Inside the ‘Accrual Anomaly’
Tzachi Zach*
Olin School of Business
Washington University in St. Louis
St. Louis, MO 63130
Tel: (314)-9354528
zach@olin.wustl.edu
First version: September 2001
Current version (1.51): June 2003
Abstract. Sloan (1996) and a number of subsequent studies present evidence that a trading strategy based
on publicly available accounting accruals earns abnormal returns of approximately 10% in the year
following its initiation. This empirical regularity has been named the ‘accrual anomaly’.
In this paper I investigate the accrual anomaly along two dimensions. First, I evaluate whether the accrual
anomaly is related to other anomalies documented in the finance literature. Second, I investigate whether
different methods for calculating long-term abnormal returns have an effect on the returns to the accrual
strategy.
My results indicate that both mergers and divestitures have an effect on the returns generated by the
accrual strategy. After excluding observations associated with either mergers or divestitures, there is a
decrease of about 25% in the strategy’s returns. Second, different calculation methods for benchmark
portfolio returns do not have a material effect on the returns of the accrual strategy. Third, when book-tomarket is added to size as a second control for normal returns, returns to the accrual strategy decrease by
approximately 20%. Fourth, the accrual strategy’s returns are much larger in a sample of Nasdaq firms.
Overall, I conclude that the accrual anomaly is sensitive to the series of tests conducted in this study,
although a substantial portion of it remains unexplained.
*
I would like to thank the members of my dissertation committee: Bill Schwert, Jerry Warner, Ross Watts (chair)
and Jerry Zimmerman for their valuable comments and suggestions. I also benefited greatly from the insights of
Sudipta Basu, Assaf Eisdorfer, Philip Joos, S.P. Kothari, Andy Leone, Jay Shanken, Michela Verardo, Charles
Wasley and doctoral workshop participants at the University of Rochester and workshop participants at University
of Chicago, Columbia University, Emory University, MIT, Northwestern University, UC Berkeley, University of
Washington and Washington University. Finally, I thank the Deloitte and Touche Foundation for generous financial
support. All errors are my own.
1.
Introduction
Capital market anomalies (or puzzles) have attracted the attention of researchers and market
participants for at least four decades. According to Schwert (2001), “anomalies are empirical results that
seem to be inconsistent with maintained theories of asset pricing”. Investors are interested in anomalies
because they potentially provide them with an opportunity to enhance portfolio returns. Academic
researchers are interested in anomalies because they present a challenge to existing theories. As a result of
these challenges new theories of asset pricing and market behavior as well as new research methodologies
are developed.
Several anomalies exist in the accounting literature. The post earnings announcement drift (Ball
and Brown, 1968; Bernard and Thomas, 1989, 1990) survived enough tests and different sample periods
to lead Fama (1998) to state that it is “above suspicion”. More recently, Sloan (1996) introduced what
was later termed as the ‘accrual anomaly’. According to Sloan (1996), investors make systematic errors in
assessing the implications of current earnings on future earnings. As a result, an investment strategy that
exploits this hypothesized processing bias earns abnormal returns of around 10% in the year following the
publication of financial statements. Unlike the post earnings announcement drift, in which investors seem
to under-react to public information and return continuation arises, in the accrual anomaly investors overreact to information in earnings and a return reversal arises. Since Sloan (1996), several studies
investigated the accrual anomaly further. I review their results in section 2 of this study.
The interpretation of the evidence in the accrual anomaly literature has been controversial. On the
one hand, some researchers accept the view that the accrual strategy’s profitability is a manifestation of
systematic pricing errors resulting from overweighting of earnings information and underweighting of
cash flow information. Others are more hesitant to adopt this interpretation and call for a series of
additional tests. For example, Kothari (2001) highlights several potential problems in calculating the
returns to the strategy. Among the issues he mentions are better measures of long-term abnormal returns,
survival biases and inference problems arising from cross-sectional dependence.
In this paper I address two broad research questions. First, I explore characteristics that cause or
are correlated with extreme accruals. In particular, I show that extreme accrual firms are more likely to
have experienced certain corporate events.1 Since these events are known to be associated with abnormal
returns, I examine to what degree the accrual anomaly overlaps with the “corporate events” anomalies.2
1
For the purpose of this paper I define the term “corporate events” narrowly to include only mergers and
acquisitions, initial public offerings, seasoned equity offerings, restructurings and divestitures.
2
The interpretation of the evidence regarding the long-term performance following corporate events is controversial
and many are hesitant to attribute the findings to the market’s inability to adequately price securities following
corporate events. Nevertheless, the empirical regularity that firms underperform or overperform several benchmarks
is still relevant to the accrual anomaly literature, which for the most part uses those same benchmarks.
1
Second, I apply new methods for calculating and drawing inference from long-term abnormal returns. For
example, Lyon et al. (1999) promote using a different calculation method for benchmark portfolio returns,
especially in samples of small firms. Since the accrual anomaly is driven by portfolios of extreme
accruals that are dominated by small firms, I investigate whether this new method has an effect on the
accrual anomaly. In addition, I examine whether the standard errors of mean abnormal returns to the
accrual strategies are affected by two related factors. The first one is the extreme skewness of long-term
abnormal returns and the second is the degree of cross-sectional as well as time-series dependence within
portfolios of extreme accruals. To address these issues I employ the calendar-time method for calculating
abnormal returns (Jaffe, 1974; Mandelker, 1974; Fama, 1998).
The results I report indicate that corporate events, in particular mergers and divestitures, are
related to the returns generated by the accrual strategy. After excluding observations associated with
mergers and acquisitions, the strategy’s hedge returns slightly decrease. The strategy’s returns generated
by the top accrual decile, where most mergers and acquisitions concentrate, decrease by about 50%.
When divestitures are excluded from the sample, returns to the accrual strategy decrease by about 20%.
When I exclude both mergers and divestitures, the returns to the accrual strategy decrease from 8.3% to
6.2%, a decline of about 25%.
Second, I show that the calculation methods promoted by Lyon et al. (1999), which hold constant
the composition of benchmark portfolios, do not have a material effect on the returns of the accrual
strategy. If anything, they cause the strategy’s returns to slightly increase. Third, I analyze the sensitivity
of the strategy’s returns to the inclusion of different benchmarks. The results indicate that when book-tomarket is added to size as a second control for normal returns, returns to the accrual strategy decrease by
approximately 20%. Combining the exclusion of corporate events and different calculation methods leads
to a decrease of about 25% in the strategy’s returns. Fourth, I show that employing the calendar-time
approach does not change the strength of the inference regarding the statistical significance of the returns
to the strategy.
I also report an auxiliary result. I show that the accrual strategy behaves differently in the two
subsamples of non-NYSE/AMEX and NYSE/AMEX firms. In particular, the hedge returns of nonNYSE/AMEX firms are higher and their composition is different. In the non-NYSE/AMEX sample, most
of the returns are attributed to the short position of firms in the top accrual deciles. This is consistent with
smaller Nasdaq firms being much more expensive to short sell.
The paper proceeds as follows. Section 2 summarizes our current knowledge about the accrual
anomaly. Section 3 describes the data and sample selection procedure. In section 4, I replicate the accrual
anomaly on my sample and in section 5 I investigate the relation between corporate events and the accrual
anomaly. The sixth section of the paper provides some additional tests that directly deal with the
2
calculation of long term abnormal returns. These include more robust calculation methods of portfolio
returns and accounting for skewness and cross-sectional dependence in calculating the standard errors of
mean abnormal returns. Although these methodological issues have been raised in the finance literature
for a number of years, they have not received attention in the accounting literature by researchers
investigating the accrual anomaly. In section 7, I combine the effects of both corporate events and return
calculation methods. Finally, section 8 concludes.
2.
Accrual anomaly: literature review
Sloan (1996) shows that portfolios of firms with high (low) accruals earn negative (positive)
abnormal returns in the year following portfolio formation.3 A hedge portfolio of annually buying firms in
the bottom decile of total accruals and selling firms in the top decile of total accruals earns a positive
return of over 10% in the year following portfolio formation.4 In addition, he shows that a large portion
of the subsequent abnormal returns is realized around future quarterly earnings announcements. Sloan
(1996) views the findings as being consistent with the market’s inability to correctly impound the
information in accruals into stock prices. More specifically, the market assigns a higher persistence to the
accrual component of earnings and as a result over (under)-prices firms whose earnings contain high
(low) accrual components.
Several studies followed Sloan (1996) to further investigate the puzzling results. These studies
can be divided into four groups: (i) studies investigating accrual components, (ii) studies exploring the
behavior of third parties such as analysts, auditors and insiders, (iii) studies searching for cross-sectional
variation in the accrual strategy’s returns and (iv) studies assessing the extent to which the accrual
anomaly is related to other anomalies. My study best fits into the latter category.
In the first line of inquiry, researchers decompose accruals into several components (e.g. changes
in inventories, changes in accounts receivables, changes in accounts payable) and observe which part is
more highly associated with the strategy’s abnormal returns. Hribar (2000) and Thomas and Zhang (2002)
follow this approach.5 Both studies show that Sloan’s results are related to accruals generated by extreme
increases or decreases in the inventory accounts. Thomas and Zhang (2002) go further and find that
changes in raw materials inventories have even higher abnormal returns associated with them. Hribar
(2000) finds no evidence of mispricing in more transparent accruals, like special items. One interpretation
3
High accruals are income increasing and low accruals are income decreasing, as is evident in the summary
statistics reported in table 1.
4
To make sure that all accounting information needed to execute the strategy is available, Sloan (1996) and other
studies that follow calculate the strategy’s returns over the 12-month period (or a longer one if appropriate) that
starts four months after the end of fiscal year.
5
A more recent study with similar results is Chan et al. (2001).
3
both studies attach to their findings is that the abnormal returns are a consequence of market’s inability to
see through earnings management.
Hribar (2000) examines the changes in working capital accounts in more detail. For example, he
decomposes the inventory and receivable accounts into discretionary and non-discretionary components,
based in part on their relation with growth in sales. He finds evidence that the discretionary components
are more associated with mispricing. Continuing on the earnings management explanation, Xie (2001)
constructs portfolios based on abnormal and normal accruals calculated using the Jones-model. He shows
that abnormal returns are earned only in portfolios of abnormal accruals, i.e. the portion of accruals that is
attributed to managers’ accounting discretion.
Richardson et al. (2001) decompose accruals to broader categories. In addition, they examine the
returns to strategies based on long-term accruals. They find that their decompositions do not yield returns
higher than the basic accrual strategy. Further, they report that long-term accruals are also associated with
future abnormal returns.
In the second category, several studies investigate the interaction between the behavior of
insiders, analysts and auditors with accrual information. Beneish and Vargus (2002) show that the
direction of insider trading by top executives is related to accrual information. More specifically, highaccrual firms that also experience insider selling experience lower future returns than high-accrual firms
that experience insider buying. Thus, it appears that insiders of high accrual firms can distinguish between
high accruals that signify an improvement and those that signify a deterioration in future firm
performance. Continuing on Xie’s (2001) point, Beneish and Vargus (2002) find some evidence that in
firms experiencing insider selling, high accruals are related to earnings management.
Two studies analyze the behavior of analysts. Bradshaw et al. (2001) employ an approach that
does not use market prices as an indicator of market’s inability to process earnings information. Instead,
they examine the behavior of two information intermediaries: sell-side analysts and independent auditors.
Abstracting away from stock returns alleviates concerns about inadequate controls for risk that naturally
arise in these types of studies. The authors find that analysts’ forecast errors are negatively correlated with
total accruals. That is, analysts’ forecast errors are more negative for high accrual firms. This means that
analysts do not incorporate into their forecasts the earnings’ reversal that is associated with high accruals.
Overall, they view their results as further evidence that strengthen inference of the market’s inability to
correctly interpret information in accruals.
Barth and Hutton (2001) also examine analysts’ role in interpreting accruals. First, consistent
with Bradshaw et al. (2001), they find that the majority of analysts do not react in response to new
information about accruals. Only 25% of analysts revise forecasts in a direction consistent with reversal
of newly announced accruals. The authors label them active analysts. Second, they report that investors
4
do not incorporate the information in the revisions of active analysts. That is, investors seem to ignore
analysts. As a result, a hedge strategy that combines information in accruals with information about active
analysts earns an astounding 27% in the year following its implementation.
In addition to their examination of analysts, Bradshaw et al. (2001) investigate the behavior of
auditors. They find that although firms with high accruals exhibit higher incidence of SEC enforcement
actions, their auditors are not more likely to issue qualified opinions. This means that auditors, too, do not
seem to take into account the implication of high accruals.
In the third category, Ali et al. (2000) explore some cross-sectional characteristics of the accrual
anomaly. They conjecture that if the accrual anomaly is a manifestation of investors’ inability to correctly
process information in financial statements, then proxies for the degree of investor sophistication should
be related to returns generated by the accrual strategy. Specifically, returns to the accrual strategy are
expected to be higher in firms that are associated with lower investor sophistication. They use variables
such as size, institutional ownership and analyst coverage to proxy for investor sophistication. Their
results are opposite to those expected under the naïve investor hypothesis. They find lower returns for
smaller firms, firms that are covered by fewer analysts, and firms that are not largely held by institutions.
They interpret their findings as contradicting the interpretations promoted by Sloan (1996) and the studies
that followed. That is, they put less weight on the story that misprocessing of information gives rise to the
returns generated by the accruals strategy.
Finally, the fourth category, to which my study belongs, contains three additional papers. First,
Collins and Hribar (2000a) find that the accrual anomaly and the post-earnings announcement drift are
distinct from each other. They present a strategy that exploits the information in both accruals and
earnings surprises. They report that the combined strategy generates higher returns than each individual
strategy. Second, Fairfield et al. (2001) argue that the accrual anomaly is a subset of a broader anomaly
related to investors’ inability to correctly impound information about growth. Third, Desai et al. (2002)
find that the accrual anomaly and the “value-glamour” anomaly proxied by the cash-flow-to-price ratio
overlap. In this study, I show that the accrual anomaly is mostly separate from anomalies related to price
behavior following corporate events.
The accrual anomaly is interesting to both accounting and finance researchers and practitioners
for at least three reasons. First, the results challenge the efficient market paradigm, suggesting that
abnormal returns can be earned by implementing a fairly straightforward strategy based on publicly
available information. Second, the anomaly highlights an important incentive for managers to engage in
earnings management. It shows that earnings manipulation can have real economic effects and that
managers can potentially influence stock prices by choosing alternative accounting methods.
