemkosky to compare
the Geske compound option model, a further development of Roi l's model, with the B-S
model. Geske's model considers options on shares to be an option on an option and
the model allows for the nonstationarity in the variance of share returns. It was found
that the Geske model overvalues out-of-the-money calls and undervalues in-the-money
calls. These results are directly opposite to those found for the B-S model (Galai
1982:58,59).
Sterk (1982), Whaley (1982) and Geske and Roll (1984)

Both Sterk and Whaley used the dividend-adjusted model of Roll and Geske and found
that it explains the time to maturity and in-the-money and out-of-the-money biases.
Geske and Roll used a super variance estimator to explain the variance in share
returns bias (Weston & Copeland 1992:464).
From the research discussed above, it can be concluded that the B-S model is as accu-
rate as any other model examined and that, in general, no other model provides better
descriptions of option pricing or option values. Most of the other models have additional
drawbacks in that they are either mathematically more complex or slower and more
cumbersome than the B-S model.
c. Hedging tests

This study proposed a method for adjusting for the risk changes that occur when an
option is held. Black and Scholes suggested creating a neutral hedge by buying the
option and at the same time selling the share, or selling the option and buying the
share (Galai 1982:59).
They examined the OTC market for the period May 1966 to July 1969 and compared
their model values with the actual prices on the date the options were issued. From this
262
they classified the options into two groups; those that the model had undervalued and
those it had overvalued compared to the market. They used this information to adjust
their neutral hedge on a daily basis and expected the excess return to be riskless. It
was found that the B-S model overvalues options on high-variance shares and under-
values options on low-variance shares. Historic data was used to estimate variance and
they concluded that if variance can be more accurately estimated, the model's perfor-
mance will be improved (Galai 1982:60,61 ).
They also observed nonstationarity in variance which may indicate market inefficiency,
but found that excess profits disappear after accounting for transaction costs (Galai
1982:61; Dobbins et al. 1994: 156).
Galai ( 1977)
Galai used the daily closing prices of all options traded on the CBOE for the period 26
April 1973 to 30 November 1973. Galai found the hedge strategy to be effective and
that deviations from the model values generated excess profit. However, almost all the
hedge returns are eliminated with the imposition of transaction costs (Galai 1982:61 ).
Chiras and Manaster (1978)

These researchers calculated the spread returns on a long position in one option and
a short position in another option on the same underlying share. A dividend-adjusted
B-S model was used and they formed 118 positions during the period July 1973 to April
1975. It was found that 93 of these positions were profitable and they concluded that,
since the model values the options more accurately than the market, this proved market
inefficiency. This conclusion, however; was qualified due to the ex-post nature of their
tests and the potential nonsimultaneity of option prices (Galai 1982:63).
Boyle and Emanuel (1980) and Galai (1983)

Boyle and Emanuel simulated the effect of discreteness, that is, non-continuous
trading, on hedge returns and found the effect to be quite small in absolute terms (Galai
1982:56). Galai also examined whether the discreteness of trading - the B-S model
263
assumes continuous trading - will affect hedge returns. He showed that it does not, as
the major component in hedge returns is changes in the deviation between model
values and market prices, while the opportunity cost and discreteness adjustments are
marginal (Brenner 1983:63).
Phillips and Smith (1980)

In an examination of the 1980 results of Chiras and Manaster, this study showed that
the profit opportunities are eliminated with the introduction of transaction costs (Galai
1982:63).
Bookstaber (1981)
A further re-examination of the Chiras and Manaster ( 1980) results by Bookstaber
provided evidence of nonsimultaneity in their data and it was concluded that the profits
observed in their study are not achievable in practice (Galai 1982:63).
Blomeyer and Klemkosky (1982)

Using the hedging technique of Black and Scholes (1972), Blomeyer and Klemkosky
compared the ability of the B-S model and Roll's model to identify overvalued and
undervalued options. This test formed part of the test previously examined under com-
parative tests and, although it was expected that Roll's model will prove to be superior,
it was found that the results indicated that the Roll model is no better than the B-S
model in identifying under- or overvalued options. They concluded that their results
support market efficiency and that the B-S model is an acceptable alternative to the
mathematically complex model of Roll (Blomeyer & Klemkosky 1982:64, 110, 111, Galai
1982:64).
The above studies have shown that, assuming market efficiency, the B-S model is able
to provide a reasonably accurate description of the pricing mechanism of options mar-
kets.
264
d. Implied volatility tests
Latam~ and Rendleman (1976)

Latane and Rendleman were the first to advocate the use of ISO (implied standard
deviation) as estimator of the volatility of share returns. They used the weekly closing
prices of the options and shares of 24 companies for the period 5 October 1973 to 28
June 1974 to calculate the individual ISDs. They used the weighted average implied
standard deviation (WISD) as estimates of volatility of returns. Latane and Rendleman
concluded that the WISD, based on the B-S model, is useful in identifying overvalued
and undervalued options and to determine the proper hedge position. Further, the B-S
model tends to overvalue options and, although it may not capture all aspects of the
option price-generating process, the model is efficient in evaluating whether individual
options are properly priced (Galai 1982:65,66).
Schmalensee and Trippi (1978)

Schmalensee and Trippi assumed the B-S model to be valid and used it to impute the
IS Os of six widely traded shares and their options from weekly data over the April 1974
to May 1975 period. They used an arithmetic average of the ISDs, based on closing
prices, as estimates of volatility of share returns. Based on the assumption that the
model is valid, they found evidence of market inefficiency, but suggested that the model
may also be inappropriate. They concluded that the B-S model is valuable in predicting
volatilities, even though it is inconsistent in its estimates (Galai 1982:66).
Beckers (1980,1981,1983)
Beckers (1980) noted that, although evidence suggests that the B-S model is valuable
in predicting future volatilities, there is a basic inconsistency in using the model to
estimate a presumably nonstationary variance. Beckers considered alternative
weighting schemes for estimating volatility and used a dividend-adjusted model on a
sample of CBOE options for the 13 October 1975 to 23 January 1976 period. The ISDs
derived from at-the-money options were found to be better predictors of volatility than
either WISDs or historic volatility estimates. Also, ISDs were found to be extremely
265
unstable over time and while this may indicate market inefficiency, Beckers attributed
the instability to the trading mechanism and a lack of market synchronization (Galai
1982:66).
In 1981, Beckers expanded the sample period to May 1975 to July 1977 and confirmed
his earlier findings, using five-day arithmetic averages of the ISDs as volatility esti-
mators. Again, this could be attributed to market inefficiency, model misspecification
or data errors. The latter seemed the most likely, since it was found that closing prices
may seriously distort the ISO estimates (Galai 1982:66,67).
Beckers (1983) tested the accuracy of a high-low variance estimator and found it to be
better than an estimator using closing prices. He suggested that the high-low variance
estimator can be improved if the cross-sectional differences between high-low data
across shares are included in the estimator (Hawinkels 1987:25).
Garman and Klaus (1980) and Parkinson (1980)

Garman and Klaus examined a number of estimators o'f price volatility and showed that
closing prices are not the most efficient estimator. Parkinson showed that the use of
high and low prices provide far superior volatility estimates than the use of closing
prices (Hawinkels 1987:24,25).
Brenner and Galai (1981)

Brenner and Galai examined the distribution properties of ISDs, using five shares for
the period 3 June 1977 to 21 October 1977. With ISDs based on daily averages, they
found significant deviations from ISDs calculated using the last transaction of the day.
Options with longer time to expiration usually have higher average ISDs than those
close to expiration and there are also significant differences across exercise prices.
They also found that average ISDs are unstable over time and concluded that these
results confirm that the joint hypothesis - that the B-S model is correct, that markets are
efficient and synchronous, and that the estimation procedure is correct - should be
rejected (Galai 1982:67).
266
Rubinstein (1981), MacBeth and Merville (1979) and MacBeth (1981)

Rubinstein used ISDs to test a number of alternative option pricing models with the aim
of establishing which model explains the B-S model biases the best. Rubinstein found
that out-of-the-money options close to expiration have higher ISDs, meaning they are
overpriced. He concluded that not one of the models are able to explain all the biases;
some explain time to expiration biases while others explain exercise price biases.
These results confirmed the findings of MacBeth and Marville (1979), in that ISDs
tended to rise with declining exercise prices for 1976 data up to October 1977.
However, towards the end of 1977 and during 1978 the bias was reversed, as was also
found by MacBeth (1981 ). Rubinstein concluded that a composite model should be
developed and, further, that any bias observed in a period should be correlated with
the level of some macro-economic variables, such as interest rates and stock market
prices and volatilities (Galai 1982:67,68).
The volatility-based studies have identified some problems with the B-S model, but
these can also be evidence of market efficiency. However, the main problem lies with
estimating volatility and before more accurate estimates can be made, no serious con-
clusions can be drawn about the validity of this parameter of the model.
e. Tests on put options
Brennan and Schwartz (1977)

Brennan and Schwartz used finite-differences methods to solve their model for divi-
dend-unprotected American puts. The model was tested on 55 observations of OTC
puts for the period May 1966 to May 1969 and although the options were dividend pro-
tected, they were not perfectly protected. Since their model had assumptions consistent
with the B-S model, they used ISDs derived from calls with the same terms for their put
valuation and used the B-S model to value the puts as if they were European options.
Only small differences between their model values and the 8-S model values were
found, although both overvalued puts by between 25% to 40%. From these findings
267
they rejected the put-call parity relationship, but concluded that the 8-S model provides
reasonably accurate values for six-month dividend-protected American put options
where the right to early exercise has no significant economic value (Galai 1982:73).
Parkinson (1977) and Farkas and Hoskin (1979)

Parkinson applied a numerical-integration approach to derive a model for valuing
American puts, with assumptions consistent with the 8-S model (Galai 1982:70).
Farkas and Hoskin tested Parkinson's 1977 model, using weekly closing prices of
options and shares for options traded on the C80E for the period 3 June 1977 to 24
December 1977. They used historical data to estimate volatility and compared the
model values with the actual prices for 365 puts. They found significant differences and
concluded that the model performs better for at-the-money puts than for out-of-the-
money puts. No significant relationship between the accuracy of values and the time
to expiration was, however, found (Galai 1982:73,74).
Studies into put option pricing have so far provided inconclusive evidence and the de-
viations observed could either be due to model invalidity, market inefficiency or lack of
market synchronization. Much research is still required on the pricing and behaviour
of put options.
4.4.7.3 South African options research
Le Plastrier and Thomas (1984)

Le Plastrier and Thomas examined the degree of correlation between the values
provided by the 8-S model and actual South African option prices. They found a high
degree of correlation and concluded that this is not surprising as most market makers,
at that stage, used the 8-S model to establish option prices. Thus, the accuracy and
validity of the 8-S model, in a South African context, remains to be tested (Hawinkels
1987:11 ).
268
Research into the efficiency of the options market and the validity of the 8-S Option
Pricing Model is only in its infancy in a South African context and the only conclusion
that can, at this stage, be reached is that more research is required.
4.4.8 Conclusion
From the review of the principles and nature of options and the 8-S Option Pricing
Model, and the review of empirical research up to the early 1980s, the following conclu-
sions can be drawn at this stage.
• Market efficiency
There does not appear to be much evidence of pricing inefficiency in options markets
and although some tests have reported profit opportunities, these are eliminated after
proper adjustment for transaction costs. Outsiders to the exchanges are not able to
consistently earn abnormal returns and even market makers, who are affected by the
bid-ask spread and opportunity costs, are unlikely to generate excess profits on any
consistent basis. The case for market synchronization is much clearer and, at this
stage, there appears to be sufficient evidence to reject the hypothesis about market
synchronization. Significant ex-post hedge returns reported in some studies indicate
a lack of either data synchronization or trading synchronization and this conclusion is
confirmed by the results of the tests on option boundary conditions (Galai 1982:69).
Market-based implied volatilities, although not that efficient in predicting future volatili-
ties, also support the hypothesis of market efficiency, in that they provide better esti-
mates than those based simply on historical data. The impact of options on the under-
lying shares and stock markets appears to be relatively small and benign, although
there appears to be a slight destabilising effect on expiration of the options. These
findings are, however, not significant enough to contemplate regulation of options
markets (Van Horne 1992:263).
269
• The 8-S Option Pricing Model

81omeyer and Klemkosky (1982:119) and Dobbins et al. (1994:156) conclude that the
8-S model generally performs well in identifying underpriced and overpriced options.
Thus, it allows traders to identify profit opportunities, though options markets are not
inefficient enough to allow profits to persist after transaction costs are considered.
According to Jarrow and Rudd ( 1983: 140, 141 ), few models in finance have such pre-
dictive accuracy as the 8-S model and evidence from empirical tests show that it is
sufficiently accurate to be used for investment decision-making. They do, however, add
that the model has some biases, but there exists a lack of consensus about the mag-
nitude and direction of these biases. It seems as if these biases change over time and
it might be because the model fails to capture all the characteristics of the underlying
share return-generating process, such as skewness and changes in variance.
While these general conclusions provide strong support for the 8-S model, it is also
useful to look at specific conclusions regarding various aspects of the model:
• The model provides good predictions of actual market prices for at-the-money
options with a medium to long time to expiration. Some consistent deviations
have, however, been found for deep-in-the-money and deep-out-of-the-money
options (Galai 1982:46,59,68).
• No other model is currently able to provide a consistently better explanation of

actual prices over time than the 8-S model. There is some evidence in favour of
Cox's constant elasticity of variance model, but these are as yet not conclusive
(Galai 1982:46,59,68).
• The biggest problem experienced with the 8-S model, or for that matter any other
model suggested so far, is the nonstationarity in the underlying share returns and
the resulting instability in the volatility estimator. The nature of the nonstationarity
is as yet not clear, but the 8-S model does provide good predictions of actual
270
prices over short periods and is even able to reveal underpriced and overpriced
options (Galai 1982:52,69).
• The B-S model does not, nor does any other model, account for transaction costs
and taxes and these may affect market prices of options (Galai 1982:69).
• The trading mechanism and lack of market synchronization may have affected
some of the test results. However, the major conclusions were reconfirmed in
tests using more detailed and accurate data (Galai 1982:69).
• The B-S model can provide incorrect prices when its assumption of no dividends
is violated, but it has been shown that the model can be used to value dividend-
unprotected American call options without significant bias (Blomeyer & Klemkosky
1982:119).
• Other violations of the model's assumptions have generally resulted in only small
and statistically insignificant deviations. These deviations also tend to decrease
with an increase in the time to expiration and an increase in the depth of being in-
the-money (Galai 1982:56).
4.5 SUMMARY
The Capital Asset Pricing Model provides the investment world with a theory that
explains the risk/return relationship, including how to determine expected returns on
shares. Although based on a number of clearly unrealistic assumptions, it provides a
better description of the share-pricing process than existed previously. The basic prin-
ciple of the CAPM is quite sensible and enlightening, that is, there exists a positive
linear relationship between risk and return and the relevant risk of a share is beta - a
measure of its risk relative to that of the market portfolio.
271
As with any other capital market theory or model, the CAPM can not be judged solely
on the realism of its assumptions. It is only through empirical testing that its validity can
be established. The CAPM has been subjected to a huge body of empirical testing and,
although there is conflicting evidence and disagreement between the findings, the main
conclusions from the early to mid-1980s tests, are the following:
• There does appear to be a positive linear relationship between risk and return,
although the slope of the SML has generally been found to be less steep than
predicted by the CAPM.
• The intercept alpha has generally been found to be positive, that is, higher than
the risk-free rate.
• Investors are only rewarded for assuming systematic risk since unsystematic risk
is not priced.
• Betas of portfolios have been found to be relatively stable over time and are
reasonable predictors of future betas, but the same cannot be said about indivi-
dual share's betas which were found to be quite unstable.
• Beta provides a reasonably good description of risk, but it seems that it is not a
complete measure of risk, and other risk factors may also be important in deter-
mining returns.
There are, however, doubts about the validity of a large number of the CAPM studies.
One of the areas of criticism is the model's dependence on the existence of a true mar-
ket portfolio. Such a market portfolio is at this stage unobservable and it may never be
observable. Testing the CAPM using the proxy for the market portfolio may not be a
true and useful test of the CAPM and there is some evidence that using different
proxies will yield varying results.
Another problem with testing the CAPM is that it is an expectational model, formulated
on an ex ante basis, while the tests are performed on an ex post basis. As investor
expectations can never be known with certainty, it is unlikely and even impossible that
there will ever be reliable tests for an expectational model. Despite these problems, the
272
CAPM has received a remarkable amount of empirical support and it can be concluded
at this stage that, until it has been totally discredited or until an improved model has
been validated, the CAPM should continue to be used in estimating future returns.
An alternative model to the CAPM, which has received wide attention, is the Arbitrage
Pricing Theory. The APT is a more general and simpler model with less restrictive
assumptions, but like the CAPM it has its limitations.
The APT also posits a relationship between expected return and risk but, unlike the
CAPM, it does not depend on the underlying market portfolio as tho only source of risk
affecting expected returns. The APT recognises that many types of risk may affect re-
turns, but this also represents a major limitation, in that these economic risk factors are
not identified by the theory.
Hence, much of the research on the APT has concentrated on the determination of the
number and identity of the macro-economic factors that affect returns. Before these fac-
tors have been identified and shown to be stable, the APT will not have any significant
practical application. By the mid-1980s, empirical studies have in general shown that
more than one risk factor is significant in explaining returns, but these studies have so
far only been preliminary. There are conflicting evidence regarding the number of
factors, and as yet no clear picture of the identity of these factors exists. The tests have
also been of such a nature that the APT can not be rejected from the information avail-
able.
For the foreseeable future, both the CAPM and APT will remain the focus of much
debate and empirical testing. It is only through continued testing and investigation that
it will be established which model explains and predicts returns the best.
Unlike the CAPM and APT. Options Theory and option pricing models, and in
particular the Black-Scholes Option Pricing Model, are concerned about total risk
273
and not only systematic risk. The price/value of an option is driven by the total volatility
of the underlying share.
Both call options and put options can be seen as investment opportunities in their own
right, or can be viewed as additions to an investment portfolio. In a portfolio sense,
options provide an inexpensive and convenient way of changing the risk/return
structure of a portfolio. Options are also useful tools for hedging against risk.
The 8-S model provides an intuitively appealing means of valuing options in efficient
markets. The first step of this valuation process requires that a neutral riskless hedge
position be established and this is achieved by combining options with the underlying
share. This hedged position is directly affected by movements in the underlying share's
price (its volatility), and it is thus necessary to rebalance the riskless hedge portfolio
on a regular basis. The value of the option, however, can be determined by using the
specifics of the option contract, the price of the underlying share and the hedged
portfolio which is assumed to earn the risk-free rate.
Empirical testing of the 8-S model has the problem that the tests are joint tests of the
hypotheses of market efficiency, market synchronization, model validity and data
accuracy. From the studies reviewed it can be concluded that, by the mid-1980s, the
status of each of these hypothesis were as follows:
• Options markets are efficient and relatively few, if any, opportunities exists for
making abnormal profits after transaction costs have been taken into account.
• The evidence tends to suggest that the hypothesis of market synchronisation is
at this stage invalid.
• While most of the parameters of the 8-S model are easily identifiable and stable,
the tests have supplied evidence that the distribution of share returns is
nonstationary. It is only through improved estimation of the underlying share's
volatility that better predictions of option values will be obtained.
274
• Although the B-S model assumes that the options to be valued are European
options on non-dividend paying shares, it has been shown to be an extremely
flexible model which can be extended to value American style options on dividend
paying shares.
• Some biases in B-S model values have been reported in many of the studies, but
the overall conclusion seems to be that it is an accurate model that provides good
predictions and explanations of option prices.
Research into the efficiency of options markets and the accuracy of option pricing
models will continue. There is a need to minimise the problem of market synchroni-
zation and to establish more accurate estimators of the volatility of the underlying
shares. Further, adjusted-B-S models and new models need to be tested in order to
explain some of the biases identified and to obtain more accurate valuations of
dividend-unprotected American put options with the possibility of early exercise.
In a South African context no significant conclusions can at this stage be formed on

the CAPM, the APT, the B-S model and the efficiency of the options market. Although
there is some evidence against the CAPM and in favour of the APT, the limited number
of studies and the resultant small body of evidence prevents the making of any definite
statements about their validity. Much research still needs to be done, especially re-
garding the relatively new options market, before any firm conclusions can be drawn.
275
CHAPTERS
RECENT DEVELOPMENTS IN THE VARIOUS CAPITAL MARKET

THEORIES AND PRICING MODELS
5.2 Recent developments in Portfolio Theory 277
iif 5.2.1 Research related to the benefits of diversification 278

f
5.3 Recent developments in market efficiency and the EMH 281
5.3.1 Research related to the EMH 281

5.3.2 Some alternatives to the EMH 337
5.4 Recent developments in the CAPM 343
5.4.1 Research related to the CAPM 344

5.5 Recent developments in the APT 361
5.5.1 Research related to the APT 362

5.6 Recent developments in Options Theory and the B-S Option Pricing
Model 375
276
5.6.1 Recent research related to the efficiency of options markets 377

5.6.2 Recent research related to the B-S Option Pricing Model 382
5.6.3 Share-index options, portfolio insurance and the market crash
of October 1987 386
5.1 INTRODUCTION
This chapter focuses on recent developments in the capital market theories and pricing
models examined in chapters 3 and 4. Since the background to, and principles un-
derlying these theories and models have already been extensively reviewed in those
chapters, this chapter only examines new developments since the mid-1980s and only
reviews and classifies the related research from the mid-1980s up to the late-1990s.
The first section examines Portfolio Theory and focuses, in particular, on how quickly
the benefits of diversification can be attained and whether international diversification
provides greater benefits than can be achieved with domestic diversification. This is
followed by a review of recent research on the three forms of market efficiency, in order
to establish whether the Efficient Market Hypothesis provides an accurate description
of the behaviour of capital markets. Some alternative efficient market theories which
have recently been developed and proposed, are also included in this review.
The third and fourth sections concentrate on the asset pricing models, namely the
Capital Asset Pricing Model and the Arbitrage Pricing Theory. Most of the recent
studies have concentrated on the validity, testability and usefulness of these models
and on whether the multiple risk factors of the APT provide a superior description of the
return-generating process than that provided by the CAPM's beta. Hence, numerous
studies have been performed to establish the identity of the priced factors which may
affect share returns.
277
The final section of the review examines Options Theory and, in particular, the
efficiency of options markets and the robustness and validity of the Black-Scholes
Option Pricing Model. More recent developments in Options Theory, such as share-
index options and portfolio insurance, and their possible role in the world stock market
crash of October 1987, are also examined and reviewed.
5.2 RECENT DEVELOPMENTS IN PORTFOLIO THEORY
Reilly and Brown (1997:251,253) conclude that a major advance in the investment field
has been the recognition that optimum investment portfolios can not be created by
simply combining shares with desirable risk/return characteristics. The development of
Portfolio Theory showed that consideration of the relationship between different shares
is essential in order to ensure that investment objectives are met. Portfolio Theory em-
phasises the importance of the risk reduction role of diversification, in particular efficient
diversification.
As discussed in chapter 3, diversification has the aim of reducing the portfolio's stan-
dard deviation (risk) and, through the addition of shares, attempts to reduce the average
covariance of returns (Reilly & Brown 1997:284 ). They (p.285) add that by including
shares which are not perfectly correlated in the portfolio, the overall standard deviation
can be reduced, although the variability of return will not be eliminated. Through effi-
cient diversification the standard deviation can be reduced to the level of the market
portfolio, that is, all unsystematic risk would have been diversified away and the only
risk remaining would be systematic risk, which can not be eliminated.
Following on the earlier research on Portfolio Theory and the benefits of diversification,
the main areas of investigation of the recent research are:
• Further investigation into how quickly the benefits of diversification can be
achieved.
• An examination of whether international diversification provides greater benefits
than domestic diversification.
278
The results of some of the recent studies are reviewed in the next section.
5.2.1 Research related to the benefits of diversification
5.2.1.1 General studies on the benefits of diversification
Tole (1982)
Tole confirmed the earlier findings that the major benefits of diversification are achieved
quickly, that is, with the introduction of the initial number of shares to the portfolio.
Around 90% of the benefits of diversification are obtained when portfolios consist of
between 12 to 18 shares (Reilly & Brown 1997:285).
Statman (1987)
Statman examined the actual data on US shares to determine the benefits of naive di-
versification. It was found that with randomly selected portfolios the average portfolio
risk can be diversified down to approximately 19%. Further, there is a dramatic re-
duction in risk when additional shares are added to the portfolio. Statman also found
that approximately half of the portfolio standard deviation can be eliminated when the
portfolio size reaches 10 shares. However, the benefits of random diversification do not
improve significantly with the addition of further shares to the portfolio. Increasing the
size of the portfolio to 20 shares eliminates only an additional 5% of portfolio standard
deviation and increasing it to 30 shares eliminates only a further 2% of standard de-
viation. The conclusion was that large number of shares are not required to achieve a
substantial benefit from diversification (Jones 1998: 181 ).
Statman also examined the benefits of diversification in comparison with the added tran-
saction costs incurred with more shares. The conclusion was that a borrowing investor
should have a well-diversified portfolio of at least 30 shares and a lending investor
should have a portfolio of 40 shares (Reilly & Brown 1997:285 ).
279
Newbould and Poon (1993)

The study of Newbould and Poon showed that a large proportion of diversifiable risk is
eliminated when portfolios consist of 50 to 60 randomly selected shares (Dobbins et al.
1994:26).
5.2.1.2 The benefits of international diversification
Ibbotson, Siegel and Love (1985)

Ibbotson et al. examined the performance of numerous assets around the world. They
constructed a value-weighted portfolio of shares, bonds, cash, real estate and precious
metals for the period 1960 to 1984. These assets were selected from the markets of
Australia, Canada, Japan, Hong Kong, Northern and Western Europe, Singapore and
the US. They computed annual returns, risk measures and correlations among the re-
turns for alternative assets. A conclusion from their study was that investors can obtain
a lower level of portfolio systematic risk by diversifying globally, rather than by only in-
vesting in their domestic market. They found that systematic risk factors in one country,
for example monetary policy, are not correlated with systematic risk factors in other
countries and, hence, through global diversification, portfolio systematic risk can
eventually be reduced to a world-systematic risk level (Reilly & Brown 1997:91,285).
Bailey and Lim (1992)

This study examined the performance of investment funds in France, Germany, Korea
and Spain to establish whether these funds enable investors to attain international
diversification. They found that the returns on the funds resemble US domestic share
returns, rather than foreign share portfolios, and concluded that these funds do not pro-
vide the expected benefits of international diversification (Reilly & Brown 1997:977).
Reilly and Brown (p.977) state that the findings of Bailey and Lim should not be seen
as evidence against the benefits of diversification, but rather as confirmation of market
the efficiency research which has shown that fund managers generally perform worse
than the market return.
280
Michaud, Bergstrom, Frashure and Wolahan (1996)

From their analysis of international equity investing over the preceding 20 years,
Michaud et al. concluded that international diversification no longer provides the same
opportunities of higher returns and lower risk as it did previously. They, however, added
that the evidence indicates that investors can still improve the risk/return characteristics
of their portfolios through thoughtful international diversification (Jones 1998: 16, 17).
Bhana (1987)
Shana studied the benefits of global diversification, and found that if South African in-
vestors are able to split their investment portfolios - half in the domestic market and half
in 17 different foreign markets - they will be able to increase their average returns and
at the same time decrease the risk substantially. It was also concluded that the dramatic
reduction in risk can in general be attributed to the low covariance of JSE returns with
the returns on foreign stock exchanges (Ross et al. 1996:312).
5.2.2 Conclusion
The evidence from these studies confirms that investors do not require a large number
of shares in their portfolios to achieve the benefits of diversification. It has also been
shown that international diversification has additional risk/return benefits, as stated by
Elton and Gruber (1995:288):
The evidence that international diversification reduces risk is uniform

and extensive... Unless there are mechanisms such as taxes and
currency restrictions that substantially reduce the return on foreign
investment relative to domestic investment, international diversification
has to be profitable for investors of some countries, and possibly all.
281
5.3 RECENT DEVELOPMENTS IN MARKET EFFICIENCY AND THE

EMH
According to Elton and Gruber (1995:406), the EMH and the concept of efficient capital
markets have been among the predominant themes in investment management and
academic literature since the 1960s. The EMH, and the testing thereof, historically con-
sists of three categories or forms of market efficiency:
• Weak form efficiency. These tests are concerned with whether all the relevant
information contained in historic share prices is fully and rapidly reflected in
current share prices.
• Semi-strong form efficiency. Testing of this level of market efficiency entails
investigation of whether all publicly available information is fully and. rapidly
reflected in current share prices.
• Strong form efficiency. This level of market efficiency requires that the tests
determine whether both publicly available and privately held information are fully
and rapidly reflected in current share prices.
The concept of market efficiency, the rationale behind the EMH and the results of
research, up to the mid-1980s, on the various forms of market efficiency have already
been reviewed in chapter 3. In the next two sections the recent research on the EMH,
and some alternative theories, will be reviewed.
5.3.1 Research related to the EMH
The tests of weak form market efficiency can be divided into the following major
groupings:
• Statistical tests of the independence between share returns. These tests attempt
to establish whether there is any dependence between successive price changes
282
and whether any such evidence can be used to earn abnormal returns (Reilly &
Brown 1997:212,213).
• Investigation of trading rules in order to establish whether investment decisions

based on trading rules render returns that differ from a simple buy-and-hold
strategy. Again, the aim is to establish whether such trading rules can be applied
to consistently earn above-average returns (Reilly & Brown 1997:212).
• Testing for any evidence of market irrationality, abnormal volatility and market
overreaction. These tests attempt to establish whether there are significant diffe-
rences between share prices and the present value of their future cash flows
(Elton & Gruber 1995:437).
The tests of semi-strong form market efficiency can be divided into the following
basic groupings:
• The first group of tests are return prediction studies, whereby researchers use
time-series analysis and publicly available information in an attempt to predict
future returns. Alternatively, researchers use cross-sectional analysis in order to
establish whether specific variables can be used to predict which shares will
yield abnormal risk-adjusted returns. It is important to note that cross-sectional
return studies are joint tests of the EMH and the asset pricing model used. This
creates the problem that anomalous results can be caused by market inefficiency
or model misspecification (Reilly & Brown 1997:216,223).
• The second set of tests are event studies. These studies examine whether ab-
normal rates of return are yielded immediately after the announcement of signi-
ficant economic events, that is, the tests attempt to establish whether investors
can earn excess risk-adjusted returns by investing after the release of informa-
tion on significant economic events (Reilly & Brown 1997:216,217).
283
Strong form tests can be divided into the following three main groupings, according
to the investors involved (Reilly & Brown 1997:234,235):
• Firstly, trading by corporate insiders is investigated to establish whether they can

consistently earn risk-adjusted excess returns.
• The second group of tests analyse whether the recommendations from stock
market specialists, such as share analysts, can be used to earn abnormal risk-
adjusted returns.
• The performance of professional money managers is the focus of the third group
of tests. These tests, too, attempt to ascertain whether these fund managers are
able to outperform the average market return on a regular basis.
The latter group of tests can be subdivided into two sub-groups:
i. Tests which examine whether certain market specialists/fund managers

have the ability to outperform others and the market on a consistent basis,
that is, whether their past performance 1s a reliable 1nd1cator of future
performance (Reilly & Brown 1997:235,979).
ii. Other tests examine whether certain market specialists/fund managers

have superior market timing and share selection ability, that is, an ability
to anticipate bull or bear markets and to correctly identify shares which
will yield abnormal returns (Reilly & Brown 1997:978: Bradfield 1999:2).
The results of some of these studies are reviewed below

