THE ACCOUNTING REVIEW

Vol. 77
Supplement 2002
pp. 35–59

The Quality of Accruals and

Earnings: The Role of Accrual
Estimation Errors
Patricia M. Dechow
University of Michigan
Ilia D. Dichev
University of Michigan

ABSTRACT: This paper suggests a new measure of one aspect of the quality
of working capital accruals and earnings. One role of accruals is to shift or
adjust the recognition of cash flows over time so that the adjusted numbers
(earnings) better measure firm performance. However, accruals require as-
sumptions and estimates of future cash flows. We argue that the quality of
accruals and earnings is decreasing in the magnitude of estimation error in
accruals. We derive an empirical measure of accrual quality as the residuals
from firm-specific regressions of changes in working capital on past, present,
and future operating cash flows. We document that observable firm charac-
teristics can be used as instruments for accrual quality (e.g., volatility of ac-
cruals and volatility of earnings). Finally, we show that our measure of accrual
quality is positively related to earnings persistence.

Keywords: quality of accruals; earnings quality; estimation errors; earnings

persistence.
Data Availability: Data are available from sources identified in the paper.

I. INTRODUCTION

T
his paper suggests a new measure of one aspect of the quality of working capital
accruals and earnings. The measure is based on the observation that accruals shift
or adjust the recognition of cash flows over time, so that the adjusted numbers
(earnings), better measure firm performance (e.g., see Statement of Accounting Concepts

We thank workshop participants at The University of Arizona, University of Chicago, Columbia University,
Indiana University, University of Michigan, The Accounting Review Conference on Quality of Earnings held at
Emory University, the 2001 Big Ten Research Conference, and especially Dan Collins, Bob Lipe, Mort Pincus,
Katherine Schipper, Peter Wysocki, and two anonymous reviewers. Professor Dechow acknowledges financial
support provided by the Michael A. Sakkinen Research Scholar Fund at the University of Michigan Business
School. Professor Dichev acknowledges the financial support of the Sanford Robertson Assistant Professorship in
Business Administration and the Norman Auerbach PricewaterhouseCoopers Faculty Fellowship.

35
36 The Accounting Review, 2002 Supplement

No. 1, FASB 1978, para. 44). For example, recording a receivable accelerates the recog-
nition of a future cash flow in earnings, and matches the timing of the accounting
recognition with the timing of the economic benefits from the sale. However, accruals are
frequently based on assumptions and estimates that, if wrong, must be corrected in future
accruals and earnings. For example, if the net proceeds from a receivable are less than the
original estimate, then the subsequent entry records both the cash collected and the correc-
tion of the estimation error. We argue that estimation errors and their subsequent corrections
are noise that reduces the beneficial role of accruals. Therefore, the quality of accruals and
earnings is decreasing in the magnitude of accrual estimation errors. Our empirical measure
of accrual quality is the extent to which working capital accruals map into operating cash
flow realizations, where a poor match signifies low accrual quality.
The intuition behind this measure is available in both theoretical and practical texts
that observe that the beneficial role of accruals is reduced by various limitations, including
estimation errors. For example, Palepu et al. (2000) discuss estimation errors as a factor
that reduces accounting quality, and suggest that estimation accuracy depends on firm char-
acteristics like complexity of transactions and predictability of the firm’s environment. We
develop this existing intuition into a practical measure of accrual and earnings quality.
Our approach to estimating the quality of accruals is related to two streams of research.
First, several studies document the benefits of the accrual process, finding that earnings is
a better measure of performance than the underlying cash flows (e.g., Dechow 1994;
Dechow et al. 1998; Liu et al. 2002). We build on this evidence by exploring the trade-
offs inherent in the accrual process. Second, a number of studies use models of ‘‘discre-
tionary accruals’’ to investigate the manipulation of accruals to achieve earnings manage-
ment goals (see Healy and Wahlen [1999] for a recent review). These studies focus on the
opportunistic use of accruals to window-dress and mislead users of financial statements.
This stream of research suggests that managerial intent affects the incidence and magnitude
of accrual estimation errors.
In contrast, we argue that even in the absence of intentional earnings management,
accrual quality will be systematically related to firm and industry characteristics. This dis-
tinction is important because such characteristics are likely to be both observable and
recurring (e.g., the volatility of operations is systematically related to the propensity to
make estimation errors) as compared to the determinants of managerial opportunism that
are often unobservable and/or sporadic (e.g., before stock offerings). For our purposes, we
do not attempt to disentangle ‘‘intentional’’ estimation errors from unintentional errors be-
cause both imply low-quality accruals and earnings.
We develop a model that examines the origination and reversal of working capital
accruals in a stylized firm. The model embodies the intuition that the timing of the firm’s
economic achievements and sacrifices often differs from the timing of the related cash flows,
and that the benefit of accruals is to adjust for these cash flow timing problems. However,
the model also reveals that the benefit of using accruals comes at the cost of including
accrual components that initiate and correct estimation errors. We focus on working capital
accruals and operating cash flows for tractability: the initiation and reversal of these accruals
occurs within a year. Our measure of accrual estimation errors is the residuals from firm-
specific regressions of changes in working capital on last year, present, and one-year ahead
cash flows from operations. These residuals are unrelated to cash flow realizations, and
include the estimation errors and their reversals. The standard deviation of these residuals
is our firm-specific measure of quality of accruals and earnings, where a higher standard
deviation signifies lower quality.
We illustrate the usefulness of our analysis in two ways. First, we explore the relation
between our measure of accrual quality and firm characteristics. The nature of the accrual
Dechow and Dichev—The Quality of Accruals and Earnings 37

process suggests that the magnitude of estimation errors will be systematically related to
business fundamentals like the length of the operating cycle and variability of operations.
We find that accrual quality is negatively related to the absolute magnitude of accruals, the
length of the operating cycle, loss incidence, and the standard deviation of sales, cash flows,
accruals, and earnings, and positively related to firm size. Our results suggest that these
observable firm characteristics can be used as instruments for accrual quality. This is im-
portant because the regression-based estimation of accrual quality demands long time-series
of data and the availability of subsequent cash flows, which makes it costly or infeasible
for certain practical applications (e.g., quality-of-accruals-based trading strategies).
Second, we illustrate the usefulness of our analysis by exploring the relation between
our measure of accrual quality and earnings persistence. Firms with low accrual quality
have more accruals that are unrelated to cash flow realizations, and so have more noise and
less persistence in their earnings. Indeed, we find a strong positive relation between accrual
quality and earnings persistence. However, our measure of accrual quality is theoretically
and empirically related to the absolute magnitude of accruals, and Sloan (1996) documents
that the level of accruals is less persistent than cash flows. Probing further, we find that
accrual quality and level of accruals are incremental to each other in explaining earnings
persistence, with accrual quality the more powerful determinant.
The remainder of the study is organized as follows. Section II presents our model of
accrual quality. Section III describes the sample and provides descriptive statistics. Section
IV derives the empirical measure of accrual quality and explores the relation between
accrual quality and both firm characteristics and earnings persistence. Our conclusions are
provided in Section V.

II. THEORETICAL MEASURE OF ACCRUAL AND EARNINGS QUALITY

Model of Accruals
Our model of accruals focuses on working capital accruals because cash flow realiza-
tions related to working capital generally occur within one year, making both the theory
and the empirics more tractable. While the intuition about errors in estimation applies to
all accruals, the long lags between noncurrent accruals and cash flow realizations practically
restrict the application of our approach to only short-term accruals.
We build our accrual framework around the observation that earnings equals cash flows
plus accruals, E ⫽ CF ⫹ Accruals. Cash flows for any period t can be categorized into
three groups:

Symbol Name
CFtt⫺1 Cash Collections or Payments of Amounts Accrued
at t⫺1 (net)
CFtt Current Cash Flows (net)
CFtt⫹1 Cash Flows Deferred to t⫹1 (net)

From an accounting perspective, for each cash flow, the two important events are the
receipt or disbursement of the cash flow and the recognition of that cash flow in earnings
(as revenue or expense). The subscript refers to the period the cash flow is received or
disbursed, and the superscript refers to the period the cash flow is recognized in earnings.
For example, CFtt–1 denotes that the cash flow occurs after the corresponding amount is
recognized in earnings (e.g., collection of an accounts receivable). CFtt refers to cash flows
received or paid in the same period as the cash flows are recognized in earnings. Finally,
CFtt⫹1 refers to cash received or paid before the revenue or expense is recognized in earn-
ings, such as cash payments for inventory.
38 The Accounting Review, 2002 Supplement

Summing up, total cash flow for period t is:

CFt ⫽ CFtt⫺1 ⫹ CFtt ⫹ CFtt⫹1 (1)