Traditionally, it has been thought that any such manipulations are transparent to investors and therefore
5
provide no benefit to managers. Third, the higher association of earnings and stock returns, compared to
the association of cash flows with stock returns, that has been investigated by accounting researchers for
almost four decades perhaps needs a second look (e.g. Dechow, 1994). The interpretation generally given
to this high association is that earnings better reflect changes in firm value. An alternative explanation
that arises in light of the accrual anomaly is that the higher association of earnings with stock returns
merely reflects investor fixation on earnings.
Several unanswered questions remain. First, researchers still worry about the calculation of
abnormal returns and whether the current procedures adequately control for risk. The long periods in
which abnormal returns are realized (up to three years after portfolio formation) raise concerns about the
interpretation of the evidence as solely against market efficiency. Kothari (2001) lists several important
issues: (i) problems of measuring expected returns in long-term return studies; (ii) mismeasurement of
expected returns arising from the characteristics of firms having extreme accruals; (iii) cross-sectional
dependence and adequate measurement of standard errors and (iv) data problems arising from survival
(see Kothari et al. 2002). In this paper I answer Kothari’s call that “The focus of future research should be
to address (these problems)” and to reevaluate the findings of current research. To address some of these
issues, I incorporate several recent methodological advancements in calculating long-term abnormal
returns. I also explore a plausible alternative explanation that relates to the long-term under or overperformance of firms following corporate events.
In the current literature, there are several results that alleviate, to some extent, the above
concerns. For example, Sloan’s evidence on the concentration of abnormal returns around earnings
announcements mitigates arguments regarding measurements of long-horizon abnormal returns and
adequate risk controls. In addition, since Bradshaw et al.’s (2001) results about the behavior of analysts
and auditors are “return-free” they are immune to some of the above problems.
The second issue is the importance of viewing the findings of both the post-earnings
announcement drift and the accrual anomaly collectively. The question is how to resolve the apparent
conflicting results that arise from them. Taking the results of the accrual anomaly and the post-earnings
announcement drift as a whole, a simple story of market participants fixating on the accrual component of
earnings is hard to reconcile.
Finally, even if we narrowly focus on the accrual anomaly results and adopt the investor fixation
hypothesis, the reasons for investors’ fixation on earnings remain unclear. Is fixation a result of a
cognitive bias like those proposed recently by various researchers to explain the book-to-market and
momentum effects? (e.g., Hong and Stein, 1999; Barberis et al., 1998; Daniel et al., 1998). Is it a result of
a fixation that is caused by other frictions (e.g., earnings management, press coverage, analysts’
incentives)?
6
3.
Data
3.1
Sample selection and variable measurement
I use two sample periods in my analysis. To make my results as comparable as possible to
previous studies I use for my main analysis NYSE/AMEX firms. However, I discuss interesting insights
for non-NYSE/AMEX firms as well. For assessing the relation between corporate events and the accrual
anomaly I use up to 15,961 NYSE/AMEX observations from the 1988-1999 period. I choose to start my
sample in 1988 because I use information from the statement of cash flows, which is only available
starting in 1988, to identify several corporate events. To study the impact of various calculation methods
of long-term abnormal returns I use 42,635 NYSE/AMEX firms from the full sample period of 19701999.
3.2
Variable measurement.
Accounting variables. Accounting variables are drawn from the COMPUSTAT primary,
secondary and tertiary files as of 1999. Following many studies in the accounting literature I measure
accruals in two ways. First, I use differences between successive balance sheet accounts according to the
following formula, (COMPUSTAT item numbers in parentheses):
ACC1: ∆current assets(#4)-∆cash(#1)-∆current liabilities(#5)+ ∆debt in current liabilities(#34)depreciation(#14))
Second, following Bradshaw et al. (2001), I calculate accruals directly from the statement of cash flows,
according to the following formula:
ACC2: Earnings before extraordinary items from CFS (#123) – Cash flows from operations
(#308).
The measurement of other accounting variables is detailed at the bottom of tables 1 and 2. All accounting
variables are deflated by average total assets (at time t and time t-1).
Stock returns. I obtain stock return data from the Center for Research in Security Prices (CRSP)
tapes as of 2001. The calculation of abnormal returns involves a few choices: (1) the type of benchmark
portfolio used to adjust firms’ raw returns (e.g. size, book-to-market), (2) the weighting scheme in
calculating benchmark portfolio returns (equal- vs. value-weighting), (3) the composition of firms in the
benchmark portfolio (uniform vs. changing) and (4) the type of abnormal returns to use for the analysis:
buy-and-hold (BHAR) or cumulative (CAR). In this paper I choose an equal-weighted, size-adjusted buy-
7
and-hold abnormal return as my reference normal return benchmark. I use NYSE cutoffs to assign firms
to 10 non-equal-sized portfolios at the end of fiscal year. Abnormal-returns calculation is done after
compounding monthly returns of both firms’ raw returns and benchmark portfolio returns. Benchmark
portfolios are implicitly rebalanced every month and their composition may change after the ranking
period. Sections 6.1 and 6.2 describe in more detail the calculation approach. In section 6 I vary the
methods for calculating abnormal returns and examine how they affect the returns on the accrual strategy.
Accrual strategy. In order to apply the accrual strategy, at the end of each year firms are ranked
into deciles based on the magnitude of their deflated accruals. Once the rankings are determined,
abnormal returns for each accrual decile are calculated by averaging the abnormal returns of all firms in a
particular accrual decile. Like other studies in the literature, I start the calculation period four full months
after the end of fiscal year. This assures that all the information in the financial statements is available to
implement the strategy.
3.3
Descriptive statistics
Table 1 reports medians and means of several variables of interest by deciles of accruals
calculated from the balance sheet (scaled by total assets). In general, the descriptive statistics are similar
to those reported in previous studies. First, the negative correlation between accruals and cash flows is
immediately apparent. Cash flows from operations decrease monotonically as we move from the lowaccrual decile (median of 0.22) to the high-accrual decile (median of 0.02).
Second, it is evident that firms with extreme accruals, those that occupy the top and bottom
accrual deciles, are smaller in terms of market capitalization. Similar pattern is also present for total assets
or total revenues. All these variables exhibit an inverted U-shaped pattern with respect to the accrual
deciles. Moreover, the median size of low accrual firms is smaller than that of high-accrual firms (p-value
smaller than 1%). Notice, however, that the mean size of low accrual firms is larger than the mean size of
high accrual firms.
Third, high-accrual firms have lower book-to-market ratios. Starting with decile 8, there is a
decrease in the median book-to-market ratio. It falls from around 0.55 to 0.51 in decile 8, 0.48 in decile 9
and 0.45 in the high accrual decile. The difference in median book-to-market ratios between the low and
high accrual deciles is significant at conventional levels.
Fourth, firm performance measured both in terms of contemporaneous stock performance and in
terms of accounting rates of return and growth, is increasing monotonically with accruals. For example,
the median raw return in the fiscal year for which accruals are measured increases from 3.2% in the lowaccrual decile to 16.6% in the high-accrual decile. The median return on assets in the low accrual decile is
5.5% and it increases to 10.5% in the high accrual decile. The median growth in sales is 1.9% for the low-
8
accrual decile while it is 23.4% in the high-accrual decile. This difference may be slightly misleading
because of the larger preponderance of mergers in the high accrual deciles and of divestitures in the low
accrual deciles. Controlling for this issue, however, median sales growth are still different across the
extreme accrual deciles (1.4% for the low decile and 20.2% for the high decile).
Fifth, compared to firms in the middle accrual deciles, forecasts of long-term growth are slightly
larger for low accrual firms (13.5% vs. 12%) while they are much larger for high accrual firms (18% vs.
12%).
Finally, I also report summary statistics on two common measure of firms’ degree of solvency.
The first is Altman’s Z, due to Altman (1968) that measures a company’s financial strength. Altman’s Z
is negatively correlated with the probability of bankruptcy. The second measure is Ohlson’s O, due to
Ohlson (1980). It is positively correlated with the probability of bankruptcy. Table 1 reveals that the exante probability of bankruptcy is monotonically decreasing as we move from the low to the high accrual
deciles. That is, low accrual firms are more risky in terms of bankruptcy probability than high accrual
firms. The median Z-score of low accrual firms is 2.7 and it reaches 3.9 in the high accrual decile. The O
measure is -0.54 for the low accrual firms and it reaches -1.14 in the high accrual decile.
All the patterns mentioned above (size, book-to-market, performance, bankruptcy risk) are also
evident in unreported summary statistics of non-NYSE/AMEX firms.
4.
Replicating the accrual anomaly
Decile returns analysis. To investigate the accrual anomaly in more depth, I first replicate it and
contrast the results with those of previous studies. Table 2 reports the size-adjusted returns on a number of
portfolios ranked on total accruals as well as on several accrual components. Basically, this table
replicates the results in Sloan (1996) and Thomas and Zhang (2002). There are several differences. First,
in panel B of Table 2 I also report returns for a sample of non-NYSE/AMEX firms. Previous studies that
considered non-NYSE/AMEX firms generally pooled them together with NYSE/AMEX firms and/or
excluded a substantial subset of them based on variables such as size.6 This sample dissection is important
because differences in the returns to the accrual anomaly that are correlated with exchange membership
can help researchers understand the phenomena better. Second, the sample period in table 2 is 1988-1999,
6
Examples include Ali et al. (2000) who use the pooled NYSE/AMEX/NASDAQ sample. Also, although they do
not state it explicitly, the number of observations in Bradshaw et al. (2001) suggests that they pool all observations
as well. Barth and Hutton (2001), Chan et al. (2001) pool NYSE/AMEX/NASDAQ firms but exclude many smaller
firms based on share price and total assets. Houge and Loughran (2001) and Xie (2001) exclude some non-NYSEAMEX firms.
9
which enables the calculation of accruals based on the cash-flow statement. The different sample period
also enables an out-of-sample replication of Sloan’s (1996) original result.7
The first column of Table 2, panel A, shows the basic Sloan (1996) result. A hedge portfolio that
is short on firms in the top accruals decile and long on firms in the low accruals decile generates a return
of 8.3% in year t+1, i.e. the 12-month period starting 4 months after the end of the fiscal year on which
the rankings are based. This return is lower than that reported in Sloan (1996).8 Out of the 12 annual
returns in the sample period, we see that 10 were positive (NPOS=10). We can reject the null hypothesis
that positive and negative returns are equally likely with an associated p-value under 1%. The second
column reports the results of a similar strategy when the ranking is based on accruals that are calculated
from the statement of cash flows. We see that the return to this strategy is 10.2%, higher than the return in
the first column. This result is similar to that of Collins and Hribar (2000b). They attribute the higher
returns to the fact that using cash-flow-based accruals leads to a more accurate ranking, whereas the
rankings that are based on balance sheet accruals are noisy.
The rest of the columns in Table 2, panel A, report returns to strategies that are based on rankings
of several accrual components. The results are generally consistent with those reported in Thomas and
Zhang (2002), although there are a few differences. First, their finding of higher returns to various
inventory strategies are not as strong in my sample. For example, I find that the returns from an
inventory-based strategy (8.1%) are slightly lower than the accrual based strategy (8.3%). In contrast,
Thomas and Zhang (2002) report these returns to be 11% and 8.8%, respectively. The differences are not
attributed to the more recent sample period I use, because similar results (unreported) are obtained for the
longer period of 1970-1997, which is used in Thomas and Zhang (2002). Second, the composition of the
hedge return is different. While most of the hedge returns (about 80%) in Table 2 are attributed to the
long portfolio (i.e. the portfolio of low accruals), the returns in Thomas and Zhang (2002) are mostly
attributed to the short portfolios (about 60%).
Panel B of Table 2 reports the returns on several accrual strategies for a sample of nonNYSE/AMEX firms. No study has separately examined this sample of firms. Ali et al. (2000), Bradshaw
et al. (2001) and Barth and Hutton (2001) examine the pooled sample of all firms. Several results are
noteworthy. First, the returns on the strategies are higher than those in panel A. For example, the return on
the basic accrual strategy is 20.5% compared to 8.3% in panel A. The higher returns continue throughout
panel B. This suggests that the accrual anomaly varies by exchange or variables correlated with exchange.
The only evidence available in the literature on the sensitivity of the accrual anomaly to firm size is in Ali
et al. (2000). Their evidence suggests that returns on an accrual strategy are lower for smaller firms. To
7
8
Other studies that focus on the later sample period are Bradshaw et al. (2001) and Collins and Hribar (2000b).
Most of the difference is explained by the unusually low returns the strategy produced during calendar year 2000.
10
the extent that non-NYSE firms in panel B are smaller the evidence here is at odds with that of Ali et al.
(2000).
Second, the returns for shorting high accrual firms are much higher (in absolute value) than those
reported in panel A. In addition, most of them are statistically significant. For the basic accrual strategy,
shorting high-accrual non-NYSE/AMEX firms earns 12.9% whereas shorting high-accrual NYSE/AMEX
firms earns only 1.3%. If we consider that smaller firms are more expensive (if not impossible) to sell
short, then the differences between the returns to high accrual firms across panels A and B are consistent
with economic intuition. That is, constraints on short selling allow for large abnormal returns to remain
unrealized in small firms.
A third observation from Panel B is that returns to strategies based on balance-sheet-accruals
(20.5%) are higher than returns to strategies based on statement-of-cash-flows-accruals (19%). A similar
pattern is apparent for inventory strategies calculated using the two methods (20.9% vs. 17.7%). These
results are in contrast to those in panel A as well as those reported in Collins and Hribar (2000b).
Regression analysis. I complement the analysis of decile hedge returns with results from
regressions of one-year-ahead size-adjusted returns on deflated accruals or ranks of deflated accruals. One
shortcoming of the regression models is that they force a linear relation between accruals and future
returns. This restriction does not apply for the hedge return results presented earlier. The results are
presented in table 3. They confirm the negative relation between accruals and future returns. All the
coefficients on the accrual variables are negative and statistically significant. The coefficients on the
ranked accrual variable in panel B can be interpreted as the average future return to a hedge portfolio of
buying low accrual firms and short-selling high accrual firms. The implied returns to hedge portfolio are
lower than those reported in the two panels of table 2. The (non-) NYSE/AMEX sample produces a return
of 6.7% (14.6%) compared to 8.3% (20.5%) in table 2. The reason for the difference is that the hedge
strategies in table 2 do not impose a linear relation between accruals and future returns.
5.