284
5.3.1.1 Classification of EMH research
Fama (1991)
In 1991 Fama reviewed his earlier classification of the three forms of the EMH and,
again, classified the tests and empirical results into three groups. The weak form cate-
gory was broadened to include some of the areas previously tested in the semi-strong
form category. According to the new classification, the three forms of EMH research
were changed as follows:
• Weak form - Tests of return predictability.
• Semi-strong form - Event studies or studies of announcement.
• Strong form - Tests for private information (Elton & Gruber 1995:407; Reilly &
Brown 1997:211; Jones 1998:262,264).
Fama also changed the basic hypothesis that the information linked to each category
be fully reflected in share prices. He referred to a weaker and more sensible version,
in that prices must reflect information to such an extent that no financial advantage can
be obtained by acting on any information (Jones 1998:257).
In this chapter the categories will still be defined as weak form, semi-strong form and
strong form, but the tests and their results will be allocated and presented according to
the new categories, as presented in Reilly and Brown (1997:211-241 ).
5.3.1.2 Global applicability of EMH research
Hawawini (1984)
Hawawini reviewed the research on the behaviour of European share prices and the
efficiency of European capital markets. The evidence showed that the behaviour of
European share prices, even though the markets are smaller and less active, are
remarkably similar to those of US shares. Most of the results on European shares
confirmed the findings on US data and hence that European markets are as infor-
mationally efficient as the US market. It can thus be concluded that investors can
285
assume that markets outside the US have the same level of information efficiency as
the US market (Reilly & Brown 1997:246,247).
5.3.1.3 General EMH research
Stoll and Whaley (1983), Thaler (1987) and LeRoy (1989)

LeRoy concluded that reported market inefficiencies have done much to damage the
reputation of the EMH, while both Thaler and Stoll and Whaley have reported that most,
if not all, of the surviving anomalies are not exploitable by ordinary investors (Shana
1994:80,81 ).
Keane (1986, 1989, 1991)

According to Keane (1986, 1989) the degree of market efficiency ranges in various de-
grees between perfect efficiency and complete inefficiency. Keane concluded that the
important issue is whether the markets are operationally efficient or inefficient, and not
whether the markets are perfectly efficient or completely inefficient. In 1991 Keane
added that it was not important to investors whether markets are efficient or inefficient,
but rather the degree of efficiency. Keane identified four degrees of efficiency:
• Perfect efficiency. No one, not even experts, will be able to earn superior returns.
• Operational efficiency. Only a few highly skilled market participants will be able
to earn superior returns and, since the market responds quickly to the release
of their information, other investors will not be able to exploit their insights.
• Moderate inefficiency. The existence of market quirks and anomalies will allow
analysts and other market researchers to develop rule-of-thumb investment
guidelines which can also be exploited by ordinary investors.
• Operational inefficiency. Well-informed ordinary investors will be able to exploit
these inefficiencies and market experts will be able to convey their knowledge
to ordinary investors. Since the markets are unable to process the information
quickly, investors can exploit the information (Shana 1994:81; Jones 1998:276).
286
Vandell and Panino (1986)

Vandell and Panino concluded that their evidence suggest operational market efficiency
and that ordinary investors are not able to spot or exploit market inefficiencies (Jones
1998:276).
Schwert (1991)
Schwart suggested that the strong support for the reported market inefficiencies is a
result of the investment community's desire to believe that markets are inefficient. It is
in their interest that markets should be proven to be inefficient, since it can then be
shown that financial analysis is of value (Shana 1994:81 ).
5.3.1.4 Weak form efficiency tests
a. Statistical tests of independence
Harris (1986), Glosten (1989) and Cambell, Grossman and Wang (1993)
These studies examined the correlation in share returns over time. The price changes
for individual transactions on the NYSE were examined and significant serial cor-
relations were found (Reilly & Brown 1997:213).
Conrad and Kaul (1988) and Lo and MacKinley (1988)

Both these studies examined whether share prices followed a random walk, using port-
folios of shares based on size, that is, the market value calculated as the number of
shares times the share price. They found a relationship between one week's and the
next week's returns and that this correlation is stronger for small shares. These results
have cast doubt on the independence of price changes in small shares, but it was noted
that the results could have been affected by the infrequent trading of small firms' shares
(Elton & Gruber 1995:423; Reilly & Brown 1997:212,213).
287
Fama and French (1988), Poterba and Summers (1988) and Fama (1991)
Poterba and Summers used variance ratios to test whether there was zero serial corre-
lation of share returns on the NYSE for the period 1926 to 1985. They found evidence
of positive correlations for the periods under one year and evidence of negative corre-
lations for periods over two years. This evidence suggests that share prices may revert
to a mean over long periods of time. They also extended their test to examine the mean
reversion in share price indices in 17 other countries and found that most countries,
except Finland, South Africa and Spain, have negative serial correlation for long-term
returns. They examined, further, whether the mean reversion could be due to discount
rates varying over time, but concluded that their findings represent evidence against the
random wal!< hypothesis (Dobbins et al. 1994: 127, 128).
Fama and French performed a more direct test of return serial correlation. They also
found evidence of the mean reversion of returns over time and found strong negative
serial correlation for the years 1926 to 1940. However, after 1941 the correlations were
found to be close to zero (Dobbins et al. 1994:128).
Fama and French argued that, since both their and Poterba and Summers' (1988) pro-
cedures had little statistical power, the results have little weight -the correlation is much
smaller after 1940 and can be due to chance. Fama (1991) added that the results can
be a combination of changing expected returns and expected returns reverting to their
mean over time (Elton & Gruber 1995:424).
Cecchetti, Lam and Mark (1990)

Cecchetti et al. challenged these findings of mean reversion and suggested that much
of the serial correlation found in historical returns are due to a small-sample bias
288
b. Tests on technical trading rules
Brush (1986) and Pruitt and White (1988)

These studies provide support for trading rules in that technical trading rules, which use
specific three-part filters or are able to adjust for the January effect, were shown to be
able to outperform a buy-and-hold strategy (Reilly & Brown 1997:215).
Fama and French (1988,1989) and Kim, Nelson and Startz (1991)
In their 1988 study, Fama and French found that dividend yield can be used to success-
fully predict returns for periods two to four years ahead. They argued that their results
indicate some market inefficiency, as high dividend yields may indicate possible high
future returns due to current share prices being irrationally low, while low dividend
yields may indicate that current share prices are irrationally high (Dobbins et al.
1994: 129).
The 1989 study of Fama and French entailed an investigation of the relationship be-
tween business conditions and expected returns on shares and bonds. They found that
these returns move together and that dividend yield and the default spread can be used
to predict share and bond returns. Both variables predict high returns when business
conditions are weak and low returns when business conditions are strong. Their re-
gression analysis also showed that they can successfully forecast returns one to four
years ahead. Although this suggests market inefficiency and the potential for profitable
trading rules, Fama and French argued that their results are consistent with a world of
changing business conditions and that this may affect the rate used to discount antici-
pated cash flows (Dobbins et al. 1994: 129, 130).
Kim et al. (1991 ), however, showed that the statistical significance of the return fore-
casting regressions of Fama and French may be much lower than they estimated
289
Jegadeesh (1990)
Through an examination of monthly share returns over the period 1934 to 1987, the
study of Jegadeesh found a pattern in share prices and, hence, provides evidence in
support of technical analysis. It was found that shares with large losses in one month
are likely to show significant gains in the next month, while shares with large gains are
likely to show significant losses in the following month (Jones 1998:521,522).
Ball, Kothari and Wasley (1995)

The study of Ball et al. casts doubt on the possibility of successfully implementing tra-
ding rules in actual real-world conditions (Reilly & Brown 1997:215; Jones 1998:261 ).
c. Tests on market volatility and market irrationality
LeRoy and Porter (1981) and Shiller (1981,1984)

These studies examined the volatility of share prices relative to the volatility of the un-
derlying fundamental variables which affect share prices. To determine whether share
prices deviate more than the deviations in these variables would imply, volatility tests
based on the following assumptions were employed:
• Share prices reflect expectations about future dividends.
• Real expected returns on shares are constant over time.
• Dividends grow at a constant rate and are described by a stationary process.
The studies found that actual share prices vary more than the theoretical prices do and
it was concluded that markets are irrational (Elton & Gruber 1995437,438).
Flavin (1983), Kleidon (1986) and Marsh and Merton (1986)

These studies criticised earlier findings on excess volatility and the study of Marsh and
Merton also showed that if the assumption about the process of how dividends are de-
termined was changed, results opposite to those of Shiller( 1981, 1984) can be found
(Dobbins et al. 1994:126; Elton & Gruber 1995:438).
290
DeBondt and Thaler (1985, 1987) and Zarowin (1989)

From their studies, DeBondt and Thaler argued that investors overreact to unexpected
and dramatic news. They found that shares which have performed poorly are underpri-
ced due to the over-pessimism of investors and that these shares provide abnormally
good returns in subsequent periods. Similar evidence was also found of investor over-
enthusiasm in that shares which have performed well are overpriced and underperform
in subsequent periods. They showed that portfolios consisting of loser shares outper-
form portfolios consisting of winners over a three to five year period. They also found
that most of the loser portfolios' abnormal returns occur in January (Shana 1994:89;
Elton & Gruber 1995:438; Jones 1998:261 ).
Zarowin argued that DeBondt and Thaler ( 1985, 1987) overstated their conclusions. Part
of the overreaction effect can be attributed to small firms being included in the loser
portfolios and that these small firms may in any case have been expected to outperform
the market, that is, firm size explains much of the overreaction effect (Dobbins et al.
1994:116).
Brown, Harlow and Tinic (1988)

Brown et al. developed the Uncertain Information Hypothesis (UIH) to show that markets
does not overreact to unanticipated information and to explain how risk-averse investors
adjust to the release of new information. They showed that the release of imperfect
information will cause risk-averse investors to adjust share prices to match the in-
creased risk and this initial adjustment will appear to be an underreaction to good news
or an overreaction to bad news. As the uncertainty disappeared with time, risk levels will
also decrease and share returns will progressively move back toward their normal levels
(Bhana 1994:90).
Shiller (1988), West (1988) and Cochrane (1991)

These studies examined the argument that the evidence of excess volatility has been
the result of changes in discount rates which, in turn, are caused by changes in econo-
mic conditions and perceived risk. Shiller and West argued that the amount of change
291
in discount rates that would be required to explain excess volatility would be implausibly
large. Cochrane, however, disagreed with this argument and showed that small changes
in discount rates have a significant effect on share prices. He added that since the risk
premium, that is, the excess of the required rate of return over the risk-free rate, is
unknown, large fluctuations in share prices can be expected (Dobbins et al. 1994: 126).
Bulkey and Tonks (1989)

This UK study also found evidence of excess volatility and Bulkey and Tonks exploited
this with a trading rule which yielded an after-tax excess return of 1,5% per annum
Cutler, Poterba and Summers (1989)

Cutler et al. argued that important political and economic news have little predictable
effect on share prices and, hence, the effect of economic factors on share prices left
room for arguments about fads unrelated to the fundamentals which drove share prices
(Weston & Copeland 1992:98).
Barsky and De Long (1989)

In their study, Barsky and De Long note that the expectations about future dividends in
support of share price movements are not fads, but are based on expected economic
performance (Weston & Copeland 1992:98,99).
Various studies
Fama (1990) showed that more than 50% of the movement in share prices are as a
result of movements in a key fundamental - changes in industrial production. Various
other studies have examined other fundamentals, time periods and countries and
showed that there is little room for markets to be driven by irrational fads. Such studies
include those of Cochrane{1989), Barra(1990), Harris and Opler{1990) and Schwert
(1990) (Weston & Copeland 1992:99).
292
Ammer (1990), Lee, Shleifer and Thaler (1991) and Levis and Thomas (1992)
These studies showed that shares of investment trusts, both in the US and the UK, are
traded at prices that are significantly different from their net asset values per share.
These differences between market value and net asset value, both higher and lower,
vary from trust to trust and over time. They concluded that this represents evidence of
market irrationality for a specific group of shares. These shares have a given dividend
stream, but are traded at prices which differ from the individual shares' dividend streams
(Dobbins et al. 1994: 116, 117).
Baldauf and Santori (1991)

Baldauf and Santori found little evidence that markets are becoming increasingly vola-
tile, but also found that, although returns are serially uncorrelated, they are not inde-
pendent. Large changes are more likely to be followed by further large changes of
either sign, while small changes are more likely to be followed by small changes of
either sign (Upsher 1993:7).
Gerety and Mulherin (1991)

Other than for the market crash of October 1987, Gerety and Mulherin found no evi-
dence that stock market volatility has increased (Upsher 1993:4).
Haugen, Talmor and Torous (1991)

Their study entailed an investigation of share price changes with respect to volatility and
found that changes in share prices occur as a result of changes in volatility, rather than
due to adjustments in expected future cash flows. They also found apparent nonlinear
behaviour since the market reacts differently to volatility increases than to volatility de-
creases (Upsher 1993:8).
Chopra, Lakonishok and Ritter (1992)

This study provided support for the findings of DeBondt and Thaler (1985, 1987). They
used five-year periods to form portfolios and found that extreme losers outperform
extreme winners by between 5% and 10% per annum over the next five-year period.
293
Even after adjusting for the size effect and time variations in beta, the overreaction
effect was found to be much stronger for smaller firms than for larger firms. They con-
cluded that this represents evidence against market efficiency, since knowing past
share returns helps to predict future share returns (Jones 1998:261,262).
Studies on the independence of share returns have generally shown insignificant cor-
relation in share returns over time and have confirmed the independence of share price
changes over time. Some recent studies have, however, provided some evidence of
autocorrelation for portfolios of small shares and that share prices do not follow a ran-
dom walk. However, it has generally not been shown that investors can use the cor-
relation and dependence of share movements to earn abnormal risk-adjusted returns
after transaction costs (Elton & Gruber 1995:417; Reilly & Brown 1997:212,213).
Most of the evidence on technical trading rules also support the weak form EMH, since
it has been shown that these trading rules can generally not outperform a buy-and-hold
strategy (Reilly & Brown 1997:215). Dobbins et al. ( 1994: 129) note, however, that some
studies on mean reversion, excess volatility and market irrationality have suggested that
it can be possible to predict share returns and that these opportunities can be exploited
to earn excess returns.
5.3.1.5 Semi-strong form efficiency tests
a. Return prediction studies
1. Dividend yield/Macro-economic indicators
Rozeff (1984) and Shiller (1984)

Both studies examined the postulate that dividend yield is a proxy for the risk premium
on shares. Their results showed a positive relationship between aggregate dividend
yield and future share returns (Reilly & Brown 1997:217).
294
Fama and French (1988,1989)

Fama and French examined the dividend yield/share return relationship for two to four
year time horizons and found that predictive ability increases with a lengthening of the
time-horizon (Reilly & Brown 1997:217).
Various studies
A number of studies have examined the relationship between share returns, dividend
yield and two variables related to the term structure of interest rates - default spread
and horizon spread. These studies, including Keim and Stambaugh (1986), Cambell
( 1987) and Chen ( 1991), found that these variables can be used to predict share and
bond returns. Cambell (1991) even found that these variables are useful for predicting
foreign share returns (Reilly & Brown 1997:217).
Salvers, Cosimano and McDonald (1990)

The study of Ba Ivers et al. showed that share prices in efficient markets need not follow
a random walk. Long-term share returns can be somewhat predictable, as long as it is
only average returns that are predictable (Reilly & Brown 1997:217).
Pesaran and Timmermann (1995)

Pesaran and Timmermann investigated the predictive power of a number of business
cycle variables and, although finding predictive power, found predictive ability to change
over time and to vary with the volatility of share returns. The predictability of share re-
turns were found to be low during the 1960s, while it was found that exploitable pre-
dictions could have been made during the more volatile 1970s, even after accounting
for transaction costs (Reilly & Brown 1997:218).
ii. Quarterly earnings and post-earnings drift
Foster, Olsen and Shevlin (1984)

Foster et al. examined a number of possible reasons for the existence of post-earnings
drift following quarterly earnings announcements. They showed that post-earnings drift
295
could be an artifact of the model used to measure expected earnings. With different
proxies for expected earnings, different results were obtained and they concluded that,
since it is unknown which earnings expectational model is correct, the findings of their
study are inconclusive (Scott 1997: 137).
Jones, Rendleman and Latam~ (1984)

Jones et al. used a sample of over 1400 shares for 36 quarters, mid-1971 to mid-1980,
and examined daily returns before, on and after the day of the announcement, to esta-
blish the response of share prices to quarterly earnings announcements. They found
a significant relationship between unexpected earnings announcements and subse-
quent excess returns (Jones 1998:268).
Mendenhall (1986)
This study showed that share prices do not adjust as rapidly to unexpected earnings
announcements as the semi-strong form of the EMH would suggest (Reilly & Brown
1997:219,220).
Clinch and Sinclair (1987)

In a study on the Australian market, Clinch and Sinclair investigated the effect of an
earnings announcement by one firm in an industry on the share returns of competing
firms. They confirmed earlier US findings, in that the competing firms' abnormal returns
vary more when another firm in the industry announces its earnings (Henderson et al.
1992:303).
They hypothesised that the last firm in an industry to announce its earnings will gene-
rally have the smallest reaction in share price. Earlier announcements by competing
firms will probably reduce the surprise element in the last firm's announcement. Simi-
larly, the firms who make the initial earnings announcements will have the biggest
reaction, and their findings confirm this hypothesis (Henderson et al. 1992:303).
296
Bernard and Thomas (1989)

The study of Bernard and Thomas also found evidence of the existence of post-
earnings drift. They examined a large sample of shares for the period 1974 to 1986 and
showed that its existence can not be attributed to changes in the risk (as measured by
beta) of the shares following the announcement. They found, further, that transaction
costs do not limit the extent of the post-earnings drift. They did, however, show that
markets take a considerable time to correctly estimate the implications of current
earnings levels for future earnings. This represents significant evidence against capital
market efficiency (Reilly & Brown 1997:137,138).
Easton and Sinclair (1989)

This Australian study examined the effect of interim earnings announcements and found
similar evidence to US studies on quarterly earnings announcements. Their results
showed significant movements in share prices on the day of the announcement and on
the day before the announcement (Henderson et al. 1992:299).
Easton (1990)
Another Australian study by Easton examined the impact of the announcement of unex-
pected extraordinary items on share returns. Using 565 announcements of extraordinary
items, no evidence of a significant association with abnormal share returns was found
(Henderson et al. 1992 :303).
iii. Price/earnings (PIE) ratios
Basu (1983)
Following on his earlier studies, Basu re-examined the relationship between share re-
turns, the PIE ratio and size effect. For the period 1963 to 1980 it was found that low
PIE shares generally have higher risk-adjusted returns than shares with high P/E ratios.
It was also shown that P/E ratios are significant after adjustment for size, and that the
size effect virtually disappears after controlling for differences in both risk and P/E ratios
(Jones 1998:270).
297
Peavy Ill and Goodman (1983) and Goodman and Peavy Ill (1985)
Their 1983 study showed that low PIE ratio shares earn higher risk-adjusted returns
than high P/E ratio shares. The 1985 study investigated whether the P/E effect is con-
fined to low-beta shares and found this not to be the case. It is, however, clear from
these studies that no matter which risk measure is used, low PIE ratio shares provide
significant abnormal returns (Reilly & Brown 1997:223; Jones 1998:270).
Banz and Breen (1986)

Banz and Breen showed that look-ahead bias, using accounting information not yet pu-
blicly released in the data sample, had a significant effect on findings that low P/E ratio
shares have higher returns. They also found that when the look-ahead bias is removed,
return differences between high and low PIE shares are largely eliminated. They con-
cluded that the introduction of this insider information into test results casts doubt on
the representativeness of prior findings (Shana 1994:89).
Jaffe, Keim and Westerfield (1989)

This study confirmed earlier findings of low PIE ratio shares yielding positive abnormal
returns and high PIE ratio shares yielding negative abnormal returns (Shana 1994:88).
They also found that there is a correlation between P/E ratios and the size effect (Page
1996:39).
Fama and French (1991)

Fama and French argued that the PIE effect disappears after the size effect and the
book value to market value ratio have been accounted for (Elton & Gruber 1995:426).
Dreman (1994)
The study of Dreman, using a sample of 1 200 small shares for the 20-year period
ending 1993, found that low PIE shares outperform high PIE shares over long periods
of time (Jones 1998:270).
298
1v. The size effect
Christie and Hertzel (1981), Reinganum (1981) and Roll (1981)

Christie and Hertzel showed that, since betas are measured using historical data, firms
that have become small have changed their risk characteristics and the use of historical
data may not capture this increased risk. Both Roll and Reinganum showed that the
betas of small firms are underestimated, since they are traded less often and non-
synchronous trading biases beta estimates downward (Elton & Gruber 1995:425).
Brown, Kleidon and Marsh (1983)

Brown et al. generally confirmed earlier findings on the size effect, but found that the
size effect is not stable over time and that, during the 1967 to 1975 period, large firms
had outperformed small firms (Reilly & Brown 1997:225).
Reinganum ( 1983, 1992)

In 1983 Reinganum showed that the performance of the shares of small firms can not
be attributed to transaction costs, and in 1992 he confirmed the findings of Brown et al.
(1983) that the size effect is unstable. It was found that, during both the 1984 to 1987
and 1989 to 1990 periods, large firms had outperformed small firms, but he still con-
tended that the size effect is a long-term phenomenon (Van Rhijn 1994:41,42; Reilly &
Brown 1997:225).
Blume and Stambaugh (1983) and Roll (1983)

Both studies showed that if small firm portfolios are only reformed annually and not
rebalanced daily, half the size effect will be eliminated through the effect of transaction
costs (Elton & Gruber 1995:426).
Stoll and Whaley (1983)

Stoll and Whaley concluded that transaction costs had not been properly considered
in most previous studies on the size effect. It was found that there exists definite diffe-
rences in transaction costs between small lower priced shares and large higher priced
299
shares and they concluded that future studies should consider transaction costs expli-
citly and with realistic holding periods (Van Rhijn 1994:40).
Amihud and Mendleson (1986)

In their study, Amihud and Mendleson showed that part of the size effect can be attri-
buted to being compensation for illiquidity. They reasoned that small shares should
have higher expected returns due to higher transaction costs and, hence, part of the
abnormal returns can be attributed to illiquidity (Elton & Gruber 1995:426).

Dimson and Marsh showed in their UK study that the shares of small firms outperform
the shares of large firms by an average of 6% per annum. They also concluded that the
size effect should be considered in any event studies using long periods of time and
where the share sample includes the shares of both small and large firms (Pike & Neale
1996:44; Reilly & Brown 1997:226).
Keim (1986)
In two 1986 studies, Keim showed that the largest abnormal returns are yielded by firms
which have recently became small, paid no dividends or have high dividend yields, have
low prices and low P/E ratios. Keim further showed that most of the size effect occur in
January (Jones 1998:271 ).
Levis (1989)
Levis examined the London Stock Exchange for the 1961 to 1985 period and found a
size effect, although smaller than found on the US market. Levis also found the size
effect to be closely related to the dividend yield effect and the P/E effect. A number of
factors were identified for the smaller size effect reported - a different time-period had
been studied, difficulty in measuring the risk of small shares and a possible sample
selection problem (Dobbins et al. 1994: 112).
300
v. Neglected firms and trading volume
Arbel and Strebel ( 1983)

Arbel and Strebel contended that the lack of attention to certain shares also has an
effect on returns. They divided the shares in their sample into three groups according
to the attention given by analysts, that is, highly followed, moderately followed and
neglected. Their results confirm the existence of the size effect, but also provide evi-
dence of a neglected firm effect. The neglected firm effect was found to exist across all
sizes of firms and was shown to be caused by a lack of information and limited institu-
tional interest (Van Rhijn 1994:42; Reilly & Brown 1997:225,226).
James and Edmister (1983)

This study examined the effect of trading volume and investigated the relationship be-
tween returns, market volume and trading activity. They found a size effect, but did not
find that trading volume can explain the size effect (Van Rhijn 1994:42; Reilly & Brown
1997:225,226).
Barry and Brown (1984)

Barry and Brown hypothesised that, due to a lack of information, neglected firms require
higher returns. The period of listing was used as a proxy for information and, after
adjusting for the size and January effects, they found a negative relationship between
the listing period and share returns (Reilly & Brown 1997:226).
v1. The book value to market value (BV/MV) ratio
Rosenberg, Reid and Lanstein (1985)

The Rosenberg et al. study found that shares with low book value to market price ratios
significantly outperform average shares and that there exists a positive relationship
between the BV/MV ratio and future share returns (Reilly & Brown 1997:226; Jones
1998:371 ).
301
Chan and Chen (1991)

Chan and Chen argued that small firms are more risky since they have lower production
efficiency and high leverage and that this causes small firms to be marginal and more
unlikely to survive economic hardship. They concluded that the size effect is a proxy for
this more fundamental risk (Elton & Gruber 1995:425).
Chan, Hamao and Lakonishok (1991)

This study of Japanese shares confirms the positive relationship between share returns
and the BV/MV ratio. (Reilly & Brown 1997:226).
Fama and French (1992, 1995)

The 1992 study of Fama and French examined the joint effects of market beta, P/E
ratios, leverage, the size effect and BV/MV ratios and provides strong support for the
latter ratio. They examined the effect of these variables on the cross-section average
share returns for the 1963 to 1990 period. They found that the positive relationship be-
tween beta and returns had disappeared between 1963 and 1990, while leverage and
P/E ratios remain significant even after the size effect has been taken into account.
However, these variables become insignificant when the BV/MV ratio are considered
and Fama and French concluded that when the size effect and BV/MV ratios are com-
bined, all cross-sectional variation in share returns is captured. Further, the BV/MV ratio
has a consistently stronger role in explaining returns (Reilly & Brown 1997:226; Jones
1998:371 ).
Fama and French's follow-up study in 1995 examined whether the relationship between
share returns, the size effect and BV/MV ratios reflect changes in earnings. They con-
centrated on high or low BV/MV shares and their relationship with profitability (as mea-
sured by ROE). They found that the BV/MV ratio remains persistent in explaining re-
turns and that the size effect is more important in portfolios consisting of small shares,
while BV/MV is more important for portfolios consisting of shares with high BV/MV ra-
tios. This confirmed that BV/MV and firm size are the two important cross-sectional va-
riables explaining cross-sectional variations in share returns (Reilly & Brown 1997:228).
302
Lakonishok, Shleifer and Vishny (1993)

The Lakonishok et al. study entailed an investigation of returns on portfolios constructed
on the basis of BV/MV ratios. They found higher returns on high BV/MV portfolios and
concluded that this is not as a result of compensation for higher risk (Elton & Gruber
1994:426).
Fairfield (1994)
Fairfield compared the BV/MV ratio (referred to in the study as the price to book (P/B)
ratio) with the P/E ratio and found that the BV/MV ratio depends on the level of expected
future profitability, while P/E ratios depend on expected changes in future profitability.
It was also shown that BV/MV ratios are more stable than P/E ratios and that firms with
high differentials between BV/MV and P/E generally maintain their classifications (Reilly
& Brown 1997:228).
Dennis, Poterba, Snow and White (1995)

The 1992 findings of Fama and French were confirmed by this study. It was found that
optimal portfolios consist of small firm shares with high BV/MV ratios. These results per-
sisted after providing for 1% transaction costs and the assumption of annual rebalan-
cing (Reilly & Brown 1997:228).
Kothari, Shanken and Sloan (1995)

Kothari et al. examined Fama and French's 1992 results by measuring beta with annual
returns rather than monthly returns. They found a significant relationship between beta
and returns and suggested that the relationship between BV/MV and returns may be
periodic and insignificant over long periods (Reilly & Brown 1997:227).
vii. The January effect
Gultekin and Gultekin (1983)

This study examined the existence of the January effect in 17 countries and found that
returns are higher in January for all the countries studied (Elton & Gruber 1995:412).
303
Reinganum (1983)
Reinganum showed that shares which had declined substantially in December show
excess returns in January. He attributed most of this effect to tax-loss selling in De-
cember, but still found some January effect for those shares which had shown gains in
December (Elton & Gruber 1995:413).
Lakonishok and Smidt (1984,1986)

Both studies found a January effect in the trading volume of small shares. The last day
of the year was shown to be the most active trading day and the abnormal trading
activity was found to continue into January (Reilly & Brown 1997:221 ).
Kato and Shallheim (1985)

Kato and Shallheim examined the relationship between firm size and the January effect
for the Japanese market. They found a strong relationship between January excess re-
turns and size excess returns, but no relationship between firm size and returns in non-
January months (Elton & Gruber 1995:412).
Keim ( 1985, 1986, 1989)

The 1985 and 1986 studies examined the relationship between share returns and divi-
dend yields and found a nonlinear relationship in January. It was also confirmed that
this relationship only exists in January and that a size effect exists in January (Reilly &
Brown 1997:221; Jones 1998:272). In the 1989 study it was shown that most small
shares trade at their bid price on the last trading day of December and at the beginning
of January. As small shares generally tend to have bigger bid-ask spreads and lower
prices, Keim concluded that these findings provide a partial explanation of the January
differences in returns (Fuller 1993:28; Elton & Gruber 1995:413).
Corhay, Hawawini and Michel (1987)

