Download Institutional Ownership and the Extent to which Stock

Institutional Ownership and the Extent to which
Stock Prices Reflect Future Earnings
James Jiambalvo
PricewaterhouseCoopers and Alumni Professor
Department of Accounting
University of Washington
Seattle, WA 98195
Tel: 206.543.9132
E-mail: [email protected]
Shivaram Rajgopal
Assistant Professor
Department of Accounting
University of Washington
Seattle, WA 98195
Tel: 206.543.7525
E-mail: [email protected]
Mohan Venkatachalam
Assistant Professor
Graduate School of Business
Stanford University
Stanford, CA 94305
Tel: 650.725.9461
E-mail: [email protected]
June 2000
We acknowledge comments from two anonymous referees, John Wild (Associate Editor), Holly Ashbaugh,
Ramji Balakrishnan, Bob Bowen, Brian Bushee, Tom Carroll, Maureen McNichols, Karen Nelson, K.R.
Subramanyam and workshop participants at the University of Arizona and the 1997 Stanford Accounting
Summer Camp. Discussions with Terry Shevlin have been especially helpful. Mohan Venkatachalam
appreciates funding from the Financial Research Initiative at Stanford University Graduate School of
Business.
Institutional Ownership and the Extent to which
Stock Prices Reflect Future Earnings
Abstract
Articles in the financial press suggest that institutional investors are overly focused on current
profitability. This suggests that as institutional ownership increases, stock prices will reflect less
current period information that is predictive of future period earnings. On the other hand,
institutional investors are often characterized in academic research as sophisticated investors.
Sophisticated investors should be better able to utilize current period information to predict future
earnings compared to other owners. According to this characterization, as institutional ownership
increases, stock prices should reflect more current period information that is predictive of future
period earnings. And consistent with this later view, we find that the extent to which stock prices
lead earnings is positively related to the percentage of institutional ownership. This result holds
after controlling for various factors that affect the relation between price and earnings. It also holds
when we control for endogenous portfolio choices of institutions (e.g., institutional investors may
be attracted to firms in richer information environments where stock prices tend to lead earnings).
Further, a regression of stock returns on changes in order backlog, conditional on the percentage of
institutional ownership, indicates that institutional owners place more weight on order backlog
changes compared to other owners. This is consistent with institutional owners using non-earnings
information to predict future earnings and explains, in part, why prices lead earnings to a greater
extent when there is a higher concentration of institutional owners.
1.
Introduction
A number of articles in the financial press suggest that institutional investors pressure managers to
achieve short-term profit goals at the expense of long-term equity value (e.g., Teitelman 1993;
Coffee 1991; Jacobs 1991; Chote and Linger 1986; Drucker 1986; Dobrzynski, et al. 1986).
According to Porter (1992), institutional investors are transient owners who are overly focused on
short-term earnings. Fearing that a short-term profit disappointment will lead institutions to
liquidate their holdings (leading to at least a temporary decline in equity value), managers are
compelled to take actions that increase short-term profit. The most commonly cited action is
cutting research and development expenditures. However, academic research has documented
results at odds with this hypothesized behavior. Both Bushee (1998) and Wahal and McConnell
(1997) find a positive relation between research and development and institutional ownership.
Further, institutional investors are often characterized as sophisticated investors who have
advantages in acquiring and processing information compared to individual investors (e.g., Hand
1990; Kim, et al. 1997; Bartov et al. 2000). If institutional investors are sophisticated, they should
be better able to utilize current period information to predict future earnings (compared to other
investors) and current period stock prices should reflect more of the information in future period
earnings as institutional ownership increases.
The purpose of our study is to provide evidence on the extent to which stock prices lead
earnings conditional on institutional ownership. We find that the extent to which stock prices lead
earnings is positively related to the percentage of institutional ownership. This result holds after
controlling for various factors that affect the relation between stock prices and earnings (e.g., firm
size, market to book, and standard deviation of earnings). It also holds when we control for
1
endogeneity related to the possibility that institutional investors are attracted to firms in richer
information environments where stock prices tend to lead earnings.
One piece of nonfinancial information that institutional owners may use to predict future
earnings is the change in order backlog. A regression of the change in order backlog on stock
returns, conditional on the percentage of institutional ownership, indicates that institutional owners
place more weight on order backlog changes compared to other owners. 1 This explains, in part,
why prices lead earnings to a greater extent when there is a concentration of institutional owners.
Overall, our results are consistent with the view that institutional owners are sophisticated
investors whose information acquisition and processing advantages are reflected in stock prices.
The results contradict the notion that institutional investors (compared to other investors) fixate on
current earnings since current stock prices are more likely to reflect future earnings as institutional
ownership increases. In addition to shedding light on the characterization of institutional investors,
the results have implications for future studies investigating the relation between earnings surprises
and stock returns. Controlling for institutional ownership may reduce measurement error in
earnings surprise proxies since institutional owners are less likely to be surprised by future earnings
realizations.