The accrual anomaly and corporate events9
This section addresses the possibility that at least part of the accrual anomaly can be traced back
to various anomalies that were previously documented in the finance literature. These include the longrun under-performance following mergers and acquisitions, initial public offerings and seasoned equity
offerings and long-run over-performance following corporate restructurings and divestitures.10 The cause
9
See footnote 2 for my definition of “corporate events”.
A non-exhaustive list includes the following studies. M&A’s: Agrawal, Jaffe and Mandelker (1992); Rau and
Vermaelen (1998); Mitchell and Stafford (2000). IPO’s: Ritter (1991); Brav and Gompers (1997). SEO’s: Loughran
10
11
for concern is that mergers and acquisitions are accompanied by high total accruals, especially accruals
that are calculated based on the balance sheet. The main reason for the correlation between accruals and
mergers is mechanical (e.g. Collins and Hribar, 2000b). The end-of-the-year balance sheet of a merged
firm is not comparable to its beginning-of-the-year balance sheet because the latter only contains the
parent firm while the former contains both the parent and the new subsidiary. As a result of the merger,
there is a mechanical increase in many balance sheet items. Since accruals are calculated as the sum of
differences in successive balance sheet accounts, they will be high if we compare the consolidated end-ofthe-year balance sheet with the unconsolidated beginning-of-the-year balance sheet. However, this is what
researchers implicitly do when comparing successive balance sheet items from Compustat.
To illustrate this, table A1 in the appendix reports selected balance sheet items of Lucent
technologies for the years 1998 and 1999. In fiscal year 1999 Lucent acquired several firms. The first two
columns of table A1 are the 1999 and 1998 balance sheets as reported in 1999. Column 3 reports the
original 1998 balance sheet as was reported in 1998. Comparing columns 2 and 3 highlights the effect of
mergers. These columns are supposed to reflect identical balance sheets. However, the addition of firms
in 1999 requires restating the 1998 comparative numbers in the 1999 balance sheet. For example, moving
from column 3 to column 2, we see that the newly acquired subsidiaries had cash balances of $469
million dollars as of December 1998 (1,154-685). On the other hand, the subsidiaries did not have any
debt maturing in 1999, which is evident in the identical $2,231 figure in columns 2 and 3. Column 4
calculates the changes in balance sheet items based on differences between successive balance sheet
items. This is the method researchers employ when using Compustat data because Compustat records its
numbers as they were reported in the original financial statements. Thus, according to column 4,
inventory increased by $1.8 billion between 1998 and 1999. However, columns 2 and 3 illustrate that
around $200 million of this increase is attributed to the addition of new firms into the 1999 balance sheet.
The real increase in inventory is $1.6 billion. It can be computed by subtracting column 2 from column 1.
Alternatively, with Compustat data, this figure is taken out of the cash flow statement (data 303). In a
similar vein, corporate restructurings and divestitures are suspected to be followed by low accruals, as
firms that were consolidated in one year no longer appear in the balance sheet in the subsequent year.11
High accruals also occur following IPOs and SEOs. The reason in this case is not mechanical.
These firms are likely to increase their investment in working capital following equity offerings. These
increased investments lead to high accruals.
It is hypothesized that portfolios of large accruals contain a higher frequency of firms that have
recently gone through an M&A, IPO or SEO, while portfolios of low accruals contain a higher frequency
and Ritter (1995); Mitchell and Stafford (2000). Restructurings/Divestitures: Cusatis et al. (1993). Fama (1998)
reviews and evaluates much of the evidence.
11
A good example of the effect of divestitures is in the 1998 financial statements of Silicon Graphics (SGI).
12
of restructured firms. Couple these conjectures with the extant results in the finance literature regarding
long-run under-performance of M&As, IPOs and SEOs and long-run over-performance of restructurings
and divestitures, it is possible that the subsequent negative (positive) returns that are experienced by high
(low) accruals firms are, at least partially, related to the anomalies previously documented in finance.
More formally, my hypotheses (in alternative form) are:
H1a: Deciles of high accruals have a larger proportion of initial public offerings, seasoned equity
offerings and mergers and acquisitions than deciles of low accruals. On the other hand, deciles of
high accruals have a smaller frequency of restructurings and divestitures than deciles of low
accruals.
H2a: Since IPO’s, SEO’s and M&A’s (restructurings and divestitures) are associated with negative
(positive) future abnormal returns, and if IPO’s, SEO’s and M&A’s (restructurings and
divestitures) are more concentrated in deciles of high (low) accruals, then the returns to the
accrual strategy, especially the returns on high (low) accruals, are expected to be lower after
excluding firms associated with these corporate events.
5.1
Frequencies of corporate events
Table 4 reports frequencies of several flags that identify corporate events. The definitions of the
different flags of corporate events are given in table A2 of the appendix as well as in the body of table 4.
The results overwhelmingly support hypothesis H1a. They indicate that when accruals are ranked
based on the balance sheet (Panel A), mergers and acquisitions, equity issuances and debt issuances
concentrate in the high deciles of accruals, and in particular the top decile. All the Z-statistics of tests that
evaluate the differences in frequencies between either the top and bottom accrual deciles or the top three
and bottom three deciles are significant at least at the 1% level. There is a fairly monotonic increase in the
frequencies of M&A’s and equity and debt issuances as we move from the lowest to the highest accrual
deciles. For example, when I flag M&A’s based on differences in inventory changes (M&A3), 26.5% of
the observations in decile 10 are M&A’s while only 7.3% are such in decile 1. The fourth flag, M&A4,
that is based on the acquisition figure in the cash flow statement exhibits lower frequencies than the other
four flags. In decile 10, 17.6% of the observations are M&A’s compared to 4.8% in the bottom decile.
The reason for the lower frequencies is that this flag only identifies cash-paid M&A’s.
A similar pattern, with an opposite direction, is apparent in the restructuring flags. As expected,
there is a statistically significant concentration of REST1, REST2 and DIVEST in the bottom deciles of
accruals.
13
Looking at panel B we see the effect of calculating accruals based on the cash flow statement on
the distribution of corporate events across accrual deciles. In general, the frequencies of corporate events
are much more uniform across the deciles compared to panel A, although in most cases there is still a
statistically significant concentration in the top (or bottom in the case of divestitures) accrual deciles. It
also seems that the concentration of REST2 in the bottom accrual deciles has become stronger. The
differences between panel A and panel B illustrate the bias that is embedded in calculating accruals from
the balance sheet and show that calculating accruals from the cash flow statement mitigates the bias,
although it does not eliminate it entirely.
Table 5 provides some information on the differences in the magnitude of accruals as a function
of various corporate events flags. As expected, in most of the flags, we observe a higher level of accruals
when firms are associated with a corporate event. For example, total accruals of firms with M&A1 are
-2.3% of total assets compared to –4.2% of total assets for firms whose M&A1 flag is off. On the other
hand, as expected, the accruals of restructured firms are much lower than those of firms that have not had
a restructuring.
5.2
Accrual anomaly after excluding corporate events
Having established that the top accrual decile contains a higher frequency of M&A’s, IPOs and
SEOs, while the bottom accrual decile contains a disproportionate number of restructurings, I now look at
hypothesis H2a. This hypothesis conjectures that a higher frequency of corporate events in top (bottom)
accrual decile, coupled with the documented under (over) performance of firms following these
transactions, leads to the possibility that some of the accrual anomaly can be traced back to the long-run
performance following corporate events. Therefore, in this section I examine how the returns to the
accrual strategy change as a result of excluding from the sample firms that experienced corporate events.
5.2.1
Frequencies of mergers and acquisitions after exclusion
First, I examine the distribution of the various M&A flags, after excluding a different M&A flag.
This will give some idea about any variation left across accrual deciles after controlling for one flag. This
information is important to assess the power of the tests. I perform the analysis only on mergers and
acquisitions because they have five different flags. The frequencies are reported in table 6 and apply to
the NYSE/AMEX sample. Similar results (unreported) are obtained when accruals are calculated based
on the cash flow statements as well as when the non-NYSE/AMEX sample is examined separately. Panel
A reports the frequencies of M&A flags across accrual deciles. These are the same frequencies reported in
table 4. In each remaining panel I exclude one of the M&A flags. Starting in panel B of table 6, where
M&A1 is excluded, we see that the variation of the other flags across accrual deciles decreases
14
substantially, but is not eliminated. For example, the Z-statistics associated with M&A3 decrease from
19.2 and 66.4 in panel A to 0.68 and 5.9 in panel B. Similar patterns are apparent in the remaining
panels.
To judge which flag has the strongest effect on smoothing the frequencies of M&A’s across
accrual deciles, I examine the average z-statistics in each panel of table 6. The baseline average zstatistics in panel A, before excluding certain firm-years, is 14.7 for z-stat1 and 52.5 for z-stat2.
Excluding M&A5, M&A3 and M&A2 leads to a similar effect on the average z-statistics. The average zstat1 falls to 0.3, 0.5 and 0.9, respectively, while the average z-stat2 falls to 8.9, 10.0 and 11.2. The rest of
the flags are ranked as follows: M&A1 (z-stat1=2.9, z-stat2=13.9); M&A4 (3.2, 21.9).
5.2.2
Returns to the accrual strategy
In this section, I test hypothesis 2 that helps us learn about the sensitivity of returns to the basic
accrual strategy, which is based on total accruals calculated from the balance sheet, to the exclusion of
firm-years associated with various corporate events. The results are reported in table 7. The ‘All’ column
reports the results without any exclusions (the same results as in the left-most column of Table 2) while
the other columns report returns to the accrual strategy when certain firms are excluded from the analysis,
based on the flags described in table A2 of the appendix and in the body of table 7. The last column
excludes firms that are associated with either M&A’s or divestitures.
M&A’s, debt and equity issues. Looking first at the effect of excluding M&A firms, we see that
the returns to the strategy decrease by up to 10%, depending on which flag is chosen. For example, when
M&A3, the inventory-based flag, is excluded, the returns decrease slightly, from 8.3% to 8.1%. Other
flags, like M&A1 and M&A2, exhibit an increase in hedge returns. Returns decrease to 6.8% for debt
issues, while they increase to 8.9% for equity issues.
Since corporate events discussed here concentrate in the top accrual deciles, it is also interesting
to examine the effects of excluding corporate events on each extreme accrual decile separately. Recall
that excluding corporate events associated with long-term underperformance is expected to increase the
returns in each decile, but to decrease the hedge returns because of the disproportionate concentration of
these events in the top accrual decile. Indeed, as expected, the returns on the top accrual decile decrease
regardless of the M&A indicator. For example, when M&A3 is used, returns on the top accrual decile
increase and turn positive, from -1.3% to 0.1%. There is also a discernible effect on the returns in the top
accrual decile when debt or equity issues are excluded. In a similar vein, returns on the bottom accrual
decile increase. For example, in the case of M&A3, returns increase from 7.0% to 8.3%.
Restructurings and divestitures. The effect of excluding firms that underwent divestitures and
restructurings is reported in the columns headed REST1, REST2 and DIVEST. The returns decrease from
15
8.3% to 7.1% and 6.6% when REST2 and DIVEST are considered, while they slightly increase to 8.5%
when REST1 is considered.
Combined effects of M&A3 and DIVEST. The last column of table 7 reports returns on the accrual
strategy when the effects of both M&A’s and divestitures are combined. After excluding firms based on
M&A3 and DIVEST, returns to the accrual strategy decrease from 8.3% to 6.2%, a reduction of about
25%. Note that these two flags are, by construction, independent.
To summarize, this section shows that excluding corporate events has some effect on the returns
of the accrual strategy in the sample of NYSE/AMEX firms. M&A’s and divestitures each account for up
to about 10% of the strategy’s returns. When combining the two, returns to the accrual strategy decrease
by about 25%, from 8.3% to 6.2%. Unreported results indicate that these effects are less evident in a
sample of non-NYSE/AMEX firms possibly because of the less frequent occurrence of mergers and
acquisitions in these firms.
5.2.3
Types of mergers and acquisitions
Several studies in the finance literature show that the post-event stock performance of mergers
varies with certain transaction characteristics. For example, Loughran and Vijh (1997), Rau and
Vermaelen (1998) and Mitchell and Stafford (2000) report that firms using stock as currency in a merger
under-perform while firms using cash over-perform. Although the reason for this empirical regularity is
still under debate, it is interesting to examine the interaction between mergers, form of payment and the
accrual anomaly.
In this section, I distinguish between the different types of mergers to explore whether excluding
any particular type leads to a more apparent reduction in the returns to the accrual anomaly. More
specifically, I separate mergers based on (i) the form of payment used in the merger and (ii) the
accounting method used (pooling or purchase). I obtain information on both the form of payment and the
accounting treatment from the SDC database. Since a particular firm can have multiple mergers in a fiscal
year, I classify firms to each category (cash, stock or pooling) if more than 50% of a firm’s mergers in a
particular year use cash, stock or pooling.
Table 8 reports the results of excluding mergers based on their characteristics. The results
indicate that after excluding cash mergers, the returns to the bottom accrual decile increase to 8.8% and
the total hedge returns increase to 9.6%. These increases are contrary to expectations formed based on
prior studies which show that cash-paid mergers do not under-perform in the period following the merger.
When excluding stock-paid mergers, the returns to the strategy decrease. However, they do not decrease
by more than they do in table 7. Not surprisingly, the pooling results are very similar to the stock results.
16
Overall, it seems that separating mergers more finely, based on the form of payment, does not
lead to a more apparent reduction in the returns to the accrual strategy. One possibility for these
unexpected results is the potential low quality of the SDC identifiers.
6.
Long-term abnormal returns and the accrual anomaly
The calculation of long-term abnormal returns has been the subject of extensive research in the
finance literature for a number of years.12 Lyon et al. (1999) describe the analysis of long-term abnormal
returns as treacherous. The problems raised in the different studies are relevant to the accrual anomaly
literature because the accrual strategy also requires calculation of long-term abnormal returns. In addition,
the success of the strategy is driven by firms in the extreme accrual deciles. These deciles predominantly
consist of small firms, which are a special cause for concern in the calculation of long-term abnormal
returns (e.g. Lyon et al., 1999; Brav and Gompers, 1997).
Despite these considerations, no study investigating the accrual anomaly has directly addressed
the issues and attempted to account for them using other calculation methods and different statistical tests.