The accounting system provides for accruals, temporary adjustments that shift the rec-
ognition of cash flows over time. When recognition of a cash flow is shifted, two accrual
entries are created, an opening and a closing accrual. The opening accrual is initiated when
either (1) revenue or expense is recognized before the cash is received or paid or (2) cash
is received or paid before it is recognized in earnings. The closing accrual is recorded when
the other element of this pair has occurred and reverses the accrual portion of the original
entry.
When cash flows occur after the corresponding revenues and expenses are recognized
in earnings, managers must estimate the amount of cash to be received or paid in the future.
To the extent that cash flow realizations differ from their accrual estimates, the opening
accrual will contain an estimation error that is corrected by the closing accrual.
We incorporate this intuition in our model using the following notation for opening
and closing accruals related to future cash flows:

Name Accrual ⫽ Amount

Accrual for future collections ACFt⫹1 / tO ⫽ CFt⫹1t ⫹ εt⫹1t
and payments—Opening
Same sign as related cash
flow
Accrual for future collections ACFt / t–1C ⫽ –CFtt–1 – εtt–1
and payments—Closing
Opposite sign of related cash
flow

For accruals, the subscript refers to the corresponding cash flow, while the superscript
indicates opening or closing accruals. For example, ACFt⫹1 / tO is an opening accrual related
to cash flow CFt⫹1t. The opening accrual at time t, which reflects the expectation of the
t⫹1 cash flow, is equal to the actual t⫹1 cash flow plus an error term that reflects the
difference between the accrual expectation and the cash flow realization. We assume that
all accruals are resolved within one period so the closing accrual for cash collections and
payments at time t offsets the opening accrual from t⫺1. The closing accrual is equal to
the actual cash flow collected or paid in t plus an error term equal to the difference between
last period’s expectation and this period’s cash flow realization. Thus, each period’s total
accruals contain an estimation error in the opening accruals (the actual value of which is
determined next period, εt⫹1t) plus a realized error in the closing accruals (determined from
current period cash flow realizations, εtt–1).1
We next provide the notation for situations where cash flows are received or paid before
their recognition in earnings. In this case, the accounting system recognizes the amount of

1
We assume that estimation errors are independent of each other and of the magnitude of the cash flow realizations,
within the firm. While estimation errors can be correlated across firms (e.g., bad debt expense proves to be
systematically understated in a recession), cross-firm dependencies are not a problem for our approach because
our tests are at the firm level. Our approach also ignores estimation errors that are correlated at the firm
level (e.g., an optimistic manager systematically underestimates expenses). This is one of the limitations of
our approach.
Dechow and Dichev—The Quality of Accruals and Earnings 39

the cash flow either as deferred revenue (for cash inflows) or a deferred cost (for cash
outflows):

Name Accrual ⫽ Amount

Accrual that defers the ACFt / t⫹1O ⫽ –CFtt⫹1
recognition of cash flows—Opening
Opposite sign of related cash flow
Accrual that defers the ACFt–1 / tC ⫽ CFt–1t
recognition of cash flows—Closing
Same sign as related cash flow

As before, accrual subscripts refer to the corresponding cash flows, while superscripts
indicate opening or closing accruals. For example, ACFt / t⫹1O is an opening accrual related
to cash flow CFtt⫹1. Since deferred cash flows occur before recognition, these accruals
contain no estimation error (in the sense of our model). For example, an inventory purchase
initiates a positive accrual equal to the purchase price cash outflow, while the sale of the
inventory triggers a negative accrual equal in magnitude to the initial cash flow. Appendix
A provides a discussion and examples of whether and how our model detects accrual
estimation errors depending on the timing of related accruals and cash flows.
Using Expression (1), and defining total accruals as the sum of opening and closing
accruals, allows us to express earnings as:

Et ⫽ CFt ⫹ Accrualst
Et ⫽ (CFtt⫺1 ⫹ CFtt ⫹ CFtt⫹1) ⫹ (ACFt⫹1 / tO ⫹ ACFt / t⫺1C ⫹ ACFt / t⫹1O ⫹ ACFt⫺1 / tC)
Et ⫽ (CFtt⫺1 ⫹ CFtt ⫹ CFtt⫹1) ⫹ (CFt⫹1t ⫹ εt⫹1t ⫺ CFtt⫺1 ⫺ εtt⫺1 ⫺ CFtt⫹1 ⫹ CFt⫺1t) (2)

Re-arranging yields:

Et ⫽ CFt⫺1t ⫹ CFtt ⫹ CFt⫹1t ⫹ εt⫹1t ⫺ εtt⫺1. (3)

Equation (3), which presents earnings as the sum of past, present, and future cash flows
plus an adjustment for estimation errors and their correction, expresses the main intuition
of the paper. The intertemporal shifting of cash flows reflected in the cash flow terms of
Equation (3) alleviates the timing problems of using current cash flows as a measure of
performance. However, this benefit comes at the cost of using estimates, with the result
that earnings includes both errors in estimates and their corrections. Both errors and error
corrections reduce the quality of earnings as a measure of performance.
Most of our predictions about the relations between the variables in Equations (2) and
(3) concern relations studied by existing research. We include them to show that our model
embodies the intuitive and well-known properties of earnings, cash flows, and accruals, and
as a benchmark and link to existing results. Using Equations (2) and (3) allows one to
make the following predictions:
a) Corr(Et, CFt) ⫽ ⫹
b) Corr(Et, Accrualst) ⫽ ⫹
c) Corr(CFt, Accrualst) ⫽ ⫺
40 The Accounting Review, 2002 Supplement

These predictions, concerning the correlations between contemporaneous earnings, cash

flows, and accruals, have been confirmed by, for example, Dechow (1994) and Dechow et
al. (1998).
d) Corr(Et, CFt⫹1) ⫽ ⫹
e) Corr(Accrualst, CFt⫹1) ⫽ ⫹
These correlations suggest that earnings, and particularly its accrual portion, anticipates
future cash flows. This result has been confirmed by, for example, Finger (1994) and Barth
et al. (2001). Note that these predictions are based on ceteris paribus assumptions, so there
is a need to control for possible confounding factors. Specifically, in examining the relation
between current accruals and future cash flows, one needs to control for the confounding
effect of current cash flows. The reason is that current accruals are negatively related to
current cash flows, and future cash flows are (empirically) positively related to current cash
flows.
f) Corr(Et, CFt⫺1) ⫽ ⫹
g) Corr(Accrualst, CFt⫺1) ⫽ ⫹
To our knowledge, the prediction that current earnings, and particularly its accrual
portion, is positively correlated with past cash flows is novel. The intuition is that accruals
defer the recognition of some past cash flows into current earnings. As mentioned above,
in examining this prediction, it is important to control for the confounding effect of current
cash flows.

Deriving an Empirical Measure of Quality of Accruals

We obtain an expression for accruals at time t by rearranging the accrual portion of
earnings in Equation (2):

At ⫽ CFt⫺1t – (CFtt⫹1 ⫹ CFtt⫺1) ⫹ CFt⫹1t ⫹ εt⫹1t ⫺ εtt⫺1. (4)

Equation (4) conveys that (1) accruals are temporary adjustments that delay or antici-
pate the recognition of realized cash flows plus an estimation error term; (2) accruals are
negatively related to current cash flows and positively related to past and future cash flows;
and (3) the error term captures the extent to which accruals map into cash flow realizations,
and can be used as a measure of accrual and earnings quality.
To derive practical measures of working capital accrual quality, we use the following
firm-level time-series regression:

⌬WCt ⫽ b0 ⫹ b1*CFOt⫺1 ⫹ b2*CFOt ⫹ b3*CFOt⫹1 ⫹ εt. (5)

Relative to Equation (4), our measure of accruals is changes in working capital, and
our proxies for cash flows related to accruals is cash flow from operations (CFO). The
residuals from the regression reflect the accruals that are unrelated to cash flow realizations,
and the standard deviation of these residuals is a firm-level measure of accrual quality,
where higher standard deviation denotes lower quality.
The theoretical specification in Equation (4) uses only the portions of past, present, and
future cash flows that are related to current accruals. Since we cannot identify these cash
Dechow and Dichev—The Quality of Accruals and Earnings 41

flow components, the empirical version in Expression (5) uses total CFOs.2 Thus, the in-
dependent variables in Expression (5) are measured with error, implying that the regression
coefficients are likely to be biased toward 0, and the R2 will be reduced. Since the theoretical
values of the coefficients from Equation (4) are b1⫽ 1, b2 ⫽ ⫺1, and b3 ⫽ 1, we expect
that 0 ⬍ b1 ⬍ 1, and ⫺1 ⬍ b2 ⬍ 0, and 0 ⬍ b3 ⬍ 1. Appendix B provides a more expansive
treatment of this issue, including simulation results that confirm that the estimated coeffi-
cients and the R2 will be biased toward zero. Appendix B also reveals that the bias is more
severe for b1 and b3 as compared to b2, because the error component in past and future
cash flows is greater than the component in current cash flows. Later in the paper (Section
IV), we implement additional tests to control for the effect of this measurement error on
our results. Finally, we add an intercept to Expression (5) to capture possible nonzero
average accruals (e.g., average positive working capital accruals due to firm growth).

