honours in accounting and financefrom the universityof
TRANSCRIPT
ii
CURRICULUM VITAE
Patricia Mary Dechow was born on July 11. 1964 in Missoula. Montana.
United States. In 1987 she received a Bachelor of Commerce degree with first class
honours in Accounting and Finance from the University of Western Australia. Dechow
entered the doctoral program at the William E. Simon Graduate School of Business
Administration at the University of Rochester in the fall of 1987. specializing in
accounting and finance. While at the Simon School, she was the recipient of an Olin
Foundation fellowship from the University of Rochester. In 1990, she received a
Masters of Science in Business Administration and was admitted to candidacy for the
Ph.D. degree. During her studies at Rochester. she taught introductory financial
accounting to undergraduate students.
111
ACKNOWLEDGEMENTS
I am grateful to the members of my dissertation committee, S. P. Kothari, Jay
Shanken, Jerry Zimmerman and especially Ross Watts (Chairman) for their advice and
encouragement throughout the project. I am also grateful to Richard Sloan for his
insightful comments and intuition on many of the issues raised in this study and for his
careful reading of many earlier drafts. The help of Philip Kearns in introducing me to
the literature on non-nested hypothesis testing and for many discussions on that topic
[in particular, Vuong (1989)] is also greatly appreciated. I also wish to thank Ray Ball
for his guidance and Ph.D. students and faculty at the Simon School who gave me
feedback on earlier drafts. Finally, I would like to thank my colleagues at the
University of Western Australia, Philip Brown, H. Y. Izan and Peter Kerby for
stimulating my interest in accounting research and encouraging me to continue my
studies.
iv
ABSTRACT
Accounting Earnings and Cash Flows as Measures of Firm Performance:
The Role of Accounting Accruals
by
Patricia Mary Dechow
This study investigates the role of accounting accruals. It identifies
circumstances under which accruals are predicted to improve earnings' ability to
measure firm performance, as reflected in stock returns. The importance of accruals is
shown to increase (i) the shorter the performance measurement interval, (ii) the greater
the volatility of the firm's working capital requirements and investment and fmancing
activities and (iii) the longer the firm's operating cycle. Under each of these
circumstances, cash flows are predicted to suffer more severely from timing and
mismatching problems that reduce their ability to reflect firm performance on a timely
basis. The results of empirical tests are consistent with these predictions.
Descriptive evidence is presented demonstrating how accruals improve
earnings' ability to reflect firm performance. Changes in cash flows are shown to
exhibit strong negative autocorrelation. This is consistent with part of the change in
cash flows being temporary and reversing over time. Consequently, cash flows suffer
v
from mismatching problems over short measurement intervals. Earnings are predicted
to have fewer mismatching problems than cash flows because accruals combine with
cash flows to mitigate these problems. Consistent with this prediction, accruals are
strongly negatively correlated with changes in cash flows. However, because a trade
off is made between reliability and timeliness, accruals do not eliminate all mismatching
problems. Consequently, earnings also exhibit negative autocorrelation, but to a far
lesser extent than cash flows.
Predictions are also made concerning components of accruals. If mitigating
mismatching problems in cash flows is the primary motivation for accruals, then ceteris
paribus, accruals that have existed for centuries are likely to mitigate the more severe
mismatching problems. This reasoning suggests that working capital accruals will be a
more important component for mitigating mismatching problems than long-term
operating accruals. In addition, special items are predicted to be less important because
their intended purpose is not mitigate mismatching problems in cash flows, but rather
reflect the reversal of past over or understatements of accruals. The empirical results
support these predictions. The study concludes that accruals play an important role in
enabling earnings to reflect firm performance on a timely basis.
TABLE OF CONTENTS
vi
Page
l. INTRODUCTION 1
2. REVIEW OF lHE LITERATURE 9
3. TESTABLE PREDICTIONS 14
3.1. The contracting roleof accruals 14
3.2. Measurement interval predictions 16
3.3. Cross-sectional predictions 20
3.3.l. Theeffectof themagnitude of accruals 20
3.3.2. Components of accruals 23
3.3.3. Theeffectof theoperating cycle 25
4. THE USEOF STOCK PRICEAS THE BENCHMARKMEASURE OF FIRM PERFORMANCE 27
4.l. Stockpriceas a benchmark 27
4.2. Stockmarketefficiency versus functional fixation 31
5. NON-NESTED MODEL SELECTION 34
5.1. Theencompassing approach to non-nested modelselection 34
5.1.1. TheJ-test 35
5.1.2. Theincremental approach 37
5.2. Using the Vuong likelihood ratio test to evaluate 38earnings versuscash flows
vii
Page
6. SAMPLE DESCRIPTION AND VARIABLE MEASUREMENT 43
6.1. Data 43
6.2. Variable definitions 44
6.3. Descriptive statistics 45
7. ANALYSIS OFMEASUREMENT INTERVAL PREDICTIONS 47
7.1. The effectof shortand longmeasurement intervals 47
7.2. Robustness tests 51
7.3. Mismatching problems withcash flows 55
7.3.1. Time-series properties 55
7.3.2. Additional evidence consistent withmismatching 57problems
8. ANALYSIS OFCROSS-SECTIONAL PREDICTIONS 60
8.1. The effectof themagnitude of all accruals 60
8.2. Analysis of the decline in thecoefficient on earnings 63acrossquintiles
8.3. Evaluation of accrual components 67
8.3.1. The effectof changes in working capital 67
8.3.2. The effectoflong-tenn operating accruals 69
8.3.3. The effectof special items 70
8.4. The effectof the operating cycle 72
9. SUMMARY ANDCONCLUSIONS 76
BffiLIOGRAPHY 79
APPENDIX 1 The relation between cash flows and earnings 90
APPENDIX 2 The effect of measurement error on the 93coefficient on earningsandcash flows
LIST OF TABLES
viii
~
Title Page
Table 1 Descriptive statistics of data for the quarterly, annual 96and four-year intervals.
Table 2 Tests comparing the association of earnings, the 97association of cash from operations and the associationof net cash flows with stock returns (adjusted formarket-wide movements) over varying measurementintervals.
Table 3 Results of the likelihood ratio test developed by Vuong 99(1989) for non-nested model selection. A significantpositive Z-statistic indicates that model 2 is rejected infavor of model 1.
Table 4 Comparison of earnings' and alternative cash flow 100measures' association with stock return over the annualinterval; 1,976 observations, 1987-1989.
Table 5 Firm-specific annual first-order autocorrelation 101coefficients and Pearson correlations for earnings, cashfrom operations, net cash flows and accrual measures.
Table 6 Tests comparing the association of earnings, cash from 103operations and net cash flows with stock returns afterincluding past and future realized values of thosevariables; 6,466 annual non-overlapping observations1964 to 1988.
Table 7 Test comparing the association of earnings and the 104association of net cash flows with stock returns acrossquintiles, where quintiles are formed based on theabsolute value of aggregate accruals. Quintile 5contains firms with the largest absolute value ofaggregate accruals.
Table 8 Examination of proxies for underlying determinants of 105earnings response coefficients across quintiles that areformed on the absolute value of aggregate accruals;annual observations 1960-1989.
IX
Title Page
Table 9 Test comparing the association of earnings and the 106association of cash from operations with stock returnsacross quintiles, where quinti1es are formed based onthe absolute value of the change in working capital.Quintile 5 contains firms with the largest absolutechanges in working capital.
Table 10 Test comparing the association of earnings and the 107association of cash from operations with stock returnsacross quintiles, where quintiles are formed based onthe absolute value of long-term operating accruals.Quintile 5 contains firms with the largest absolute long-term operating accruals.
Table 11 Tests comparing the association of earnings before tax 108and the association of earnings before tax and specialitems with stock returns over different measurementintervals.
Table 12 Descriptive statistics on firm level operating and trade 110cycles (measured in days) and firm level and industrylevel Pearson correlations between the absolute valueof the change in working capital and the length of theoperating and trade cycle; annual observations, 1960-1989.
Table 13 Pearson correlations (probabilities) between the R2s 111(from regressions performed by industry of stockreturns on cash from operations or stock returns onearnings) with the average length of the industryoperating and trade cycles; annual observations 1960-1989.
1
1 . Introduction
This study examines the production of accounting earnings. Earnings are the
summary measure of firm performance produced under the accrual basis of accounting.
The production of earnings is important since the earnings number is used as a
summary measure of firm performance by a wide range of users. For example, the
Conference Board's (1990) survey of top-executive compensation reveals that
earnings-based bonuses are almost universal for executives of larger U.S.
manufacturing corporations. In addition, earnings are used in debt covenants to
monitor a firm's ability to meet its future payments [Smith and Warner (1979)].
Detailed earnings information is also reported in the prospectuses of firms seeking to go
public. As a fmal example, when firms announce preliminary results of performance to
the stock market, they report earnings. The components of earnings are revealed later
[Wilson (1987)]. The objective of this study is to better understand the role of accruals
in producing earnings as one of the key outputs of the accounting process. The study
develops and tests explanations for why accrual accounting is used to produce an
internally generated measure of firm performance.
The view adopted in this study is that the accrual process evolved to overcome
problems with measuring firm performance over finite intervals when the firm is in
continuous operation (i.e., a going concern). Information asymmetries between the
2
firm's management and other contracting parties create a demand for a measure of firm
performance to be produced at regular intervals. The common interest of all contracting
parties is to determine the firm's ability to generate cash flows. However, over finite
intervals, reporting the firm's net cash receipts and disbursements (realized cash flows)
is not necessarily informative. This is because realized cash flows have timing and
mismatching problems that cause them to be a relatively noisy measure of firm
performance, To mitigate these problems, parties contracting with the firm require
management to use generally accepted accounting principles to alter the timing of cash
flow recognition in earnings.
The accounting principle that guides the timing of cash flow recognition is the
revenue recognition principle. The revenue recognition principle requires revenues to
be recognized when a firm has performed all, or a substantial portion, of services to be
provided and cash receipt is reasonably certain. Similarly, if cash is received in
advance, revenue is recognized only as the service is performed. Closely linked to the
revenue recognition principle is the matching principle. The matching principle requires
that cash outlays associated directly with revenues be expensed in the period in which
the finn recognizes the revenue. By having such principles, the accrual process is
designed to mitigate timing and mismatching problems inherent in cash flows and
produce a more timely measure of firm performance.' In this context, a timely measure
of finn performance is one that reflects the effects of events or actions on the finn's
cash generating ability in the period during which those events or actions occur.
1 Accounting principles are discussed in Paton and Littleton [(1940), pp. 69-72] and in mostaccounting texts books, e.g., Stickney, Weil and Davidson [(1990), pp. 86-87] and Kieso andWeygandt [(1989), pp. 88-89].
3
Therefore, the acceptable set of accruals is predicted to improve the ability of earnings
to measure firm performance relative to realized cash flows.2
Requiring management to make accrual adjustments, however, introduces a
new set of problems. Management typically have some discretion over the recognition
of accruals. This discretion can be used by management in two ways: they can use
their discretion to signal their private information about firm performance, or they can
use it to opportunistically manipulate earnings. Signaling is expected to improve the
ability of earnings to measure firm performance since management presumably have
superior information concerning their firm's cash generating ability [Holthausen and
Leftwich (1983), Watts and Zimmerman (1986), Holthausen (1990) and Healy and
Palepu (1991)]. Therefore, a credible signal will reduce information asymmetries and
result in more efficient contracting. However, to the extent that management use their
discretion to opportunistically manipulate earnings, it will become a less reliable
measure of firm performance.' Therefore, it is an empirical question as to whether the
net effect of management discretion is to improve or reduce earnings' ability to reflect
firm performance relative to realized cash flows.
The concern that management will use their informational advantage to
opportunistically manipulate accruals is consistent with the allowable set of accruals
being limited by a set of accounting conventions [Watts and Zimmerman (1986)].
2 The term accrual is used in a general sense and includes both accruals, e.g., accounts receivable anddeferrals, e.g., prepaid subscriptions.
3 Watts and Zimmerman (1986) and Schipper (1989) provides a review on earnings management,Healy (1985), DeAngelo (1988) and Clinch and Magliolo (1992) provide evidence of earningsmanagement 'Earnings management' is often discussed in the popular press, see for example "GM 3rdquarter net jumped as change in accounting erased operating loss," Wall Street Journal, October 10,1987, p. 2, "Will "FASBEE' pinch your bottom line," Fonune, December 19, 1988,93-108.
4
Accounting conventions, such as objectivity, verifiability and the use of the historical
cost valuation model, limit the flexibility of management to manipulate revenue and
expense recognition. In the absence of problems with information asymmetries, such
conventions would be dysfunctional since their effect is to reduce the timeliness of
earnings. However, since management manipulation is not always detectable (at least
over short measurement intervals), contracting parties desire a performance measure
that is reliable (and verifiable by auditors) so that there are bounds on the manipulation
that can occur. The accrual process is therefore the result of a trade-off between
timeliness (or relevance) and reliability.'
This study designs empirical tests to compare earnings with both net cash flow
and cash from operations. These comparisons are made because the accrual process
can be thought of as starting with cash flows and then applying accrual adjustments to
produce earnings. By comparing earnings to realized cash flows in various
circumstances, insights can be obtained into the importance of the accrual process. Net
cash flow is the change in the firm's cash balance between two points in time. This
measure is chosen because it excludes all accrual adjustments and can therefore be
thought of as the most 'primitive' cash-based measure of firm performance. Cash from
operations is chosen as the second cash-based measure because it is often prescribed as
a competing measure of firm performance [e.g., Reimann (1990) and Skala (1991)].
This measure has also been used in previous research [e.g., Rayburn (1986), Wilson
(1986), Bowen, Burgstahler and Daley (1987), and Bernard and Stober (1989)]. Cash
from operations excludes accruals relating to the finn's operating activities. However,
4 BaH (1989) and Watts and Zimmerman [(1986) p. 206] discuss this trade off. The FASB alsoidentifies relevance and reliability as qualities that increase the value of accounting earnings inStatement of Financial Accounting Concepts No.2, paragraph 90.
5
as with earnings, cash from operations does not contain the direct cash flow effects of
the firm's capital investment and financing activities.
In order to compare the relative ability of earnings and cash flows to measure
finn performance, it is necessary to obtain a benchmark measure of finn performance,
Stock returns (net of market-wide movements) are used as the benchmark because they
are assumed to reflect a wider information set than either earnings or realized cash
flows concerning the finn's ability to generate future cash flows. A sufftcient condition
for this assumption to hold is that the stock market "correctly" evaluates the information
in earnings and cash flows so that it is "fully reflected" in stock price [Fama (1976,
133)].5 Market-wide movements are removed from stock returns to improve the power
of the empirical tests. Previous research suggests that market-wide movements are
dominated by macro-economic factors that have little impact on earnings or realized
cash flows [Sloan (1992a)]. In addition, if market-wide movements are 'noise' from
the perspective of evaluating firm-specific performance, then their removal will produce
a 'stronger' signal of the measure desired in contracts [Holmstrom (1982)]. Thus, the
measure (earnings or cash flows) that is more highly associated with market-adjusted
stock returns is judged to be the more useful measure of firm performance.
This study hypothesizes that accruals mitigate mismatching and timing problems
inherent in cash flows so as to produce a more useful summary measure of firm
performance. A mismatching problem exists when a cash outlay is made in one
measurement interval but the cash inflow associated with that outlay is received in
5 The assumption that the stock market is efficient with respect to publicly available information isimplicit in most capital market research performed in accounting. beginning with Ball and Brown(1968).
6
another measurement interval. For example, assume a firm pays cash for inventory and
then sells the inventory on account within one measurement interval. Cash flows will
reflect the cost but not the benefit associated with the sale. However, if the cash receipt
is reasonably certain, then earnings will reflect the net benefit, A timingproblem exists
when a firm performs a service in one measurement interval but the net cash receipt
(although reasonably certain) is received in another measurement interval. In the
example discussed above, assume that the firm both purchases and sells the inventory
on account so that there are no cash flow effects. Again, earnings will reflect the net
benefit of the sale. Whereas, cash flows will reflect the net benefit only in the next
measurement interval when cash is paid to the supplier and cash is received from the
customer. Cash flows do not suffer from mismatching problems in this circumstance
but they are an untimely measure of firm performance. Timing and mismatching
problems are predicted to become more severe, the shorter the performance
measurement interval.
Consistent with the prediction that accruals mitigate the timing and mismatching
problems inherent in cash flows, realized cash flows are demonstrated to be a poor
measure of firm performance relative to earnings. An examination of cash flows time
series' properties indicates that changes in cash flows exhibit strong negative
autocorrelation. This is consistent with part of the change in cash flows being
temporary and reversing over time. Consequently t cash flows suffer from
mismatching problems over shorter measurement intervals. Earnings are predicted to
have fewer mismatching problems than cash flows because accruals combine with cash
flows to mitigate these problems. Consistent with this prediction, accruals are strongly
negatively correlated with changes in cash flows, while earnings itself, displays little
negative autocorrelation.
7
The study builds on the results of Easton, Harris and Ohlson (1992) who show
that the association of earnings with stock returns improves over longer measurement
intervals. The results presented here indicate that both cash flows and earnings
improve as measures of firm performance over longer measurement intervals, but that
cash flows improve relatively more than earnings. This result suggests that because
accrual accounting trades-off timeliness and reliability, accruals do not eliminate all
timing and mismatching problems in cash flows. Therefore, earnings also suffers from
timing and mismatching problems over short measurement intervals. However,
earnings suffer relatively less than cash flows because they include accruals.
The study also makes cross-sectional predictions concerning the importance of
accruals. The study predicts that for firms in 'steady state' (i.e., firms with cash
requirements for working capital, investments and financing that are relatively stable),
cash flows will have few mismatching problems and will be a relatively useful measure
of firm performance. However, for firms operating in volatile environments with large
fluctuations in their working capital requirements and their investment and financing
activities, cash flows will have more severe mismatching problems. Thus, cash flows'
ability to reflect firm performance will decline as the volatility of firms' working capital
requirements and investment and financing activities increases. Accruals are predicted
to mitigate mismatching problems in cash flows so that earnings' ability to reflect firm
performance is expected to vary less on these dimensions. The results are consistent
with this prediction. This finding is then used to obtain additional cross-sectional
predictions. The volatility of firms' working capital requirements is predicted to
increase with the length of their operating cycle. Thus, accruals are predicted to be
more important for firms in industries with long operating cycles. The results are
consistent with this prediction.
8
Finally, the study investigates specific accrual components. Short-term
working capital accruals are predicted to be relatively more important than long-term
operating accruals for mitigating the timing and mismatching problems inherent in cash
flows. Working capital accruals include the earliest developed accruals [Littleton
(1966)]. If mitigating mismatching problems in cash flows is the primary motivation
for accruals, then ceteris paribus, these accruals are predicted to mitigate the more
severe mismatching problems. In addition, special items (including write-downs and
write-offs of receivables, equipment, etc.) are investigated. Special items reflect the
cumulative effect of past over or understatement of accruals. These adjustments are not
made to mitigate mismatching problems in contemporaneous cash flows, rather they are
included in earnings primarily to make management accountable for past misstatements
of accruals and to maintain 'clean surplus' (e.g., Paton and Littleton [(1940), pp. 98
102] and APB Opinion No. 30). Therefore, special items are not expected to improve
earnings' ability to reflect firm performance. The results are consistent with these
predictions.
The next chapter summarizes previous empirical research of relevance to this
study and describes the contribution of this study. Chapter 3 develops the research
hypotheses. Chapter 4 evaluates the role of price as the benchmark measure of firm
performance. Chapter 5 introduces and describes a recently developed statistical test
that is used to evaluate the research hypotheses. Chapter 6 discusses sample selection
issues and variable measurement. Chapter 7 and 8 analyze the empirical results.