In this section I fill this void. I discuss several issues that are relevant for the calculation of abnormal
returns and the statistical tests performed based on them. Section 6.1 describes the advantages and
disadvantages of the two main methods for calculating long-term abnormal returns: (i) the buy-and-hold
method (BHAR) and (ii) the cumulative abnormal returns (CAR) method. Section 6.2 focuses on different
methods for calculating benchmark portfolio returns under the BHAR method. The section draws mainly
from Lyon et al. (1999). Section 6.3 discusses the “bad model” problem. The above three issues are
concerned with an appropriate calculation of abnormal returns. Section 6.4 is concerned with the effects
of skewness and cross-sectional dependence on inference from t-tests. Both of these issues relate to an
appropriate calculation of the standard errors of abnormal returns. In section 6.5 I present the results.
6.1
BHAR vs. CAR
Buy-and-hold abnormal returns. All studies in the accrual anomaly literature calculate buy-
and-hold abnormal returns (BHAR) of firm i during period T in the following way:13
m
m
t =1
t =1
BHARiT = ∏ (1 + Rit ) − ∏ (1 + R pt )
(1),
where Rit is the return of firm i in month t, p is the portfolio (usually size-based) to which firm i is
assigned in the beginning of period T, and m=12, 24 or 36, depending on the length of the period under
12
Studies that directly and indirectly deal with the calculation of and inference from long-term abnormal returns
include: Barber and Lyon (1997), Kothari and Warner (1997), Fama (1998), Lyon et al. (1999), Brav et al. (2000),
Loughran and Ritter (2000) and Mitchell and Stafford (2000).
13
Only Thomas and Zhang (2002) provide an explicit formula. Other studies (e.g. Sloan, 1996; Xie, 2001) provide a
sentence or paragraph that describe the process.
17
study. Typically, the calculation period starts four months after the end of the fiscal year whose accruals
are ranked.
Statistical inference is conducted on two quantities. First, some studies evaluate the statistical
significance of the returns to each accrual decile separately. This examination is possible using two
approaches, pooled and time-series. Second, all studies evaluate the statistical significance of the returns
on the hedge strategy, i.e. returns to the low accrual decile minus returns to the high accrual decile, using
a time-series approach. Table A3 of the appendix summarizes the approaches and provides the formulas
for each statistic. In sections 6.4 and 6.5 I discuss the assumptions underlying each of the above methods
and assess their validity in the context of the accrual anomaly literature.
Cumulative abnormal returns. The literature on long-term abnormal returns also proposes to
calculate cumulative abnormal returns (CAR). Under this method, monthly abnormal returns are summed
over the period under study, according to the following formula:
CARiT = ∑ (Rit − R pt )
m
(2),
t =1
where p is defined in the same manner as in the BHAR. Dividing CAR by the number of months,
m, results in an average abnormal returns (AAR) for the period.
Each of the methods has advantages and disadvantages. The main advantage of the BHAR
method is that it correctly tracks the returns earned by a buy-and-hold investor because it takes into
account the compounding of returns. In contrast, the CAR method does not compound returns and
therefore its returns do not reflect the profits (or losses) of investors. If researchers are specifically
interested in returns to trading strategies, like in the case of the accrual anomaly, then BHAR is preferred.
However, BHAR introduces serious problems. First, as I discuss in section 6.3, the measurement
of normal returns is exposed to the “bad model” problem. Because of compounding, BHAR is likely to
magnify measurement errors, which increase with the calculation horizon. Second, BHAR produces much
more skewed returns and as a result is more likely to be exposed to skewness biases. Third, BHAR is
more exposed to cross-sectional dependence problems, which increase in sample size. In fact, one of the
solutions for cross-sectional dependence I discuss in section 6.4 is the calendar time approach, which by
its nature requires CAR’s. Because of all these disadvantages, Mitchell and Stafford (2000) as well as
Fama (1998) prefer CAR’s over BHAR’s. Lyon et al. (1999) prefer the BHAR method. However, they
recommend using several methods to have a clearer perspective of the phenomenon under study.
18
6.2
Methods of calculating returns of benchmark portfolios
In this section I discuss the procedure of calculating the returns on the benchmark portfolio (the
second term in equation 1). This issue is relevant regardless of the asset-pricing model being used or the
type of benchmarks employed (e.g. size, BM, Size-BM and more). It is also important for both calculation
methods (CAR or BHAR). Barber and Lyon (1997) demonstrate that the returns of benchmark portfolios
are subject to two biases that relate to the portfolio composition at the time of an event (or at the point of
ranking, in the case of portfolio-based strategies): (i) rebalancing bias and (ii) new listing bias. The
rebalancing bias refers to the fact that the returns on the benchmark portfolios are calculated assuming
monthly rebalancing while the returns on the sample portfolio are not rebalanced. Monthly rebalancing
means that firms with previous high (low) returns are implicitly sold (bought). Since it is known that
returns in adjacent months are negatively correlated due to, for example, the bid-ask bounce, rebalancing
causes the benchmark returns to be higher than they otherwise would have been.
The new listing bias refers to the fact that the composition of the benchmark portfolio changes as
a result of new firms being added to it after the event date. Since new firms underperform various
benchmarks (e.g. Ritter, 1991), this generates deflated benchmark returns.
For the BHAR method, which is more widely used in the accrual anomaly literature, Lyon et al.
(1999) outline a procedure that is able to eliminate these biases. According to it, returns of benchmark
portfolios are first compounded, over period t (1,3 or 5 years), for each individual firm that is in the
portfolio at the beginning of the cumulating period(s) and only then are averaged across each portfolio.14
The formula is the following and is labeled RpBH:
R psBHτ
 s +τ

(1 + Rit ) − 1
N s ∏
 ,
= ∑  t =s
Ns
i =1
(3),
where Ns is the number of firms in the benchmark portfolio at the time of the event. The
alternative and more widely used procedure, RpREB is the following:
Nt
R
REB
psτ
s +τ
∑R
t=s
Nt
= ∏ (1 +
14
i =1
it
),
(4),
I treat firms that dropped at one point during period t like Lyon et al. (1999). Specifically, for each month
following the disappearance of a firm, I substitute the equal-weighted monthly return of the size portfolio to which
the firm belonged. For more detail, see the discussion on page 169 in Lyon et al. (1999).
19
where Nt is the number of firms in the benchmark portfolio at any month t, after the event data.
The new listing and rebalancing biases are shown to be more prominent in small firms. Over the sample
period used in Lyon et al. (1999), RPBH results in benchmark returns that are lower than those obtained by
RPREB. This indicates that the rebalancing bias, which is expected to cause benchmark returns to be higher,
is the dominant one.
This is relevant to the accrual anomaly because small firms dominate the extreme accrual deciles.
My goal is to investigate how much of an effect this issue has on the top and bottom accrual deciles.
Since both of these deciles are dominated by small firms, as is evident in the summary statistics presented
in table 1, it would be expected, based on the findings in Lyon et al. (1999), that abnormal returns in both
deciles will be higher as a result of using lower benchmark returns. The overall effect on the hedge
returns to the accrual strategy is left for empirical examination.
6.3
Bad model problems
Every calculation of long-term abnormal returns requires a model for expected returns. All
models have errors inherent in them but these errors get exacerbated as we extend the calculation period,
especially when using the BHAR method. A number of firm characteristics have been identified in the
literature to capture the risks that are priced by the market. These include firm size and book-to-market
ratio. However, these characteristics are still problematic for extremely small firms. Fama and French
(1993) point out, and Mitchell and Stafford (2000) reemphasize, the fact that for 3 out of 25 Size-BM
portfolios, the three-factor model does not explain fully the cross-section of stock returns.
In the context of the accrual anomaly, Kothari (2001) emphasizes that errors in the model of
expected returns are magnified when long-term returns are involved. In addition, Kothari (2001) argues
that the determinants of normal returns are prone to change in samples of firms with extreme
characteristics, like extreme accruals.
Recent advances in asset-pricing studies have identified another factor that explains the crosssection of stock returns. Carhart (1997), following the momentum findings of Jegadeesh and Titman
(1993), promotes the use of an additional factor that tracks returns of recent winners and losers.15
According to the momentum effect, firms that have had high returns in recent months (winners) continue
to earn high returns. In a similar vein, bad performers (losers) continue to earn low returns. Going back to
table 1, we see that recent stock returns (RET12) of firms in the bottom accrual decile are low, while they
are high for firms in the top accrual decile. Therefore, the momentum factor predicts high returns for top
accrual firms and low returns for low accrual firms. The opposite, however, is true. Therefore, ex-ante it
15
The momentum factor has been used in studies that evaluate mutual fund performance like Kothari and Warner
(2001) and Daniel et al. (1997).
20
seems that if the momentum factor is accounted for, the returns to the accrual strategy will increase. It is
still an empirical question, however, in part because of the length of time for which the momentum effect
lasts.
Thus, in this study I examine whether a finer size partitioning as well as an adjustment for other
known firm characteristics, like book-to-market and momentum, are able to partially explain the returns
from the accrual strategy.
6.4
Skewness and cross-sectional dependence
The t-statistics in table A3 of the appendix rely on several assumptions. In this section I discuss
these assumptions in light of both skewness and cross-sectional dependence that are likely to be present in
the context of the accrual anomaly.
The statistical significance of the mean abnormal returns in each accrual decile can be assessed
using both a pooled and a time-series method. Looking at the pooled method first, the distributions of
BHARd and BHARdT are asymptotically normal according to the central limit theorem. (The same is true
of the distribution of BHARdT in equation IIa.) This approximation, however, depends on the rate of
convergence to normality, which in turn is a function of the number of observations underlying the mean,
the degree of non-normality of the underlying variable that is being averaged (in this case one-year
abnormal returns) and the degree of cross-sectional dependence in the sample under study (Brav 2000;
Sutton 1993). Since the degree of non-normality and cross-sectional dependence varies, the use of this
approach is problematic.
The second approach, based on time-series averages and standard deviations, can be used for both
individual deciles and hedge portfolio returns. Two underlying assumptions are worth mentioning. First,
the components of the time series average need to be approximately normal. Otherwise, their mean is not
assured to be normally distributed because of the relatively small number of components underlying it.
This is true for both equation IIb and IIIb. With respect to the hedge strategy, another related assumption
is that the linear combination of the returns to the top and bottom accrual deciles is also normal. However,
it is not clear whether the abnormal returns of each accrual decile achieves a reasonable convergence to
normality. This depends, as mentioned above, on sample size, the degree of non-normality and the degree
of cross-sectional dependence in each accrual decile.
The second assumption underlying the time-series approach is the independence of each
component underlying the time-series mean. This assumption requires that all returns are calculated
during the same calendar period. However, as I demonstrate below, this assumption is not innocuous
when firms whose fiscal-year-ends differ are pooled together into a single portfolio.
21
Most studies in the accrual anomaly literature rank firms based on their accruals regardless of
fiscal-year ends. As a result, the returns of each decile, calculated over one year starting four months after
fiscal-year-end, are not perfectly aligned in calendar time. Moreover, because of Compustat conventions,
it is possible that returns of two firms in the same calendar month are included in different Compustat
fiscal years. Figure 1 illustrates the problem. Consider the 1998 financial statements of 12 firms, each
with a different fiscal-year-end. The returns to the accrual strategy of each firm are calculated over a
different calendar period as depicted by the box-filled area. The returns based on the 1997 financial
statements are depicted with the slash-filled areas. As can be seen, there are calendar months with
different shading types. As a result, there is some dependence between the returns of 1997 and the returns
of 1998, violating the independence assumption needed for correct specification of the time-series mean
statistics (IIb and IIIb).
Calendar-time approach. To overcome the problem of cross-sectional dependence, I use the
calendar-time approach promoted by Fama (1998) and Mitchell and Stafford (2000). This method was
first used in Jaffe (1974) and Mandelker (1974) and has recently gained popularity in the finance
literature. Because this method requires tracking of monthly returns in calendar time by using CAR’s, it
can also be helpful in overcoming the skewness inherent in buy-and-hold abnormal returns. Another
advantage of this method is that it contains more observations than a time-series method based on annual
returns. Details on the approach are available upon request and are also outlined in Jaffe (1974) and in
Mandelker (1974).
6.5
Results
In table 9, I report the returns on the accrual strategy when different benchmarks and calculation
methods of normal returns are employed. Since there are no special data requirements, I use the full
sample period, 1970-1999. The results do not differ when the more recent sample period is used. In the
left part of table 9, I use the traditional approach to calculating benchmark portfolio returns, RpREB.
Consistent with Lyon et al. (1999) I label it the ‘rebalancing’ method. As previously discussed in section
6.2, this method suffers from both the rebalancing and the new listing biases. The right part of table 9 uses
RpBH, “buy-and-hold” benchmark returns that control for the new listing and rebalancing biases.
In addition, I use several benchmarks to calculate normal returns as follows: (i) Size subtracts
from each firm’s annual raw return, the equal-weighted return of one of 10 size portfolios based on NYSE
cutoffs. This benchmark is the one used in tables 2 and 6; (ii) BM subtracts from each firm’s annual raw
return, the equal-weighted return to 10 book-to-market portfolios based on NYSE cutoffs; (iii) BM+SIZE
subtracts from each firm’s annual raw return, the return of one of 70 size-book-to-market portfolios based
22
on 14 size portfolios (where the lowest size decile is partitioned to 5 quintiles) that are further
decomposed into 5 BM quintiles; (iv) Momentum subtracts from each firm’s annual raw return, the equalweighted return to 10 momentum portfolios based on the 12-month return up to one month before the
implementation of the accrual strategy; (v) BM+SIZE+momentum subtracts from each firm’s annual raw
return, the return of one of 125 size/book-to-market/momentum portfolios.
Effect of different portfolio calculation methods. I examine whether differences exist between the
returns to the accrual strategy as a result of switching between portfolio calculation methods, holding the
benchmark for normal returns fixed. When the buy-and-hold method is used, returns to the strategy
decrease or increase, depending on the normal-return benchmark employed. For example, when size is the
benchmark, returns slightly decrease from 9.1% to 8.9%. When BM+Size+momentum is used as a
benchmark, returns increase from 8.3% to 8.5%.
Unreported results on the subsample of non-NYSE/AMEX firms show that the effects of the buyand-hold methodology on abnormal returns are similar. The only exception is when BM+size is the
benchmark. In that case, hedge returns increase substantially from 8.4% to 10.2%.
Effect of different normal-return benchmarks. I now turn to the left part of table 9 and compare
the results obtained from the use of several benchmarks for normal returns. Most studies of the accrual
anomaly use a size-based adjustment to calculate abnormal returns. Some studies add an adjustment for
book-to-market. To see the effect of adding both a book-to-market and a momentum adjustment in each
calculation method, I report the returns from the different benchmarks in separate columns of table 9.