III. SAMPLE SELECTION, DESCRIPTIVE STATISTICS,

AND CALIBRATION TESTS
Table 1 summarizes our sample selection. Our sample is obtained from the Compustat
annual industrial and research files over 1987 to 1999. Given Collins and Hribar’s (2002)
result that the balance-sheet approach to deriving CFO leads to noisy and biased estimates,
we use CFO as given in the Statement of Cash Flows reported under the Statement of
Financial Accounting Standards No. 95 (SFAS No. 95, FASB 1987).3 The sample is re-
stricted to firms with complete data for assets, earnings, cash flow from operations, changes
in accounts receivable, and changes in inventory. The last two requirements ensure sample
firms have significant working capital accruals. Truncating the most extreme 1 percent of
cash from operations, earnings, and changes in working capital, and requiring at least one

2
Property-casualty insurance firms provide information about their accrual estimates, the subsequent cash flow
realizations, and the resulting estimation errors (e.g., Petroni 1992; Anthony and Petroni 1997; Beaver and
McNichols 1998). We do not pursue an industry-specific analysis because the cross-industry variation in the
magnitude of estimation errors is likely to be larger than the typical within-industry variation, and because our
interest is in more generalizable results.
3
This Standard required the presentation of the statement of cash flows for fiscal years ending after July 15, 1988.
However, some firms adopted this standard early in 1987.

TABLE 1
Derivation of Sample, 1987 to 1999

Firm-years with available cash from operations, earnings, and changes in accounts 59,360
receivable and inventory and total assets reported in the statement of cash flows
Firm-years after truncation of the most extreme 1 percent of earnings, cash from 55,850
operations, and changes in working capital
Firm-years with both lead and lag values of cash from operations 30,317
Firms with 8 or more annual observations 1,725
Firm-years available for the 1,725 firms used in the remaining analysis 15,234
Three-digit SIC groups with over 50 observations 136
Firm-years available for the 136 industries 27,204

All data is from Compustat.

42 The Accounting Review, 2002 Supplement

year of past and future cash flows and earnings, yields a sample of 30,317 firm-years. In
addition, we require at least eight years of data to estimate firm-specific regressions. This
restriction reduces our sample to 15,234 firm-year observations for 1,725 firms.4 For some
of our industry-level analyses, less stringent data requirements yield a sample of 27,204
firm-year observations over 136 three-digit SIC industries.
Cash flow from operations is Compustat item 308. The change in working capital from
year t⫺1 to t (⌬WC), is computed as ⌬AR ⫹ ⌬Inventory ⫺ ⌬AP ⫺ ⌬TP ⫹ ⌬Other Assets
(net), where AR is accounts receivable, AP is accounts payable, and TP is taxes payable.
Specifically, ⌬WC is calculated from Compustat items as ⌬WC ⫽ ⫺ (item 302 ⫹ item 303
⫹ item 304 ⫹ item 305 ⫹ item 307).
We calculate earnings after short-term accruals but before long-term accruals (Earn) as
Earn ⫽ CFO ⫹ ⌬WC. We report Earnings before extraordinary items (Prof) as Compustat
item 123, and Accruals (Prof-CFO) to provide comparability with other research. All var-
iables are scaled by average total assets. We also calculate the length of the operating cycle
(OC) as:

360/(Sales/Average AR) ⫹ 360/(Cost of Goods Sold)/(Average Inventory)

where Sales is Compustat item 12, cost of goods sold is item 41, AR is item 2, and
Inventory is item 3.
Descriptive statistics and correlations are provided in Table 2. An examination of Panel
A reveals that descriptive statistics are in line with those of other studies using similar
variables and time period (e.g., Barth et al. 2001). Earn exceeds CFO, implying that short-
term accruals are mostly positive. This is not surprising, given that most firms are growing
and therefore increasing their working capital. As expected, average accruals are negative
(⫺0.046), primarily because of depreciation.
The Pearson correlations in Panels B and C of Table 2 illustrate the relations between
our sample variables and provide comparability with previous research (results for Spear-
man correlations are similar). These empirical correlations are in agreement with existing
findings and the predictions of the model. Specifically, there is a positive contemporaneous
correlation between Earn and CFO (0.73), and between Earn and ⌬WC (0.33), and a neg-
ative correlation between CFO and ⌬WC (⫺0.41). We also find that Accruals and ⌬WC
are highly positively correlated (0.75), suggesting that working capital accruals capture
much of the variation in total accruals.
Consistent with Barth et al. (2001), we find that earnings and changes in working capital
anticipate future cash flows from operations. Note that the simple correlation between ⌬WCt
and CFOt⫹1 is ⫺0.01 and statistically insignificant. As previously discussed, the reason is
that ⌬WCt is negatively correlated with CFOt, and CFOt is positively correlated with CFOt⫹1
(0.56), which counteracts the expected positive relation between ⌬WCt and CFOt⫹1. In Panel
C of Table 2, we report that the partial correlation between ⌬WCt and CFOt⫹1, controlling
for CFOt, is 0.29 (significant at the 0.0001 level). We also find that working capital accruals
are positively related to past cash flows, implying that accruals defer the recognition of
some past cash flows. The simple correlation between ⌬WCt and CFOt⫺1 in Panel B is only
0.008 but the partial correlation controlling for CFOt in Panel C is 0.31 (significant at the
0.0001 level). Summarizing, our descriptive statistics and correlation results are in line with
predictions and existing results, indicating that our model captures reasonably well some
of the key features of accrual accounting.

4
481 firms have eight available observations, 1,054 have nine observations, and 190 firms have ten observations.
Dechow and Dichev—The Quality of Accruals and Earnings 43

TABLE 2
Descriptive Statistics and Correlations for 15,234 Firm-Year Observations
1987 to 1999

Panel A: Descriptive Statistics

Standard Lower Upper
Mean Deviation Quartile Median Quartile
Cash flow from operations (CFO) 0.075 0.097 0.033 0.082 0.131
Change in working capital (⌬WC) 0.015 0.070 ⫺0.015 0.010 0.044
Earnings before long-term accruals (Earn) 0.091 0.093 0.057 0.096 0.141
Earnings before extraordinary items (Prof) 0.030 0.113 0.009 0.042 0.081
Accruals (Prof-CFO) ⫺0.046 0.098 ⫺0.084 ⫺0.044 ⫺.0035
Total Assets (in millions) 2,436 10,878 50 240 1,215

Panel B: Pearson Correlations

Earnt CFOt ⌬WCt CFOt⫹1 CFOt⫺1 Earnt⫹1 Accrualst
Proft 0.816* 0.575* 0.292* 0.458* 0.470* 0.540* 0.585*
Earnt 0.728* 0.325* 0.571* 0.575* 0.674* 0.220*
CFOt ⫺0.411* 0.558* 0.549* 0.597* ⫺0.327*
⌬WCt ⫺0.011 0.008 0.072* 0.745*
CFOt⫹1 0.482* 0.735* ⫺0.024
CFOt⫺1 0.491* ⫺0.001
Earnt⫹1 0.031*

Panel C: Partial Correlations (controlling for the effect of CFOt )

Pearson
Correlation CFOt⫹1 CFOt⫺1
⌬WCt 0.289* 0.306*

* Significant at the 0.0001 level.

Variable definitions:
Cash flow from operations (CFO) ⫽ item 308 from the Compustat Statement of Cash Flows;
Change in working capital (⌬WC) ⫽ ⌬AR ⫹ ⌬Inventory ⫺ ⌬AP ⫺ ⌬TP ⫹ ⌬Other Assets (net),
where AR is accounts receivable, AP is accounts payable, and
TP is taxes payable;
Earnings before long-term accruals (Earn) ⫽ CFO ⫹ ⌬WC;
Earnings before extraordinary items (Prof) ⫽ Compustat item 123; and
Accruals ⫽ Prof ⫺ CFO.
All variables are scaled by average total assets.