Finally, chapter 9 summarizes the study and provides its conclusions.
9
2 • Review of the Literature
The issue of whether earnings or cash flows is a more useful measure of firm
performance is at the heart of determining if the accruals made by accountants are
useful. Thus, when Ball and Brown (1968) first examine the usefulness of earnings by
evaluating its association with stock returns, they also addressed this concern. Ball and
Brown find that earnings are more successful than their cash flow measure in predicting
the sign of annual stock returns. Similar results are documented by Beaver and Dukes
(1972). Both studies however, use crude cash flows proxies (earnings before
depreciation). Therefore, one explanation for their findings is that the cash flow proxy
is inappropriate. For example, Bowen Burgstahler and Daley (1986) show that
earnings before depreciation is more highly correlated with earnings than other more
finely defined measures of cash flows. Their evidence also indicates that for most of
the cash flow measures that they investigate, past realizations of the cash flow measures
predict future realizations of the cash flow measures better than do earnings.
Recent research has adopted a slightly different perspective from the earlier
studies to investigate the usefulness of accrual or cash flow information. Following
Patell and Kaplan (1977), recent work has focused on determining the components of
earnings that provide incremental information. This is referred to here as the
10
incremental approach. The question of interest is to determine which financial variables
the market finds informative when assessing firms' future cash flows. Various models
have been used to obtain abnormal returns and different techniques have also been used
to obtain the unexpected components. A regression is then performed of abnormal
stock returns on the various unexpected components of earnings. The interpretation of
the results is that if the coefficients on the unexpected components are significant, then
the market finds these components incrementally informative.
Patell and Kaplan (1977) investigate whether total funds from operations has
incremental information over earnings. They find that this surrogate for cash flows
does not provide any incremental information beyond earnings. Rayburn (1986)
extends Patell and Kaplan by determining whether a more refined measure of cash
flows provides incremental information, She decomposes earnings into cash from
operations, current and noncurrent accruals. Rayburn finds that current accruals and
cash flows have about equal incremental information content but fmds weaker evidence
for noncurrent accruals. Bowen, Burgstahler and Daley (1987) build on the results of
Patell and Kaplan by showing that their more refined measure of cash flows has
incremental information beyond earnings. They include both earnings and cash from
operations in a regression and show that both cash flows and earnings are significant.
Similar results are also reported by Schaefer and Kennelley (1986). Jennings (1990)
provides further analysis on the results of Rayburn and Bowen, Burgstahler and Daley.
He suggests that Rayburn's results are consistent with cash flows and current accruals
having similar incremental information content but that Bowen, Burgstahler and
Daley's results are consistent with these components being valued differently. Finally,
Livnat and Zarowin (1990) analyze whether cash flows relating to investing, fmancing
and operations have incremental information content over earnings. Their results
11
suggest that decomposing earnings into these finer partitionings provides incremental
information over earnings.
Wilson (1987) addresses the question of incremental information content in a
novel manner. The contribution of Wilson (1987) is to provide a test that specifically
controls for the information content of earnings. This is important since one concern
with Patell and Kaplan (1977) is that when the component of interest is highly
correlated with earnings (as is total funds from operations), it is difficult to disentangle
the information content of the component from the information content of earnings.
Wilson observes that the Wall Street Journal usually publishes the earnings number
prior to the release of the financial statements. This means that the components of
earnings are publicly available only when the financial statements are released.
Therefore, by analyzing the returns around the date when the financial statements are
released, he can control for the information in earnings and focus on the incremental
information content of the components.
Wilson's evidence is consistent with there being incremental information in the
cash from operations (and the current accrual) component. Wilson (1986) using a
different method of aggregating returns and assessing information content, provides
evidence consistent with the total accrual component being incrementally informative
beyond the cash component. Bernard and Stober (1989) perform similar tests to
Wilson using a more comprehensive data set. However, their evidence suggests that
Wilson's findings cannot be generalized. They find little evidence of incremental
information content for cash flows or accruals apart from the years that Wilson
examines. They also test whether the valuation implications of cash flows vary with
the economy or with a model that links the mix of currerit accruals to future sales.
However, no evidence is found that supports either prediction.
12
When summarizing this area of the literature, both Bernard (1991) and Neill,
Schaefer, Bahnson and Bradbury (1991) conclude that the results are mixed. It is not
clear which components of earnings are relatively more important for revising stock
prices. Studies using annual holding periods and refined cash flow measures generally
find that cash flows have incremental information content. However, studies
employing daily holding periods that have less problems with collinearity have obtained
inconsistent results. The conclusion generally drawn from this line of research is that
the incremental value of cash flows or accruals may depend on the specific contexts
analyzed.
The perspective adopted in this study differs from evaluating the incremental
information of accruals and cash flows and is similar to that of earlier studies by Ball
and Brown (1968) and Beaver and Dukes (1972). That is, the objective of this study is
to determine which measure, earnings or cash flows, is a more useful summary
measure of finn performance, This perspective is adopted because it is hypothesized
that the role of accruals is to mitigate timing and mismatching problems in cash flows in
order to produce earnings as an alternative, more useful summary measure of finn
performance. Evaluating the incremental content of cash flows or accruals does not
provide a direct test of which measure, earnings or cash flows, is a more useful
summary measure of finn performance. This is because observing that accruals and
cash flows have incremental information when included separately in a regression does
not tell us which of the two summary measures, earnings or cash flows is superior. In
summary, this study builds on previous research by (i) demonstrating the role of
accruals in mitigating temporary mismatching problems in cash flows, (ii) formally
establishing that earnings is a superior summary measure of firm performance and (iii)
13
identifying the determinant of mismatching problems in cash flows, thus highlighting
the circumstances under which accruals playa more important role.
14
3. Testable Predictions
3.1. The contracting role of accruals
The existence of information asymmetries between firms' managers and outside
parties contracting with the finn creates a demand for an internally generated measure of
firm performance, This measure can be used to evaluate management and as a source
of information to investors and creditors on the firm's cash generating ability. The
problem faced by contracting parties is that although management is the most informed
party to report on the firm's performance, they are also evaluated and rewarded based
on the finn's performance. Therefore, in the absence of objective procedures to
determine performance, external parties have difficulty assessing the reliability of
signals produced by management. On the one hand, contracting parties could demand
that managers report realized cash flows. These can be objectively measured but are
influenced by the timing of cash receipts and outlays. For example, management
would be penalized for purchasing inventory (above beginning inventory levels) even if
this was a positive net present value decision." On the other hand, management could
6 In the analysis undertaken here it is assumed that if the only numbers reported were cash flows, thencash flows would not change. However, management can also manipulate the timing and recognition
15
attempt to determine the change in the net present value of the firm's expected future
cash flows in each reporting period. The use of net present values however, could
provide management with so much reporting flexibility that any signal produced would
be difficult toverify and would result in an unreliable measure of finn performance.
The accrual process can be viewed as trading-off these two problems when
producing earnings as a summary measure of firm performance. The accrual process
provides rules on the recognition of cash flows so that earnings will be a more timely
measure of firm performance than realized cash flows. However, accruals are also
required to be objective and verifiable. For example, expenditures can only be
capitalized when there is objective and verifiable evidence that the expected cash flows
will be realized. Similarly, cash received in advance is not recognized as revenue until
there is verifiable evidence that the firm has performed the service. Requiring
objectivity and verifiability limits management's discretion in reporting a measure of the
firm's cash generating ability. This will reduce the timeliness of reported earnings in
circumstances where management is more informed and wishes to signal private
information. However, it will also reduce the ability of management to falsely signal.
Ifexisting accruals are the outcome of efficient contracting, then the use of accruals is
expected, on average, to improve the ability of earnings to measure firm performance
relative to cash flows. Alternatively, if the dominant effect of accruals is to provide
management with flexibility to manipulate earnings, then cash flows will provide a
relatively more useful summary measure of firm performance over short measurement
intervals.
or cash flows. The economic implications of manipulating cash flows could be greater than that ofmanipulating accruals andis an additional explanation for theexistence of accruals [seeBall (1989)].
16
This study compares the ability of earnings to reflect finn performance relative
to the ability of net cash flows and cash from operations. Net cash flows measure the
change in the finn's cash balance over the period. This measure will fluctuate with
cash inflows and outflows associated with the finn's investment and financing activities
as well as the finn's operating activities. No accrual adjustments are made to net cash
flows. Therefore, 'aggregate accruals' refers to the difference between earnings and
net cash flows. Cash from operations reflects the net cash flows generated by the
finn's operating activities. This measure includes accruals that are 'long-term' in nature
that control for timing and mismatching problems associated with the firm's investment
and financing activities." For example, cash from operations excludes the cash flow
effects associated with purchasing capital investments, paying dividends or issuing new
stock. It is, however, sensitive to decisions by management to capitalize rather than
expense cash outlays for assets. Earnings include additional accruals beyond those
contained in cash from operations. In particular, it includes accruals that are short-term
in nature that control for temporary fluctuations in firms' working capital requirements.
Appendix 1 provides a detailed break-down of the accruals made to net cash flows to
obtain earnings.
3.2. Measurement interval predictions
If the accrual process evolved so as to produce a more timely measure of finn
performance than realized cash flows, then net cash flows are, on average, expected to
be a relatively poor measure of finn performance. Net cash flows have no accrual
adjustments and will suffer severely from timing and mismatching problems. Cash
7 Accruals relating to thefirm's investments andfinancing areconsidered long-term since they are notexpected to becompletely reversed for at least oneyear.
17
from operations are also expected to be a poor measure of firm performance because
they suffer from timing and mismatching problems associated with changes in firms'
working capital requirements. Earnings are expected to be, on average, a more useful
measure of firm performance than either cash flow measure. Earnings contain accruals
that mitigate the mismatching problems associated with firms' investment and fmancing
cash flows as well as working capital accruals that mitigate temporary fluctuations in
operating cash flows.
Stock returns are assumed to encompass the information about firm
performance provided in cash flows and earnings. Therefore, the measure (either cash
flows or earnings) that explains more of the variation in stock returns is assumed to be
a more useful measure of firm performance over the period. This generates the first
prediction:
Hypothesis 1: Overshortmeasurement intervals, there is a stronger association
between contemporaneous earnings and stock returns than
between either contemporaneous cash from operations and stock
returns or contemporaneous net cash flows and stockreturns.
The alternative view is that the accrual basis of accounting has gone 'too far.'
That is, management has been given so much flexibility to manipulate accruals that any
signal produced is primarily noise. Under this view, if a performance measure is
desired, then the measure of choice is cash flows. This view is often expressed in the
popular press, for example, Skala (Chemical Week, May 8,1991) states:
"Many financial analysts regard operating cash flow as a better gauge ofcorporate financial performance than net income, since it is less subject todistortion from differing accounting practices."
As a second example, Stem (Planning Review, JanlFeb 1988) states:
18
.. Think cash and risk - forget earnings per share."
As a final example, Wilson (1987) provides the following quote from a Chase
Financial brochure:
"... It is cash-in minus cash-out, or Free Cash Flow, that counts - notaccounting earnings or their growth."
Thus, it is an empirical question whether the first-order effect of accruals is to
produce a more timely summary measure of finn performance (as predicted under
efficient contracting) or a more noisy measure of finn performance (as predicted under
management manipulation).
Additional implications of the relation between earnings and cash flows can be
obtained by considering the following simplified example of a finn with only one
accrual, accounts receivable. [For a more comprehensive model of accruals the reader
is referred to Jones (1991)].
Let Ct = cash collected in accounting period 1.
St = revenues generated from sales made during accounting period t.
q> =proportion of accounting period t-l sales for which cash is not collected
until the next accounting period t. q> is assumed constant for each
accounting period and all cash is collected by t.
\
Now C] =(1- q»St + q>St-I· (3-1)
Thus, earnings will differ from cash flows in each period to the extent that credit sales
are excluded from current performance under the cash basis of accounting. Consider
doubling the time interval over which performance is measured. Under these
19
circumstances, the importance of accounting accruals diminishes. Originally, the
proportion of uncollected sales equaled cpo With a doubling of the time interval, total
revenues are (St + St- I). Total accruals at the end of the period are still cpSt. So the
proportion of uncollected sales cp' is:
, cpStcp =St + Sj; I' (3-2)
As long as St-l > 0, then cp > cp' and, ceteris paribus, increasing the time interval will
always result in accruals constituting a smaller proportion of the total revenues
generated over the period. Equation (3-2) reveals that over longer intervals earnings
and realized cash flows are expected to converge as measures of firm performance.
Hypothesis 2 builds on hypothesis 1 by predicting the direction of convergence:
Hypothesis 2: The contemporaneous association of realized cash flows with
stock returns improves relative to the contemporaneous
association of earnings with stock returns as the measurement
interval is increased.
The alternative hypothesis is that due to the manipulation of accruals, earnings
are a noisy measure of firm performance over short intervals. Under the alternative
hypothesis, earnings will improve relative to cash flows over longer measurement
intervals. Note however, that the ability of earnings to reflect firm performance is also
expected to improve over longer measurement intervals [see, Easton, Harris and
Ohlson (1992)]. This is because accruals do not completely mitigate all short-term
timing and mismatching problems in realized cash flows. An empirical investigation of
hypothesis 2 can provide insights into the economic importance of accruals. Evidence
that cash flows' ability to measure firm performance is poor over intervals that are
20
commonly used to report finn performance (e.g., one year) and converge to that of
earnings only over long measurement intervals, would confirm the economic
significance of accruals.
3.3. Cross-sectional predictions
3.3.1. The effect of the magnitude of accruals
Equation (3-1) can also provide insights into determinants of cross-sectional
variation in the usefulness of accruals. Recall that in equation (3-1):
Ct =(1- lp)St + <PSt-I.
If a steady state firm is defined as one that is neither growing nor declining
(i.e., neither increasing nor reducing sales), then this implies that,
St =St-l.
Substituting St for St-l in equation (3-1):
and so Ct =St·
Therefore, in a steady state firm there will be no difference between the
numbers reported under a cash system or an accrual system. Thus, this is not a
situation where accruals are important Consider instead, the case of a second firm that
has identical sales in period t-l to the steady state firm but has an increase (or decrease)
in sales in period 1. In this case St :/: St-l and:
Ct =St - <p(St - St-l),
21
and so
(3-3)
where 65t =(5t - 5t-l)·
Equation (3-3) reveals that the magnitude of the difference between revenues
and cash flows for any period will be greater (i) the larger cp, i.e., the proportion of
sales on credit; and (ii) the larger the magnitude of the change in revenues (65t).
Alternatively stated, the difference between revenues and cash flows over the period is
increasing in the absolute magnitude of the change in the balance of accounts
receivable, cp65t ' over the period.
This analysis highlights where accruals are expected to play an important role in
measuring firm performance. The accrual process is most important for firms that have
had large changes in the net balance of their non-cash accounts. Consider for example,
a ship building firm that obtains a lucrative construction contract. The construction
takes several accounting periods and the payment by the customer occurs on completion
of the contract. Under generally accepted accounting principles, revenue recognition
for this contract is based on an engineer's estimate of the degree of completion. If cash
collection is reasonably certain, then the actual timing of the cash collection is not
relevant for reporting purposes. Cash flows for the firm could easily be negative in the
early periods because of purchases required for the construction contract. Revenues on
the other hand (through an increase in accounts receivable) are positive, and the
application of the matching principle will lead to positive earnings. Thus, earnings will
better reflect the contract's value and indicate that the firm has performed well in each of
the periods. This highlights an important feature of the accrual process. If accruals
reduce timing and mismatching problems, then earnings are expected to reflect
22
relatively more value-relevant events when earnings and cash flows differ by the
greatest magnitude.
While the analysis focuses on accounts receivable, it is readily generalizable to
other accruals (changes in other non-cash accounts). When the net change in all non
cash accounts is large in magnitude (either positive or negative), earnings will more
closely reflect firm performance than realized cash flows. Equation (3-3) suggests that
cash flows are not a poor measure of firm performance for firms that are in 'steady
state'. However, when firms undertake new investment and financing activities or
experience large changes in their working capital requirements, (when ASt is large in
absolute magnitude) cash flows are expected to be a relatively poor measure of firm
performance. Under such circumstances, cash flows suffer from timing and
mismatching problems and are less able to reflect firm performance. Accruals are
predicted to reduce these problems in earnings. This leads to the following hypothesis:
Hypothesis 3 The larger theabsolute magnitude ofaggregate accruals madeby
a firm, the lower the contemporaneous association betweennet
cash flows and stock returns relative to the association of
earnings andstockreturns.
Hypothesis 3 predicts that as aggregate accruals increase in magnitude, net cash
flows will suffer more greatly from timing and mismatching problems. Thus, in firms
with relatively small changes in aggregate accruals, net cash flows will have few
mismatching problems and will be a useful measure of firm performance. Whereas,
when aggregate accruals are relatively large in magnitude, net cash flows will have
more timing and mismatching problems and will be a poor measure of firm
performance. As earnings includes accruals that mitigate these timing and mismatching
23
problems, earnings association with stock returns is not expected to decline as much
with increases in the magnitude of accruals. Note that by considering the magnitude of
accruals irrespective of the sign, a more powerful test of the relative importance of the
accrual process is obtained.
3.3.2. Components of accruals
This study investigates several components of accruals to provide insights into
which accruals are relatively more important for reflecting firm performance. One
subset of accruals investigated is short-term working capital accruals. Watts (1977)
suggests that accruals evolving in an unregulated economy are more likely to be the
outcome of efficient contracting. Many of the components of working capital (for
example, accounts receivable, inventory and accounts payable) have existed for
centuries [see Littleton (1966)]. Therefore, ceteris paribus, if accruals evolved to
mitigate mismatching problems in realized cash flows, then working capital accruals are
likely to mitigate the more acute mismatching problems. Therefore, working capital
accruals are predicted to be important for reflecting firm performance over short
measurement intervals.
Hypothesis 4(a): The larger the absolute magnitude of the change in working
capital, the lower thecontemporaneous association between cash
from operations and stock returns relative to the association of
earnings andstock returns.
In contrast, long-term operating accruals (such as depreciation), are predicted to
be relatively less important for two reasons. First, long-term operating accruals are
more recent accrual innovations and therefore, ceteris paribus, are likely to mitigate less
severe mismatching problems in cash flows. Second, many of these accruals were
24
introduced following the regulation of accounting practices and so the motivation for
their introduction can be influenced by other factors [see Watts (1977) and Watts and
Zimmerman (1979)].8 Therefore, the association of cash from operations with stock
returns is not expected to be a declining function of long-term operating accruals. Cash
from operations is used as the cash-based measure of firm performance because the
objective of the test is to focus on working capital and long-term operating accruals (the
accruals made to cash from operations to obtain earnings).
The final component of accruals investigated is special items. Special items
represent the cumulative effect of previous under or overstatement of accruals. Prior to
APB No. 30 special items were predominantly classified as 'extraordinary' since they
are non recurring in nature and are not expected to be relevant for measuring current
performance [e.g., Nichols (1974) and Barnea, Ronen and Sadan (1975)]. However,
management had discretion to determine what special items would be classified as
'extraordinary' and therefore excluded from earnings (from continuing operations).
This discretion was perceived to be used by management to manipulate earnings. For
example at the time APB No. 30 was released the Wall StreetJournal notes than?
"critics ...contend the proliferation of special items has confused investors andenabled corporate managements to gloss over bad business decisions. Toooften, they charge, routine losses are allowed to pile up. Then, when finallywritten off as an "extraordinary" loss, the write-off is inflated, sweeping all the"garbage" off the books and setting up a healthy profit rebound."
8 Accruals introduced following regulation willnotnecessarily reduceearnings' ability to measure fumperformance. Manytransactions that resultin accruals did not occurprior to regulation. Therefore, thesameprocedures mayhaveevolved in an unregulated environment and beconsistentwith theobjectivesof contracting parties.
