Quarterly Cash Flows and Accruals round3 final

Specifically, future quarterly earnings are more highly associated with current net operating cash flows than with accruals, because accruals have lower persistence than net operating ca

Quarterly Cash Flows, Accruals and

Future Returns

Joshua LivnatProfessor of AccountingDepartment of AccountingStern School of Business Administration

New York University

55 Water St

NYC, NY 10041(212) 438 3934

massimo_santicchia@sandp.com

First Draft: December 2004Current Draft: October 19, 2022

The authors gratefully acknowledge the preliminary and original Compustat quarterly

data provided by Charter Oak Investment Systems Inc and the SEC filing dates provided

by Compustat The authors also thank Chiara Szczesny for copy editing This research project is an outcome of work done by the S&P Equity Research Group

Quarterly Cash Flows, Accruals and

Future Returns

Abstract

This study shows that the famous accruals anomaly is also present in quarterly earnings and accruals Specifically, future quarterly earnings are more highly associated with current net operating cash flows than with accruals, because accruals have lower

persistence than net operating cash flows Consequently, firms with extremely high (low) current quarterly accruals have significant and negative (positive) abnormal returns through the subsequent four quarters A hedge portfolio that combines short positions in firms with extremely high quarterly accruals and long positions in firms with extremely low quarterly accruals is shown to generate consistently positive and economically significant returns for the 53 quarters in the study The study is different from prior studies in that it uses the originally reported, or un-restated, quarterly data to calculate accruals and SEC filing dates to identify the precise day on which investors first obtain information about accruals

Quarterly Cash Flows, Accruals and

Future Returns

In a rigorous study, Sloan (1996) documents that net operating cash flow is more closely associated with future income and stock returns than accruals; which is the difference between income and net operating cash flows Sloan attributes this

phenomenon to the fact that accruals reverse faster than earnings in subsequent periods and are less persistent than net operating cash flows Investors who focus on total income and not on operating cash flows tend to overestimate the persistence of accruals and underestimate the persistence of net operating cash flows A trading strategy that holds long positions in companies with extremely low accruals and short positions in

companies with high accruals obtains significant abnormal returns over the subsequent three years Sloan also shows that a substantial portion of the abnormal returns to the accrual strategy occurs around subsequent quarterly earnings announcement dates, consistent with an initial market mispricing of accruals that is corrected when future quarterly earnings become known Sloan’s findings have been confirmed by many ensuing studies

To better understand what accruals are and why they are likely to be less

persistent than net operating cash flows, the appendix highlights two specific accrual cases of inventory and accounts receivable In the first case, management is overly optimistic about future sales, building up excessive inventories that are reduced in future periods Because the current accounting system treats inventory increases as investments

in assets which are not expensed on the income statement, earnings are initially higher but lower in subsequent periods when inventories are drawn down The first case does notassume any sinister managerial motives; it is based on the reasonable assumption that some managers may at times be too optimistic about future demand leading to excess inventories The second case is based on channel stuffing1, when a manager inflates current income by booking current sales that are not collected until future periods and that reduce future sales The accounting system again records the increase in receivables

as an investment, and the income statement is based on all sales whether collected during the period or not, leading to higher current income at the expense of lower future income

Of course, managers can also affect current income through mere accounting choices (unlike the channel stuffing described in the second case) such as the selection of

aggressive accounting methods and estimates For example, if bad debt expense in the current period is too low it will likely lead to higher bad debt expense in the future and lower future income

Whether accruals are actively managed to produce a desired level of current income, or are just due to genuine but erroneous managerial expectations about the future, the current accounting system nevertheless includes accruals in current earnings, which may be less persistent into future earnings Careful financial statement users such

as financial analysts, investors and creditors should closely scrutinize the firm’s quarterly accruals for the possibility of future reversals However, if a sufficiently large proportion

of investors are fixated on earnings, ignoring information in net operating cash flows and

1 Channel stuffing is a deceptive business practice used by a company to inflate sales by deliberately sending retailers along its distribution channel more products than they are able to sell to customers during the normal course of business.

accruals, prices may not accurately reflect the economic situation of the firm, leading to apotential mispricing and the accruals anomaly.