IV. RESULTS
An Empirical Measure of Accrual Quality
Table 3 presents results of regressions of working capital accruals on past, present, and
future cash flows from operations. First, we present firm-level regressions (Panel A) because
44 The Accounting Review, 2002 Supplement

TABLE 3
Regressions of the Change in Working Capital on Past, Current, and Future Cash Flow from
Operations for Firm-Years between 1987 to 1999

⌬WCt ⫽ b0 ⫹ b1CFOt⫺1 ⫹ b2CFOt ⫹ b3CFOt⫹1 ⫹ ␧t

Intercept b1 b2 b3 Adjusted R 2
Panel A: Firm-Specific Regressions (1,725 firms)
Mean 0.04 0.17 ⫺0.62 0.09 0.47
(t-statistic) (23.03) (19.38) (⫺57.06) (10.38)
Lower quartile 0.001 ⫺0.02 ⫺0.91 ⫺0.10 0.23
Median 0.04 0.14 ⫺0.65 0.09 0.55
Upper quartile 0.08 0.35 ⫺0.35 0.28 0.80

Panel B: Industry-Specific Regressions (136 industries)

Mean 0.03 0.19 ⫺0.51 0.15 0.34
(t-statistic) (16.09) (21.10) (⫺35.77) (15.33)
Lower quartile 0.01 0.11 ⫺0.63 0.08 0.22
Median 0.03 0.18 ⫺0.52 0.15 0.34
Upper quartile 0.04 0.26 ⫺0.40 0.23 0.45

Panel C: Pooled Regression (15,234 firm-year observations)

Coefficient 0.03 0.19 ⫺0.51 0.18 0.29
(t-statistic) (39.43) (32.12) (⫺78.29) (29.18)

The t-statistics in Panel A are determined based on the distribution of the 1,725 coefficients obtained from the
firm-specific regressions requiring a minimum of eight observations per firm. T-statistics in Panel B are determined
based on the distribution of the 136 coefficients obtained from three-digit SIC grouping regressions requiring a
minimum of 50 observations per grouping.
All variables are defined as in Table 2.

our theory is defined and most naturally applied on a firm-level basis. In addition, we
expect that a firm-level specification is superior to cross-sectional specifications because
the regression coefficients are likely to differ across firms. The reason is that our operating
cash flow proxies contain measurement error, which is likely to be systematically related
to firm characteristics. For example, a long operating cycle implies that substantial future
cash flows are recognized in current earnings (e.g., by accruing receivables). As a result,
when future cash flows related to present accruals are a large part of total future cash flows,
we expect that total future cash flows will load with a larger coefficient in Expression (5).
Since such effects are likely related to firm-level characteristics, we believe that a firm-
level specification is superior. However, we also present industry-specific and pooled results
(Panels B and C), because our firm-specific time-series is short, and we are concerned about
noisy estimation at the firm level.
Results for the firm-specific regressions in Panel A of Table 3 are consistent with the
theory and the univariate results in Table 2. As predicted, current changes in working capital
Dechow and Dichev—The Quality of Accruals and Earnings 45

are negatively related to current cash flow from operations, and positively related to past
and future cash flow from operations. The mean coefficient on current cash flows is ⫺0.62,
while the mean coefficients on past and future cash flows are 0.17 and 0.09, respectively.
As discussed earlier (and detailed in Appendix B), the absolute magnitudes of the coeffi-
cients are less than the theoretical value of 1 because of the measurement error in the
independent variables, and the downward bias is greater for the coefficients on past and
future cash flows. Based on the cross-sectional distribution of the firm-specific coefficients,
all of these means are highly statistically significant, with t-statistics ranging in absolute
value from 10 to 57. Adjusted R2s indicate that this specification provides reasonable ex-
planatory power (mean R2 ⫽ 0.47 and median R2 ⫽ 0.55).
Results for industry-specific and pooled regressions in Panels B and C of Table 3 are
consistent with the firm-specific results. The mean coefficients on current CFO are ⫺0.51
for both the industry-specific regressions and the pooled regression compared with –0.62
for the firm-specific mean. The coefficients on past and future cash flows are also compa-
rable in magnitude with the firm-specific results, on the magnitude of 0.15 to 0.19, and are
reliably positive. The adjusted R2s are lower, with an average of 0.34 for the industry
specification and 0.29 for the pooled regression, perhaps because the implications of our
firm-specific model are less descriptive in cross-sectional specifications. Summarizing, re-
sults for firm-specific, industry, and pooled specifications are similar. Based on better the-
oretical grounding and better empirical fit, we proceed with the firm-level specification.
We perform several tests of the robustness of our results. Recall that the cash flow
variables (CFOs) are noisy estimates of the theoretical cash flow variables (cash flows
related to accruals). Measurement error in the independent variables leads to both biased
coefficient estimates and biased estimates of residuals. Our sensitivity checks rely on the
observation that CFO is likely to be a good proxy for the theoretical cash flow variable
when the firm is in a steady state, because in these circumstances, cash flows related to
accruals are likely to be a fairly constant proportion of CFO.
We implement this observation on two dimensions. First, we rerun the regressions
reported in Table 3, controlling for the effect of sales growth, either by including a growth
term in the regression or by running the regression only on low-growth firms (defined as
percentage sales growth between ⫺5 and 5 percent). The tenor of the results remains the
same for both specifications (results not included). In addition, controlling for growth does
not appreciably affect the relations between accrual quality, firm characteristics, and per-
sistence of earnings (discussed later). Second, we analyze the effect of cash flow volatility
on accrual quality. If cash flow volatility is low, then we expect a relatively stable relation
between total cash flows and cash flows related to accruals. We control for cash flow
volatility either by including a cash flow volatility term in the regressions or by running
the regression on low-volatility firms only (firms with volatility in the lowest quintile).
Results remain qualitatively unchanged for both of these specifications.5
Finally, we investigate the effect of special items, which often contain long-term ac-
cruals, and therefore affect both our measure of short-term accruals and our measure of
cash flow from operations (we assume that cash flow from operations is related only to
short-term accruals). Specifically, sometimes Compustat includes restructuring charges in

5
Because we believe that accrual quality is theoretically negatively related to cash flow volatility, the results after
controlling for cash flow volatility are probably more correctly interpreted to indicate that the quality of accruals
concept has construct validity beyond that captured by cash flow volatility. However, that does not imply that
one should control for cash flow volatility in deriving accrual quality. Theoretically, high cash flow volatility
causes low accrual quality because of large forecast errors in volatile environments, and the effect of this causal
variable should not be excluded from the empirical construct. For this reason, we present the main results without
a control for cash flow volatility. A similar argument applies to sales growth.
46 The Accounting Review, 2002 Supplement

‘‘other assets and liabilities’’ (item 307), contaminating our measure of short-term accruals.
A portion of the Compustat cash flows from operations can also be related to restructuring
charges, contaminating our measure of cash flows. When we replicate the results in Panel
A of Table 3 for firms whose special items are less than 5 percent of earnings and 1 percent
of assets, the tenor of the results is unchanged. When we exclude ‘‘other assets and liabil-
ities’’ (item 307) from our measure of working capital accruals, the tenor of the results is
also unchanged.

The Relation between Accrual Quality and Firm Characteristics

We use the standard deviation of the residuals as a firm-specific measure of accrual
quality, where a higher standard deviation signifies lower quality.6 This empirical measure
of accrual quality can be utilized for a variety of purposes. For example, it can be used in
market-based tests of the relation between stock prices and earnings. Another application
is to adapt our measure of accrual quality to devise alternative tests of earnings manage-
ment. Accruals manipulated by management in an opportunistic manner often behave em-
pirically just like accruals that are the result of unavoidable errors in estimation. For ex-
ample, from the point of view of the accounting system, booking a fictitious receivable and
not collecting it looks similar to booking a regular receivable and not collecting it.7
In this section, we illustrate two applications of our measure of accrual quality. First,
we explore the relation between our measure and selected firm characteristics. While we
assume that the realization of individual estimation errors is random, the average magnitude
of these errors for a particular firm is likely to be systematically related to the firm’s
characteristics such as the volatility of operations or the length of the operating cycle. For
example, even with good skills and the best intentions, managers of firms in volatile in-
dustries are likely to make larger accrual estimation errors.
Mapping the relation between firm characteristics and estimation errors is important
for research, practical, and pedagogical purposes. The reason is that a firm-specific regres-
sion estimate of the standard deviation of the residuals requires a lengthy time-series. In
addition, the regression approach requires information about future cash flows, which re-
duces its usefulness in many settings (e.g., abnormal return strategies based on quality of
earnings). Therefore, it is valuable to identify observable firm characteristics that act as
instruments for the propensity to make estimation errors. We note that some firm charac-
teristics we examine are likely to be correlated with the standard deviation of the residuals
by construction (e.g., the standard deviation of accruals). This is not a concern, because
our main interest is in identifying strong relations between the unobservable estimation
errors and observable firm characteristics, regardless of the source of the relation.
Based on existing theory, results, and economic intuition, we expect that:
● The longer the operating cycle, the lower accrual quality.
Longer operating cycles indicate more uncertainty, more estimation and errors of
estimation, and therefore lower quality of accruals.
● The smaller the firm, the lower accrual quality.

