Professional Documents
Culture Documents
Accounting Earnings and Security Valuation: Empirical Evidence of the Fundamental Links
Author(s): Peter D. Easton
Source: Journal of Accounting Research, Vol. 23, Studies on Accounting Earnings and
Security Valuation: Current Research Issues (1985), pp. 54-77
Published by: Wiley on behalf of Accounting Research Center, Booth School of Business,
University of Chicago
Stable URL: http://www.jstor.org/stable/2490689
Accessed: 24-06-2016 17:45 UTC
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted
digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about
JSTOR, please contact support@jstor.org.
This content downloaded from 128.135.240.154 on Fri, 24 Jun 2016 17:45:55 UTC
All use subject to http://about.jstor.org/terms
1. Introduction
Considerable accounting research has been devoted to analyzing the
1 This conceptual framework is formally presented in Ohlson [1979] and Garman and
Ohlson [1980]. These studies examine the implications of the nature of the information
and valuation links for the form of the contemporaneous link.
54
This content downloaded from 128.135.240.154 on Fri, 24 Jun 2016 17:45:55 UTC
All use subject to http://about.jstor.org/terms
"reduced-form" characterization of the information link between accounting earnings and future benefits. Since security prices are observ-
able variables while the fundamental variables-expected future benefits-are unobservable, it is not surprising that tests of the information
dend realizations. Dividends are chosen as the valued future attribute for
two reasons: (i) authoritative accounting statements explicitly identify
future cash receipts as the variable about which accounting data should
The empirical tests begin (in section 3) with a test of the risk-adjusted
This content downloaded from 128.135.240.154 on Fri, 24 Jun 2016 17:45:55 UTC
All use subject to http://about.jstor.org/terms
popular press and the academic literature. The cross-sectional correlations between accounting earnings and the present value of the future
dividend realizations are used (in section 4) as an indication of the
information content of accounting earnings with respect to future divi-
dends. The correlation between security price and the present value of
the future dividend realizations is used as a benchmark against which to
rize significant information in addition to that implicit in current dividends. The coefficient on dividends in this multiple regression is negative
and significant at the 5% level. That is, conditional on the level of
accounting earnings, the higher the dividend payment in the current
year, the lower the dividend payments in the future. However, in the
simple regression of present value on current dividends, the coefficient
This content downloaded from 128.135.240.154 on Fri, 24 Jun 2016 17:45:55 UTC
All use subject to http://about.jstor.org/terms
Accounting earnings for fiscal year 19623 and security prices three
months after the end of 19624 are evaluated as predictors of dividend
payments and the capital gain over the next 19 years. The empirical
estimation procedures are also conducted for samples 1962 through 1971,
1962 through 1979, 1963 through 1979, 1963 through 1980, and 1972
through 1980. For these latter samples, only those results that are
qualitatively different from the 1962 through 1980 sample are reported.
The source of the accounting earnings and annual dividend data is the
the sample. Security return data are obtained from the CRSP Monthly
Stock Return File. Securities that do not have a complete record of
returns on the file for the entire five-year period ending fiscal year 1962
are excluded.5 These selection criteria result in a sample of 349 firms.6
Market return data are obtained from the CRSP Monthly Stock Index
File.
5 Financial and insurance companies and utilities are also excluded from the sample.
Regulation and the nature of the assets of these companies may result in noticeably
different valuation and information relationships from those of other types of companies.
6 Since the sample includes only large and stable firms, the results may not be generalizable to a wider sample. However, the aim of this paper is to provide empirical evidence
of the fundamental links associated with the contemporaneous link between accounting
earnings and security price that has been observed in the information content studies. It is
appropriate, therefore, to use this limited sample.
This content downloaded from 128.135.240.154 on Fri, 24 Jun 2016 17:45:55 UTC
All use subject to http://about.jstor.org/terms
58 PETER D. EASTON
T=t
(1)
T
where Pj*t denotes the present value of the dividend stream at time t,
E[djs I Zt] is the expected dividend for security j at time s conditional on
the vector of available information Zt, E[rjT I Zt] is the conditional
expectation of the T period return, and E[dT I Zt] is the conditional
expectation of the liquidating dividend at the end of the holding period.
