Country Risk and the Global CAPM -

Evidence about Beta from the MSCI ACWI

Term Paper∗

University College London

January 11, 2018

Abstract

It is widely known, that there exists no general definition of country risk, that this is probably because
country risk is a multi-dimensional and complex concept and that the effects of the same country risks can be
both benficial and harmful dependent on the industry or firm. Nonetheless, adding a country risk premium
to the CAPM is conceived as a plausible measure of risk. This essay argues against this view and especially
against an argument against applying an unadjusted CAPM to emerging market countries, namely that
the beta of emerging countries is generally lower that the beta of developed countries. After arguing for not
adjusting the CAPM when it comes to valuation in emerging markets, this will be achieved by calculating
current emerging and developed country betas against the MSCI All Country World Index as well as their
rolling betas over time. It is found that the emerging markets’ average beta is not in generally higher both
currently and throughout time. Overall, the essay may contributes to a more transparent way of doing global
business valuations by being clear about the arbitrariness of the differet risk concepts used when trying to
adjust the CAPM.

Contents

I Introduction 2

II Should there be a Country Risk Premium? 2

III Estimating Betas for a Global CAPM 3

i Benchmark Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
ii Calculating Current Betas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
iii Calculating Betas over Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

IV Concluding Remarks 9

∗ Thispaper was written for the course SEESGS50: Cor-

porate Finance and investment in Emerging Markets, held † StudentNumber: 17066036, M.A. Comparative Busi-
by Dr. Eugene Nivorozhkin. ness Economics.

1
 Country Risk and the Global CAPM • January 2018 • Florian Aichinger

I. Introduction [Estrada, 2007]). For this, the discussions about

whether or not to extend the CAPM and there-
he capital asset pricing model (CAPM)

T
fore the discount rate with country risk will
is still the workhorse model when it shortly be summarized. Following from that,
comes to determining the cost of eq- the view of applying an unadjusted CAPM to
uity. Although there are alternatives such as emerging markets can be seen as consistent,
multi-factor models or arbitrage pricing the- especially when assuming that the marginal in-
ory, none of those has replaced the CAPM vestor is globally diversified. Thereafter, it will
and its dominant role in asset pricing yet be tried to explore whether or not the beta of
([Brealy et al., 2011]). Its standard version sim- emerging market countries is generally lower
ply puts forth the linear relation between risk then the one of develped countries, a hypoth-
and return as esis and common critique of applying an un-
adjusted CAPM to emerging markets, as pro-
r i = r f + β i (r m − r f ), (1)
posed e.g. by [Baker and Filbeck, 2015]. With
where ri , rm , r f and β i denote the expected risk the MSCI All Country World Index (ACWI)
premium on equity, the expected risk premium as a proxy for the market portfolio, the betas
on the benchmark (the market), the risk free of all developed and emerging MSCI country
rate and the risk factor β respectively. The indexes which are part of the ACWI are esti-
CAPM’s risk factor, β, is therefore a measure of mated both currently and over time in order to
how sensitive the equity’s excess returns, rie = explore the betas’ average development. The
ri − r f , is to market excess return movements, results, namely that the emerging markets’ av-
e
rm = rm − r f . With this, to estimate beta, a erage beta is not generally smaller, together
simple OLS regression can be applied to with its explanations and implications are dis-
e e cussed. A further hypothesis, that the vari-
ri,t = αi + β i rm,t + ε i,t (2)
ability of the emerging markets’ average beta
with t being a time subscript, αi being the re- shrinks over time is put forth and might be
gression constant and ε i,t being the error term. subject to further research.
The risk factor can therefore be interpreted as
the average rise or fall of equity i’s excess re-
turns, when the market rises or falls an extra II. Should there be a Country Risk
1%. Premium?