6
An alternative measure of accrual quality at the firm-year level is absolute value of the residual for that year.
The tenor of the results is similar for this alternative specification, but the relations to firm characteristics and
earnings persistence are weaker.
7
Thus, as discussed earlier, the magnitude of estimation errors is also likely to be related to variables capturing
managerial skill and managerial opportunism, e.g., meeting analysts’ forecasts, auditor quality, stock offerings,
and closeness to debt covenants.
Dechow and Dichev—The Quality of Accruals and Earnings 47

We expect that large firms have more stable and predictable operations and, there-
fore, fewer and smaller estimation errors. In addition, large firms are likely to be
more diversified and various portfolio effects across divisions and business activities
reduce the relative effect of estimation errors.
● The greater the magnitude of sales volatility, the lower accrual quality
Sales volatility indicates a volatile operating environment and the likelihood of
greater use of approximations and estimation, with corresponding large errors of
estimation and low accrual quality.
● The greater the magnitude of cash flow volatility, the lower accrual quality
High standard deviation of cash flows is another measure of high uncertainty in the
operating environment.
● The greater the magnitude of accrual volatility, the lower accrual quality
Since our measure of accrual quality is derived as a residual from accruals, accrual
volatility and accrual quality are at least partly related by construction.
● The greater the magnitude of earnings volatility, the lower accrual quality
Earnings is the sum of cash flows and accruals. Since the volatility of both com-
ponents is predicted to be negatively related to earnings quality, we expect that
greater volatility in earnings signifies lower accrual quality.
● The greater the frequency of reporting negative earnings, the lower accrual
quality
Losses are indicative of severe negative shocks in the firm’s operating environment.
Accruals made in response to such shocks are likely to involve substantial estimation
error (e.g., restructuring charges). Thus, losses are indicative of low accrual quality.
● The greater the magnitude of accruals, the lower accrual quality
More accruals indicate more estimation and errors of estimation, and therefore lower
quality of accruals.

Table 4 provides results for these hypothesized relations. Panel A presents descriptive
statistics where some of the variables are first averaged on a firm basis to be consistent
with the rest of the variables and our firm-level measure of accrual quality. The distribution
of Average operating cycle (mean 141 days, standard deviation 62 days) indicates short
operating cycles. (In unreported tests, we find that all firms in our sample have average
operating cycles of less than one year.) Thus, our assumption that most working capital
accruals reverse within one year seems reasonable for this sample.
Panel B of Table 4 presents Pearson correlations between our measure of accrual quality
(sresid) and firm characteristics (results for Spearman correlations are similar). While all
correlations have the predicted signs, and all are significant at the 0.0001 level, there is a
large range of magnitudes. The correlations for length of the operating cycle (0.28) and
standard deviation of sales (0.34) are comparatively low. The correlations for all remaining
variables exceed 0.50, with highest correlations for the standard deviation of earnings (0.82),
the standard deviation of accruals (0.75), and proportion of negative earnings (0.63). Such
strong correlations suggest that these variables can be used as reliable instruments for
accrual quality. The high correlation with the average level of working capital accruals
(0.69) suggests a strong relation between sresid and the measure of accrual quality in Sloan
(1996). We explore this relation in more detail later in the paper.
In Panel C of Table 4 we investigate whether parsimonious combinations of these firm
characteristics capture accrual quality better than any individual variable. Our baseline spec-
ification regresses sresid on Std. Dev. Earn (the variable with the highest correlation in
Panel B) and shows an adjusted R2 of 0.67 (Model 1). Next, we decompose the volatility
48 The Accounting Review, 2002 Supplement

TABLE 4
Descriptive Statistics and the Correlation between Quality of Working Capital Accruals
(sresid ) and Selected Firm Characteristics for 1,725 Firms between 1987 to 1999

Panel A: Descriptive Statistics

Standard Lower Upper
Mean Deviation Quartile Median Quartile
Standard deviation of the 0.028 0.025 0.011 0.020 0.037
residuals (sresid )
Average operating cycle 141.08 61.807 90.229 131.505 184.81
Log (Total Assets) 5.50 2.16 3.88 5.44 7.06
Std. Dev. of Sales 0.215 0.205 0.097 0.166 0.273
Std. Dev. of CFO 0.060 0.040 0.031 0.051 0.078
Std. Dev. of ⌬WC 0.056 0.038 0.025 0.045 0.076
Std. Dev. of Earn 0.050 0.044 0.021 0.037 0.063
Proportion of earnings that 0.100 0.189 0 0 0.111
are negative for each
firm
Average 兩⌬WC兩 0.048 0.033 0.023 0.040 0.067

Panel B: Pearson Correlations between the Standard Deviation of the Residuals (sresid) and
Selected Firm Characteristics (p-values in parentheses)
Average Log Std. Std. Std. Prop. of
Op. (Total Dev. Std. Dev. Dev. Dev. Negative Average
Cycle Assets) Sales CFO ⌬WC Earn Earnings 兩⌬WC兩
0.28 ⫺0.55 0.34 0.60 0.75 0.82 0.63 0.69
(0.0001) (0.0001) (0.0001) (0.0001) (0.0001) (0.0001) (0.0001) (0.0001)

Panel C: Regressions Where the Dependent Variable is the Standard Deviation of the Residuals
(sresid) and the Independent Variables are Firm Characteristics (n ⫽ 1,725)
Std. Dev. Std. Dev. Std. Dev. Adj.
Intercept Earn ⌬WC CFO R2
(1) Coefficient 0.004 0.462 0.67
(t-statistic) (9.29) (58.63)
(2) Coefficient 0.001 0.480 0.008 0.57
(t-statistic) (1.34) (29.32) (0.56)
(3) Coefficient ⫺0.002 0.317 0.257 0.76
(t-statistic) (⫺4.10) (36.73) (26.17)
Prop. Std. Average Log
Negative Dev. Average Op. (Total Adj.
Intercept Earnings Sales 兩⌬WC兩 Cycle Assets) R2
(4) Coefficient 0.012 0.049 0.005 0.334 0.00 ⫺0.001 0.61
(t-statistic) (6.03) (21.62) (2.63) (21.79) (0.07) (⫺5.22)
(Continued on next page)
Dechow and Dichev—The Quality of Accruals and Earnings 49

TABLE 4 (Continued)

The standard deviation of the residuals (sresid ) is calculated based on the residuals from the following firm-specific
regressions:
⌬WCt ⫽ b0 ⫹ b1CFOt⫺1 ⫹ b2CFOt ⫹ b3CFOt⫹1 ⫹ εt
where:
Cash flow from operations (CFO) ⫽ item 308 from the Compustat Statement of Cash Flows;
Change in working capital (⌬WC) ⫽ ⌬AR ⫹ ⌬Inventory ⫺ ⌬AP ⫺ ⌬TP ⫹ ⌬Other Assets (net),
where AR is accounts receivable, AP is accounts payable, and
TP is taxes payable; and
Earnings before long-term accruals (Earn) ⫽ CFO ⫹ ⌬WC.
All variables are scaled by average assets.
The standard deviations of Sales, Earn, CFO, and ⌬WC are calculated at a firm level. Operating cycle is equal to
360 / (Sales Average AR) ⫹ 360 / (Cost of Goods Sold) / (Average Inventory). Proportion of earnings that are negative
is calculated as the number of firm-years with negative earnings divided by the total number of firm-years for each
firm. For this table, we use a log(Total Assets) specification to correct for right-tail skewness in assets.

of earnings into cash flow and accrual volatility, both of which are highly correlated with
sresid in Panel B. Model (2) indicates that Std. Dev. ⌬WC subsumes the explanatory power
of Std. Dev. CFO. For a likely explanation, recall that sresid is defined as the variation of
the component of accruals unrelated to cash flow realizations. Model (3) includes the vol-
atility of working capital accruals and the volatility of earnings, with the R2 increasing to
0.76. Finally, Model (4) regresses sresid on all remaining firm characteristics. The R2 is
0.61, lower than the R2 for the baseline regression with earnings volatility (Model 1).
Summarizing, a simple practical way to gauge the quality of accruals is to assess the
volatility of both earnings and accruals.