The remainder of our paper is organized as follows. In the next section we briefly review
related studies and develop our hypotheses. Section three outlines our research methodology while
section four describes the data used in tests of competing views. Empirical findings are presented in
section five and concluding comments are presented in the final section.
1
Our results obtain after controlling for managerial ownership. We assume that other than institutional investors and
managers, the only major investor category is individuals. A similar assumption is made in Bushee (1998) and in Lang
and McNichols (1997).
2
2.
Hypothesis development
Institutional owners as transient investors focused on current earnings
Institutional investors are often characterized as transient owners who are overly focused on current
earnings. For example, Porter (1992) notes:
“Perhaps the most basic weakness in the American system is transient ownership in which
institutional agents are drawn to current earnings, unwilling to invest in understanding the
fundamental prospects of companies, and unable and unwilling to work with companies to
build long-term earnings power.”
According to Jacobs (1991) firms cannot hope to establish a viable dialogue with such transient
investors over the future prospects of the corporation. Graves (1988) argues that fund managers
cannot afford to take a long-term view in their investment decisions since they are reviewed and
rewarded on the basis of quarterly, or at most, annual performance measures. Consistent with
institutional transience, Potter (1992) and Kim, et al. (1997) report greater stock return volatility and
trading volume surrounding earnings announcements of firms with higher institutional ownership.
Further, Lang and McNichols (1997) find that institutional trading is related to the market’s reaction
to earnings announcements. However, Eames (1997) finds only limited evidence of changes in
institutional holdings after earnings announcements.
In a recent study, Bushee (1999) regresses institutional ownership on firm value components
related to expected near-term earnings and expected long-term earnings. He finds that the mean
coefficient on near-term earnings is positive and the mean coefficient on long-term earnings is
negative consistent with institutional investors having preferences for value in near-term earnings
compared to long-run value. However, the results are not statistically significant. For institutional
investors classified as transient investors, Bushee reports convincing evidence of a preference for
near-term earnings.
3
Institutional owners as sophisticated investors
Contrary to the view that institutional owners are transient investors, who are overly focused on the
short-term, is the view that they are sophisticated investors with advantages in acquiring and
processing information. This perspective is supported by a survey conducted by Shiller and Pound
(1989) who find that institutional investors spend more time performing investment analysis. Lev
(1988) argues that well-endowed investors have access to information that is too costly for others to
acquire. And given that institutional investors, on average, have more resources than individual
investors, they are likely to be better informed.
Walther (1997) finds that in regressing stock returns on unexpected earnings measures, the
weight on analysts’ forecasts of earnings increases and the weight on seasonal random-walk
forecasts of earnings decreases with the percentage of stock held by institutional investors. Since
analysts’ forecasts are more accurate than seasonal random-walk forecasts, her result is consistent
with the idea that institutional owners are more sophisticated than other owners. Further evidence
on the sophistication of institutional owners is provided in a recent study conducted by Bartov,
Radhakrishnan, and Krinsky (2000).2 They show that the pattern of observed post-earningsannouncement abnormal returns documented by Rendleman, Jones and Latane (1987), Freeman and
Tse (1989), and Bernard and Thomas (1990) is reduced as the proportion of firm shares held by
institutional investors increases. Thus, institutional owners are more sophisticated than other
owners in that they are more likely to appreciate the time-series implications of a quarterly earnings
innovation.
2
Our study was developed independently of the work by Bartov et al. (2000).
4
Hypotheses
It is well known that stock prices lead earnings (Kothari and Sloan, 1992). This occurs
because not all economic actions taken by the firm are reflected in current period earnings and
information about such actions (e.g., information about customer orders, long-term sales contracts,
and investment activities) are available to investors. Such economic actions, though not reflected in
current earnings, will eventually be reflected in future period earnings. We posit that sophisticated
investors with advantages in acquiring and processing information will impound relatively more
information about future earnings, not currently reflected in current earnings, in their investment
decisions. If investors impound value relevant information not reflected in current earnings, current
stock prices will reflect information about future earnings after controlling for current earnings. To
the extent that institutional investors are more sophisticated than other investors, the proportion of
information in future earnings relative to current earnings, reflected in current period stock prices,
should increase with institutional ownership.3 On the other hand, if institutional investors are
myopically focused on current period earnings, this relation will not hold (or may even be reversed).
Accordingly, we test the following null hypothesis:
HYPOTHESIS 1: The proportion of information in future earnings relative to current earnings,
reflected in current stock prices, is not related to the level of institutional ownership.
An alternative approach to testing whether institutional investors are more sophisticated than
other investors is to investigate whether the pricing of forward-looking information that will be
reflected in future earnings is influenced by the level of institutional ownership. We consider order
backlog as one such item of forward looking information that will be incorporated in future
earnings. To the extent that institutional investors are more sophisticated than other investors, we
3
This assumes, as in Walther (1997) and Bartov et al (2000), that the likelihood that the marginal investor is
sophisticated increases with the percentage ownership by institutional investors.
5
posit that information in order backlog reflected in current period stock prices should increase with
institutional ownership. Accordingly, we test the following hypothesis (stated in null form):
HYPOTHESIS 2: The extent to which order backlog changes are reflected in stock prices is not
related to the level of institutional ownership.