In this paper, the actual ex post dividend payments (djs) are used as
surrogates for dividend expectations E[djs I ZJ.7 The date of the dividend
payment s is taken as the date on which the check was mailed. The
where r T is the estimate of expected security return over the time period
t to T, -fT and rmT are, respectively, estimates of the risk-free rate and the
expected return on the market over time period t to T, and fjt is the firmspecific estimate of systematic risk (beta) at time t.8 The term structure
of the risk-free rate is estimated from available riskless investments9 at
7 It is inconceivable (and contrary to the model being tested) that a good model of
dividend expectations would not rely on security prices, accounting earnings, and/or current
dividends (as predictors). However, the use of these predictors in this paper would be
circular.
predicted relationships. Obviously, Pjt is measured with an error that is difficult to define.
This error in measurement has a considerable impact on the empirical estimation procedures in this paper, and some sensitivity analyses are conducted to examine the potential
effect of this measurement error.
'These investments include U.S. government and governmental agency bonds. The term
structure estimation employs an optimization procedure that minimizes both the effect of
This content downloaded from 128.135.240.154 on Fri, 24 Jun 2016 17:45:55 UTC
All use subject to http://about.jstor.org/terms
Confronting this problem is, however, beyond the scope of this paper. As
an attempt to avoid the problem the sensitivity of the results to the
T=t
Tt
studies (for example, LeRoy and Porter [1981], Shiller [1981], and
Grossman and Shiller [1981]) have constructed tests of market ration-
Pj*t = Pi t + bi t (5)
and hence:
observations. An algorithm developed by Hoag [1982] is used to calculate the term structure.
10 Ibbotson and Sinquefield [1979] provide historical return data from January 1926December 1978. Based on this data, the average (compound) monthly return on long-term
government bonds over the period January 1926-March 1963 was 0.00269, and the average
monthly return on the CRSP-value-weighted market return index was 0.00780. The average
market premium, therefore, was 0.00511. This average market premium is the estimate of
the expected market premium used in the calculation of expected security returns for firms
with December fiscal year-ends. The expected market premium is calculated similarly for
firms with fiscal years ending in months other than December.
" See Shiller [1981, p. 422] for a detailed discussion of this step.
This content downloaded from 128.135.240.154 on Fri, 24 Jun 2016 17:45:55 UTC
All use subject to http://about.jstor.org/terms
60 PETER D. EASTON
(where aot and ait are regression coefficients and cjt is the regression
disturbance term.)12
The disturbance term ejt in regression (7) includes the difference (bjt)
between the ex post rational price (Pjt) and the security market price
(Pjt). The difference may be interpreted as error in measurement of Pat.
This error arises due to errors in measurement of each of the components
of Pitt (the risk-free rate, the market return, beta, expected dividend
payments, and the expected liquidating dividend). The disturbance term
ejt is assumed to have zero mean but is unlikely to have constant variance.
To overcome this problem the regression parameter covariance matrix
present value per share).14 These choices are a result of the method of
where (xi, ei) is a sequence of independent, not (necessarily) identically distributed random
vectors such that xi (a 1 x k vector) and ci (a scalar) satisfy E(x 'ei) = 0. ,0 is a finite
unknown k x 1 parameter vector to be estimated. Under additional assumptions that
essentially ensure that the (X'X/n) matrix (where X is the k x n matrix made up of the
xi vectors) is nonsingular and finite, White proves that the following covariance estimator
is consistent:
n
The regression results reported in the paper are OLS estimates of the regression coefficients
with the t-statistics based on White's consistent variance-covariance estimator.
14 All regressions reported in this paper use "per share" data. The "per share" specification is not motivated by a need to choose a common "deflator" of "whole firm" data. Rather
the motivation is that an equity share is the fundamental unit of value to the shareholder.
