1. INTRODUCTION
Accounting information is one of the most significant sources of
financial information for analysis especially when it comes to
establishing the intrinsic value of the firm against which active
portfolio managers would compare the market price to initiate buy, sell
or hold decisions. If accounting information is an important component
of intrinsic value determination, we would expect accounting earnings
and share prices to be closely connected. Furthermore, accounting
fundamentals (book values and earnings) purport to measure value (i.e.
capital) and changes in value (i.e. profit) of the firm. Similarly, the
price of a share at one point in time also reflects value (i.e. market
capitalisation) and the change in price from one point in time to
another, the return to investors. In so far as accounting profits and
share returns attempt to measure the same characteristic, the change in
wealth of the firm, there should be some association between accounting
earnings and share prices. If this is the case, then earnings, as
reported by companies, should be useful to investors, especially foreign
investors, in predicting the return on shares. Indeed, because foreign
investors may not be physically present in the investee country, they
may not be fully aware of all information, especially non-earnings
related information, which may be circulated in the local market and
which may have an impact on share prices. These foreign investors will
probably, due to their geographical distance, have to rely on published
or reported earnings to assess the earnings potential of a particular
firm. A review of the literature shows that there that been a scant
amount of literature on the accounting earnings-foreign investors
hypothesis. In fact the pioneering work stems from Ball and Brown (1968)
who studied the usefulness of accounting earnings as an explanatory
variable for explaining stock returns or change in share prices. Since
then relatively few studies have followed among which features Beaver et
al. (1980), Hopwood and McKeown(), Lustgarden and Lev(), Brief-Zarowin
()
The objective of this paper is therefore to assess the extent of
the usefulness of accounting earnings to foreign investors by examining
the significance of the correlation between earnings and share prices.
If earnings can explain a reasonable proportion of the change in share
prices, then earnings will have some value relevance in predicting
future securities returns. This work is believed to supplement the
literature which has so far only scantly dealt with the accounting
earnings-foreign investors hypothetised link.
2. Methodology and Analysis: Many of the studies have regressed
stock price revisions on unexpected earnings. However, this method to
evaluate the value relevance of accounting earnings in explaining stock
return depends entirely on the proper measurement of expected earnings.
It was also pointed out by Lev ()that unexpected earnings may only
convey a small part of the information in earnings and that total
earnings might be a better variable as it would capture the impact of
the entire information set about earnings that is released throughout
the year. Studies by Ohlson and Brief and Zarowin have established the
pairing of earnings and book value to explain stock market values. This
study proposes to assess the value relevance of earnings to foreign
investors using two key independent variables, namely, the level of
earnings and the change in earnings, following Ohlson (1991)
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
Ohlson arrived at a model which expresses price as weighted average
of book value and total earnings. The weighting attached to book value
at time t, [y.sub.t] is [1- [alpha] (R-1)]. The weighting attached to
total earnings is [alpha] (R-I). Total earnings are represented by the
term [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] i-e. total
earnings is arrived at by multiplying clean surplus earnings, [X.sub.t]
by a multiple, [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] and
subsequently deducting the current dividend, [d.sub.t]. Ohlson's
above earnings valuation model shows that the relevance of book value
and total earnings depend on the magnitude of the w term which captures
the extent to which the current level of residual income is likely to
persist in the future.
Modelling the relation between returns and earnings
Having established the relationship between earnings and market
value according to Ohlson's earnings valuation model, we can now
turn to modelling the relationship between earnings and changes in
market value. Since the relevance of earnings in explaining market value
depends on the term w, which reflects the extent to which current level
of residual income is likely to persist in the future, i.e. on the
degree to which earnings follow a permanent or transitory process, it
seems logical to keep the same distinction when modelling the
relationship between earnings and return.