The CAPM’s simplicity made it not only Following from Markowitz’s portfolio theory,
the workhorse model, but also an object of a rational investor holding an efficient, well-
constant criticism. Assuming that the reader diversified portfolio is only subject to non-
is somewhat familiar with the myriad of rea- diversifiable or market risk ([Markowitz, 1952].
sonable critiques, this essay focuses on one in Especially given that investors are only inter-
particular, namely the common view that the ested in a portfolio’s expected return and its
CAPM has to be adjusted by a country risk pre- standard deviation, given borrowing and lend-
mium (CRP) when valuing projects or compa- ing possibilities at r f and given an efficient,
nies in emerging markets ([Damodaran, 2003], competitive market, every investor would hold

2
 Country Risk and the Global CAPM • January 2018 • Florian Aichinger

a combination of the market portfolio and a Non-diversifiability in portfolio theory simply

risk-free loan. Within this framework, beta, states that if holding a well-diversified (ran-
as a well-defined risk concept, measures the dom) portfolio, the variance will be the same
contribution of an individual security to the as the one of the market. This holds true even
risk of a well-diversified portfolio. Diversi- if stock returns are correlated, as the market
fiable risk does not affect the discount rate. is simply a weighted average of these returns.
Someone proposing to adjust the CAPM with The argument therefore that country risk can-
country risk therefore needs to ask at least not be diversified due to correlation seems to
two initial questions: First, whether there is refer to the concept of systemic risk rather than
any reasonable and applicable definition of to systematic, i.e. non-diversifiable risk as used
country risk and second, whether this notion in asset pricing. With regard to valuation, I
of country risk is diversifiable. Regarding will therefore argue that country risk belongs
the first, most authors implicitly or explic- to the non-academic realm of calculating the ex-
itly refer to risk factors specific to the coun- pected cash flows. There, no defined concepts
try at interest such as political-, government are washed out, the assignment of probabili-
policy-, social-, macro- and microeconomic- ties requires both thorough justification and
or natural risks [Bouchet et al., 2003]. In a takes into the relevant factors for the individ-
broad sense with the focus on political risk, ual company or operation and is not depen-
[Haendel et al., 1975] defined country risk as dent on a general definition of a CRP. Over-
the “probability of occurrence of political all, adding a general country risk premium
events that will change the prospects for prof- to the CAPM without a general definition of
itability of a given investment”, defining it as country risk thus makes the undertaking ar-
a performance variance risk concept. In this bitrary [Kruschwitz et al., 2012]. Adding an
sense e.g. a policy change could be harmful an ill-defined CRP to the CAPM would trans-
for one industry or firm while being beneficial late to an ill-defined discount rate, ignoring
for another. Other authors in turn define coun- the well-defined concept of the beta. The sub-
try risk as negative outcomes, i.e. as events sequent analysis therefore starts with two as-
which, if they occur, negatively effect a firm’s sumptions: First, that the CAPM does not have
operations or investments [Roy and Roy, 1994]. to be adjusted by a CRP and second, that the
Whether defined as performance variance or marginal investor is globally diversified.
negative outcomes, three caveats stand out:
First, there exists no general definition of coun-
III. Estimating Betas for a Global
try risk, second, this is probably because coun-
CAPM
try risk is a multi-dimensional and complex
concept and third, the effects of country risks
i. Benchmark Data
can be differ across industries and firms (see
also [Bouchet et al., 2003]). Especially due to The latter assumption makes this analysis a
the last point a general definition is hard to theoretical one. Following from [Lintner, 1965],
imagine. But even if there were one, how could the market portfolio is an undetermined, gen-
we conclude that it would be non-diversifiable? eral market factor, which might, however, be
approximated by a stock index [Bimberg, 2009].