3. Research method
We adopt an approach suggested by Kothari and Sloan (1992) to assess the extent to which
stock prices reflect a greater proportion of information in future versus current earnings for
institutional investors.4 They note that stock prices impound information that will only later be
reflected in accounting earnings and suggest an equation of the following form:
Rit,t-τ = ω 0 + ω1(τ ) Eit Pit −τ + ε it
(1)
where Rit,t-τ is the buy-and-hold return for firm i over the period t-τ to t, Eit is income before
extraordinary items for the accounting period ended at time t and Pit-τ is the stock price at end of
period t-τ. ω1(τ) represents the market’s response during the period t-τ to t to earnings information
for the period ended at time t.
Because stock prices lead accounting earnings, as τ increases, it is more likely that the
information contained in earnings at time t will be incorporated in the return over the period t-τ to t.
This is because capital market participants likely impound information about current earnings in
earlier time periods. To the extent that information contained in current accounting earnings has
already been incorporated in stock prices of a previous period, the coefficient ω1(τ) will get smaller
(larger) as the time interval τ gets smaller (larger), i.e., ω1(τ=2) > ω1(τ=1). Furthermore, the ratio of
4
This approach has also been used by Jacobson and Aaker (1993) to demonstrate that stock prices in Japan reflect a
greater proportion of the information in future as opposed to current earnings. Here we use the approach to analyze
whether, within the U.S., the proportion of information in future versus current earnings is related to institutional
ownership.
6
ω1(τ) obtained for a longer time interval to that obtained for a shorter time interval (ω1(τ=2) / ω1(τ=1))
provides a measure of the extent to which information in current earnings has been impounded in
prices in an earlier time period (see Figure 1). A higher ratio indicates that more information in
current earnings has already been incorporated in past stock prices.
To examine the influence of level of institutional ownership on the extent of information
about future earnings impounded in earnings, we conduct two tests. First, we interact the earnings
term with the level of institutional ownership and examine whether stock prices of firms with
greater institutional ownership incorporate more information about future earnings. That is, we
modify equation (1) as follows:
Rit,t-τ = ω 0 + ω 1(τ ) E it Pit −τ + ω 2(τ ) E it * INSTit Pit −τ + ε it
(2)
where INST represents the percentage of institutional ownership. Equation (2) is estimated for both
τ =1 and τ =2 using a seemingly unrelated regression (SUR) approach. By comparing ω2(τ=2) and
ω2(τ=1) we examine the influence of institutional ownership on the extent to which information in
future periods are incorporated in current stock prices. If current stock prices of firms with higher
institutional ownership incorporate more information about future earnings then we should expect
ω2(τ=2) to be greater than ω2 (τ=1).5
Next, we partition institutional ownership into quintiles and examine the ratio
(ω1(τ=2) / ω1(τ=1)) across the ownership quintiles. If institutional owners are sophisticated investors,
then the stock price for firms with large institutional ownership should impound information in
current earnings earlier than the stock prices of firms with small institutional ownership. Therefore,
5
This result would not be expected to be hold if institutions are indexers with no incentives to gather information on the
firm.
7
we expect the ratio (ω1(τ=2) / ω1(τ=1)) to be higher for firms with large institutional ownership
compared to firms with small institutional ownership.
The tests suggested above assess whether institutional investors are sophisticated investors
who incorporate information about future earnings in determining share prices. However, they do
not directly test whether institutional investors incorporate forward looking information, that will be
reflected in future earnings. To address this, we consider order backlog changes as a plausible piece
of forward looking information.6 We estimate the following equation to examine the extent to
which stock prices incorporate order backlog information:
Rit,t −1 = γ 0 + γ 1 Eit + γ 2 ∆BACKLOG it + γ 3 ∆BACKLOG it * INSTit + υ it
(3)
where ∆Backlog represents change in order backlog from period t-1 to t.
If the market impounds forward looking information such as backlog changes, we expect γ2
to be positive. And if institutional owners are sophisticated investors in processing such forward
looking information, then the stock price for firms with large institutional ownership are more likely
to reflect the impact of backlog changes than firms with smaller institutional ownership. Thus we
expect γ3 to be positive. However, if institutional investors are myopically focused on short-term
accounting earnings, then the likelihood of stock prices reflecting backlog changes (that are
expected to be reflected in future earnings) should decline with higher institutional ownership.
Under that scenario, we expect γ3 to be negative.
4. Data
To conduct the empirical tests outlined in the previous section we obtain data from two
6
A regression of earnings in t on changes in order backlog in t-1 indicates a significant positive relation between the
two variables (p < 0.01, R-squared =0.03).
8
sources. First, we gather ownership data from the Disclosure Database distributed by Disclosure
Incorporated for the years 1989 to 1995. The database reports the percentage of outstanding shares
owned by institutions and corporate owners. These ownership data are reproduced in the database
from SEC filings 13-F, 13-D, 13-G, 14-D, and Forms 3 and 4. In a comparative study of the
reliability of ownership data from several databases, Anderson and Lee (1996) conclude that the
Disclosure Database ranks very favorably over peer databases. Next, we obtain financial data and
stock price data from the 1995 COMPUSTAT annual tapes.