This content downloaded from 128.135.240.154 on Fri, 24 Jun 2016 17:45:55 UTC
All use subject to http://about.jstor.org/terms
the investment period is short, the cash received when the security is
liquidated will be by far the largest cash receipt to the shareholdererrors in the estimate of the price at which the security can be sold at
the end of the holding period will be a primary source of error in the
estimate of present value. With a longer holding period, there will be
80. The difference between P~t+l and Pjt (implicitly, if not explicitly,
over the holding period are common to both P)*t+l and P~t.
The null hypotheses based on regression (7) are aot = 0 and alt = 1.
Rejection of these null hypotheses leads to the conclusion that the
tor behavior; dividends are not the valued future attribute; or the securities market is not rational.
value (specifically aot = 0 and alt = 1) may be too ambitious. For this
Using equation (3), the present value of cash receipts over six different
holding periods (1962-71, 1962-79, 1962-80, 1963-79, 1963-80, and 197115 The price of the security at the end of the holding period is the cash receipt associated
with the liquidation of the investment.
This content downloaded from 128.135.240.154 on Fri, 24 Jun 2016 17:45:55 UTC
All use subject to http://about.jstor.org/terms
62 PETER D. EASTON
80) are determined for each security satisfying the selection criteria
identified in section 2.
The results for the analyses based on regression (7) are reported in
table 1. Although there were some differences across the six sample
periods (1962-71, 1962-79, 1962-80, 1963-79, 1963-80, and 1971-80),
only the results based on the 1962-80 sample are reported in the table
because the differences do not affect the overall conclusions. Since the
data in each of the samples are not independent, we would expect similar
results. On the other hand, the coefficients from cross-sectional regressions of this form have been shown to be unstable over time, and some
instability was evident across the samples. This instability has been
interpreted as a manifestation of cross-sectional dependence in the
disturbance term.16 Qualitative differences in the results across the six
sample periods are reported in the text. When reading the results in this
table note that rA4j is the a priori best estimate of expected return,
discount rates ri3j, through rljr are progressively higher than r4jr, while
discount rates rP5j through ri8j, are progressively lower. Discount rate ri8jr
is the estimate of the risk-free rate.
Consider the results for tests of the null hypothesis ait = 1.17 The
discount rate for which the null hypothesis at = 1 is not rejected (F41r)
is the discount rate based on the a priori best estimate of market return.18
As the discount rate increases from the best estimate, the coefficient ait
becomes increasingly significantly less than one, and as the discount rate
table 1) when Pj*) is based on high discount rates, and more particularly,
the null aot = 0 is rejected when the estimate Pj*] is based on the a priori
best estimate of expected return. Rejection of the null aot = 0 when the
17 t-statistics for tests of the null hypothesis alt = 1 are reported in square brackets.
18 In each sample period, the discount rate for which alt was closest to one was based on
the a priori best estimate of the return on the market or a small digression from it (K =
1.2 or K = 0.8).
This content downloaded from 128.135.240.154 on Fri, 24 Jun 2016 17:45:55 UTC
All use subject to http://about.jstor.org/terms
Results from the Regression of the Present Value of the Ex Post Dividend Stream(Pi) on
Security Price (PjJ
ait
Discount
rij,
(K
Rate
tart)
12.96
1.6)
(6.03)**
(t
0.52
1)
0.55
(28.45)**
[-26.18]**
r2jT
12.96
0.64
0.58
r3jT
(K
12.73
1.2)
(4.07)**
0.78
0.59
(20.67)**
[-5.93]**
r4jT
(K=
1)
12.12
(3.14)**
0.95
0.60
(18.19)**
[-1.03]
r5jT
10.95
1.15
0.61
0.6)
9.04
1.39
(1.53)
(14.90)**
6.22
1.67
0.62
[4.19]**
r7jT
0.63
r81T
(K=
0)
rfT
(-0.25)
-2.74
2.37
0.64
(12.09)**
[6.99]**
(to1t) is the t-statistic for the test of the null hypothesis a1t = 0.