In their analysis of share returns on the Brussels, London, New York and Paris stock
exchanges, Corhay et al. found a January effect on all the exchanges (Shana 1994:87).
304
Jones, Pearce and Wilson (1987)

Jones et al. examined the hypothesis that tax-selling might explain the January effect
and found evidence that refute this hypothesis. They found that in the 1821 to 1917
period, before income taxes had existed, a January effect had existed which is similar
to the effect after the introduction of income taxes. They also found that, although
Belgium and Japan has no capital gains taxes, the January effect still exists on these
markets. For Australia, which has a non-December tax year, excess returns are found
in January (Reilly & Brown 1997:413).
Fama (1991)
Fama examined the January effect and its relationship with the size effect for two
periods, 1941 to 1981 and 1982 to January 1991. It was found that the January effect
is closely related to the size effect and that most of it occurs during the first few days of
January. It was also found that the size effect had been more pronounced during the
1941to1981 period than in the second period (Elton & Gruber 1995:411,412).
Arnott, Dorian and Macedo (1992)

Arnott et al. found evidence that suggests that the size effect and the link between the
size effect and the January effect have disappeared. However, they still found a
significant January effect, in that shares which had performed poorly recover signi-
ficantly in January, while those shares which had performed exceptionally drop back in
January. They further found that the year-end BV/MV effect remains strong (Arnott
1993: 19).
Bhardaj and Brooks ( 1992)

This study provided evidence that previous studies on the relationship between the size
effect and the January effect have been biased by incomplete consideration of a price
effect and transaction costs. They examined share returns for the 1967 to 1986 period
and found that the January effect is not so much a small firm effect, but rather a low-
price phenomenon. Further tests showed that when transaction costs are considered,
portfolios of low priced shares almost always perform worse than the market portfolio.
305
They concluded that excess January returns on low priced shares can be explained by
a bid-ask bias and higher transaction costs and, further, that the January anomaly is not
persistent and unlikely to be exploited by most investors (Jones 1998:273).
Haugen and Jorion (1996)

Haugen and Jorion found that the January effect still exists on the NYSE, that its magni-
tude has not changed significantly and that there is no evidence that it has disappeared
(Jones 1998:272,273).
Various studies
Ritter (1988) suggested that portfolio-rebalancing affected returns in January, since it
was found that a large number of small investors sell small shares in early December
and buy small shares in late December and early January. The study of Jones, Lee
and Apenbrink (1991) showed that tax-loss selling does not account for part of the
January effect, while Dyl and Maberly (1992) found that odd-lot sales of shares follow
a similar pattern to that reported by Ritter (1988), and that this can not entirely be attri-
buted to tax-loss selling (Fuller 1993:28).
viii. Other calendar effects
Various studies
French (1980) reported a weekend effect. Rogalski (1984) found that the weekend
effect can be decomposed into two effects, Friday close of trading to Monday opening
and a Monday trading effect. Smirlok and Stacks (1986) found evidence of a time of
the day effect (Reilly & Brown 1997:222), while the studies of Keim and Stambaugh
(1984) and Keim (1989) confirmed the existence of a weekend effect (Elton & Gruber
1995:411 ). The study of Ball and Bowers (1988) provide evidence of the existence of
a day of the week effect (Dobbins et al. 1994: 115).
306
Gibbons and Hess (1981)

Gibbons and Hess examined the period 1962 to 1978 and found negative share returns
for Mondays. They also reported large positive returns on Wednesdays and Fridays
(Elton & Gruber 1995:411 ).
Harris (1986)
Intra day and day of the week patterns for the period December 1981 to January 1983
were examined and Harris found large negative returns for Mondays, while the other
four days of the week yield positive returns which are largely equal. Harris also found
that half of Monday's negative returns occurs between Friday's close and Monday's
opening of trade, while the other half occurs during the first 45 minutes of Monday's
trading. Hereafter Monday's returns resemble those of the other days of the week. It
was also shown that share prices tend to rise during the last half hour of trading of each
day (Elton & Gruber 1995:411 ).
Ariel (1987)
Ariel's study found significant time of the month effects, whereby most of the market's
cumulative advances occur during the first half of the month's trading (Reilly & Brown
1997:222).
Kim (1988)
Kim found a persistent trend of Monday's returns being low or negative, while those of
Friday's tends to be positive and high. Kim also reported that this effect is evident on
most, if not all, stock exchanges of developed countries and also on smaller markets
like those of Hong Kong, Korea and Taiwan (Shana 1994:87,88).
Lakonishok and Smidt (1988)

The study of Lakonishok and Smidt used 90 years of daily data and confirms the exis-
tence of anomalies in the pricing of shares around holidays, turn of the week, turn of the
month and at the end of the year (Dobbins et al. 1994: 115).
307
Connolly (1989)
Connolly used superior technology and corrected for the statistical flaws in previous
studies of the weekend effect on the NYSE. He concluded that the Monday effect is very
weak and had actually disappeared during the 1980s (Shana 1994:88).
b. Event studies
i. Share-splits
Grinblatt, Masulis and Titman (1984)

This study examined share price reactions to a sample of share-splits and reported po-
sitive higher returns for shareholders on the day of the announcement and for several
days beyond the announcement date (Reilly & Brown 1997:230; Jones 1998:263).
ii. Stock exchange listings
Sanger and McConnell (1986) and McConnell and Sanger (1987, 1989)
All three studies provided evidence of potential profit opportunities immediately after the
announcement of applications for listings and possible excess returns immediately after
the actual listing (Reilly & Brown 1997:232).
Dharan and Ikenberry (1995)

Dharan and Ikenberry confirmed that share prices decline after the initial listing and that
this provides some opportunities for abnormal profits. They did, however, show that the
degree of decline depends on the size of the firm and that the largest results are yielded
by small firms which list prior to a decline in performance (Reilly & Brown 1997:232).
308
iii. Initial public offerings
Ritter (1983,1991) and Levis (1993)

In his 1983 study, Ritter showed that initial public offerings do not provide long-term
abnormal returns and generally substantially underperform, on a risk-adjusted basis,
after the first day (Elton& Gruber 1995:434). The 1991 study confirmed this conclusion
and Ritter showed that investing in initial public offerings on the first day of public
trading and holding the investment for three years, results in a wealth of only 83% of
what can be attained by investing in a portfolio of similar shares on the NYSE or AMEX.
The 1993 study of Levis on the UK market yielded findings which confirm the results of
Ritter's studies (Reilly & Brown 1997:232).
Various studies
The studies of Miller and Reilly (1987), Chalk and Peavy Ill and Hanley (1993) have
shown that the market corrects the initial underpricing of new share issues within one
day after the initial offering. They concluded that initial public offerings are generally
underpriced, but that only a few of the investors who are allocated shares in the original
issue will be able to take advantage of the initial underpricing. Hanley and Wilhelm
(1995) showed that institutional investors are able to capture most of the initial profits
(Reilly & Brown 1997:230).
Ibbotson, Sindelar and Ritter (1988)

Ibbotson et al. reviewed some of the earlier studies on initial public offerings and con-
cluded that new share issues are generally underpriced. The evidence show that
investors who are able to purchase the new issues can earn abnormal returns. It was
also shown that the market adjusts prices quickly and that investors who buy shortly
after the offering will not be able to earn excess returns (Jones 1998:263).
Hanley, Lee and Seguin (1996)

The study of Hanley et al. examined the behaviour of market prices and trading volume
following the initial public offerings of 65 closed-end funds. Their study included nearly
309
all the new closed-end funds on the NYSE and AMEX during the period January 1988
to May 1989. They found that the shares of the newly issued closed-end funds drop in
price and yield negative returns during the first few months of trading. However, upon
further investigation they found that the decline in share prices only starts 30 days or
more after the issues and that the initial drop is usually significant (Scott 1997: 141 ).
Their findings also suggest that large institutional investors start selling large number
of shares immediately after trading starts. The reason for the prices not beginning to
decline immediately seems to be due to price management by the underwriters of the
new issues. The underwriters over-allot the initial public offerings to small investors and
use the sales by the large institutional investors to fill the over-allotted orders and,
hence, supply and demand is kept in balance during the early days of trading. In this
way the underwriters maintain their reputations with large investors and, with the time-
delay, it seems unlikely that their reputations would be significantly tarnished in the
eyes of small investors (Scott 1997: 141 ).
This suggests that there are different classes of investors with different degrees of mar-
ket power and market prices can, therefore, be influenced by other factors than supply
and demand. In addition, it also casts doubt on the hypothesis that market prices fully
reflect all available information. As the poor performance of closed-end funds during the
months following the initial issue is a matter of public record, it would seem rational that
small investors would wait three to four months before investing. However, they do not
and this fact represents another anomaly of market efficiency (Scott 1997: 141, 142).
1v. Corporate unbundling and spin-offs
Miles and Rosenfeld (1983)

These researchers argued that, in efficient markets, spin-off announcements should not
effect the value of a firm, unless the spin-off is expected to affect future cash flows.
They did, however, find that the market has a systematic positive reaction to spin-offs
and unbundling -their study showed that 14% positive abnormal returns are associated
310
with spin-off announcements (Blount & Davidson 1996:67,68).
Hite and Owens (1983)

Hite and Owens, too, showed that positive abnormal returns are associated with spin-off
announcements. They found 7% abnormal returns for the period 50 days prior to the
announcement date until completion of the spin-off. They argued that these results are
caused by the spin-offs and unbundling, allowing the parent firms and subsidiaries to
concentrate on activities, wherein they would each have their respective comparative
advantages or could not enter into as a conglomerate (Blount & Davidson 1996:68).
Cusatis, Miles and Woolridge (1993,1994)

Both studies showed that the abnormal returns continue on a long-term basis and that
the market's reaction to these announcements confirms investor expectations about the
advantages of corporate downsizing. These studies measured the share returns for the
spin-off firms, parent firms and parent-spin-off combinations and found that the positive
abnormal returns last for the three years following the announcement date. They con-
cluded that these positive results are due to the spin-offs eliminating the inefficiencies
associated with a lack of strategic fit or synergy between the parentfirm and its sub-
sidiaries (Blount & Davidson 1996:68).
v. Accounting changes
Brennan (1995)
Brennan reviewed the numerous studies on the impact of accounting changes on share
prices. He concluded that these studies show that the markets react quickly to accoun-
ting changes and adjusts share prices to their true value, that is, the shares are valued
on the basis of economic events and not on the basis of cosmetic accounting informa-
tion (Reilly & Brown 1997:233).
311
vi. Corporate events (mergers. acquisitions. etcetera)
Smith (1986) and Jensen and Warner (1988)

Both these studies reviewed the earlier studies on corporate events. The evidence indi-
cates that market prices are adjusted on the basis of the underlying economic impact
of the events and that the changes in prices are quite rapid, that is, usually completed
within three days of the announcement (Reilly & Brown 1997:234).
Agrawal, Jaffe and Mandelker (1990)

Agrawal et al. re-examined the earlier study of Jaffe and Mandelker in 1976 and found
similar results. They showed that the post-merger performance of the acquiring firms
anomaly (post-announcement drift) is still significant and that firms which acquire other
firms show abnormal returns that generally last over the next five years (Elton & Gruber
1995:434).
Healy, Palepu and Ruback (1990)

This study showed that the reaction of stock markets to takeover announcements are
a good indicator of the operating performance of the companies in the period after the
takeover had taken place (Weston & Copeland 1992:100).
vii. Unexpected economic news or world events
Pearce and Roley (1985)

Pearce and Roley examined the response of share prices to announcements of money
supply, real economic activity, inflation and the Federal Reserve's discount rate. They
found evidence in support of semi-strong market efficiency, as it was shown that the
impact on share prices does not persist beyond the announcement day (Reilly & Brown
1997:233; Jones 1998:264).
312
Jain (1988)
In an even more in-depth analysis, Jain examined the hourly response of share prices
to surprise announcements of money supply, prices, industrial production and unem-
ployment rate. It was found that only money supply and prices have a significant effect
on share prices and that the effect is reflected in share prices within one hour (Reilly &
Brown 1997:233; Jones 1998:264).
Pound and Zeckhauser (1990)

Pound and Zeckhauser examined the effect of public takeover rumours as published in
the 'Heard in the Street' column in The Wall Street Journal. They found the market to
be efficient at responding to the analysts' opinions on shares and the features on the
companies published in the column. On average no abnormal returns can be earned on
the day of the publication, although it was shown that abnormal returns do occur in the
calender month before publication (Jones 1998:264).
v111. Non-earnings-related information
Ou and Penman (1989)

Ou and Penman investigated whether non-earnings-related financial statement informa-
tion can be used to devise an investment strategy whereby abnormal returns can be
earned. They used a large sample of firms and applied 68 different financial ratios for
the years 1965 to 1972 with the aim of establishing which ratios best predict increases
and decreases in net income for the next year. They used the 16 best ratios to estimate
a multivariate regression model and applied the model to predict the earnings changes
of their sample firms for the period 1973 to 1983. They used these predictions to deve-
lop the following investment strategy. They bought the shares, for each year and each
firm, three months after financial year-end in cases where increases in net income were
predicted and sold short the shares of the firms with negative changes in net income
predictions. The shares bought were held for two years and then sold at the ruling mar-
ket price, while those sold short were bought after two years at the ruling market price
(Scott 1997: 138, 139).
313
The results of their study showed that the strategy yields significant risk-adjusted ex-
cess returns (14,53% in excess of market returns) before transaction costs. They con-
cluded that transaction costs can not fully explain these abnormal returns. Their findings
represent another anomaly and evidence against market efficiency, since it appears as
if all financial information is not fully incorporated into share prices. It also seems as if
the market only adjusts prices with the actual announcement of the next two years'
earnings changes and by then their investment strategy had already yielded its excess
returns (Scott 1997: 139, 140).
Tinic (1990) and Ball (1992)

Both studies argued that Ou and Penman's 1989 results might be due to misspecifica-
tion of the CAPM, that is, beta does not fully capture all the factors determining expec-
ted returns. It is thus incorrect to conclude that markets are inefficient (Scott 1997: 140).
Greig (1992)
Greig re-examined Ou and Penman's 1989 study and concluded that the reported
excess returns are not due to market inefficiency in processing accounting information,
but rather due to the effect of firm size on expected returns. By controlling the results
of Ou and Penman (1989) for firm size, it was found that all excess returns are elimi-
nated. He also showed that Holthausen and Larcker's 1992 findings might be explained
by firm size (Scott 1997: 140).
Holthausen and Larcker (1992)

In their study, Holthausen and Larcker also showed that excess returns can be earned
with the use of a financial statement ratio-based investment strategy. However, when
they used Ou and Penman's (1989) strategy over the 1978 to 1988 period, they found
no excess returns and concluded that this suggests that the earlier results may be
restricted to the time period examined (Scott 1997: 140).
314
Stober (1992)
Stober also applied Ou and Penman's (1989) investment strategy and confirmed their
findings of excess returns. Stober, however, found that the excess returns last up to six
years and that it seems unlikely that it would take the market that long to adjust share
prices in line with their fundamental values. This suggests that Ou and Penman's results
may have been due to some permanent differences in expected returns, that is, firm
size or risk, rather than to the inefficient incorporation of financial information into share
prices (Scott 1997: 140).
Lev and Thiagarajan (1993)

According to Rees (1995:145-147), Lev and Thiagarajan used a more recent sample
for their study, 1974 to 1988, and differed from Ou and Penman ( 1989) in that they did
not base their investment strategy model on all available line item data, but used indi-
cators of high-quality earnings employed by analysts. These indicators included, among
others, items such as:
• Increases in inventory at a rate greater than sales increases.
• Cutbacks in capital expenditure and research and development.
• Disproportionate increases in debtors and provisions for doubtful debts.
• Disproportionate drops in gross profit and changes in selling and administrative
expenses.
• Changes in the effective rate of taxation.
• Changes in the workforce and qualifications of the audit report.
They found that most of these variables are statistically significant when share returns
are modelled over the accounting year. Their model also showed that firms with high-
quality earnings tend to have greater increases in earnings than those with low quality
earnings. They concluded, similar to Ou and Penman ( 1989), that fundamental analysis
of financial statements can contribute to superior earnings forecasts.
Reilly and Brown (1997:234) conclude that the evidence on the semi-strong form of the
EMH is mixed. The results of most of the studies on events such as share-splits, initial
315
public offerings, accounting changes, corporate events and unexpected economic news
or world events tend to provide strong support for semi-strong market efficiency. It is
only event studies on stock exchange listings and non-earnings-related information
which yield some conflicting results. Return prediction studies, on the other hand, pro-
vide much evidence in conflict with semi-strong form market efficiency. The time-series
studies on dividend yield, macro-economic indicators, quarterly earnings announce-
ments and calendar patterns, and the cross-sectional return prediction studies on P/E
ratios, the size effect, neglected firms, trading volume and the BV/MV ratio have yielded
anomalous results.
Elton and Gruber (1995:424) provide five possible explanations for the anomalous
results which suggest the existence of a positive relationship between excess returns
and the characteristics of firms.
• Firstly, that the observed observations are not real and that the tests overstate
the significance of the statistical relationships.
• Secondly, that the firm characteristics are proxies for risk variables which are not
included in the asset pricing models (CAPM or APT).
• Thirdly, that the CAPM is misspecified and, hence, betas are underestimated for
small firms.
• The fourth explanation addresses the continued existence of these anomalies,
in that trading costs preclude the profitable exploitation thereof.
• The fifth and final explanation is that markets are inefficient.
The consistent evidence of post-earnings drift suggests that quarterly earnings sur-
prises can be used to earn excess returns, and the importance of this anomaly led The
Wall Street Journal to publish a section on 'earnings surprises' in quarterly earnings
reports (Reilly & Brown 1997:220).
Reilly and Brown (1997:221,222) also note that the other calendar effects, for example
a monthly effect, weekend effect, day-of-the-week effect and intra day effect, are not as
significant as the January effect. They find the January anomaly to be fascinating due
316
to its pervasiveness and close relationship with the size effect. They add that explana-
tions for the January effect include that of a dividend yield effect, a trading volume effect
and tax-loss selling, but that these only receive mixed support. They conclude by stating
that the numerous studies on the January effect only succeed in producing as many
questions as answers about this anomaly.
5.3.1.6 Strong form efficiency tests
1. Insider trading by corporate insiders
Kerr (1980)
Kerr tested the trading rule identified in the earlier study of Jaffe (1974) and concluded
that the market had eliminated the inefficiency (Reilly & Brown 1997:235).
Nunn, Madden and Gombola (1983)

The Nunn et al. study confirm the 1974 findings of Jaffe and found that corporate in-
siders are able to consistently earn excess returns on their share transactions (Jones
1998:265).
Givonly and Patmon (1985)

In a review of the studies on insider trading in the US, the researchers found that the
evidence generally show that corporate insiders are able to outperform the market, in
particular when a large number of insiders trade in a specific share (Shana 1994:85).
Rozeff and Zaman (1988)

This study found that the investors who act on publicly available information on insider
trading are also able to earn abnormal profits. However, they showed that these ab-
normal profits are substantially reduced when the size effect and the PIE ratio effect are
taken into account and completely eliminated when transaction costs are included. They
also showed that the excess returns earned by the corporate insiders are substantially
reduced with the inclusion of transaction costs. In such cases abnormal returns average
317
only 3% to 3,5% per annum (Jones 1998:265).
Peers (1992)
Peers examined insider trading by chairpersons, presidents and other top officials of
companies over the 1975 to 1989 period. He found that they are able to earn substantial
abnormal profits, especially on large trades. Most other insiders are only able to perform
slightly better than the market return (Jones 1998:265).
Pettit and Venkatesh (1995)

The Pettit and Venkatesh study confirm the significant relationship between insider
trading and longer-term share performance (Reilly & Brown 1997:235).
Various studies
Seyhun (1986) supported the principle that it should be considered which group of in-
siders (chairpersons, directors, top officials or other insiders) are doing the trading. The
study of Lee and Solt (1986) show that it is generally not possible to use the trading
activities of insiders as the basis of a market timing strategy, and the studies of Seyhun
(1988) and Chowdhury, Howe and Lin (1993) provide support for their conclusion
(Reilly & Brown 1997:235,236).
11. The performance of analysts
Dimson and Marsh (1984,1988)

In their 1984 study, Dimson and Marsh investigated the forecasting ability of 35 share
analyst firms. They found a small degree of forecasting ability from the 47 000 recom-
mendations examined, but also that these recommendations do not outperform the mar-
ket when transaction costs are included. They also showed that past forecasting ability
does not necessarily mean future forecasting ability and that investors, therefore, can-
not predict which analysts will be able to provide superior forecasts in the future (Shana
1994:91,92; Elton & Gruber 1995:435).
318
Dimson and Marsh did, however, show that combining the analysts' forecasts leads to
improved predictions of the future, but also found that more than half of the information
content of these forecasts are incorporated in share prices within the first month after
the forecast. This means that investors need to act quickly on these recommendations
(Elton & Gruber 1995:435). Their 1988 study confirmed that differences in forecasting
ability between analysts do not seem to persist over time, that is, no correlation between
long-term and short-term forecasting ability could be found (Shana 1994:93).
Stickel (1985), Hall and Tsay (1988) and Huberman and Kandel (1990)
Stickel used event-study methods to evaluate the performance of the Value Line invest-
ment strategy, whereby Value Line publishes weekly share rankings by dividing shares
into five groups - group one being the shares with the best prospects and group five
being the shares with the worst (Shana 1994:92; Elton & Gruber 1995:436). The study
revealed an announcement effect in ranking changes, which suggests that Value Line
has information that is not fully reflected in share prices. It takes the market up to three
days to adjust prices to changes in rankings and these price changes are of a perma-
nent nature. This evidence suggests either access to additional information not fully in-
corporated into share prices or superior forecasting ability. However, it was also noted
that the price effects of the changes in rankings are too small to outperform the market
after transaction costs (Shana 1994:92; Elton & Gruber 1995:436; Jones 1998:273).
Huberman and Kandel confirmed Stickel's (1985) results (Jones 1998:274), while Hall
and Tsay showed that the superior performance of Value Line has diminished in recent
·years, to a level whereby the returns barely cover the transaction costs of a buy-and-
hold strategy (Shana 1994:92).
Elton, Gruber and Grossman (1986)

Elton et al. examined the database of a large bank which also ranks shares into five
groups, namely best buys, buys, holds and two classes of sells. It was found that both
the classification and changes in classification have information content and that
abnormal risk-adjusted returns can be earned by buying upgraded shares or shares with
319
a high ranking, or selling downgraded shares or shares with a low ranking. The excess
returns were found to exist in the month of the ranking and for two months after the
ranking or change in ranking. However, no superior forecasting could be identified and
larger excess returns are associated with acting on changes in rankings than with acting
on the recommendations themselves (Elton & Gruber 1995:435,436).
Hulbert (1990)
Hulbert provided further evidence on the Value Line rankings and showed that the sys-
tem has performed less well after 1983. In addition, the evidence suggest that these an-
nouncements do not yield abnormal returns after the consideration of transaction costs,
as was evidenced by the fact that the Value Line Centurion Fund consistently under-
performed the market during the 1980s (Reilly & Brown 1997:238).
Affleck-Graves and Mendenhall (1992)

These researchers argued that abnormal returns associated with the Value Line ran-
kings are not due to superior ability, but are caused by the post-earnings drift asso-
ciated with the quarterly earnings announcement anomaly (Reilly & Brown 1997:238).
Various studies
Both Lloyd-Davies and Canes (1987) and Lin, Smith and Syed (1990) suggested that
analysts possess private information. They found that the prices of shares mentioned
in the 'Heard on the Street' column of The Wall Street Journal change significantly on
the day of publication. Opposite to these findings, Desai and Jain (1995) examined the
recommendations of the money managers at Barron's round table and found no abnor-
mal returns after the publication date, although abnormal returns do appear in the pe-
riod between the meeting date and the date of publication. Womack (1996) analysed
analysts' recommendations and found that they have both market timing and share
selection ability (Reilly & Brown 1997:238,239).
320
iii. The performance of professional money managers
Various studies
Shukla and Trzcinka (1994) showed that mutual funds perform inconsistently and that
inferior performance is their only consistency. These findings are supported by the stu-
dies of Henriksson (1984) and Chang and Lewellen (1984). The studies of Munnell
(1983) and Brinson, Hood and Beebower (1986) also found that pension funds per-
form worse than the aggregate market, while Berkowitz, Finney and Logue (1988)
showed that the performance of endowment funds are no better than a buy-and-hold
strategy (Reilly & Brown 1997:239,240).
Lehman and Modest (1987)

Although Lehman and Modest reported significant differences between the various
benchmarks (market indices) used to measure performance, they concluded that mutual
funds have on overage performed consistently worse than the overall market (Reilly &
Brown 1997:977).
Grinblatt and Titman (1989,1993) and Connor and Korajczyk (1991)

In their 1993 study, Grinblatt and Titman examined the performance of money mana-
gers, using portfolio holdings which do not require a market benchmark. They found that
money managers of growth shares had earned significant positive risk-adjusted returns
during the 1976 to 1985 period (Reilly & Brown 1997:977). This confirmed their 1989
findings and those of Connor and Korajczyk (1991) regarding superior performance
Ippolito (1989)
Ippolito found that mutual funds are able to outperform the market on a risk-adjusted
basis. However, he also showed that once transaction costs and management fees are
considered, all evidence of superior performance disappear (Bhana 1994:86).
321
Cumby and Glen (1990)

Cumby and Glen measured the performance of international funds against a US index
and the Morgan Stanley world equity index. They used two risk-adjusted performance
measures and found no evidence that these international funds are able to provide
superior results (Reilly & Brown 1997:977).
Elton, Gruber, Das and Aklarka (1991, 1994)

In their 1994 study, Elton et al. used a three-factor model to measure the risk during the
period studied by Ippolito in 1989 and found that the abnormal returns are actually
negative with the use of more extensive risk measures (Reilly & Brown 1997:239,240).
Their 1991 study confirmed that mutual fund managers are generally unable to perform
well enough to cover their management fees and expenses (Elton & Gruber 1995:437).
1v. Persistence of performance
Ang and Chua (1982)

Ang and Chua examined the consistency of funds with different objectives and found
that, although the funds generally meet their objectives, they are not able to do so on
a consistent basis (Reilly & Brown 1997:979).
Dunn and Theisen (1983)

This study confirmed earlier evidence that a fund's past performance should not be
used to predict future performance. They examined institutional portfolios over a ten-
year period and found no consistency in performance and concluded that past results
do not explain future results (Reilly & Brown 1997:979).
Grinblatt and Titman (1992)

Grinblatt and Titman, however, found that differences between the performance of va-
rious funds persist over time and, hence, past performance does provide useful infor-
mation about future performance (Reilly & Brown 1997:979).
322
Hendricks, Patel and Zeckhauser (1993)

The persistence of performance was confirmed in their study, especially over a short-
term one-year evaluation period (Reilly & Brown 1997:979).
Bauman and Miller (1994)

Rather than examine calender periods, Bauman and Miller used full market cycles to
predict performance rankings. They found that there is a positive and meaningful corre-
lation between the portfolio performance rankings of one cycle with the next (Reilly &
Brown 1997:979).
Brown and Goetzmann (1995)

Brown and Goetzmann examined the persistence of performance, taking into account
any survivorship bias, and found evidence confirming performance persistence. They
did, however, acknowledge that their results were dependent on the time period of the
study and the correlation of performance of successful funds with funds having similar
themes (Reilly & Brown 1997:979).
Malkiel (1995)
Malkiel confirmed the conclusions of Brown and Goetzmann (1995) regarding the time
period studied, since he found evidence of performance persistence in the 1970s, but
no such evidence for the 1980s (Reilly & Brown 1997:980; Jones 1998:73). Malkiel
examined all diversified US equity funds for the 1971 to 1991 period and showed that
the top 40 funds generally have a year-to-year pattern in their success rates. He con-
cluded that past performance is no guarantee offuture performance, although there are
some funds which had outperformed the market over the 20-year period of the study
(Jones 1998:72, 73).
Clements (1996) and Droms and Walker (s.a.)