Focusing on the BM adjustments, we see that hedge returns decrease by around 20% from 9.1% to 7.4%
(7.1%) when BM (together with size) is used as a benchmark. When momentum is separately accounted
for, we see that hedge returns, as expected, slightly increase. Overall, when all three factors are
considered, returns decrease to 8.3%. Unreported results of the non-NYSE/AMEX sample reveal a similar
pattern, with BM+size having a larger impact, leading to a decline of approximately 40% in hedge
returns.
Calendar-time approach. Table 10 reports results of the calendar-time approach. In contrast to
the results so far, in this approach I report monthly returns to accrual portfolios that are formed every
month, based on newly available accrual information. Firms are assigned to accrual portfolios based on
cutoff points from the previous year. This approach overcomes some of the difficulties arising from
skewness and cross-sectional dependence. Table 10 indicates that returns to the accrual strategy are still
high and statistically significant. For example, the basic accrual strategy earns an average monthly return
of 0.74% over the 1988-1999 period. The strategy produced positive returns in 88 of the 144 months. The
23
hypothesis that the accrual strategy is equally likely to result in positive and negative returns is strongly
rejected (a p-val of 0.0029). Table 10’s Results are consistent with those in table 2. Overall, I conclude
that problems of cross-sectional dependence and skewness do not seem to influence the inference about
the statistical significance of the accrual strategy.
Summary. The analysis in this section shows that implementing a more robust method for
calculating benchmark portfolios’ returns has little effect on the returns to the accrual strategy. In
addition, a book-to-market adjustment accounts for about 20% of the accrual strategy’s returns (this affect
is larger in the NASDAQ sample). The momentum factor is responsible for a slight increase in the accrual
strategy’s returns. Finally, applying a calendar-time approach, which enables a more accurate calculation
of standard errors, does not diminish the statistical significance of the strategy’s returns.
7.
Combined effects of corporate events and normal return benchmarks
In section 5 I demonstrate the effect of corporate events on the returns to the accrual strategy. All
the results in that section are based on size-adjusted abnormal returns, to make them comparable to most
studies in the literature. In section 6, I report the effects of different calculation methods and different
normal return benchmarks on the accrual strategy. In this section I combine the two effects. The purpose
of this exercise is to provide an upper bound on the degree to which return measurement and corporate
events influence the returns to the accrual strategy.
Table 11 reports hedge returns (returns to the low accrual decile minus returns to the high accrual
decile) for a sample of NYSE/AMEX firms during 1988-1999. The first row is also reported in table 7.
The left-most column is in the spirit of table 9, although the sample period is different to accommodate
the data requirements of some corporate events identifiers. Note that in this table I employ both
approaches for calculating benchmark returns, the rebalancing method and the buy-and-hold method. As
can be seen from table 11, a combination of adjustment for book-to-market, size and momentum together
with exclusion of mergers and divestitures decreases the hedge returns from 8.3% to 6.6%. Overall, after
both issues are considered, returns to the accrual strategy decrease from 8.3% to a range of between 4.8%
and 7.5%. This constitutes an average decrease of about 25%.
8.
Summary
The accrual anomaly, first documented by Sloan (1996), introduces a profitable trading strategy
that is based on easily obtained accounting information. The interpretation of the evidence has been
controversial. On the one hand, some researchers accept the view that the strategy’s profitability is a
24
manifestation of systematic pricing errors resulting from overweighing of earnings information and
underweighting of cash flow information. Others are more hesitant to adopt this interpretation and call
for a series of additional tests. For example, Kothari (2001) highlights several potential problems in
calculating the returns to the strategy. Among the issues he mentions are better measures of long-term
abnormal returns, survival biases and inference problems arising from cross-sectional dependence.
In this paper I accomplish two goals. First, I establish a link between the accrual anomaly and
several corporate events (e.g. mergers and divestitures). Firms that underwent these events have been
shown in the finance literature to underperform or overperform several benchmarks in the years following
the events. I demonstrate that when firms associated with these events are excluded from the sample,
returns to the accrual strategy decrease by up to 25%. The fact that corporate events and the accrual
anomaly are correlated is silent with respect to the sources of either the accrual anomaly or the
over/under-performance following corporate events. One interpretation, still consistent with the evidence,
is that investors are misled by accruals and cause the drift in prices following corporate events.
Alternatively, an unidentified mutual source for both the accrual anomaly and the drift following
corporate events is also consistent with the evidence.
Second, I expose the accrual anomaly to a series of new tests that have recently arisen in the
finance literature. My results indicate that the accrual anomaly is immune to a more sophisticated method
of constructing benchmark portfolios. In addition, I quantify the effect of controlling for book-to-market
and momentum, in addition to size. I find a reduction of about 20% in the strategy’s returns as a result of
the book-to-market control and a decrease of up to 25% after controlling for both corporate events and
different return calculation procedures. Finally, I find that applying the calendar-time approach, which
addresses cross-sectional dependence and skewness, does not change the statistical significance of the
returns to the accrual strategy.
Overall, I conclude that the accrual anomaly is sensitive to the series of tests conducted in this
study. However, a substantial portion of it remains unexplained.
25
References
Agrawal, A., Jaffe, J., Mandeleker, G., 1992, The post-merger performance of acquiring firms: A reexamination of an anomaly, Journal of Finance 47, 1605-1621.
Ali, A., Hwang, L., Trombley, M., 2000, Accruals and future stock returns: Tests of the naïve investor
hypothesis, Journal of Accounting, Auditing and Finance, 161-181.
Altman, E., 1968, Financial ratios, discriminant analysis and the prediction of corporate bankruptcy,
Journal of Finance, 23, 589-609.
Ball, R., Brown, P., 1968, An empirical evaluation of accounting income numbers, Journal of Accounting
Research, 159-178.
Barber, B., Lyon, J., 1997, Detecting long-run abnormal stock returns: The empirical power and
specification of test statistics, Journal of Financial Economics 43, 341-372.
Barberis, N., Shleifer, A., Vishny, R., 1998, A model of investor sentiment, Journal of Financial
Economics 49, 307-344.
Barth, M., Hutton, A., 2001, Financial analysts and the pricing of accruals, Stanford University, working
paper.
Beneish, M., Vargus, M., 2002, Insider trading, earnings quality and accrual mispricing, The Accounting
review 77, 755-791.
Bernard, V., 1987, Cross-sectional dependence and problems in inference in market-based accounting
research, Journal of Accounting Research 25(1), 1-48.
Bernard, V., Thomas, J., 1989, Post earnings announcement drift: Delayed price response or risk
premium?, Journal of Accounting Research 27 (Suppl), 1-36.
Bernard, V., Thomas, J., 1990, Evidence that stock prices do not fully reflect the implications of current
earnings for future earnings, Journal of Accounting & Economics 13, 305-340.
Bradshaw, M., Richardson, S., Sloan, R., 2001, Do analysts and auditors use information in accruals,
Journal of Accounting Research 39(1), 45-74 .
Brav, A., 2000, Inference in long-horizon event studies: A Bayesian approach with application to initial
public offerings, Journal of Finance 55, 1979-2016.
Brav, A., Geczy, C., Gompers, P., 2000, Is the abnormal return following equity issuances anomalous?,
Journal of Financial Economics 56, 209-249.
Brav, A., Gompers, P., 1997, Myth or reality? The long-run underperformance of initial public offerings:
Evidence from venture and nonventure capital-backed companies, Journal of Finance 52, 17911821.
Carhart, M., 1997, On persistence in mutual fund performance, Journal of Finance 52, 57-82.
26
Chan, K., Chan, L., Jegadeesh, N., Lakonishok, J., 2001, Earnings quality and stock returns: The evidence
from accruals, University of Illinois at Urbana-Champaign, working paper.
Collins, D., Hribar, P., 2000a, Earnings-based and accrual-based market anomalies: one effect or two?
Journal of Accounting & Economics 29, 101-123.
Collins, D., Hribar, P., 2000b, Errors in estimating accruals: Implications for empirical research,
University of Iowa, working paper.
Cusatis, P., Miles, J., Woolridge, J., 1993, Restructuring through spinoffs, Journal of Financial
Economics 33, 293-311.
Daniel, K., Grinblatt, M., Titman, S., Wermers, R., 1997, Measuring mutual fund performance with
characteristic-based benchmarks, Journal of Finance 52, 1035-1058.
Daniel, K., Hirshleifer, D., Subrahmanyam, A., 1998, Investor psychology and security market under- and
over-reactions, Journal of Finance 53, 1839-1885.
Dechow, P., 1994, Accounting earnings and cash flows as measures of firm performance: The role o
accounting accruals, Journal of Accounting & Economics 18, 3-42.
Desai, H., Rajgopal, S., Venkatachalam, M., 2002, Value-glamour and accrual mispricing, Duke
University, working paper.
Elliott, J., Shaw, W., 1988, Write-offs as accounting procedures to manage perceptions, Journal of
Accounting Research 26 (Suppl), 91-119.
Fairfield, P., Whisenant, S., Yohn, T., 2001, Accrued earnings and growth: Implications for earnings
persistence and market mispricing, Georgetown University, working paper.
Fama, E., French, K., 1993, Common risk factors in the returns on stocks and bonds, Journal of Financial
Economics 33, 3-56.
Hong, H., Stein, J., 1999, A unified theory of underreaction, momentum trading and overreaction In asset
markets, Journal of Finance 54, 2143-2184.
Houge, T., Loughran, T., 2000, Cash flow is king: Cognitive errors by investors, Journal of Psychology
and Financial Markets 1, 161-175.
Hribar, P., 2000, The market pricing of component of accruals, Cornell University, working paper.
Jaffe, J, 1974, Special information and insider trading, Journal of Business 47, 411-428.
Jegadeesh, N., Titman S., 1993, Returns to buying winners and selling losers: Implications for stock
market efficiency, Journal of Finance 48(1), 65-91.
Kothari, S.P., 2001, Capital markets research in accounting, Journal of Accounting & Economics 31, 105231.
Kothari, S.P., Sabino, J., Zach, T., 2002, Implications of survival and data restrictions on tests of market
efficiency, Massachusetts Institute of Technology, working paper.
27
Kothari, S.P., Warner, J., 1997, Measuring long-horizon security price performance, Journal of Financial
Economics 43, 301-339.
Kothari, S.P., Warner, J., 2001, Evaluating mutual fund performance, Journal of Finance 56(5), 19852010.
Lee, C., 2001, Market efficiency and accounting research: A discussion of 'Capital Market Research in
Accounting' by S.P. Kothari, Journal of Accounting & Economics 31, 233-253.
Loughran, T., Ritter, J., 1995, The new issues puzzle, Journal of Finance 50, 23-51.
Loughran, T., Ritter, J., 2000, Uniformly least powerful tests of market efficiency, Journal of Financial
Economics 55, 361-389.
Lyon, J., Barber, B., Tsai, C., 1999, Improved methods for tests of long-run abnormal stock returns,
Journal of Finance 54, 165-201.
Mandelker, G., 1974, Risk and return: The case of merging firms, Journal of Financial Economics 1, 303335.
Mitchell, M., Stafford, E., 2000, Managerial decisions and long-term stock price performance, Journal of
Business 73, 287-329.
Ohlson, J., 1980, Financial ratios and the probabilistic prediction of bankruptcy, Journal of Accounting
Research, 18, 109-131.
Rau, R., Vermaelen T., 1998, Glamour, value and the post-acquisition performance of acquiring firms,
Journal of Financial Economics 49, 223-254.
Richardson, S., Sloan, R., Soliman, M., Tuna, I., Information in accruals about the quality of earnings,
University of Michigan, working paper.
Ritter, J., 1991, The long-run performance of initial pubic offerings, Journal of Finance, 46, 3-27.
Schwert, W., 2001, Anomalies and market efficiency, in Handbook of the Economics of Finance, chapter
17.
Sloan, R., 1996, Do stock prices fully reflect information in accruals and cash flows about future
earnings? The Accounting Review 71, 289-315.
Sutton, C., 1993, Computer-intensive methods for tests about the mean of an asymmetrical distribution,
Journal of the American Statistical Association 88/423, 802-810.
Thomas, J., Zhang, H., 2002, Inventory changes and future returns, Review of Accounting Studies, 7,
163-187.
Xie, H., 2001, The mispricing of abnormal accruals, Accounting Review, 76(3) 357-373.
28
Appendix
Table A1
Lucent Technologies
(selected financial statement data, Numbers in millions of US dollars)
The first two columns of the table are the 1999 and 1998 balance sheets as reported in 1999. That is, column (2) is the 1998
comparative numbers. Column (3) reports the original 1998 balance sheet as was reported in 1998. Column (4) is the
difference between column (1) and column (3). Column (5) is the difference between column (1) and column (2). It reflects
changes in balance sheet items controlling for the effect of the mergers.
Balance sheets
1999
1998
BS of 1999 BS of 1998 BS of 1998
(1)
(2)
(3)
ASSETS
Cash and cash equivalents
Receivables, Net
total inventories
Deferred income taxes – net
Other current assets
Total current assets
Property, plant and equipment – net
Prepaid pension costs
Deferred income taxes – net
Capitalized software development
costs
Other assets
Total assets
LIABILITIES
Accounts payable
Payroll and benefit-related liabilities
Postretirement and postemployment
benefit liabilities
Debt maturing within one year
Other current liabilities
Total current liabilities
Postretirement and postemployment
benefit liabilities
Long-term debt
Other liabilities
Total liabilities
SHAREOWNERS’ EQUITY
Common stock
Additional paid-in capital
Guaranteed ESOP obligation
Retained earnings
Accumulated other comprehensive
income (loss)
Total shareowners’ equity
Total liabilities and shareowners’
equity
Changes in selected
balance sheet items
BS
(4)
CF
(5)
$1,816
10,438
6,151
1,583
1,943
21,931
$1,154
7,405
4,538
1,775
912
15,784
$685
6,939
4,340
1,623
491
14,078
$1,131
$3,499
$1,811
($40)
$1,452
$7,853
$662
$3,033
$1,613
($192)
$1,031
$6,147
6,847
6,175
–
5,693
3,754
761
5,403
3,754
750
$1,444
$2,421
($750)
$1,154
$2,421
($761)
470
3,352
$38,775
298
3,073
$29,363
298
2,437
$26,720
$172
$915
$12,055
$172
$279
$9,412
$2,878
2,300
$2,157
2,592
$2,040
2,511
$838
($211)
$721
($292)
137
2,864
3,599
11,778
187
2,231
3,718
10,885
187
2,231
3,459
10,428
($50)
$633
$140
$1,350
($50)
$633
($119)
$893
6,305
4,162
2,946
$25,191
6,380
2,409
1,980
$21,654
6,380
2,409
1,969
$21,186
($75)
$1,753
$977
$4,005
($75)
$1,753
$966
$3,537
31
7,731
-33
6,099
30
6,589
-49
1,422
13
4,485
-49
-279
$18
$3,246
$16
$6,378
$1
$1,142
$16
$4,677
-244
$13,584
-283
$7,709
1,364
$5,534
($1,608)
$8,050
$39
$5,875
$38,775
$29,363
$26,720
$12,055
$9,412
29
Table A2:
Definitions of corporate events flags
Definition
Corporate event
Flag
Mergers and
Acquisitions
M&A1
Equals one if Compustat footnote code 1 indicates the
existence of a merger transaction. Footnote item 1 is
defined as an increase of at least 50% in sales as a result
of a merger
M&A2
Equals one if the firm is identified as an acquirer in a
completed merger in the SDC database.