Modelling earnings as a transitory process If earnings follow a
transitory process, w is deemed to be zero and market value would equal
book value. In other words, the market to book value ratio would be
equal to one. If we represent [V.sub.jt] as the value of firm j at the
end of year t, then a market-to-book value of unity would imply that
[B.sub.jt] (the book value of firm j at the end of year t) is equal to
[V.sub.jt]. Since [V.sub.jt] is equal to [B.sub.jt], then the return of
firm j for year t, [R.sub.it] would be given by:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
Where:
[E.sub.jt] is the total clean surplus earnings for firm j for year
t
[D.sub.jt] is the dividend for firm j for year t
[V.sub.jt-1] is the value of firm j at the start of year t
The above expression defines return for year t simply as the
capital gain over year t plus the dividend received for year t,
expressed as a ratio of opening market value of equity. Since market
value tracks book value, the return for year t would be the increase in
book value of equity from year t-1 to year t plus the dividend for year
t (i.e. clean surplus earnings for year t) expressed as a ratio of
opening book value of equity, [B.sub.jt-1] As previously stated,
defining earnings according Ohlson's clean surplus expression is in
line with IASC's definition of profit, namely that profit is the
net increase in economic benefits in the form of net inflows or
increases in net assets that result in increases in equity without
considering distribution to equity participants as a decrease in
economic benefits. Modelling earnings as a permanent process
If earnings are deemed to follow a permanent process, Ohlson's
earnings-based valuation model suggests that the cum div value of equity
capital would be a multiple of clean surplus earnings as follows:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
The proposition that share price in the market is a multiple of
earnings is similar to the concept that the market value of firm j at
the end of year t is the product of reported earnings after tax of
company j for year t and the price-earnings ratio for firm j. Therefore,
where earnings follow a permanent process, value can be expressed as:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
Where [beta] represents the price earnings ratio or multiple for
company j. The return of firm j for year t is:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
The equation captures the idea that if earnings increase by Rs1 a
[beta] value of say 10 would imply that market value would increase by
Rs10. Because earnings are permanent, i.e current earnings persist into
the future, earnings become less volatile and more predictable.
Consequently for such higher quality earnings, investors would be
prepared to pay'S' times earnings to buy shares in the
company.
Regression Equations
The regression equations which would model equation whereby
earnings are deemed to follow a transitory process and equation whereby
earnings are deemed to follow a permanent process can be specified as
follows:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] [E2]
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] [E3]
Because in reality the value of w in Ohlson's residual income
valuation model is likely to be somewhere between 0 and 1, earnings
could consist of both transitory and permanent components. To capture
this possibility, equation [E3] includes two independent variables, the
total earnings level and the change in earnings.
Findings and an analysis
Table 1 summarizes the regression results of the three regression
equations together with key statistical significance tests, namely the t
and the F statistics.
The following observations can be made from the information
contained in Table above. First, it appears that Model 3 produces the
highest [R.sup.2] and, therefore, is more useful in explaining stock
returns. However, an important property of [R.sup.2] is that it is a
non-decreasing function of the number of explanatory variables present
in the model. Thus, as the number of explanatory variables increases,
R" almost invariably increases and never decreases. Because Model 3
contains two independent variables, its [R.sup.2] cannot be compared
with those of Models 1 and 2 which contain only one independent
variable. It is better to use the adjusted [R.sup.2] to compare the
explanatory power of several models because the adjusted [R.sup.2]
calculates a coefficient of determination which takes into account the
number of independent variables present in the model. On the basis of
adjusted [R.sup.2], Model 1 has the highest explanatory power at 19.48
%.
Model 1 models earnings as following a transitory process to
explain changes in stock prices. Model 2 on the other hand models
earnings as following a permanent process to explain stock returns. The
[R.sup.2] of Model 1(20.78 %) is significantly higher than that of Model
2 (3.6 %). Therefore, the market considers that it is earnings level as
a whole rather than change in earnings which are, to a certain extent,
relevant in explaining stock market returns on the SEM. This implies
that book value or change in book value has partial value relevance in
explaining stock returns. The term 'partial' is used because
Model 1 reports an [R.sup.2] of only 20.78 % which means that 20.78 % of
the total variation in stock returns is explained by earnings level
deflated by market capitalisation at the start of the 12-month return
period. Indeed, if earnings are deemed to follow a transitory process,
i.e. there is little or no permanence in the composition of earnings,
market value would track book value and therefore stock returns,
measured by change in market values, would be 'proxied' by
changes in book value. Changes in book value or net assets are
essentially caused by the earnings or profits after tax generated by the
company during the year. Therefore, modelling change in stock prices
with the earnings level as an independent variable is in fact mirroring
the value relevance of book value. This confirms the conclusion arrived
at by Brief and Zarowin, namely, that when earnings are transitory, as
opposed to permanent, book value has the greatest value relevance.