The study of Drams and Walker, as reported by Clements in 1996, examined the perfor-
mance of 151 funds which had existed over the entire 20-year study period, ending in
1990. It was found that only 40 of these funds had outperformed the market in more
323
than 10 of the years, while none of the funds had been able to outperform in all of the
five-year subperiods examined. In addition, no evidence was found that funds which had
performed well in the first 1O years were more likely than other funds to do so in the next
10 years (Jones 1998:74).
Elton, Gruber and Blake (1996) and Gruber (1996)

Both these studies provided evidence in support of the persistence of performance
between different funds (Reilly & Brown 1997:979; Jones 1998:72).
v. Market timing and share selection ability
Kon and Jen (1979)

Kon and Jen examined the ability of mutual funds to select undervalued shares and to
time the market. Although they found evidence of market timing ability, none of the
funds examined yielded consistent superior returns (Reilly & Brown 1997:978).
Shawky (1982) and Veit and Cheney (1982)

Shawky found that mutual funds are generally able to alter their risk characteristics in
line with their market timing ability and that they are able to improve the diversification
of their portfolios. Despite this, most funds still generate inferior results (Reilly & Brown
1997:978). Veit and Cheney, on the other hand, examined 74 mutual funds and con-
cluded that the funds are not able to change their risk characteristics in line with their
market timing strategies (Reilly & Brown 1997:978; Jones 1998:397).
Chang and Lewellen (1984)

This study found little evidence of successful market forecasting. Chang and Lewellen
concluded that mutual funds show no special market timing or share selection ability
and, hence, mutual funds are not be able to outperform a buy-and-hold strategy (Reilly
& Brown 1997:978; Jones 1998:397).
324
Henriksson ( 1984)
Henriksson examined 116 mutual funds and found no evidence of consistent share
selection and market timing ability. The study showed that managers are not able to
forecast large changes in the market and that those who show superior share selection
ability have poor market timing ability (Reilly & Brown 1997:978; Jones 1998:397).
Alexander and Stover (1990)

This study confirmed the earlier findings that fund managers have poor market timing
ability (Brevis 1998:4).
Lee and Rhaman (1990)

From their investigation of mutual fund performance, Lee and Rhaman concluded that
some funds show superior share selection and market timing ability. The study covered
the performance of 93 funds over the period January 1977 to March 1984. It was found
that 15% of the fund managers showed significant positive selection ability, 10,8% ne-
gative selection ability and 17,2% showed significant positive market timing ability. Their
study ignored negative timing ability (Bradfield 1998: 11 ).
Chan and Chen (1992)

Chan and Chen examined the market timing and share selection performance of asset
allocation funds. They found no evidence of any special abilities and, in fact, showed
that the funds generally perform worse than their portfolio benchmark (Reilly & Brown
1997:978,979).
Wagner, Shellans and Paul (1992)

The Wagner et al. study compared the market timing performance of 25 mutual funds
against a buy-and-hold strategy, whereby dividends were re-invested in three-month
Treasury bills. Their examination covered the January 1985 to November 1990 period
and found that the risk-adjusted returns of their market timing strategies yielded superior
results to that of a buy-and-hold strategy. They found that 92% of the fund managers
had outperformed the market during the crash of October 1987 (Brevis 1998:34,35).
325
Pond (1994)
Pond showed that there was considerable risk attached to market timing and if investors
are not in the market at critical times, their returns would be significantly reduced (Jones
1998:397).
Hulbert (1996)
The results of Hulbert's study showed that only 3% of share timing strategies used over
the most recent five-year and eight-year periods had outperformed a buy-and-hold
strategy on a pure market timing basis (Jones 1998:396).
Bello and Janjigian (1997)

Bello and Janjigian examined the performance of investment funds which follow an
active investment strategy. Their results show that these funds have a statistically
significant market timing ability (Brevis 1998:36).
Reilly and Brown ( 1997:241) conclude that the results from these studies on strong form
market efficiency provide mixed support for the EMH. The evidence provided by studies
on insider trading clearly provides no support for the hypothesis, while those on stock
exchange specialists indicate that some professionals do have monopolistic access to
important information. They are, therefore, able to earn excess returns on some con-
sistent basis. Against that, the evidence on the performance of professional money
managers provides support for strong form market efficiency. The evidence on their
inability to consistently outperform a buy-and-hold strategy. their lack of persistent
performance and their poor market timing abi Iity provide s1gnif 1cant support for the EMH.
As both professional money managers and average investors do not generally have
access to inside information, the results of these studies are quite supportive of the
strong form EMH as applied to investors in general.
326
5.3.1. 7 South African research
a. General research
Seneque ( 1987)
Seneque reviewed the efficiency of the JSE and concluded that the issue is not to es-
tablish whether it is efficient or inefficient, but rather to establish the degree of efficien-
cy. He noted that the fact that the JSE is a small market with relatively few listed shares
and, more importantly, the fact that many shares are closely held and thinly traded com-
plicates the issue of establishing the degree of efficiency of the JSE (Van Rhijn 1994:7).
Bradfield (1989)
Bradfield concluded that the problem of thin trading is probably the main cause of diffi-
culties experienced with risk measurement in small markets and concluded that this is
also the cause of the problems experienced with estimating beta on the JSE (Van Rhijn
1994:43).
Shana (1994)
In a review of the efficiency of the JSE, Shana concluded that only a few professional
investors are able to achieve superior performance and that only company officers tra-
ding on inside information are able to consistently outperform the market. Some evi-
dence of calender effects and the PIE effect was found, but the magnitude of the abnor-
mal price behaviour was found to be small. Shana concluded, further, that the JSE can
therefore be regarded as an oper~tionally efficient market (Shana 1994:93-95).
b. Weak form tests
1. Independence of share price changes
Brummer and Jacobs (1981)

Brummer and Jacobs examined the pattern of share price changes on the JSE and
327
found that, although price changes are not completely random, the degree of depen-
dence is marginal and would not enable investors to consistently outperform the market
(Shana 1994:83).
Van Rhijn (1994)

The study of Van Rhijn used the industrial sector of the JSE to examine weak form mar-
ket efficiency. The industrial sector was divided into two groups, a group consisting of
companies which traded less than 250 000 shares per annum and, to rule out thin tra-
ding, a second group of companies with share trading of more than 250 000 per annum.
Van Rhijn concluded that the JSE is not weak form efficient for individual shares, but
confirmed weak form efficiency for portfolios. The study also showed that a thin trading
effect exists on the JSE (Van Rhijn 1994:9, 10, 175, 176,242,243,340).
ii. Market irrationality
Shana (1989, 1993)

Shana found in 1989 that, during the1970 to 1984 period, the market had overreacted
to unexpected favourable or unfavourable company-specific events. In the 1993 study,
it was shown that the market had overreacted to earnings announcements during the
1975 to 1989 period. These results indicate that positive abnormal returns can be
earned in the year following negative earnings announcements (Shana 1994:89).
Page and May (1992)

Page and May replicated the study of DeBondt and Thaler (1985, 1987) to test for mar-
ket overreaction on the JSE. They found evidence in support of the Overreaction Hypo-
thesis, in that it was shown that loser portfolios significantly outperform winner portfolios
(Shana 1994:89).
Shana (1994)
In this 1994 study, the Uncertain Information Hypothesis (UIH) of Brown, Harlow and
Tinic (1988) was used to examine whether the JSE had overreacted to the arrival of
328
unanticipated information during the 1976 to 1991 period. It was shown that, for both
good and bad news, price adjustment patterns were generally significantly positive after
the initial reaction. Further, the volatility of prices also increased significantly after the
announcements. He concluded that these findings provide strong support for the UIH.
Shana added that these findings do not support market irrationality, since it would
appear that the market reacts to the uncertain information in an efficient, if not quite
instantaneous, manner. Further, previous studies which had provided support for the
Overreaction Hypothesis were incorrect in their suggestion that there are predictable
patterns in post-event share returns, since the possibility that the dramatic news might
increase the risk of shares had not been taken into account (Shana 1994:90).
c. Semi-strong form tests
i. General
Firer, Ward and Teeuwisse (1987)

The study of Firer et al. implied that the JSE is efficient at the semi-strong level (Van
Rhijn 1994:7).
ii. Earnings announcements
Knight and Affleck-Graves (1985)

This study entailed an analysis of the speed and direction of share price changes fol-
lowing the release of preliminary earnings announcements. Knight and Affleck-Graves
found that the market reacts fairly quickly to these announcements, while there is also
evidence that the market anticipates poor performance. They concluded that their
findings support semi-strong market efficiency (Shana 1994:84).
329
iii. Calendar effects
Shana (1985)
Shana examined share price behaviour on various days of the week and found that
Mondays yield significant average negative returns. The other trading days show posi-
tive average returns, with Wednesdays having the higher returns. He, however, showed
that these effects are not exploitable when transaction costs are considered and con-
cluded that the results support market efficiency (Shana 1994:88).
Bradfield (1989, 1990)

The studies of Bradfield provide no evidence of a January effect on the JSE, but did
show that there exists a significant December effect. Bradfield argued that the Decem-
ber effect is caused by thin and lacklustre trading during the holiday season in South
Africa, leading to uncharacteristically low volatility. The excess return for December was
found to be less than 3% and Bradfield concluded that this does not represent an
exploitable inefficiency after transaction costs (Shana 1994:87; Van Rhijn 1994:45,46).
Davidson and Meyer (1993)

From their re-examination of the Monday effect on the JSE during the period 1986 to
1991, Davidson and Meyer reported that the Monday effect had disappeared. They sug-
gested that earlier findings had been as a result of methodology which biased results
in favour of finding a Monday effect (Shana 1994:88).
Shana (1994)
In one of a series of 1994 studies, Shana examined the impact of a public holiday effect
on share returns on the JSE during the 1975 to 1990 period. He showed that, on the
last trading day before the holiday, shares advance with a disproportionate frequency
and that mean abnormal returns average five times more than during the rest of the
year. He, however, showed that these abnormal return opportunities are not exploitable
as the returns are insufficient to cover transaction costs (Shana 1994:88).
330
iv. Size effect. dividend yield. PIE ratios and liquidity
De Villiers, Lowings, Pettit and Affleck-Graves (1986)

De Villiers et al. concluded that, during the period 1973 to 1983, shares of large· firms
earned higher risk-adjusted returns than did small firms. Although this provides evi-
dence of a size effect on the JSE, the nature of the effect is opposite to that found in
international (mostly US) studies (Page 1996:30).
Bradfield, Barr and Affleck-Graves (1988)

From their CAPM test, using 100 shares over the 1973 to 1984 period, Bradfield et al.
investigated the size, dividend yield and liquidity effects on the JSE (Bradfield, et al
1988:14; Van Ransburg 1998:35). They found that the dividend yield effect is statis-
tically insignificant, implying that the risk-adjusted expected returns on high-dividend
yield shares are not statistically different from those on low-dividend yield shares on the
JSE. They also found a statistically insignificant negative value for the size effect,
implying that the size effect does not exist on the JSE. Finally, they used the bid-ask
spread as a proxy for the liquidity effect and also found this value to be negative, but
insignificant. Although this may imply that the liquidity effect is also not significant for
the JSE, they did qualify their conclusion by stating that their bid-ask spread estimation
may not have captured the true implicit bid-ask spread on the JSE with sufficient accu-
racy (Bradfield et al. 1988:16-18).
Page and Palmer (1991)

Page and Palmer found no significant size effect on the JSE during the 1978 to 1988
period, but did show that size-related signs and coefficients are positive and consistent
with the earlier findings of De Villiers et al. (1986/. They did, however, find evidence of
a significant PIE effect and showed that low PIE shares yield significantly more than do
high PIE shares (Shana 1994:88; Van Rhijn 1994:38; Page 1996:30).
331
Page (1996)
The study of Page confirmed the 1991 results of Page and Palmer regarding the size
effect. Page used the data of 145 industrial companies for the period January 1978 to
September 1991 and, using both the CAPM and APT, found a persistent PIE effect
(Page 1996:34,39).
v. Event studies - Accounting changes
Knight and Affleck-Graves (1983,1988)

This 1983 study examined the movements in share prices of 21 companies that had
announced changing from the FIFO to LIFO method of stock valuation. Knight and
Affleck-Graves found short-term negative impacts on share prices and concluded that
the market had not interpreted the change in stock valuation correctly. The JSE there-
fore showed signs of being an inefficient market. They showed in their 1988 study that
the negative impact of the change in stock valuation method had been as a result of the
expectation that the companies' future earnings would be reduced. They concluded that
the negative impact on share prices is, therefore, an efficient market response and
evidence in support of the efficiency of the JSE (Shana 1994:84 ).
Gevers (1990)
With the abolishment of South African tax concessions on LIFO stock valuation in 1984,
Gevers examined the effect of the resultant higher taxes and reduced cash flow on
share prices. It was found that the market's reaction had been positive for companies
which had reverted from a LIFO to FIFO policy, that is, the share prices had reacted
positively to the expected increase in future earnings. It was also found that the market
had already discounted the positive effect on earnings by the time of the announcement
of the change in stock valuation policy (Shana 1994: 84)
332
vi. Event studies - share issues. share-splits and share-dividends
Biger and Page (1992)

Biger and Page examined the market reaction to share-splits and share-dividends
(capitalisation issues) during the 1980 to 1990 period and found evidence in support of
semi-strong form efficiency. The market reacts positively to both these announcements
and positive abnormal returns can be observed before and during the week of the
announcement. Biger and Page noted that the systematic risk of shares which are to be
split has been higher than previously perceived by the market and, hence, that the
increase in share prices can be attributed to an increase in systematic risk (Shana
1994:84,85).
Thompson and Ward (1995)

In their review of the efficiency of the JSE, Thompson and Ward found mixed evidence,
although they concluded that the market's reaction to new issues of shares, capitalisa-
tion issues and share-splits generally supports semi-strong form efficiency (Ross et al.
1996:293).
vii. Event studies - Corporate unbundling
Picton (1993)
Picton argued that the motives for South African corporate unbundling are negative by
nature and, hence, should not provide any benefits to shareholders. Where firms are
moving away from their optimal capital structures, firm values are expected to drop and
negative returns should thus be expected (Blount & Davidson 1996:64, 72).
Blount and Davidson (1996)

In their empirical study on corporate unbundling, Blount and Davidson examined the
market's response to unbundling announcements. They performed two event studies,
the first examining the effect of the announcements on parent firms and the second exa-
mining the impact on the subsidiaries. They investigated the possibility of abnormal re-
333
turns occurring during the 60 days prior to and the 60 days following the announcement
of the restructuring. They found that parent firms show greater abnormal returns after
the announcement, while the abnormal returns for the subsidiaries are not statistically
significant, both before and after the announcement. They found, further, that there is
no consistent trend in the cumulative abnormal returns of parent firms, while the subsi-
diaries show significant cumulative negative abnormal returns at the end of the obser-
vation period (Blount & Davidson 1996:64,69,70).
These share price reactions are opposite to US findings of positive abnormal returns.
Blount and Davidson concluded that the results are either due to unbundling repre-
senting a movement away from efficient structures and the motives thereof not being
market related, or that the South African market is not sophisticated enough to ensure
that unbundling enhances shareholder wealth (Blount & Davidson 1996:63).
d. Strong form tests
1. Insider trading
Shana {1987, 1990)

In his 1987 study, Shana showed that there is substantial evidence of insider trading
connected to company takeovers. It was found that insiders earn substantial returns by
trading in the shares of the acquired companies prior to the announcement. The insider
trading is also accompanied by a marked increase in trading volume. The 1990 study
concentrated on whether insider trading is also linked to companies who substantially
change their dividend policies. Again; significant insider trading was found in the six
months preceding the announcement of the change and insiders are able to earn signi-
ficant excess returns (Bhana 1994:85,86).
334
ii. The performance of professional money managers and analysts
Carter, Affleck-Graves and Money (1982)

This study entailed an evaluation of the South African unit trust industry over the 1965
to 1980 period. Carter et al. concluded that unit trusts underperform the market average
on a consistent basis (Shana 1994:86).
Gilbertson and Vermaak (1982)

Gilbertson and Vermaak examined the performance of 11 unit trusts over the 197 4 to
1981 period. They found that the unit trusts generally outperform the market, but the
reliability of the risk-adjusted portfolio returns are questionable since the risk measure
used, beta, is nonstationary and unstable. They also found that one unit trust had been
able to consistently outperform the others (Shana 1994:86).
Knight and Firer (1989)

Knight and Firer examined the performance of 10 unit trusts over the 1977 to 1986
period. Although five unit trusts were found to have outperformed the market average
before transaction costs and management fees, their average annual risk-adjusted
returns, after taking these expenses into consideration, were found to be less than the
market return (Shana 1994:86; Srevis 1998:37).
Lambrechts (1989,1994)
In his 1989 study, Lambrechts found that several unit trusts are able to outperform the
market in certain periods, but that the industry as a whole is not able to outperform the
market average. The results of the 1994 study confirmed the earlier results, since it was
found that only three out of 37 unit trusts had been able to outperform their respective
market indices over the period April 1991 to March 1994 (Shana 1994:87).
Bhana (1990)
Shana examined the recommendations of stockbrokers and investment advisory ser-
vices and found that they are of limited value to investors. Profit opportunities are
335
available prior to the publication of the recommendations, but these disappear within
the week after publication. Shana concluded that only those investors who have access
to private and unpublished advice will be able to take advantage of profit opportunities.
Shana also noted that the study had only investigated a small number of stockbrokers
and investment advisory services and, hence, may not provide representative results
(Shana 1994:92,93).
Biger and Page ( 1993)

Biger and Page examined the performance of unit trusts, using an APT multi-factor
approach. They examined 25 unit trusts over the period February 1988 to March 1992
and found that unit trusts are not able to outperform a simple buy-and-hold strategy
(Srevis 1998:37).
Garvin (1995) and Meyer (1997)

Meyer examined the performance of unit trusts over the July 1985 to June 1995 period
and also found that unit trusts are not able to outperform a buy-and-hold strategy.
These results were confirmed in Garvin's study (Srevis 1998:39).
iii. Market timing and share selection ability
Carter, Affleck-Graves and Money (1982)

From this 1982 study of the unit trust industry, covering the period 1965 to 1980, Carter
et al. concluded that unit trust managers have no special share selection or market
timing ability (Shana 1994:86).
Smith and Chapman (1994)

Smith and Chapman examined 28 actively managed unit trusts and their study covered
the March 1973 to December 1992 period. They found that the unit trust managers have
poor market timing and share selection ability (Srevis 1998:37,38).
336
Bradfield (1998)
In a recent study, Bradfield confirmed earlier findings that local fund managers exhibit
no positive share selection or market timing ability. Bradfield examined monthly prices
of 13 unit trusts over the June 1985 to June 1995 period and found that none of the fund
managers show any significant share selection ability. It was also shown that only one
of the funds exhibits statistically significant positive timing ability, while six exhibit statis-
tically significant negative timing ability (Bradfield 1998: 1, 7,9, 11 ).
Brevis (1998)
Brevis examined the industrial sector of the JSE for the periods January 1970 to Sep-
tember 1987 and January 1989 to June 1997, in order to compare the performance of
a buy-and-hold strategy with a market timing strategy. The results indicate that, with the
application of a market timing strategy within an APT framework, active fund managers
should, from both a theoretical and practical viewpoint, be able to earn higher risk-
adjusted returns than passive managers. However, Brevis noted that both dividends and
transaction costs were ignored in the study and that the study was limited to the
industrial index. Only by taking all indices and the effect of dividends and transaction
costs into consideration, can it be conclusively established whether an active market
timing strategy can outperform a buy-and-hold strategy on the JSE (Brevis 1998:220,
253,254).
Brevis (1998:45,46,242) concludes that, as yet, no definite conclusions can be drawn

regarding the degrae of efficiency of the JSE. It seems as if it can be viewed as being
operationally efficient, but that there also exists opportunities for market specialists to
outperform the market. The average investor who does not have the knowledge and
ability of market specialists is, however, unlikely to be able to outperform the market on
a consistent basis. Hence, the conclusion currently seems to be that the JSE is efficient
for some investors, but inefficient for others.
337
5.3.2 Some alternatives to the EMH
With the stock market crash of October 1987, the US stock market declined 23% on a
Monday in October 1987, following a substantial decline on the prior Friday. The UK
stock market also fell by a third in three days and similar declines occurred in most
other world markets. No major release of bad news had been associated with these
declines and it could thus not be ascribed to rational changes in investors' expectations
(Dobbins et al. 1994:21; Elton & Gruber 1995:438).
Numerous reasons have been postulated as being the causes for the crash, including
panic, failure of the trading mechanism and formula trading. If the reasons are found to
be unpredictable events it implies that the hypothesis about market information effi-
ciency remains unchallenged. It does, however, suggest some market irrationality (Elton
& Gruber 1995:438).
Pike and Neale (1996:45) add that the October 1987 crash has contributed to the cre-
dence of the Speculative Bubble Theory.
5.3.2.1 The Speculative Bubble Theory

According to this theory, stock market behaviour is not based on the fundamental analy-
sis of new information, but rather on the inflating of prices and the subsequent bursting
of these speculative bubbles. The price inflation is caused by investors who believe that
other investors will always be willing to pay more for the shares at a later stage. Bull
markets are created by their trading activities, but in the end the bubble bursts and, de-
pending on the size of the bubble, the market corrects itself with a crash (Pike & Neale
1996:45).
The crash of October 1987 has also strengthened the belief that share prices do not
follow a random walk. Investigation of the pattern of past share prices and the realisa-
tion that shares had been overvalued before the crash, has led to the uncovering of
evidence that share price movements do not follow a random walk. It has, for example,
338
been found that significant market downturns occur more frequently than significant
upturns. The principles of a new branch of mathematics, Chaos Theory, have been
employed in order to explain these patterns (Pike & Neale 1996:45,46).
5.3.2.2 Chaos Theory

Chaos Theory is a branch of mathematics that is used to analyse natural systems, such
as ocean currents, weather patterns and river systems. These systems appear to be
chaotic, but chaos theorists suggest that their random and unpredictable patterns do
follow a set of rules. Once a trend begins in such a system, it continues and the pattern
becomes seemingly highly predictable until it is stopped by something. From this, it was
found that these systems can be modelled and the behaviour thereof could be fore-
casted. However, the predictions of the behaviour of chaotic systems are extremely
sensitive to the accuracy of the conditions specified at the beginning of the estimation
period, that is, any small errors in the specification can lead to major errors in the
forecast (Pike & Neale 1996:46; Jones 1998:275).
Peters ( 1991 )
Peters suggests that stock markets are chaotic in the manner described by Chaos
Theory, in that markets do have memories, often have major price swings and move-
ments in share prices are not completely random. He found that today's price move-
ments in UK share prices are affected by price changes which had occurred a number
of years ago. He also found that the most recent price changes have the biggest effect
on current price changes and that the impact decreases the longer the changes are
traced backwards. Further, it was shown that these patterns 1n price movements are
persistent and if prices have changed upwards, the next change is more likely to be up
than down. However, Chaos Theory also suggests that persistent upward movements
will be stopped and, more than likely, result in major corrections or even market
crashes. Peters suggests that stock markets do exhibit patterns which are overlaid with
random noise and the more noise, the more inefficient the market. He found that the US
market is more efficient than the markets of the UK and Japan (Pike & Neale 1996:46).
339
Others have suggested that stock markets are basically rational and efficient, with only
occasional chaos associated with speculation activity. It is currently not clear whether
markets are efficient, chaotic or somewhere in between. Hopefully further research will
provide some evidence to bring some clarity to the matter (Pike & Neale 1996:46).
Vanderwicken (1994)
Vanderwicken found that high-tech share traders have used the powerful computers
which are currently available, together with recent non-financial discoveries, in order to
attempt to discover how stock markets really work. These high-tech traders have found
evidence that markets follow the same principles as natural systems such as weather
patterns, physical structures such as molecules or biological structures such as the
human brain (Jones 1998:275).
Vanderwicken identified four basic ideas on which these high-tech traders base their
models:
• Chaos Theory is based on the idea that investors buy shares because the
prices are rising and, hence, prices continue to rise with the continued buying.
This pattern continues until it is stopped by some surprising occurrence, for
example, unexpected economic events or news. The chaos in the market then
resumes until something happens to cause a new trend to start (Jones
1998:275).
• Neural networks computer programs which attempt to mimic the human brain,
that is, they learn from experience and can be taught to react in certain ways.
Similarly to technical analysis, only much more thorough and much faster, these
networks can be used to seek patterns in share price movements and to predict
future movements in the market. Vanderwicken cites evidence about a fund that
had used neural networks for share selection since 1989 and has outperformed
the Standard & Poor 500 index by between 2% and 7% per quarter over a three-
year period (Jones 1998:275).
340
• Genetic algorithms are computer programs which mimic the evolution of living
organisms and is based on the principle that markets also have memories. The
algorithms find patterns in stock market behaviour and then predict how the
market will behave under different conditions (Jones 1998:275).
• Microstructure analysis uses computer programs which are derived from

particle physics and are used to analyse the relationship between stock market
changes and environmental changes (Jones 1998:275).
Vanderwicken notes that by 1994, 10% to 15% of stock market trading volume can be
accounted for by these new technologies and that the field of investing will be revo-
lutionised if these high-tech traders were to be shown to be consistent in their success
(Jones 1998:276).
5.3.3 Conclusion
Capital market efficiency has significant implications for investment analysis and port-
folio management. The activities of rational profit-seeking investors who react quickly
to new information should keep the market relatively efficient, resulting in market prices
that show true intrinsic values. This will ensure that the risk/return relationship remains
consistent (Reilly & Brown 1997:247).
The results of most weak form efficiency tests tend to support the hypothesis that share
prices reflect all market information and this means that technical trading rules which
use historical data cannot be used to successfully and consistently predict future share
returns (Reilly & Brown 1997:247).
The test results on semi-strong form efficiency are mixed. Most of the event studies on
economic events, such as share-splits, initial public offerings, accounting changes, cor-
porate events, unexpected news and non-earnings-related information, to a large de-
341
gree provide consistent support for semi-strong form market efficiency. However, the
return prediction studies have produced conflicting results, which generally do not
support the semi-strong hypothesis. Anomalous evidence from quarterly earnings
announcements, PIE ratios, the size effect, neglected firms, various calendar effects
and BV/MV ratios have indicated that different variables can be used to predict diffe-
rential rates of share returns (Reilly & Brown 1997:247).
Strong form efficiency implies that all information is reflected in share prices and this
hypothesis is not supported by the evidence from the tests on corporate insiders and
stock exchange specialists. The studies on analysts provide mixed evidence and it
seems as if some analysts have either superior skills or access to private information.
However, the performance of portfolio managers and the tests on their market timing
and share selection ability generally support strong form efficiency, in that these
managers are generally unable to outperform a buy-and-hold strategy (Reilly & Brown
1997:247).
Elton and Gruber ( 1995:439) add that evidence from the studies on weak form efficiency
and semi-strong form event studies to a large degree confirm that all publicly available
information is rapidly incorporated in share prices. The anomalous results on share re-
turn patterns, firm characteristics and post-earnings drift are reliant on the asset pricing
model chosen and, hence, the implications of these results remain controversial. They
conclude that anomalous results of calendar patterns in share returns do not support
market efficiency, but add that the findings on the inability of professional money
managers to outperform the market raises serious doubts about the usefulness and
significance of these patterns.
The recent evidence from the tests on the EMH hold some very important implications
for investors, analysts and portfolio managers. Reilly and Brown (1997:247) provide the
following analysis of these implications:
• It seems as if technical analysis is of no value.

342
• Fundamental analysis can be useful, but its use is limited as it requires the ability
to accurately predict estimated future values for the relevant economic variables.
• Superior analysis is possible if accurate projections of different variables can be
provided. These projections should also be different from the consensus.
• If investors do not have access to superior analysis, their portfolios should be
managed like index funds. Those with superior analysis ability and knowledge
should generally concentrate on neglected firms and medium-capitalised firms.
These alternatives provide the biggest likelihood of being mispriced, especially
when market value/book value ratios and firm size are also taken into consi-
deration.
Reilly and Brown (1997:247) conclude that the evidence indicates that capital markets
are fairly efficient and, as a consequence, most analysts and portfolio managers find it
extremely difficult to achieve superior results in any consistent manner.
Some alternative theories to the Efficient Market Theory, including Speculative Bubble
Theory and Chaos Theory, together with the development of some sophisticated com-
puter programs - based on discoveries in mathematics, physics and biology - suggest
that the EMH does not hold and that there is consistent and/or predictable patterns in
share price movements (Pike & Neale 1996:46; Jones 1998:274-276).
Although it is not known which of these descriptions of capital markets are correct,
market prices can not be ignored, even if it is thought that they may not represent the
true value of the companies. Only the examination of long-term trends will provide in-
sight into the behaviour of stock markets (Pike & Neale 1996:46,47). Since the issue of
market efficiency remains unresolved, investors should probably assume that the
markets are reasonably, but not totally, efficient (Jones 1998:274,277).
343
5.4 RECENT DEVELOPMENTS IN THE CAPM
Blume ( 1993:6, 7) notes that since the development of the CAPM, the use of beta as a
measure of portfolio risk has been widely applied in the investment community, even
though empirical studies have identified some discrepancies between the CAPM pre-
dictions and empirical data. These early tests of the CAPM found that the basic rela-
tionship between beta and expected returns is positive, although not as steep as what
was predicted. It was generally found that low-beta shares have higher returns than pre-
dicted and that the returns of high-beta shares are lower than predicted. The testability
and usability of the CAPM had also been questioned by Roll in 1977. The main area of
criticism was centred around the CAPM's reliance on a market portfolio of all risky
assets and since this market portfolio is unobservable, market proxies have to be used
in the tests and application of the CAPM. Empirical tests have shown that when incom-
plete measures of the market portfolio are used, beta is mismeasured, resulting in
inaccurate predictions of share returns. Despite these problems, it was shown at the be-
ginning of the 1980s that the CAPM was the best pricing mechanism available at the
time and that it was widely used for investment analysis.
The CAPM has been the subject of a large number of empirical tests and the results of
the earlier tests have generally supported the CAPM. It was found that beta is a valid
and reasonably complete measure of risk and that the CAPM, although an expectational
model, generally provides useful explanations of actual events. However, since the mid-
1980s, more recent studies have provided evidence that seriously challenges the vali-
dity of the CAPM (Mclaney 1997: 177). Most of the recent studies of the CAPM have
concentrated on the following aspects:
• An examination of the relationship between beta and expected return, that is,
determining whether the CAPM describes share returns and whether the rela-
tionship between beta and return is positive and linear. These tests have not
only examined the relationship between beta and expected return, but have also
attempted to establish whether other variables provide additional or better
344
explanations of the return-generating process. Skewness in the return distri-

bution, firm size, PIE ratios, leverage, the book value/market value (BV/MV)
ratios, dividend yield and liquidity have also been examined to establish whether
these variables are additional risk factors which, along with beta, explain returns
(Reilly & Brown 1997:312-315).
• Another area of investigation is an examination of the stability of beta and the

effect of stock market volatility. These tests have attempted to establish whether
historical betas are useful in estimating future betas and also if there exists a
relationship between share returns and stock market volatility (Upsher 1993:6;
Reilly & Brown 1997:310).
• As the principle of the existence of a market portfolio consisting of all risky assets
is central to the CAPM, many studies have examined the effect of the use of a
proxy for the market portfolio. These studies have also provided controversial
arguments regarding the testability and usability of the CAPM (Ross 1993:11;
Reilly & Brown 1997:293).
The results of these recent studies are reviewed in the next section.
5.4.1 Recent research related to the CAPM
5.4.1.1 Beta as a measure of risk
a. Skewness
Sears and Wei (1988) and Lim (1989)

Both these studies have confirmed the importance of skewness in the return distribution
as a means of explaining why the CAPM seems to undervalue low-beta shares and
overvalue high-beta shares (Reilly & Brown 1997:314).
345
b. Other variables. including market capitalisation. leverage, P/E ratios and liquidity
Bhandari (1988)
The study of Bhandari found that financial leverage, as measured by the debt/equity
ratio, provides some additional explanation of cross-sectional average returns. This
implies that not only beta and firm size are important in explaining returns, but also that
a CAPM with three risk variables (beta, size and financial leverage) will provide a better
explanation of the return-generating process (Reilly & Brown 1997:314).
Baillie and Gennero (1990)