The Relation between Accrual Quality and Earnings Persistence

Earnings persistence is frequently discussed as a measure of earnings quality (e.g.,
Penman 2001, 623; Revsine et al. 2002, 245). The theory in this paper provides a natural
link between our measure of accrual quality and earnings persistence. Specifically, an ex-
amination of Equation (3) reveals that, holding the time-series properties of realized cash
flows constant, adding more accrual errors results in lower earnings persistence. Thus, we
expect that firms with low accrual quality will also have low earnings persistence.
The results on the empirical relation between accrual quality and earnings persistence
are presented in Panel A of Table 5. We report portfolio results to maintain comparability
with earlier studies (e.g., Sloan 1996; Barth et al. 2001) and to examine for potential
nonlinearities in the quality-persistence relation. The results have the same tenor in a re-
gression specification (not tabulated). In Panel A, firm-years are first sorted into quintile
portfolios based on the standard deviation of the firm-specific regression residuals (sresid)
from estimating Expression (5). Within each portfolio, we regress future earnings on current
earnings, and report the slope coefficient (called Persistence) and the adjusted R2. We find
a negative relation between standard deviation of the residuals and persistence; the slope
coefficient declines from 0.94 to 0.55 between quintiles 1 and 5. This decline is monotonic
but the relation is steeper in the tails. Since these are univariate regressions, the decline in
adjusted R2 from 0.83 to 0.28 parallels that of the slope coefficient. Thus, these results
50 The Accounting Review, 2002 Supplement

TABLE 5
The Relative Information Content of Accrual Quality and Level of Accruals for Earnings
Persistence for Firm-Years between 1987 to 1999

Panel A: Portfolios Based on the Magnitude of the Standard Deviation of the Residuals (sresid)
Earnings
Std. Dev. Persistence
Resid Persistence Regression Number of
Portfolio (sresid) 兩⌬WCt兩 (␦1) Adj. R2 Observations
1 0.006 0.023 0.943 0.830 3,047
2 0.013 0.033 0.816 0.651 3,043
3 0.020 0.042 0.799 0.619 3,049
4 0.032 0.058 0.756 0.545 3,045
5 0.069 0.086 0.551 0.280 3,050

Panel B: Portfolios Based on the Absolute Value of Accruals (兩⌬WCt兩)

Earnings
Persistence
Std. Dev. Resid Persistence Regression Number of
Portfolio 兩⌬WCt兩 (sresid) (␦1) Adj. R2 Observations
1 0.005 0.014 0.814 0.566 3,046
2 0.015 0.016 0.791 0.548 3,047
3 0.031 0.019 0.809 0.584 3,047
4 0.056 0.024 0.747 0.466 3,047
5 0.135 0.035 0.550 0.329 3,047

Accrual quality is measured as the standard deviation of working capital residuals from the following firm-specific
regression:
⌬WCt ⫽ b0 ⫹ b1CFOt⫺1 ⫹ b2CFOt ⫹ b3CFOt⫹1 ⫹ εt
Level of accruals is measured as the absolute value of current change in working capital.
Earnings persistence (␦1) is measured for each portfolio from the following regression:
Earnt⫹1 ⫽ ␣ ⫹ ␦1Earnt ⫹ vt
Cash flow from operations (CFO) ⫽ item 308 from the Compustat Statement of Cash Flows;
Change in working capital (⌬WC) ⫽ ⌬AR ⫹ ⌬Inventory ⫺ ⌬AP ⫺ ⌬TP ⫹ ⌬Other Assets (net),
where AR is accounts receivable, AP is accounts payable, and
TP is taxes payable; and
Earnings before long-term accruals (Earn) ⫽ CFO ⫹ ⌬WC.
All variables are scaled by average assets.
Dechow and Dichev—The Quality of Accruals and Earnings 51

confirm the hypothesized positive relation between accrual quality and earnings
persistence.8
Panel A of Table 5 also reports mean level of working capital accruals for each sresid
portfolio. Similar to the results in Table 4, these portfolio results reveal a positive relation
between standard deviation of the residuals and level of accruals. This relation is important
because Sloan (1996) shows that level of accruals is less persistent than cash flows. Thus,
at this point it is difficult to distinguish empirically between the effects of accrual quality
and the level of accruals on earnings persistence. However, as mentioned earlier, there are
a priori reasons to believe that our measure of accrual quality and the level of accruals are
both proxies for the same aspect of the unobservable ‘‘true’’ accrual quality. Thus, to a
large extent, the real issue is which of these two measures performs better empirically for
the task at hand—in this case, explaining earnings persistence.
We pursue two approaches to distinguish between the explanatory power of sresid and
level of accruals with respect to earnings persistence. In Panel B of Table 5, we form
quintiles based on the absolute magnitude of accruals and measure persistence for each
quintile.9 Results confirm the expected negative relation between the level of accruals and
earnings persistence; the slope coefficient declines from 0.81 to 0.55, and the adjusted R2
from 0.57 to 0.33, between quintiles 1 and 5. However, these declines are less pronounced
than those in Panel A. The relative decline in Persistence is 0.39 (0.55 for adjusted R2)
across the accrual quality quintiles vs. 0.26 (0.24 for adjusted R2) across the level of accrual
quintiles. In addition, the relation between level of accruals and persistence in Panel B is
nonmonotonic, and essentially flat for low- to medium-level accruals. The overall impres-
sion from Panel B is that level of accruals indicates low earnings persistence only for
extreme accrual realizations. Thus, the relation between accrual quality and earnings per-
sistence is stronger than the relation between level of accruals and persistence, especially
for firms with high accrual quality.
Our second approach is provided in Table 6, which reports portfolio results on the
relation between sresid and earnings persistence, holding the level of accruals constant
(Panel A); and the relation between the level of accruals and earnings persistence, holding
sresid constant (Panel B). To hold the level of accruals constant (as in Panel A), we first
sort the sample into decile portfolios based on level of accruals, and then within each level-
of-accruals decile, we further sort the observations into five portfolios based on standard
deviation of the residuals (subportfolios 1 to 5). We pool all ten subportfolios 1 together
into one portfolio, all subportfolios 2 together into another portfolio, and so on. The result
is five portfolios, with substantial variation of sresid but nearly the same level of accruals.
As shown in Panel A of Table 6, this two-pass portfolio construction controls quite
well for level of accruals, with mean 兩⌬WC兩 ranging from 0.047 to 0.051 across portfolios
1 to 5. Despite the strong positive correlation between sresid and level of accruals at the
firm level, the portfolios in Panel A exhibit large variation in sresid: the conditional range
between mean sresid of portfolios 1 and 5 in Table 6 is 0.55 compared to the unconditional
range of 0.63 in Table 5. This result indicates that the variation in sresid independent of

8
Our measure of accrual quality has some ‘‘look-ahead’’ bias in explaining earnings persistence because good
accrual quality means a positive relation between current accruals and future cash flows. An alternative specifi-
cation, which uses residuals from regressing accruals on only past and present cash flows produces similar
portfolio results. This robustness in results is probably explained by the relative stability of our measure of
accrual quality over time.
9
Our measure of accruals includes only changes in working capital, while the measure of accruals in Sloan (1996)
includes both changes in working capital and depreciation. However, most of the variation in Sloan’s measure
is due to variation in accounts receivable and inventory (Sloan 1996, 297).
52 The Accounting Review, 2002 Supplement

TABLE 6
The Incremental Information Content of Accrual Quality and Level of Accruals for Earnings
Persistence for Firm-Years between 1987 to 1999

Panel A: Portfolios Based on the Magnitude of the Standard Deviation of the Residuals (sresid),
Controlling for the Absolute Value of Accruals (兩⌬WCt兩)
Earnings
Persistence
Std. Dev. Persistence Regression Number of
Portfolio Resid (sresid) 兩⌬WCt兩 (␦1) Adj. R2 Observations
1 0.008 0.047 0.913 0.783 3,044
2 0.015 0.047 0.819 0.642 3,049
3 0.022 0.047 0.742 0.542 3,046
4 0.032 0.049 0.666 0.426 3,052
5 0.063 0.051 0.602 0.328 3,043

Panel B: Portfolios Based on the Absolute Value of Accruals (兩⌬WCt兩), Controlling for the Standard
Deviation of the Residuals (sresid)
Earnings
Persistence
Std. Dev. Resid Persistence Regression Number of
Portfolio 兩⌬WCt兩 (sresid) (␦1) Adj. R2 Observations
1 0.006 0.027 0.788 0.502 3,042
2 0.020 0.028 0.762 0.513 3,049
3 0.036 0.028 0.743 0.438 3,050
4 0.059 0.028 0.707 0.457 3,049
5 0.120 0.029 0.574 0.408 3,044