[t,1,I is the t-statistic for the test of the null hypothesis ait = 1.
tions from the predicted relationships. Furthermore, each of the components of expected return (beta, the riskless rate, and the market premium)
are estimated with error. Thus there are several reasons (in addition to
market irrationality) why the null aot = 0 may be rejected. Section 4
In table 1, the t-statistics for the test of the null hypothesis that ait =
0 are highly significant for every method of calculation of Pjtt. Evidently,
This content downloaded from 128.135.240.154 on Fri, 24 Jun 2016 17:45:55 UTC
All use subject to http://about.jstor.org/terms
64 PETER D. EASTON
future dividends are aggregated into a present value using the riskadjusted dividend capitalization formula. Errors in estimating the ex-
pected security returns that are used as the discount rate in this formula
prohibit further conclusions regarding the form of this valuation link.
given time and the present value of future cash receipts are, in this
earnings to present value is constant across firms. Consider the extreme case where
This content downloaded from 128.135.240.154 on Fri, 24 Jun 2016 17:45:55 UTC
All use subject to http://about.jstor.org/terms
dends is shown by the t-statistic for tests of the null hypothesis, b1t = 0.
The relation between accounting earnings and the future stream of
dividends, depicted in regression (8), may not necessarily represent the
information mapping implicitly used by investors. Since accounting
earnings are fundamentally undefined as intrinsic economic variables,
the form of the "true" relationship between accounting earnings and the
future dividend stream is unknown. Unlike the valuation link examined
in section 3, there is no theoretical foundation which suggests the form
of the information link examined in this section. Note, however, that the
purpose of this section is to determine whether accounting earnings are
correlation between security price and the present value of future cash
receipts is an appropriate benchmark against which to compare the
information content of accounting earnings. The following regression is
used for this purpose:
Note that regression (9) is not used to test any notion of a "true"
economic model underlying the relation between accounting earnings
and the present value of the future dividend stream or between security
price and the present value of the future dividend stream. If this were
Pj*t =bljtEjt.
Forcing bljt to be constant cross-sectionally (as in regression (8)) will result in an error
term with variance equal to u2Elt, where a2 is the variance of b1jt. (A similar point is made
by Christie [1984].) The appropriate procedure to overcome this second form of heteroscedasticity is to replace regression (8) with:
variance-covariance estimator proposed by White (and used for regression (7)) was also
used for hypothesis tests based on regression (8).
20 Again, White's consistent variance-covariance estimator was used in the hypothesis
testing.
This content downloaded from 128.135.240.154 on Fri, 24 Jun 2016 17:45:55 UTC
All use subject to http://about.jstor.org/terms
66 PETER D. EASTON
nificance of the bit coefficient are all very significant, showing that
earnings are significantly correlated with present value. The significance
between accounting earnings and present value are of the same order of
magnitude as the correlations between security price and present value
(see table 1), implying that accounting earnings and security price are
calculated using a high discount rate, but c1t is not significantly different
from zero when P}t is calculated using a low discount rate. The significant
incremental explanatory power of accounting earnings is somewhat surprising. The fact that this significance declines as the discount rate
decreases suggests that accounting earnings provide significant incre21 A significant c1t coefficient suggests error in measurement of Pjt and warrants further
investigation of the method of estimation of this variable.
22 The similarity of the results across samples is not surprising in view of the fact that
the observations are by no means independent. In a sense there is only one observation
and the results may not be generalizable beyond the 1962-80 observation period. Since the
time series of Pjt, Ejt, and Pj*t is nonstationary, it is possible that the relationship between
these variables might have been different had the cross-section been based on a different
slice across history. On the other hand, the similarity of the results is a sign that the
multicollinearity of Ejt and Pjt does not seriously affect the validity of conclusions drawn
from these results.