The final sample consists of 10,500 firm-year observations. The sample selection
procedure, summarized in Table 1, consists of three stages. We initially obtain an initial sample of
institutional ownership data for 38,221 firm-years during the period 1989-95 from the Disclosure
database. Of these, a number of firms-years are not found in the 1995 COMPUSTAT tapes reducing
the sample to 18,014 firm-years. We eliminate 1,056 observations that are in the top and bottom 1%
of the returns and earnings distributions. We also drop 6,458 firm-years for which data on analyst
following are not available from the IBES tapes. This filter is necessary to control for the
possibility that analyst following may be correlated with institutional ownership and the presence of
analyst forecasts would facilitate the market’s pricing of future earnings or backlog changes in
current periods. The selection procedure yields 10,500 firm year observations for our empirical
analysis.
Table 2 presents the descriptive statistics for the variables used in empirical tests. Of
particular interest is the level of institutional ownership as measured by the percentage of their
stockholdings relative to the total shares outstanding. The mean (median) percentage institutional
ownership is 41% (40%). The distribution of percentage institutional ownership is comparable to
that reported in prior research (e.g., Eames, 1997, Bushee, 1998). For the median firm, institutional
9
owners collectively own four times as much equity as managers (40 % stake for institutions
compared with 9% of managers). Hence, institutions are likely to have substantial incentives to
invest in gathering information about the firms in which they invest. The mean and median total
assets (unreported) are $3510 million and $317 million respectively, indicating a positively skewed
distribution of firm size. The average income before extraordinary items scaled by lagged market
value is 0.05. Among the cross-correlations examined (not reported), two correlation statistics are
especially noteworthy. Firm size (SIZE), as measured by the logarithm of market value of equity, is
highly associated with percentage institutional ownership and the correlation between the two
variables is 0.54 (p < 0.01). Hence, any analysis that does not control for firm size will face a
potentially large omitted-variable bias. Similarly, the strong negative correlation (-0.30, p < 0.01)
between institutional ownership and managerial ownership highlights the importance of controlling
for managerial ownership while examining the effects of institutional ownership.
5.
Results
Analysis of the Extent of Future Earnings Reflected in Stock Prices
The results of estimating equation (2) using a pooled fixed-year effects model is reported in
Table 3.7 Recall that we need to estimate equation (2) for two time periods (τ = 1 and 2). To
facilitate comparison of coefficients across time periods we estimate equation (2) for both time
periods as a seemingly unrelated regression. Recognizing that several factors affect the relation
between earnings and returns in addition to institutional ownership we control for such factors by
interacting earnings with those variables in our empirical analyses. Specifically, we control for firm
7
To address the cross-correlation and serial correlation inherent in a panel data approach, we also estimated all
regressions in the paper year-by year and computed Z statistics to test coefficient estimate significance across years as
suggested by Bernard (1987). The tenor of our results was similar to those reported in the main body of the paper.
10
size, analyst following, market to book, lagged earnings, earnings variability, leverage and
managerial ownership. We include firm size and number of analysts to proxy for the information
environment while market to book proxies for growth opportunities. Lagged earnings is included to
control for auto-correlation in earnings and earnings variability (measured by the standard deviation
of annual earnings divided by lagged total assets) is included because studies of firm valuation have
shown it to affect the earnings-return relation (e.g., Collins, et al. 1987; Kormendi and Lipe 1987;
Collins and Kothari 1989). Leverage (total debt divided by market value of equity) and managerial
ownership are included because they are related to accounting choices that affect the
informativeness of earnings (Warfield, et al. 1995).
Table 3 presents results of estimating equation (2) with and without including the control
variables. Note that the coefficient on earnings for stock returns estimated over the longer window
(τ =2), i.e., ω1(τ =2) is 2.75 while the coefficient estimated for the shorter window (τ =1), i.e., ω1(τ=1)
is 1.12. This is consistent with the notion that stock prices lead accounting earnings (Kothari and
Sloan, 1992). In other words, since the information in period t earnings is partially impounded in
stock prices from period t-2 to t-1, the reaction to period t earnings will be greater for return
window t-2 to t compared to return window t-1 to t. Our primary focus, however, is comparing
coefficients, ω2 (τ=2) and ω2 (τ=1), viz., the coefficients on the interaction of institutional ownership
and earnings. Note that ω2 (τ=1) is 0.81 whereas ω2 (τ=2) is 3.81. These results are consistent with the
hypothesis that institutional owners are sophisticated investors since stock prices incorporate
information earlier when institutional ownership is relatively high. If institutional owners were
more focused on current earnings, compared to other investors, opposite results would obtain. Our
inferences are unaltered when we control for the variables discussed earlier. In particular, the
11
coefficient on the interaction term between institutional ownership and earnings increases when the
stock return window increases (i.e., ω2 (τ=2) = 3.69 > ω2 (τ=1) = 0.83). The coefficient on the control
variables are all statistically significant with the predicted sign with the exception of the number of
analysts variable.8
In addition to the interaction analysis reported above, we focus on the ratio of coefficients
for short and long intervals, not the coefficients themselves. That is, we test the prediction that the
ratio of ω1(τ=2) to ω1(τ=1) increases across quintiles of institutional ownership. To conduct these tests,
we modify equation (1) in the following way. We drop the earnings variable and the earnings institution interaction term from the regression.9 Instead, we allow the coefficient on earnings to
vary as a function of the quintile membership of institutional ownership. As before, we account for
control variables described above and estimate the regression across quintiles as a seemingly
unrelated regression. Results of estimating the modified version of equation (1) across quintiles are
reported in Table 4. As expected, we find that the coefficient on earnings increases with the length
of measurement interval in which returns are measured, i.e., ω1(τ=2) > ω1(τ=1) across all quintiles.