Baillie and Gennero found almost no evidence in support of the relationship between
mean portfolio returns and the variance of those returns. They concluded that a simple
mean-variance model, the CAPM, is inappropriate and that alternative risk measures
must be found (Upsher 1993:7).
Fama and French (1992,1993,1996)

In their 1992 study, Fama and French examined the viability and usefulness of the
CAPM through an evaluation of the combined roles of market beta, firm size, P/E ratios,
financial leverage and the BV/MV of equity ratio. The impact of these variables was
examined using the cross-section of average share returns on the NYSE, AMEX and
NASDAQ (Reilly & Brown 1997:315).
They found a correlation between beta and returns, but showed that this is due to the
correlation of both beta and returns to the other factors. They also showed that, even
when beta alone is used to explain average share returns, the relationship between
beta and average returns had disappeared during the 1963 to 1990 period. It was, how-
ever, shown that the other variables are significant in explaining average returns (Rees
1995:172; Reilly & Brown 1997:315).
346
They found, further, that the negative relationship between firm size and average re-
turns persists after the inclusion of the other variables, while the positive relationship
between the BV/MV ratio and average returns also persists when the other variables are
included. Also, with the inclusion of both the BV/MV ratio and firm size, the BV/MV ratio
has been shown to have a more consistent and significant role in explaining returns. It
was also shown that returns increase with an increase in the BV/MV ratio and decrease
with an increase in size and, hence, that the highest average return can be obtained
with portfolios consisting of smallest size and highest BV/MV shares (Reilly & Brown
1997:315).
From these findings, Fama and French launched a severe and controversial attack on
the basic single-factor CAPM, in that they concluded that beta is not the correct
measure of risk and, thus, that the CAPM does not explain average share returns.
Critics of their conclusions have, however, accused them of mining their data until they
could find something against the CAPM (Nichols 1993:68).
In their 1993 study, Fama and French extended their analysis to include both shares
and bonds, and concluded that there are at least three factors which are important in
explaining share returns and two in bond returns. They found that a market factor, size
factor and BV/MV factor provide a good explanation of the cross-section of average
share returns, while a term premium and default premium explain most of the variation
in bond returns (Rees 1995: 172).
From these findings, they suggested that a three-factor CAPM should be used. They
applied such a model in their 1996 study and concluded that the model explains a
number of anomalies from earlier studies (Reilly & Brown 1997:315).
Chan and Lakonishok (1993) and Lakonishok (1993)

Using data for the 1926 to 1991 period, Chan and Lakonishok examined the relationship
between beta and returns in order to establish whether Fama and French (1992) had
been correct in their conclusions about the importance of beta. They found that the
347
prices of high-beta shares had declined more than those of low-beta shares. This
provides proof that high-beta shares are in fact more risky than low-beta shares and,
thus, that investors should expect compensation for this type of downside risk. They
also found that, in both up and down markets, the estimated slope of the compensation
per unit of beta had been close to the expected slope as described by the CAPM. This
confirmed that high-beta shares are more risky when investors are more concerned
about downside risk (Lakonishok 1993:39).
Lakonishok (1993:40) concluded that Fama and French (1992) probably went too far
when they suggested that there exists no relationship between beta and return. The
available evidence does not support such an unqualified statement.
Fouse (1993)
Fouse also criticised the 1992 conclusions of Fama and French and suggested that they
seemed to have forgotten that the CAPM is an expectational model. Fama and French
(1992) did not distinguish between expected and realized returns and in fact seemed
to believe that these two returns are the same. Fouse emphasised that investment ma-
nagers know that this is not true. He added that most of the criticism of beta is based
on evidence that it is a technical signal that does not work without fail. Sharpe (1970)
suggested that an appropriate test of the CAPM depends on the intended use of the
model. Fouse concluded that beta theory provides investors with a robust framework
for the understanding of share prices and share price movements, and without the in-
sights of the CAPM investors will be unable to fully capitalise on discounted cash flow,
the value of factor analysis and the insights of the APT (Fouse 1993: 110, 117).
Dennis, Perfect, Snow and Wiles (1995)

The Dennis et al. study confirmed the results from Fama and French's (1992, 1993)
studies. They confirmed that portfolios of small firms with high BV/MV ratios provide the
optimal combination, even after considering transaction costs and annual rebalancing,
and that the best results can be obtained with rebalancing every four years (Reilly &
Brown 1997:317).
348
Kothari, Shanken and Sloan (1995)

While Fama and French (1992, 1993, 1996) had measured beta with monthly returns, the
study of Kothari et al. used annual returns in order to avoid the trading problems asso-
ciated with using monthly data. Using these annual betas, they found substantial com-
pensation for beta risk and suggested that the relationship between returns and the
BV/MV ratio may be a time-period phenomenon, and that this may not be significant
over a long period (Reilly & Brown 1997:317).
Pettengill, Dundaram and Matthur (1995)

Pettengill et al. also noted the Fama and French (1992, 1993) problem of using realized
returns to test the CAPM. They showed that when returns are adjusted for expectations
regarding negative market returns, a consistent and significant relationship exists be-
tween beta and returns(Reilly & Brown 1997:317).
Grundy and Malkiel (1996) and Jagannathan and Wang (1996)

Grundy and Malkiel contended that beta is very useful in measuring risk in declining
markets - which is when it is at its most important to assess risk. The Jagannathan and
Wang study used a conditional CAPM, which allows for changes in beta and the market
risk premium, and showed that it performs well in explaining cross-sectional returns
Various studies
Handa, Kothari and Wasley (1989) and Kothari, Shanken and Sloan (1992) found
correlation between market capitalisation and errors in beta estimates. Banz and Breen
(1986) and Kothari, Shanken and Sloan (1992) also showed that selection bias has
an important effect on beta estimates and, thus, that apparent anomalies can be as a
result of poor implementation of the CAPM. Fuller (1993) argued that the small firm
effect can to a large degree be attributed to such mismeasurement, and in particular the
effect of the bid-ask spread and differentials in order flow across time (Harrington
1993:2,3).
349
5.4.1.2 Measuring beta
Some studies have examined published estimates of beta for comparability, in particular
the betas published in Merrill Lynch's Security Risk Evaluation Report and Value Line's
Investment Swvey. These beta estimates differ in the data used, in that Merrill Lynch
uses the Standard & Poor 500 as market proxy with 60 monthly observations, while
Value Line uses 260 weekly observations with the NYSE composite series as market
proxy (Reilly & Brown 1997:311,312).
Statman ( 1981 )
Statman examined the betas for 195 firms over comparable periods and found small but
significant differences between their beta estimates (Reilly & Brown 1997:312).
Reilly and Wright ( 1988)

Reilly and Wright examined 1 100 shares for three non-overlapping periods and found
the same differences as had been reported by Statmari in 1981. They showed that the
differences are due to the different time intervals used, while both the size and the di-
rection of the differences are affected by the shares' market value. It was found that the
weekly interval yields larger betas for large firms and smaller betas for small firms. They
concluded that the time interval and the size of the firm should be carefully considered
when betas are estimated (Reilly & Brown 1997:293,312).
5.4.1.3 Stability of beta and the relationship between share returns and market
volatility
Pindyck (1984, 1988)

In the 1984 study, Pindyck investigated the decline in real share prices over the 1965
to 1981 period. It was found that profitability and the variance of returns (volatility) are
significant in explaining positive share price changes. The 1988 study showed that vo-
latility provides more of an explanation of the market decline in 1974 than profitability
or interest rates could (Upsher 1993:6).
350
Poterba and Summers (1986)

Poterba and Summers found that shocks to volatility must occur over a long period of
time in order to have any significant effect on share prices. They argued, however, that
volatility shocks are short lived and that expected share returns are therefore not affec-
ted by volatility changes (Oosthuizen 1992:4; Upsher 1993:6).
Taylor (1986)
Taylor examined the stochastic process generating share returns and found that the
process generating share returns and prices is not linear. He concluded that this non-
linearity is as a result of changes of the variance of returns, and that this is caused by
movements in the general level of market activity (Oosthuizen 1992:5).
French, Schwert and Stambaugh (1987)

From their examination of the relationship between share returns and market volatility,
French et al. found a positive relationship between expected volatility and the expected
market risk premium. It was also found that a negative relationship exists between share
returns and unexpected changes in volatility (Oosthuizen 1992:5: Upsher 1993:6).
Chou (1988)
Chou used data from the same source as had been used by French et al. in their 1987
study, and claimed that the methodology of the Poterba and Summer (1986) study had
been limited and could have provided misleading results. Chou found persistence in
volatility and a strong relationship between excessive volatility and a decline in share
returns (Oosthuizen 1992:6; Upsher 1993:7).
Ball and Kothari (1989)

This study showed that firms which currently have high betas and are involved in mer-
gers or acquisitions, are likely to experience a lowering in their betas as the merged or
acquired firms are likely to have lower beta values. This provides additional evidence
that beta values can change over quite short periods of time (Rees 1995: 170).
351
Schwert (1990)
Schwart investigated whether computer trading and options trading have caused an in-
crease in market volatility, but found no evidence to support this assertion (Upsher
1993:4).
Bodertha ( 1991) and Lilian ( 1991)

Lilian found significant time variability in the return to risk ratio (index of relative risk
aversion), while Bodertha tested the CAPM with time varying risk and returns and found
evidence of the size effect, monthly effect and quarterly effect (Upsher 1993:8).
Poon and Taylor (1992)

Using UK data and a methodology similar to that used by French et al. in 1987, Poon
and Taylor found no statistical significant relationship between share market volatility
and share returns. They concluded that earlier studies' assumption that share returns
are normally distributed, caused an overstatement of the significance of the relationship
(Upsher 1993:9).
5.4.1.4 The market portfolio - testability and usability of the CAPM
Roll and Ross (1993,1994) and Ross (1993)

In their 1994 study, Roll and Ross confirmed that the use of an incorrect index as proxy
for the market portfolio can lead to invalid results and conclusions (Ward 1994: 101 ).
The paper of Ross, based on the 1993 review of the academic literature by Roll and
Ross (1993: 11-13), provides the following criticisms and conclusions about the testa-
bility and usability of the CAPM:
• No empirical systematic relationship between expected return and beta can be

discerned.
• Most empirical studies do not support the theoretical relationship between
expected return and beta, whereby the CML should rise at a rate that is equal to
the slope of the return of the market portfolio in excess of the risk-free rate.
352
• Beta is of limited use in estimating expected share returns and beta has nothing
to say about the CAPM. It is therefore an illusion to conclude that the CAPM is
the same as finding that expected returns and beta are related to each other.
• The CAPM states that the market portfolio is mean-variance efficient and all
knowledge about the CAPM is contained in this statement, that is, expected
returns should be linearly related to the beta of the market portfolio. Ross added
that this was not found to be the case. The empirical evidence does not support
the linear relationship of expected share returns, whereby share portfolios with
higher betas has higher returns.
• Although the CAPM provides an elegant explanation of the risk and return issues
and how they are related, it can not be tested in any practical way and, hence,
can not be proved to be either valid or invalid.
• Since the CAPM is not useful in providing expected returns, it is a wonderful, but
rather useless, theory. Beta, however, does have some practical use, in that it
can provide some information on the relative risk of different portfolios, and that
it shows that the prices of shares with high betas tend to increase more when the
market goes up and decrease more when the market went down than, do the
share prices of shares with low betas. It can, however, not be stated with confi-
dence that high beta portfolios will yield long-term higher expected returns.
Reilly and Akhtar (1995)

Reilly and Akhtar confirmed that the choice of market proxy can have a serious impact
on the testing of the CAPM. They found significant differences in beta estimates for 30
shares for three alternative periods, using three different market proxies. They conclu-
ded that this can have a significant effect on expected returns. They also showed that,
since the covariance is much lower, while the variance is only marginally smaller, betas
tend to be lower with a world share index than with a domestic market index (Reilly &
Brown 1997:296, 320,321 ).
Various studies
The results of the studies of Amihud, Christensen and Mendelson (1992) and Kandel
353
and Stambaugh (1993) confirmed that, even though a relationship between beta and
expected returns may not exist, proxies for the market portfolio may actually be close
to being mean-variance efficient. They also showed that this may be true, even when
there exists an exact linear relationship between beta and expected returns for optimal
mean-variance portfolios. Jagannathan and Wang (1993), hence, concluded that the
CAPM may actually be true and the observed anomalous results may have been caused
by the true market portfolio not being used (Harrington 1993:2).
a. Beta as a measure of risk
Seally and Knight (1987)

Seally and Knight used 107 shares, allocated to 1Oportfolios, over an eight-year period,
starting 26 January 1973, to examine the systematic effect of dividend policy on JSE
share prices. They included a dividend yield variable in the CAPM to test the impact of
dividend policies on realised share returns. They found that the CAPM's positive risk-
return relationship does not hold and that the actual risk premium offered for bearing
risk is not statistically different from zero, or is at best only marginally positive. They did,
however, find a significant positive relationship between share returns and dividend
yield over the full period of their study (Seally & Knight 1987:33-41 ).
Westwell (1987)
Westwell showed that the movement of the market, as expressed by the JSE All Share
Index, is the most important factor influencing share returns (Oosthuizen 1992:7).
Bradfield, Barr and Affleck-Graves (1988)

Their study entailed an investigation of the applicability of the CAPM for the JSE and,
secondly, whether variables other than beta has significant additional influences on JSE
share returns. They used a sample of 100 shares, with weekly prices for the period
January 1973 to December 1984 and a series of returns taken at four-weekly intervals
354
for their study. They found a linear risk/return relationship and that ex ante estimates
of beta are successful in predicting share returns on the JSE, although it was found that
beta is not as successful with predicting returns on gold shares. They concluded that
their findings provide support for the validity of the CAPM and that the CAPM should be
accepted as a reasonable model in the context of the JSE (Bradfield et al. 1988: 14; Van
Rhijn 1994:143,144).
The second part of their study found no evidence that the additional variables exa-
mined, namely dividend yield, firm size, market capitalisation and liquidity, have any sig-
nificant additional explanatory power of JSE share returns. They concluded that these
effects do not exist on the JSE (Barr 1990: 18; Ward 1994: 100).
Bradfield and Barr (1989)

Bradfield and Barr confirmed that the CAPM stands up well to South African empirical
testing and that the CAPM can be relied upon for valuing JSE shares (Ward 1994: 100).
Viljoen (1989)
Viljoen examined the performance of the CAPM on the JSE during the market crash of
October 1987. It was found that the CAPM does explain the relationship between risk
and return, although the relationship was not found to be completely linear or static.
Further, some weaknesses in using beta as a risk measure were identified and it was
found that no significant correlation between share returns and betas exists for the
sample of shares examined (Viljoen 1989:80).
Ward (1994)
Ward used monthly index and dividend data from 38 JSE sectoral indices over the
January 1983 December 1992 period and found the CAPM to be an acceptable model
for both mining and industrial shares. It was also found that firm size and financial
gearing represent a major dimension of risk in these sectors. Some evidence was found
in support of the 1992 findings of Fama and French, in that it was shown that there is
a significant relationship between share returns and market capitalisation. However, the
355
relationship between market price and net asset value was not found to be statistically
significant (Ward 1994:99, 103, 111, 112).
b. The stability of beta and the relationship between share returns and market
volatility
Oosthuizen (1992)
Oosthuizen used 13 years of daily closing levels of the JSE All Share Index, Industrial
Index and All Gold Index as data for his investigation of the relationship between share
returns and the volatility of share returns. The time periods January 1979 to August
1985 and September 1985 to January 1992 were examined and it was found that the
market responds quickly to volatility and the best results are obtained with the use of
a five-day variance to determine volatility (Oosthuizen 1992:6, 17,48).
Upsher (1993)
The study of Upsher extended the work of Oosthuizen (1992) and found that share
returns do not comply with the assumptions of the CAPM. Upsher, also, used the daily
closing prices for the JSE All Share Index, Industrial Index and All Gold Index, but over
the October 1978 to February 1993 time period. A strong negative relationship between
expected returns and unpredictable volatility was found, although it was shown that for
the All Share Index there exists a strong positive relationship between expected returns
and predictable volatility. Upsher concluded that this provides some support for the
CAPM, but that the overall evidence from the study was inconclusive (Upsher
1993:iv,v, 16,39,68).
Keogh ( 1994)
This study examined the stability of beta and the usability of the CAPM and beta in a
South African context, using seven industrial sectors over the 1965 to 1993 period. It
was found that beta does not remain stable over time and that this has a negative effect
on the usability of beta and CAPM in the South African context (Keogh 1994:1,2,205).
356
Van Rhijn (1994)

Van Rhijn also examined the stability of beta and the usability of the CAPM for financial
and investment management under South African conditions. He examined the shares
of the companies listed in the industrial sector of the JSE over period January 1980 to
December 1992, using monthly return data. Dividends were only included to establish
the beta coefficients. The findings on the stability of beta confirmed the results obtained
by other researchers, in that portfolio betas were found be much stabler than those of
individual shares. However, they were also found to have a higher degree of instability
during the years after 1990. Van Rhijn identified some major considerations which
should be taken into account when estimating beta on the JSE:
• The thinly traded nature of the JSE.
• The choice of market proxy.
• The distribution of share returns.
• The time varying element of systematic risk.
• The inclusion of prior beliefs about systematic risk measures.
The major conclusions from this study were that the relationship between average re-
turns and risk for the JSE supports the CAPM and that the CAPM can be used to test
the risk/return relationship on the JSE. Further, higher returns are associated with
higher betas and this, too, is consistent with the theoretical beta/return relationship (Van
Rhijn 1994:4,5,9, 175, 176,276,297,338).
Bowie and Bradfield (1997)

Bowie and Bradfield (1997:1-4) measured the predictability of betas on the JSE from
one five-year period to the next, namely January 1977 to December 1981 , January 1982
to December 1986 and January 1987 to December 1991. in order to establish the stabi-
lity of beta on the JSE. They found that, once the distortion of thin trading on the JSE
was eliminated, the levels of beta stability are similar to those reported for UK and US
stock markets. Betas were also found to have a tendency to regress to some average
mean over time.
357
c. The risk-free rate
Bradfield, Barr & Affleck-Graves (1988)

Bradfield et al. found that the 12-month fixed deposit rate of major South African com-
mercial banks provides a reasonable proxy for the risk-free rate and that, for the JSE,
the intercept of the CAPM is not significantly different from this rate (Bradfield et al.
1988: 15; Oosthuizen 1992:46).
Firer (1993)
Firer (1993:29) reviewed the South African CAPM research, in particular the interest
rate used as proxy for the risk-free rate. He found that the choice of rate had varied from
the 90-day Treasury bill rate, the 360-day Treasury bill rate to the 12-month fixed de-
posit rate. He concluded that this diversity in choice reflects the fact that interest rates
are to a degree managed by the authorities and that no financial instrument is com-
pletely riskless. He added that the lack of clarity regarding this component of the CAPM
detracts from its practical application.
Various studies
The study of Laing (1988) used the 91-day Treasury bill rate as proxy for the risk-free
rate (Van Rhijn 1994:172), while Ward (1994) used an index of monthly returns based
on the 91-day Banker's Acceptance discount rate as proxy (Ward 1994: 103). On the
other hand, Page (1996) used Black's (1972) zero-beta portfolio approach in preference
to a Treasury bill rate as proxy for the risk-free rate (Page 1996:31 ).
d. The market portfolio
Venter, Bradfield and Bowie (1992)

Venter et al. investigated the predictive power of the CAPM using different market
indices as proxy for the market portfolio. They showed that the error between predicted
returns and actual returns can be reduced with the selection of an appropriate market
sector index (Ward 1994: 101 ).
358
Various studies
Bowie and Bradfield (1993) concluded that the use of an incorrect market proxy results
in beta being underestimated, causing a reduction in the predictive power of the CAPM.
Ward (1994) found that the JSE All Share Index is an acceptable proxy for the market
portfolio and that the All Share Index is preferable to the major sectoral index (Ward
1994:100, 111 ). Laing (1988) also used the All Share Index as a surrogate for the mar-
ket portfolio (Van Rhijn 1994: 172). Page (1996:31 ), on the other hand, used an equally
weighted index of all the shares examined in the study as market proxy.
e. The CAPM versus the APT
Page (1989)
Page compared the performance of three single-factor CAPM-based models with the
performance of a two-factor APT-based model and concluded that the APT approach
provides superior results on the JSE (Ward 1994: 101 ).

Biger and Page found that the observed superior performance of South African unit trust
managers disappears when their performance are evaluated using a multi-factor APT
approach rather than a CAPM-based approach (Page 1996:41 ).
Ward (1994)
Ward concluded that his findings provide some support for the APT in preference to a
single factor CAPM and that, for the JSE, an APT framework with two or more factors
could be considered in place of the CAPM (Ward 1994: 111, 112).
Page (1996)
Page (1996:28,29,41) undertook a comparative analysis of the CAPM and APT frame-
works to test whether observed results and conclusions on efficient market size and
earnings anomalies are substantially changed by using an APT multi-factor framework.
He found that the use of the APT has no significant impact on the results obtained with
359
the CAPM and concluded that the APT does not remove any possible misspecifications
within the CAPM.
5.4.2 Conclusion
Although it has been found that high-beta shares earn lower returns than expected and
low-beta shares higher returns than expected, most of the empirical evidence regarding
the beta/return relationship have tended to support the CAPM at the beginning of the
1980s. More recent studies, however, have considered other variables in order to esta-
blish whether a better explanation of the risk/return relationship could be obtained
(Reilly & Brown 1997:315,316).
Several studies have provided support for the inclusion of skewness in the distribution
of returns as a variable, since it was found that there exists a correlation between po-
sitive skewness and high beta values. The inclusion of skewness is advocated on the
assumption that investors prefer positive skewness and that they would be willing to
accept lower average returns in exchange for the opportunity to earn very high returns
Based on the results from EMH research, some studies have also examined other varia-
bles -firm size, financial leverage, PIE ratio and BV/MV of equity ratio - in order to esta-
blish whether these variables, in addition to beta, can provide an explanation of cross-
sectional returns. Some of these studies have shown that these variables have explana-
tory power beyond beta. Firm size and BV/MV, in particular, have been identified as the
most significant variables. These findings are, however, still controversial - some of the
recent studies have supported these variables, while others have differed and found a
significant and consistent relationship between beta and share returns (Reilly & Brown
1997:317,330).
Other studies have examined the risk-free rate and market portfolio, in order to establish
the validity of using proxies for these parameters or to establish whether the use of pro-
360
xies invalidates the tests of the CAPM (Mclaney 1997: 177; Reilly & Brown 1997:321 ).
Ross ( 1993) and Roll and Ross ( 1993, 1994) extended the earlier critique of Roll in 1977
and they contend that, without using the true market portfolio, testing of the CAPM and
using it to evaluate portfolio performance is meaningless. Other studies support this
contention and showed that the use of different market proxies can yield significantly
different beta values and, hence, significantly different expected returns (Reilly & Brown
1997:330).
Reilly and Brown (1997:321 ), however, contend that the problems and arguments as-
sociated with the market portfolio do not invalidate the CAPM. The dilemma surrounding
the market portfolio and choice of correct market proxy only represents a measurement
problem in testing the CAPM or in using the model to evaluate portfolio performance.
Therefore, the challenge lies in identifying or developing better proxies for the market
portfolio and/or to find improved measures whereby portfolio performance measurement
can be adjusted to reflect this dilemma.
According to Karnosky (1993:56), criticism of the CAPM on the grounds that it provides
unreliable forecasts of short-term market conditions, is misdirected. The CAPM is an
expectational model that attempts to explain the relationship among variables and
should not be viewed as a trading mechanism or a means for active portfolio manage-
ment. The CAPM only serves as the basis for the development of models that can be
used to predict future share prices and returns. Comparing the CAPM and APT on the
basis of their ability to replicate history and to forecast market prices is a futile exercise.
These two models should not be seen as alternatives, since the APT attempts to pro-
vide an explanation of current market conditions , while the CAPM attempts to describe
the fundamental relationships underlying the market.
Karnosky (1993:56) concludes that the CAPM should be judged on the basis of the in-
sight it provides on the risk/return relationship and that without the CAPM, knowledge
of the market and market conditions will be severely limited.
361
5.5 RECENT DEVELOPMENTS IN THE APT
Given the evidence of beta instability and the arguments regarding the CAPM's depen-
dence on the existence of a true market portfolio, Ross(1976) developed an alternative
asset pricing theory - the Arbitrage Pricing Theory. The APT requires less assumptions
and contends that share returns are affected by multiple factors, not only beta. The
APT, however, does not specify the number or identity of the factors affecting returns
and most empirical studies have, firstly, attempted to establish whether three, four or
five factors are significant in share returns and, secondly, whether factors such as infla-
tion, growth in GNP, changes in interest rates, etcetera are the relevant variables (Reilly
& Brown 1997:297-299).
While some recent studies have attempted to establish the number offactors, most stu-
dies were concerned with establishing the identity of these factors. Brevis (1998:98)
notes that the first group of tests are known as unspecified factor tests, while these-
cond group of tests are known as specified factor tests.
• Unspecified factor tests. These tests generally use factor analysis to determine
the number of statistically significant factors in a sample of returns. Factor ana-
lysis does, however, have the disadvantage that it is used to determine the num-
ber of factors and the sensitivity of shares to these factors. This is a complex pro-
cess and the number of shares which can be analysed at any one time are limi-
ted. A further disadvantage with the application of factor analysis is that the signs
of the factors and the sensitivity of each share to each factor can not be de-
termined in any unique way. Further, a unique hierarchy of factors for different
shares can also not be determined (Brevis 1998:98-100).
Hence, Brevis (p.99) notes that the usability of the APT depends on the reliability of the
statistical methods used to apply it. If these statistical methods are unreliable, and even
though the APT might be correct, it can not be successfully applied in the investment
process. Any tests of the APT are, therefore, combined tests of the APT and the statis-
362
tical methods that are used to apply the APT.
• Specified factor tests. Since the unspecified factor tests do not provide an in-
dication of the nature of priced systematic risk factors, researchers generally use
a set of hypothetical factors to establish whether these variables are priced.
These hypothetical factors can either be firm characteristics (dividend yield, size,
etcetera), macro-economic factors (interest rates, inflation, etcetera) or a group
of portfolios used as indices that combine the factors which may affect share re-
turns (Brevis 1998: 102, 103).
The results of some recent international and South African APT research are reviewed
in the next section.
5.5.1 Recent research related to the APT
5.5.1.1 The number of priced factors
Conway and Reinganum (1988)

These researchers established that there are two priced factors, a maior priced factor
and a minor priced factor. They found that, when applying factor analysis, no positive
relationship exists between the number of factors analysed and the number of priced
factors identified (Barr 1990: 18).
Connor and Korajczyk (1993)

Connor and Korajczyk examined the cross-sectional monthly share returns on the NYSE
and AMEX over the 1967 to 1991 period. They concluded that there are between one
and six factors generating share returns on these markets and that all the factors after
the first one have a particularly significant effect during January (Van Rhijn 1994: 170).
363
Fama and French (1993)

Rather than using macro-economic variables, Fama and French used a set of portfolios
in their model to determine the number of factors which explain share and bond returns.
They found that a five-factor model performs well in explaining cross-sectional average
returns and variations in share and bond returns (Elton & Gruber 1995:385,386; Brevis
_1998: 103).
5.5.1.2 The identity of the priced factors
Burmeister and Wall (1986), Burmeister, Wall and Hamilton (1986), McElroy and
Burmeister (1988) and Berry, Burmeister and McElroy (1988)
These studies identified five macro-economic variables that are significant in explaining
returns and found that such a five-factor APT model has greater explanatory power than
the CAPM (Brevis 1998:105). Brevis (1998:105,106) and Jones (1998:245) note that
these studies have identified the following five priced factors:
• Unanticipated changes in the risk premium.

