Fundamental and Macroeconomic Information for

the Security Prices Valuation: The French Case

by Isabelle Martinez, ESUG - University of Toulouse I
Abstract
This research aims at assessing the usefulness of fundamental and macroeconomic information
to the French market.Arelationship between abnormal returns and fundamentals is
analyzed between 1992 and 1997. The empirical findings show that the strength of the
returns-fundamentals association increases when two macroeconomic variables are added
to the explanatory model. The pooled regression R2 varies from 14 to 29% depending upon
whether the abnormal returns are regressed on the fundamentals only or on the fundamentals
and the two macroeconomic variables. The pooled regression R2 is also strongly affected.
This coefficient varies from 14 to29%depending upon whether the stock returns are
regressed on the three factors only or on the three factors and the twomacroeconomic variables.
These results clearly show that financial statements and macroeconomic data provide
informations to value security prices.

1. Introduction
Several functions are commonly assigned to accounting. One of them is the publication of
useful information to enable investors to value securities. Since the seminal work of Ball
and Brown (1968) and Beaver (1968), a large body of financial research has been devoted
to the usefulness of accounting earnings. The market valuation approach is commonly used
to study the association between earnings and stock prices. As outlined by Lev (1989), Cho
and Jung (1991), Dumontier and Labelle (1994) or Martinez (1994), the findings suggest
that market returns are explained in part by accounting earnings. This is why recent research
emphasizes the role of other financial statement information such as inventory or
capital structure (Ou and Penman, 1989; Holthausen and Larcker; 1992, Lev and Thiagarajan,
1993; Riahi-Belkaoui, 1997).
The purpose of our research is to assess to what extent accounting data published by
French firms are useful in security valuation.Weidentify a set of financial variables
(fundamentals)
claimed to be used by investors to evaluate the security prices. Our findings suggest
that the fundamentals are value-relevant to explain stock returns.Wealso show that the
explanation of returns is considerably strengthened when two macroeconomic variables
(the annual change in the Consumers’ Price of Index and the annual change in real Gross
National Product (GNP) are added to the explanatory model.
The next section examines the literature concerning the usefulness of accounting data.
Sections three and four describe the methodology used and the research findings, respectively.
The last section is devoted to concluding remarks.

2. A survey of empirical studies on the usefulness of accounting data

Research studies can be classified into two groups: the return - earnings research and the
return-fundamentals research.
The return-earnings research aims at assessing the usefulness of accounting earnings
to investors.Arelation is measured between a stock price variation for a given period and an
accounting variable representing earnings generated by the firm during the same period.
The regression is estimated cross-sectionally as follows:
Ri = 1 + iVi +

i (1)
Ri, Vi and i stand for the stock return of firm i, the variable representing earnings generated
by firm i and the regression noise, respectively.
The R-square coefficient measures the degree to which the stock returns are explained
by the accounting earnings. The coefficient i is defined as the effect of one US dollar (or
one French franc) of earnings on stock returns and is typically called the earnings response
coefficient (ERC). Several metrics are commonly used to compute the stock returns1, e.g.
the stock price change during the period, the period’s raw return and the period’s abnormal
return. Three accounting variable measures are also commonly used, namely the earnings
deflated by the beginnings of the period stock price, the unexpected earnings deflated by the
beginning of the period stock price and the unexpected earnings deflated by previous earnings.
Table 1 summarizes various characteristics and findings of a sample of return earnings
studies. The R2 coefficients obtained by regressing stock returns on accounting
earnings are low. They do not exceed 10% except for the French study (Dumontier and Labelle,
1994). There are several drawbacks with previous studies mentioned in table 1:
- Firstly, they link annual earnings to contemporaneous stock returns and use one-year period
data. However, the choice of this particular time period may not be optimal to test the
association between returns and earnings. Some people believe that there is probably a lag
between the period in which value-relevant events occur and the period in which these
events are integrated into earnings (Dumontier and Labelle, 1994). Lev (1989), Warfield
and Wild (1992), Easton, Harris and Ohlson (1992) or Dumontier and Labelle (1994)
recommend
increasing the interval over which the return-earnings relation is examined. Their
results show that the R2 coefficients increase with the interval length. For instance, the R2
coefficient in the Lev’s model is3%for a one-year interval and35%for a five-year interval.
- Secondly, these studies are based upon the pooled data of many firms under the assumption
that the return-earnings relation is homogeneous across firms (i.e. investors react identically
to earnings of all firms). A new line of research introduces also firm characteristics
such as firm size (Atiase, 1985) or stock exchange market (Grant, 1980).
- Thirdly, in these researches, earnings are believed to be the main information item provided
in financial statements. However, the low R2 coefficients obtained by regressing
stock returns on accounting earnings suggest that other financial variables, named fundamentals,
may provide more useful information to investors than the earnings data.
The association between stock returns and fundamentals was first examined by Ou
and Penman (1989). These authors used a statistical search procedure in the determination
of relevant fundamentals. They examined the ability of accounting information to generate
profitable trading strategies by developing amodel to predict the sign of unexpected annual
earnings-per-share (a logit model). The Ou and Penman’s (1989) study was expanded by
Holthausen and Larcker (1992). Later, Lev and Thiagarajan (1993) used a guided search
procedure in the determination of candidate fundamentals. This procedure is based on the
analysts’ claim of the usefulness of financial ratios in security valuation. The authors identified
a set of 12 fundamentals claimed by analysts to be useful, and examined these claims by
estimating the incremental value-relevance of these fundamentals over earnings. Two
cross-sectional regressions were compared. The first is the conventional return-earnings
regression
given by Eq. 1. The second includes the fundamentals as follows:

(2)
where Ri and Sij are the return of firm i and the fundamental j of firm i, respectively.
This regression model assumes that the fundamentals are uncorrelated but this may
not always be the case. The results of Lev and Thiagarajan (1993) show an improvement in
R2. For the 1980s, the 12 selected fundamentals add on average approximately 70% to the
explanatory power of earnings with respect to excess returns. They also show that the
return-fundamentals relation is strengthened when it is conditioned on macroeconomic
variables. Recently, Riahi-Belkaoui (1997) used a similar methodology to examine the role
of fundamentals in stock valuation. The author identified a set of 11 fundamentals and estimated
the regression models with Eqs. 1 and 2 for each of the years from 1973 to 1991. The
results show an improvement in R2 for each year indicating that the fundamentals contributed
significantly to the explanation of stock returns beyond accounting earnings. Similarly
to Lev and Thiagarajan (1993), Riahi-Belkaoui (1997) extended the analysis to evaluate the
impact of changes in inflation and GNP growth and showed that the significance of fundamentals
is conditioned by macroeconomic variables.
The aim of our study is to extend the work of Lev and Thiagarajan (1993) and Riahi-
Belkaoui (1997) to the French market, by using principal components analysis (PCA) on
the original set of fundamentals. This strategy will allow the determination of a small
number of uncorrelated factors to be included in the regression model, unlike previous studies
which assume a priori uncorrelated fundamentals.
3. Methodology
Following the procedure of Lev and Thiagarajan (1993) and Riahi-Belkaoui (1997), we
identified a set of candidate fundamentals used by analysts to evaluate firms’ performance
and estimate future earnings. These signals are the different financial ratios summarized in
appendix 1: profitability, growth, activity and financial structure ratios.
To compute profitability ratios, we used two different approaches. The « accounting
approach » is based on the book value of assets and the «monetary approach » ignores
accumulated
depreciation and amortization. The later enables the correction of accounting data.
Indeed, it is frequent for managers to manipulate accounting data by means of depreciation
and amortization, with the unique aim of improving or lowering accounting earnings. Al-though
we hypothesized that the monetary approach is more relevant than the accounting
approach to value securities, we used the two approaches to test this hypothesis.
Moreover, we split each profitability ratio (return on invested capital, return on assets
and return on equity) into two parts: the margin and the turnover ratios. Because the product
of several variables lose information according to this principle, we hypothesized that the
margin and turnover ratios provide more information than the profitability ratios. Finally,
we selected a set of 28 fundamentals claimed by analysts to be useful to value securities.
To examine empirically the usefulness of our candidate fundamentals, we proceeded
to two steps. In the first, we usedPCAto study the set of fundamentals. In the second, we
assessed
the usefulness of accounting and macroeconomic information.
The goal of the PCA is to represent the relationships among the 28 retained fundamentals
parsimoniously. That is, we aimed at obtaining a relatively small number of uncorrelated
factors useful in characterizing the set of candidate fundamentals.
In PCA, factors are estimated as linear combinations of the observed variables (fundamentals):
F a X j jk k
k
(3)
where ajk is the score coefficient for the factor Fj and the observed variable Xk.
Then, two regressions were estimated. In the first, stock returns are regressed on the
factors previously obtained with Eq. 3. This regression model is estimated cross-sectionally
as follows:
R F i j ij i
j

0 (4)
Ri is the annual stock return of firm i computed as the annual average of weekly raw returns.
The errors are assumed to be normally distributed, independent, random variables with zero
mean and standard deviation.
The second regression includes two macroeconomic variables: the annual change in
the Consumers’ Price Index (an inflation indicator) and the annual change in real GNP (a
state-of-the-economy variable). That is based on the results of Lev and Thaigarajan (1993)
and Riahi-Belkaoui (1997) which showed that the association between stock prices and
accounting
information is strengthened when it is conditioned on macroeconomic variables.
The second regression model is estimated as follows:
R F INFL GNP i j ij i
j

0 (5)
where INFL and GNP are the annual change in the Consumers’ Price Index and the annual
change in real GNP, respectively.