9 See n Accounting Panel seen curbing the use of special items," the Wall Street Journal, March 15,1973, p. 2.
25
APB No. 30 reduced the discretion of management by requiring that 'special
items' flow through income from continuing operations so that 'clean surplus' is
maintained and cumulative earnings is measured objectively.l? Therefore, special items
are an accrual adjustment that is predicted to reduce earnings' ability to reflect firm
performance.
Hypothesis 4(b): There is a stronger association between earnings (before special
items) and stock returns than between earnings (after special
items) andstockreturns overshortmeasurement intervals.
3. 3.3. The effect of the operating cycle
The final hypothesis predicts the type of firms in which the volatility of accruals
will be large and hence realized cash flows will be a poor measure of firm performance.
Equation (3-3) reveals that the change in accounts receivable, <pASt, is composed of two
components <p, and ASt. Therefore, the magnitude of accruals is larger, the greater the
change in the level of sales ASt, and the larger the proportion of sales on credit, cpo
Generalizing from sales, St can be proxied by the level of operating activity and cp can
be proxied by the length of the operating cycle. The operating cycle measures the
average time elapsing between the disbursement of cash to produce a product and the
receipt of cash from the sale of the product. Firms with longer operating cycles are
expected to have larger working capital requirements for a given level of operating
10 Prior to APB No.9 (issued in 1966) the primary concern had been whether to disclose extraordinaryitems in retained earnings or in the net income figure. APB No.9 defined extraordinary items as 'of acharacter significantly different from the typical or customary business activities of the entity' and'which would not be expected to recur frequently.' APB No. 30 (issued in 1973) tightened thisdefmition by requiring items to be classified as extraordinary only if they were both unusual in natureand infrequent in occurrence.
•26
activity. Therefore. in firms with longer operating cycle. a given change in the level of
operating activity [t\Stl. is expected to translate into a larger change in the required level
of working capital [cpt\Stl. Thus. the length of the operating cycle is predicted to be an
underlying determinant of the volatility of working capital. Cash from operations
excludes accruals relating to the finn's operating activities. Hence the ability of cash
from operations to measure firm performance is expected to decline as the length of the
operating cycle increases. II
This leads to the following hypothesis:
Hypothesis 5: The longer afirm's operating cycle, the more variable thefirm's
working capital requirements and the lower the
contemporaneous association between cashfrom operations and
stockreturns.
Short-term working capital accruals are hypothesized to reduce the timing and
mismatching problems inherent in cash from operations. Hence, the ability of earnings
to reflect finn performance is not expected to be as sensitive to the length of the
operating cycle.
II The analysis focuses on cash from operations rather than net cash flows because cash fromoperations is directly affected by the firm's operating activities (and hence the operating cycle). Netcash flows is only indirectly affected by the firms operating cycle since it will also fluctuate with theinvestment and fmancing activities undertaken by thefirm,
(4-1)
27
4. The Use of Stock Price as the BenchmarkMeasure of Firm Performance
4.1. Stock price as a benchmark
The question of interest in this study is: given an internally generated measure
of finn performance is desired for contracting, which measure, earnings or cash flows
better reflects the firm's performance over a given measurement interval? This study
assumes that stock markets are efficient in the sense that stock prices encompass the
information provided in cash flows, earnings and other sources of information about
firms' expected stream of future cash flows. Therefore, stock price performance is
used as the benchmark measure of firm performance, because it is assumed to reflect in
a timely manner, the net cash flow effects of current events or actions taken by the firm.
In the empirical analysis each performance measure is scaled by beginning of
period price. This reduces problems with heteroskedasticity [see Christie (1986)].
Thus,
p. M·~ =a+~_l_t.Pit-l Pit-l
where Pit is the price of firm i at time t, Mit is either earnings-per-share, net cash flows-
per-share or cash from operations per-share for firm i measured from t-l to 1.
28
Subtracting Pp~t-l from both sides results in a regression of stock returns (R) onIt-l
earnings or cash flows-per-share scaled by price:
M·Rot = (n-I) + A~.
I p Pit-l (4-2)
This study compares each measures' association with stock returns (the
benchmark measure of firm performance) and judges the one with the highest
association to be a more useful measure of firm performance. In equation (4-2), stock
returns are regressed on the 'level' of earnings or cash flows per-share.P Thus, if
either earnings or cash flows contain temporary components that are not reflected in
stock returns, then these are not removed when assessing the ability of each measure to
reflect firm performance. Note that some portion of this temporary component may be
predictable based on the past time-series of each variable.P However, these
predictable components are not removed because one of the hypothesized roles of
accruals in this study is to remove both the predictable and non-predictable temporary
components in cash flows. Therefore, in this context, attempting to eliminate
predictable temporary components in cash flows would ignore the fact that accruals
already incorporate this adjustment.
In the empirical analysis stock returns net of market-wide movements are used
as the dependent variable. The reason for removing market-wide movements is to
12 Existing research has used the 'level' specification of earnings [Easton and Harris (1991), Easton,Harris and Ohlson (1992) and Kothari and Sloan (l992)]. Ohlson (1991) and Kothari (1992)analytically demonstrate that the earnings 'level'specification improves the explanatory powerof theregression overtherust difference specification.
13 It may also be possible to predict temporary components using other accounting data, see, forexample, Lev andThiagarajan (1991).
29
obtain a 'stronger' signal of the performance measure desired by contracting parties. In
this study it is assumed that market-wide movements are predominantly driven by
macro-economic factors that are beyond management's control (such as changes in
discount rates). Therefore, they can be considered 'noise' from the perspective of
evaluating the firm's relative performance [see Holmstrom (1982)]. Sloan (1992a)
provides evidence that market-wide movements are not reflected in earnings or cash
flows. This is consistent with both of these internally generated measures of firm
performance excluding factors such as changes in discount rates. It is therefore
assumed that the desired performance measure to be used in contracts excludes these
macro-economic factors.
From an empirical view point there is an added advantage of removing market
wide movements. Bernard (1986) points out that removing market-wide movements
reduces problems with cross-sectional dependence in regressions. Note however, that
removing market-wide movement is not critical to the analysis since, empirically, both
earnings and cash flows have a low association with market-wide movements over each
of the measurement intervals examined. Therefore, similar results are obtained when
raw stock returns are substituted as the dependent variable.
The approach of evaluating the usefulness of accounting information such as
earnings by determining its association with stock returns is not new and is widely used
in accounting research. For example, when Lev (1989) reviews the state of accounting
research he evaluates the usefulness of earnings' by determining its association with
stock returns. Ball and Brown (1968) and Beaver and Dukes (1972) evaluate the
relative usefulness of earnings and earnings before depreciation by comparing each
measure's association with stock returns. In addition, Beaver and Landsman (1983)
evaluate the usefulness of FASB Statement No. 33 earnings by determining if it can
30
explain any of the residual variation from a regression of stock returns on historical cost
earnings. As Beaver and Landsman [(1983), p. 2] point out, obtaining alternative
measures of expected cash flows other than stock price based measures is difficult to
achieve and any other proxy (such as actual future cash flows) introduces additional
assumptions and noise into the analysis.
Even though stock prices are assumed to encompass the information in earnings
and cash flows about firm performance, an internally generated measure of firm
performance is not necessarily redundant. As discussed in chapter 1, stock markets
react to the release of earnings information and to forecasts of earnings [e.g., Foster
(1977) and Patell (1976)]. Thus, the production of financial information such as
earnings is an integral part of price formation. If it is assumed that for competitive
reasons it is costly for firms to disclose all information [see Black (1992)], then a
summary measure is the most desirable format to release information to the stock
market. Alternatively, aggregating information into a summary measure of firm
performance could be the most efficient way for management to communicate with
users when there are information processing costs [e.g., Beaver and Demski (1979)
and Beaver (1981, p. 167)].
Another important reason why internally generated measures of firm
performance, such as earnings are not redundant even when stock prices are available is
highlighted by Sloan (1992b). Sloan (1992b) presents evidence consistent with
earnings-based incentives being incrementally useful over stock price-based incentives
for rewarding management. He shows that earnings reflect firm-specific changes in
value, but are less sensitive to market-wide movements in equity values. Thus,
including earnings-based performance measures in compensation contracts helps shield
executives from fluctuations in firm value that are beyond their control [see Banker and
31
Datar (1989)]. Smith and Watts (1982) also point out that within firms or in private
firms, where there are no stock prices, alternative measures are required for evaluating
and rewarding management. Finally, unlike stock prices, earnings can be decomposed
into components, this is useful in debt covenants since debt holders can provide
management with incentives to increase overall firm performance as well as to protect
their claims [e.g., Smith and Warner (1979)].
4.2. Stock market efficiency versus functional fixation
An assumption made in the analysis is that the stock market is efficient. That is,
price is an unbiased measure of expected future cash flows and so any information in
earnings or cash flows is fully revealed in price. However, if price deviates from
fundamentals and contains noise, then the implicit assumption is that the noise in price
is uncorrelated with either cash flows or earnings. This assumption implies that no bias
in favor of either measure is introduced by using price as a proxy for firm performance.
A violation of this assumption would occur if the market is functionally fixated on one
of the two measures. For example, if the market is functionally fixated on earnings,
then the noise in stock price is correlated with earnings but not with cash flows. Under
this scenario, a higher association between earnings and stock returns would be found
even if earnings are not a more 'useful' measure of firm performance. The higher
association would be a consequence of using an inappropriate benchmark of firm
performance. A similar problem would exist if the market is functionally fixated on
cash flows.
Studies investigating functional fixation have focused on determining whether
the market is fixated on earnings. To date, however evidence on functional fixation has
been mixed. Early studies investigating functional fixation of earnings examined
32
whether the market can 'see through' accounting procedure changes. Both Kaplan and
Roll (1972) and Ball (1972) do not find statistically significant stock price reactions to
the accounting procedure changes. However, these studies suffer from industry
clustering and other confounding event that makes interpretation difficult [see Watts and
Zimmerman (1986)]. Hand (1990) tests for functional fixation by determining whether
the market reacts to earnings that contain gains from previously disclosed debt-equity
swaps. Hand suggests that his evidence is consistent with the stock market reacting to
these previously disclosed 'paper' gains. However, Ball and Kothari (1991) argue that
Hand's results are indistinguishable from the previously documented size effect in
stock returns at announcement dates. Although, Hand (1991) argues that he presents
additional evidence that is inconsistent with Ball and Kothari's explanation.
Recent work by Sloan (l992c) provides evidence consistent with the functional
fixation hypothesis. Sloan shows that high operating accruals are followed by negative
accrual reversals that cause earnings reductions (and vice versa). He argues that if the
market is fixated on earnings, these predictable earnings reversals caused by accruals
will not be fully anticipated by the share price. Therefore, future returns will be
predictable based on the magnitude of accruals. Sloan presents evidence consistent
with this prediction. Sloan finds that at most, ten percent of accruals reverse (cause
predictable changes in subsequent earnings) while the remaining 90 percent of accruals
capture permanent changes in firms' earnings generating ability. Therefore, consistent
with the results presented in this study, the overriding effect of accruals appears to be
an improvement in the ability of earnings to reflect firm performance. However,
because the market does not fully incorporate the predictable earnings reversals, excess
returns are documented by Sloan, for up to three years following the original earnings
release.
33
One test performed in this study directly addresses the concern of functional
fixation. This study examines a subset of accruals that are made to correct for past over
or understatement of accruals. The subset of accruals is special items. Special items
include write-off or write-downs of assets, any adjustment applicable to prior years etc.
These accruals are not predicted to be made to improve earnings' ability to measure finn
performance but rather to maintain 'clean surplus' and to ensure that management is
accountable for the effect of such adjustments. Therefore, their inclusion in earnings is
predicted to reduce the ability of earnings to measure firm performance. The evidence
is consistent with earnings before special items having a higher association with stock
returns than earnings from continuing operations over short measurement intervals.
This evidence is inconsistent with functional fixation, since the market 'backs out' this
component when assessing finn performance.l?
14 A direct test controlling for the evidence of functional fixation documented by Sloan is alsoperformed. Sloan's (l992c) evidence is consistent with predictable accrual reversals taking up to threeyears to be fully incorporated intoprice. Therefore, returns cumulated from the beginning of yeart tothe end of year t+3 will incorporate the effectof any reversals. The resulting R2s for regressions ofearnings and cash from operations using this return metric as the dependent variable are 14.61% forearnings and 6.01% for cash from operations (the sample consists of 4,175 non-overlapping finnobservations). When contemporaneous annual stock returns are used, the R2 on earnings is 17.23%and 4.20% for cash from operations. Therefore, the tenor of the results do not change using thisalternative four-year return metric. Earnings havemore explanatory power thancashflowsevenaftercontrolling for thepotential effectof functional fixation,
34
5 • Non-Nested Model Selection
5.1. The encompassing approach to non-nested model
selection
The research question addressed in this study is: which measure, earnings or
cash flows is a 'better' estimate of firm performance as reflected in stock returns?
Thus, the objective of the study is to determine the conditions under which earnings
explain more of the variation in stock returns than cash flows. If earnings do explain
more of the variation than cash flows, then earnings can be thought of as a more useful
measure of firm performance.
A simple approach to selecting the 'best' model is to compare the coefficients of
determination (R2s) from the separate regressions of stock returns on earnings and
stock returns on cash flows, choosing the model with the highest R2. This seemingly
ad hoc approach has been justified by Gaver and Geisel (1974) and Pollak and Wales
(1991). Gaver and Geisel show that a comparison of R2s can be justified on Bayesian
grounds when the prior information on the distribution of the model parameters is
diffuse (noninformative). Pollak and Wales using a classical approach, show that
35
when both models have the same number of parameters, the likelihood dominance
criterion implies a comparison ofR2s.
An alternative way of addressing this question is the encompassing approach to
model selection [for a review of this literature, see Mizon and Richard (1986)]. The
encompassing principle is based on the idea that a model can be considered superior to
rival models if it can account for the salient features of rival models. The approach
adopted is to test for misspecification of one model in the direction of competing
models. The procedure is done in two steps. First, one model is assumed to be true
and the test involves determining if this model 'encompasses' or explains the salient
characteristics of the competing model. This occurs if we cannot reject the specification
of the null model in the direction of the alternative. The second step is to do the same
procedure in reverse, i.e., determine if the second model explains the relevant
characteristics of the first model (when the second model is assumed true). If the first
model explains the relevant characteristics of the second model but the second model
does not explain the relevant characteristics of the first model, then we can
unambiguously conclude that the first model 'encompasses' the second model and is
the model of choice. The problem with this approach is that ambiguous results are
obtained when both models reject each other or neither model rejects the alternative.
Under these circumstances, such tests do not indicate which model is closer to the 'true
data generating process.'
5.1.1. The .l-test
One 'encompassing test' is the I-test developed by Davidson and MacKinnon
(1981). This test is discussed and used in Biddle and .Seow (1990). Consider
applying the I-test to determine the relative ability of earnings and cash flows to explain
36
stock returns. In these circumstances, the J-test is simply a multiple regression of stock
returns on earnings (E) and cash flows (C), (for simplicity the time subscripts are
excluded):
(5-1)
The test then determines whether the coefficients on either earnings or cash flows is
significant. Thus, tests of the hypothesis that earnings relative to cash flows has a
stronger association with stock returns can lead to the four potential outcomes, as
illustrated on the following table.
Hypothesis:
~=o
Not Rejected
Not Rejected
Cannot Distinguish
Hypothesis: "( =0
Rejected
Reject Cash Flows inFavor of Earnings
Rejected Reject Earnings inFavor of Cash Flows Cannot Distinguish
When applied to the actual data used in this study, only the coefficient on earnings is
significant at conventional levels over the quarterly interval (an outcome in the top right
cell). This can be interpreted as earnings 'encompassing' the salient features of cash
flows and therefore earnings is selected as the 'superior' model. However, over the
annual and four-year intervals, the coefficient on both cash flows and earnings are
significant at conventional levels (the outcome is in the bottom right cell). Under these
circumstances, the J-test cannot rank the two competing models (earnings or cash
flows) or indicate which is 'better.' This occurs since under the null hypothesis, one
37
model (e.g., earnings) is assumed to be 'true' and thus if cash flows provide any
additional information the original presumption that earnings is the true model must be
incorrect. A similar rejection of cash flows in favor of earnings occurs when cash
flows are assumed under the null to be 'true.' Therefore, this test lacks power at
conventional levels, to distinguish earnings from cash flows as competing non-nested
models. [See Maddala (1988, pp. 443-447) for additional discussion on the Davidson
and MacKinnon test].
5.1.2. The incremental approach
Consider decomposing earnings into cash flows and accruals (C + A), so that
equation (5-1) is now restated in terms of these components:
or
(5-2)
where 0 = ~+'Y.
Ifwe are only interested in determining if accruals have incremental information
content, then all that is necessary is to determine if'Y is significant in the regression.
However, in this study it is hypothesized that the reason for accruals is so that earnings
relative to cash flows is a more useful summary measure of firm performance..
Producing earnings therefore, not only involves determining the set of accruals that will
be included in earnings but also the form the accruals will take so that they can be
aggregated to produce earnings. This places additional structure on accruals since they
must not only be informative (have incremental information content) but also be of a
38
form that any information lost through aggregation is not so large as to render earnings
a poorer summary measure of firm performance than cash flows. Therefore, if we are
interested in determining if earnings explains relatively more of the variation in stock
returnsthancash flows, then two conditions are required:
(a)yis significantly different from zero
(b) the linear restriction, 0=y,cannotbe rejected
Since equation (5-2) is identical to (5-1) the results of this test are obvious.
Although, accruals have incremental content [as documented by Rayburn (1986)], the
linearrestriction is rejected over the annual andfouryear intervals. Therefore, although
accruals have incremental information over that of cash flows, no statistical inference
can be made concerning which measure earnings or cash flows is a superior summary
measure of firm performance,
5.2. Using the Vuong likelihood ratio test to evaluate
earnings versus cash flows
A recent development in model selection techniques is Vuong (1989). Vuong
has provided a statistical test to determine which of two models better explains the
dependent variable. The difference between the Vuong test and other non-nested tests,
such as the f-test, is that Vuong has derived the distribution of the likelihood ratio
statistic without the assumption under the null hypothesis that either model is 'true.'
This test is directional and can be used to determine which of two imperfect models is
closer to the true 'data generating process.' The advantage of the Vuong test in the
circumstance outlined above is that Vuong's test statistic allows both models to have
explanatory power(be in thebottom rightcell) but provides direction concerning which
39
of the two is closer to the 'true data generating process'. Therefore. the Vuong test
allows rejection of cash flows in favor of earnings in situations where the tests
described above would not
Consider first. the model of stock returns depending on earnings (again. for
simplicity the time subscripts are excluded):
(5-3)
This implies that Ri are independently and normally distributed with mean (lE + ~EEi
and a common variance O'E2. The joint density of the observations is.
(5-4)
The log likelihood function L«l,~.0'2) is:
(5-5)
When maximizing L with respect to c, ~ and 0'2, as noted in Maddala (1988),
the maximum likelihood estimators of (l and ~ are the same as the least squares1\ 1\
estimators of (l and ji, Substituting (lEand ~E for (lE and ~E we can see that for each
i, Ri - ~E - ~EEi = eEio while the maximum likelihood estimate of O'E2 is R~SE , where
RSSE is the residual sum of squares from the regression of stock returns on earnings.
Now consider the model of stock returns depending on cash from operations:
(5-6)
40
A similar log likelihood function can be obtained for cash from operations:
n n 1 1 ]log L(RcFO) = .L log L(RCi) =.Lf. 2 10g (27tO'c2) -~2Rr - (Xc - ~cCFOi)2
1=1 1=11. 20'c
(5-7)
Again, the maximum likelihood estimate for 0'2 is R;SC and eci = Ri -~ - ~cCFOi.