• Unanticipated changes in the term structure of interest rates.
• Unanticipated changes in inflation or deflation.
• Unanticipated changes in the long-term growth rate of the economy.
• The residual market risk not captured by the other factors.
Berry et al. ( 1988) found that the significance of these factors changes between indus-
tries and various economic sectors. They suggested that there is no correct collection
of macro-economic factors and that a collection of equivalent factors can yield results
similar to that of the five identified factors. The explanatory power of such an equivalent
collection will, however, not be greater than the identified factors (Brevis 1998: 107).
Berry et al. also identified three characteristics which risk factors must possess in order
for them to be priced APT factors. It must be noted that it is the unexpected changes in
these factors that are important.
364
• Firm-specific factors are not APT risk factors, since they do not have a pervasive
influence on share returns.
• The factors must be priced, that is, influence expected returns.
• The factors must be unpredictable to the market as a whole. Hence, since the
macro-economic variables are at least partially predictable, they themselves are
not the APT factors, and it is only the unexpected (unpredictable) changes there-
in which are the APT risk factors (Jones 1998:243).
Beenstock and Chan (1988)

In their UK study, Beenstock and Chan examined share returns on the London Stock
Exchange over the period 1977 to 1983. They did not apply factor analysis and avoided
the anonymity of the market risk factors. Instead, they specified an economic model
generating share returns and used an iterative procedure to establish the parameters
of the model. They found that four macro-economic variables explain share returns, na-
mely interest rates, money supply, inflation and fuel and material costs (Barr 1990: 18;
Pike & Neale 1996:288,289).
Fama (1991)
Fama concluded that rather than using factor analysis to identify factors, macro-eco-
nomic variables should be tested to determine whether these variables are correlated
with share returns and, then, whether the loadings of returns on these factors explain
cross-sectional share returns. The use of factors with an economic motivation will,
therefore, ensure that multi-factor pricing models improve the understanding of share
prices (Van Rensburg 1998:21 ).
Poon and Taylor (1991)

When Poon and Taylor extended the techniques used by Chen et al. in 1986 and
applied it to the UK market, they did not find that the same macro-economic variables
affect UK share returns. They concluded that this may be due to other macro-economic
factors being important, inadequate methodology of pricing relationships, or a combina-
tion of the two (Rees 1995: 173).
365
McGowan and Francis (1991)

This study found that changes in industrial production, inflation, personal consumption,
money supply and interest rates are important APT factors in the US market (Pike &
Neale 1996:288).
Clare and Thomas (1994)

The UK study of Clare and Thomas examined 56 portfolios, each consisting of 15
shares, sorted firstly by beta and then according to firm size. The important factors iden-
tified by the beta-ordered portfolios are oil prices, corporate default risk, private sector
bank lending, current account bank balances, the Retail Price Index and the redemption
yield on UK corporate loan stock. However, for the size-ordered portfolios, only default
risk and the Retail Price Index are important factors. They found some degree of corre-
lation among the variables, but as with McGowan and Francis (1991 ), showed that the
return on the market index is not an explanatory factor (Pike & Neale 1996:289).
5.5.1.3 The APT versus the CAPM
Connor and Korajczyk (1986)

Connor and Korajczyk tested a five-factor APT model and concluded that it explains the
abnormal return results of the size effect and January effect better than does the CAPM
Gultekin and Gultekin (1987)

This study investigated the January anomaly within an APT framework and found that
the APT can not explain the January anomaly any better than can the CAPM (Linley
1992:28,29; Reilly & Brown 1997:328).
Burmeister and McElroy (1988)

They, too, concluded that the APT can not explain the January effect any better than
the CAPM. However, when their test went beyond the January effect, they concluded
that the APT is superior to the CAPM in explaining returns (Reilly & Brown 1997:328).
366
Elton and Gruber (1988, 1989)

According to Elton and Gruber, studies on the Japanese market have clearly demon-
strated that the APT is superior over the CAPM in selecting shares and explaining past
returns. Their studies found that a five-factor APT model explains and predicts expected
returns better than the CAPM. Unlike other markets, small shares on the Japanese
market have smaller betas than large shares, and they suggested that this should imply
lower expected return using the CAPM. It was, however, found that small shares still
show significantly higher excess returns. They concluded that when small shares on the
Tokyo Stock Exchange are defined as anything but the largest 100 shares, the multi-
factor APT explains returns much better than the CAPM. Elton and Gruber added that
the APT is almost universally used in the Japanese market as a replacement for the
CAPM (Elton & Gruber 1995:380).
Lehman and Modest (1988)

In order to determine the sensitivity of each share to each factor, Lehman and Modest
constructed portfolios which imitated the returns of systematic risk factors. They con-
cluded that their findings show that the APT is able to explain certain phenomena which
the CAPM does not explain (Brevis 1998: 100).
Mei (1993)
The study of Mei confirmed the findings of earlier studies, in that the APT provides a
slightly better description of share returns than does the CAPM (Van Rhijn 1994: 167).
a. Number of priced factors
Van Rensburg and Slaney (1997)

Van Ransburg and Slaney examined monthly JSE data over the January 1985 to De-
cember 1995 period and concluded that there are at least two, but not more than three,
priced APT factors on the JSE. It was found that the first two factors explained more
367
than 50% of share returns and they suggested that the first two factors can be a
gold/mining factor and an industrial factor. It was also suggested that the third factor
can be from a non-precious metal mining source (Van Ransburg 1998:21-23).
b. The identity of the priced factors
Jordaan (1986)
In a study outside the APT framework, Jordaan investigated whether exchange rate risk
should be included as an additional variable in the CAPM, that is, whether there exists
a correlation between share returns and the Rand/US dollar exchange rate and whether
the inclusion of such a variable can increase the explanatory power of the CAPM. The
sample used for the study consisted of 90 JSE shares from the mining, financial and
industrial sectors, covering the January 1983 to May 1986 period. He found that only
17 of the shares show significant correlations between share returns and the exchange
rate. It was also found that these correlations are unstable. Jordaan concluded that
there exists no statistically significant relationship between share returns and the
Rand/US dollar exchange rate (Brevis 1998: 109, 110).
Westwell (1987)
Westwell examined share returns on the JSE over two periods, July 1975 to May 1980
and June 1980 to February 1985. He found that, although these factors are not the only
significant ones and that their relative importance changes over time, the market index,
the gold price and possibly the Rand/US dollar exchange rate are significant factors in
explaining share returns (Laing 1988:37,80; Brevis 1998: 110).
Correia and Wormald (1988)

The study of Correia and Wormald investigated whether there exists a relationship
between share returns, inflation and expected inflation. Their study covered the 1960
to 1986 period and they found no significant relationship between share returns and
inflation variables for the JSE. However, when they used contemporaneous changes
in the Bankers Acceptance rate, rather than the Treasury bill rate, as a proxy for
368
changes in expected inflation, they found a significant negative relationship between

JSE share returns and inflation variables (Van Rensburg 1998:27).
Barr (1990)
The study of Barr (1990: 18,20), following on the earlier study of Page (1986), attempted
to establish the identity of the two most significant APT factors on the JSE. The data
used in the factor analysis consisted of the month-to-month annualised returns of 26
non-gold JSE Actuarial Indices over the August 1978 to July 1987 period. The data of
indices was used to eliminate the problems associated with thin trading.
Barr concluded that the two main factors affecting JSE share price movements are an
industrial type index and a financial type index. It was also found that the gold price and
short-term interest rates are the two variables affecting the industrial type index, while
local business confidence and the performance of overseas stock markets are the main
variables affecting the financial type index (Barr 1990:25; Brevis 1998: 11 O; Van Rens-
burg 1998:26).
Both Barr (1990:20) and Van Rensburg (1998:26) note that gold indices were omitted
from the study since the gold price has been shown to be the macro-economic variable
which dominates the pricing of gold shares.
Bradfield (1990)
Bradfield examined the monthly return data of 30 shares listed on the JSE over the Sep-
tember 1978 to November 1987 period with the aim of establishing the correlation be-
tween JSE share returns and those on the London, New York and Tokyo stock exchan-
ges. It was shown that the correlation coefficients between the international markets
are positive and statistically significant, while the NYSE was found to be the most influ-
ential (Brevis 1998: 119, 120; Van Rensburg 1998:27).
369
Reese ( 1993)
Reese examined share returns over the period January 1980 to December 1989 to esta-
blish whether the Rand/US dollar exchange rate, gold price, term structure of interest
rates, growth rate, market return and corporate default risk are priced APT factors on
the JSE. The effect of these factors on both mining and industrial shares were tested
and it was found that shares in these two sectors are influenced by different priced fac-
tors. It was, for example, shown that the combinations of gold price and market return,
exchange rate and market return, and inflation and market return have the same expla-
natory power of industrial share returns as the combination of default risk, gold price
and market return. The results of the study indicate that the Rand/US dollar exchange
rate, inflation, gold price, market returns and corporate default risk are priced APT
factors on the JSE, while the term structure of interest rates and growth rate are insig-
nificant in explaining JSE share returns (Brevis 1998:111-123).
Van Rensburg (1994,1995,1998)

Van Rensburg's 1995 study entailed an examination of the simultaneous effect of the
gold price, inflation rate, term structure of interest rates and the return on the Dow-
Jones Industrial Index on the 1980 to 1989 monthly share returns of four sectors of the
JSE. He found that there is a statistically significant negative relationship between the
rate of inflation and share returns of all four sectors. He also identified the gold price
and Dow-Jones returns as priced factors, and concluded that Reese (1993) had not
measured the unexpected changes in the term structure of interest rates and, hence,
had been incorrect in concluding that it is not a priced factor. Van Rensburg found a
significant negative relationship between share prices and the term structure of interest
rates (Brevis 1998:114-120).
The 1996 study found that the unexpected changes in the Dow-Jones Industrial Index,
short-term interest rates, the term structure of interest rates and the residual market
factor are significant priced factors on the JSE (Van Ransburg 1998:28). The 1998
study reviewed some recent unpublished APT research and concluded (Van Rensburg
1998:36) that growth rates of the following variables are also significant in explaining
370
JSE share returns:
• The level of gold and foreign reserves.

• The money market shortage.
• The balance on the current account as a percentage of GDP.
• Future indexed sectoral accounting earnings.
In addition, the study also examined monthly JSE share returns over the period January
1965 to December 1995. It was found that Dow-Jones returns and changes in money
market rates affect returns on industrial shares, while they have an insignificant effect
on gold shares. He also showed that both the gold price and changes in long-term inte-
rest rates affect gold and industrial shares, and that the gold shares are three times
more sensitive to the gold price than industrial shares (Van Ransburg 1998:33,35).
Buijs (1996)
The study of Buijs used monthly return data over the 1988 to 1993 period and confirmed
the significant relationship between gold sector share returns and the US dollar gold
price (Van Ransburg 1998:27).
Du Plessis (1996)
In a study outside the APT framework, Du Plessis attempted to establish the determi-
nants of industrial share prices on the JSE. The study covered the 1985 to 1994 period
and it was found that the Rand/US dollar exchange rate has the biggest effect on share
prices. The results showed a strong negative correlation between share prices and the
exchange rate and a significant correlation between inflation and share prices. Both
interest rates and economic production were found to have no significant correlation
with share prices (Brevis 1998:111 ).
Van Rensburg and Slaney (1997)

These researchers concluded that there are at least two and maybe three priced factors
on the JSE and suggested that the first two factors are a gold/mining factor and an
371
industrial factor. The third factor was thought to be a non-precious metal mining nature.
They found that the returns on the All Gold and Industrial/ Financial & Industrial Indices
can be legitimate proxies for the first two factors (Van Rensburg 1998:21-23).
Brevis (1998)
Brevis (1998:107,163,218) examined the relationship between 15 variables and the
industrial index of the JSE. The variables used in the study had either been identified
in earlier studies or were based on economic theory. The study covered the periods
January 1970 to September 1987 and January 1989 to July 1997 and it was found that
the relationship between the industrial index and some of the factors had changed
between the two periods. She concluded that this is proof that the nature of the priced
factors affecting share returns changes over time. The results of the examination of
unexpected changes in the potentially priced APT factors are summarised by Brevis
(1998:163-170) as follows:
• Both periods of the study show a positive, though statistically insignificant, re-
lationship between the JSE Industrial Index and the Dow-Jones Industrial Index.
• The first period show a statistically significant positive relationship between the
Rand/US dollar exchange rate and the industrial index, while the relationship is
negative and insignificant in the second period.
• The relationship between the industrial index and company liquidations was also
found to be positive, but statistically insignificant, for both periods.
• Both periods show a statistically significant relationship between the gold price
and industrial index, although the relationship weakened in the second period.
Brevis ( 1998: 165) suggests that this is due to gold having a less significant role
in the South African economy.
• A statistically significant negative relationship between short-term interest rates
and the industrial index was found in the first period, which changes to an
insignificant positive relationship in the second period.
• A similar relationship to the one of short-term interest rates was found for long-
term interest rates, although the positive relationship in the second period re-
372
mains statistically significant.

• The relationship between the term structure of interest rates and the industrial
index was also found to be similar to that found with short-term interest rates, ex-
cept that the first period yields a significant positive relationship and the second
period an insignificant negative relationship.
• GNP was found to have a statistically insignificant relationship with the industrial
index, the first period's being positive and the second negative.
• The industrial index shows a statistically significant relationship with business
confidence in both periods, although it was found to have weakened in the se-
cond period.
• The relationship between money supply and the industrial index was found to
have been negative and insignificant in the first period, which changes to a sta-
tistically significant relationship in the second period.
• Both periods show a statistically insignificant positive relationship between retail
sales and the industrial index.
• The relationship between wholesale sales and the industrial index was also
found to be positive, but statistically significant in the second period.
• The relationship between building plans approved and the industrial index was
shown to be positive, though statistically insignificant, in both periods.
• A similar, though negative relationship was found for buildings completed.
• The relationship between new car sales and the industrial index was found to be
positive and statistically significant in both periods, although with some weake-
ning of the rslationship in the second period.
Since the sensitivity of share returns to the various factors has been shown to change
over time, Brevis (1998:249,250) concludes that the research into the priced APT
should be performed on a continual basis. She adds that the factors identified above
should also be applied to all other sectors of the JSE and other factors, not yet tested,
should be identified in order to obtain even better explanations of share returns.
373
c. The APT versus the CAPM
Page (1987, 1989)

In his 1987 study, Page used JSE share data that did not include any known events and
concluded that APT-based models significantly outperform CAPM-based models (Laing
1988:vii,xi,82). The 1989 study used the weekly data from 30 randomly selected non-
thinly traded shares over the 1981 to 1984 period. He compared three CAPM-based
models with a two-factor APT-based model and concluded that the APT approach is su-
perior in measuring share price performance on the JSE (Page 1989:79,81; Van Rhijn
1994: 173, 174; Van Ransburg 1998: 18).
Laing (1988)
The study of Laing examined whether APT-based models can better explain share re-
turns than does the CAPM. The study consisted of two parts: the first part attempted to
establish whether significant differences in abnormal return estimates are obtained by
the different models and, the second part, whether the APT models provide smaller
differences between actual and predicted returns than the CAPM. Laing examined 79
shares which had experienced mergers activity during the March 1973 to May 1985
period and ignored the effect of dividends in the return calculations. The results of the
study are very different from the 1987 study of Page, in that it was not found that APT
models significantly outperform the CAPM (Laing 1988:xi,4; Van Rhijn 1994:171-173).
Laing (1988:xii,82,83), however, noted that the data used in the study contained real
events, while the 1987 data of Page data did not contain any known events. Thus, some
uncertainty exists regarding the fluctuations and distortions caused by the events and
the effect of the inclusion of some thinly traded shares. It was shown that the CAPM
minimises abnormal returns in comparison with the APT models, but Laing concluded
that the findings are inconclusive.
374

Biger and Page evaluated the performance of 25 South African unit trusts over the
February 1988 to March 1992 period and concluded that APT models with three to five
factors perform better in explaining the pricing and return of shares on the JSE, than
does a single-factor model (Brevis 1998: 102; Van Ransburg 1998: 19).
Page (1993, 1996)

The 1993 study confirmed earlier findings that the APT is superior to the CAPM in its
description of the return-generating process on the JSE. Page also suggested that the
APT provides significant insights into share price behaviour, even though the relevant
priced macro-economic factors have not been properly identified (Brevis 1998: 102).
Contrary to the 1993 findings of Biger and Page, the 1996 study of Page did not find
that the APT can explain the size and earnings anomalies on the JSE any better than
the CAPM can (Page 1996:27,31 ).
Various studies
The studies of Knight and Firer (1989), Smith and Chapman (1994), Garvin (1995)
and Meyer ( 1997) also suggested that APT-based models are able to provide better de-
scriptions of returns on JSE share portfolios than the CAPM can (Brevis 1998:6).
5.5.2 Conclusion
Brevis (1998:102, 103, 122, 123) summarises the results from the specified and unspeci-
fied APT factor research as follows:
• The combined tests of the APT and the statistical methods applied to determine
the number of factors and the sensitivity of shares to these factors generally
show that multiple systematic risk factors are priced and that they affect share
and portfolio returns.
• The specified factor tests, using firm characteristics and groups of portfolios,
provide some evidence that a multi-factor model has greater explanatory power
375
than a single-factor model.

• The use of hypothetical macro-economic variables in empirical tests provides the
strongest support for the APT's explanatory ability.
• There exists no single and unique group of APT factors, since these factors
change between markets and over time.
Elton and Gruber (1995:388) add that the evidence that groups of macro-economic
variables or portfolios are priced differently than implied by the CAPM, is significant
from both a theoretical and practical viewpoint, since it provides a more complete de-
scription of share returns. They warn, however, that these additional variables may have
been found to be priced, not because the APT correctly describes expected returns, but
because the market has been incorrectly identified in the construction of the APT
models used in the studies. They conclude that the APT provides a better explanation
of returns than the CAPM does and, hence, the use of multi-factor models in share
selection, portfolio management and portfolio evaluation is growing.
Adding to these conclusions, Reilly and Brown (1997:330) conclude that, since it has
as yet not been conclusively shown that one model is superior to the other, both the
CAPM and APT will probably be used and tested for the foreseeable future.
5.6 RECENT DEVELOPMENTS IN OPTIONS THEORY AND THE B-S

OPTION PRICING MODEL
Both Black ( 1990: 17) and Tompkins ( 1992: 11, 12) note that, since the breakthrough in-
sights into option pricing by Black and Scholes (1973) and their development of the B-S
model, the growth in options trading has escalated enormously and share options are
now traded in many exchanges around the world.
Except for the recent development of aspects such as share-index options and portfolio
insurance, the recent research has generally continued to focus on those areas investi-
376
gated in earlier studies. Again, the review of research is divided between studies on
options market efficiency and those on the validity of the 8-S model. In addition to these
studies, studies into the more recent introduction of share-index options and portfolio
insurance, and their possible role in the market crash of October 1987, are also exa-
mined.
Gemmill (1993:257-260) notes that the following aspects of options market efficiency
continue to be examined:
• Arbitrage tests looking for general pricing inefficiency.

• Model-based efficiency tests, whereby it is investigated whether option pricing
models can be used to identify mispriced options.
• Implied volatility tests, which attempt to establish whether implied volatilities pro-
vide good predictions of future volatilities and, hence, whether information is fully
reflected in option prices.
• Tests on whether options in general and the impact of their expiration have any
destabilizing effect on stock markets.
Recent examination of the 8-S model has to a large degree concentrated on implied vo-
latility tests. The stability of volatility is a key assumption underlying the model and the
behaviour of volatility has continued to be examined. Direct comparisons between
model values and actual option prices have also been performed on an ongoing basis.
The aim of the volatility tests is largely to establish the validity of the model, while the
direct comparisons are more concerned about possible misspecification.
Numerous studies have also been done on the methods and procedures for estimating
volatility and valuing newer exotic share options, interest rate options and currency op-
tions. These studies, however, fall outside the scope of this study and are not included
in the next section's review of recent developments in options theory and options
pricing.
377
5.6.1 Recent research related to the efficiency of options markets
a. Simple arbitrage tests
Chance (1988)
Chance examined Standard & Poor 500 index options and found that, after taking tran-
saction costs into account, the number of violations of the minimum bounds on call and
put prices to be insignificant (Gemmill 1993:257).
b. Model-based efficiency tests
Gemmill and Dickins (1986)

The UK study of Gemmill and Dickins found that the London market for individual op-
tions is only efficient when transaction costs are ignored (Gemmill 1993:258).
Kemna (1989)
Kemna examined the Amsterdam options market and found persistent biases between
market prices and prices provided by the 8-S model. Kemna, however, was unable to
establish whether these biases are due to market inefficiency or model misspecification
(Gemmill 1993:259).
Dawson and Gemmill (1991)

The study of Dawson and Gemmill found that the bid-ask spreads of London share-
index options are inefficiently set by market makers (Gemmill 1993:258).
Gemmill (1992)
Gemmill (1993:258,259) also found evidence of inefficiency in the pricing of London
share-index options. It was shown that the prices had not reflected the 1987 opinion-poll
information on the election and that almost-riskless arbitrage opportunities had existed
between the individual options and the index options (Gemmill 1993:258,259).
378
c. Implied volatility efficiency tests
Figlewski (1989)
Figlewski provided a warning about the use of implied volatilities. It may reflect the true
level of volatility, but may also reflect any misestimations inherent in the valuation pro-
cess (Reilly & Brown 1997:901 ).
Kemna (1989) and Scott and Tucker (1989)

Both these studies examined the forecasting performance of implied volatilities. Kemna
concentrated on the Dutch options market and Scott and Tucker studied currency
options. Their results confirmed earlier findings that implied volatilities provide better
projections of volatility than do historical volatilities (Gemmill 1993:259).
Leong (1990) and Robb (s.a.)

Leong (1990:60) noted that Robb had identified two implied volatility anomalies in
options markets. Firstly, a weekend effect, whereby the implied volatilities of at-the-
money call options on treasury bond features generally tends to rise from before to after
the weekend. Secondly, an economic statistics effect, whereby it has been observed
that the implied volatilities of listed options tends to drop, rather than increase, after the
release of important economic statistics.
Leong concluded that both these observed anomalies can be explained by the simpli-
fying assumptions of the B-S model. Most people trading in listed options use the basic
B-S model with calendar time as an input. However, economic time would be a more
practical measure, since it would provide implied volatilities which are more time-homo-
geneous. Weekends represent lapses in calendar time, but not of economic time, and
option values should therefore not change when everything else remains the same. The
lapses in calendar time, however, cause an artificial time decay in the B-S model and
result in lower option values, but if the markets are efficient, such time decays should
not affect option prices and implied volatilities.
379
The economic statistics effect was explained by the fact that the estimation of implied
volatilities, using the constant volatility method, are the market norm, while the moving-
average method provides a more exact picture of reality. The difference between these
two methods can be especially significant for short-dated options (most listed options)
and with the passage of important economic events, the moving-average volatilities over
the remaining lives of options should be substantially reduced. In efficient markets, any
anticipated increase in volatility should be incorporated into the pricing volatility and,
hence, implied volatility should not rise after the release of economic statistics which
have already been discounted by the market. Further, as the market applies the moving-
average volatility method, implied volatilities should drop after the release of the news
(in order to reflect the fact that one important day in the averaging process had passed)
(Leong 1990:60,61 ).
An alternative and equivalent explanation is linked to the weekend effect which has al-
ready been discussed. The release of the economic statistics represents a passage of
economic time, but not calendar time, and the implied volatilities has to drop to simulate
the fact that the 8-S model does not recognise the tim& decay in economic time (Leong
1990:61 ).
d. The impact of options trading on market efficiency
Snelling (1987)
This study examined the price behaviour of options and their underlying shares for 135
earnings announcements during the period August 1979 to November 1979. It was
found that the option prices lead the share prices in anticipating and reacting to the
announcements, but only by minutes rather than hours (Hodges 1988:27).
Stoll and Whaley (1987,1990)

Stoll and Whaley investigated the impact of the expiration of share-index options and
their 1987 study examined data for the years 1984 and 1985. They examined both
single days, on which only Standard & Poor (S&P)100 options expired, and triple days,
380
on which S&P 500 futures, S&P 500 futures options and S&P 100 options expired. It
was found that prices fell by 0,48% during the last trading hour of triple Fridays and that
this drop was reversed during the first half hour of trading on the following Monday.
There was also an increase in volatility on the triple Fridays, but the single Fridays
showed no significant changes in price or volatility. They did a follow-up study in 1990,
after the June 1987 switch of the settlement time of US contracts from the Friday closing
period to the Friday opening period. They found a drop in prices of around 0,3% on the
Friday opening, which is about half of the bid/ask spread of S&P 100 options (Gemmill
1993:262).
Nabar and Park (1988)

Using daily price data over the 1973 to 1985 period, Nabar and Park examined the im-
pact of the listing of 390 new share options. They found a drop in the volatility of the un-
derlying shares of between 4% and 8%, with the most significant impact four months
after the listing of the options (Hodges 1988:27; Gemmill 1993:261 ).
Van den Bergh and Kemna (1988)

In an examination of the expiration of options on 13 shares during the period January
1984 to October 1987 on the Amsterdam market, these researchers found that volatility
actually decreases on expiration (Gemmill 1993:263).
Conrad (1989)
Conrad used daily data to examine the impact of 96 new options listed during the 1974
to 1980 period. She found that when the options start to trade, the underlying shares'
prices on average increase relative to the other shares. In a further examination,
Conrad studied 76 of the underlying shares to establish whether there is any increase
in share price at the time of the announcement that trading of the options will com-
mence. She found that the share price volatility decreases after the introduction of the
options (Gemmill 1993:261 ).
381
Gemmill (1989)
In a study on the London market, Gemmill examined 1O shares and found a 17% drop
in the volatility of the share prices after the introduction of the options. In addition, 18
shares with options were compared with 18 shares, of similar size, without options.
Gemmill used three months of daily data around the crash of October 1987 and found
that the shares with options had a larger trading volume and higher volatility. Regres-
sion analysis was applied to establish the impact of these two variables and it he found
that options trading causes a 4% to 12% reduction in volatility (Gemmill 1993:261 ).
Skinner (1989)
Skinner examined 304 new options for the 1976 to 1986 period and found a significant
decline in share price volatility. Similar to Conrad in 1989, Skinner attributed the drop
in volatility to a reduction in company-specific risk since the beta values of the shares
remain unaffected. Skinner also found an increase in trading volume after the intro-
duction of the options and concluded that the drop in volatility had been caused by the
total market becoming larger and an improvement in liquidity (Gemmill 1993:261 ).
Lamoureaux (1990)
Lamoureaux suggested that a failure of earlier US studies had been that they had not
considered market-wide movements in specific risk over time. It was found that most of
the options examined in these studies had been introduced in the mid-1970s, prior to
a drop in company-specific risk. He concluded that, when the results were adjusted for
this drop in specific risk, the introduction of options did not have a destabilizing effect
(Gemmill 1993:262).
Pope and Yadav (1992)

Pope and Yadav examined the impact of individual options expiring on the London
market during the October 1982 to September 1987 period. Their data consisted of 46
expiration dates and 465 options of individual firms expiring and it was found that prices
fell by an average of about 0,5% at the time of expiration (Gemmill 1993:262).
382
5.6.2 Recent research related to the 8-S Option Pricing Model
a. Assumptions - Distribution and volatility of returns
Various studies
The studies of Fama and French (1987) and Poterba and Summers (1988) showed
that long-term volatilities have a lower mean than short-term volatilities and, hence,
long-term options may be priced with lower implied volatilities than short-term options.
They concluded that share prices revert towards a long-term mean over time. Kim,
Nelson and Startz (1991) suggested the results from these studies could have been
due to peculiarities of the 1926 to 1946 period examined (Gemmill 1993: 117).
Galai (1987)
From a review of empirical studies on option pricing, Galai concluded that, although the
B-S model can provide incorrect values when its assumptions are violated, these
deviations in value appear to be small and generally insignificant (Gemmill 1993: 118).
Hull and White (1987) and Johnson and Shapiro (1987)

Both these studies found that volatility does not appear to be constant over time. They
also showed that, if the changes are random, it will only have a small impact on option
prices (Gemmill 1993:117).
Jarrow (1987)
Jarrow concluded that when implied volatilities are used in the B-S model, these implied
volatilities would ensure that the model takes any generalisations of its assumptions into
consideration (Hodges 1988:26; Gemmill 1993:118).
Hodges (1988)
Hodges suggested that Jarrow's (1987) conclusion had been overly optimistic and that,
although implied volatility reflects market consensus, the market's consensus might ac-
tually be wrong (Hodges 1988:26).
383
Kemna (1989)
This study on Dutch options confirmed prior findings that return distributions, rather than
being normal, show fat tails. He did, however, not find any evidence of an exercise-
price-implied-volatility-bias (Gemmill 1993: 113, 116).
Gemmill (1991)
In a study on the London market, Gemmill found that low-exercise-price share-index
options generally tend to have higher implied volatilities than high-exercise-price share-
index options. It was also found that the degree of skewness changes on a month-to-
month basis (Gemmill 1993:116).
Leong ( 1991)
Leong noted that there are three basic difficulties associated with the estimation of vo-
latility. These are non-stationarity (volatility changes over time), non-uniformity (volatility
is higher on certain days than others) and mean reversion (the volatility time-series
moves away from extremely high or low values and reverts to some long-term mean).
Further, that there generally exists two approaches to estimating volatility, using histo-
rical data, in order to overcome the B-S model's problem with the fact that volatility is
not constant over time. These two approaches are:
• Using as much historical data as possible to ensure that sampling errors are
reduced.
• Using the most recent 30-day to 90-day data in order to eliminate the problem
of non-stationarity.
Leong, however, recommended that estimates should be made using both approaches.
These estimates should then be compared in order to establish whether volatility is rea-
sonably stable and, hence, whether it would be reasonable to estimate volatility using
historical data (Leong 1991 a:63,66).
384
Day and Lewis ( 1992)

Day and Lewis examined the implied volatilities which are derived from share-index
options and concluded that these implied volatilities are extremely sensitive to model
misspecification (Upsher 1993:8).
b. Assumptions - Transaction costs
Figlewski (1989)
Figlewski examined the effect of the rebalancing interval on Leland's (1985) conclusions
and showed that, for US share-index options, rebalancing risks are not significant,
except in illiquid markets where hedges are required for long periods. Thus, narrow
spreads will be found in active options markets, while the spreads demonstrated by
Leland can be found in some OTC markets (Gemmill 1993:123,124).
Hodges and Neuberger (1989)

These researchers concluded that Leland's (1985) formula leads to an increase in the
values of call options and that it overestimates transaction costs (Gemmill 1993:123).
c. Comparative tests
Geske and Torous (1988)

Geske and Torous (1988:50-58) examined 2 323 call options on non-dividend paying
shares over the period August 1976 to October 1977, in order to establish whether the
B-S model misprices options. They used a random sample of CBOE call option tran-
sactions data for one date per month to minimise any problems of non-simultaneity
between share and option prices. The following summarises Geske and Torous's re-
sults:
• When the underlying share sample's standard deviation of returns is used to

estimate volatility, the B-S model systematically undervalues calls with respect
to volatility and skewness.
385
• The 8-S model undervalues options on low-volatility shares and overvalues

options on high-volatility shares.
• The 8-S incorrectly values call options with longer than 30 days to expiration.
• The 8-S model provides incorrect values for deep-in-the-money and deep-out-of-
the-money call options, although the degree of error does not appear to be
economically significant.
d. South African research
Hawinkels (1987)
Hawinkels (1987:9, 12-17,26,35) examined close to 160 price relatives over a seven
month period and found that returns are not normally distributed. Although this can have
significant implications in applying the 8-S model, the extent to which this result affects
the accuracy of the model had not been established in the study. The study also exa-
mined the estimation of volatility in the South African market and found that the high-low
estimator appears to provide better predictions of volatility than does the close-to-close
estimator. He also identified a definite need for comprehensive comparative option pri-
cing model research for the South African market.
Beaven (1990)
Beaven compared actual with theoretical option prices for options with less than 60
days to maturity. Only small differences were found and it was concluded that these
differences cannot be attributed to market efficiency (Bird & Page 1991: 14 ).
Bird and Page (1991)

Bird and Page (1991 :14,20,21) found large differences between theoretical and actual
option prices in the South African market and concluded that the market is inefficient in
the pricing of options. They also found that the closer the options come to maturity the
less inefficient the market becomes in its pricing, and concluded that this explained
Beaven's (1990) findings. They suggested that, although this inefficiency provides profit
opportunities for arbitrageurs, if the market takes the unavailability of realistic arbitrage
386
strategies into account, then its pricing is actually efficient as the differences will not be
exploitable. It was also suggested that the thin trading of these options is a possible
cause of the perceived inefficiency.
Wood (1997)
Wood noted in her article that the trading in share options, which had recently become
available on the South African Futures Exchange (SAFEX}, had reached the R13,5 mil-
lion mark and that it was expected that the number of options would be increased from
six to ten in the near future (Wood 1997:93).
Delport (1998, 1999)