M&A3
Equals one if changes in inventory calculated from the
balance sheet are at least 10% larger than changes in
inventory calculated from the cash flow statement
M&A4
Equals one if the acquisitions item in the cash flow
statement (Item 129) is at least 10% of lagged total assets.
M&A5
Equals one if the change in goodwill is positive.
Equity issuance
EQUITY
Equals one if cash flows from equity financing (Item 108)
are greater than 10% of lagged total assets
Debt issuance
DEBT
Equals one if cash flows from debt financing (Item 111)
are greater than 10% of lagged total assets. The flags that
are based on the statement of cash flows are only
available after 1988.
Restructurings and REST1
divestitures
Equals one if restructuring charges (Item 66) exceed
$10,000 (e.g. Collins and Hribar, 2000b).
REST2
Equals one if negative special items (item 17) exceed 1%
of lagged total assets (e.g. Elliott and Shaw, 1988).
DIVEST
Equals one if changes in inventory calculated from the
balance sheet are at least 10% smaller than changes in
inventory calculated from the cash flow statement
30
Table A3
Summary of t-statistics used in the accrual anomaly literature
Pooled
Mean
Each
decile
separately
∀d = 1,2,...,10
Nd
BHARd =
∑ BHARiT
t-statistic
t − statdPooled =
i∈d , i =1
Nd
Time series
Mean
BHARd
std d N d
(
)
N dT
BHARdT =
∑ BHARiT
i∈d , i =1
N dT
(I b)
(I a)
t-statistic
T
BHARh =
(III a)
Nd = The number of firm-years in accrual decile d during the sample period (across all years).
Stdd = standard deviation of abnormal returns in decile d across all years
NdT = The number of firms in accrual decile in year T.
StdT = Standard deviation of the time-series of BHARdT.
Stdh = Standard deviation of the time-series of hedge returns, BHARhT.
31
t − stat
TS
d
=
∑ BHAR
(std
dT
T =1
T
T
(II b)
(II a)
Hedge returns
T
∑ BHAR
hT
T =1
T
t − stat hTS =
(III b)
BHARh
(std
h
T
)
T
)
Table 1
The table reports medians and (means) of various variables partitioned by deciles of total accruals calculated
using the balance sheet approach. The sample consists of all NYSE/AMEX firms with available data during the
1988-1999 period.
NI
CFO
ACC
BM
Size
Low
2
3
4
5
6
7
8
9
High
0.02
(-0.00)
0.22
(0.23)
-0.15
-0.18
0.04
(0.03)
0.18
(0.17)
-0.10
-0.10
0.05
(0.04)
0.16
(0.17)
-0.07
(-0.07)
0.05
(0.04)
0.15
(0.15)
-0.06
(-0.06)
0.05
(0.05)
0.14
(0.15)
-0.04
(-0.04)
0.05
(0.05)
0.13
(0.13)
-0.03
(-0.03)
0.05
(0.06)
0.12
(0.12)
-0.02
(-0.02)
0.05
(0.05)
0.10
(0.11)
0.00
(-0.00)
0.06
(0.07)
0.09
(0.09)
0.02
(0.03)
0.07
(0.08)
0.02
(0.01)
0.10
(0.13)
0.55
0.54
0.53
0.55
0.55
0.55
0.55
0.51
0.48
(0.70)
(0.67)
(0.66)
(0.67)
(0.62)
(0.63)
(0.63)
(0.62)
(0.59)
153.02
388.41
494.77
621.20
726.15
607.40
513.39
354.73
304.01
(1,683.8) (2,877.4) (3,707.5) (3,101.9) (3,862.0) (3,259.1) (3,281.1) (2,267.1) (1,881.5)
0.45
(0.54)
166.80
(968.1)
3.2%
(11.4%)
5.5%
ROA
(4.0%)
1.9%
Sales grwth
(52.0%)
7.5%
(15.7%)
8.2%
(7.2%)
4.9%
(11.8%)
7.7%
(13.5%)
8.7%
(8.9%)
5.4%
(20.5%)
8.3%
(13.1%)
8.9%
(8.9%)
5.3%
(10.2%)
10.8%
(16.0%)
9.2%
(9.6%)
6.5%
(12.3%)
10.5%
(16.8%)
9.0%
(9.1%)
7.0%
(13.0%)
11.5%
(16.6%)
9.3%
(9.9%)
8.0%
(15.4%)
10.4%
(17.9%)
9.6%
(9.8%)
10.3%
(21.1%)
9.3%
(18.5%)
10.5%
(10.5%)
13.4%
(20.6%)
16.6%
(35.4%)
10.5%
(12.3%)
23.4%
(90.5%)
LTG
13.50
(14.61)
12.00
(13.78)
12.00
(13.32)
12.00
(12.73)
12.00
(12.51)
12.00
(12.45)
12.50
(13.19)
14.00
(14.88)
15.00
(16.34)
18.00
(19.30)
Z
2.72
(3.60)
-0.54
(-0.40)
2.81
(3.80)
-0.92
(-0.84)
2.90
(3.72)
-1.03
(-0.99)
2.83
(3.54)
-1.04
(-0.98)
2.84
(3.66)
-1.06
(-1.02)
2.81
(3.55)
-1.03
(-1.05)
3.04
(4.13)
-1.08
(-1.12)
3.32
(5.08)
-1.16
(-1.18)
3.90
(6.05)
-1.32
(-2.02)
3.89
(6.26)
-1.14
(-1.13)
RET12
O
Variable definitions (NI, CFO and ACC are scaled by total assets):
NI: Income before extraordinary items (Compustat # 18) scaled by total assets.
CFO: Operating income after depreciation (#178) – ACC.
ACC (∆current assets(#4)-∆cash(#1)-∆current liabilities(#5)+ ∆debt in current liabilities(#34)-depreciation(#14)).
BM: Book-to-market ratio (#60/Size).
Size: Market capitalization at end of fiscal-year (#25*#199).
RET12: Annual raw stock returns over the firm’s fiscal year.
ROA: return on assets (#178 / #6).
Growth in Sales (#12 / lag(#12) – 1).
LTG: Long-term sales growth forecasts from IBES (in percentage points).
Z: Altman’s (1968) Z-score. 1.2*(#179/ #6)+1.4*(#36/ #6)+3.3*(#18+#16+#15)/ #6+0.6*size/ #181+#12/ #6.
O: Ohlson’s (1980). -1.32-0.407*log(#6)+6.03*(#181/ #6)-1.43*(#179/ #6)+0.076*(#72/ #4)-1.72*TLTA-2.37*(#172/ #6)-1.83*(#217/
#181)+.285*NL-0.521*((#172-lagged#172)/((abs(#172)+abs(lagged#172))));
Where TLTA=1 when #181>#6. NL=1 when #172 < and lagged #172 <0.
32
Table 2: Panel A
The table reports size-adjusted 12-month returns to portfolios of firms ranked based on the level of accruals or components of accruals specified in the head of each
column. Size adjustment is calculated using returns of 10 size portfolios based on NYSE cutoffs. Sample period is 1988-1999. Returns are measured starting four months
after fiscal year end to make sure that financial information to form the strategy is available. This panel reports results of NYSE or AMEX firms. NPOS is the number of
years (out of 12) in which each strategy resulted in positive returns. Pval is the p-value associated with NPOS, testing the hypothesis that it is equally likely to get positive
and negative returns in a given year. *,**,*** signify statistical significance at the 10%,5% and 1% levels, respectively, based on the time-series of annual returns of each
decile or of the hedge strategy (portfolio 1 – portfolio 10).
Variable
Changes in Changes in Changes in
Accruals Accruals Changes in Changes in
Changes in Changes in
Change in Change in
raw
work in
finished
from
from cash Inventory inventory
current
current
accounts
accounts
Deprec.
materials progress
goods
balance
flow
(balance (cash flow
liabilities
assets
receivable payable
inventory inventory inventory
sheet
statement
sheet)
statement)
Decile
16,100
Low
2
3
4
5
6
7
8
9
High
Hedge
NPOS
Pval
15,815
15,961
14,075
16,100
16,100
16,100
16,030
16,044
9,270
9,207
9,111
7.0% **
8.2% ***
5.9% ***
6.1% ***
3.5%
6.1% ***
5.9% *
5.0% ***
4.5% *
5.4%
5.8% **
6.2% **
3.2%
7.5% ***
4.6% ***
4.6% **
2.9%
4.9% *
0.0%
2.8%
2.9%
4.4% *
1.4%
4.1% **
4.7% *
0.7%
1.3%
3.4%
3.1%
2.2%
2.1%
1.7%
3.1%
1.2%
3.5%
3.3%
3.2%
3.1%
4.0%
4.4%
1.0%
0.3%
3.5% *
0.3%
0.3%
6.7% **
4.1%
1.9%
-0.1%
1.1%
2.8%
0.5%
-0.6%
0.0%
2.8%
1.8%
1.2%
3.9%
2.8%
5.8% **
0.5%
2.7%
-0.2%
0.5%
1.3%
0.7%
2.2%
0.6%
1.1%
-0.5%
-0.9%
-0.2%
1.4%
-0.6%
1.1%
0.3%
-0.5%
3.4%
1.1%
2.7%
2.5%
2.1%
1.8%
1.4%
0.3%
0.4%
0.0%
-0.1%
2.9%
1.3%
-0.5%
3.1%
0.7%
0.2%
0.0%
1.9%
0.1%
-1.0%
0.2%
1.3%
0.4%
0.0%
0.3%
-0.7%
0.1%
-0.6%
2.7%
1.5%
1.5%
1.7%
2.5%
-0.6%
1.9%
-1.5%
4.4%
3.3%
1.9%
6.0%
4.0%
7.6% *
-1.3%
8.3%
-2.0%
**
10.2%
-2.2%
***
8.1%
-1.4%
***
7.6%
4.9%
**
-1.4%
**
0.0%
6.1%
*
10
9
9
10
6
9
8
10
5
10
9
7
0.0032
0.0193
0.0193
0.0032
0.3872
0.0193
0.0730
0.0032
0.6128
0.0032
0.0193
0.1938
33
Table 2: Panel B
The table reports size-adjusted 12-month returns to portfolios of firms ranked based on the level of accruals or components of accruals specified in the head of each
column. Size adjustment is calculated using returns of 10 size portfolios based on NYSE cutoffs. Sample period is 1988-1999. Returns are measured starting four months
after fiscal year end to make sure that financial information to form the strategy is available. This panel reports results of non-NYSE and non-AMEX firms. NPOS is the
number of years (out of 12) in which each strategy resulted in positive returns. Pval is the p-value associated with NPOS, testing the hypothesis that it is equally likely to
get positive and negative returns in a given year. *,**,*** signify statistical significance at the 10%,5% and 1% levels, respectively, based on the time-series of annual
returns of each decile or of the hedge strategy (portfolio 1 – portfolio 10).
Variable
Changes in Changes in Changes in
Accruals Accruals Changes in Changes in
Changes in Changes in
Change in Change in
raw
work in
finished
from
from cash Inventory inventory
current
current
accounts
accounts
Deprec.
materials progress
goods
balance
flow
(balance (cash flow
liabilities
assets
receivable payable
inventory inventory inventory
sheet
statement
sheet)
statement)
Decile
23,417
23,134
23,572
22,226
23,417
23,417
23,417
23,332
23,415
18,502
18,083
18,363
Low
7.6%
5.9%
7.9%
**
8.0%
-4.8%
8.6%
4.5%
5.0%
3.6%
11.2%
8.0%
6.5%
5.7%
-0.3%
5.0%
5.5%
6.9%
5.4%
4.4%
1.7%
7.8%
3.4%
2
6.2%
5.3%
5.2%
3
3.7%
4.7%
-0.3%
-1.2%
-2.5%
4.4%
6.5%
-0.1%
0.9%
5.0%
2.3%
4
2.7%
2.2%
3.6%
5.3%
-0.6%
2.4%
-2.1%
-1.9%
3.9%
0.9%
4.1%**
5
-1.2%
0.6%
10.7%
-1.6%
0.8%
-0.5%
0.8%
-2.1%
2.0%
2.2%
1.5%
3.1%
6
0.1%
0.7%
0.0%
0.0%
1.1%
-0.3%
-0.3%
-1.3%
-4.2%*
-5.9%
-1.0%
0.8%
7
-0.4%
-0.8%
-1.6%
-3.1%
-0.8%
-1.7%
-4.4%
0.3%
0.8%
-1.3%
-6.2%
4.7%
8
-0.5%
-2.5%
-3.1%
-0.9%
-0.8%
-4.4%
-3.4%
0.0%
-4.3%
0.5%
1.5%
-0.6%
9
-9.2% ***
-6.3% **
-6.0% **
-7.0% ***
-1.0%
-4.7%*
-5.9%**
-2.5%
-4.5%
-1.7%
3.0%
1.0%
***
3.8%
-8.8%**
-9.4%*** -7.7%***
High
-12.9%
***
-13.1%
***
-13.0%
***
-9.7%
-13.1%*** -6.2%**
Hedge
20.9% *** 17.7% ***
-8.6%** 21.7%*** 10.6%**
13.8%** 12.9%*** 18.9%**
20.5% *** 19.0% ***
NPOS
11
10
12
12
3
11
9
11
10
11
Pval
0.0002
0.0032
0.9270
0.0002
0.0193
0.0002
0.0032
0.0002
Variable definitions:
Accruals from balance sheet: ∆current assets(#4)-∆cash(#1)-∆current liabilities(#5)+ ∆debt in current liabilities(#34)-depreciation(#14))
Accruals from cash flow statement: Earnings before extraordinary items, CFS (#123) – Cash flows from operations (#308).