This content downloaded from 128.135.240.154 on Fri, 24 Jun 2016 17:45:55 UTC
All use subject to http://about.jstor.org/terms
Relation Between Accounting Earnings, Security Prices, and the Present Value of the Next
19 Years of Dividend Realizations
Regression (8) Regression (9)
Discount Rate pi*,= bot + b1tEjt + ejt Pjt = Cot + c1tEjt + C2tP t + Pit
bDt bt RR2 Cot Cit C2t R2
(tbo) (tblt) (t.Ot) (t."t) (t-12d
realizations in the near future. If this is the case, then the apparently
significant incremental explanatory power may simply be an artifact
resulting from the use of a discount rate in the calculation of Pj*t that is
too high. A possible explanation for the high cross-sectional correlation
cit coefficient is obtained by replacing Pj*t in regressions (7), (8), and (9)
with the annual dividend realization in each year of available data (that
is, 1962 through 1980). These regressions may be written as follows:
68 PETER D. EASTON
5.0 -
IN
4.6 - --accounting earnings
42 -
6 3.8 -
2.6-
2.2
FIG. 1.-Significance of the association between (1) security price and future annual
dividend realizations and (2) accounting earnings and future annual dividend realizations.
Based on regressions (7') and (8'):
Ho: a, = 0, and
Ho: b, = 0.
Figure 1 graphs the t-statistics for tests of the null hypothesis; a', = 0
and b1' = 0 for estimation intervals 1 to 18 years (that is, predictions for
23 The analyses were repeated using security prices and accounting earnings for each of
the years 1963 to 1979 as predictors of dividend realizations over the remaining years of
available data (that is, 17 years of dividend realizations were predicted using 1963 accounting earnings, etc.). The analyses were also repeated for a cross-section and time series of
all of these data. The results are qualitatively equivalent to those based on the 1962
accounting earnings and security price data reported here. These regressions were also
conducted for portfolios made up of pairs of securities weighted such that the portfolio beta
is one. The coefficient relating price to future annual dividends is a function of both risk
and dividend policy. Aggregation into the portfolios had the aim of, at least, reducing the
error associated with cross-sectional variation in the coefficient. The results were qualitatively equivalent to those reported here.
This content downloaded from 128.135.240.154 on Fri, 24 Jun 2016 17:45:55 UTC
All use subject to http://about.jstor.org/terms
0.8 CM
_o0
Inre
etal
explanation
0.
1962
1965
1968
1971
1974
1977
1980
FIG. 2.-Decomposition of the coefficient of determination of (R2) from multiple regressions of future annual dividend realizations on accounting earnings and security price.
Regressions
RT 2 R2 =RE2
-incremental explanation provided security price alone
each of years 1963 to 1980). The correlation between security price and
annual dividends and between accounting earnings and annual dividends
and c2T are significantly different from zero (that is, statistically significant incremental explanatory power in accounting earnings and security
This content downloaded from 128.135.240.154 on Fri, 24 Jun 2016 17:45:55 UTC
All use subject to http://about.jstor.org/terms
70 PETER D. EASTON
respectively, denoted R12 and R22, then RT2 _ R12 = RP2 represents the incremental
explanation of the cross-sectional variation in annual dividends provided by security price,
while RT2 _ R22 = RE2 represents the incremental explanation provided by accounting
earnings. The remaining component of RT2 (i.e., RT2 _ RE2 _ RP2) represents the information
common to both earnings and price.
This content downloaded from 128.135.240.154 on Fri, 24 Jun 2016 17:45:55 UTC
All use subject to http://about.jstor.org/terms
equity investment.26 Tests of the information link between current dividends and future cash receipts are based on the following regression:
the bit coefficient from regression (8) (see table 2) provides an indication
of the relative strength of the information links between accounting
earnings and future cash receipts and current dividends and future cash
receipts. A more precise comparison is based on the regression:
[1956] and Fama and Babiak [1968]), current dividends will be good
predictors of dividend realizations in the short-run future. Since higher
avoids problems associated with measurement of P5t. However, the issue of whether
accounting earnings are a good summary of expected future benefits is not addressed.
26 Staubus [1965] examined the information content of accounting data from this point
of view.