More importantly, we find that the ratio for institutional ownership quintile 5 (2.02) is higher than
the ratio for quintile 1 (1.37). To evaluate the statistical significance of that ratio, we conduct a
simulation analysis as in Jacobson and Aaker (1993). Essentially, we generate 1,000 normal
observations with the mean and variance characteristics reported in Table 4 for ω1(τ=2) and ω1(τ=1).
Using these simulated observations of ω1(τ=2) and ω1(τ=1), we obtain 1,000 ratios of the coefficients
for the highest and lowest institutional ownership quintile. We find (results not reported) that the
8
We are unable to offer an explanation for the anomalous result related to this control variable. Given the attention
paid to analysts forecasts in accounting research, future work aimed to an explanation seems clearly warranted.
9
The earnings term must be dropped to prevent the regression from overidentification problems.
12
mean ratio of the simulated coefficients for the high institutional ownership quintile is greater than
the mean ratio of the simulated coefficients for the low institutional ownership quintile at the 0.01
level (two-tailed test).10
Analysis of order backlog changes
Next we estimate equation (3) to examine the relation between order backlog information
and stock returns, conditional on institutional ownership. Table 5 presents results of estimating
equation (3). The drop in sample size from 10,500 to 3,426 firm-years is due to the absence of
order backlog information for the excluded firms. To maintain consistency with the equations (1)
and (2), we augment equation (3) by the determinants of the returns-earnings regression discussed
above as control variables. As expected, the coefficient on change in order backlog is positive and
statistically significant (coefficient = 0.08, p < 0.01). More importantly, we find that the pricing of
order backlog information increases with the level of institutional ownership. The coefficient on the
interaction of changes in order backlog and percentage institutional ownership is 0.20 and
statistically significant at the 0.01 level. These results provide additional support for the idea that
prices of firms with higher levels of institutional ownership impound more information about future
earnings. The coefficients on all control variables are significant and have the same signs observed
in Tables 2 and 3.
Endogeneity
One plausible explanation for obtaining the results documented above is that institutions
seek to invest in firms with richer information environments. For example, Gompers and Metrick
(1998) show that, on average, institutions prefer firms that are large and have high book-to-market
10
Table 1 shows that the distribution of stock returns is right-skewed. To check whether such skewness impacts
inferences, we re-ran our tests using a logarithmic transformation of returns (i.e., ln (1+Rit)) as the dependent variable.
The tenor of the conclusions is, however, unchanged.
13
ratios. And, O’Brien and Bhushan (1990) report that institutional investors typically invest in firms
that are widely followed by security analysts. One way to address such endogeneity is to include
interactions of earnings with size (SIZE), market-to-book (MB), and analyst following (NANAL) in
regressions (1) and (2). Note that the empirical specification reported in Tables 3 and 4 already
include these interaction terms. Turning to the backlog regression, we introduced the interactions of
backlog changes with SIZE, MB and NANAL to regression (3) and found that inferences from these
revised specifications (untabulated) are similar to those already reported in Table 5.
These sensitivity checks, however, do not address the possibility that institutions and returns
may be simultaneously determined. To address that possibility, we conducted a two-stage least
squares regression where equations (1) and (2) and (3) were simultaneously estimated in turn with a
specification where institutional ownership (INST) was the dependent variable and the chosen
instrumental variables were firm size, market-to-book, number of analysts, managerial ownership
and leverage while stock returns (Rit,t-τ ) was the hypothesized endogenous variable. The tenor of
our inferences remains unchanged from those reported in Tables 3, 4, and 5.
6. Conclusion
In this paper, we test two competing views of institutional owners. One view is that
institutional owners are overly focused on current financial performance. If this is the case,
compared to other investors, institutions are less likely to consider future period earnings in pricing
securities. An opposing view is that institutional owners are sophisticated investors with better
information processing capabilities and hence, stock prices of firms with higher institutional
ownership will tend to reflect a relatively greater proportion of the information in future period
earnings. We find results consistent with the latter view. Specifically, we find that for firms with
14
higher levels of institutional ownership relatively more future earnings information is impounded in
stock prices in comparison to firms with lower institutional ownership. Studies of firm valuation
and contracting suggest a number of factors that need to be controlled when examining earningsreturn relations. Controlling for these factors, however, does not change the inferences with respect
to institutional ownership in our study.