3
 Country Risk and the Global CAPM • January 2018 • Florian Aichinger

The choice of the benchmark portfolio has a tributes to the index nearly 5 times as much as
large effect on the estimates of the beta and the emerging markets in total, an unbalance also
subsequent analysis should therefore serve as driven by the US focus of MSCI Inc. Note that
one possible case, which can be easily modified the results presented are as of November 2017
by using or constructing other benchmark in- only, a complete reconstruction of the index
dexes with relevant weights of other marginal would require the weights of all (changing)
investors [Brown and Brown, 1987]. Here, the constituent indexes since the creation of the in-
MSCI ACWI is used as a proxy for the mar- dex. This not being necessary and lacking the
ket portfolio. The index comprises 47 coun- market share data over time, the MSCI ACWI
tries of which 23 are considered developed and price index provided by Datastream is used
24 considered emerging markets according to instead of constructing the index. In the next
MSCI’s own classification [MSCI, 2017]. Each section the returns, both monthly and weekly,
Index includes about 85% of the free float ad- of all country indexes will be regressed against
justed market capitalization in each market the returns of the market in order to determine
1. As the index is constructed using market the country betas. The country beta of coun-
weights, the country shares differ widely with try x can then be interpreted as the average
the United States (US) representing 52.67 % beta of the companies (in country x) against
of the index and 632 out of 2490 constituent the ACWI.
companies, therefore being by far the largest
component[MSCI, 2017]. Although the con- ii. Calculating Current Betas
struction of the ACWI is not known in de-
First, simple and logarithmic (log) returns
tail, evidence suggests that the ACWI is sim-
are calculated for all country indexes and the
ply a market capitalization weighted sum of
ACWI, both weekly and monthly, by taking the
its comprising indexes, as reconstructed in ta-
first and last value of the period. I.e. simple
ble 1 on page 5. The sum of the countries’
returns for period t are calculated as
reported market shares and number of con-
stituents is identical to the numbers reported l −p f
s
pi,t i
for the MSCI ACWI2 . Emerging markets make ri,t = f
(3)
pi,t
up only 11.53% in terms of the total capital-
ization, resulting from 33.61% of the contained l and
for weekly and monthly periods, where pi,t
companies. An average developed market com- f
pi,t denote the last and first value of period t
pany in the ACWI has therefore a market capi- respectively. Log returns can then be written
talization 4 times higher than the market capi- as
l
pi,t
talization of an average emerging market com- l
ri,t og = log( f
). (4)
pany (24.114 vs. 6.208 billion). The US con- pi,t
For an initial application of regression 2 a time
1 MSCIdefines free float as total shares outstanding ex-
span of two years for weekly returns and a
cluding shares held by strategic investors such as govern-
ments, corporations, controlling shareholders, and manage- time span of five years for monthly returns is
ment, and shares subject to foreign ownership restrictions used in order to have a reasonable amount of
[MSCI, 2000]. data.
2 Ignoring the tiny difference of 53 million.

4
 Country Risk and the Global CAPM • January 2018 • Florian Aichinger

Table 1: Contructing the ACWI: Developed and Emerging Markets Indexes with Market Share and Number of
Constituents. As of November 2017. Datasource: Thomson Reuters via Datastream. Own summary and
calculations.

5
 Country Risk and the Global CAPM • January 2018 • Florian Aichinger

However, choosing the time span in practice and for 5 countries when considering monthly
should also reveal how much weight is put returns. For the subsequent analysis, the fact
on recent or past events, which should be de- that the returns are not normally distributed
termined by the valuation object. The price as described should be kept in mind.
development and the respective returns can
be regarded in figure 1 for the MSCI ACWI. Next, country betas are calculated for both
As expected, the return series are graphically weekly and monthly log returns, for 2 years
close to a random walk and are mean-reverting. and 5 years respectively, of which the results
Regarding the return distributions, a first intu- can be seen in table 3. The betas are pre-
ition can be gained by looking at a histogram sented also market weighted in order to show
of both weekly and monthly returns and com- that the average weigthed beta equals 1. It
paring it no normally distributed data. As the is striking that the variability among the de-
return distributions do not differ in any rel- veloped countries’ betas is much lower then
evant way when comparing log and simple those of the emerging countries’ betas, rang-
returns, only log returns are depicted. As can ing from 0.262 (Pakistan) to 1.952 (Greece) for
been seen in figure 2, both weekly and monthly monthly returns, and from 0.461 (Russia) to
log returns of the MSCI ACWI have more ob- 3.034 (Greece) for weekly returns. The aver-
servations around the mean. Expected long age non-weighted beta of developed markets is
tails cannot be detected in figure 2. The graph- slightly higher compared to the average beta of
ical analysis gives reason to look at the distri- emerging markets. This might seem “strange”
butions of all constiuent indexes in more de- or “unexpecte” at first, but remembering that
tail. Table 2 therefore summarizes relevant de- the beta simple measures return movements
scriptive statistics, including the skewness and in comparison to the benchmark it cannot be
kurtosis and shows the results of the Jarque- held responsible to be an all-comprising risk
Bera normality test. The Jarque-Bera test tests factor, as referred to in section II. If the Pak-
whether sample data have the skewness and istan stock market tumbles due to some event
kurtosis matching a normal distribution. Go- and the benchmark is not effected due to Pak-
ing back to figure 2, the distirbution of the istan’s relatively low market capitalization or
MSCI ACWI’s returns resemble a double expo- more systematic if there are factors benefit-
nential distribution for which we would expect ting for the US market while threatening to
the skewness to be around 0 and the kurtosis pakistani stocks, then Pakistan’s beta will be
to be greater than 3. Table 2 reveals that this lower. While [Baker and Filbeck, 2015] state
is indeed true not only for the benchmark re- that “risky emerging markets report low be-
turns but also for most other countries3 . With tas”, data reveals that this is not generally true.
a significant level of 1% the null-hypothesis of What is true, is that the variability of emerging
normally distributed returns is rejected for 17 market betas is higher between the countries.
out of 48 countries when considering weekly, The average however resembles those of devel-
oped markets.
3 Note that the exceptional kurtosis for egypt’s weekly
returns is due a massive dip in November 2016. As the
MSCI Egypt only comprises 3 stocks this is not representa-
tive for the country.