This content downloaded from 128.135.240.154 on Fri, 24 Jun 2016 17:45:55 UTC
All use subject to http://about.jstor.org/terms
72 PETER D. EASTON
examine this contention, regressions (10) and (11) are repeated with
present value replaced by the annual dividend realization in each year of
available data. That is:
and:
where dj, is the dividend realization in future period i-. The reported
results are for the 18 regressions in which t = 1962 and X- 1963, ...
1980.
30.0 -
26.0
22.0 -
CO 18.0 .f
.
CO 14.0
10.0
6.02.0
Ho: a, = 0.
This content downloaded from 128.135.240.154 on Fri, 24 Jun 2016 17:45:55 UTC
All use subject to http://about.jstor.org/terms
0.8
cc
? ~. . . . . . . N~.. . . . . / provided by
o 0.6
Incromental explanation
m Eoplanation provided by
0 2- W c urrent d ividend s
0.0 0 IJI/
1962 1965 1968 1971 1974 1977 1980
FIG. 4.-Decomposition of the coefficients of determination (R2) from multiple regressions of future annual dividend realizations on current annual dividend realizations and
security prices.
Regressions
__
a',
',Djt
Zin
(10
')R-22
di,=a',+aaDPt+zej (81') R2
Decomposition of RT2 (the "total" explanation of cross-sectional variation in dj,):
-incremental explanation provided by current dividends alone
RT2 R2 = RD2
-incremental explanation provided accounting earnings alone
RT 2 R 2 RE2
-explanation provided by both accounting earnings and security price
RT 2 _RD2 RE2.
counting earnings as predictors of the next four years of dividend realizations, but accounting earnings provide significant incremental explanatory power for dividend realizations over the years 1967 through 1980.
The results from regressions (10) and (11) are summarized in table 3.
As expected from the analyses of regressions (10') and (11'), these
results are sensitive to the choice of the discount rate in the present
value calculation. Despite the high correlation between current dividends
and dividend realizations for a considerable length of time into the future,
the correlation between current dividends and Pt (see table 3) is low-
This content downloaded from 128.135.240.154 on Fri, 24 Jun 2016 17:45:55 UTC
All use subject to http://about.jstor.org/terms
74 PETER D. EASTON
TABLE 3
Relation Between Accounting Earnings, Current Dividends, and the Present Value of the
Next 19 Years of Dividend Realizations
Regression (10) Regression (11)
Discount Rate Pi*t = aot + atDjt + zjt Pi*, = Sot + #ItEit + f2tDit + Yjt
aot
ait
R2
/3ot
#it
02t
and PR (see table 2). The t-statistics on the coefficient alt are positive
and they increase as the discount rate increases. This result may be
explained by the observation (in figure 3) that the correlation between
current dividends and future dividend realizations declines as the time
interval between current dividends and the future dividend realization
increases.
provide significant incremental explanatory power (of the cross-sectional variation in PRt) over current dividends. It is interesting to note
that current dividends provide significant incremental explanation of the
results in figures 3 and 4 which suggest that a higher discount rate would
for all discount rates used in the calculations of P)t, the coefficient /2t is
negative. That is, conditional on the level of accounting earnings, the
higher the current dividends, the lower the present value of the future
dividends. This result is a reflection of the trade-off between dividends
This content downloaded from 128.135.240.154 on Fri, 24 Jun 2016 17:45:55 UTC
All use subject to http://about.jstor.org/terms
capital gains).27 The negative 02t coefficient lends support to the claim
that accounting earnings are a good information number in the following
sense. If all other information is ignored, higher current dividends imply
higher future dividends (a,, is positive), but for a given level of accounting
earnings, higher current dividends imply lower future dividends (12t is
negative).28 These results are consistent with the notion that accounting
earnings reflect the dividend-paying ability of the firm and that these
earnings may be either paid in the current period or reinvested.29
In summary, the results in this section support the contention that
accounting earnings are a good summary information number. The
information summarized in accounting earnings is by no means subsumed
by the information in current dividends. In fact, information in accounting earnings is apparently useful in interpreting the information in
current dividends.
cial statement data (principally accounting earnings and current dividends) and abnormal security returns. This paper examines empirically
the proposition that the contemporaneous association reflects the information link between financial statement data and future benefits (meas-
Tut = lit, Var(TIt) = Var(flt) and T2t = (Olt + #2t)From the results in table 3 it can be seen that TI2t > 0 for every method of calculating P]*t.