4. Results
The sample firms were selected for the period 1992-1996 according to the availability of
consolidated accounting data on the Dafsapro database and the availability of weekly security
prices on the Datastream File. Financial firms were not included in our analysis because
accounting standards are not the same than for non-financial firms. Finally, our
sample set contained 50 industrial firms summarized in appendix 2.
In the first step, we used PCA to study the fundamentals. Regardless of the Scree Test
rule, the first three principal factors were retained to represent the set of 28 original variables.
Table 2 contains the initial statistics for each factor. The three factors account for
60% of the total variance. The proportion of the total variance explained by each factor is
30.2%, 21.9% and7%respectively. To judge how well the three-factor model describes the
original variables, we computed the proportion of variance (called the communality) of
each variable explained by the factor model. The communalities for the variables are shown
in Table 3. A majority of accounting variables is correctly represented by the three-factor
model (communalities > 0.5).
Table 4 contains the coefficients (called factor loadings) used to express a variable in
terms of the factors. These coefficients indicate how much weight is assigned to each factor.
To identify the first three factors, variables that have large loadings for the same factor are
pooled together while variables that have small factor loadings (< 0.5 in absolute value) are
omitted. Thus, factor 1 is the factor with the largest loading (in absolute value) for the margin
and turnover ratios given by the decomposition of the profitability ratios (return on invested
capital, return on assets and return on equity) and for the equity and the total assets
turnovers. Factor 2 is composed by the return on invested capital, the return on assets, the
return on equity and the leverage ratio. Finally, factor 3 is the factor with the largest loading
(in absolute value) for the rate of working capital requirements and the financial balance ratio.
After the PCA, conclusions are three-fold:
- Firstly, the most important variables for our study are the ratios which constitute the three
principal factors, namely the profitability ratios (the return on invested capital, the return on
assets and the return on equity and their decomposition into margin and turnover ratios), the
leverage ratio, the rate of working capital requirements and the financial balance ratio.
- Secondly, the monetary approach seems to be more relevant than the accounting approach.
The factor loadings are indeed slightly higher when the profitability ratios are computed
using the monetary approach. For example, the factor-one loadings are for the return
on assets of 0.90 and 0.88 depending upon whether the monetary approach or the accounting
approach is used.
- Thirdly, the margin and turnover ratios reflected by factor 1 appears to be more relevant
than the profitability ratios reflected by factor 2. Indeed, the first factor explains about 30%
of the total variance against 21% for the second factor. This result can be related to several
studies such as Hopwood and McKeown (1985), Bublitz and Ettredge (1989), Swaminathan
and Weintrop (1991) which show that the variables used to compute earnings (taxes,
operating charges and products,...) provide supplementary information beyond earnings.
In the second step, we tested the relationship between stock returns and the three principal
factors obtained previously. The regression model given by Eq. 4 was estimated for
each of the years from 1992 to 1996 and across all years. Table 5 contains the regression results.
For every year (except 1994), the adjusted R2 is statistically significant (0.05 level)
and varies from8%to 35%. Factor 2 contributes significantly to the explanation of stock return
variance (the regression coefficients are positive and statistically significant at 0.05
level, except for 1994). This result indicates that the fundamentals (return on invested capital,
return on assets, return on equity and leverage ratio) represented by factor 2 are the most
informative for explaining stock prices.
The regression model given by Eq. 5 combines the three principal factors and the two
macroeconomic variables, GNP and INFL. Across-years coefficient estimates and R2 are
reported in Table 6. The R2 is strongly affected by adding the macroeconomic variables. It
is two times higher than the R2 obtained when the three factors appear alone in the pooled
regression on stock returns. Factor 2 that was statistically significant in Eq. 4 is also significant
in Eq. 5 and the two macroeconomic variables provide additional information to explain
the return variability. For the inflation indicator and the state-of-the-economy
variable, the coefficient estimates are statistically significant at 0.05 level. These coefficients
are negative and positive, respectively. This was expected because the security market
appears to decrease in case of inflation and to rise in case of economical growth.
5. Concluding remarks
This paper aimed at assessing the usefulness of financial statement information to the
French market. The identification of value-relevant fundamentals was guided by analysts’
descriptions. Our study differs from previous works because the association is not directly
tested between stock returns and fundamentals. The fundamentals identified as valuerelevant
were processed by a principal components analysis. This allows us to determine a
small number of uncorrelated factors to be included in the regression on stock returns. This
way, the analysis was not affected by multicollinearities in the data. The empirical findings
showed that the relationship between returns and the first three factors (used to represent the
fundamentals claimed by analysts to be useful) is statistically significant for each of the
years from 1992 to 1996. The adjusted R2 varies from 8% to 35%. According to the tstatistics,
the factor with the largest loading (in absolute value) for the return on invested
capital, the return on assets, the return on equity and the leverage ratio is value-relevant to
explain return variability. In order to complete the regression on stock returns, we added
twomacroeconomic variables (an inflation indicator and a state-of-the-economy variable).
The pooled regressionR2 is also strongly affected. This coefficient varies from 14% to 29%
depending upon whether the stock returns are regressed on the three factors only or on the
three factors and the two macroeconomic variables. These results clearly show that financial
statements and macroeconomic data provide informations to value security prices.