A regression of returns on changes in order backlog, conditional on institutional ownership,
indicates that institutional investors place more weight on changes in order backlog compared to
other investors. Since changes in order backlog are related to future earnings, this analysis provides
additional support for the contention that institutional owners are sophisticated investors. Recent
work by Bartov et al. (2000) which suggests that inefficient pricing of earnings is reduced when
institutional ownership is high, also supports this view.
This is not to say however, that all institutional owners are sophisticated investors who are
more likely to consider future earnings in pricing securities. Indeed, research by Bushee (1998)
suggests that a sub-category of institutional investors (specifically those with high portfolio
turnover, diversification and following a momentum trading strategy) may overly focus on current
period earnings. However, our research contributes to the debate regarding institutional ownership
by demonstrating that on average stock prices are more likely to reflect future earnings when
institutional ownership is high which is consistent with institutional investors possessing
information acquisition and information processing advantages.
15
References
Anderson, R.C. and D.S. Lee. 1996. Ownership databases: Which brand is right for you? Working
Paper, Texas A & M University.
Bernard. V. 1987. Cross-sectional dependence and problems in inference in market-based
accounting research. Journal of Accounting Research: 1-48.
Bernard, V.L. and J. Thomas. 1990. Evidence that stock prices do not fully reflect the implications
of current earnings for future earnings. Journal of Accounting and Economics: 305-340.
Bartov, E. S. Radhakrishnan and I. Krinsky. 2000. Investor sophistication and patterns in stock
returns. The Accounting Review: 43-63.
Bushee, B. 1998. Institutional investors, long term investment, and earnings management. The
Accounting Review: 305-333.
Bushee, B. 1999. Do institutional investors prefer near-term earnings over long-run value? Working
paper, Harvard University.
Chote, P. and J.K. Linger. 1986. Business and the short term syndrome. Washington Post (June 15):
F1-F2
Clikeman, P.M. 1996. Constraints to earnings management: An empirical study of auditor type and
institutional ownership. Working Paper, University of Richmond.
Coffee, J.C., Jr. 1991. Liquidity versus control: The institutional investor as a corporate monitor.
Columbia Law Review: 1277-1368.
Collins, D.W. and S.P. Kothari. 1989. An analysis of the intertemporal and cross-sectional
determinants of earnings response coefficients. Journal of Accounting and Economics: 143-182.
Collins, D.W., S.P. Kothari and J.D. Rayburn. 1987. Firm size and the information content of prices
with respect to earnings. Journal of Accounting & Economics: 111-139.
Dobrzynski, J. H., Z.Schiller, G. L. Miles, J.R. Norman and R.W. King. 1986. More than ever it’s
management for the short term. Business Week (November 24):92-93
Drucker, P.F. 1986. To end the raiding roulette game. Across the Board: 30-39.
Eames, M.J. 1997. Institutional investor myopia, ownership, earnings and returns. Working Paper,
Santa Clara University.
16
Freeman, R.N. and S. Tse. 1989. The multiperiod information content of accounting earnings:
Confirmations and contradictions of previous earnings reports. Journal of Accounting Research
(Supplement): 49-79.
Gompers, A. and A. Metrick. How are large institutions different from other investors? Why do
these differences matter? Working paper, Harvard University and National Bureau of Economic
Research.
Graves, S.B., 1988. Institutional ownership and corporate R&D in the computer industry. Academy
of Management Journal: 417-428.
Hand, J. 1990. A test of the extended functional fixation hypothesis. The Accounting Review: 740763
Jacobs, M.T. 1991. Short term America: The Causes and Cures of Our Business Myopia, Boston,
MA: Harvard Business School Press.
Jacobson, R., and D., Aaker. 1993. Myopic management behavior with efficient, but imperfect
financial markets: A comparison of information asymmetries in the U.S. and Japan. Journal of
Accounting & Economics: 383-405.
Kim, J., I. Krinsky, and J. Lee. 1997. Institutional holdings and trading volume reactions to
quarterly earnings announcements. Journal of Accounting, Auditing and Finance: 1-14.
Kormendi, R and R. Lipe. 1987. Earnings innovations, earnings persistence and stock returns.
Journal of Business: 323-345
Kothari, S.P and R.G. Sloan. 1992. Information in prices about future earnings: Implications for
earnings response coefficients. Journal of Accounting and Economics: 143-171.
Lang, M., and M. McNichols. 1997. Institutional trading and corporate performance. Working
Paper, Stanford University.
Lev, B. 1988. Toward a theory of equitable and efficient accounting policy. The Accounting Review:
1-22.
O’Brien, P.C., and R.Bhushan. 1990. Analyst Following and Institutional Ownership. Journal of
Accounting Research (Supplement): 55-82.
Porter, M.E. 1992. Capital disadvantage: America's failing capital investment system. Harvard
Business Review: 65-82.
Potter, G. 1992. Accounting earnings announcements, institutional investor concentration, and
common stock returns. Journal of Accounting Research: 146-155.