The purpose of this study is to investigate whether quarterly accruals exhibit the same pattern as that of annual accruals in prior studies As the literature review in the next section illustrates, most accruals studies to date focused on annual accruals, which are shown to have lower levels of persistence than net operating cash flows However, if extreme quarterly accruals contain valuable information about future earnings and stock return reversals, users of financial statements should focus on quarterly cash flows and accruals to obtain an early warning signal that future earnings may reverse Financial analysts tend to revise earnings forecasts after quarterly earnings are released; therefore, their ability to improve forecasts after a careful consideration of quarterly accruals and cash flows is likely to be extremely important for them Similarly, investment and credit managers are unlikely to wait around a whole year for the next annual earnings

announcement to examine changes in accruals, rebalance their portfolios, or take actions about their outstanding credit positions, if such information is available and is useful on aquarterly basis

Why then did most previous studies examine annual instead of quarterly accruals?There are two primary reasons for that First, accruals typically become known to the market when firms file their 10-Q/10-K forms with the SEC, unless they disclose net operating cash flows in their preliminary earnings announcements, which only about 10%

do Unlike preliminary earnings release dates which are available in the Compustat Quarterly database, SEC filing dates are not part of computerized databases used by academics, hence quarterly accruals have not been studied by academics in earlier studies

due to the massive necessary hand-collection Instead, these studies focused on annual accruals, beginning the return accumulation period four months after fiscal year-end, when firms have already filed their 10-K forms with the SEC The second reason for not studying quarterly accruals is that quarterly accruals and subsequent reversals may be very natural and likely in companies that are subject to seasonality in their operations It

is unclear whether quarterly accruals may exhibit the same pattern as annual accruals if most firms are affected by within-year seasonality in operations

Our study shows that quarterly net operating cash flows are more persistent than quarterly accruals, and that future quarterly earnings are more strongly associated with current quarter’s net operating cash flows than with current accruals Furthermore, there

is a negative and statistically significant association between current accruals and

subsequent abnormal stock returns through all four subsequent quarters We also show that a hedge portfolio that holds long positions in companies with the most extreme negative accruals and short positions in companies with the most extreme positive

accruals yields positive and significant abnormal returns, whether held for one, two, three

or even four quarters These results indicate that analysts, investors and credit managers need to carefully examine net operating cash flows and accruals when interpreting currentquarterly earnings

erroneous levels of current accruals In particular, managers may use positive accruals to elevate earnings when operating cash flows are low and negative accruals that reduce high levels of earnings when operating cash flows are high Indeed, Sloan (1996) and others document a negative association between net operating cash flows and accruals, aswell as higher earnings for companies with higher levels of accruals than those with lower accruals Xie (2001) and DeFond and Park (2001) provide evidence that the accrualanomaly is driven by the mispricing of abnormal accruals, i.e., those that are subject to managerial discretion In contrast, Beneish and Vargus (2002) find that the accrual anomaly can mainly be attributed to mispricing of income-increasing accruals, regardless

of whether total or discretionary accruals are used Thomas and Zhang (2002) attribute the accruals anomaly to investors’ failure to correctly understand the role of inventory changes, which is just one although important component of total accruals Richardson, Sloan, Soliman and Tuna (2004) provide evidence that the accruals anomaly can be attributed to those accounts that have low earnings quality and potentially high

managerial discretion

Several studies attempt to determine whether the accrual anomaly is unique, supplemented, or subsumed by other known anomalies Collins and Hribar (2000) show that the accruals anomaly is separate from the post-earnings-announcement drift, and thatcombining both anomalies yields greater abnormal returns than each of them alone Barthand Hutton (2003) show that the accruals anomaly can be combined with analysts’ earnings revisions to yield greater abnormal returns Desai, Rajgopal and Venkatachalam (2004) provide evidence that the accruals anomaly is another manifestation of the value-glamour (growth) anomaly Finally, Fairfield, Whisenant and Yohn (2003) suggest that

the accrual anomaly is partially driven by mispricing the implications of growth in net operating assets for future profits and returns