/\ /\Substituting eEi for Ri - (XE - f3EEi in equation (5-4) and similarly substituting
/\ /\and eCi for R; - (Xc - f3cCFOi in equation (5-7), and obtaining estimates of O'c2 and
O'E2, we can determine which of the two models explains relatively more of the
dependent variable. This is done by first forming a likelihood ratio test comparing cash
flows to earnings:
(5-8)
An estimate of the variance, ro2, of LR is given by [see Vuong (1989, equation 4.2)]:
The statistic is then formed as:
Z=_1 LR...[0/\
CJ)
(5-9)
(5-10)
which tends in distribution to a standard normal random variable. This test is
directional in the sense that if the Z-statistic is positive and significant the test indicates
that earnings explain significantly more of the variation in stock returns than cash from
41
operations, whereas if the Z-statistic is negative and significant the opposite conclusion
can be drawn.
In the case at hand a simpler approach to estimating the Z-statistic is available.
Mter substituting estimates R~SE for O'E2 and eEi for Ri - ~E - ~EEi in equation (5-4)
and R~Sc for O'c2 and eCi for Ri -~ - acCFOi in equation (5-7), and then substituting
(5-5) and (5-7) into (5-8) we can obtain for each observation i,
LRj = 10g[L(REi)] - 10g[L(RCi)]
1 (27t ) 1 (27t ) n 2 n 2=210g n RSSc - 2log n RSSE + 2RSSc (eci) - 2RSSE (eEi) .
Simplifying we can obtain mj for each observation:
which if summed results in the likelihood ratio statistic (5-8).
(5-11)
(5-12)
The next step is to estimate the standard deviation of LR. Vuong notes (see p.
318), instead of estimating the standard deviation ofLR directly to form a Z-statistic, in
this simple case we can obtain the Z-statistic by regressing mj on unity. The coefficient
in this regression will equal ~ 10g[~~~~J and tells us the mean difference in
explanatory power between earnings and cash flows. The standard error, from the
regression tells us whether the relationship is unusual, i.e., if the difference is
significant. The Z-statistic can be obtained by multiplying the t-statistic from the
regression by (n~1f. Note that a positive Z-statistic implies that the residuals
produced by the cash from operations' regression are larger in magnitude than those
from the earnings regression. Hence, a positive and significant Z-statistic indicates that
42
earnings explain significantly more of the variation in stock returns than cash from
operations.
6. Sample DescriptionMeasurement
and
43
Varia hie
6.1. Data
The sample consists of firms listed on the New York Stock Exchange or the
American Stock Exchange. Three measurement intervals are examined: quarterly,
annual and four yearly. Firms are required to have accounting data available on either
the 1990 versions of the COMPUSTAT Merged Expanded Annual Industrial file, the
COMPUSTAT Merged Research Annual Industrial file, or the COMPUSTAT Quarterly
Industrial file. Firm observations are excluded if they do not have data to calculate
earnings-per-share, cash from operations per-share or net cash flows per-share. This
results in a sample of 26,793 firm-quarter observations, 30,489 firm-year observations
and 10,041 firm-four-year observations. The firms are also required to have monthly
returns available on the CRSP tapes. This reduces the sample of firms to 20,716 firm
quarter observations from 1980 to 1989,28,647 firm-year observations from 1960 to
1989 and 5,454 firm-four-year (non-overlapping) observations from 1964 to 1989.
The sample excludes firm observations with the most extreme one percent of earnings-
44
per-share, cash from operations-per-share or net cash flows-per-share. 15 This reduces
the sample to 19,733 firm-quarter observations, 27,308 firm-year observations and
5,175 firm-four-year observations.
6.2. Variable Definitions
All financial statement variables used in the empirical tests are on a per-share
basis and scaled by beginning of period price. The variables are defined as follows:
E = earnings per-share (excluding extraordinary items and discontinued operations),
scaled by beginning of period price.
AWC = change in (non-cash) working capital per-share, scaled by beginning of period
price, where the change in (non-cash) working capital is defined as:
AWorking capital =AM +Mnv +AOthCA - AAP - ATP - AOthCL
where A is the change in each variable from period t-1 to t, AR is accounts
receivable, Inv is inventory, OthCA is other current assets, AP is accounts
payable, TP is tax payable and OthCL is other current liabilities. 16
15 Initial testsshowed thatoutliers wereoftenmorethanfiveor six standard deviations from the meanand in some cases unduly affected the regression results, The decision to exclude observations isconsistentwith previous research in this area. Collins and Kothari (1989) truncate their sample byexcluding observations where earnings changes are greaterthan 200 percent Bowen, Burgstahler andDaley (1987) also reportresults for a winsorized sample where extremeobservations are scaled to thesame value. The cut-off point is however, arbitrary; footnotes will report the main results for thesampleincluding outliers.
16 Other current assets includes items such as prepaid insurance and suppliesapart from inventory.Other current liabilities includes items such as dividends payable. deferred income tax payable andcontracts payable. As in Bowen, Burgstahler andDaley (1987) and Rayburn (1986), short-term accrualssuch as notes payable and the currentportion of long-term debt are excluded under the premise thattheserelatemore to financing activities than to operations.
45
CFO =cash from operations per-share, scaled by beginning of period price.
{Operating income before depreciation - interest - taxes - ~working caPital} IP
number of common shares outstanding t-l
NCF =the change in the balance of the cash account on a per-share basis, (net cash
flow per-share) scaled by beginning of period price.
LTA = Earnings - Cash from operations - AWorkingcapital
is long-term operating accruals per-share scaled by beginning of period price.
AA =Earnings - Net cash flows
is the net change in all non-cash accounts (aggregate accruals) on a per-share
basis scaled by beginning of period price.
Special items = COMPUSTAT item 32 on the quarterly tape and item 17 on the annual
tape. Special items include adjustments applicable to prior years, any
significant non recurring item, non recurring profit or loss on sale of assets,
investments etc., write-downs or write-offs of receivables, intangibles etc.,
. flood, fire, and other natural disaster losses.
Rit = CRSP buy-and-hold stock return (including dividends) for firm i over time
interval t, where t is the contemporaneous quarter, year or four-year period, less
the CRSP value-weighted market index (including dividends) over the
corresponding fiscal period.
6.3. Descriptive Statistics
Table 1 presents descriptive statistics on the variables used in the analysis.
Eamings-per-share scaled by price has a median value of 0.016, 0.090 and 0.383 for
46
the quarterly. annual and four-year intervals respectively. Annual earnings are greater
than four times the quarterly number. while the four-year earnings are greater than four
times the annual earnings. A similar pattern is observed for cash from operations and
net cash flows.I? Since the variables are scaled by beginning of period price. average
reported values will tend to increase over longer intervals due to the reinvestment of
earnings. A second difference between the quarterly and annual observations is the
shorter time period available for quarterly data. In table 1, net cash flows have a lower
mean than earnings. This is expected since a firm's financing and investment policies
affect net cash flows but not earnings. For example. when a firm pays dividends it
reduces retained earnings and cash but not reported earnings.
Table I indicates that over the quarterly interval. relative to earnings. both cash
from operations and net cash flows have more negative realizations. In addition. both
cash flow measures have higher standard deviations than earnings. Over longer
intervals. both the standard deviation and the proportion of negative realizations for the
cash flow measures decline relative to earnings. One explanation for this pattern is that
accruals off-set extreme negative and positive cash flow realizations associated with
mismatched cash receipts and disbursements over short measurement intervals.
Evidence that accruals and cash flows are negatively correlated would support this
explanation. This explanation is investigated in more detail in the next chapter.
17 Table 1 indicates that on average changes in working capital are positive. The components of thechange in working capitalwereinvestigated for the annual data. Of the total changein currentassets,47 percent is due to increases in accounts receivable and 48 percentis due to increases in inventory.For current liabilities, 47 percentof the change is due to accounts payable, and 10 percent is due toincreases in tax payable. Cbanges in current assetsare on average twice as large as changes in currentliabilities. Also note thatdespite changes in working capital beingpositive, cash from operations hasa larger mean and median than earnings. This is because long-term operating accruals, such asdepreciation, are largerin magnitude than changes in working capital and theirnet effecton earnings isnegative.
47
7. Analysis of Measurement Interval Predictions
Discussion of the results is divided into two chapters. This chapter provides
tests of hypotheses 1 and 2. Evidence of cash flows' time-series properties is also
presented that is consistent with cash flows suffering from mismatching problems.
Chapter 8 provides the results of the cross-sectional predictions on where accruals are
expected to be relatively more important in improving earnings' association with stock
returns.
7.1 .. The effect of short and long measurement intervals.
Hypothesis 1 predicts that earnings will have a stronger association with stock
returns relative to net cash flows or cash from operations over short measurement
intervals. Hypothesis 1 is examined by performing three pooled regressions: (1) stock
returns on earnings (2) stock returns on cash from operations and (3) stock returns on
net cash flows, then determining which measure has significantly more explanatory
power using Vuong's (1989) test. The regression performed is:
(7-1)
where (X) is either earnings (E), cash from operations (CFO) or net cash flows (NCF).
48
Table 2 presents the results of tests of hypothesis 1. For each measurement
interval, the R2 is larger in regressions including earnings relative to the regressions
including cash flows. The R2 for the quarterly observations is 3.24 percent for
earnings, 0.01 percent for cash from operations and 0.01 percent for net cash flows.
The R2s increase to 16.20 percent for earnings, 3.18 percent for cash from operations
and 2A7 percent for net cash flows over the annual interval. Over the four-year
interval the R2s increase further to 40.26 percent for earnings, 10.88 percent for cash
from operations and 6.12 percent for net cash flows.18 Over each measurement
interval, earnings are more strongly associated with stock returns than either cash flow
measure. These results support hypothesis 1. They are also consistent with Easton,
Harris and Ohlson (1992), who show that earnings' association with stock returns
improves over longer measurement intervals.
Hypothesis 2 predicts that as the measurement interval increases, cash flows
relative to earnings will capture a greater proportion of the variation in stock returns.
This occurs because over longer measurement intervals the mismatching problems with
cash flows are less acute and so accruals are relatively less important for improving
earnings' association with stock returns. Consider the following ratios:
Since the dependent variable is the same in both the cash flows and earnings
regressions these ratios measure the relative explanatory ability of earnings and cash
18 When the outliersare included in the regressions the coefficients and R2s decline in magnitude forboth earningsand cash flows, but the tenorof theresultsis unchanged. Regressions for the annualandfour-year interval werealsoperformed usingonly the1980to 1988period (equivalent to the period usedfor the quarterly interval). The tenorof the resultsis unchanged.
49
flows over different measurement intervals. A ratio less than one indicates that
earnings explain more of the variation in stock returns than cash flows. If these ratios
increase as the measurement interval is lengthened, then this is consistent with
hypothesis 2. Table 2 indicates that as the measurement interval increases from one
quarter to one year to four years, both the earnings and the cash flow measures have a
higher association with stock returns. This suggests that both earnings and cash flows
suffer less from timing and mismatching problems over longer measurement intervals.
However, in table 2 the ratio r~%Q) increases from 0.003 for quarterly, to 0.20 for
annual, to 0.27 for the four-year interval. A similar but smaller effect is observed for
net cash flows. This is consistent with cash flows relative to earnings, having more
ability to explain the variation in stock returns as the measurement interval is
lengthened.
The adjusted R2s reported in table 2 are consistent with the hypothesis that
earnings are more strongly associated with stock returns than cash flows. However,
simply comparing R2s does not provide statistically reliable evidence that earnings is
superior to cash flows. In order to formally discriminate between the two competing
specifications, they are evaluated as competing non-nested models. One of the
premises of the paper is that neither earnings nor cash flows are a perfect measure of
firm performance because both suffer to varying degrees from timing and mismatching
problems. As discussed in chapter 5, Vuong (1989) provides a likelihood ratio test for
model selection without presuming under the null that either model is 'true.' This
allows a directional test indicating which of the competing hypotheses, if either, is
closer to explaining the data. Intuitively, Vuong's tests allows us to determine which
model has relatively more explanatory power. An alternative test for non-nested model
selection is the J-test by Davidson and MacKinnon (1981). This test simplifies to
50
performing a multiple regression including both cash flows and earnings. However,
when both variables have incremental explanatory power (as is typically the case in the
tests performed here), the J-test lacks power and cannot distinguish between the
competing hypotheses. Intuitively, Vuong's test is a more powerful test than the J-test
because it can reject one hypothesis in favor of the alternative, in circumstances where
the J-test cannot.
Table 3 reports the results of Vuong's test of non-nested models. Over each
interval, earnings explain significantly more of the variation in market-adjusted stock
returns than either cash from operations or net cash flows. Since earnings explain
significantly more of the variation in stock returns than cash from operations, accruals
made to cash from operations to obtain earnings are relatively important for controlling
for the timing and mismatching problems inherent in cash from operations. Cash from
operations, however, does not explain significantly more of the variation in stock
returns than net cash flows over the quarterly interval, although the Z-statistic does
increase in significance over longer intervals. Since the difference between net cash
flows and cash from operations is the long-term investment and financing accruals,
these accruals seem relatively less important for reflecting firm performance. One
explanation for this result is that firms use their internally generated cash from
operations to finance their investment decisions so that the net cash inflow (outflow)
from these two sources is relatively small. Thus, the role of accruals in controlling for
these mismatching problems is relatively less important. Consistent with this
explanation cash from operations is found to have a strong negative correlation of
approximately -0.8 with cash from investment and financing.
51
7.2. Robustness tests
An alternative explanation of the results reported in table 2 is that cash flows
suffer more greatly from a random coefficients problem than earnings. If the
coefficient on cash flows varies across firm and a pooled regression is performed, then
the reported R2 will be understated. If the coefficient on cash flows varies much more
than the coefficient on earnings, then this could lead to an incorrect inference that cash
flows have a lower association than earnings with stock returns. To investigate this
issue, firm-specific regressions are performed on all firms that have more than 10
observations. The average adjusted R2 (in percent) from these regressions for the
quarterly interval for earnings is 5.43, for cash from operations is 0.84 and for net cash
flows is 0.86. For the annual interval, the average adjusted R2 for earnings is 21.87,
for cash from operations is 7.28 and for net cash flows is 4.19. This analysis was not
conducted over the four year interval because of the few time-series observations
available per firm. The results for firm-specific regressions are consistent with
hypothesis 1 and 2. Performing firm-specific regression improves the R2 for both cash
flows and earnings. Thus, the results reported in table 2 are not due solely to cash
flows suffering more greatly from a random coefficients problem.
A second concern is that the Vuong Z-statistic reported in table 3 are unduly
influenced by a few unusual years. To determine if this is the case, the Vuong test was
repeated over the annual interval for each of 23 years that had more than 100
observations (1967 - 1989). The results (not reported) indicate that in all 23 years,
earnings outperforms both cash flow measures, therefore the results are not driven by a
few unusual years. In addition, the Z-statistics are significant at the 0.01 level in 21 out
of 23 years when earnings are compared to cash from operations, and 20 out of 23
52
years when earnings are compared to net cash flows. However, when cash from
operations is compared to net cash flows, the Z-statistic is only positive in 12 of the 23
year and significant in four of the 23 years. Therefore, less reliability can be placed on
cash from operations relative to net cash flows being significantly more strongly
associated with stock returns.
A third potential concern with the tests in table 2 and 3 is that the cash flow
proxies are inappropriate. Prior to 1988, firms were not required to report cash from
operations but instead could report 'funds' under a variety of different formats. To
avoid problems with comparability across firms, one consistent definition is used for
cash from operations over the entire sample period. This approach has also been
adopted in existing research [e.g., see Rayburn (1986) and Bernard and Stober
(1989)]. However, since balance sheet data is used to obtain the changes in working
capital, this number will contain noise. The effect of the noise in this measure is
reduced by eliminating outliers, however the noise may still be of a large enough
magnitude that the observed association with stock returns is understated. The
consequence of this could be an incorrect conclusion that earnings has a higher
association with stock returns than cash from operations.
A fourth concern is that alternative cash flow measures may be more consistent
with firm performance. For example, theoretical models such as Fama and Miller
(1972, chapter 2) suggest that cash flows after investment expenditures may be the
variable of interest. In addition, it could also be argued that since earnings includes a
charge for the effect of investment decisions (depreciation), the appropriate cash flow
proxy should be one that includes cash outlays for investments. Alternatively, if
investment decisions are assumed to be exogenous, investors may be primarily
concerned with cash flows relating to the firms' financing decisions.
53
To provide some insights into the third and fourth concerns table 4 compares
the cash flow measures reported in table 2 to various alternative definitions of cash
flows. The alternative definitions of cash flows are obtained from The Financial
Accounting Standards Board's FAS 95, Statement of Cash Flows. FAS 95 requires
firms to separately report cash flows relating to operations, investments and financing
and therefore mitigates problems with obtaining cash flow proxies. FAS 95 also
requires replacement of the old funds statement with a new cash flow statement for
fiscal years ending after July 15, 1988. Some firms adopted the standard in 1987,
therefore data is obtained for all firms using the cash flow statement format. Table 4
provides seven different cash flow measures and compares each of them to earnings.
The seven measures are:
(i) cash from operations as required by FAS 95 (CF095)(ii) cash from operationas measured in table 2 (CFO)(iii) CF095 plus investments cash flows (CF095 + Inv CF)(iv) investmentcash flows (negative for investment outlays) (Inv CF)(v) financing cash flows (positive for financing inflows) (Fin CF)(vi) investment cash flows plus financing cash flows (Inv CF + Fin CF)(vii) net cash flows as measured in table 2 (NCF).
The results in table 4 indicate that earnings has a higher association with stock
returns than any of the cash flow measures. The R2 on earnings is 14.58 percent,
while the largest R2 for the cash flow variables is 5.20 percent (for CF095). The
Vuong Z-statistics comparing earnings to each cash flow measure are all significant at
the one percent level and indicate that earnings has significantly more explanatory
power than the cash flow measures.
The cash flow measure most comparable to earnings is (CF095+Inv CF)
because as with earnings, this measure includes charges for investment outlays.
However, in the production of earnings the investment charge is allocated as
'depreciation' over the life of the investment. The effect of this is to better match the
54
cost of the investment outlay with the revenue that it generates. The results indicate that
(CF095+Inv CF) has a zero association with stock returns. Combining cash from
operations with investment outlays produces a poor measure of firm performance. The
reason for this can be seen by noting that (Inv CF) has a negative association with stock
returns. When firms have negative investment cash flows (make investment outlays)
this is generally associated with positive expected future cash flows. Therefore,
allocating all of the negative investment cash flow to one period and not recognizing the
future benefit, produces a measure that suffers severely from mismatching problems.
The results in table 4 indicate that CF095 has an adjusted R2 of 5.20 percent
where as CFO has an adjusted R2 of 2.48 percent. The Vuong Z-statistic is 2.76
(significant at the 0.01 level, two-tailed test) indicating that the cash flow measure as
described by FAS 95 has significantly more explanatory power than CFO. This is
consistent with CFO being a noisier measure of cash from operations than CF095.
This noise could be the result of including certain accounting entries with no direct cash
flow consequences in CFO. For example, the book-keeping effects of divestiture or
mergers will be included in CFO but not CF095. Alternatively, the higher R2 for
CF095 could be due to a sample selection bias; early adopters of FAS 95 anticipated
that providing CF095 would be more informative than CFO.19 However, as noted
above, CF095 still has significantly less explanatory power than earnings. In
summary, the evidence in table 4 is consistent with hypothesis 1 and suggests that the
low explanatory power of cash flows is not due solely to the choice of cash flow
measures.
19 When 418 'early adopters' are excluded from the sample. there is no significant difference betweenthe explanatory power of CF095 and CFO. For the 418 'early adopters' the explanatory power ofCF095 is over seven percent while the explanatory power of CFO is less than one percent.