The higher explanatory power of Model 1 with respect to Model 2 can
be confirmed by carrying out statistical tests to check whether the
regression earnings response coefficients (slopes) are significantly
different from the hypothesized coefficient of zero. A hypothesized
coefficient of zero (the null hypothesis) implies that the independent
variable has no explanatory power. For Model 1, since the observed t
value (4) is greater than the critical value at the 5 % level of
significance (2) with 61 degrees of freedom, it can be concluded that b
= 222.02 is significantly different from zero. For Model 2 on the other
hand, since the observed t-statistic for the slope (1.5) is less than
the critical value of t at 5 % level of significance, it can be
concluded that d = 106 is not significantly different from zero. This
means that for Model 2 we should accept the null hypothesis namely that
the coefficient for d, change in earnings, is statistically zero.
When combining earnings and change in earnings as independent
variables (Model 3), the regression coefficient for earnings (b' =
218.72) is statistically significant because its t statistic (3.61) is
beyond the critical value for t at 5 % level of significance. On the
other hand, the regression coefficient for change in earnings in Model 3
(d' = 9.87) is statistically insignificant meaning that the true
coefficient of d is in fact zero and that the fact that d, the
regression coefficient, is different from zero is merely due to chance.
This conclusion is supported by the fact that the t statistic for d is
0.14 which is less than the critical t value of 2. It is quite clear
that if Model 3 generates an adjusted [R.sup.2] of 18.16 % it is
entirely because of the explanatory power of earnings level alone. The
larger the F value, the more meaningful is the regression model in
explaining stock returns. Because the F value of Model 2 (2.28) is below
the critical value of F at the 5 % level of significance (4), Model 2 is
not statistically significant in explaining stock returns implying that
the slope of the model is zero. Model 1 has the greatest F value (16)
suggesting that it is the most useful in explaining stock return: the F
value of Model 1 (16) is very much higher than the critical F value of
4. This result is not surprising as Model 1 also has the greatest
adjusted [R.sup.2].
Furthermore, the fact that adding change in earnings as an
additional independent variable to Model 1 produces a significant drop
in the F value from 16 to 7.88 in spite of the fact that the critical
value also shifts from 4 to 3.15, suggests that this variable should not
be added. This conclusion can be confirmed by assessing the incremental
contribution of the variable, change in earnings, after allowing for the
contribution of total earnings level. The F statistic to assess this
incremental contribution would be:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
'newk' represents the number of new regressors. At 5%
level of significance, the addition of the variable 'change in
earnings' does not increase the explanatory power of the model.
Interpreting the regression results in the context of the local
market setting
Skewness of the stock market: The Mauritius stock market can be
considered quite atypical of other more developed stock markets for two
key reasons. Firstly, key components of the Mauritius stock market do
not, unlike other developed stock markets, mirror those of the economy.
Textiles represent a key contributor to GDP for the Mauritian economy,
yet not a single textile company is listed. Secondly, the top ten
companies make up 75% of the total market capitalisation as at 31st
December 1999 and the top six make up 61% of total market capitalisation
as at the same date.
Family-owned companies: Furthermore, since most Mauritian companies
are family owned, most of the shares will be locked in the sense that
they will be held for family reasons and for the long term. Therefore,
it is evident that any change in earnings will not be considered as a
permanent change warranting investors to revise their forecasts and
change their shareholding which would cause a revision in stock prices.
The fact that only a small portion of the shares in issue are traded
suggests that shareholders do not consider changes in earnings form one
year to the next as a permanent change since the shareholders will hold
on to most of their shares and only trade a small portion.