As for most anomalies, a question remains why the accruals anomaly persists and are not arbitraged away Collins, Gong and Hribar (2003) provide evidence that firms with a greater proportion of institutional investors, who are supposed to be more

sophisticated, are subject to a weaker accruals anomaly Mashruwala, Rajgopal and Shevlin (2004) show that the annual accruals strategy does not earn positive abnormal returns in all twelve months after portfolio formation, and that it is indeed weaker for companies with lower arbitrage risk Lev and Nissim (2004) show that firms with extreme accruals have attributes that institutional investors shun Finally, Kraft, Leone and Wasley (2004) argue that the accruals anomaly is likely driven by relatively few extreme observations and the failure to properly use delisting returns

Most accrual studies have followed Sloan’s (1996) use of annual cash flows and accruals, which are associated with annual buy and hold returns There are just a few studies that examine whether the accruals anomaly is present in quarterly data Collins

and Hribar (2000) provide such evidence, but they accumulate returns from the assumed SEC filing date through the subsequent two quarters, and not from the actual date on

which investors obtain access to the SEC filings DeFond and Park (2001) examine whether investors seem to interpret quarterly earnings surprises correctly, depending on abnormal working capital accruals for a sample of observations in the years 1992-1995 They show that abnormal returns for good earnings news firms with income-decreasing accruals during the 80-day period after the preliminary earnings announcement exceed those with income-increasing accruals However, they essentially assume that the market

has the accrual information at the time of the preliminary earnings announcement.2 Thus, there is very little research into whether quarterly accruals follow the same patterns as those observed for annual accruals Financial analysts, investors and creditors are likely

to be interested in applying the accruals strategy using quarterly data instead of waiting a whole year for the next annual earnings, if quarterly accruals exhibit similar

characteristics to annual accruals

3 Data and Sample

3.1 The Preliminary and Un-restated Compustat Quarterly Data

Data entry into the Compustat databases has been performed in a fairly structured manner over the years When a firm releases its preliminary earnings announcement, Compustat takes as many line items as possible from the preliminary announcement and enters them into the quarterly database within 2-3 days The preliminary data in the database are denoted by an update code of 2, until the firm files its Form 10-Q (10-K) with the SEC or releases it to the public, at which point Compustat updates all available information and uses an update code of 3 Unlike the Compustat Annual database, which

is maintained as originally reported by the firm (except for restated items), the CompustatQuarterly database is further updated when a firm restates its previously reported

quarterly results For example, if a firm engages in mergers, acquisitions, or divestitures

at a particular quarter and restates previously reported quarterly data to reflect these events, Compustat inserts the restated data into the database instead of the previously

2 In robustness checks, they mention that their results are weaker for a sub-sample of firms with balance sheet information in the preliminary earnings announcement They also report that their results do not change when they use abnormal returns cumulated from the preliminary earnings announcement through the next 20 trading days, which presumably include the SEC filing date They report that their results hold for the first three fiscal quarters but not for the fourth.

reported numbers Similarly, when the annual audit is performed and the firm is required

to restate its previously reported quarterly results by its auditor as part of the disclosure contained in Form 10-K, Compustat updates the quarterly database to reflect these restated data

Charter Oak Investment Systems, Inc (Charter Oak) has collected the weekly original CD-Rom that Compustat sent to its PC clients, which always contained updated data as of that week From these weekly updates, Charter Oak has constructed a database that contains three numbers for each firm for each Compustat line item in each quarter The first number is the preliminary earnings announcement that Compustat inserted into the database when it bore the update code of 2 The second number is the “As First Reported” (AFR) figure when Compustat first changed the update code to 3 for that firm-quarter The third number is the number that exists in the current version of Compustat, which is what most investors use The Charter Oak database allows us to use the first-reported information in the SEC filing, so that our quarterly earnings, cash flows and accruals correspond to those reported originally by the firms, which are also available to market participants at the time of the SEC filing Using the restated Compustat Quarterly database may induce a hindsight bias into our back-tests, since we may have used

restated earnings, cash flows or accruals that were not known to market participants on the SEC filing dates

quarter at the time of the study The only limitation on the initial selection of

firm-quarters is that market value at quarter end is in excess of $50 million, yielding an initial population of 368,378 firm-quarters From this initial population, we delete observations

if the originally reported income before extraordinary items and discontinued operations (Compustat Quarterly item No 8) is missing; or the originally reported quarterly net operating cash flow (Compustat Quarterly item No 108) is missing; if market value at the end of the prior quarter is unavailable; or if total assets (Compustat Quarterly item