Accrual quality is measured as the standard deviation of working capital residuals from the following firm-specific
regression:
⌬WCt ⫽ b0 ⫹ b1CFOt⫺1 ⫹ b2CFOt ⫹ b3CFOt⫹1 ⫹ εt
Level of accruals is measured as the absolute value of current change in working capital.
Earnings persistence (␦1) is measured for each portfolio from the following regression:
Earnt⫹1 ⫽ ␣ ⫹ ␦1Earnt ⫹ vt
Cash flow from operations (CFO) ⫽ item 308 from the Compustat Statement of Cash Flows;
Change in working capital (⌬WC) ⫽ ⌬AR ⫹ ⌬Inventory ⫺ ⌬AP ⫺ ⌬TP ⫹ ⌬Other Assets (net),
where AR is accounts receivable, AP is accounts payable, and
TP is taxes payable; and
Earnings before long-term accruals (Earn) ⫽ CFO ⫹ ⌬WC.
All variables are scaled by average assets.
Dechow and Dichev—The Quality of Accruals and Earnings 53

the level of accruals is substantial and empirically important. Further, the regression results
reported in Table 6, Panel A reveal that the conditional variation in sresid is strongly related
to earnings persistence. The slope coefficient varies from 0.91 in portfolio 1 to 0.60 in
portfolio 5, and adjusted R2 varies from 0.78 in portfolio 1 to 0.33 in portfolio 5. These
results compare well to those from the unconditional specification in Table 5 (differences
in Persistence of 0.31 vs. 0.39, and differences in R2s of 0.45 vs. 0.54). In other words,
the relation between accrual quality (sresid) and earnings persistence is largely preserved
after controlling for the level of accruals.
Panel B in Table 6 contains the results for the relation between level of accruals and
earnings persistence after controlling for accrual quality, using the same portfolio derivation
method as in Panel A. First, we rank all observations into ten portfolios based on sresid,
then within each decile portfolio we rank observations into five subportfolios on level of
accruals, and finally we pool all subportfolios with the same rank together. At the end, we
have five portfolios that preserve most of the variation in level of accruals, while holding
sresid nearly constant across portfolios. Results reveal a slight decline in the variation of
the slope coefficient across portfolios; the difference between portfolios 1 and 5 is 0.22 in
Table 6 vs. 0.26 in Table 5. However, there is a substantial erosion in the range of R2, the
difference is 0.09 in Table 6 vs. 0.24 in Table 5. In addition, the differences in slope
coefficients and R2 across portfolios for Panel B in Table 6 are substantially smaller than
their counterparts for sresid in Panel A.
Summarizing, the combined evidence of Tables 5 and 6 suggests that our measure of
accrual quality captures variation in earnings persistence incremental to the variation ex-
plained by level of accruals. In fact, our results suggest that the relation between accrual
quality and earnings persistence is stronger than that between level of accruals and per-
sistence. However, relative to our measure of accrual quality, level of accruals is simpler
and easier to use, and requires no extensive identification and prediction process. Thus, the
comparison presented here is biased against the performance of the level of accruals, es-
pecially for settings that require ex ante measures of quality of earnings (e.g., real-time
trading strategies).

V. CONCLUSION
This study suggests a new approach to assessing accrual and earnings quality, based
on the intuition that accruals are temporary adjustments that resolve timing problems in the
underlying cash flows at the cost of making assumptions and estimates. Precise estimates
imply a good match between current accruals and past, present, and future cash flow re-
alizations, while imprecise or erroneous estimates reduce the beneficial role of accruals.
Following this intuition, we define accrual quality as the extent to which accruals map into
cash flow realizations. In the empirical domain, we operationalize this notion of accrual
quality as the standard deviation of the residuals from firm-specific regressions of working
capital accruals on last-year, current, and one-year-ahead cash flow from operations.
One important feature of our approach is that the notion of estimation errors includes
both intentional and unintentional errors. This distinction is important because most existing
research assumes that accrual and earnings quality is only affected by management intent
to manipulate, while such intent is unobservable, and likely idiosyncratic and sporadic. In
contrast, our approach reveals that accrual quality is likely to be systematically related to
observable and recurring firm characteristics like volatility of operations because higher
volatility is associated with higher incidence of unavoidable estimation errors.
A theoretical limitation of our approach is that it captures the extent to which accruals
map into the related cash flows but it provides little insight into the proper timing of these
54 The Accounting Review, 2002 Supplement

accruals with respect to the cash flows. For example, our approach cannot be used to decide
whether to expense or capitalize R&D. Limitations in the empirical portion of our analysis
include a restriction to working capital accruals and an assumption of no serial correlation
in the estimation errors.
We illustrate the usefulness of our approach in two contexts. First, we document that
observable firm characteristics can be used as instruments for accrual quality. Specifically,
we find that earnings volatility and the volatility of accruals are good proxies for our
measure of accrual and earnings quality. Second, we find a positive relation between accrual
quality and earnings persistence, which remains after controlling for the level of
accrual effect documented by Sloan (1996). This positive relation suggests there are im-
portant practical benefits from identifying and measuring accrual quality.
Our investigation of the interrelations between accrual quality, level of accruals, and
earnings persistence also suggests a reconciliation of the findings of Dechow (1994) and
Sloan (1996). Dechow (1994) finds that accruals improve earnings’ ability to measure per-
formance relative to cash flows. Sloan (1996) finds that the accrual portion is less persistent
than the cash flow portion of earnings, which suggests that firms with high levels of accruals
have low quality of earnings. Our reconciliation is based on the observation that a high
level of accruals signifies both earnings that are a greater improvement over underlying
cash flows, and low-quality earnings. The reason is that accruals are largest when the
underlying cash flows have the most timing and mismatching problems, so more accruals
signify greater improvement over the underlying cash flows. However, this benefit comes
at the cost of incurring estimation errors, and there will be a positive correlation between
levels of accruals and the magnitude of these estimation errors. Thus, everything else equal,
large accruals signify low quality of earnings, and less persistent earnings.

APPENDIX A
Examples of Cash Flows and Accruals Types, and the Resulting Estimation Errors
The following table outlines the recording conventions when cash is received or paid
before or after it is recognized in earnings:

Cash Flow Occurs at t⫹1 Cash Flow Occurs at t–1

(after it is recognized in earnings) (before it is recognized in earnings)
Case (1) Case (3)
Recognized in Cash Inflow: Collection Cash Inflow: Deferred Revenue
earnings at t as Accrual created and reversed: Accrual created and reversed:
revenue Asset Liability
e.g., Accounts Receivable e.g., Deferred Revenue
Case (2) Case (4)
Recognized in Cash Outflow: Payment Cash Outflow: Deferred Cost
earnings at t as an Accrual created and reversed: Accrual created and reversed: Asset
expense Liability e.g., Inventory, Prepaid Rent
e.g., Warranty Liability

The examples below show the link between accruals and the related cash flows, and whether
an estimation error will be detected by our model. We also include the notation introduced
in Section II.
Dechow and Dichev—The Quality of Accruals and Earnings 55

Case (1): Cash Receipt Follows Revenue Recognition

Record $100 of credit sales
t Dr Accounts Receivable (ACFt⫹1 / tO) 100
Cr Sales Revenue 100
No error situation
t⫹1 Dr Cash (CFt⫹1t) 100
Cr Accounts Receivable (ACFt⫹1 / tC) 100
Error situation: an unanticipated $30 is not collected
t⫹1 Dr Bad debt write-off (εt⫹1t) 30
Dr Cash (CFt⫹1t) 70
Cr Accounts Receivable (ACFt⫹1 / tC) 100
Cash increases by $70 at t⫹1 but accounts receivable at t increases by $100. Our model
will detect an error of $30.

Case (2): Cash Payment Follows Expense Recognition

Estimate $100 for warranty expense
t Dr Warranty Expense $100
Cr Warranty Liability (ACFt⫹1 / tO) $100
No error situation
t⫹1 Dr Warranty Liability (ACFt⫹1 / tC) 100
Cr Cash (CFt⫹1t) 100
Error situation: Warranty work amounts to $135
t⫹1 Dr Warranty Liability (ACFt⫹1 / tC) 100
Dr Warranty expense (εt⫹1t) 35
Cr Cash (CFt⫹1t) 135
Warranty liability increases by $100 at t, and is not equal to the cash flow of $135 at t⫹1.
Our model will detect a $35 estimation error.

Case (3): Cash Receipt Precedes Revenue Recognition

Receive $100 for a prepaid magazine subscription
t⫺1 Dr Cash (CFt–1t) 100
Cr Deferred Revenue (ACFt–1 / tO) 100
No error situation
t Dr Deferred Revenue (ACFt⫺1 / tC) 100
Cr Sales Revenue 100
Error situation: Customer demands a refund of $10 at time t
t Dr Deferred Revenue (ACFt–1 / tC) 100
Cr Cash 10
Cr Sales Revenue 90
56 The Accounting Review, 2002 Supplement

Our model will not detect an estimation error because the deferred revenue accrual and
cash flow at t–1 are equal in magnitude, and the deferred revenue reversal at t is equal to
the cash flow received at t–1. Note that accrual earnings contains no ‘‘error’’ since only
$90 is recognized in earnings, while cash basis revenue recognition would overstate revenue
(by $10) at t–1 and understate revenue by the same amount at t. This type of error will
affect our empirical version of the model because we cannot identify specific cash flows
relating to accruals, and our measure of cash flows at t will include the $10. This effect is
similar to that induced by including cash expenses CFtt, in our measure of cash flows CFOt
(see Appendix B).