Theoretical relevance of total earnings in stock valuation: If
shareholders wish to maintain their shareholding in the blue chip
companies, it can be shown theoretically that under certain assumptions,
it will be total earnings that will be relevant in stock valuation.
Clearly, earnings are important to stockholders because earnings provide
the cash flow necessary for paying dividends. Dividends however are also
important because they are what stockholders actually receive from the
firm and are the focus of dividend valuation models. Earnings of a firm
can either be used to pay dividends or to fund positive net present
value (NPV) projects and this could perhaps explain why total earnings
have greater explanatory power in explaining stock return movements: the
dividends accruing to shareholders depend on total earnings minus
investment opportunities financed from such earnings.
The impact of foreign investors :Since mid 1994, foreign investors
have been authorised to invest in listed as well as in newly issued
shares, with a limit however for sugar companies: no foreign investor,
be it an individual or a company, can hold more than 15% of the issued
share capital of a sugar company. Although there is no legal restriction
to the proportion of shares that can be held in companies outside the
sugar sector, some companies in their Memorandum and Articles of
Association or in their prospectus limit the proportion of total shares
that can be subscribed by foreign investors. Analyzing foreign net
purchases/sales from 1995 to 1999 will shed further light on the reasons
behind the greater explanatory power of deflated total earnings as an
independent variable in the return-earnings regression. Analysis of the
top five companies shows relatively good earnings momentum However,
except for SBM and MCB, the earnings per share (EPS) and the dividend
per share (DPS) tend to fluctuate, increasing at times and decreasing at
others. Do shareholders view these signals as permanent changes in
future profits requiring a change in shareholding? The answer is of
course, no since as mentioned before most of the shares are not actively
traded. Shareholders tend to hold on to their shares for the long term
and seek capital growth. This seems to be confirmed by the low dividend
payout ratio.
One can argue that the fact that the absolute level of PAT is
positive and undergoing steady growth has prompted foreign investors to
buy stocks of the top ten companies on the SEM. There is no doubt that
foreign investors have had a positive impact on the growth of the stock
market and have also contributed to making the Mauritian stock market
more efficient. Because foreign investors tend to bring technical
expertise it is not uncommon for local investors to mimic the behaviour
of foreign investors. Indeed the SEMDEX dipped from January 1995 to
November 1995, it subsequently picked up in December 1995 and January
1996 probably because of the sustained net foreign purchases from May
1995 onwards. The significant foreign net buying impetus from May 1995
to May 1998 gave local investors greater confidence to buy shares and
this obviously resulted in a rise in the SEMDEX from November 1996 to
August 1998.
It must be realised, however, that restrictions on foreign
ownership prevent the market from being more liquid. For instance,
foreign investors cannot hold more than 3% of the shares of SBM and 2.5%
of the shares of MCB. Limited foreign intervention and the absence of a
strong and active domestic investor base--most of the shareholders hold
on to their shares and only trade a small portion--are constraints which
prevent the market from gaining liquidity. Low liquidity leads to high
price volatility and renders entry and exit conditions difficult. An
investor may have to pay an unduly high price on just a small number of
shares if little sellers are forthcoming which makes entry to the market
difficult. Similarly, a shareholder wishing to sell may need to unduly
lower the price if buyers are not forthcoming which makes it difficult
for the shareholder in question to sell his shares and exit the market.
We should not be too surprised therefore if the 12-month cumulative
returns for all companies exhibit considerable volatility over time.
This seems to be confirmed by the negative intercept of 21.31% of
regression Model 1 which is significant at the 5% level (the observed t
statistic for the intercept is -2.44 which is greater than the critical
t value of 2 at 5% level of significance). Because the volatility in
return is induced by an illiquid market, we cannot expect returns to
fully reflect the information contained in reported earnings. However,
total earnings (profits after tax) run a greater chance of explaining
some movement in stock returns. The increase in net foreign purchases
from 1994 to June 1998 contributed to increasing the SEMDEX which meant
that there were more positive cumulative 12-month returns observations
from 1994 onwards to match positive absolute earnings figures. In fact
from 1994 onwards, there were 53 cumulative return observations, of
which 31 were positive. In other words, although earnings may be subject
to fluctuations, the overall trend is for the absolute level of total
earnings to increase. This is perhaps why deflated total earnings can
explain around 20% of the movement in 12-month cumulative stock returns.