No 44) at the end of the prior quarter or the at the end of the current quarter are missing This yields a reduced population of 242,292 firm-quarters

For each firm-quarter in this reduced population, we obtain the SEC filing date of the 10-Q/K forms, which is supplied to us by Compustat for the calendar years 1991-

2004 This reduces the sample to 176,864 observations with SEC filing dates We then use the GVKEY from Compustat to match the observation to the CRSP database We compute Buy and Hold excess Returns (BHR) from two days after the SEC filing date through the next four subsequent preliminary earnings announcements We assume that investors get access to the SEC filings on the day after the filing date, and that after estimating accruals, they take portfolio positions on the following day We use holding periods through one day after subsequent earnings announcements since prior studies indicate that a substantial portion of the accrual strategy returns are generated around subsequent earnings announcements, presumably when investors understand their earlier mispricing

To reduce the survival bias, we use holding periods of 90, 180, 270, or 360 days after the SEC filing date if subsequent quarterly earnings announcements are missing If a

security is delisted from an exchange before the end of the holding period, we use the delisting return from CRSP if available, and -100% if the stock is forced to delist by the exchange or if the delisting is due to financial difficulties After delisting, we assume the proceeds are invested in the benchmark size and B/M portfolio This is the procedure used by Kraft, et al (2004) We first calculate the buy and hold return on the security during the holding period; then subtract the buy and hold return on a similar size and B/Mbenchmark portfolio for the same holding period The benchmark returns are from Professor Kenneth French’s data library, based on classification of the population into six (two size and three B/M) portfolios.3 To make sure that our results are not driven by observations with extreme returns, we delete all observations with buy and hold excess returns in any of the four return periods at the top or bottom 0.5% of the distribution Thisreduces the sample to 155,985 firm-quarters

Consistent with the accruals literature, we estimate accruals as earnings minus netoperating cash flows, and scale by average assets during the quarter To be consistent in our scaling, all current and subsequent quarters’ variables (i.e., earnings, net operating cash flows and accruals in quarters t, …,t+4) are scaled by average total assets in quarter

t.4 To eliminate the undue influence of extreme observations, earnings, net operating cash flows and accruals are winsorized to fall in the range [-1,+1] This procedure keeps all observations, even when some earnings, cash flows, or earnings surprises are extreme

Table 1 provides summary statistics about our sample As can be seen from the table, mean (median) quarterly earnings are about -0.07% (1.03%) of average total assets,with over 75% of the firm-quarters having positive earnings The mean (median)

3 http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html.

4 Scaling by net operating assets in subsequent quarters may introduce a bias due to growth in total assets as

is shown by Fairfield, et al (2003).

quarterly net operating cash flow is even higher at 1.25% (1.78%) of average total assets, but accruals have a negative mean (median) of 1.33% (1.0%) of average assets, mostly due to depreciation and amortization which is included in income but excluded from net operating cash flow Note that in spite of the negative mean and median accruals, over 25% of the observations have positive, income-increasing accruals Finally, the mean (median) buy and hold abnormal return from the SEC filing through the subsequent preliminary earnings announcements are all negative, indicating that the average firm in our sample has been likely to disclose unfavorable news from the SEC filing through the subsequent earnings announcement.

(Insert Table 1 about here)

To focus on firms which are more likely to have extreme accruals that will be less persistent, we further require that all observations in the bottom two deciles of accruals, i.e., the most negative accruals, have both positive income and positive net operating cashflow Such firms have either decided to manage earnings downwards to reduce already high positive earnings and operating cash flows, or have future expectations that are too pessimistic on the average.5 We further require that firms in the top two accrual deciles, i.e., the most positive accruals, also have positive current earnings, consistent with either income-increasing earnings management or with too optimistic forecasts on the average These restrictions reduce the sample further to 137,012 observations for the remaining analyses

5 The assignment to accrual deciles is based on all available observations even if they do not have SEC filing dates or abnormal returns As is well documented by Sloan (1996) and others, firms in the extreme accrual deciles tend to be smaller than other firms