6
 Country Risk and the Global CAPM • January 2018 • Florian Aichinger

Figure 1: Weekly and Monthly Log Returns and Prices. Datasource: Thomson Reuters via Datastream. Own depictions
and calculations.

Figure 2: A Graphical Impression: Histogramm with Normal Curve for Weekly and Monthly Log Returns for the
MSCI ACWI. Datasource: Thomson Reuters via Datastream. Own depictions and calculations.

7
 Country Risk and the Global CAPM • January 2018 • Florian Aichinger

Table 2: Testing the Normality of Returns: Descriptive Statistics. Datasource: Thomson Reuters via Datastream. Own
summary and calculations.

8
 Country Risk and the Global CAPM • January 2018 • Florian Aichinger

As a next step it will be analysed whether While the first point is a strong one, the second
the variability of the beta is not only the case one should be considered as an educated hy-
between countries, but also within countries, pothesis. Further research would be necessary
i.e. it will be looked at how the average beta to hold it.
of emerging markets evolved over time in com-
parison to developed markets. IV. Concluding Remarks

If the marginal investor is globally diversified,

iii. Calculating Betas over Time
so it was argued here, he should apply a global
Rolling regressions are applied to all the avail- CAPM. This does not mean that it is assumed
able index data. The regressions are the same that beta contains country risk. Beta as a sim-
as before with the only difference that they are ple “return comparing” instrument is not con-
used repeatedly while moving in time. Before structed to be an all-encompsasing risk mea-
continuing the analysis, some caveats of false sure. If country risk was present in the past,
interpretation are put forth. As the benchmark then it might be to some extend reflected by
index started earlier than many of the country the returns and therefore beta. However, es-
indexes, the averages do vary in the number of pecially when it comes to valuation, adjusting
countries incorporated. The latest country in- the cashflows with the best available probabili-
dexes started in May 2005 however (Quatar and ties for country risk factor should be applied.
the United Arab Emirated), setting the date How this should be done is a separate question,
from which on all current constituent countries for which to provide an answer was not the
contribute to the average. Furthermore, current goal of this essay. Instead, after speaking in
classifications are used, not be appropriate for favor for applying a global CAPM, the beta
all the times analysed. The assumption is there- of emerging and developed markets was scru-
fore, that this does not influence the outcomes tinized. Adding to the argumentative results
in any significant way. put forth against incorporating an ambiguous
CRP, quantitative results were found. Most of
The results are depicted in figure 3. Despite all that it cannot be concluded that the average
the mentioned caveats, the analysis may under- emerging markets’ beta is lower, although the
line at least two major points. First, it supports betas between the individual emerging market
the finding of the previous section, namely that countries vary more. A result also underlined
it cannot be concluded that emerging market when observing the evolution of the average
betas are generally lower than those of devel- emerging markets’ beta over time in compari-
oped markets over time. This is more evident son to developed markets. Overall, the essay
when regarding the results for monthly return might help to take a global CAPM a bit more
data. The statement that market capitalization seriously by disregarding at least one main ar-
lowers beta might therefore be questioned. Sec- gument quantitatively, namely that the beta of
ond, an increased market capitalization over emerging markets is generally lower, and by ar-
time might result into a lower variance of the guing against other commonly hold views such
average beta, as markets become more ma- as that correlation among emerging markets
ture and uncertainty is eventually diminished. adds to non-diversifiable risk. In the end,