Case (4): Cash Payment Precedes Expense Recognition

Pay $100 for inventory
t⫺1 Dr Inventory (ACFt⫺1 / tO) 100
Cr Cash (CFt⫺1t) 100
No error situation
t Dr Cost of goods sold 100
Cr Inventory (ACFt–1 / tC) 100
Error situation: Write down of $20
t Dr Write down 20
Dr Cost of goods sold 80
Cr Inventory (ACFt–1 / tC) 100

Our model will not detect an estimation error for the write-down because the inventory
accrual and cash flow at inception are equal, and the inventory reversal at t is equal to the
cash flow paid at t–1. Our model, therefore, provides no insights into the timing of expense
recognition.

APPENDIX B
Simulations Examining the Effect of Mismeasuring Cash Flows in Our Regressions
Our theoretical model of accruals listed in equation (4) is the following:

At ⫽ CFt⫺1t – (CFtt⫹1 ⫹ CFtt⫺1) ⫹ CFt⫹1t ⫹ εt⫹1t – εtt⫺1. (4)

Our empirical approximation of this model is:

⌬WCt ⫽ b0 ⫹ b1*CFOt–1 ⫹ b2*CFOt ⫹ b3*CFOt⫹1 ⫹ εt. (5)

The cash flow terms in Equation (5) capture the constructs in Equation (4) with measure-
ment error, where the measurement error is given by bolded components in parentheses
reported below:

CFOt⫺1 ⫽ CFt⫺1t ⫹ (CFt⫺1t⫺1 ⫹ CFt⫺1t⫺2)

CFOt ⫽ (CFtt⫹1 ⫹ CFtt⫺1) ⫹ (CFtt)
CFOt⫹1 ⫽ CFt⫹1t ⫹ (CFt⫹1t⫹1 ⫹ CFt⫹1t⫹2)
Dechow and Dichev—The Quality of Accruals and Earnings 57

For CFOt⫺1, the noise consists of cash flow received and recognized in earnings in the
same period (CFO t⫺1t⫺1) plus cash payments/collections (e.g., receivables) for transactions
recognized in earnings at t⫺2. For CFOt the noise stems from cash flows received and
recognized in earnings in the same period. For CFOt⫹1 the noise stems from cash flows
received and recognized in earnings in the same period (CFOt⫹1t⫹1) plus deferred cash flows
to be recognized in the following period (CFOt⫹1t⫹2). Thus, with respect to current accruals,
the noise component for CFOt is smaller than the noise component for CFOt⫺1 and CFOt⫹1.
We use simulations to examine the effect from this type of noise on the coefficients
and R2 in Equation (5). In these simulations there is no estimation error. All credit sales
are collected, all deferred revenues are recognized the following period, and there are no
write-downs of inventory. Thus, any change in the R2 and the coefficients is entirely attrib-
utable to mismeasurement of the desired cash flows.
Our model firm generates sales in three ways:
(1) Credit sales are made at time t and collected in t⫹1 (creates accounts receivable)
(2) Cash sales are made at time t
(3) Deferred revenue is recognized at time t for cash received at t-1 (creates deferred
revenue)
We use a random number generator to generate each type of sales for 200 periods.
Sales randomly range from 0 to 200. The firm starts business with an initial endowment of
inventory and shareholders’ equity. The inventory policy is to replace in period t inventory
sold in period t⫺1, and there are no price changes. For simplicity, we assume that the
inventory required for cash sales at t is purchased at t so that cash from operations and
earnings are the same for cash sales (i.e., no accruals are made for cash sales). The initial
endowment is set large enough such that the firm never has stock-outs of inventory. In this
firm:

Working capital ⫽ Accounts Receivable ⫹ Inventory ⫺ Deferred

Revenue;
Accruals ⫽ ⌬Accounts Receivable ⫹ ⌬Inventory ⫺
⌬Deferred Revenue;
Cash from operations in period t (CFOt) ⫽ Cash from credit sales made at t⫺1 (CFtt⫺1)
⫹ Cash for cash sales made at t (CFtt) ⫹ De-
ferred Revenue (CFtt⫹1) ⫺ Cash outlays for in-
ventory (CFtt); and
Earnings ⫽ Cash from operations ⫹ Accruals ⫽ Sales
⫺ Cost of Goods Sold.

Exhibit 1 provides regression results for a benchmark case, where all cash flows are
measured correctly, as in Equation (4). Since we have perfect knowledge of the related
cash flows, and there is no estimation error, we expect that b1 ⫽ 1, b2 ⫽ ⫺1, b3 ⫽ 1, and
the R2 is 100 percent. The simulation confirms this intuition.
Exhibit 2 documents the effect of using cash flow from operations (CFO, as defined
above) instead of the theoretically correct cash flow variables. Since the cash flow variables
are measured with error, we expect R2 and the coefficients to decrease in absolute magni-
tude. In addition, we expect that the larger measurement error for CFOt⫺1 and CFOt⫹1 will
force b1 and b3 more toward zero as compared to b2. These predictions are confirmed by
the results in Exhibit 2.
58 The Accounting Review, 2002 Supplement

EXHIBIT 1

Regression Results of Simulations Where There Is No Mismeasurement of Cash Flows at tⴚ1,

t, or tⴙ1

⌬WCt ⫽ b0 ⫹ b1(CFt⫺1t) ⫹ b2(CFt⫺1

t ⫹ CFt⫹1
t ) ⫹ b3(CFt⫹1 ) ⫹ ␧t
t

Intercept b1 b2 b3 Adjusted R2
Coefficient 0.00 1.00 ⫺1.00 1.00 1.00
(p-value) (⬍ 0.001) (⬍ 0.001) (⬍ 0.001) (⬍ 0.001)

EXHIBIT 2

Regression Results of Simulations Where Cash Flows from Operations (CFO) for tⴚ1, t, and
tⴙ1 Are Used in the Regressions

⌬WCt ⫽ b0 ⫹ b1*CFOt⫺1 ⫹ b2*CFOt ⫹ b3*CFOt⫹1 ⫹ ␧t

Intercept b1 b2 b3 Adjusted R2
Coefficient ⫺0.75 0.49 ⫺0.94 0.49 0.74

(p-value) (0.623) (⬍ 0.001) (⬍ 0.001) (⬍ 0.001)

Summarizing, our simulations confirm that using cash flow from operations (as in
Equation (5)) instead of the theoretically prescribed cash flow variables (as in Equation (4))
results in reduced R2 and regression coefficients biased toward zero, with less pronounced
effect for the coefficient of current CFO.

REFERENCES
Anthony, J., and K. Petroni. 1997. Accounting estimation disclosures and firm valuation in the prop-
erty-casualty insurance industry. Journal of Accounting, Auditing, and Finance 12 (Summer):
257–281.
Barth, M. E., D. P. Cram, and K. K. Nelson. 2001. Accruals and the prediction of future cash flows.
The Accounting Review 76 (January): 27–58.
Beaver, W. H., and M. McNichols. 1998. The characteristics and valuation of loss reserves of property
casualty insurers. Review of Accounting Studies: 73–95.
Collins, D. W., and P. Hribar. 2002. Errors in estimating accruals: Implications for empirical research.
Journal of Accounting Research 40: 105–134.
Dechow, P. M. 1994. Accounting earnings and cash flows as measures of firm performance: The role
of accounting accruals. Journal of Accounting and Economics 18 (July): 3–42.
———, S. P. Kothari, and R. L. Watts. 1998. The relation between earnings and cash flows. Journal
of Accounting and Economics 25 (May): 133–168.
Financial Accounting Standards Board (FASB). 1978. Objectives of Financial Reporting by Business
Enterprises. Statement of Accounting Concepts No. 1. Norwalk, CT: FASB.
———. 1987. The Statement of Cash Flows. Statement of Financial Accounting Standards No. 95.
Norwalk, CT: FASB.
Finger, C. A. 1994. The ability of earnings to predict future earnings and cash flow. Journal of
Accounting Research 32: 210–223.
Dechow and Dichev—The Quality of Accruals and Earnings 59

Healy, P. M., and J. M. Wahlen. 1999. A review of the earnings management literature and its
implications for standard setting. Accounting Horizons 13 (December): 365–383.
Liu, J., D. Nissim, and J. Thomas. 2002. Equity valuation using multiples. Journal of Accounting
Research 40 (March): 135–172.
Palepu, K. G., P. M. Healy, and V. L. Bernard. 2000. Business Analysis and Valuation. Cincinnati,
OH: South-Western College Publishing,
Penman, S. 2001. Financial Statement Analysis and Security Valuation. New York, NY: McGraw-
Hill, Irwin.
Petroni, K. R. 1992. Optimistic reporting in the property-casualty insurance industry. Journal of Ac-
counting and Economics 15 (December): 485–509.
Revsine, L., D. W. Collins, and W. B. Johnson. 2002. Financial Reporting and Analysis. Upper Saddle
River, NJ: Prentice Hall.
Sloan, R. G. 1996. Do stock prices fully reflect information in accruals and cash flows about future
earnings? The Accounting Review 71 (July): 289–315.