4 Results

Table 2 shows the distribution of excess Buy and Hold Returns (BHR) from two days after the SEC filings through one day after the subsequent earnings announcements for sample observations sorted according to the level of accruals each quarter The

accruals literature provides evidence that firms with income-decreasing, or low, accruals have larger future subsequent abnormal returns than firms with income-increasing, or high, accruals The evidence in Table 2 is consistent with those prior findings The bottomaccruals decile (decile 0) always has higher subsequent returns than the top accruals decile (decile 9) Note that this difference in BHR is 3.4% for the first quarter, rising to 6.1% through quarter t+2, 8.4% through quarter t+3, and 9.9% through quarter t+4 The latter result is comparable to Sloan’s (1996) hedge portfolio return for the subsequent year The table also shows that the long portfolio (extreme negative accruals, decile 0) has BHR that exceeds the expected transaction costs of 2% Thus, the strategy yields positive abnormal returns net of transaction costs, even if only the long position is used

(Insert Table 2 about here)The accruals literature attributes the negative association between accruals and returns to the lower persistence of accruals as compared to net operating cash flows Table 3 examines the relationship between earnings in the subsequent four quarters and current earnings, net operating cash flows and accruals As can be seen in the table, there

is a positive and statistically significant association between current earnings, and

earnings in the subsequent four quarters, where about 78-79 cents of each dollar of current quarter’s earnings is carried over to earnings in the subsequent four quarters (79,

78, 78, and 83 cents for the subsequent four quarters, respectively)6 The table also shows that net operating cash flow is more persistent than accruals, with 81 cents of every dollar of cash flow carried over to the subsequent four quarters (81, 81, 81 and 87 cents, respectively), whereas only about 68-74 cents of a dollar of accruals are carried over to earnings in the subsequent four quarters (74, 70, 68, and 73, respectively) Thus, consistent with findings about annual data, quarterly net operating cash flows and

accruals exhibit the same pattern of a greater persistence of net operating cash flows than accruals into future earnings.7

(Insert Table 3 about here)Table 4 provides direct evidence about the persistence of accruals It shows the associations between accruals in the current quarter t and those in the subsequent four quarters The table shows a negative and statistically significant association between current accruals and accruals at quarter t+1 (-0.025, p<0.0001), which is then completely reversed by quarter t+2 (0.025, p<0.0001), no association with accruals in quarter t+3 (0.001, p=.831), and a very high and significant positive association with accruals in quarter t+4 (0.419, p<0.0001) Thus, Table 4 shows that quarterly accruals tend to reverse

in the immediately subsequent quarter, reverse again (becoming a positive association) in quarter t+2, and exhibit a high level of seasonality at quarter t+4 This evidence may explain why accruals are less persistent than net operating cash flows in predicting future quarterly earnings The lower level of persistence in quarterly accruals explains why a strategy of long positions in firms with low accruals and short positions in firms with high accruals can yield significant positive returns Earnings in firms with extreme

6 The higher persistence of earnings at the current quarter t into quarter t+4 is evidence of seasonality, which is expected in quarterly data.

7 As noted at the bottom of Table 3, the coefficients on net operating cash flows and accruals are

statistically different from each other in all four quarterly regressions.

quarterly accruals will tend to be less persistent than earnings of firms with low levels of accruals When investors learn the surprising actual quarterly earnings in the future, stockprices decline for high accruals firms with lower future earnings and rise for low accrualsfirms that experience higher earnings in future quarters.

(Insert Table 4 about here)Table 5 provides regression results that examine the association between future returns and current earnings, net operating cash flows and accruals It shows that future BHR’s are positively and significantly associated with current earnings for every holding period through the subsequent four quarters An even stronger association is observed between future excess returns and current net operating cash flows, likely due to the higher persistence of net operating cash flows In contrast, there is a significantly lower incremental association between accruals in quarter t and future excess returns, after controlling for the levels of net operating cash flows, likely due to the lower persistence

of accruals These results indicate that future returns are more strongly related to net operating cash flows than to accruals