The [R.sup.2] of 20.8% generated by Model 1 would have been higher
had it not been for the financial crises experienced by Asia, Latin
America and Russia from mid 1997 to January 1999. Since most global
funds experienced huge losses from the financial turbulence which hit
these emerging markets, it is not surprising that these funds decided to
liquidate Mauritian holdings and realise some profits to compensate the
above-mentioned losses. Furthermore, the meltdown in such emerging
markets could cause investors to revise upwards the risk factor in other
emerging markets like Mauritius even though the economic fundamentals of
these other emerging markets had not altered significantly. Because
information is costly, it could be optimal for one investor to follow
the lead of another perceived to have an information advantage. Such
herd behaviour would cause many foreign investors to disinvest and even
cause local investors to do the same.
4. Conclusion: The above analysis has shown that reported
accounting earnings are relevant to foreign investors in explaining both
stock prices and changes in stock prices However, it is important when
quantifying the extent of the usefulness of accounting earnings, to
appropriately model the return-earnings relationship. Indeed, when
earnings are modelled as following a transitory process, an [R.sup.2] of
20.78% is produced with an earnings response coefficient which is
significant at 5% level of significance. When earnings are modelled as
following a permanent process, an [R.sup.2] of only 3.6% is produced
with an ERC not significantly different from zero at 5% level of
significance. These regression results were achieved by pooling 63 pairs
of return-earnings observations for the period 1990-1999 for 12
companies which have been at least once listed on the SEM 7. Because the
SEM 7 includes the largest 7 Mauritian companies by market value which
satisfy the liquidity criteria--both in terms of value traded and
trading frequency--laid down by the Index Management Committee, the
share prices of listed companies which have been included in the SEM 7,
either regularly or intermittently, run a greater chance of capturing
the information contained in accounting earnings. The marked difference
in the [R.sup.2] of the two models confirms that the model used to
evaluate the significance of the contribution of earnings in explaining
stock returns does have an impact on the regression results and,
therefore, on the estimation of the value relevance of accounting
earnings.
REFERENCES
(1.) Ball, R. and Brown, P. (1968) "An Empirical Evaluation of
Accounting Numbers", Journal of Accounting Research, Autumn.
(2.) Beaver, W. (1989) " Financial Reporting: An Accounting
Revolution", Second Edition, Prentice Hall.
(3.) Brief, R. and Zarowin, P. "The Value Relevance of
Dividends, Book Value and Earnings". Working Paper, New York
University.
(4.) Lev. B 1989) "On The Usefulness of Earnings and Earnings
research: Lessons and Directions from Two Decades of Empirical
Research", Journal of Accounting Research.
(5.) Ohlson. J. (1991) "Earnings, Book Values and Dividends in
Security Valuation", Working Paper.
(6.) Rees. B. "Financial Analysis", Prentice Hall, Second
Edition.
(7.) Samuels, J. M. Wilkes, F. M. and Brayshaw, R. E.
"Management of Company Finance", Chapman and Hall, Fifth
Edition.
(8.) Strong, N. (1993) "The Relation between Returns and
Earnings: Evidence from the UK", Accounting and Business Research.
Matthew Lamport, University of Mauritius, Reduit, Mauritius
Table 1: Summary of regression results and significance statistics
1(E1) 2(E2) 3(E3)
Intercept a -21.31 C 36 a'
ERC b 222.02 D 16 b' 72
d' 7
SE of:
Intercept E(a) 8.73 SE(c) 17 SE(a') 8
ERC E(b) 55.50 SE(d) 73 SE(b') 57
SE(d') 48
[R.sup.2] 20.78% 0% 0%
Adjusted [R.sup.2] 19,48% 2% 6%
No. Of 63 63 63
observations
Observed t statistic 1(b) 4 t(d) 51 t(b') 1
for ERC t(d') 4
Critical t value 2 2 2
Observed F statistic 16 2.28 7.88