17
Rendleman, R.J., C.P. Jones, and H.A. Latane. 1987. Further insight into the standardized
unexpected earnings anomaly: Size and correlation effects. Financial Review 22: 131-144.
Shiller, R.J. and J. Pound. 1989. Survey evidence on the diffusion of interest and information among
investors. Journal of Economic Behavior and Organizations: 44-66.
Schipper, K. 1989. Commentary on earnings management. Accounting Horizons: 91-102.
Teitelman, R. 1993. Wall Street and the new economic correctness. Institutional Investor
(February).
Wahal, S. and J. McConnell. 1997. Do institutional investors exacerbate managerial myopia?
Working paper, Purdue University and Emory University, Goizueta Business School.
Walther, B. 1997. Investor sophistication and market earnings expectations. Journal of Accounting
Research: 157-192.
Warfield, T., J.J.Wild and K.Wild. 1995. Managerial ownership, accounting choices, and
informativeness of earnings. Journal of Accounting and Economics: 61-91.
18
Figure 1
Timeline underlying the research design related to hypothesis 1
Et
t-2
t-1
t
ω1(τ=1)
Rt,t-1
ω1(τ=2)
Rt,t-2
Rt,t-1
= stock return over the period t-1 to t
Et
= earnings for period ended at time t
ω1(τ=1) = proportion of earnings information at time t (Et)
impounded in stock prices during t-1 to t.
ω1(τ=2) = proportion of earnings information at time t (Et)
impounded in stock prices during t-2 to t.
Ratio =
ω1(τ = 2)
= proportion of information in Et impounded
ω1(τ =1)
in stock prices during period (t-2,t-1) relative to (t-1,t).
19
Table 1
Sample selection criteria
Firm observations in the Disclosure database (1989-95)
Less: Firms with no financial data or price data available in the 1995
Compustat database
Less: Extreme values for earnings and returns
Less: Analyst following data not available from IBES database
Final sample
20
38,221
20,207
18,014
1,056
16,958
6,458
10,500
Table 2
Descriptive Statistics
Variable
INST
MGR
SIZE
LEV
EVAR
MB
NANAL
E
R
Mean
Std.dev.
0.41
0.17
5.70
0.21
0.06
2.43
8.63
0.05
0.22
0.22
0.20
1.86
0.17
0.07
2.75
8.69
0.10
0.54
Median
0.40
0.09
5.58
0.19
0.04
1.76
5.00
0.07
0.12
(N=10500)
First
Third
Quartile
Quartile
0.23
0.02
4.32
0.06
0.02
1.23
2.00
0.03
-0.11
0.58
0.24
6.99
0.33
0.06
2.78
12.50
0.10
0.41
INST is the percent of equity shares held by institutional investors, MGR represents percentage of shares held
by inside owners, i.e., individuals (officers, directors, and principal owners) who can exercise significant
influence over corporate affairs, SIZE is natural logarithm of market value of equity, LEV indicates the ratio of
total debt scaled by lagged total assets, EVAR is the standard deviation of annual earnings divided by lagged
total assets, NANAL is the number of analysts following the firm, MB is the market to book ratio, E represents
income before extraordinary items scaled by lagged market value of equity and R is the annual stock return
measure over the fiscal year. The sample is comprised of firm-year observations drawn from the 1989-1995
fiscal years.
21
Table 3
Seemingly unrelated regression (SUR) results of the relation between earnings and stock
returns conditional on the percentage of institutional ownership
Rit −τ,t = ω 0 + ω 1(τ ) E it + ω 2 (τ ) E it * INSTit + ω 3(τ ) E it * MGRit + ω 4 (τ ) E it * NANALit + ω 5(τ ) E it * MBit + ω 6(τ ) E it * LEVit
+ ω 7 (τ ) E it * EVARit + ω 8(τ ) E it −1 + ω 9 (τ ) E it * SIZE it + yeardummies + ε it (τ )
Pred.
Sign
Eit
+
Eit*INSTit
+
Eit*MGRit
+
Eit*NANALit
+
Eit*MBit
+
Eit*LEVit
-
Eit*EVARit
-
Eit-1
-
Eit*SIZEit
+
N
SUR Adj. R2
Dependent Variable
Rt-1,t
1.12
(9.36)
0.81
(2.78)
10500
Rt-2,t
F-statistic
[p-value]
2.75
(25.33)
3.81
(15.35)
101.21
[0.00]
63.41
[0.00]
10500
Dependent Variable
Rt-1,t
1.35
(4.61)
0.83
(2.37)
0.84
(2.67)
-0.07
(-4.99)
0.01
(0.58)
-1.17
(-3.67)
-2.71
(-3.80)
-0.86
(-14.61)
0.22
(3.60)
10500
23.12%
Rt-2,t
1.31
(4.88)
3.69
(10.88)
3.04
(10.88)
-0.18
(-15.82)
0.23
(13.72)
-4.40
(-15.09)
-9.25
(-13.53)
-0.22
(-3.13)
0.65
(11.79)
F-statistic
[p-value]
0.01
[0.91]
38.99
[0.00]
10500
27.60%
We do not report the coefficients on the intercept and the year-dummies for expositional convenience. t-statistics are in
parenthesis. F-statistics relate to comparing coefficients under the two dependent variables. Rt-1,t (Rt-2,t) is the stock
return measured over the period t-1 (t-2) to t. See Table 2 for the definitions of other variables. The sample is
comprised of firm-year observations drawn from the 1989-1995 fiscal years.