55
7.3. Mismatching problems with cash flows
7.3.1. Time-series properties
Tables 2 and 3 provide evidence consistent with cash flows being a relatively
poor measure of firm performance. The next set of tests examine whether cash flows
have time-series properties consistent with them suffering from mismatching problems.
If cash flows suffer from temporary mismatching of cash receipts and disbursements,
then this suggests that (i) changes in cash flows will exhibit negative autocorrelation,
i.e., a large cash outflow this period is more likely to be followed by a large cash
inflow next period. For example, in the ship building firm discussed in chapter 3, cash
flows are expected to be negative in the first period, when expenditures are outlaid and
positive in the next period when cash is collected. Therefore, changes in cash flows are
likely to contain temporary components that are reversed over time. If accruals are used
to match cash receipts and disbursements associated with the same economic event,
then this suggests that (ii) changes in accruals will also exhibit negative autocorrelation.
If accruals are used to match cash receipts and disbursements, then this suggests that
(iii) accruals will be negatively correlated with changes in cash flows. Finally, if
mismatching problems are less important over longer intervals, then (iv) cash flows and
accruals will have a smaller (in magnitude) negative correlation over longer intervals.
Evidence in support of these predictions is provided in table 5. Panel A of table
5 presents firm-specific first-order annual autocorrelations for each performance
measure: net cash flows, cash from operations and earnings. The results indicate that
changes in net cash flow per share exhibit average negative autocorrelations of -0.523.
Changes in operating cash flow per share exhibit slightly smaller (in magnitude)
average negative autocorrelations of -0.434, while changes in earnings-per-share has an
56
even smaller negative autocorrelation of -0.175.20 The change in each accrual measure
also exhibits negative autocorrelation. These result are consistent with predictions (i)
and (ii) above. They are consistent with cash flows containing larger temporary
components than earnings.
Turning to the third point discussed above. If accruals are used to smooth
temporary fluctuations in cash flows, then changes in cash flows and accruals will be
negatively correlated. In addition, if mismatching problems are more acute over short
measurement intervals, then the correlations will be more negative over short intervals.
The results for this test are reported in panel B of table 5. The results are consistent
with these predictions. The average correlation between changes in net cash flows and
aggregate accruals is -0.876 over the quarterly interval, -0.553 over the annual interval
and -0.407 over the four-year interval. Similar result are reported for correlations
between changes in cash from operations and changes in working capital. Table 5
panel B also reports the correlation between changes in cash from operations and
changes in earnings. Over longer intervals, as the temporary components in cash flows
'cancel each other out', changes in earnings and changes in cash from operations are
expected to have a higher positive correlation with each other. The results indicate that
the correlation increases from 0.059 for the quarterly interval, to 0.132 for the annual
interval, to 0.300 for the four-year interval.
The results presented in table 5 are consistent with the matching principle. They
are also consistent with the alternative view that management use accruals to 'smooth'
20 Both quarterly and the four year data exhibit a similar pattern to theannual data. Changes in cashflows and accruals exhibit negative autocorrelations, while earnings exhibit close to zeroautoeorrelations.
57
earnings regardless of whether this improves earnings' ability to measure firm
performance. However, the empirical results in tables 2 and 3 demonstrate that
accruals on average, improve the association of earnings with contemporaneous stock
returns. This suggests that earnings are not 'arbitrarily' smoothed but are smoothed in
a way that increases their ability to accurately reflect firm performance. The results in
table 5 also suggest that assuming all changes in cash flows and accruals are permanent
(a random walk model for expected cash flows and accruals) is not necessarily
appropriate [see for example, DeAngelo (1988), Rayburn (1986) and Pourciau (1992)].
This could be particularly important in studies attempting to determine if accrual
manipulation has occurred, since last period's accruals are unlikely to be the best
estimate of this period's accrual. Identifying the mean reverting component in cash
flows and accruals could also be important when modeling the unanticipated
components of each variable in market reaction studies.
7.3.2. Additional evidence consistent with mismatching problems
One extension of the tests performed in table 2 is to show that part of cash
flows' low association with stock returns is driven by its time-series properties, i.e.,
because cash flows have temporary components (or mismatching problems). If these
temporary components are 'canceled out' with past and future realizations of cash
flows, then by taking the time-series average of the cash flow variables or by including
additional realizations of cash flows in the regression with annual stock returns, a
higher association will be observed. Note that if the role of accruals is to reduce these
mismatching problems, then taking the average of earnings is not expected to result in
as much improvement as that observed for cash flows. The point of this section is not
to obtain the highest possible R2 for the cash flow regressions. The objective is to
show that by allowing the regression procedure to perform the role of accruals, (i.e.,
58
place weights in the current period, on past and future cash flows) an improved
association will be documented between cash flows and stock returns.
In table 6 the following regressions are performed:
(7-2)
(7-3)
(7-4)
where Rit is the adjusted stock return measured over the fiscal year for each firm, X is
either earnings, cash from operations or net cash flows on a per share basis scaled by
price at t-l. The regressions are performed every three years beginning in 1964. This
is done so that there are no overlapping observations, since including overlapping
observations can lead to overstated R2s. However, the results are similar even when
overlapping observations are included in the regressions. The analysis is limited to
three realizations ofeach variable in order to maintain a sufficiently large sample size.
The results in table 6 indicate that the average cash from operations variable has
a higher association with stock returns than the regression that only includes the current
realization. The R2 increases from 4.79 when only CFO t is included in the regression,
to 5.63 when the average of the three years is included in the regression. In addition,
when the coefficients on the past and future realizations are allowed to vary, the R2
increases to 6.73 percent. Similar results are documented for net cash flows. When
the average of three realizations are included in the regression, the R2 increases from
3.18 to 4.58 percent. These results are consistent with cash flows having temporary
components that 'cancel out' since the average cash flow variable is a better proxy of
future cash flows then the current realization. In contrast, when average earnings is
59
included in the regression instead of the current realization, the R2 declines. The R2 is
17.26 when E t is included in the regression, and 14.68 for average earnings.
However, as also shown by Collins, Kothari, Shanken and Sloan (1992), including
future realizations of earnings (separately) in the earnings regressions does improve the
R2. This is consistent with earnings exhibiting timing problems.
These results are consistent with accruals mitigating mismatching problems in
cash flows. When average cash flows are included in a regression with stock returns,
the R2 increases over that obtained for the current realization. This is consistent with
average cash flows exhibiting fewer mismatching problems and therefore being a less
noisy measure of firm performance than currently reported cash flows. This occurs
since the overall effect of averaging is to mitigate the (large) temporary components in
cash flows. In contrast, since earnings includes accruals and suffers to a lesser extent
from mismatching problems, the overall effect of taking the average is to reduce
earnings' association with stock returns.
60
8 . Analysis of Cross-sectional Predictions
This chapter provides cross-sectional tests of hypotheses 3 and 4. Section 8.1
analyzes the effect of aggregate accruals on the ability of earnings and cash flows to
measure firm performance. Results of tests concerning accrual components are then
presented. Finally, tests concerning the operating cycle are performed to determine if
this is an important economic determinant of where cash flows and earnings are likely
to differ in their ability to measure firm performance.
8.1 .The effect of the magnitude of aggregate accruals
The results in the previous chapter suggest that, relative to cash flows, earnings
are more strongly associated with stock returns over short measurement intervals.
These results are consistent with the predictions of hypotheses 1 and 2. Hypothesis 3
predicts that in firms with more volatile investment, financing and operating activities,
net cash flows will be a relatively poor measure of firm performance. In these
circumstances, cash flows suffer to a greater extent from timing and mismatching
problems. Since accruals attempt to control for these problems, cash flows are
expected to have a relatively low association with stock returns when accruals are large
in magnitude.
61
Table 7 provides results of a test to determine whether the ability of net cash
flows to reflect firm performance declines as the absolute value of aggregate accruals
[abs(AA)] increases. In this test, all firm-period observations are ranked on the basis
of abs(AA), quintiles are then formed and separate regressions of stock returns on
earnings and stock returns on net cash flows are performed for each quintile. Quintile I
contains firm-observations for which the magnitude of abs(AA) is small, while quintile
5 contains firm-observations for which the magnitude of abs(AA) is large. For firms
that are in 'steady state', where abs(AA) is small, earnings and cash flows are expected
to have a similar association with stock returns (a similar R2). However in quintile 5,
where abs(AA) is large, indicating more timing and mismatching problems with cash
flows, cash flows relative to earnings are predicted to have a lower association with
stock returns. Since accruals are expected to be relatively more important over short
time intervals, the decline in R2 is expected to be greatest over the quarterly and annual
intervals. Cash flows are also more likely to suffer from errors in variables in quintile
5 relative to quintile 1. If excluding abs(AA) is an important source of measurement
error in the cash flow regressions, then the coefficient on net cash flows will decline
across quintiles (see Appendix 2).
The empirical implications of hypothesis 3 are supported in table 7. Over the
quarterly interval, in quintile 1, where abs(AA) is small and earnings and cash flows
are most similar, cash flows are expected to suffer from fewer mismatching problems.
In this quintile, the R2 on cash flows is 3.44 and the R2 on earnings is 3.55.
However, when abs(AA) is large (quintile 5), net cash flows has an R2 of 0.15 percent
while earnings has an R2 of 4.96 percent. Thus when earnings and cash flows differ
by the greatest magnitude, earnings has a higher association with stock returns than
cash flows. The same finding is observed over the annual and four-year interval.
62
When abs(AA) is large in magnitude, indicating greater timing and mismatching
problems with cash flows, earnings relative to cash flows, has a higher association
with stock returns. Moving from quintile 1 down to quintile 5, the R2 on quarterly
cash flows declines monotonically from 3.55 to 0.15 percent. A close to monotonic
decline is observable for cash flows over the annual and four-year intervals. In
contrast, earnings shows no obvious decline across quintiles. This is consistent with
cash flows suffering from more timing and mismatching problems as accruals increase
in absolute magnitude.
The coefficient on cash flows also declines in size across quintiles over each
measurement interval. For the quarterly interval the coefficient on cash flows is 1.60 in
quintile 1 and an insignificant -0.08 in quintile 5. This is consistent with cash flows
being a 'noisier' measure of firm performance in quintile 5 relative to quintile 1 because
it suffers more greatly from timing and mismatching problems. If accruals mitigate but
do not eliminate these mismatching problems in cash flows, then a similar (but smaller)
decline will be observable for earnings. However, although the coefficient on quarterly
earnings declines from 1.69 in quintile 1 to 0.60 in quintile 5, the R2 on earnings does
not decline. This result suggests that the decline in the coefficient on cash flows and
earnings is not due solely to 'noise' but inter-portfolio variation in underlying
determinants of the coefficients. This issue is investigated in the next section.
Table 7 also reports the Z-statistic and probability from the Vuong test for each
quintile. The results indicate that as expected, in quintile 1, earnings does not explain
significantly more of the variation in stock returns than cash flows. However, in
quintile 5, where abs(AA) is large in magnitude, the Z-statistic is significant. This
suggests that earnings explain more of the variation in stock returns than cash flows in
63
quintile 5. This is consistent with accruals mitigating mismatching problems in cash
flows so that earnings is a more timely measure of firm performance.
The results of the tests presented in table 7 support hypothesis 3. The results
identify the conditions under which cash flows are less effective in measuring firm
performance. They demonstrate that cash flows are not a poor measure of firm
performance per se. In 'steady state' firms where the magnitude of accruals is small
and cash flows and earnings are most similar, cash flows are a relatively 'accurate'
measure of firm performance (have similar R2s to that of earnings over each interval as
reported in table 2). However, when the magnitude of accruals is large, cash flows'
association with stock returns declines and earnings are significantly more strongly
associated with stock returns. In other words, the results indicate that accruals increase
the ability of earnings to reflect performance on a timely basis. Overall, the results are
consistent with the hypothesis that accountants accrue revenues and match expenditures
to revenues so as to produce a performance measure (earnings) that better reflects firm
performance than realized cash flows. This suggests that management manipulation of
accruals is a second-order effect and the first-order effect of the accrual process is to
produce a more timely measure of firm performance.
8.2. Analysis of the decline in the coefficient on earnings
across quintiles
The results in the previous section indicate that the coefficients and R2s on cash
flows decline across quintiles formed on the absolute value of aggregate accruals
[abs(AA)]. One reason for this is that as the abs(AA) increases, cash flows suffer more
greatly from timing and mismatching problems and are a 'noisy' measure of firm
performance. The effect of this noise is to bias down the coefficient and R2 on cash
64
flows (see appendix 2). However, the results in table 7 indicate that although there is
no observed decline in the R2s on earnings, there is a decline in the coefficient (the
earnings response coefficient) across quintiles. If the decline in the earnings response
coefficient is due to noise, then both the coefficient and the R2 are expected to decline
across quintiles. This suggests that the decline in the earnings response coefficient is
due to inter-portfolio variation in the 'true' value of the coefficient Existing literature
has identified a number of determinants of cross-sectional variation in 'true' earnings
response coefficients [see Kormendi and Lipe (1985), Collins and Kothari (1986) and
Easton and Zmijewski (1986)]. This section demonstrates that the decline in the
earnings response coefficient can be explained by these determinants.
Existing research investigating earnings response coefficients predicts that,
ceteris paribus, the earnings response coefficient is inversely related to the expected
return on a security [see Kormendi and Lipe (1985), Collins and Kothari (1986) and
Easton and Zmijewski (1986)]. Earnings response coefficients are also expected to be
greater for firms with growth opportunities, since a given shock in earnings is expect to
be magnified in future earnings [see Collins and Kothari (1986)].21
The following proxies for risk and growth opportunities are investigated to
determine if they vary systematically across quintiles formed on the absolute value of
aggregate accruals [abs(AA)]:
21 Konnendi and Lipe (1985) and Collins and Kothari (1986) hypothesize that earnings responsecoefficient will vary based on the persistence of earnings. However. if earnings have more temporarycomponents (a given change does not persist) then this is expected to affect both the R2 and theearnings response coefficient. Therefore persistence is unlikely to explain the observed decline inearnings response coefficients across quintiles. Consistent with this prediction, temporary componentsin earnings (as measured by fum-specific autocorrelation coefficients, see section 7.3) did not varysignificantly across quinliles.
65
Risk (i)
(ii)
(iii)
(iv)
Growth (v)opportunities
(vi)
(vii)
EJP = earning per share from t-2 to t-I scaled by price at t-1.
Leverage =assets at t-I scaled by book value of equity at t-1.
Mkt/Book =market value of equity scaled by book value ofequity both measured at t-I.
Size = market value of equity at t-1.
R&DIAssets =research and development expenditures scaled byassets at t-1.
Capital explDepr = capital expenditures scaled by depreciationexpense.
Net/Gross = the net book value of fixed assets scaled by thegross book value of fixed assets.
Variables (i) to (iv) are proxies for the risk attributes of firms. Firms with high
EJP ratios or high leverage are expected to have higher expected returns [see Beaver and
Morse (1978) and Kothari (1988)]. Fama and French (1992) also suggests that firm
size and market to book ratios are inversely related to firms' expected returns.
Variables (v) to (vii) are proxies for growth opportunities. Firms that spend a relatively
large proportion of their assets on R&D are often assumed to have a greater proportion
of their value in growth opportunities [see Smith and Watts (1992)]. Firms with
growth opportunities are likely to outlay more on capital expenditure relative to their
depreciation. They are also likely to be relatively 'young' firms (have newer assets).
Table 8 reports mean and median values for each of these variables across
quintiles. Medians are reported since they are less sensitive to the effects of outliers.
Consistent with inter-portfolio variation in risk. firms in quintile 5 relative to other
quintiles, have higher FJP ratios. are more highly levered. have lower market to book
values and are smaller in size. Results from difference in means tests (not reported)
indicate that quintile 5 differs significantly from quintile 1 and all other quintiles based
on these four variables. This is consistent with the decline in the earnings response
66
coefficients across quintiles being due to risk differences. There is also evidence in
table 8 consistent with growth opportunities varying in a systematic manner across
quintiles. Median R&D scaled by assets is 0.026 in quintile 1 and declines
monotonically to 0.015 in quintile 5. However, there is less variation across quintiles
in Capital explDepr or Net/Gross.
The results in table 8 can also be used to reconcile why the earnings response
coefficients decline across quintiles but the R2 on earnings increase in quintile 5. If the
release of financial statements is a relatively more important source of information for
firm-observations in quintile 5, then the explanatory power of earnings is expected to
be greater. Table 8 reveals that the mean and median firm size in quintile 5 is lower
than for other quintiles. This is consistent with the predictions of Atiase (1985) and the
empirical results of Collins, Kothari and Rayburn (1986). Earnings explains a
relatively greater proportion of the variation in stock returns for small firms since there
are fewer competing sources of information.
. In summary, the results in table 7 are inconsistent with the decline in the
earnings response coefficients across quintiles being due to 'noise' because there is no
observable decline in the R2 on earnings. This section presents evidence to explain the
observed decline in the earnings response coefficients. The results in table 8 indicate
that cross-sectional determinants of earnings response coefficients vary systematically
across quintiles formed on the absolute magnitude of accruals.22
22 A regression was also performed to determine if abs(6AA) still explained earnings responsecoefficients after controlIing for variables (i) to (vii), To obtain the observations for the regression,fum-specific earnings response coefficients are determined for 926 flrms that had between 10 and 27annual observations; then the time-series average value of variables (i) to (vii) and abs(6AA) isobtained for each fum. Thesevariables are then included in a multiple regression where the earningsresponse coefficient is the dependent variable. The resultsof the regression indicate that even aftercontrolIing for variables (i) to (vii), abs(6AA) still has explanatory power. It should be noted
67
8.3. Evaluation of accrual components
This section provides the empirical results concerning accrual components. The
analysis is divided into three sections. The first section presents the results for working
capital accruals. These accruals are predicted to be relatively important for mitigating
the timing and mismatching problems in cash flows. The second section provides
additional support for this prediction by showing that cash flows' ability to reflect firm
performance does not vary systematically with the magnitude of long-term operating
accruals. Finally, a specific accruals is identified that in contrast to previous results
actually reduces earnings' association with stock returns.
8.3.1. The effect oJ changes in working capital
The previous results suggest that on average, aggregate accruals as a group,
improve the ability of earnings to reflect firm performance. Table 9 examines a subset
of accruals, specifically, short-term working capital accruals. A comparison is made
between earnings and cash from operations since both measures include the effects of
long-term investment and financing accruals. Working capital accruals are expected to
reverse within one year and hence these accruals are predicted to directly mitigate
temporary fluctuations in firms' cash from operations. If working capital accruals
mitigate temporary mismatching problems in operating cash flows, then firms with
more volatile working capital requirements are expected to have a relatively low
association between stock returns and cash from operations.
however, that the results from this regression are notdirectly comparable to the resultsreported in table7 sincein that table quintiles are formed based on pooled observations.
68
Table 9 presents results of quintile regressions where observations are ranked
based on the absolute value of the change in working capital [abs(~WC)]. For each
measurement intervals, the R2s on cash from operations are more similar in magnitude
to those of earnings in quintile 1. However, as abs(~WC) increases in magnitude, the
R2 on cash from operations declines and in quintile 5, is close to zero. For example,
the R2 on quarterly cash flows is 1.98 in quintile 1 and declines to 0.16 in quintile 5.