9
 Country Risk and the Global CAPM • January 2018 • Florian Aichinger

Table 3: Calculating Betas for all Countries using Weekly and Monthly Log Returns. Note that the sum equals not
exactly one due to rounding differences. Datasource: Thomson Reuters via Datastream. Own summary and
calculations.

10
 Country Risk and the Global CAPM • January 2018 • Florian Aichinger

Figure 3: Rolling Regressions: Average Betas for Emerging and Developed Markets for Both Monthly and Weekly Log
Returns. Datasource: Thomson Reuters via Datastream. Own depictions and calculations.

11
 Country Risk and the Global CAPM • January 2018 • Florian Aichinger

the essay may contribute to a more transparent [Kruschwitz et al., 2012] Kruschwitz, L.; Loef-
way of doing global business valuations by fler A.; Mandl, G. (2012) DamodaranâĂŹs
being clear about the arbitrariness of differet Country Risk: A Serious Critique Business
concepts. Valuation Review Vol. 31

[Lintner, 1965] Lintner, J. (1965) The Valuation

References
of Risk Assets and the Selection of Risky
[Baker and Filbeck, 2015] Baker, H.; Filbeck, Investments in Stock Portfolios and Cap-
Greg (2015) Investment Risk Management ital Budgets The Review of Economics and
Oxford University Press p. 164 Statistics Vol. 47 No.1 pp. 13-37

[Bimberg, 2009] Bimberg, M. (2009) Acquisi- [Markowitz, 1952] Markowitz, H. (1952) Port-
tions and Takeover Offers in Germany folio Selection The Journal of Finance Vol.
(translated title) Springer Fachmedien only 7, No.1 p. 77 - 91
available in German p. 107
[MSCI, 2000] MSCI to Adjust for Free Float
[Brealy et al., 2011] Brealy, R.; Allen, F.; Myers, and to Increase Coverage to 85% MSCI
S. (2011) Principles of Corporate Finance press release Geneva December 10, 2000
- Tenth Edition McGraw-Hill/Irwin New
[MSCI, 2017] Global Market Accessibility Re-
York
view June 2017
[Brown and Brown, 1987] Damodaran, A.
[MSCI, 2017] MSCI ACWI Index (USD) fact-
(2003) Does the composition of the
sheet November 2017
market portfolio really matter? Journal of
Portfolio Management Winter p. 26 - 32 [MSCI, 2017] MSCI "Country" Index (USD)
Factsheets November 2017
[Bouchet et al., 2003] Bouchet, M.; Clark, E.;
Groslambert, B. (2003) Country Risk As- [Roy and Roy, 1994] Roy, A.; Roy P. (1994) De-
sessment - A Guide to Global Investment spite Past Debacles, Predicting Sovereign
Strategy John Wiley & Sons Ltd Chichester Risk Still Presents Problems Commercial
[Damodaran, 2003] Damodaran, A. (2003) Lending Review 9 (3) 92âĂŞ5.
Country risk and company exposure: the-
ory and practice Journal of Applied Finance
Vol. 13 p. 64 - 78

[Estrada, 2007] Estrada, J. (2011) Discount

Rates in Emerging Markets: Four Mod-
els and an Application Journal of Applied
Corporate Finance Vol. 19 p. 72 - 77

[Haendel et al., 1975] Haendel, D.; West G.;

Meadow, R. (1975) Overseas Investment
and Political Risk Foreign Policy Research
Institute Monograph Series No. 21

12