22
Table 4
Seemingly unrelated regression (SUR) results of the relation between earnings and stock
returns conditional on the quintile membership of the percentage of institutional ownership
Rit −τ,t = ω 0 + ω 1(τ ),1 E it * Quintile1 + ω 1(τ ), 2 E it * Quintile2 + ω 1(τ ),3 E it * Quintile3 + ω 1(τ ), 4 E it * Quintile4
+ ω 1(τ ),5 E it * Quintile5 + ω 3(τ ) E it * MGRit + ω 4 (τ ) E it * NANALit + ω 5(τ ) E it * MBit + ω 6 (τ ) E it * LEVit
+ ω 7 (τ ) E it * EVARit + ω 8(τ ) E it −1 + ω 9(τ ) E it * SIZE it + yeardummies + ε it (τ )
Pred.
Sign
Eit*Quintile1
Eit*Quintile2
Eit*Quintile3
Eit*Quintile4
Eit*Quintile5
+
+
+
+
+
Eit*MGRit
+
Eit*NANALit
+
Eit*MBit
+
Eit*LEVit
-
Eit*EVARit
-
Eit-1
-
Eit*SIZEit
+
N
Dependent variable
Rt-1,t
Rt-2,t
1.58
(5.31)
1.36
(4.40)
1.56
(4.62)
1.98
(5.61)
2.00
(5.47)
0.84
(2.68)
-0.07
(-5.21)
0.01
(0.62)
-1.24
(-3.87)
-2.82
(-3.95)
-0.86
(-14.59)
0.23
(3.74)
2.17
(7.88)
1.71
(6.11)
2.14
(7.13)
2.94
(9.33)
4.04
(12.49)
2.97
(10.64)
-0.19
(-16.52)
0.23
(13.56)
-4.57
(-15.61)
-9.84
(-14.32)
-0.24
(-3.34)
0.72
(13.05)
10500
SUR Adj. R2
Ratio
(ω1(τ=2) / ω1(τ=1))
1.37
1.26
1.37
1.48
2.02
10500
27.66%
Quintile1 through Quintile5 assume a value of one to indicate membership to a particular quintile of institutional
ownership and zero otherwise. Rt-1,t (Rt-2,t) is the stock return measured over the period t-1 (t-2) to t. We do not report
the coefficients on the intercept and year dummies for expositional convenience. t-statistics are in parenthesis. Rt-1,t
(Rt-2,t) is the stock return measured over the period t-1 (t-2) to t. See Table 2 for the definitions of other variables. The
sample is comprised of firm-year observations drawn from the 1989-1995 fiscal years.
23
Table 5
OLS regression results of the relation between earnings, order backlog and stock returns
conditional on the percentage of institutional ownership
Rit −1,t = γ 0 + γ 1 E it + γ 2 ∆BACKLOGit + γ 3 ∆BACKLOGit * INSTit + γ 4 E it * INSTit + γ 5 E it * MGRit
+ γ 6 E it * NANALit + γ 7 E it * MBit + γ 8 E it * LEVit + γ 9 E it * EVARit + γ 10 E it −1 + γ 11 E it * SIZE it + yeardummies + υ it
Pred. Sign
Eit
+
∆BACKLOGit
+
∆BACKLOGit*INSTit
+
Eit*INSTit
+
Eit*MGRit
+
Eit*NANALit
+
Eit*MBit
+
Eit*LEVit
-
Eit*EVARit
-
Eit-1
-
Eit*SIZEit
+
Coefficient Estimates
0.83
0.81
(7.69)
(2.13)
0.08
0.02
(5.37)
(0.72)
0.20
(2.15)
1.28
1.19
(2.40)
(2.21)
1.08
1.10
(2.80)
(2.85)
-0.13
-0.13
(-5.95)
(-5.97)
0.32
0.32
(9.86)
(9.82)
-2.62
-2.61
(-5.89)
(-5.85)
-3.74
-3.69
(-2.98)
(-2.94)
-0.69
-0.69
(-10.83)
(-10.82)
0.38
0.39
(4.32)
(4.42)
N
Adj. R2
3426
3426
25.22%
25.30%
We do not report the coefficients on the intercept and the year dummies for expositional convenience. t-statistics are in
parenthesis. Rt-1,t is the stock return measured over the period t-1 to t. ∆BACKLOG is the change in order backlog
scaled by lagged market value of equity. Rt-1,t (Rt-2,t) is the stock return measured over the period t-1 (t-2) to t. See
Table 2 for the definitions of other variables. The sample is comprised of firm-year observations drawn from the 19891995 fiscal years.
24