This contrasts with the results for earnings for each interval. The R2s in quintile 5 are
of a similar magnitude to those reported in quintile 1. For example, the R2 on quarterly
earnings is 2.81 in quintile 1 and increases to 3.71 in quintile 5. As in table 7, the
coefficients on cash from operations and earnings decline across quintiles, suggesting
that not all of the decline in the coefficient on cash from operations can be attributed to
the exclusion of working capital accruals. The results of the Vuong test indicate that
over the quarterly interval, the Z-statistic is insignificant in quintile 1 and significant in
quintile 5. This is consistent with working capital accruals being relatively important
over short time intervals. Over the annual and four-year interval the Z-statistic is
significant and positive for all quintiles, with the Z-statistic being largest in magnitude
in quintile 5.23
The results in table 9 provide support for hypothesis 4(a). They are consistent
with working capital accruals mitigating the timing and mismatching problems in cash
flows. Thus, working capital accruals play an important role in improving earnings'
ability to measure firm performance.
23 When outliers are included similar results are report in tables 7 and 9 in that the R2on cash flowsis always lowest in quintile 5 relative to quintile I. Table9 wasalso replicated using annualcash fromoperations as reported in FAS 95. The R2 on CF095 declined across quintiles, being smallest inmagnitude in quintile 5. However, the decline is not perfectly monotonic since quintile4 exhibited ahigherR2 than quintile 1.
69
8.3.2. The effect of long-term operating accruals
This section focuses on the effect of long-term operating accruals on the
association between cash from operations and stock returns. In the previous section it
was hypothesized that working capital accruals are made specifically to reduce timing
and mismatching problems with realized cash flows. Long-term operating accruals by
defmition, take several years to reverse. They are included in earnings primarily so that
in the long run, cash-in minus cash-out will be equal to earnings, e.g., depreciation,
gains on sales. They are not expected to be as important for mitigating mismatching
problems in currently realized operating cash flows. Thus, cash from operations'
association with stock returns is not expected to be a declining function of the
magnitude of these long-term operating accruals.
Table 10 presents results of quintile regressions where observations are ranked
based on the absolute value of the long-term operating accruals [abs(LTA)]. Quintile 1
contains observations where abs(LTA) is small in magnitude, while quintile 5 contains
observations where abs(LTA) is large. There are three things worth noting in this
table. First, for each measurement intervals, the R2s on cash from operations are no
longer most similar in magnitude to those of earnings in quintile 1. Second, the R2 on
cash from operations are not a declining function of [abs(LTA)]. For example, over the
annual interval the R2 on cash from operation is 2.24 percent in quintile 1 and increases
to 4.00 percent in quintile 5. Third, over each interval, the R2 on cash from operations
observed for each quintile is of a similar magnitude to that observed in the pooled
regressions reported in table 2 (approximately zero over the quarterly interval, three
percent over the annual interval and ten percent over the four-year interval). Therefore,
any variation across quintiles seems random. This contrasts with the working capital
70
results reported in table 9. In that table over each interval, quintiles 1 to 4 have R2s on
cash from operations that are larger in magnitude than those reported for the pooled
regressions in table 2. The evidence is consistent with long-term operating accruals
being less important for mitigating timing and mismatching problems in cash from
operations. Thus, the results are consistent with working capital accruals playing an
important role (and long-term operating accruals playing a less important role) in
improving earnings' ability to measure firm performance.P
8.3.3. The effect of special items
The section tests hypothesis 4(b) that predicts that special items are an accrual
adjustment that will not improve earnings' ability to reflect finn performance over short
measurement intervals. Results of several empirical studies suggest that special items
or subsets of special items are less useful for measuring current performance. For
example, Jones and Bublitz (1991) find that there is a greater frequency of surprise
write-offs found in the fourth quarter and there is a smaller market reaction to
unexpected earnings in that quarter. Dechow, Huson and Sloan (1992) find that a
subset of special items (those relating to restructuring charges) are filtered out of
management compensation. Finally, DeAngelo, DeAngelo and Skinner (1992) find
that by excluding such items from earnings, they are able to obtain a better model of
expected future earnings and dividends.
24 Tests comparing net cash flows to cash from operations as long-term investment and financingaccruals increase in magnitude were also performed. The difference in R2s acrossquintiles werenotsignificant, therefore consistent with the findings in table 3, these long-term accruals seemrelativelylessimportant for improving theability of earnings tomeasure fum performance.
71
Table 11 contains results for two separate samples. The first sample consists of
firm-observations reporting a special item. the second sample consists of firm
observations where no special items are reported.25 Earnings before tax are reported
in the table so that both special items and earnings are measured on a consistent basis.
Hypothesis 4(b) predicts that if special items are included in earnings primarily to
reduce management discretion. then over short measurement intervals their inclusion in
earnings will reduce earnings' association with stock returns. Over longer intervals.
such adjustments are likely to be less important since the choice of depreciating an asset
versus a one time write-off will have zero impact on total net income (if clean surplus
holds).
The results in table 11 indicate that over the quarterly interval. the R2 on
earnings is lower than the R2 on earnings before special items. Similar results are
found for the annual interval. This is consistent with special items reducing the ability
of earnings to measure firm performance over short time intervals. The Z-statistics
from the Vuong test are significant in favor of earnings before special items at the three
percent level for the quarterly interval and at the one percent level for the annual
interval. In addition. when special items are included as a separate explanatory variable
in regressions including earnings before special items. the coefficient on special items is
insignificant over both the quarterly and annual intervals (not reported). Over the
longer four year interval however. there is no significant difference between earnings
and earnings before special items.
25 The second sample is provided for comparative purposes because results are reported withoutremoving the extreme one percent of observations. Special items usually result in unusual earningsnumbers so that many of the observations for the 'with special items sample' are from the extremepercentiles.
72
These results are consistent with special items reducing the ability of earnings to
measure firm performance over short measurement intervals. The results also indicate
that over the annual and four-year interval, the sample of firms with special items
generally has a lower association between stock returns and earnings (before special
items) than the sample of firms with no special items. One explanation for this is that
firms with special items are undergoing significant changes in their operating,
investment and financing activities and therefore the earnings number is a poorer
predictor of these firms' expected future cash flows.
To summarize, the results presented for components are consistent with the
predictions. Working capital accruals are shown to be relatively important for
improving earnings' ability to measure firm performance, while long-term operating
accruals are shown to be relatively less important. This is consistent with short-term
working capital accruals mitigating the mismatching problems inherent in cash flows.
The results for special items suggest that not all accruals are included in earnings with
this objective in mind. Special items do not improve earnings' ability to measure firm
performance over short measurement intervals. This is because special items do not
mitigate mismatching problems in realized cash flow but instead reflect the cumulative
effect of past under or overstatement of accruals.
8.4. The effect of the operating cycle
The results in the previous section indicate that cash flows have a lower
association with stock returns in firms experiencing large changes in their working
capital requirements. This section identifies an underlying economic determinant that
influences the variability of firms' working capital requirements. Hypothesis 5 predicts
that if the variability of working capital is associated with the length of a firm's
73
operating cycle, then cash flows will have a weaker association with stock returns in
firms with long operating cycles.
To investigate the association between firms' working capital requirements and
the length of the operating cycle, two proxies for the length of the operating cycle are
calculated for the annual interval [see Bernstein (1990), p. 104]:
. I (ARt + ARt_l)/2) ( (Invt + Invt_l)12 )operating cyc e = Sales/360 + Cost of goods sold/360
d I _(ARt + ARt_ d/2 ) ( (Invt + Invt.l)12 ) (APt + APt-l)/2 )tra e cyc e - Sales/360 + Cost of goods sold/360 - Purchases/360 .
Where AR is accounts receivable, Inv is inventory and AP is accounts payable. The
first component measures the number of days' sales in accounts receivable. The
second component measures the number of days it takes to produce and sell the
product. The third component of the trade cycle is the number of days credit obtained
from suppliers.w
Panel A of table 12 provides descriptive statistics on firms' operating and trade
cycles. A minimum of 10 observations per finn is required to calculate each firm's
operating and trade cycle. This results in a sample of 1,252 firms with 20,115 annual
observations. It takes the average (median) firm in the sample 146 (138) days from
taking delivery of inventory to produce a product, to sell the product and receive the
cash from the trade receivable. The average (median) trade cycle is 108 (101) days.
Measures of the operating and trade cycle are sensitive to outliers. To reduce the effect
of outliers the data is aggregated by industry. The aggregated industry operating (trade)
26 Purchases are calculated as ending inventory pluscostof goodsoldless beginning inventory. Noadjustment is madefor depreciation in costof goodsoldsinceCOMPUSTATdoesnot recordthisas aseparate item.
74
cycle is obtained by: (i) estimating firm-specific operating (trade) cycles and (ii)
averaging the operating (trade) cycles for the firms in each industry. The minimum
industry operating cycle is 28 days (Eating and Drinking Places) and the maximum is
277 (Real estate).27
Panel B of table 12 analyzes firm-level correlations between the length of the
operating (trade) cycle and the variability of the change in working capital. The results
suggest that firms with longer operating and trade cycles also tend to have more volatile
working capital requirements. The results in panel C indicate that by aggregating the
data by industry, a stronger association is obtained between the length of the operating
(trade) cycle and the variability of working capital requirements. Thus, in industries
where the operating or trade cycle is long, working capital requirements also tend to be
more volatile. This is consistent with the length of the operating cycle being an
economic determinant of the volatility of working capital requirements.
The results in panel B and C of table 12 suggest that cash flows' association
with stock returns will vary in cross-section with the length of the operating (trade)
cycle. To investigate the effect of the operating (trade) cycle on cash from operations'
and earnings' association with stock returns, the following procedure is performed (i)
all firms with an estimated operating (trade) cycle are categorized by two digit SIC
code, (ii) separate regressions of returns on earnings and returns on cash from
operations are performed for each industry and (iii) Pearson correlations are obtained
between the length of the operating (trade) cycle for the industry and the size of the R2
27 A trade-off between obtaining more firms per industry and a more accurate measure of averageoperating cycles perfum resulted in thechoice of tenobservations perfum. The tenor of theresults isunchanged when thenumber of observations perfum range from eightto 20.
75
for earnings and cash flows from these industry-specific regressions.
Table 13 reports Pearson correlations between the length of the average
operating (trade) cycle for the industry and the R2 from regressions of stock returns on
cash from operations performed for each industry. A negative correlation indicates that
industries with long operating cycles tend to have weaker associations between stock
returns and cash from operations. The results indicate a significant negative correlation
of -0.483 between the R2 from the cash flow regressions and the length of the
operating cycle. Similar results are found for the trade cycle. Earnings also exhibit
negative correlations, however they are not significant. This suggests that accruals play
a relatively more important role in firms with long operating cycles.28 The evidence in
table 13 supports hypothesis 5. The results are consistent with the length of the
operating cycle being an important economic determinant of the size of the change in
firms' working capital requirements and hence the ability of cash flows to measure finn
performance. The results also suggest that accruals mitigate this problem in earnings.
Earnings therefore reflect finn performance on a more timely basis than cash flows, for
firms in industries with long operating cycles.29
28 The volatility of firms' (industry) operating and trade cycle was also investigated since firms thathave large changes in the length of their operating cycle may also have large changes in their workingcapital requirements. However the relation is weak, the correlation between the volatility of theoperating cycle with the R2 from regressions ofretums on cash flows is -0.040 (probability=0.15) on afirm basis and -0.230 (probability=0.08) on an industry basis. While the correlation between thevolatility of the trade cycle with the R2 from regressions of returns on cash flows is -0.028(probability=0.33) on a firm basis and -0.144 (probability=0.28) on an industry basis. Thus, thevolatility of the operating and trade cycle does not explain cross-sectional variation in cash flows'.asscciation with stock return.
29 The analysis was also conducted on a firm basis for 1,252 firms that had at least 10 observations.The correlation for cash from operations between the firm-specific R2s and the operating cycle is-0.109 (probability=O.OOO) and the trade cycle -0.084 (probability=0.OO2). For earnings, the correlationbetween the R2s and the operating cycle is 0.049 (probability=0.078) and the trade cycle 0.080(probability=0.OO5).
76
9 • Summary and Conclusions
This study hypothesizes that the role of accounting accruals is to provide a
measure of short-term performance that more closely reflects expected future cash
flows than realized cash flows. The results support this prediction. First, earnings,
relative to cash flows, are more strongly associated with stock returns over short
intervals. In addition, the ability of cash flows to measure firm performance improves
relative to earnings as the measurement interval is lengthened. Second, earnings,
relative to cash flows, have a higher association with stock returns in firms
experiencing large changes in their working capital requirements and their investment
and financing activities. Under these conditions, cash flows suffer more greatly from
timing and mismatching problems and are less able to reflect finn performance.
This study also predicts that although accruals on average, improve earnings
association with stock returns, certain accruals are less likely to mitigate mismatching
problems in cash flows. Evidence is presented that indicates that long-term operating
accruals playa less important role in this respect. In addition, the inclusion of special
items in earnings is shown to reduce earnings' association with stock returns over short
intervals. However, this does not imply that special items should be excluded from
77
earnings from continuing operations. It may be important to include special items in
earnings so that management is accountable for such charges.
The contribution made by this study is to document the benefits of accrual
accounting and to provide an explanation for its prevalence. In particular, the study
helps explain why earnings are more often reported and used in contracts than cash
from operations or net cash flows. The approach taken here differs from that typically
adopted in previous research. This study assumes that cash flows are a more
'primitive' performance measure than earnings and the value added by accountants is in
accruing cash receipts and disbursements so as to attain a more useful measure of firm
performance over short measurement intervals. Therefore, the set of existing accruals
are expected, on average, to improve earnings' ability to measure firm performance.
The results demonstrate that isolating the conditions under which earnings and cash
flows are expected to differ by the greatest magnitude provides more powerful tests of
the usefulness of accounting accruals. Overall, the evidence suggests that accruals play
an important role in improving the ability of earnings to reflect firm performance on a
timely basis.
Finally, the study also provides new insights into two branches of existing
research. First, recent research has documented inter-industry variation in the
association between cash flows and stock returns [e.g. Biddle and Seow (1990)]. This
study identifies an underlying determinant of this variation. Specifically, the strength
of the association between cash flows and stock returns is shown to be negatively
related to the length of the operating cycle at the industry level. Second, this study
contributes to the long window analysis performed by Easton, Harris and Ohlson
(1992). They show that the association between earnings and stock returns improves
over longer measurement intervals. This result is consistent with earnings' suffering
78
from timing and mismatching problems over short measurement intervals. The
evidence in this study indicates that cash flows suffer more greatly than earnings in this
respect and that accruals help to mitigate these problems.
79
B:ibliography
American Institute of Certified Public Accountants, Accounting Principles Board,
"Reporting the Results of Operations," Accounting Principles Board Opinion
No.9 (1966).
American Institute of Certified Public Accountants, Accounting Principles Board,
"Reporting the Results of Operations - Reporting the Effects of Disposal of a
Segment of a Business, and Extraordinary; Unusual and Infrequently Occurring
Events and Transactions," Accounting Principles Board Opinion No. 30
(1973).
Atiase, R., (1985), "Predisclosure Information, Firm Capitalization and Security Price
Behavior Around Earnings Announcements," Journal ofAccounting Research,
23, 21-35.
Ball, R., (1972), "Changes in Accounting Techniques and Stock Prices," Empirical
Research in Accounting: Selected Studies," Journal ofAccounting Research,
10, (Supplement), 1-38.
80
Ball, R, (1989), "The Firm as a Specialist Contracting Intermediary: Application to
Accounting and Auditing," Working Paper, (University of Rochester,
Rochester, NY).
Ball, R and P. Brown, (1968), "An Empirical Evaluation of Accounting Numbers,"
Journal ofAccounting Research, 6, 159-78.
Ball, R. and S. P. Kothari, (1991), "Security Returns Around Earnings
Announcements," The Accounting Review, 66, 295-330.
Banker, R. and S. Datar, (1989), Sensitivity, Precision, and Linear Aggregation of
Signals for Performance Evaluation, Journal ofAccounting Research, 27, 21
39.
Barnea, A., J. Ronen and S. Sadan, (1975), The Implementation of Accounting
Objectives: An Application to Extraordinary Items, The Accounting Review,
50, 58-68.
Beaver, W., (1981), Financial Reporting: An Accounting Revolution, (Prentice Hall,
Englewood Cliffs, N.J.).
Beaver, W. and J. Demski, (1979), "The Nature of Income Measurement," The
Accounting Review, 54, 38-46.
Beaver, W. and R, Dukes, (1972), 'Interperiod Tax Allocation, Earnings
Expectations, and the Behavior of Security Prices," The Accounting Review,
47, 320-333.
Beaver, W. and W. Landsman, (1983), "Incremental Information Content of Statement
33 Disclosures," Financial Accounting Standards Board Research Report.
81
Beaver, W. and D. Morse, (1980), "What Determines Price-Earnings Ratios,"
Financial Analysts Journal, 36,65-76.
Bernard, V., (1987) "Cross-Sectional Dependence and Problems in Inference in
Market-Based Accounting Research," Journal ofAccounting Research, 25, 1
48.
Bernard, V., (1989) "Capital Markets Research in Accounting During the 1980's: A
Critical Review," in Thomas J. Frecka, Ed., The State ofAccounting Research
as We Enter the 1990's (Department of Accountancy, University of lllinois,
Urbana-Champaign, IL), 72-120.
Bernard, V. and T. Stober, (1989), "The Nature and Amount of Information in Cash
Flows and Accruals," The Accounting Review, 64, 624-652.
Bernstein, L., (1990), Analysis of Financial Statements, third ed., (Dow Jones-Irwin,
Homewood, 11.).
Biddle, G. and G. Seow, (1990), "The Incremental and Relative Information Content
Across Industries of Three Performance Measures: Net Income, Cash Flows,
and Sales," Working Paper, (University of Washington, Seattle, WA).
Black, F., (1991), "Choosing Accounting Rules," Working Paper, (Goldman, Sachs
& Co, New York, NY).
Bowen, R., D. Burgstahler and L. Daley, (1986), "Empirical Evidence on the
Relationship Between Earnings and Various Measures of Cash Flows," The
Accounting Review, 61, 713-25.
82
Bowen, R., D. Burgstahler and L. Daley, (1987), "The Incremental Information
Content of Accrual Versus Cash Flows," The Accounting Review, 62, 723
747.
Christie, A., (1987), "On Cross-Sectional Analysis in Accounting Research," Journal
ofAccounting and Economics, 9,231-258.
Clinch, G. and J. Magliolo, (1992), "Executive Compensation Effects of Accounting
Orientated Discretionary Transactions in Commercial Banks," Journal of
Accounting and Economics, (forthcoming).
Collins, D. and S.P. Kothari, (1989), "An Analysis of Intertemporal and Cross
Sectional Determinants of Earnings Response Coefficients," Journal of
Accounting and Economics, 11, 143-181.
Collins, D., S.P. Kothari, and J. Rayburn, (1987), "Firm Size and the Information
Content of Prices with Respect to Earnings," Journal of Accounting and
Economics, 9, 111-138.
Collins, D., S.P. Kothari, J. Shanken and R. Sloan, (1992), "Further Evidence on the
Relative Importance of Current versus Future Earnings in Explaining Current
Returns," Working Paper, (University of Rochester, Rochester, NY).
The Conference Board, (1990), Top Executive Compensation: 1990 edition (New
York, NY).
Davidson, R. and J. MacKinnon, (1981), "Several Tests for Model Specification in the
Presence of Alternative Hypotheses," Econometrica, 49, 781-793.
83
DeAngelo, L., (1988), "Managerial Competition, Information Costs, and Corporate
Governance: The use of Accounting Performance measures in proxy contests,"
Journal ofAccounting and Economics, 10,3-36.
DeAngelo, L., H. DeAngelo and D. Skinner, (1992), "Dividend Losses, and the
Quality of Earnings," Journal ofAccounting and Economics, (forthcoming).
Dechow P., M. Huson and R. Sloan, (1992), "The Effect of Restructuring Charges on
CEO Compensation," Bradley Policy Research Center, Managerial Economics
Research Studies Working Paper Series, MR 91-09.