Inventory from balance sheet: ∆inventories(#3).
Inventory from cash flow statement: -data303
Current liabilities: ∆current liabilities(#5)- ∆debt in currnt liabilities(#34)-∆deferred taxes(#71)
Current assets: ∆current assets(#4)-∆cash(#1)
Depreciation: -data14
Accounts receivable: ∆data2; Accounts payable: ∆data70
Raw material inventories: ∆data76; Work-in-progress inventory: ∆data77; Finished-goods inventory: ∆data78
34
-4.9%
0.0%
-11.0%***
8.0%
6
0.3872
17.5%***
12
-
Table 3
This table reports results of regressing one-year ahead size adjusted returns on deflated
balance-sheet accruals (panel A) and ranks of deflated balance-sheet accruals (panel B).
Regressions are run separately for each year during 1988-1999. The table reports average
coefficients, average adjusted R2’s and average number of observations (N) from the annual
regressions. T-statistics in parentheses are calculated based on the standard deviation of the
time-series of coefficient estimates.
*,**,***
signify statistical significance at the 10%, 5% and 1% levels, respectively.
Sample
Intercept
Accruals
Adj R2
N
Panel A: Regular regression
Non-NYSE/AMEX
-0.03
(-1.18)
-0.31***
(-3.60)
0.33%
2,260
NYSE/AMEX
0.01
(0.59)
-0.21***
(-2.06)
0.37%
1,510
Pooled
-0.01
(-1.15)
-0.30***
(-3.71)
0.33%
3,770
0.30%
2,260
Non-NYSE/AMEX
Panel B: Rank regressions
0.06
-0.1464***
(1.66)
(-4.25)
NYSE/AMEX
0.05***
(3.13)
-0.0667***
(-2.94)
0.33%
1,510
Pooled
0.06***
(2.85)
-0.1143***
(-4.28)
0.27%
3,770
35
Table 4
The table reports frequencies of corporate events stratified by accrual deciles calculated based on the balance sheet (Panel A) or
the cash flow statement (Panel B). The sample consists of all firms with available data in NYSE and AMEX in the 1988-1999
period. M&A’s are identified by several flags as follows: M&A1 includes firms if footnote code 1 in Compustat identifies them
as an acquirer due to an increase of more than 50% in sales. M&A2 includes firms that are identified by SDC as acquirers in a
completed merger in a certain year. M&A3 identifies a merger if changes in inventory calculated from the balance sheet are at
least 10% larger than changes in inventory calculated from the statement of cash flows. M&A4 identifies a merger if Compustat
item 129, acquisitions from the statement of cash flows, exceeds 10% of total assets. M&A5 identifies a merger if the change in
goodwill is positive. Other corporate events are identified as follows: EQUITY identifies an equity offering if Compustat item
108, sale of common stock, is at least 10% of total assets. DEBT identifies a debt issue if Compustat item 111, issuance of long
term debt, is at least 10% of total assets. REST1 identifies a restructure based on whether restructuring charges, Compustat
item 66, exceed $10,000. REST2 identifies a restructuring based on whether negative special items, Compustat item 17, exceed
1% of lagged total assets. DIVEST identifies a divestiture if changes in inventory calculated from the balance sheet are at least
10% smaller than changes in inventory calculated from the statement of cash flows.
Z-stat1 is for a test comparing the frequencies of corporate events in Decile 10 and Decile 1. Z-stat2 is for a test comparing the
frequencies of the combined deciles 8,9 and 10 with the frequencies of the combined deciles 1,2, and 3.
Accrual Deciles
Low
2
3
4
5
6
7
8
9
High
Z-stat 1
Z-stat 2
Panel A: Balance sheet accruals
M&A1
14.2%
11.9%
13.6%
12.5%
14.6%
15.9%
16.8%
19.3%
23.3%
29.3%
13.67
45.77
M&A2
14.2%
16.5%
19.4%
19.4%
20.5%
21.2%
24.0%
24.4%
26.4%
29.0%
13.38
37.70
M&A3
7.3%
8.6%
10.0%
12.1%
12.1%
15.1%
16.7%
19.2%
20.1%
26.5%
19.18
66.37
M&A4
4.8%
5.1%
5.5%
5.1%
6.1%
7.2%
7.8%
9.8%
11.5%
17.6%
15.16
61.66
M&A5
7.9%
7.4%
9.0%
8.9%
10.1%
11.7%
12.6%
14.9%
16.0%
19.0%
12.17
51.08
DEBT
21.8%
24.6%
23.3%
23.5%
21.7%
21.9%
22.4%
25.8%
27.3%
32.4%
8.92
17.82
EQUITY
22.6%
16.8%
17.6%
15.3%
17.7%
21.1%
20.4%
23.2%
30.6%
44.6%
17.39
46.49
REST1
12.2%
8.5%
7.1%
5.6%
4.7%
3.9%
4.6%
4.2%
4.4%
4.8%
(9.96)
(48.64)
REST2
44.4%
32.8%
28.6%
23.5%
23.6%
20.6%
21.8%
20.7%
21.1%
21.3%
(18.35)
(45.42)
DIVEST
23.2%
15.1%
15.3%
13.8% 12.7% 11.1% 10.5%
9.1%
Panel B: Cash flow statement accruals
7.7%
5.4%
(18.94)
(61.14)
M&A1
18.2%
13.5%
14.8%
14.8%
17.1%
17.0%
16.6%
19.3%
20.4%
19.1%
0.86
18.10
M&A2
18.6%
17.4%
19.8%
21.9%
22.7%
22.2%
23.1%
23.7%
24.3%
21.9%
3.05
18.08
M&A3
10.8%
11.6%
13.6%
14.6%
15.7%
16.9%
16.4%
18.0%
17.4%
14.2%
3.75
23.54
M&A4
7.6%
6.5%
7.2%
7.5%
8.1%
8.4%
8.5%
8.7%
10.0%
8.1%
0.65
15.93
M&A5
8.8%
7.5%
9.7%
11.0%
11.4%
12.4%
12.8%
14.8%
15.4%
14.1%
6.23
37.94
DEBT
22.2%
23.8%
25.0%
25.3%
24.7%
23.8%
21.9%
24.0%
26.9%
27.8%
4.85
8.81
EQUITY
28.3%
20.5%
19.2%
16.7%
17.9%
18.6%
18.8%
22.3%
25.3%
43.5%
11.72
25.39
REST1
7.6%
6.0%
5.7%
5.5%
5.5%
5.3%
5.5%
6.3%
5.7%
6.9%
(0.94)
(1.18)
REST2
57.6%
38.5%
30.0%
24.5%
20.8%
19.1%
16.8%
18.8%
17.1%
15.9%
(32.13)
(76.30)
DIVEST
18.2%
12.3%
12.5%
13.1%
12.4%
12.0%
12.4%
11.7%
11.0%
9.9%
(8.85)
(20.09)
36
Table 5
The table reports medians of total accruals for firms that are associated with corporate events (Yes) and firms that are not
associated with corporate events (No). M&A’s are identified by several flags as follows: M&A1 includes firms if footnote
code 1 in Compustat identifies them as an acquirer due to an increase of more than 50% in sales. M&A2 includes firms that
are identified by SDC as acquirers in a completed merger in a certain year. M&A3 identifies a merger if changes in inventory
calculated from the balance sheet are at least 10% larger than changes in inventory calculated from the statement of cash
flows. M&A4 identifies a merger if Compustat item 129, acquisitions from the statement of cash flows, exceeds 10% of total
assets. M&A5 identifies a merger if the change in goodwill is positive. Other corporate events are identified as follows:
EQUITY identifies an equity offering if Compustat item 108, sale of common stock, is at least 10% of total assets. DEBT
identifies a debt issue if Compustat item 111, issuance of long term debt, is at least 10% of total assets. REST1 identifies a
restructure based on whether restructuring charges, Compustat item 66, exceed $10,000. REST2 identifies a restructuring
based on whether negative special items, Compustat item 17, exceed 1% of lagged total assets. DIVEST identifies a
divestiture if changes in inventory calculated from the balance sheet are at least 10% smaller than changes in inventory
calculated from the statement of cash flows.
P-val is for a wilcoxon test for the difference between the medians across the two groups.
M&A1
M&A2
M&A3
M&A4
M&A5 EQUITY
DEBT
REST1
REST2 DIVEST
No
-0.0425
-0.0409
-0.0439
-0.0412
-0.0414
-0.0411
-0.0402
-0.0375
-0.0352
-0.0349
Yes
-0.0233
-0.0348
-0.0230
-0.0070
-0.0237
-0.0322
-0.0172
-0.0528
-0.0516
-0.0556
p-val
< 0.01
< 0.01
< 0.01
< 0.01
< 0.01
< 0.01
< 0.01
< 0.01
< 0.01
< 0.01
37
Table 6
The table reports frequencies of mergers and acquisitions stratified by accrual deciles calculated based on the balance sheet. The
sample consists of all firms with available data in NYSE and AMEX in the 1988-1999 period. M&A’s are identified by several
flags as follows: M&A1 includes firms if footnote code 1 in Compustat identifies them as an acquirer due to an increase of more
than 50% in sales. M&A2 includes firms that are identified by SDC as acquirers in a completed merger in a certain year. M&A3
identifies a merger if changes in inventory calculated from the balance sheet are at least 10% larger than changes in inventory
calculated from the statement of cash flows. M&A4 identifies a merger if Compustat item 129, acquisitions from the statement
of cash flows, exceeds 10% of total assets. M&A5 identifies a merger if the change in goodwill is positive.
Z-stat1 is for a test comparing the frequencies of corporate events in Decile 10 and Decile 1. Z-stat2 is for a test comparing the
frequencies of the combined deciles 8,9 and 10 with the frequencies of the combined deciles 1,2, and 3.
Accrual Deciles (calculated from the balance sheet)
Low
2
3
4
5
6
7
8
9
High
Z-stat 1
Z-stat 2
Panel A: All flags in
M&A1
14.2%
11.9%
13.6%
12.5%
14.6%
15.9%
16.8%
19.3%
23.3%
29.3%
13.67
45.77
M&A2
14.2%
16.5%
19.4%
19.4%
20.5%
21.2%
24.0%
24.4%
26.4%
29.0%
13.38
37.70
M&A3
7.3%
8.6%
10.0%
12.1%
12.1%
15.1%
16.7%
19.2%
20.1%
26.5%
19.18
66.37
M&A4
4.8%
5.1%
5.5%
5.1%
6.1%
7.2%
7.8%
9.8%
11.5%
17.6%
15.16
61.66
M&A5
7.9%
7.4%
9.0%
8.9%
10.1%
11.7%
12.6%
14.9%
16.0%
19.0%
12.17
51.08
Panel B: M&A1 excluded
M&A2
8.9%
10.9%
12.9%
13.6%
14.4%
13.1%
13.8%
13.4%
13.3%
12.2%
3.58
11.42
M&A3
5.4%
6.4%
8.0%
8.4%
8.9%
10.0%
9.3%
8.7%
7.3%
5.8%
0.68
5.90
M&A4
1.1%
1.5%
2.2%
1.7%
1.5%
1.7%
1.2%
1.3%
1.9%
2.2%
3.14
7.28
M&A5
3.5%
3.7%
4.4%
5.7%
5.7%
6.2%
6.5%
6.7%
6.7%
6.1%
4.06
31.11
Panel C: M&A2 excluded
M&A1
8.5%
6.7%
7.5%
5.8%
8.2%
7.4%
6.6%
8.4%
8.9%
9.0%
0.64
9.44
M&A3
6.6%
7.5%
8.8%
8.7%
9.4%
11.0%
9.5%
10.0%
9.0%
7.3%
0.91
8.58
M&A4
3.2%
2.6%
3.4%
2.9%
3.2%
3.0%
2.6%
2.9%
3.5%
2.7%
(1.11)
(1.27)
M&A5
4.5%
3.9%
5.5%
6.2%
6.4%
6.4%
6.2%
7.7%
7.8%
6.8%
3.30
27.96
Panel D: M&A3 excluded
M&A1
13.2%
8.4%
9.3%
8.6%
10.3%
10.1%
9.5%
10.2%
10.7%
11.3%
(2.06)
2.35
M&A2
14.8%
13.5%
15.4%
16.4%
16.8%
16.7%
16.7%
16.3%
16.6%
15.7%
0.88
7.35
M&A4
4.4%
3.6%
3.2%
3.6%
4.1%
4.5%
3.6%
3.6%
4.0%
4.1%
(0.58)
2.13
M&A5
6.3%
5.0%
5.9%
7.6%
7.1%
8.1%
7.6%
8.9%
9.0%
9.1%
3.56
28.22
Panel E: M&A4 excluded
M&A1
12.4%
8.9%
10.3%
9.4%
11.1%
10.9%
10.0%
12.7%
13.3%
14.0%
1.65
16.32
M&A2
14.8%
14.0%
16.6%
18.0%
18.6%
17.6%
18.2%
18.8%
18.8%
17.3%
2.49
14.18
M&A3
7.8%
8.9%
9.9%
11.0%
12.1%
13.3%
12.1%
13.4%
11.8%
10.5%
3.33
20.05
M&A5
6.0%
5.2%
6.9%
7.9%
8.4%
9.0%
8.8%
10.9%
10.9%
10.1%
5.44
36.93
16.32
Panel F: M&A5 excluded
M&A1
13.5%
9.9%
M&A2
14.8%
14.2%
16.2%
M&A3
8.5%
9.2%
10.0%
M&A4
4.8%
4.2%
4.3%
4.2%
9.9%
9.8%
11.7%
11.2%
10.5%
11.6%
12.2%
11.6%
1.65
17.6%
18.3%
16.9%
17.3%
17.3%
17.5%
15.2%
0.44
7.36
11.3%
11.6%
12.8%
11.5%
12.3%
11.1%
9.1%
0.86
10.88
5.0%
4.9%
4.2%
4.5%
5.2%
3.8%
(1.68)
0.91
38
Table 7
The table presents size-adjusted 12-month returns to portfolios of NYSE/AMEX firms ranked based on the level of accruals (calculated from the balance sheet).