Thus the negative /2t coefficient is due to the inclusion of dividends in the earnings variable.
28 The negative sign is also obtained when Pj*t in regression (9) is replaced by Pjt. For
this regression using 1962 data, the coefficient /2t is significant at the 10% level. Although
the purpose of this section is to examine the information link, the fact that similar results
those results nullify the argument that the results based on Pj*t may simply be an artifact
of the method of calculating Pj*t.
29 The negative sign of the f2t coefficient is not robust across time periods or across
subsamples in any one time period.
This content downloaded from 128.135.240.154 on Fri, 24 Jun 2016 17:45:55 UTC
All use subject to http://about.jstor.org/terms
76 PETER D. EASTON
earnings are a good predictor of the future stream of cash receipts from
an equity investment and that there is a strong correlation between
security price and the present value of future cash receipts. Considered
AHARONY, J., AND I. SWARY. "Quarterly Dividend and Earnings Announcements and
BAR-YOSEF, S., AND H. E. LELAND. "Risk Adjusted Discounting." Working paper no. 134,
Institute of Business and Economic Research, University of California, Berkeley, December 1982.
BEJA, A. "Capital Markets with Delayed Learning." Ph.D. dissertation, Stanford University,
1967.
CHAMBERS, A. E., AND S. H. PENMAN. "Timeliness of Reporting and the Stock Price
Reaction to Earnings Announcements." Journal of Accounting Research 22 (Spring 1984):
21-47.
COPELAND, B. L., JR. "Do Stock Prices Move Too Much to Be Justified by Subsequent
Changes in Dividends?: Comment." American Economic Review 73, no. 1 (1983): 234-35.
FAMA, E. F. "Risk Adjusted Discount Rates and Capital Budgeting Under Uncertainty."
Journal of Financial Economics 5 (1977): 3-24.
, AND J. OHLSON. "Information and the Sequential Valuation of Assets in ArbitrageFree Economies." Journal of Accounting Research 18 (Autumn 1980): 420-40.
This content downloaded from 128.135.240.154 on Fri, 24 Jun 2016 17:45:55 UTC
All use subject to http://about.jstor.org/terms
ments: Critique and Extension." Journal of Financial and Quantitative Analysis (June
1981): 193-206.
LAUB, P. M. "Some Aspects of the Aggregation Problem in the Dividend-Earnings Relationship." Journal of the American Statistical Association 67 (September 1972): 552-59.
LEROY, S. F., AND R. D. PORTER. "The Present-Value Relation: Tests Based on Implied
Variance Bounds." Econometrica 49, no. 3 (1981): 555-74.
MARSH, T. A., AND R. C. MERTON. "Dividend Variability and Variance Bounds Tests for
the Rationality of Stock Market Prices." Working paper, Massachusetts Institute of
Technology, August 1984.
Ross, S. A. "The Arbitrage Theory of Capital Asset Pricing." Journal of Economic Theory
13 (1976): 341-60.
RUBINSTEIN, M. "The Valuation of Uncertain Income Streams in the Pricing of Options."
Bell Journal of Economics (1976): 407-25.
SHILLER, R. J. "Do Stock Prices Move Too Much to Be Justified by Subsequent Changes
in Dividends?" American Economic Review 71, no. 3 (1981): 421-36.
STAUBUS, G. J. "The Association of Financial Accounting Variables with Common Stock
Values." The Accounting Review (January 1965): 119-34.
THEIL, H. Principles of Econometrics. New York: Wiley, 1971.
. "Comments on the Impact of Dividends on Earnings Announcements: A Reconciliation." Journal of Business 49 (January 1976): 97-106.
This content downloaded from 128.135.240.154 on Fri, 24 Jun 2016 17:45:55 UTC
All use subject to http://about.jstor.org/terms