Easton, P. and T. Harris, (1991), "Earnings as an Explanatory Variable for Returns,"
Journal ofAccounting Research, 29, 19-36.
Easton, P., T. Harris, and J. Ohlson, (1992), "Aggregate Accounting Earnings Can
Explain Most of Security Returns: The Case of Long Return Intervals," Journal
ofAccounting and Economics, 15, 119-142.
Easton, P. and M. Zmijewski, (1989), "Cross-sectional Variation in the Stock Market
Response to Accounting Earnings Announcements," Journal ofAccounting and
Economics, 11, 117-142.
Fama, E., (1976), Foundations of Finance, (Basic Books Inc., New York, N.Y.)
Fama, E. and K. French, (1992), "The Cross-Section of Expected Stock Returns," The
Journal ofFinance, 47, 427-65.
Fama, E. and M. Miller, (1972), The Theory ofFinance, (Holt, Rinehart and Winston,
New York, NY).
84
Foster, G., (1977), "Quarterly Accounting Data: Time-Series Properties and
Predictive-Ability Results," The Accounting Review, 52, 1-21.
Gaver, K. and M. Geisel, (1974), "Discriminating Among Alternative Models:
Bayesian and Non-Bayesian Methods," in P. Zarembka, ed., Frontiers in
Econometrics, (Academic Press, New York, NY), 49-77.
Hand, J., (1990), "A Test of the Extended Functional Fixation Hypothesis," The
Accounting Review, 65, 587-623.
Hand, J., (1991), "Extended Functional Fixation and Security Returns Around
Earnings Announcements: A Reply to Ball and Kothari," The Accounting
Review, 66, 739-746.
Healy, P., (1985), "The Effect of Bonus Schemes on Accounting Decisions," Journal
ofAccounting and Economics, 7, 85-107.
Healy, P. and K. Palepu, (1991), "Financial Reporting and Finn Value," Working
Paper, (Massachusetts Institute of Technology, Cambridge, MA).
Holmstrom, B., (1982), "Moral Hazard in Teams," Bell Journal of Economics, 13,
324-340.
Holthausen, R., (1990), "Accounting Method Choice: Opportunistic Behavior,
Efficient Contracting and Information Perspectives," Journal ofAccounting and
Economics, 12, 207-218.
Holthausen, R. and R. Leftwich, (1983), "The Economic Consequences of Accounting
Choice: Implications of Costly Contracting and Monitoring, Journal of
Accounting and Economics, 5, 7-118.
85
Jennings, R., (1990), "A Note on Interpreting 'Incremental Information Content'," The
Accounting Review, 65, 925-32.
Jones, C. and B. Bublitz, (1991), "Market Reactions to the Information Content of
Earnings over Alternative Quarters," Journal of Accounting, Auditing and
Finance. 10, 549-566.
Jones, J., (1991), "Earnings Management During Import Relief Investigations,"
Journal ofAccounting Research, 29, 193-228.
Kaplan, R. and R. Roll, (1972), "Investor Evaluation of Accounting Information:
Some Empirical Evidence," Journal ofBusiness, 45, 225-257.
Kieso, D. and J. Weygandt, (1989), Intermediate Accounting, (John Wiley & Sons,
New York, NY).
Kormendi, R. and R. Lipe, (1987), "Earnings Innovations, Earnings Persistence and
Stock Returns, Journal ofBusiness, 60,323-345.
Kothari, S. P., (1988), "An Analysis of Earnings to Price Ratios and Systematic
Risk," Working Paper, (University of Rochester, Rochester, NY).
Kothari, S. P., (1992), "Price-Earnings Regressions in the Presence of Price Leading
Earnings: Earnings Level versus Change Specifications and Alternative,
Deflators," Journal ofAccounting and Economics, 15,173-202.
Kothari, S. P. and R. Sloan, (1992), "Information in Prices about Future Earnings:
Implications for Earnings Response Coefficients," Journal ofAccounting and
Economics, 15, 143-172.
86
Lev, B., (1989), "On the Usefulness of Earnings and Earnings Research: Lessons and
Directions from Two Decades of Empirical Research," Journal ofAccounting
Research, 27 (Supplement), 153-192.
Lev, B. and S. Thiagarajan, (1991), "Financial Statement Information," Working
Paper, (University of California, Berkeley, CA).
Littleton, A., (1966), Accounting Evolution to 1900, (Russell and Russell, New York,
NY).
Livnat, J. and R. Zarowin, (1990), "The Incremental Information Content of Cash
Flow Components," Journal ofAccounting and Economics, 13,25-46.
Maddala, G, (1988), Introduction to Econometrics, (Macmillan Publishing Company,
New York, NY).
Mizon, G. and 1.Richard, (1986), "The Encompassing Principle and Its Application to
Testing Non-Nested Hypotheses in Empirical Econometrics,' Econometrica,
54, 657-678.
Neill, J., T. Schaefer, R. Bahnson and M. Bradbury, (1991), "The Usefulness of
Cash Flow Data: A Review and Synthesis," Journal ofAccounting Literature,
10, 117-150.
Nichols, D., (1974), "The 'Never-to-Recur Unusual Item" - A Critique of APB
Opinion No. 30," CPA Journal, March, 45-48.
Ohlson, J., (1991), "The Theory and Value of Earnings and Introduction to the Ball
Brown Analysis," forthcoming in Contemporary Accounting Research.
87
Patell, J., (1976), "Corporate Forecasts of Earnings Per Share and Stock Price
Behavior: Empirical Tests," Journal ofAccounting Research, 14,246-276.
Patell, J. and R. Kaplan, (1977), "The Information Content of Cash Flow Data
Relative to Annual Earnings," Working Paper, (Stanford University, Stanford,
CA).
Paton, W. and A. Littleton, (1940), An Introduction to Corporate Accounting
Standards, (American Accounting Association,Sarasota, FL).
Pollak, R. and T. Wales, (1991), "The Likelihood Dominance Criterion: A New
Approach to Model Selection," Journal ofEconometrics, 47, 227-242.
Pourciau, S., (1993), "Earnings Management and Nonroutine Changes," Journal of
Accounting and Economics, 16, (forthcoming).
Pratt, J., (1990), Financial Accounting, (Scott, Foresman and Company, Glenview,
ll.).
Rayburn, J., (1986), "The Association of Operating Cash Flow and Accruals with
Security Returns," Journal ofAccounting Research, 24, (Supplement) 112
133.
Reimann B., (1990), "A Session for Students of Shareholder Value Creation,"
Planning Review, 18, May-June, 42-44.
Schaefer, T. and M. Kennelley, (1986), "Alternative Cash Flow Measures and Risk
Adjusted Returns," Journal ofAccounting, Auditing and Finance, 1,278-87.
Schipper, K., (1989), "Earnings Management," Accounting Horizons, 3, 91-102.
~~---------- -
88
Skala, M., (1991), "U.S. Industry's Cash Flow Maintains Uptrend," Chemical Week,
148, 28.
Sloan, R., (1992a), "Accounting Earnings and Top Executive Compensation,"
Doctoral thesis, (University of Rochester, Rochester, NY).
Sloan, R, (1992b), "Accounting Earnings and Top Executive Compensation," Journal
ofAccounting and Economics, 16, (forthcoming).
Sloan, R, (1992c), "Is the Stock Market Fixated on Reported Annual Earnings?"
Working Paper, (University of Pennsylvania, Philadelphia, PA).
Smith C. and J. Warner, (1979), "On Financial Contracting: An analysis of Bond
Covenants," Journal ofFinancial Economics, 6, 117-161.
Smith C. and R Watts, (1982), "Incentive and Tax Effects of Executive Compensation
Plans," Australian Journal ofManagement, 7,139-157.
Smith C. and R Watts, (1992), "The Investment Opportunity Set and Corporate Policy
Choices," Journal ofFinancial Economics, (forthcoming).
Stem J., (1988), "Think Cash and Risk -- Forget Earnings per Share," Planning
Review, 16, 6-9,48.
Stickney, C; R. Weil and S. Davidson, (1990) Financial Accounting: An Introduction
to Concepts, Methods, and Uses, 6th Ed., (Harcourt, Brace Jovanovich, New
York, NY).
Vuong, Q., (1989), "Likelihood Ratio Tests for Model Selection and Non-Nested
Hypotheses," Econometrica, 57, 307-333.
89
Watts, R., (1977), "Corporate Financial Statements, a Product of the Market and
Political Process," Australian Journal ofManagement, 2, 53-78.
Watts, R. and J. Zimmerman, (1979), "The Demand for and Supply of Accounting
Theories: The Market for Excuses," The Accounting Review, 54, 273-305.
Watts, R. and J. Zimmerman, (1986), Positive Accounting Theory, (Prentice-Hall
Englewood Cliffs, N.J.).
Wilson, G. P., (1986), "The Relative Information Content of Accruals and Cash
Flows: Combined Evidence at the Earnings Announcement and Annual Report
Release Date, Journal ofAccounting Research, 24, (Supplement), 165-200.
Wilson, G. P., (1987), "The Incremental Information Content of the Accrual and
Funds Components of Earnings After Controlling for Earnings," Th e
Accounting Review, 62, 293-322.
90
Appendix 1.
The Relation Between Cash Flows and Earnings
This appendix provides a conceptual description of the changes in cash in terms
of all other (non-cash) accounts recorded on firms' balance sheets. It indicates the
types of accruals that are made to obtain cash from operations and earnings. The
relation between cash flows and earnings is discussed in most accounting texts books,
see for example, Stickney, Weil and Davidson [(1990), pp. 172-201] and Pratt
[(1990), pp. 747-749].
The basic accounting equation can be stated as (assets = liabilities +
stockholders' equity).
A=L+SE (A-I)
Total assets can be divided into cash (CASH) and non-cash assets (NA), and
shareholders' equity into contributed capital (CC) and retained earnings (RE).
CASH + NA = L + CC + RE (A-2)
Liabilities (L) can be separated into current liabilities (CL) and long-term
liabilities (LTL) while non-cash assets can be divided into current non-cash assets
(CNA) and long-term non-cash assets (LTNA)
CASH + CNA + LTNA = CL + LTL + CC + RE. (A-3)
Rearranging the above equation gives
CASH=CL+LTL+CC+RE -CNA-LTNA. (A-4)
91
Since the balance sheet equation holds at every point in time, it must also hold
between points in time. Therefore, the change in the cash balance can be expressed as:
ACASH =ACL + ALTL + ACC + ARE - ACNA - ALTNA. (A-5)
The change in retained earnings is equal to revenues (R) less expenses (EX),
less dividends (D), while the change in long-term liabilities can be separated into those
affecting cash (ALTLc) and those not affecting cash (ALTLnC). Similarly, the change in
long-term assets can be separated into those affecting cash (ALTNAc) and those not
affecting cash (ALTNAnc). Substituting and rearranging, the change in the cash
balance [net cash flows (NCF)] can be expressed in terms of changes in non-cash
accounts.
NCF =R - EX + ~CL - ~CNA + M;rLnc - ~TNAnc - ~TNAc + ~TI..c + ~CC - D
1 I' IEarnings Changes in
workingcapital
Changes innoncurrent accounts(e.g. depreciation,gains andlosses andspecial items)
I IPurchasesaxlretirementsoflongterm assets
1-----Long-termborrowings andrepayments, stockissuances andrepurchases anddividends
Cash from operations Cash from invesunent andfmancing activities
(A-6)
Thus, the difference between earnings and net cash flows is equal to the net
change in the balance of all non-cash accounts (AA). This is termed 'aggregate
accruals' in this study.
From the above equation, cash from operations (CFO) is equal to:
CFO =R - EX +ACL - ACNA +ALTLnc - ALTNAnc (A-7)
92
and the difference between earnings (E) and cash from operations each period is:
E - CFO = IJ.WC - !J.L'fLoc + !J.LTNAnc.
or,
E - CFO = IJ.WC + LTA.
(A-8)
(A-9)
This difference is equal to the change in working capital accruals (IJ.WC =
ACNA - IJ.CL) as well as long-term operating accruals (LTA =!J.LTNAnc - ALTLnd.
Long-term operating accruals include depreciation, amortization, gains and losses on
debt retirement. gains and losses on long-term assets retirements and other adjustments
generally classified as special items by the firm.
93
Appendix 2.
The Effect of Measurement Error on the Coefficient on
Earnings and Cash Flows
The role of earnings is to provide a measure of firm performance over a given
measurement interval. However, due to problems with information asymmetries
between management and other contracting parties, earnings are also required to be
objective and verifiable. As a result, earnings reflect realized and reasonably certain
cash flows related to the firm's activities during the measurement interval. The
implications of uncertain future cash flows for the period's performance are not
reflected in earnings. This can be viewed as earnings imperfectly measuring firm
performance. Therefore, we can describe earnings for a given firm i in period t as:
(A-lO)
where x is the expected net cash flows associated with the period's activities. Cash
flows can be viewed as a more 'noisy' signal of firm performance than earnings since
accruals are hypothesized to help reflect the cash flow implications of the finn's current
operating activities (by reducing timing and mismatching problems). Cash flows can
be defined as:
(A-ll)
where n is the effect of excluding the information in accruals. Assume initially that
cov(x,u) =0 and cov(x,n) =0.
Therefore in a regression of stock returns on earnings per-share scaled by price:
94
since u is assumed not to be value relevantcov(R,u)=0
then:
where Wit =eit + Uit - bUit·
This is a classical errors in variables problem since cov(w,E);t: 0
cov(w,E) = -bou2
(A-12)
(A-13)
d li cov(Rx)an p irn bRE = var(x) + var(u)
and similarly, the R2 on cash flows, RRC2
and for cash flows per-sharescaled by price:
plim bRC = B1 + ( au 2 + an2 + 2aun)/ax2
Where O'un is the covariance betweenUand n , which is initially assumed to be equal to
zero.
Therefore ~ > bRE > bRC.
The R2 will also decline becauseof the noise terms. The R2for earnings will be:
R 2 _ bRE2 var(x)RE - var(R)
bRC2 var(x)var(R)
95
If quintiles are formed on the basis of the magnitude of the change in net
accruals, then the coefficient on cash flows will decline across quintiles. This will
occur since on2 is expected to be larger where the change in accruals is large, and
smaller where the change in accruals is small. Now, when cov(u,n) =0, earnings
relation with stock returns will not vary with the magnitude of the change in net
accruals. However, if cov(u,n) > 0, firms that have large changes in accruals also have
more uncertainty associated with their expected future cash flows (or more 'noisy'
earnings), then the coefficient on earnings will also decline as n (the magnitude of the
change in accruals) increases, but less so than for the coefficient on cash flows.
96
Table 1
Descriptive statistics of data for the quarterly, annual and four-year intervals.
mean standard median upper lower percentdeviation quartile quartile negative
Quarterly (observations = 19,733)
Earnings(E)8 0.009 0.044 0.016 0.258 0.006 17.61Change in
0.033 -0.023 44.93workingcapital (AWC) 0.003 0.097 0.004Cash fromoperations (CFO)8 0.025 0.098 0.023 0.059 -0.009 30.32Net cash flows (NCF)8 0.000 0.058 0.000 0.017 -0.017 50.29Stock returns'' -0.007 0.160 -0.020 0.079 -0.111 55.60
Annual (observations =27,308)
Earnings(E)8 0.086 0.134 0.090 0.147 0.046 11.40Change inworkingcapital (AWC) 0.026 0.203 0.023 0.094 -0.030 36.84Cash fromoperations(CFO)8 0.138 0.234 0.107 0.222 0.025 19.13Net cash flows (NCF)8 0.016 0.110 0.004 0.044 -0.024 44.75Stock returns" 0.027 0.431 -0.030 0.213 -0.238 53.63
Four YearC (observations =5,175)
Earnings(E)8 0.405 0.453 0.383 0.638 0.175 11.90Change inworkingcapital (AWC) 0.145 0.451 0.086 0.267 -0.022 29.41Cash fromoperations(CFO)8 0.555 0.589 0.464 0.834 0.195 10.76Net cash flows (NCF)8 0.059 0.191 0.018 '0.098 -0.025 38.07Stock returns'' 0.172 1.203 -0.099 0.592 -0.581 54.24
a All financial statement variables are on a per share basis and scaled by beginning of period price.Observations for the quarterly intervalare from 1980 to 1989, the annual intervals from 1960 to 1989and for the four year interval from 1964 to 1989.b Stock returnsafter adjustingfor the CRSP value-weighted index.c The four-yearobservationsarenon-overlapping and are the cumulatedone year values per share (afteradjusting for the numberof commonsharesoutstanding) scaledby beginningperiod price.
97
Table 2
Tests comparing the association of earnings. the association of cash from operationsand the association of net cash flows with stock returns (adjusted for market-wide
movements) over varying measurement intervals.s
Rit =a. + ~ (X)it + £it
Independent variable (X)
Eamings(E)Cash from
Operations (CFO)Net Cash
Hows (NCF)
Quarterly
Intercept(t-statistic)
Coefficient(t-statistic)
Adjusted R2 (%)
Annual
Intercept(t-statistic)
Coefficient(t-statistic)
Adjusted R2 (%)
Four Year
Intercept(t-statistic)
Coefficient(t-statistic)
Adjusted R2 (%)
-0.013(10.40)
0.742(25.71)
3.24
-0.084(-29.58)
1.297(72.65)
16.20
-0.510(-29.42)
1.686(59.06)
40.26
-0.008 -0.007(-5.75) (-5.51)
0.022 0.036(1.70) (1.61)
0.01 0.01
R2CFO =0.003R2NCF = 0.003
R2E R2E
-0.017 0.018(-5.87) (6.86)
0.328 0.614(29.98) (26.31)
3.18 2.47
~=0.20R2NCF = 0.15
R E . R2E
-0.199 0.081(-9.19) (4.76)
0.675 1.56(25.16) (18.40)
10.88 6.12
R2c FO =0.27R2NCF = 0.15
R7E R7E
98
a Reported parameter estimates are from pooled regressions. Rit is the stock return adjusted for theCRSP value-weighted index for finn i calculated over the time interval 1, where t is equal to onequarter, one year or four years. E is earnings per share, CFO is cash from operations per share and NCFis net cash flows per share. All variables are scaled by beginning of period price. The four-year valuesare the cumulated one year values per share (after adjusting for the number of common sharesoutstanding) scaled by beginning period price. The total number of observations is 19,733 for quarterly,27,308 for annual and 5,175 for four years. Observations for the quarterly interval are from 1980 to1989, the annual intervals from 1960 to 1989 and for the four year interval from 1964 to 1989.