Returns are measured starting four months after fiscal year end to make sure that financial information to form the strategy is available. Sample period is 19881999. The column “All” presents returns to all firms that belong to the sample. The rest of the columns exclude certain firms from the sample based on various
criteria, while keeping the accrual rankings from the column ‘All’. M&A’s are identified by several independent flags as follows: M&A1 includes firms if
footnote code 1 in Compustat identifies them as an acquirer due to an increase of more than 50% in sales. M&A2 includes firms that are identified by SDC as
acquirers in a completed merger in a certain year. M&A3 identifies a merger if changes in inventory calculated from the balance sheet are at least 10% larger
than changes in inventory calculated from the statement of cash flows. M&A4 identifies a merger if Compustat item 129, acquisitions from the statement of
cash flows, exceeds 10% of total assets. M&A5 identifies a merger if the change in goodwill is positive. Other corporate events are identified as follows:
EQUITY identifies an equity offering if Compustat item 108, sale of common stock, is at least 10% of total assets. DEBT identifies a debt issue if Compustat
item 111, issuance of long term debt, is at least 10% of total assets. REST1 identifies a restructuring based on whether restructuring charges, Compustat item
66, exceed $10,000. REST2 identifies a restructuring based on whether negative special items, Compustat item 17, exceed 1% of lagged total assets. DIVEST
identifies a divestiture if changes in inventory calculated from the balance sheet are at least 10% smaller than changes in inventory calculated from the
statement of cash flows.
*,**,*** signify statistical significance at the 10%,5% and 1% levels, respectively, based on the time-series of annual returns of each decile or of the hedge
strategy (portfolio 1 – portfolio 10).
All
Low
2
3
4
5
6
7
8
9
High
Hedge
M&A1
M&A2
M&A3
M&A4
M&A5
DEBT
EQUITY
REST1
REST2
DIVEST
DIVEST &
M&A3
7.0% **
8.5% ***
9.1% **
8.3% ***
7.7% ***
7.9% ***
8.1% ***
8.1%***
7.2%**
5.2%*
5.8%**
7.4%**
3.2%
3.9%
4.3%
3.9%
3.6%
3.3%
3.9%
4.0%*
2.9%
1.6%
2.5%
3.3%
4.7%
*
4.9%
*
3.8%
5.0%
*
4.4%
*
5.0%
*
5.0%
**
4.5%
*
*
*
4.5%
4.8%
5.3%
**
5.8%**
3.2%
3.0%
4.5% *
3.6%
3.2%
3.1%
4.2% **
3.5%
3.2%
3.0%
3.3%
4.0%
-0.1%
-0.4%
0.0%
0.0%
-0.1%
0.0%
0.3%
0.4%
-0.2%
-0.9%
0.0%
0.2%
0.5%
0.6%
0.7%
0.3%
0.2%
1.0%
1.5%
0.6%
0.4%
0.6%
0.7%
0.4%
1.4%
2.4%
2.0%
2.5%
1.9%
2.1%
1.9%
1.5%
1.8%
0.5%
1.7%
3.1%*
0.3%
1.4%
2.0%
1.1%
0.6%
1.2%
1.3%
1.0%
0.6%
-0.9%
-0.9%
-0.7%
0.1%
-0.3%
2.3%
1.4%
0.7%
1.0%
-0.1%
-1.1%
0.3%
-0.1%
0.8%
2.9%
-1.3%
-0.7%
-0.8%
0.1%
-0.5%
0.5%
1.3%
-0.8%
-1.3%
-1.9%
-0.9%
1.2%
8.3% **
9.2% ***
9.9% **
8.1% ***
8.2% ***
7.4% **
6.8% *
8.9%***
8.5%**
7.1%*
6.6%**
6.2%**
39
Table 8
The table presents size-adjusted 12-month returns to portfolios of NYSE/AMEX firms ranked based on the
level of accruals. Returns are measured starting four months after fiscal-year end to make sure that financial
information to form the strategy is available. Sample period is 1988-1999. The column “All” presents
returns to all firms that belong to the sample. The rest of the columns exclude from the sample certain firms
that are associated with different types of mergers, while keeping the accrual rankings from the column
‘All’. CASH excludes all firms that had cash-paid mergers during a fiscal year. STOCK excludes all firms
that had a stock-based merger during the year. POOL excludes all firms that chose to treat their merger
using the pooling method. I include firms with multiple mergers in a fiscal year in the CASH, STOCK or
POOL subsamples if over 50% of their mergers were paid in cash, stock or were treated using the pooling
method.
All
Cash
Stock
Pooling
7.0%**
8.8%**
7.2% ***
7.1%***
2
3.2%
3.0%
3.1%
3.0%
3
4.7%*
3.1%
4.4% *
4.6%*
4
3.2%
2.3%*
3.3%
3.3%
5
-0.1%
-1.1%
-0.2%
-0.2%
6
0.5%
-0.1%
0.7%
0.6%
7
1.4%
2.0%
1.9%
1.4%
8
0.3%
1.4%
-0.1%
0.3%
9
0.1%
0.2%
0.6%
0.1%
High
-1.3%
-0.8%
-0.8%
-1.0%
Low
Hedge
8.3%**
9.6%**
40
8.0% ***
8.1%**
Table 9
The table reports 12–month returns to portfolios of NYSE and AMEX firms ranked annually based on the level of their total accruals that are calculated from the
balance sheet. Sample period is 1970-1999. Rebalancing and Buy-and-hold refer to two methods for calculating the long-run returns of benchmark portfolios.
Rebalancing method is exposed to the new listing and the rebalancing biases because: (i) the composition of the reference portfolios changes throughout the
holding period as a result of new firms being added to the universe of firms; (ii) the reference portfolio weights are periodically rebalanced. Buy-and-hold
method constrains the reference portfolio to contain the same firms that were in it at the time when the strategy was first implemented. This method also makes
sure that the portfolio weights are not rebalanced periodically. For details, see Lyon, Barber and Tsai (1999). Several “normal return” adjustments are employed
to compute abnormal returns. On December of each year, firms are assigned to size and book-to-market portfolios. Assignments to momentum-based portfolios
are done one month before the implementation of the strategy. The returns from the following reference portfolios are subtracted from each firm’s annual
returns: Size subtracts the equal-weighted returns of one of 10 size portfolios based on NYSE cutoffs. BM subtracts the equal-weighted returns of one of 10
book-to-market portfolios based on NYSE cutoffs. BM+size subtracts the equal-weighted returns of one of 70 size-book-to-market portfolios based on NYSE
cutoffs (5 BM portfolios * 14 size portfolios, where the lowest size decile is further partitioned into 5 equal quintiles). Momentum subtracts the equal-weighted
returns of one of 10 momentum portfolios based on returns in the previous 12 months. BM+size+Momentum subtracts the equal-weighted returns of one of 125
Size-BM-momentum portfolios (5*5*5). Returns are measured starting four months after fiscal year end to make sure that financial information to form the
strategy is available.
*,**,***
signify statistical significance at the 10%,5% and 1% levels, respectively, based on the time-series of annual returns of each decile or of the hedge
strategy (portfolio 1 – portfolio 10).
Rebalancing method
Decile
Size
BM
BM+Size
Buy and Hold method
Momentum
BM+Size+
Momentum
Size
BM
BM+Size
Momentum
BM+Size+
Momentum
Low
5.7% ***
3.4% ***
3.9% ***
4.3% ***
3.2% ***
5.2% ***
3.6% ***
3.7% ***
4.6% ***
3.4% ***
2
4.7% ***
1.5%
2.0% *
2.6% **
1.6%
4.0% ***
1.6%
1.7%
3.1% **
1.9%
3
4.5%
***
4
2.5% **
5
0.9%
6
2.0%
7
1.7%
8
0.8%
9
*
1.7%
2.5%
-0.4%
0.3%
***
2.1%
1.7%
0.6%
-0.5%
*
*
3.9%
***
**
2.7%
*
2.1% **
1.8%
2.3%
2.1% *
-0.2%
0.4%
1.0%
-0.1%
0.4%
-1.8%
-0.9%
-1.0%
-1.4%
-2.0%
-0.9%
-1.5%
-1.7%
-0.7%
0.4%
-0.4%
-0.4%
1.5%
-0.5%
0.3%
0.1%
-0.2%
-0.7%
0.1%
-0.7%
-1.0%
1.1%
-0.7%
-0.2%
-0.3%
-0.7%
-0.3%
-1.2%
0.4%
-1.0%
-0.6%
-1.2%
-0.4%
-1.9%
High
-3.4% **
Hedge
9.1% ***
*
-1.3%
***
-1.3%
-3.0%
-3.9% ***
-3.2% **
-5.7% ***
7.4% ***
7.1% ***
10.0% ***
-2.9%
***
-0.7%
*
-2.8% ***
-1.0%
-2.0%
-1.7%
-5.0% ***
-3.7% **
-4.0% **
-3.6% **
-4.9% ***
-5.1% ***
8.3% ***
8.9% ***
7.6% ***
7.3% ***
9.5% ***
8.5% ***
41
-2.4%
-1.0%
**
Table 10
The table reports averages of size-adjusted monthly returns to portfolios of NYSE or AMEX firms ranked based on the level of accruals or
components of accruals specified in the head of each column. Size adjustment is calculated using returns of 10 size portfolios based on NYSE
cutoffs. Sample period is 1988-1999. Each month, portfolios are rearranged as a result of firms going in and out of accrual portfolios. The
assignment is done based on accrual cutoffs from the previous year. NPOS is the number of months (out of 144) in which each strategy resulted in
positive returns. Pval is the p-value associated with NPOS, testing the hypothesis that it is equally likely to get positive and negative returns in a
given year. Annual ret is a multiplication of Hedge returns by 12, to give a sense of the annual return associated with the monthly hedge returns.
*,**,***
signify statistical significance at the 10%,5% and 1% levels, respectively, based on the time-series of annual returns of each decile or of the
hedge strategy (portfolio 1 – portfolio 10).
Changes in Changes in
Accruals
Changes in Changes in
Accruals
Inventory
inventory Change in Change in
Changes in
from cash
raw
work in
from
accounts
accounts
finished goods
from
from cash
flow
materials
progress
balance
receivable
payable
inventory
balance
flow
statement
inventory
inventory
sheet
sheet
statement
Low
2
3
4
5
6
7
8
9
High
Hedge
NPOS
Pval
Annual Ret
0.64%
***
0.38%
0.05%
0.41%
***
0.05%
0.05%
0.10%
-0.04%
0.10%
0.00%
0.09%
0.07%
-0.15%
-0.04%
0.09%
-0.12%
0.02%
-0.18%
0.13%
0.05%
0.05%
0.02%
0.01%
-0.09%
-0.13%
-0.06%
0.08%
0.00%
-0.08%
-0.11%
-0.07%
0.02%
0.06%
-0.10%
-0.25%
-0.27%
0.29%
-0.20%
-0.01%
-0.06%
-0.15%
-0.10%
-0.11%
-0.04%
-0.01%
-0.08%
-0.19%
-0.18%
-0.23%
*
-0.14%
-0.07%
-0.07%
-0.23%
*
-0.17%
0.00%
*
-0.18%
-0.32%
**
-0.38%
0.68%
96
0.0000
8.15%
0.52%
***
-0.23%
*
-0.27%
**
-0.22%
*
-0.31%
**
-0.26%
**
-0.37%
0.75%
93
0.0002
9.06%
-0.22%
0.74%
88
0.0029
8.87%
***
-0.34%
0.98%
94
0.0001
11.75%
***
***
***
0.30%
*
**
***
42
0.26%
-0.04%
0.30%
79
0.1056
3.61%
*
**
0.45%
**
*
0.45%
0.46%
**
**
0.33%
*
0.44%
-0.16%
0.12%
-0.13%
-0.20%
*
-0.09%
-0.12%
-0.09%
-0.26%
**
-0.12%
-0.01%
-0.13%
0.00%
0.45%
79
0.1056
5.46%
**
-0.38%
0.83%
91
0.0005
10.01%
*
***
-0.35%
0.69%
96
0.0000
8.22%
*
***
-0.32%
0.76%
87
0.0048
9.14%
*
***
Table 11
The table presents 12-month hedge returns to an accrual strategy resulting from buying firms in the low accrual decile and selling firms in the high accrual
decile. Firms belong to NYSE/AMEX. Returns are measured starting four months after fiscal-year-end to make sure that financial information to form the
strategy is available. Sample period is 1988-1999. The column “All” presents returns to all firms that belong to the sample. The rest of the columns exclude
certain firms from the sample based on various criteria, outlined in table 6, while keeping the same accrual rankings from column ‘All’. Return adjustments are
described in table 8.
Return
DIVEST
Method
All
M&A1 M&A2 M&A3 M&A4 M&A5 DEBT EQUITY REST1 REST2 DIVEST
benchmark
& M&A3
Size
BM
REB
method
BM+Size
Momentum
BM+Size+Mom
Size
BM
BH
method
BM+Size
Momentum
BM+Size+Mom
8.3%
9.2%
9.9%
8.1%
8.2%
7.4%
6.8%
8.9%
8.5%
7.1%
6.6%
6.2%
7.2%
7.9%
7.7%
6.2%
7.0%
5.9%
5.3%
7.3%
7.3%
6.2%
6.2%
4.8%
7.1%
8.5%
8.5%
6.8%
7.3%
6.3%
5.7%
7.5%
7.3%
6.0%
6.3%
5.7%
9.2%
10.1%
10.4%
9.0%
9.4%
8.3%
7.9%
9.5%
9.3%
8.6%
7.9%
7.5%
7.8%
9.1%
8.9%
7.9%
8.3%
7.4%
6.2%
8.1%
7.8%
7.9%
6.8%
6.6%
7.9%
8.5%
8.5%
6.9%
7.6%
6.5%
6.1%
7.9%
8.0%
6.7%
6.7%
5.3%
7.5%
8.9%
8.7%
6.9%
7.6%
6.5%
6.2%
7.9%
7.7%
6.6%
6.5%
5.6%
8.4%
9.3%
9.9%
8.1%
8.3%
7.4%
6.9%
9.0%
8.5%
7.2%
6.7%
6.1%
8.8%
9.9%
10.1%
8.7%
9.1%
8.1%
8.0%
9.1%
8.9%
8.2%
7.3%
7.0%
8.0%
10.0%
9.2%
8.4%
8.9%
8.1%
6.8%
8.3%
7.9%
8.4%
7.0%
7.2%
43
Figure 1
This figure depicts the return cumulation periods of the accrual strategy that are based on financial statements for the 1997 and 1998 fiscal-years.
Each row represents firms with different fiscal-year-ends. Cumulation periods start four months after the end of fiscal-year.
1998
1999
2000
FYR 1 2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6 7 8 9 10 11 12
1
2
3
4
5
6
7
8
9
10
11
12
Compustat fiscal year
1997
1998
44