Delport noted in 1998 that 375 million options with a value of R367 million had traded
to date. It was also noted that options have increased in popularity with South African
investors and that the options section is one of the most active sectors on the exchange
(Delport 1998:5). In the first of two 1999 articles, Delport concluded that it was likely that
more than half of the 10 listed options that will reach maturity on 16 September 1999
will expire out-of-the-money. Most of these options are call options and it was found
that, similar to the Australian market where around 80% of options are sold before
maturity, most investors sell their options before expiration (Delport 1999a:57).
In the follow-up article, Delport noted that trading in listed options had increased from
four million in October 1997 to 112 million in June 1999. The number of listed options
had also increased to 66 and it was found that investors still prefer call options over put
options (Delport 1999b:71 ).
5.6.3 Share-index options, portfolio insurance and the market crash of October
1987
Since 1973 , both put and call options have been available on organised exchanges
and, although the way they have been viewed has since changed, they have been po-
pular with investors ever since. They were initially viewed as speculative instruments
387
and were generally purchased to obtain leverage benefits. In 1983, share-index options
were introduced and these also appealed to investors, since they provided investors
with the opportunity to trade in industries or the market as a whole, while also providing
the benefits of leverage. Since 1984, portfolio insurance, using share-index options,
have been used to increase portfolio yields and many investors sold put options to take
advantage of the rising share prices of the 1980s. These strategies worked wel I until the
crash of October 1987, after which, due to the losses incurred, options were again
viewed as purely speculative instruments. It was only in the early 1990s that options
trading had reached its pre-crash levels and were again considered to be an essential
element in strategic portfolio management (Gemmill 1993:149; Jones 1998:551 -555).
5.6.3.1 Share-index options

Share-index options are options on stock market indices and provide investors with the
opportunity to invest in options on a portfolio of shares (Levy & Sarnat 1994:633). They
can be used to increase (speculate) or decrease (hedge) the exposure to a particular
market (Gemmill 1993:151; Jones 1998:555).
Share-index options also have fixed exercise prices and expiration dates and allow in-
vestors who are bullish about the market to buy a call on the market index, while those
who are bearish can buy a put on the market index. This has the advantage that in-
vestors do not have to make an industry or individual share decision, but only a market
decision. They do, however, differ from options on individual shares in that they do not
require the delivery of shares on expiration, but the buyers receive cash from the sellers
upon exercise of the'contract. This settlement amount is the difference between the
closing price of the index and the strike price of the option (Jones 1998:552).
As with the valuation of individual share options, the 8-S model needs to be adjusted
for dividends, transaction costs and the return distribution of the index when valuing
share-index options (Gemmill 1993:155).
3BB
a. Dividends
Since an index consists of a number of individual shares, the level of the index will drop
on the shares' ex-dividend dates. Also, the larger the number of shares in the index and
the more frequent the dividend payments, the smaller will be the effect on the level of
the index at any one time. Thus, early exercise of index call options will not really be
worthwhile where the dividends are a relatively smooth series of payments. In the va-
luation of share-index options, however, the values based on continuous dividends may
actually be quite different from those based on actual dividends and, also, the problem
of undervaluation of American put options still remains (Gemmill 1993:155-157).
b. Transaction costs
Since the 8-S model is based on an arbitrage position between the option and the un-
derlying share, and since it is difficult to simultaneously trade in all the shares in the
index, transaction costs have a larger impact on share-index option values than options
on individual shares. Hence, market makers in options generally hedge their portfolios
with index futures which have very low transaction costs (Gemmill 1993:157).
Yadav and Pope (1990)

Yadav and Pope found that, for both the UK and US markets, the present value of the
futures price are lower than the spot index. This is opposite to what had been found in
Japan, and this disequilibrium creates the problem of knowing whether option values
should be based on the present value of the futures or upon the spot index. It was found
that market makers generally use the present value of the futures price, since they are
also using the futures to hedge their positions in options (Gemmill 1993: 157).
c. Return distribution of the share index

An assumption of the 8-S model is that share returns are lognormally distributed and
if share returns do not follow such a distribution, then the returns on the share index
(the sum of the log normal distributions) can in theory not be lognormally distributed. The
same principle also applies to the situation where share returns are normally distributed
(Gemmill 1993: 157).
389
Badrinath and Chatterjee (1988)

Their study provided evidence that the distribution of share-index returns is close, if not
closer, to a normal distribution than those of individual shares (Gemmill 1993: 158).
Gemmill (1993: 158, 159) concludes that, where dividend payments are a smooth series,
Merton's dividend-adjusted model should be used to value European share-index op-
tions. But, since share-index options are American by nature and dividend payments are
lumpy, a pseudo-American adjusted 8-S model or the binomial model would be pre-
ferred for valuing calls. The 8-S model has the advantage that it is much faster than the
more accurate binomial model. For puts, however, it is advised that the binomial model
be used.
d. Share-index options, share-index futures and the market crash of October 1987
As previously noted, share-index options prices are more likely to wander around their
fair values than would individual option prices, because the higher transaction costs
associated with the index options makes arbitrage with share-index options more expen-
sive. Hence, market makers prefer using futures for hedging their option positions and
futures index prices may consequently have a strong influence on index option prices.
Therefore, by implication, if it was found that share-index futures had a role in the crash
of October 1987, share-index options, through their close relationship with share-index
futures, may also have played a role.
Where there is drop in the prices of share-index futures, investors can use index arbi-
trage - an arbitrage between share-index futures and the underlying shares - to sell the
underlying shares and buy the futures and, thus, gaining the difference in price between
the two. Through this action a link between the share prices and share-index futures
prices is created and index arbitrage has the benefit that it brings the cash market and
the futures into line (Levy & Sarnat 1994:645).
Levy and Sarnat (p.645,646) summarise that, despite this positive role of index arbi-
trage, futures have been blamed for the October 1987 crash, in that the futures and
390
share markets had become unhinged because brokers had not been able to cope with
the number of sell orders. This had been caused by the fact that shares had not been
traded on the NYSE by the morning of 20 October 1987, resulting in the sell orders
being transferred to the futures market and ending with share-index futures selling at
a discount of 22%. This had provided significant profit opportunities for arbitrageurs,
which could not be exploited since the shares could not be sold, with the result that
arbitrageurs did not enter the market. Thus, the drop in the futures market had not been
the cause of the crash, but had only reflected the panic in the stock market. There is
general agreement that the futures market plays a positive economic role and index
arbitrageurs play a beneficial role in linking the futures and cash markets.
Levy and Sarnat (p. 633) add that share-index options also play a positive economic
role, in that, rather than buying numerous individual options, a portfolio of options can
be obtained with the purchase of a share-index option. Share-index options also play
a significant role in facilitating portfolio insurance, which is examined in the next section.
5.6.3.2 Portfolio insurance

Portfolio insurance is an investment strategy whereby portfolios are hedged in such a
way that the portfolios will yield a minimum return, while also providing the opportunity
to gain the additional benefits of rising share prices. Portfolio insurance is usually ob-
tained with either share-index options or share-index futures (Levy & Sarnat 1994:633;
Jones 1998:541 ), but in this section only the use of options will be considered.
In situations where investment funds invest in large numbers of shares and fund mana-
gers wish to protect their funds against share price declines, put options can be bought
on every share in the portfolio to protect the fund. However, this is usually not possible
in practice since options are not available on all shares and, hence, the insurance effect
is attained with the purchase of share-index put options. The higher the correlation
between the returns of the share index and the share portfolio, the greater the protec-
tion offered by the portfolio insurance (Levy & Sarnat 1994:633).
391
Jones (1998:541) adds that the principle of portfolio insurance is quite simple. A protec-
tive put option is bought to ensure that the portfolio can be sold at a price sufficient to
provide the minimum return. The remaining portfolio funds are invested as normal and
the put option protects it against declines in share prices, since the value of the portfolio
must be equal to or exceed the exercise price of the put option at the end of the invest-
ment period - otherwise the put option will be exercised, yielding the minimum return.
Leland (1988:119, 120) notes a further way by which portfolio insurance can be attained
with the use of options. This technique is based on the Black and Scholes (1973) option
pricing theory, whereby options can be priced through arbitrage and options can be
replicated through dynamic rebalancing between shares and the risk-free asset. Shares
are sold when prices drop, meaning that losses are limited, and shares are bought when
share prices rise, that is, upside opportunity cost is limited to the equivalent of the
option price. This strategy, therefore, represents a series of stop-loss orders and these
stop-loss orders are reversed when the market starts moving up again. Under ideal con-
ditions such a hedging strategy would provide the same portfolio insurance as that of
buying a put option.
• Portfolio insurance and the market crash of October 1987

To ensure that the desired level of protection is achieved, portfolio insurance policies
are programmed to buy shares when prices rise and sell shares when prices drop.
These automatic computer-generated buy and sell orders have also been blamed for
the crash in the stock markets. Levy and Sarnat ( 1994:634) provide the following sum-
mary of the possible role of portfolio insurance in the October 1987 crash.
To protect against losses, portfolio insurance can be achieved by selling share-index

options and/or shares after the market has dropped and buying them after the market
has started to rise. By 19 October 1987, $60 to $90 billion's worth of assets had been
insured and all these assets had been managed using program trading, which is quite
similar to stop-loss selling (a prearranged order to sell shares when their prices had
fallen to a certain level). Many fund managers had used program trading and when all
392
the sell orders came through on the same day, not enough buyers had been available
and the share prices dropped sharply. This had increased the number of sell orders,
causing even further drops in prices and, hence, portfolio insurance had not been able
to provide the required protection. It is still unclear how big a role portfolio insurance
played in the crash, but it still remains a powerful investment tool.
Constantinides (1987)
Constantinides suggested that an increasing number of investors had decided to adopt
hedging strategies prior to the market crash of October 1987 and found that many of the
portfolio insurers' clients had "ratcheted" up their protection during the preceding bull
market (Leland 1988:112,121).
Shiller ( 1987)
Shiller found that more investors had been using informal hedging, such as stop-loss
orders, rather than formal portfolio insurance on 19 October 1987 (Leland 1988: 105).
Leland (1988)
According to Leland (1988: 105, 109), the evidence suggests that only about 10% to 15%
of the trading on 19October1987 had been related to portfolio insurance. The fact that
portfolio insurance had also not been available with the crash of28October1929would
seem to suggest that it can not be blamed for the 1987 crash. However, although port-
folio insurance had not been available in 1929, closely related hedging techniques such
as stop-loss selling had been used by investors. Leland suggested that a large number
of hedgers had increased the volatility of the market and showed that, even though no
new information had been released, such an increase in volatility could have led to a
market crash.
Leong ( 1991)
According to Leong (1991 b:85}, portfolio insurance would still have failed as a hedge
strategy, even if the cash and futures market had not become decoupled and the market
had not become illiquid with the crash of October 1987. The reason, Leong suggested,
393
was that portfolio insurance would have failed anyway with such an increase in market
volatility, since the principle of a risktess hedge has not been designed to perform well
in extremely volatile markets.
5.6.4 Conclusion
The evidence from simple arbitrage tests show that options markets can generally be
considered as being efficient (Gemmill 1993:258). Both model-based studies and im-
plied volatility efficiency studies have presented evidence of pricing inefficiency, but
since the models used (mostly the 8-S model) are based on simplifying assumptions
and volatility needs to be estimated, these results are inconclusive since they are joint-
tests of the model and market efficiency or the method of estimating volatility and mar-
ket efficiency (Jones 1998:550). The results from the empirical studies on the effect of
options trading and the expiration of options on the underlying shares have provided
some evidence of inefficiency. Gemmill (1993:261,262) summarises that options trading
has generally not been found to significantly increase volatility and that it has led to a
decrease in specific risk. However, there is some evidence, though not strong, that
options trading has led to an increase in trading volume and market liquidity. Van Horne
(1992:263) adds that there is evidence that the expiration of options has a temporary
effect on share prices and volatility, but also that the effect is rather small and soon
reversed after expiration.
From the evidence provided by the studies on the B-S Option Pricing Model, Gemmill
( 1993: 118, 124) concludes the following:
• There is clear evidence that the returns distribution has fat tails and that options
are valued accordingly.
• The evidence on skewness of the distribution of returns is less clear. There may
be a case for higher volatilities for low-exercise-price options in order to reflect
the effect of leverage, but the studies have also shown that skewness tends to
change from period to period.
• There is also evidence that short-term volatilities are higherthan long-term vola-
394
tilities, and that volatility may revert to some mean over time.
• Bid-ask spreads on options appear to be relatively large, but do not necessarily
indicate inefficiency in pricing. Most of the problems associated with transaction
costs and the distribution of returns can be eliminated with adjustments to the
volatility estimates.
• The B-S model is valid and robust, and more complicated models have parame-
ters which are not well-defined and difficult to estimate.
Further, Jones (1998:550) concludes that the empirical studies generally support the
efficiency of options markets and the validity of the B-S model. Although some biases
have been identified, these can largely be attributed to difficulties experienced with the
estimation of volatility and, hence, can not be seen as conclusive evidence of model
invalidity or market inefficiency.
Very few empirical studies on the South African options market exist and, since there
is clear evidence that options are becoming more popular and that options trading has
shown significant increases, there is a distinct need for studies on both options market
efficiency and the robustness of the B-S model under South African conditions.
Reilly and Brown (1997:907) conclude that the B-S model, in a dividend-adjusted form,
is also quite suitable for the valuation of share-index options. Since share-index options
are well-diversified portfolios, the volatility of the share index's price is usually signifi-
cantly lower than that of individual shares. The applicable dividend yield might also be
readily available, since it can be assumed to be the average of the annualised yield on
the index during the options life.
It has been shown that share-index futures and, due to their close relationship, by im-
plication also share-index options, had not played a significant role in the market crash
of October 1987. However, significant evidence has been provided from which it can be
argued that portfolio insurance, and the use of share-index options therein, had played
a significant role in the market crash.
395
CHAPTERS
SUMMARY, CONCLUSIONS AND RECOMMENDATIONS
6.2 Summary of the literature study 397
6.2.1 Accounting Theory and investment decision-making 397
6.2.2 Portfolio Theory 398
6.2.3 Market efficiency and the EMH 400
6.2.4 The CAPM 401
6.2.5 The APT 403
6.2.6 Options and the 8-S Option Pricing Model 404
6.3 Conclusions 406
6.3.1 General 406
6.3.2 Portfolio Theory 407
6.3.3 Market efficiency and the EMH 407
6.3.4 The CAPM 408
6.3.5 The APT 409
6.3.6 Options and the 8-S Option Pricing Model 409
6.4 Recommendations 410
6.5 Summary 411

396
6.1 INTRODUCTION
The purpose of this study has been to establish whether the existing capital market
theories and pricing models provide a reliable description and explanation of the wor-
king of the capital market; whether the recent studies and research have led to any new
developments in the field of capital market theory and pricing models; whether the re-
search can be summarised in order to provide a framework for future research; and into
which areas future research, especially South African research, should be directed in
this field of study.
To this end, the literature study, firstly (chapter 2), examines the capital market theories
and pricing models in the context of accounting theory and investment decision-making.
The role of these theories and models in the formulation of accounting theory and their
importance for investment decision-making are reviewed in this first section. The
second part of the study (chapters 3 and 4) entails a review and examination of the
historical background to and the principles and implications of the various theories and
pricing models. Included in these sections of the study, are the reviews and summaries
of the research conducted since the development of these theories and pricing models
up to the early 1980s. The final section of this study (chapter 5) concentrates on a
review and summary of the more recent research and empirical studies. An examination
of the recent developments in the capital market theories and pricing models is also
included in this final section of the literature study.
This chapter consists of a summary of the main aspects of the literature study, Followed
by a summary of the main conclusions drawn from the review of the research con-
ducted. It ends with recommendations for future research studies in this field of study,
especially under South African conditions.
397
6.2 SUMMARY OF THE LITERATURE STUDY
The whole study consists of a literature study of the main capital market theories and
pricing models, in the context of accounting theory. Before conclusions can be drawn
and recommendations made from the review of empirical studies and other research,
it is first useful to briefly summarise the current knowledge on these theories and
models.
6.2.1 Accounting theory and investment decision-making
Although there exists no single comprehensive theory of accounting, a collection of

theories has resulted from the use of different approaches to the formulation of an
accounting theory. No matter which approach is followed, an accounting theory must
be confirmed before it can be accepted (Belkaoui 1992:58,65) and a theory can gene-
rally only be confirmed through empirical testing. The aim of the empirical testing is not
so much to establish whether the theory is true, but rather if it works. As Rees (1995:39)
notes, it is not the validity of a theory or model's assumptions that 1s important, but
rather its descriptive and predictive power. When applied to accounting and investment
decision-making, the process begins with the formulation of a theory and ends with the
empirical testing thereof to establish its validity or falsity (Kam 1990: 511,512).
One of the approaches to the formulation of accounting theory is known as the predic-
tive approach, that is, it uses the principle of predictive ability to choose among the op-
tions which are available to predict the events that are of interest to investors and other
related parties. One stream of the predictive approach to the formulation of an accoun-
ting theory is known as the market-based approach. This approach 1s concerned with
the ability of accounting information to explain and predict capital market reaction to
accounting disclosure (Belkaoui 1992: 139).
Rees (1995:153) identifies two crucial aspects of the market-based approach to the for-
mulation of an accounting theory. Firstly, there are the pricing models (CAPM, APT and
398
the B-S model}, which are based on the principles of Portfolio Theory. These models
are used by investors, analysts, portfolio managers and academics to make sense of
the important investment decision-making variables, that is, security prices, risk and ex-
pected return. The second aspect is the EMH, which asserts that market prices rapidly
incorporate all publicly available information, including accounting information, in an
unbiased manner. The pricing models also have a role to play in the empirical testing
of the EMH, since they provide the benchmark values against which market prices are
measured.
The investment decision is a complex one and the solution to the problem may be a
product of many variables. However, the essence of the decision is that an investment's
expected return should provide adequate compensation for the risk involved (Broadbent
1992:8). Portfolio Theory and the asset pricing models are concerned with risk and ex-
pected return and the main advantage of the CAPM, the APT, the B-S model and the
supporting EMH and Portfolio Theory is that they are empirically testable (Hendriksen
& Van Breda 1992: 185). These pricing models and related capital market theories, as
elements of the market-based approach to the formulation of an accounting theory, and
the empirical testing thereof are the objects and main focus of this study.
6.2.2 Portfolio Theory
The implications of Portfolio Theory, and in particular the benefits of diversification and
the insights it provides into the pricing of shares, are key fundamentals for investment
decision-making, portfolio management and investment risk management.
Portfolio Theory deals primarily with investors' main concern about how risk will affect
their total wealth. Hence, Portfolio Theory only evaluates the risk of individual invest-
ments to the extent to which they affect or contribute to the overall risk of investment
portfolios. The concept of diversification flows directly from this principle and from the
fact that there is a degree of independence among the risk sources which affect the
shares in portfolios. Hence, according to the diversification concept, the risk of a port-
399
folio will show a rapid decline with the initial addition of shares to the portfolio and, also,
that only a small number of shares will provide substantial risk reduction benefits (Jones
1998:178-181).
Since the principle of risk reduction through diversification allows investors to limit or
minimise their exposure to risk without adversely affecting their returns, the concept of
diversification can be viewed as providing the key to the management of investment
portfolios and portfolio risk. However, it is important to note that, since all shares are to
some degree affected by common sources of risk, risk cannot be completely eliminated.
Risk can thus be divided into two components, namely systematic (non-diversifiable)
risk and unsystematic (diversifiable) risk (Jones 1998:180, 181 ).
Although the concept of diversification has significant benefits, it does not provide a di-
rect method for investors to select efficient and optimal portfolios. Efficient portfolios
are evaluated according to expected return and risk (as measured by the standard de-
viation) and are those portfolios which provide the highest return for a given level of risk
or the lowest level of risk for a given return. From the set of available efficient portfolios,
investors are then able to select the optimal portfolio, which best suits their needs and
risk preferences, using the principles of utility theory (Jones 1998:204-206).
Apart from the risk reduction and portfolio selection benefits, Portfolio Theory also has
important implications for the pricing of shares. Since only part of the risk of shares can
be eliminated through diversification, investors need to focus on the non-diversifiable
systematic component of risk, because this is the component of risk that should be
priced in stock markets. The risk that is relevant to an individual share can be estimated
by measuring its contribution to the overall risk of a well-diversified portfolio. Against
this, the share returns that can be expected on the basis of this contribution to risk need
to be estimated using the CAPM or APT, and in the case of share options the B-S
Option Pricing Model (Jones 1998:219-221).
400
6.2.3 Market efficiency and the EMH
Some of the most interesting and important empirical research has analysed whether
capital markets can be regarded as efficient. Although the efficiency of capital markets
remains a controversial area in investment research, it has real-world implications for
investors, analysts and portfolio managers (Reilly & Brown 1997:208).
The modern version of market efficiency states that market prices adjust rapidly, al-
though not instantaneously, to new information becoming available and that the price
adjustments occur in a correct and unbiased manner, that is, the errors in adjustments
on average balance out. The EMH is concerned with the extent to which market prices
fully and quickly reflect new information and can be divided into three forms of efficien-
cy, namely weak form, semi-strong form and strong form efficiency. Each of the three
levels of market efficiency relies on different levels of information-processing efficiency
and each requires different types of tests to establish its validity. The key to the empiri-
cal testing of the EMH is to establish whether investors can consistently earn abnormal
returns, that is, returns in excess of those commensurate with the level of risk involved.
Short-lived random inefficiencies and one-off short-term unusual returns do not consti-
tute evidence of market efficiency in an economic sense (Jones 1998:255-259).
Pike and Neale ( 1996: 42, 43) and Jones ( 1998: 266-269) summarise the fol lowing impor-
tant implications of EMH market efficiency for investment management:
• Market efficiency implies that fundamental analysis, that is, the study of the in-
trinsic value of securities, will have no value unless the analyst has access to
inside information or has superior analytical ability.
• The EMH also casts doubt on the value of technical analysis, that is, the study
of patterns or trends in price movements over time.
• Active investment management will also have no value in efficient markets and
passive buy-and-hold strategies will, on average, outperform active investment
strategies. This, however, does not imply that investment and portfolio manage-
ment has no value. The principles of efficient diversification and risk assessment
401
must still be applied and the level of transaction costs must still be monitored.
It should, however, be noted that the strong form of the EMH cannot hold completely,
since it is clearly unrealistic to assert that market prices can fully reflect all information,
including non-public inside information. Further, although the evidence on market effi-
ciency largely supports the EMH, especially at the weak form and semi-strong form
levels, several anomalies have as yet to be satisfactorily explained. These anomalies
tend to suggest that there are opportunities to exploit for those investors and investment
managers who have share selection abilities (Jones 1998:258,267).
Despite some anomalous evidence of market inefficiency and irrationality, a large body
of research tends to suggest that market prices are fair and generally reflect the value
of shares, given the available information. This implies that market prices do appear to
respond very quickly to new information, that market prices are difficult to predict based
on publicly available information and, if the market misprices some shares, that there
is no obvious way of identifying mispriced shares (Ross et al. 1996:294).
6.2.4 The CAPM
The fundamental principle underlying the CAPM is that there is a linear relationship
between systematic risk (as measured by beta) and expected returns (Arnott 1993: 16).
The CAPM, building on the insights of Portfolio Theory, attempts to explain this rela-
tionship between expected return and risk and assumes that investors will only invest
in risky assets if a sufficient return can be gained as compensation for the risk taken
Another fundamental assumption of the CAPM is that investors hold well-diversified

portfolios and that they can construct optimal portfolios by combining investment in the
risky market portfolio with borrowing or lending at the risk-free rate of interest. Thus it
has been concluded that investors are concerned only with how a share contributes to
the risk of the portfolio and, based on the principles of Portfolio Theory, that with the
402
assessment of the risk of a portfolio, investors are concerned only with the non-diversi-
fiable systematic component of risk. The CAPM uses beta as the measure of risk that
cannot be eliminated through diversification and describes expected returns as being
the return on the riskless asset plus a risk premium. This risk premium equals the
market return in excess of the risk-free rate times the share/portfolio's beta. Beta,
therefore, explains the differences between shares/portfolios' expected returns (Harring-
ton 1993: 1).
The CAPM has been the subject of a vast amount of theoretical investigation and
empirical research and the controversy surrounding it continues. Part of this controversy
is concerned with whether beta is the only measure that explain returns, and some
evidence suggests that variables such as dividend yield, P/E ratios, firm size and the
BV/MV ratio provide additional explanatory power. Even more controversial is that some
studies have failed to show a significant relationship between beta and expected returns
(Harrington 1993:1 ).
Another problematic aspect of the CAPM is whether the use of proxies for the market
portfolio casts doubt on the usefulness of the results obtained from empirical tests of the
CAPM. There are two dimensions associated with the fact that the true market portfolio
is unobservable. In the first instance, the market proxy may be close to being mean-
variance efficient and the tests may thus yield results in support of the CAPM, even
though the CAPM relationships may not hold. Secondly, the choice of market proxy may
be the cause of the observed anomalous results and beta may in fact provide a true
description of expected returns and. Hence the CAPM might be true (Harrington
1993:2).
Another question is whether the CAPM is correct in its description of expected returns
or whether the alternative model, the APT, may describe returns better. The compa-
rative tests between the two models and their ability to describe anomalous results
depend, however, largely on the efficiency of the markets, hence no definite conclusions
can be drawn from such tests. Closely linked to this problem, is the fact that some
403
anomalies have not persisted consistently over time and that some of the evidence
brought against the CAPM may have been unearthed through data mining or data
snooping (Harrington 1993:2).
An even more fundamental problem is that the statistical tests which have been used
to examine the CAPM may not have been implemented or interpreted correctly. This
relates to problems whereby parameters of the CAPM are incorrectly measured or esti-
mated and whereby the statistical significance of the results is either overstated or
underemployed in comparison with their economic significance (Harrington 1993:2,3).
Important to the whole controversy surrounding the CAPM, is Arnott's (1993:22) conclu-
sion that, since the CAPM is an expectational model, its linear relationship between
beta and expected returns can never be proved or disproved.
6.2.5 The APT
The APT was developed a number of years after the CAPM, and this alternative asset
pricing model suggests that share returns are a function of various common risk factors,
not only beta. The APT is a more general theory than the CAPM, but it has been found
that the two models can be shown to be identical in conditions where there is only one
risk factor (Arnott 1993: 16; Jones 1998:247).
One of the advantages of the APT is that it is not dependent on the existence of the
CAPM's market portfolio, but has the problem that the priced risk factors are not well
specified, especially on an ex ante basis. The evidence from empirical testing suggests
that several risk factors are priced, generally between three and five, and also that
some of the factors identified are not compatible with the theory of market efficiency
(Arnott 1993: 16; Jones 1998:247). Arnott, however, notes that the APT, rather than
being seen as an asset pricing model, can be viewed as a risk model, that is, it explains
the risk factors that affect returns, but is somewhat silent on the pricing of shares and
expected share returns.
404
Various criticisms have also been levelled at APT studies that have used factor analysis
methodology. These criticisms are based, firstly, on the fact that there exists no ade-
quate factor extraction rule, and hence, different numbers of risk factors are extracted
and found to be priced for different sample sizes and different numbers of observations.
The second criticism is that there is no guarantee of risk factor consistency across
different portfolios. Thirdly, and possibly the most serious criticism, is that factor ana-
lysis provides statistical risk factors which may not have any observable relationship
with any economic variables (Van Ransburg 1998:20,21 ).
These criticisms have led to alternative tests of the APT, whereby macro-economic vari-
ables are preselected and tested to establish whether significant risk premiums are
associated with these macro-economic factors. The benefit of these tests is that the
identified priced risk factors have economic significance (Van Ransburg 1998: 16).
6.2.6 Options and the 8-5 Option Pricing Model
Since the 1970s, and with the development of sophisticated capital markets, options
have become indispensable financial instruments, not only to manage risk but to also
fulfill the need to manage uncertainty. The management of risk has led to the develop-
ment of Portfolio Theory, the CAPM and APT. However, although these theories and
models are elegant, they rely on investors' expectations of risk, expected return and the
correlation between risk and return. These theories and models did, however, not solve
the problem of uncertainty, since investor expectations about uncertainty are generally
not the same, hence the problem remained of how to determine the value of uncertain
securities (options) upon which all investors could agree (Tompkins 1992:11, 12).
The problem of valuing options, both call (buy) options and put (sell) options, of either
an American or European nature, was solved with the development of the B-S Option
Pricing Model. The model simply removes investor expectations about uncertainty, that
is, share prices are based on expectations about uncertainty and options on those
shares are based on expectations about the share prices, and by simply subtracting the
405
one from the other, the expectations about uncertainty are eliminated. This procedure
ensures that it is a relatively simple matter to determine option values, that is, the price
of the share minus the price of the option on that share should equal a riskless in-
vestment (an investment with no uncertainty) and when this equation is rearranged the
value of the option can be calculated (Tompkins 1992: 12).
Empirical testing has shown that it is rare for B-S option values to be equal to the actual
market prices of the options. Black (1988:51) notes three reasons, other than errors in
calculation, that generally account for such differences, and adds that a combination of
the three usually explains the differences between value and price:
• The option values calculated using the B-S model may be correct and the market
prices may be out of line.
• Some of the different inputs, especially volatility, used when applying the B-S
model may be wrong.
• The B-S model in itself may be wrong.
A particular problem, other than the effect of the early exercise of American options, as-
sociated with the B-S model, is estimating the underlying share's volatility. This has a
tremendous impact on the values of certain options and is a significant area of empirical
research (Black 1988:52,53). Another important topic related to the B-S model is the
efficiency of the options market, and with it that of the stock market, and the related
effect on option prices compared with option values (Reilly & Brown 1997:873). Signi-
ficant empirical testing has been performed on this and the related implications for
option pricing theory, and this represents an area of vital importance for the field of
derivative financial instruments (Hodges 1988:8).
Although the B-S model is based on several unrealistically simple assumptions, Black
(1990: 17) notes that this weakness is also its strength since it allows the average inves-
tor to use it. Black (1988:51) adds, however, that it is not the realism of its assumptions
that is important, but rather how well it works, and as yet no other model has been
developed that has the descriptive and predictive power of the B-S model.
406
6.3 CONCLUSIONS
6.3.1 General
With reference to the problems defined in chapter 1 of this study, t.he following general
conclusions can be drawn:
• Although empirical studies have provided some evidence of anomalies and while
there are some controversial issues still to be resolved, the body of evidence
generally confirms that the main capital market theories and pricing models
reviewed in this study do provide a reasonably accurate description of reality.
• The literature study has been able to provide an exposition of how each of the
theories and models describes and explains the nature and working of the capital
market.
• Although some imperfections have been identified, research has shown that the
theories and models are valid and do have descriptive and predictive ability.
• Recent research has made significant contributions to the available body of