99
Table 3
Resultsof the likelihood ratio test developed by Vuong (1989) for non-nested modelselection. A significant positive Z-statistic indicates thatmodel2 is rejected in favorof
model l.s
Comparison ofModell vs Model 2:
Quarterly
Earnings vs Cash from operations
Earnings vs Net cash flows
Cash from operations vs Net cash flows
Annual
Earnings vs Cash from operations
Earnings vs Net cash flows
Cash from operations vs Net cash flows
Four Year
Earnings vs Cash from operations
Earnings vs Net cash flows
Cash from operationsvs Net cash flows
Z-statisticfor difference in
explanatory power''
7.60
7.62
0.06
24.16
23.27
2.27
17.04
18.03
3.67
Probability
« 0.001)
«0.001)
(0.954)
«0.001)
« 0.001)
(0.013)
« 0.001)
« 0.001)
«0.001)
a The number of observations in thepooled regressions is 19,733 for quarterly, 27,308 for annual and5,175 for four years. Observations for the quarterly interval are from 1980 to 1989, the annualintervals from 1960to 1989 andfor thefour yearinterval from 1964 to 1989.b For more detail on the test developed by Vuong (1989) see chapter 5. Reported probabilities arefrom a two-tailed test
Table 4
Comparison of earnings' and alternative cash flow measures' association with stock returns over the annual interval; 1,976observations, 1987-1989.a
Rit =0. + J3 (X)it + tit
Independent Variable (X)
Earnings CF095 CFO CF095+ InvCF FinCF Inv CF + NCFInvCF FinCF
Intercept -0.087 -0.103 -0.074 -0.047 -0.070 -0.047 -0.082 -0.048(t-statistics) (11.00) (-10.55) (-8.14) (-5.60) (-7.68) (-5.63) (-8.40) (-5.80)
Coefficient 1.016 0.478 0.283 0.019 -0.192 0.037 -0.300 0.458(t-statistics) (18.39) (10.46) (7.16) (0.65) (-6.17) (1.26) (-6.84) (5.70)
Adjusted R2 (%) 14.58 5.20 2.48 0.00 1.84 0.00 2.27 1.57
Vuong Z-statisticb
comparing: model 1vs model 2
Earnings vs cash flows measures 5.40 6.45 7.89 6.97 7.77 6.88 6.81
CF095vsCFO 2.76
a Reported parameter estimates are from pooled regressions. Rit is the stock return adjusted for the CRSP value-weighted index for ftrm i calculatedover the contemporaneous annualinterval. Earningsis earnings from continuing operations, CFOis cash fromoperations, NCFis net cash flows, all asdescribed in table2. CF095 is cash from operations as required by FAS95 (COMPUSTAT # 308), Inv CF is investment cash flows (COMPUSTAT #311) and is negative for investment outlays, Fin CF is cash inflows from ftnancing (COMPUSTAT # 313),. All variables are on a per sharebasis andscaled by beginning of period price.b A positiveZ-sratistic indicates thatmodel2 is rejected in favorof model I (all Z-statistics are signiftcant at the0.01 level). -8
101
Table 5
Finn-specific annual first-order autocorrelation coefficients and Pearson correlations forearnings, cash from operations, net cash flows and accrual measures.a
PanelA: Annualfirst-order autocorrelation coefficientsr'
Perfonnance measures Mean Median
dNet cash flow per share -0.523 -0.53960perating cash flow per share -0.434 -0.4396Earnings per share -0.175 -0.177
Accrual measures
Mggregate accruals per share -0.438 -0.4706Change in working capital per share -0.465 -0.472
Panel B: Firm specific Pearson correlations:Mean Median
corr(6Net cash flow per sharecjj, All accruals per sharej)
Quarterly -0.876 -0.952Annual -0.553 -0.602Four years -0.407 -0.698
corr(6Cash from operations per Sharet-l,b6Working capital per sharej)
QuarterlyAnnualFour years
-0.702-0.646-0.362
-0.720-0.672-0.616
corr(6Cash from operations per Sharet-l,t, 6Eamings per sharet-l,V
Quarterly 0.059Annual 0.132Four years 0.300
corr(6Net cash flow per sharet-l,b 6Earnings per sharet-l,J
0.0600.3980.545
QuarterlyAnnualFour years
0.3840.3520.230
0.4230.4110.469
a A sample of 595 firms (from a total of 1222) is used to calculatethe correlations and over thequarterly interval; each firm is required to have at least 16 quarterly observations. A sample of 827(from a total of 2761) rums is used to calculate the correlations and autoeorrelations over the annual
102
interval; each firm is required to have at least 16 annual observations. A sample of 715 firms (from atotal of 2024) are used to calculate the correlations over the four year interval; each firm is required tohave at least 4, four-year observations. Observations for the quarterly interval are from 1980 to 1989,the annual intervals from 1960 to 1989 and for the four year interval from 1964 to 1989.b To obtain fust-order autocorrelation coefficients the following regression is performed for each fum:~t,t+l= a + P~t-l,t + £t,t+l. Where X is earnings per share, net cash flows per share, cash fromoperations per share, change in working capital per share or aggregate accruals per share, ~t,t+1 is thechange in X from time interval t to t+1. All accrualsi is measured as the difference between earningsand net cashflows per share in time period 1.
103
Table 6
Tests comparing the association of earnings, cash from operations and net cash flowswith stock returns after including past and future realized values of those variables;
6,466 annual non-overlapping observations 1964 to 1988.a
Rit =a + ~IXit + Eit
Rit =a + ~ave(Xit + Xit-l + Xit+l)/3 + eu
Rit =a + ~IXit + ~2Xit·l + ~3Xit+l + Eit
a ~1 ~2 ~3 ~ave AdiR2
Cash from -0.043 0.446 4.79operations (-6.87) (18.06)
-0.070 0.637 5.63(-10.12) (19.67)
-0.071 0.386 0.281 0.02 6.73(-10.37) (20.06) (11.27) (0.78)
Net cash flows 0.009 0.737 3.18(1.67) (14.60)
-0.003 1.564 4.58(-0.59) (17.64)
-0.003 1.500 0.481 0.341 5.39(-0.65) (18.45) (10.10) (7.57)
1.490(36.74)
Earnings -0.125(-19.94)
-0.136(-20.02)
-0.137 1.253 -0.167 0.503(-20.62) (33.36) (-3.69) (11.79)
1.624(33.36)
17.26
14.68
19.11
a Rit is the fiscal stock return measured for year t adjusted for the CRSP value-weighted index. X iseither earningsper share,cash from operations per share or net cash flows per share. All variables arescaledby the price at t-I. Observations are obtained from everythirdyearbeginning in 1964.
104
Table 7
Testscomparing the association of earnings and the association of net cash flows withstock returns across quintiles, wherequintiles are formed based on the absolutevalueof aggregate accruals. Quintile 5 contains finn-observations with the largest absolute
valueof aggregate accruals,"
Rit = an + ~nNCFit + enit
Rit = a e + ~eEit + £eit
Quintile 1 = small absolute aggregate accruals, abs(AA)
Z-statisticTlDle Net cash flows Earnings fordifference in Probability
Interval ~n Adj R2(%) ~ Adj R2(%) explanatory power''
Quarterly
Quintile 1 1.60 3.44 1.69 3.55 0.58 0.561Quintile2 1.09 1.88 1.44 2.82 2.35 0.019Quintile3 0.67 1.17 1.19 3.03 3.28 0.001Quintile4 0.26 0.32 1.24 4.69 5.99 0.000Quintile5 -0.08 0.15 0.61 4.96 4.78 0.000
Annual
Quintile 1 2.37 16.20 2.50 16.84 1.72 0.085Quintile2 1.89 12.23 2.18 15.44 4.96 0.000Quintile3 1.47 8.76 1.91 14.49 6.18 0.000Quintile4 1.03 6.51 1.56 14.82 7.25 0.000Quintile5 0.14 0.24 0.97 21.98 15.65 0.000
Four Year
Quintile 1 2.50 27.25 2.09 26.99 -0.16 0.446Quintile2 1.96 18.45 1.83 22.66 1.78 0.074Quintile3 2.32 18.97 1.77 26.79 3.00 0.003Quintile 4 1.73 8.99 1.51 23.84 5.43 0.000Quintile5 1.13 2.83 1.29 31.78 10.37 0.000
a Rit is the stock return adjusted for the CRSP value-weighted index for firm i calculated over thetime interval t. where t is equal to one quarter, one year or four years. E is earnings per share, NCF isnet cash flows per share and Abs(AA) is the absolute value of the aggregate accruals per share. Allvariablesare scaledby beginning of periodprice.The four-year valuesare the cumulatedone year valuesper share (after adjusting for the number of common shares outstanding) scaled by beginning periodprice. The total number of observations is 19,733 for quarterly, 27,308 for annual and 5,175 for fouryears. Observationsfor the quarterly interval are from 1980 to 1989, the annual intervals from 1960 to1989and for the four year intervalfrom 1964 to 1989.b Vuong's (1989)Z-statisticcomparesearningsand net cash flowsas competingnon-nestedmodels. Asignificant positive Z-statistic indicates that net cash flow is rejected in favor of earnings. Reportedprobabilitiesfor the Z-statisticsare from a two-tailedtest
105
Table 8
Examination of proxies for underlying determinants of earnings response coefficientsacrossquintiles formed on the absolute valueof aggregate accruals; annual observations
1960-1989.a
Small abs(AA) Large abs(AA)
Quintile 1 Quintile 2 Quintile 3 Quintile 4 Quintile 5
Coefficient on 2.50 2.18 1.91 1.56 0.97Earnings
Predictedchange invalue across quintiles
Mean ValuesRisk
EJP + 0.054 0.058 0.076 0.091 0.077Leverage + 2.387 1.973 2.316 2.436 2.847MktIBook 2.325 1.880 1.553 1.184 0.862Size - 653.984 762.803 863.653 521.710 251.744
Growth Opportunities
R&D/Assets 0.038 0.034 0.031 0.028 0.024Capital explDepr 2.041 2.151 2.104 2.081 1.982Net/Gross 0.593 0.597 0.593 0.585 0.572
Median ValuesRisk
EJP + 0.055 0.065 0.085 0.109 0.140Leverage + 1.985 1.964 2.010 2.099 2.229MktIBook 1.600 1.588 1.349 1.028 0.741Size 88.241 125.843 116.736 64.017 27.696
Growth Opportunities
R&D/Assets 0.026 0.023 0.021 0.018 0.015Capital explDepr 1.522 1.685 1.718 1.696 1.609Net/Gross 0.592 0.597 0.591 0.582 0.572
a Quintiles are formed based on fum-year observations of the absolute value of aggregate accrualsAbs(AA). Size is measured as market value of equity at t-I. FJP is the earnings from t-2 to t-I toprice at t-I. Leverage is measured as assets at t-l scaled by book value of equity at t-I. MktIBook isthe firms market value of equity scaled by book value of equity both measured at t-I. R&D/Assets isresearch and development expenditures scaled by assets at t-I. Capital explDepr is capital expendituresscaled by depreciation expense. Net/Gross is the net book value of fixed assets scaled by the grossbook value of fixed assets.
106
Table 9
Tests comparing the association of earningsand the association of cash from operationswithstock returnsacrossquintiles, wherequintilesare formed based on the absolute
valueof thechange in working capital. Quintile 5 containsfirms with the largestabsolute changes in working capital.a
Rit =<:xc + ~cCFOit + £Cit
Rit =(Xe + ~eEit + Eeit
Quintile 1 = small absolute change in working capital, abs(L\We)
Z-stansucTime Cash from operations Earnings fordifference in Probability
Interval ~c Adj R2(%) ~e Adj R2(%) explanatorypower''
Quarterly
Quintile 1 0.81 1.98 1.16 2.81 1.49 0.135Quintile2 0.66 1.64 1.21 3.99 2.55 0.011Quintile3 0.51 1.74 1.01 3.72 2.50 0.012Quintile4 0.18 0.41 0.90 3.94 5.13 0.000Quintile 5 -0.05 0.16 0.60 3.71 3.88 0.000
Annual
Quintile 1 0.88 7.72 1.71 14.24 6.93 0.000Quintile 2 0.90 8.69 1.71 15.73 7.44 0.000Quintile3 0.76 8.24 1.57 16.54 9.00 0.000Quintile 4 0.59 7.34 1.47 19.08 10.24 0.000Quintile 5 0.11 0.79 1.02 18.19 15.09 0.000
Four Year
Quintile 1 1.16 27.19 2.12 45.69 5.92 0.000Quintile 2 1.05 26.06- 1.80 42.08 4.45 0.000Quintile 3 1.06 23.20 1.71 34.31 4.09 0.000Quintile4 0.85 19.01 1.70 45.22 9.52 0.000Quintile 5 0.28 2.25 1.42 34.46 11.37 0.000
a Rit is the stock return adjusted for the CRSP value-weighted index for firm i calculated over the timeinterval t, where t is equal to one quarter, one year or four years. E is earnings per share, CPO is cashflows from operations per share and Abs(6.WC) is the absolute value of the change in working capitalper share. All variables are scaledby beginning of periodprice. The four-yearvalues are the cumulatedone year values per share (after adjusting for the number of common shares outstanding) scaled bybeginning period price. The total numberof observations is 19,733 for quarterly, 27,308 for annual and5,175 for four years. Observations for the quarterly interval are from 1980 to 1989, the annualintervals from 1960 to 1989 and for the four year interval from 1964 to 1989.b Vuong's (1989) Z-statistic compares earnings and cash from operations as competing non-nestedmodels. A signlficant positive Z-statistic indicates that cash from operations is rejected in favor ofearnings. Reportedprobabilities for the Z-statisticsare from a two-tailedtest.
107
Table 10
Tests comparing the association of earnings and the association of cash from operationswith stock returns across quintiles, where quintiles are formed based on the absolute
value of long-term operating accruals. Quintile 5 contains firms with the largestabsolute long-term operating accruals.a
Rit = <Xc + ~cCFOit + £Cit
Rit = <Xe + ~eEit + £eit
Qulntile 1 = small absolute long-term operating accruals, abs(LTA)Z-statistic
Time Cash from operations Earnings for difference in Probability
Interval ~c Adj R2(%) ~e Adj R2(%) explanatory power''
Quarterly
Quintile 1 0.01 0.00 1.44 3.28 4.42 0.000Quintile 2 0.02 0.00 1.51 4.42 3.75 0.000Quintile 3 0.05 0.05 1.15 3.34 3.87 0.000Quintile 4 0.04 0.02 1.30 6.03 5.85 0.000Quintile 5 0.01 0.00 0.56 3.95 4.23 0.000
Annual
Quintile 1 0.43 2.24 1.88 12.68 9.32 0.000Quintile 2 0.38 2.21 2.01 17.66 13.45 0,000Quintile 3 0.35 . 2.34 1.82 18.49 9.25 0.000Quintile 4 0.36 3.17 1.60 20.65 12.51 0.000Quintile 5 0.28 4.00 0.96 20.46 14.10 0.000
Four Year
Quintile 1 1.07 11.18 2.37 45.00 8.44 0.000Quintile 2 0.79 7.88 2.09 44.75 6.99 0.000Quintile 3 0.89 10.17 2.08 44.01 9.39 0.000Quintile 4 0.98 18.04 1.67 48.51 10.14 0.000Quintile 5 0.55 8.86 1.30 34.39 7.95 0.000
a Rit is the stock return adjusted for the CRSP value-weighted index for fum i calculated over the timeinterval 1, where t is equal to one quarter, one year or four years. E is earnings per share, CFO is cashflows from operations per share and Abs(LTA) is the absolute value of long-term operating accruals pershare. All variables are scaled by beginning of period price. Long-term operating accruals are obtainedby deducting cash from operations and changes in working capital from earnings (see Appendix 1). Thefour-year values are the cumulated one year values per share (after adjusting for the number of commonshares outstanding) scaled by beginning period price. The total number of observations is 19,733 forquarterly, 27,308 for annual and 5,175 for four years. Observations for the quarterly interval are from1980 to 1989, the annual intervals from 1960 to 1989 and for the four year interval from 1964 to1989.b Vuong's (1989) Z-statistic compares earnings and cash from operations as competing non-nestedmodels. A significant positive Z-statistic indicates that cash from operations is rejected in favor ofearnings. Reported probabilities for the Z-statistics are from a two-tailed test.
108
Table 11
Testscomparing the association of earnings before taxes and the association of earningsbefore special itemsand taxes withstock returns over different measurement intervals.a
Quarterly
Earnings beforetax(including special items) Adj R2(%)
Earnings before taxand special itemsAdj R2 (%)
Vuong Z-statistic(probability)b
Numberof observations
Annual
Earnings before tax(including special items) Adj R2 (%)
Earnings before taxand specialitems Adj R2 (%)
Vuong Z-statistic(probability)b
Numberof observations
Four Year
Earnings before tax(including special items) AdjR2 (%)
Earningsbeforetaxand special itemsAdj R2(%)
Vuong Z-statistic(probability)b
Numberof observations
Sample of firmobservations with
special items
0.92
2.96
2.14(0.03)
2,535
6.34
9.34
2.77(0.01)
5,832
34.48
33.95
-0.73(0.47)
2,083
Sampleof finnobservations withno specialitems
1.31
18,181
18.70
22,815
41.07
3,369
109
a Stock returns are adjusted for the CRSP value-weighted index and calculated over the time interval t,where t is equal to one quarter,one year or four years. Earnings before tax and earnings before tax andspecial items are on a per share basis and scaled by beginning of period price. The four-year values arethe cumulated one year valuesper share (after adjusting for the numberof common shares outstanding)scaled by beginning period price. Observations for the quarterly interval are from 1980 to 1989, theannual intervals from 1960 to 1989and for the four year interval from 1964 to 1989.b Vuong's (1989) Z-statistic comparesearningsbefore tax and special items and earnings before tax ascompeting non-nested models. A significant positive Z-statistic indicates that earnings before tax isrejected in favor of earnings before tax and special items. Reported probabilities are from a two-tailedtest.
110
Table 12
Descriptive statistics on firm level operating and trade cycles (measured in days) andfirm level and industry level Pearson correlations between the absolute value of thechange in working capital and the length of the operating and trade cycle; annual
observations, 1960-1989.a
Panel A: Mean Standard Median Upper Lower NumberDeviation Quartile Quartile of Firms
Operatingcycle 146.02 72.99 138.03 184.62 96.35 1,252
Trade cycle 108.11 78.86 101.07 147.49 60.02 1,252
Panel B: Finn-level correlations (probabilities inparentheses)
(Number of observations = 1,252)
corr (abs(~WC),operating cycle) 0.187(0.000)
corr (abs(~WC),trade cycle) 0.163(0.000)
Panel C: Industry-level correlations (probabilities inparentheses)
(Number of observations = 58)
corr (abs(~WC),operatingcycle)
corr (abs(~WC),tradecycle)
0.405(0.002)
0.450(0.000)
a abs(~WC) is the absolute value of the change in non cash working capital per share scaled by beginningof period price. Industry operating cycles, trade cycles and abs(~WC) are calculated by (i) averaging thetime-series of fum specific values and (ii) taking the average of the fum specific values across two digitSIC classifications.
. I (ARt + ARt-I)I2) ( (Invj + Invt-I)12 )operating cye e = +Sales/360 Cost of goods sold/360
trade cle=(ARt + ARt-I)/2 ) + ( (Invt + Invt-l)12 ). (APt + APt-I)12 )cy Sales/360 Cost of goods sold/360 Purchases/360
Table 13
Pearson correlations (probabilities) between the R2s (from regressions performedby industry of stock returns on cash from operations or stock returns on earnings)
with the average length of the industry operating and trade cycles; annualobservations 1960-1989.a
Correlation between the R2 from 58 industry specific regressions ofstockreturnson cash from operations (CFO) with theaverage industry operating or trade cycle
III
corr (R2.operating cycle)
corr (R2.trade cycle)
-0.483(0.001)
-004.18(0.001)
Correlation between theR2 from 58 industry specific regressions ofstockreturnson eaminu (E) with the average industry operating or trade cycle
corr (R2.operating cycle)
corr (R2.trade cycle)
-0.083(0.538)
-0.012(0.926)
a The R2s for cash from operations are obtained by performing separate regressions of returns(adjusted for the CRSP value-weighted market return) on cash from operations for each of 58 twodigit SIC industry classifications. The R2s for earnings are obtained in a similar manner. Cashfrom operations and earnings are on a per share basis and scaled by beginning of period price. Theoperating cycle for each industry is obtained by first calculating firm-specific mean operatingcycles and then taking the average across firms in the same two digit SIC industry classifications.A similar procedure is used to obtain the trade cycle. The minimum number of observationsrequired per firm is 10.
. I (ARt + ARt-l)/2) ( (Invt + Invt-l)/2 )operating eyc e = +Salesl360 Cost of goods sold/360
tradec cle=(ARt + ARt-l)/2 ) + ( (Invt + Invt-l)/2 ) _ (APt + APt-l)/2 )y Sales/360 Cost of goods sold/360 Purchasesl360