(Insert Table 5 about here)Table 6 examines the differential future returns to the extreme decile portfolios using a methodology that is used extensively in post-earnings-announcement drift

studies Firms are sorted into accrual deciles every quarter and assigned the decile rank (with the lowest accrual decile obtaining the rank of 0 and the highest the rank of 9) The decile rank is divided by 9, which is then subtracted by 0.5, so that the scaled decile rank now ranges between -0.5 and 0.5 When future BHR’s are regressed on the scaled decile rank, the intercept measures the average excess return across all deciles, while the slope

coefficient measures the differential return between the highest accrual decile and the lowest accrual decile As can be seen in Table 6, the intercept is almost never significant, whereas the slope coefficient is always negative and statistically different from zero This

is expected since the highest accrual decile is expected to yield future returns that are lower than the lowest accrual decile The table also shows that the “hedge” portfolio of the lowest accrual decile minus the highest accrual decile yields an excess return of 2.71% (-0.0271 slope coefficient) for the first subsequent quarter, 5.19% through the second quarter, 6.72% through the third quarter, and 7.82% through the fourth subsequentquarter These are both economically and statistically significant excess returns, even after transaction costs

(Insert Table 6 about here)Table 7 examines the average returns on extreme accruals and hedge portfolios across the 53 quarters, from the first quarter of 1991 through the first quarter of 2004, using a testing methodology similar to Fama and MacBeth (1973) As the table reports, the long portfolio (lowest decile accruals) has a mean return from two days after the SEC filing through one day after the next earnings announcement of 2.62% over the 53

quarters, statistically significant at the 0.0001 level In 41 of 53 quarters, the portfolio yields positive returns In 32 of the 53 quarters, the portfolio yields positive returns above the presumed transaction costs of 2% For the equivalent short portfolio (the highest decile accruals), the mean return is much lower at 98% although statistically significant at the 0.0056 level, with 36 quarters of above zero returns and 19 quarters of above 2% returns The table shows that the hedge portfolio, which combines the long andshort positions, has a mean return of 3.59%, is positive in 48 of the 53 quarters, and has

above 2% returns in 37 of the 53 quarters The table also reports the results of longer holding horizons (2-4 quarters after the SEC filings) The table shows that the benefits of the accruals strategy are increasing with longer holding returns with short positions, which provide larger and more significant returns over a period that spans at least two quarters after the SEC filings Thus, the accruals strategy seems to be yielding abnormal returns that are both statistically and economically significant.

(Insert Table 7 about here)Figure 1 shows the BHR abnormal returns from two days after the SEC filing through a day after the subsequent earnings announcement by quarter for the lowest (long) and highest (short) accrual decile, as well as for the hedge portfolio that combines both positions As can be seen in the figure, most long returns are positive (41 out of 53 quarters), and most short returns are negative (36 out of 53 quarters) Furthermore, the hedge portfolio yields positive BHR in 48 out of 53 quarters, and the magnitude of the most negative BHR is only about 5% Thus, sorting firms by accruals is successful in predicting the returns through the next quarterly announcements Figure 2 shows

cumulative abnormal returns to the long and short accrual deciles, as well as the hedge portfolio that combines both throughout the 53 quarters used in the study As can be seen

in the figure, the cumulative returns of the long accrual decile are greater than the short accrual decile, but both are clearly dominated by the hedge portfolio Thus, using the information inherent in accruals can help investors obtain consistent abnormal returns over a lengthy period of time

(Insert Figures 1 and 2 about here)

Robustness Checks:

1 Large Firms – We repeat the analysis in Table 7 for firms with a market value of

equity in excess of $1 billion For these firms, the long portfolio generally yields returns that are not significantly different from zero, but the short portfolio yields returns that are statistically and economically significant The hedge portfolio yields returns that are larger than those reported in Table 7 for all firms

2 Recent Data and Post-EDGAR data – We repeat the analysis in Table 7 with

data from quarters after 1999 The hedge portfolio returns are better than those provided in Table 7, but mostly due to the long portfolio position The short portfolio yields returns that are insignificantly different from zero We also examined the results of Table 7 for the post-EDGAR period (after 1996) The results are similar to those in Table 7 for both the short and long portfolios, and for the hedge portfolio

3 Balance Sheet or Cash Flow Information is not in the Preliminary Earnings Announcement – We repeat the analysis for firms that did not report information

in their preliminary earnings announcements on the main current accruals

(inventories, accounts receivable and accounts payable) or net operating cash flows The results are very similar to those reported in Table 7

4 Fourth fiscal quarter versus other quarters – The main results of Table 7 are

similar for the first three fiscal quarters and the fourth quarter, except that returns

in the fourth quarter are slightly lower than those in Table 7 although statistically significant, and those for the first three quarters are slightly higher