knowledge and, as regards the EMH, some alternative theories have been
posited. It is, however, still unclear whether the CAPM or APT is superior and
which provides the best description and explanation of share returns.
• With the review of empirical research, especially recent research, it has been
possible to provide a framework and establish specific areas for future South
African research in this field of study.
It is, however, also useful to draw specific conclusions about each of the main capital
market theories and pricing models reviewed in this study. The following sections deal
with these specific conclusions, starting with Portfolio Theory.
407
6.3.2 Portfolio Theory
• Investors can reduce the level of investment risk through diversification and can
thus reduce the level of risk associated with a given expected return.
• The major risk reduction benefits of diversification can be achieved relatively

quickly, that is, most of the benefits can be obtained when portfolios consist of
between 10 to 20 shares.
• Global diversification enables investors to improve the risk/return characteristics

of their investment portfolios. Investors can obtain lower risk levels by diversi-
fying globally or can earn higher returns for a given level of risk, compared with
the returns that can be achieved by investing only in their domestic markets.
6.3.3 The EMH
• The empirical evidence generally supports the weak form of the EMH. Despite
some evidence of mean reversion, excess volatility and market irrationality, it is
reasonable to conclude that capital markets, including the JSE, are weak form
efficient.
• Regarding semi-strong market efficiency, the empirical evidence is mixed and

less conclusive. For both the South African market and international markets,
results from event studies provide strong support for semi-strong market effi-
ciency. Against this, the return prediction studies, that is, studies on calendar
effects, the size effect, the BV/MV effect, etcetera, have yielded anomalous
results.
• The results from empirical studies on strong form market efficiency are also
mixed. The evidence on the general inability of professional money managers to
show consistent superior performance and market timing ability provides
408
significant support for the EMH. The inability of these professionals to con-
sistently outperform the market also raises doubts about the significance and
usefulness of the semi-strong form anomalies. Against this, the evidence from
insider trading activities and those of stock market specialists clearly pro"vides
no support for the EMH.
• Current knowledge therefore suggests that capital markets, especially markets

of developing countries like the JSE, are operationally efficient. This means that
the markets are efficient for investors and professional money managers in
general, but stock market specialists and insiders are consistently able to
outperform the market return.
• Alternative market efficiency theories, such as the Speculative Bubble Theory

and Chaos Theory, have been proposed, but these still have to be shown to
have better explanatory power and predictive ability than the EMH.
6.3.4 The CAPM
• There is a positive linear relationship between risk and return, but the slope of
the SML appears to be less steep than predicted by the CAPM.
• Although it has been found that betas of individual shares are far less stable than
those of portfolios, it can generally be concluded that beta is a valid and a fairly
complete measure of risk. The relative stability of portfolio betas means that the
CAPM is more useful in structuring share portfolios than in estimating returns on
individual shares.
• The CAPM is an expectational model and should not be judged on the basis of
the realism of its assumptions, but rather on how well it explains and predicts. It
is, however, the difficulty of measuring investor expectations that has caused
some researchers to conclude that the CAPM may basically be untestable.
409
• The CAPM , although not misspecified, may be inadequate, because other risk
factors (other than beta) may provide a better description of the risk/return rela-
tionship.
• The CAPM has received a considerable amount of empirical support, and it is

reasonable to conclude that the model should continue to be used in estimating
expected returns until it has either been discarded or an improved model has
been validated.
6.3.5 The APT
• The APT is by nature descriptive and may or may not be able to explain what
should be.
• There is evidence that expected returns are affected by more than one risk factor
and that the APT may be superior to the CAPM. but these results are as yet not
conclusive. It is therefore reasonable to conclude that both models will be conti-
nued to be used and that only continued empirical testing will determine whether
one model is superior to the other.
• There is as yet no consensus about the number of APT risk factors, nor about
their identity. Different studies have found different factors for different periods
and datasets and have also found evidence that the factors and their impact on
returns can change quite rapidly. Thus ongoing research is required to establish
a comprehensive and stable list of factors and their impact on returns.
6.3.6 Options and the B-5 Option Pricing Model
• The put-call parity relationship generally holds and options markets can be
regarded as being efficient in their pricing of options. The evidence on the hypo-
thesis of market synchronization is less clear and requires further research.
410
• The 8-S model is valid and is an accurate and flexible model that provides good
predictions and explanations of option prices. No other option pricing model
currently provides better descriptions of option pricing or option values and the
8-S model can even be used to identify underpriced and overpriced options.
6.4 RECOMMENDATIONS
The recommendations listed below are aimed at future South African research into the
capital market theories and pricing models under local conditions. However, these
recommendations can largely also be applied internationally.
• Research is needed to establish whether and to what degree international di-

versification still has risk/return benefits for South African investors.
• Ongoing research is required into the issue of stock market efficiency. It is ne-
cessary to establish whether the level or degree of efficiency of the JSE has
changed, to determine its exact level or degree of efficiency and to ascertain
whether other theories, such as Chaos Theory or the Speculative Bubble Theory,
provide complementary or better descriptions of the efficiency of the JSE.
• Since it has not yet been conclusively shown whether either the CAPM or APT
is superior to the other, both models should continue to be tested. The aim of the
research should be to improve the parameter estimation of both models, to
compare their predictive and explanatory power and to establish whether the two
models, when used jointly, provide better results than when a model is used on
its own.
• Since the sensitivity of share returns to the various APT risk factors has been
shown to change over time, research into these factors should also continue.
Other factors should also be identified and tested in order to obtain an even
better explanation of share returns.
411
• Very few empirical studies of the South African options market have been
conducted and there is a distinct need for research into the efficiency of the
South African options market, the degree of synchronization between the stock
market and the options market and the accuracy and robustness of the B-S
Option Pricing Model under South African conditions.
6.5 SUMMARY
The review of the principles and nature of the main capital market theories and pricing
models, together with the review of empirical research thereon, have shown that these
theories and models have contributed much to the understanding of the working of
capital markets and of investment risk and return. All the~e theories and models have
predictive ability and explanatory power, but, with possibly the exception of Portfolio
Theory, current knowledge of the validity, accuracy and robustness of these theories
and models has much scope for improvement. Ongoing research is required to improve
understanding of the efficiency and pricing mechanism of capital markets and to
establish whether other theories and models, some yet to be developed, can provide
better explanations and descriptions of capital markets and investor behaviour. This
broadening of knowledge can only be to the benefit of investors, as investment
practitioners, and academics, as investment theoreticians and contributors to the deve-
lopment of an accounting theory.
411
• Very few empirical studies of the South African options market have been
conducted and there is a distinct need for research into the efficiency of the
South African options market, the degree of synchronization between the stock
market and the options market and the accuracy and robustness of the 8-S
Option Pricing Model under South African conditions.
6.5 SUMMARY
The review of the principles and nature of the main capital market theories and pricing
models, together with the review of empirical research thereon, have shown that these
theories and models have contributed much to the understanding of the working of
capital markets and of investment risk and return. All these theories and models have
predictive ability and explanatory power, but, with possibly the exception of Portfolio
Theory, current knowledge of the validity, accuracy and robustness of these theories
and models has much scope for improvement. Ongoing research is required to improve
understanding of the efficiency and pricing mechanism of capital markets and to
establish whether other theories and models, some yet to be developed, can provide
better explanations and descriptions of capital markets and investor behaviour. This
broadening of knowledge can only be to the benefit of investors, as investment
practitioners, and academics, as investment theoreticians and contributors to the deve-
lopment of an accounting theory.
412
BIBLIOGRAPHY
A. ARTICLES
Alison, S. 1991. Is one beta good enough? Management Accounting, March:36-40.

Barr, GDI. 1990. Macroeconomic identification of the pricing factors on the Johannes-
burg Stock Exchange. South African Journal of Business Management, 21 (1 },
June: 17-26.
Shana, N. 1994. A review of the efficiency of the Johannesburg Stock Exchange. De
Ratione, 8(2), Summer:79-98.
Bird, WW & Page, MJ. 1991. The pricing of South African equity linked financial instru-
ments. Journal for Studies in Economics and Econometrics, 15(3): 1-21.
Blount, G & Davidson, S. 1996. Stock market responses to corporate unbundling in
South Africa. De Ratione, 10(1 }, Winter:63-73.
Bowie, DC & Bradfield, DJ. 1997. Some evidence on the stability of beta coefficients on
the JSE. SA Journal of Accounting Research, 11(2):1-20.
Bradfield, DJ. 1998. Measuring the selection and timing abilities of South African fund
managers. SA Journal of Accounting Research, 12(1 ):1-13.
Bradfield, DJ, Barr, GDI & Affleck-Graves, JF. 1988. Asset pricing in small markets -the
South African case. The South African Journal of Business Management, 19(1 ),
March: 11-21.
Broadbent, JM. 1992. The capital asset pricing model: its theory and practice. CIMA
Student, 6, February:B-10.
Delport, R. 1998. Waar die geleentheid in opsies le. Finansies & Tegniek, 50(22),
Junie:64.
Delport, R. 1999a. Die spekulatiewe aard van opsies. Finansies & Tegniek, 51 (21 ),
Junie:57.
Delport, R. 1999b. Moenie te lank vashou aan opsies. Finansies & Tegniek, 51 (24 ),
Julie:71.
Fielding, J. 1989. Is beta better? Management Accounting, November:38-40.
413
Firer, C. 1993. Estimating the return parameters of the capital asset pricing model. De
Ratione, 7(1 ), Winter:23-39.
Firer, C. & Thomson, TA. 1993. A note on the estimation of betas for single industry
companies. De Ratione, 7(2), Summer:24-28.
Gourley, B., Maguire, W., Miller, J. & Seneque, P. 1987. The cost of equity revisited.
Accountancy SA, 4(7), July:219-225.
Gregory, A. 1990. The usefulness of beta in the investment appraisal process. Manage-
ment Accounting, January:42-43.
Nichols, NA. 1993. Efficient? Chaotic? What's the newfinance? Harvard Business Re-
view, March-April:50-60.
Page, MJ. 1989. Model selection for measuring security price performance. South Afri-
can Journal of Business Management, 20(2), June:78-81.
Page, MJ. 1996. Further evidence of firm size and earnings anomalies on the Johan-
nesburg Stock Exchange. De Ratione, 10( 1), Winter: 27 -44.
Sealy, NR. & Knight, RF. 1987. Dividend policy, share price and return: a study on the
Johannesburg Stock Exchange. The Investment Analysts Journal, 29, May:33-
47.
Seneque, PJC. 1987. Recent developments in the pricing of financial assets. De Ra-
tione, 1(2), Summer:28-40.
Van Ransburg, P. 1998. Unifying the factor analytic and prespecified variable ap-
proaches to APT factor identification on the Johannesburg Stock Exchange. SA
Journal of Accounting Research, 12(1):15-45.
Viljoen, B. 1996. Option dealing strategies and the related risk management proce-
dures. Raurek Bulletin, 5(6), August:40-46.
Ward, M. 1994. Risk on the Johannesburg Stock Exchange. De Ratione, 8(2), Summer:
99-114.
Wood, S. 1997. Derivatives: investor options open up. Financial Mail, 146( 12), Octa
ber: 1997: 93.
414
8. BOOKS
Belkaoui, AR. 1992. Accounting theory. 3rd edition. London:Academic Press.

Bicksler, JL. 1977. Capital market equilibrium and efficiency. Lexington, Massachusetts:
Lexington Books.
Brenner. M. 1983. Option pricing - theory and applications. Lexington, Massachusetts:
Lexington Books.
Clark, JJ, Hindelang, T J & Pritchard, RE. 1979. Capital budgeting - planning and control
of capital expenditure. Englewood Cliffs, New Jersey: Prentice-Hall.
Correia, C, Flynn, D, Uliana, E & Wormald, M. 1993. Financial Management. 3rd edition.
Cape Town:Juta.
Cragg, JC & Malkiel, BG. 1982. Expectations and the structure of share prices. Chicago:
University of Chicago Press.
Dickerson, B, Campsey, BJ & Brigham, EF. 1995. Introduction to financial management.
4th edition. Orlando, Florida: Dryden.
Dobbins, R, Witt, S & Fielding, J. 1994. Portfolio theory and investment management.
2"d edition. Oxford:Blackwell.
Drury, C. 1992. Management and cost accounting. 3rd edition. London: Chapman & Hall.
Elton, EJ & Gruber, MJ. 1995. Modern portfolio theory and investment analysis. 5th edi-
tion. New Yori<:Wiley.
Fama, EF. 1976. Foundations of finance - portfolio decisions and security prices. New
York:Basic Books.
Fewings, DR. 1979. Corporate growth and common stock risk. Greenwich, Connecticut:
Jai Press.
FINEX. 1992. From Black-Scholes to black holes: new frontiers in options. London: Risk
Magazine.
Firth, M. 1977. The valuation of shares and the efficients-markets theory. London:
MacMillan.
Friend, I & Bicksler, JL. 1977. Risk and return in finance volume I. Cambridge, Massa-
chusetts: Ballinger.
415
Gemmill, G. 1993. Option pricing - an international perspective. Maidenhead, Berkshire:

McGraw-Hill.
Griffin, PA. 1987. Usefulness to investors and creditors of information provided by finan-
cial reporting. 2nd edition. Stamford:FASB.
Harrington, DR 1987. Modern portfolio theory, the capital asset pricing model and arbi-
trage pricing theory: a user's guide. 2nd edition. Englewood Cliffs, New Jersey:
Prentice-Hal I.
Harrington, DR & Korajczyk, RA. 1993. The CAPM controversy: policy and strategy im-
plications for investment management. ICFA Continuing Education. New York:
Association for Investment Management and Research (AIMR).
Henderson, S, Peirson, G & Brown, R. 1992. Financial accounting theory - its nature
and development. 2nd edition. Melbourne: Longman Cheshire.
Hendriksen, ES & Van Breda, MF. 1992. Accounting theory. 5th edition. Homewood,
Boston: Irwin.
Hodges, S. 1990. Options: recent advances in theory and practice. Manchester:Man-
chester University Press.
Jarrow, RA & Rudd, A. 1983. Option pricing. Homewood, lllinois:Dow Jones-Irwin.
Jones, CP. 1998. Investments: analysis and management. 6th edition. New York: Wiley.
Kam, V. 1990. Accounting theory. 2nd edition. New York:Wiley.
Keane, SM. 1983. Stock market efficiency- theory, evidence and implications Dedding-
ton, Oxford:Phillip Allen.
Levy, H & Sarnat, M. 1994. Capital investment and financial decisions. 5th edition.
Heme I Hempstead, Hertfordshire: Prentice-Hall.
Lumby, S. 1994. Investment appraisal and financial decisions. 5th edition. London: Chap-
man & Hall.
Markowitz, HM. 1991. Portfolio selection - efficient diversification of investments. 2nd edi-
tion. Cambridge, Massachusetts:Basil Blackwell.
Mclaney, EJ. 1997. Business finance: theory and practice. 4th edition. London: Pitman.
Milne, F. 1995. Finance theory and asset pricing. Vienna Institute for Advanced Studies.
Oxford:Claredon Press.
416
O' Brien, J & Srivastava, S. 1995. Investments: a visual approach - modem portfolio
theory and CAPM tutor. Cincinnati, Ohio: South-Western.
Pike, R & Neale, B. 1996. Corporate finance and investment - decisions and strategies.
2nd edition. Hemel Hempstead, Hertfordshire:Prentice-Hall.
Rees, B. 1995. Financial analysis. 2nd edition. Hemel Hempstead, Hertfordshire: Pren-
tice-Hall.
Reilly, FK & Brown, KC. 1997. Investment analysis and portfolio management. 5th edi-
tion. Orlando, Florida: Dryden.
Ross, SA, Westerfield, RW & Jaffe, JF. 1990. Corporate finance. 2"d edition. Home-
wood, Illinois: Irwin.
Ross, SA, Westerfield, RW, Jordan, BD & Firer, C. 1996. Fundamentals of corporate
finance. 1st South African edition. London: Irwin.
Scott, WR 1997. Financial accounting theory. Upper Saddle River, New Jersey:Pren-
tice-Hall.
Sharpe, WF. 1985. Investments. 3rd edition. Englewood Cliffs, New Jersey:Prentice-
Hall.
Van Horne, JC. 1992. Financial management and policy. gth edition. Englewood Cliffs,
New Jersey:Prentice-Hall.
Viscione, JA & Roberts, GA. 1987. Contemporary financial management. Columbus,
Ohio: Merrill.
Weston, JF & Copeland, TE. 1992. Managerial finance. 9th edition. Orlando, Florida:
Dryden.
Wolk, HI, Francis, JR & Tearney, MG. 1989. Accounting theory: a conceptual and insti-
tutional approach. 2nd edition. Boston, Massachusetts:PWS-Kent.
417
C. COLLECTIVE WORKS
Arnott, RD. 1993. The CAPM and equity management, asset allocation, and perfor-
mance management, in The CAPM controversy ... , edited by Diana R Harrington
and Robert A Korajczyk. 1993. New York:AIMR:16-24.
Beaver, W, Kettler, P & Scholes, M. 1970. The association between market determined
and accounting determined risk measures, in Capital market ... , edited by James
L Bicksler. 1977. Lexington, Massachusetts:Lexington Books:253-294.
Bicksler, JL. 1973. The usefulness of beta risk for estimating the cost of capital, in Risk
and return ... , edited by Irwin Friend and James L Bicksler. 1977. Cambridge,
Massachusetts: Ballinger: 81-100.
Black, F. 1988. The holes in Black-Scholes, in From Black-Scholes to ... , FINEX. 1992.
London:Risk Magazine:51-56.
Black, F, 1990. Living up to the model, in From Black-Scholes to ... , FINEX. 1992.
London: Risk Magazine: 17-21.
Blomeyer, EC & Klemkosky, RC. 1982. Tests of market efficiency for American call
options, in Option pricing ... , edited by Menachem Brenner. 1983. Lexington,
Massachusetts: Lexington Books: 101-120.
Blume, ME. 1973. Ex ante and ex post measures of return, in Risk and return ... , edited
by Irwin Friend and James L Bicksler. 1977. Cambridge, Massachusetts:Ballin-
ger:35-48.
Blume, ME. 1975. Betas and their regression tendencies, in Risk and return ... , edited
ger:21-34.
Blume, ME. 1993. The capital asset pricing model and the CAPM literature, in The
CAPM controversy ... , edited by Diana R Harrington and Robert A Korajczyk.
1993. New York:AIMR: 1-4.
Blume, ME & Friend, I. 1974. Risk, investment strategy and the long-run rates of return,
in Risk and return ... , edited by Irwin Friend and James L Bicksler. 1977. Cam-
bridge, Massachusetts: Ballinger: 1-20.
418
Burton, ET. 1982. Observations on the theory of option pricing on debt instruments, in
Option pricing ... , edited by Menachem Brenner. 1983. Lexington, Massachu-
setts: Lexington Books:33-44.
Cox, JC & Rubinstein, M. 1982. A survey of alternative option-pricing models, in Option
pricing ... , edited by Menachem Brenner. 1983. Lexington, Massachusetts:
Lexington Books:3-32.
Fama, EF. 1970. Efficient capital markets: a review of theory and empirical work, in Ca-
pital market ... , edited by James L Bicksler. 1977. Lexington, Massachusetts:
Lexington Books:41-82.
Fama, EF & MacBeth, D. 1974. Tests of the multi period two-parameter model, in Risk
and return ... , edited by Irwin Friend and James L Bicksler. 1977. Cambridge,
Massachusetts: Ballinger:297-326.
Fouse, WL. 1993. What have we learned? What does it mean? And what does the
future hold?, in The CAPM controversy ... , edited by Diana R Harrington and
Robert A Korajczyk. 1993. New York:AIMR: 110-117.
Friend, I & Blume, ME. 1975. The demand for risky assets, in Risk and return ... , edited
ger: 101-139.
Fuller, R. 1993. The small-firm anomaly, liquidity premiums, and time-varying returns
to size, in The CAPM controversy .. ., edited by Diana R Harrington and Robert
A Korajczyk. 1993. New York:AIMR:25-32.
Galai, D. 1982. A survey of empirical tests of option pricing models, in Option pricing ... ,
edited by Menachem Brenner. 1983. Lexington, Massachusetts: Lexington Books:
45-77.
Geske, R & Torous, W. 1988. Black-Scholes option pricing and robust variance estima-
tion, in Options: recent advances ... , edited by Stewart Hodges. 1990. Manches-
ter: Manchester University Press:3-9.
Harrington, DR. 1993. The CAPM controversy: an overview, in The CAPM controversy
... , edited by Diana R Harrington and Robert A Korajczyk. 1993. New York:AIMR:
1-4.
419
Hodges, S. 1988a. Financial options: new markets and new valuation techniques, in
Options: recent advances ... , edited by Stewart Hodges. 1990. Manchester: Man-
chester University Press:3-9.
Hodges, S. 1988b. Valuation and hedging: theoretical issues, in Options: recent advan-
ces ... , edited by Stewart Hodges. 1990. Manchester: Manchester University
Press:22-32.
Jensen, MC. 1972. Capital markets: theory and evidence, in Capital market ... , edited
by James L Bicksler. 1977. Lexington, Massachusetts: Lexington Books: 111-164.
Karnosky, OS. 1993. Global investment in a CAPM framework, in The CAPM contro-
versy ... , edited by Diana R Harrington and Robert A Korajczyk. 1993. New York:
AIMR:56-61.
Lakonishok, J. 1993. Is beta dead or alive?, in The CAPM controversy ... , edited by
Diana R Harrington and Robert A Korajczyk. 1993. New York:AIMR:38-41.
Leland, HE. 1988. On the stock market crash and portfolio insurance, in Options: recent
advances ... , edited by Stewart Hodges. 1990. Manchester: Manchester Universi-
ty Press: 105-122.
Leong, K. 1990. Exorcising the demon, in From Black-Scholes to ... , FINEX. 1992.
London: Risk Magazine:57-62.
Leong, K. 1991 a. Estimates, guesstimates and rules of thumb, in From Black-Scholes
to ... , FINEX. 1992. London:Risk Magazine:63-67.
Leong, K. 1991 b. Solving the mystery, in From Black-Scholes to ... , FINEX. 1992.
London:Risk Magazine:83-88.
Mandelker, G. 1974. Risk and return: the case of merging firms, in Capital market ... ,
edited by James L Bicksler. 1977. Lexington, Massachusetts: Lexington Books:
389-424.
Merton, RC. 1973. A reexamination of the capital asset pricing model, in Risk and re-
turn ... , edited by Irwin Friend and James L Bicksler. 1977. Cambridge, Massa-
chusetts: Ballinger: 141-159.
Myers, SC. 1973. The relationship between real and financial measures of risk and
return, in Risk and return ... , edited by Irwin Friend and James L Bicksler. 1977.
Cambridge, Massachusetts: Ballinger:49-80.
420
Ross, SA. 1973. Return, risk and arbitrage, in Risk and return ... , edited by Irwin Friend
and James L Bicksler. 1977. Cambridge, Massachusetts:Ballinger:189-218.
Ross, SA 1993. Is beta useful?, in The CAPM controversy ... , edited by Diana R Har-
rington and Robert A Korajczyk. 1993. New York:AIMR: 11-15.
Rozeff, MS. 1974. Money and stock prices, in Capital market ... , edited by James L
Bicksler. 1977. Lexington, Massachusetts:Lexington Books:517-579.
Simister, G. 1988. Practical issues in options trading, in Options: recent advances ... ,
edited by Stewart Hodges.1990. Manchester: Manchester University Press: 10-21.
Tompkins, R. 1992. Introduction, in From Black-Scholes to ... , FINEX. 1992. London:
Risk Magazine: 11-13.
Treynor, JL & Black, F. 1973. How to use security analysis to improve portfolio selec-
tion, in Capital market ... , edited by James L Bicksler. 1977. Lexington, Massa-
chusetts: Lexington Books:581-603.
421
D. DISSERTATIONS, THESES AND RESEARCH REPORTS
Anderson, AS. 1978. A multi-period capital asset pricing model. PhD thesis, University
of Arkansas.
Berges-Lobera, A 1982. The arbitrage pricing theory: estimation and applications. PhD
thesis, Purdue University.
Brevis, T. 1998. Tydsberekening binne 'n APT raamwerk. DCom tesis, Universiteit van
Suid-Afrika, Pretoria.
Casabona, PA 1979. Duration, beta and the capital asset pricing model: an integration
of risk measures for the analysis of security returns. PhD thesis, City University
of New York, New York.
Ferreira, JDaS. 1981. Characteristics of efficient portfolios of common stocks: implica-
tions for the testing of the CAPM. PhD thesis, University of Pennsylvania.
Gay, RC. 1982. The structure of security returns in an imperfect capital market: global
perspectives and evidence. PhD thesis. Harvard University, Cambridge.
Hawinkels. 1987. Risk versus return for option based strategies in the capital market.
MBA research report, University of Witwatersrand, Johannesburg.
Hedge, KG. 1983. Some empirical tests of the arbitrage pricing model. PhD thesis, The
Ohio State University.
Keogh, WJ. 1994. The stability of beta and the usability of the capital asset pricing
model in the South African context. MEcon thesis, University of Orange Free
State, Bloemfontein.
Laing, TA 1988. Abnormal return measurement in event studies - the arbitrage pricing
theory contrasted with methods used in previous studies. MBA research paper,
University of Cape Town, Cape Town.
Linley, PM. 1992. An evaluation of the capital asset pricing model and arbitrage pricing
theory in the pricing of assets. Technical report, University of Cape Town, Cape
Town.
Mulder, IJ. 1998. Accounting for financial instruments in corporate treasuries. DCompt
thesis, University of South Africa, Pretoria.
422
Mynhardt, RH. 1996. Die gebruik van opsies in vasterentedraende effekte om beleg-
gingsrisko te beperk. MCom verhandeling, Universiteit van Suid-Afrika, Pretoria.
Oosthuizen, CP. 1992. Measuring share market volatility within the framework of the
capital asset pricing model. MBA study project, University of Stellenbosch, Stel-
lenbosch.
$aenger, E. 1991. An investigation into a future-orientated approach to financial re-
porting in South Africa. DCom thesis, University of South Africa, Pretoria.
Spoede, CW. 1982. An investigation of security price performance and market model
methodologies. PhD thesis, University of Texas, Austin.
Upsher, SM. 1993. Expected share market returns and volatility: the South African ex-
perience including a critique of the capital asset pricing model. MBA study pro-
ject, University of Stellenbosch, Stellenbosch.
Van Rhijn, HJP. 1994. The capital asset pricing model for financial decision-making
under South African conditions. OBA thesis, University of Pretoria, Pretoria.
Viljoen, T. 1989. A critical analysis of the performance of various risk measures during
and after the stock market crash of 1987. MCom dissertation, University of Wit-
watersrand, Johannesburg.

43176319

Uploaded by

Copyright:

Available Formats

43176319

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

43176319

Uploaded by

Copyright:

Available Formats

CAPITAL MARKET THEORIES AND PRICING MODELS:

EVALUATION AND CONSOLIDATION OF THE

EUGENE RUDOLPH LAUBSCHER

submitted in fulfilment of the requirements

UNIVERSITY OF SOUTH AFRICA

SUPERVISOR: PROF E SAENGER

• Prof BL Steyn for his contribution to the study.

• Me YB van Stuyvenberg, librarian, for her assistance in accessing and obtaining

• Me EM van Deventer for her assistance with the technical wordprocessing

• Me ME Joubert for her editorial assiatance.

• My family and colleagues for their encouragement.

• My wife, Mariana, whose endless love, understanding and patience supported

CAPITAL MARKET THEORIES AND PRICING MODELS:

Eugene Rudolph Laubscher

2.3 The historical background to the development of capital

4.1 Introduction 118

5.1 Introduction 276

6.1 Introduction 396

Formula 3.1 Standard deviation of an investment 48

Formula 3.2 Actual return on an investment 49

Formula 3.3 Expected return on an investment 49

Formula 3.4 Standard deviation of a portfolio 63

Formula 3.5 Expected return on a portfolio 64

Formula 3.6 The lower confidence limit 68

Formula 3.7 Maximum utility of wealth 70

Formula 4.1 The Capital Asset Pricing Model 135

Formula 4.2 Alpha 136

Formula 4.3 The Arbitrage Pricing Theory factor-model 189

Formula 4.4 Put-call parity 224

Formula 4.5 The Black-Scholes Option Pricing Model 233

Figure 3.1 Standard deviation as a measure of risk 47

Figure 3.2 The relationship between risk and return 50

Figure 3.3 The relationship between income (wealth) and utility 54

Figure 3.4 Risk reduction through diversification 60

Figure 3.5 Efficient frontier 65

Figure 3.6 Risk-averse investors' utility indifference curves 69

Figure 3.7 Selection of the optimal portfolio 71

Figure 3.8 Reactions of capital markets to new information 79

Figure 4.1 The capital market line 131

Figure 4.2 The security market line 134

Figure 4.3 Boundaries of call option values 221

THE SCOPE AND PURPOSE OF THE STUDY

1.2 Background to the study 2

1.3 Definition of the problem 4

1.4 Purpose of the study 6

1.5 Motivation for and importance of the study 6

1.6 The research approach 7

1. 7 Organisation of the study 9

1.2 BACKGROUND TO THE STUDY

A further important development was Black and Scholes's (1973) contribution of an

The development of the Arbitrage Pricing Theory by Ross ( 1976) as a rival/alternative

1.3 DEFINITION OF THE PROBLEM

This gives rise to the definition of the problem of this study:

Does the available body of knowledge and evidence from em-

it filters and translates a large body of work written mostly by aca-

The sub-problems of the study can be summarised as follows:

• Can the recent research and developments be summarised and consolidated to

• Where should future research, especially under South African conditions, be

1.4 PURPOSE OF THE STUDY

1.5 MOTIVATION FOR AND IMPORTANCE OF THE STUDY