8/9/2019 Earnings and Operating Cash Flow
1/5
The Difference between Earnings and Operating Cash Flow 765
COMPUST AT for earnings before extraordinary items (data
18),
depreciation and
amortization expense (data 14), total assets (data 6). data for computing operating
cash tlow (data 1. 4, 5, 34, 106, 126, and 217), and data for computing control
variables (receivables, 2: inventory, 3; long-term
debt
9;
sales,
12; common shares,
25;
adjustment factor, 27; retained earnings. 36; sale of equity securities, 108;
capital expenditures, 128; auditor and audit opinion, 149; share price, 199; and
exchange listing, ZLIST) for the year the fraud was discovered and the preceding
year.' The sample consisted of 56 firms.
Data were examined for these firms for up to five years, from three years up
to the fraud discovery year to the year following the discovery year. The number
of firm-years varied during this period because of missing data, usually because a
firm did not exist in certain years prior or subsequent to the fraud discovery.
Sample sizes vary for fraud firms across tests reported in this study because of
missing data for one or more variables included in some of the tests. The
availability of earnings and cash fiow data does not guarantee that data for all
control variables are available.
The firms were distributed across 21 two-digit SIC industries, though those
with SIC codes from 3000 to 3999 w ere prominent (27 firms). Most of these were
technology firms. Event years were distributed across the sample period, with no
more than eight observations falling in the same calendar year. Fraud was
discovered for many of the firms shortly after they became publicly traded. Fraud
was discovered within the first three years of the listing period for 28 of the firms.
Twenty-three of the firms were deleted from the COMPUST AT nonresearch files
within three years of the discovery year. Sixteen firms were deleted from the non-
research files because of bankruptcy or liquidation. Another
23
of the firms merged
or went private. Only 17 of the firms continued to exist in 1993 as publicly traded
firms.
Variables and descriptive statistics
Earnings were measured as earnings before extraordinary items plus depreciation
and amortization expense. Depreciation and amortization were added to make the
earnings number comparable with operating cash flow . Expenditures for plant
assets are deducted in computing investing cash flow and do not enter the
computation of operating cash flow. Consequently, if all other operating
transactions were for cash, operating cash fiow would exceed earnings by the
amount of depreciation and amortization expense.
Operating cash fiow was computed as
O E
= NI, DAE, E,^G,+
T
{CL, - CL, ^)-
CA,
- CA, _ ,), (1)
where OCE, is operating cash fiow in year /, N I is earnings before extraordinary
items,
DAE is depreciation and amortization expense, E is equity in earnings, G is
gain (or loss) from sale of long-term assets, T is deferred taxes, CL is current
liabilities (less short-term debt), and CA is current assets (less cash and equiva-
lents).
Earnings (adjustment for depreciation and amortization is assumed in future
8/9/2019 Earnings and Operating Cash Flow
2/5
766 Contemporary Accounting Research
references) minus operating cash flow was the primary measure used in this study
(eamings-cash flow hereafter). Both eamings and cash flow were scaled hy
beginning-of-period assets to control for size. Om itting
E G
and
T
from the
equation had minimal effect on empirical results.
Other variahles examined in the analysis included financial variables and re-
lated items considered in prior studies (see Tahle 1). A list of these variables and
their computations is provided in Table 2. We selected these variables as being
representative of
the
variables considered in prior studies that were available from
COMPUST AT. The large numher of firms in our analysis precluded the inclusion
of other variahles.
Table provides descriptive statistics for continuous explanatory and control
variables for the sample of fraud firms and for a sample of 60,453 nonfraud firm-
years for 1974-93. The statistics for fraud firms are for 150 firm-years for the 3
years immediately prior to the year of fraud detec tion. Data were not availahle for
all fimi.s for all firm-years. The existence of a firm on COMPUSTAT in a
particular year did not guarantee that all variahles were available for that year. The
nonfraud sample included all firm-years for nonfraud firm s for which all variables
were availahle from COMPUSTAT annual industrial, full, and research files. These
files include essentially all publicly traded New York Stock Exchange (NYSE),
American Stock Exchange (AMEX), and over the counter (OTC) firms in the
United States, including those deleted from the nonresearch files because of
bankmptcy, liquidation, and merger.
Data also are provided in Table 3 for those firm-years in the fraud sample in
which the excess of eamings over cash fiows were the largest. Determining the
specific periods in which the frauds affected eamings or cash fiows and the
magnitudes of the effects is problematic. The discovery of fraud does not
necessarily result in specific identification of the period in which the fraud
occurred. Because many of the firms in our sample existed for a limited period as
publicly traded companies prior to the discovery, we assume the frauds affected
financial statements in the years immediately prior to discovery. We limit the
analysis to no more than three years prior to the discovery to ensure a reasonable
number of observations in which the effects of the fraud were likely to have
occurred. Using several years in the prediscovery period in the analysis will
understate the relation between fraud and accmals to the extent the fraud did not
affect the accmals throughout this period. Using the year in which the magnitude
ofthe accrual variahle was largest in the analysis assum es that the existence of high
accmals in any year is a signal of potential fraud. This assumption leads to an
overstatement of the relation only if a high accmal resulted for a fraud firm for
some reason other than the fraud. We examine both the three-year sample and the
maximum-year sample because of potential understatement or overstatement of
results. The maximum-value years were distributed across
the
prediscovery period.
Twenty-seven percent of the maximum years were three years prior to discovery,
41 percent w ere two years prior
to
discovery, and 32 percent were in the year prior
to discovery. ̂
8/9/2019 Earnings and Operating Cash Flow
3/5
The Difference between Eamings and Operating Cash Flow 767
TABLE 2
Variables included in current study
Variable
Computation
Eamings
Operating cash flow
Earnings - cash
Free cash flow
Amount of securities issued
Leverage
Auditor change
Audit opinion
Inventory
Change in inventory
Receivables
Change in receivables
Sales growth
Market retum
Retained eamings
Market value
Sales to assets
Assets
Sales
Years listed
Industry
Stock exchange listing
Eam ings before extraord inary items, + Depreciation,) /
Assets,, I
NI, + D, +
E, G,
T + CL,
- CL,
, -
CA, - CA, ^ )*
As defined above
[Operating cash flow^ - Cap exp,^,g , -
2 to < -
i)J
Current assets, _ ,
Equity securities issued,/ M arket value, _,
CurTent assets,
Long-term debt,/ Assets,
1
if
auditor, not equal to auditor, _
,
1 if opinion not unqualified
Inventory,/Sales,
Inventory,/Sales,)/ Inventory,_ |/Sales,_ 1) . ,
Receivables,/Sales,
Receivables, /Sales,) / Receivables, _ / Sales, _ ,)
Sales , /Sa les , . ,
Price, - Price, _ ,)/Pdce, _ , [adjusted for splits and divi-
dends]
Retained eamings,/Assets,
Market value^/Assets,
Sales,/Assets,
_ ^
Assets, [log of assets used in analysis]
Sales, [log of sales used in analysis]
Current year - first year on COMPUSTAT
1 if SIC > 2999 and
8/9/2019 Earnings and Operating Cash Flow
4/5
768
Contemporary Accounting Research
TABLE 3
Desctiplive statistics and comparisons for fraud and non-fraud firms
Variable
Eamings
Operating cash flow
Eamings-cash f low
Free cash flow
Equity securities
Leverage
Inventory
Change in inventory
Receivables
Change in receivables
Sales growth
Market retums
Retained earnings
Market value
Sales to assets
Assets
Sales
Fraud firms
(prediscovery)
N
Mean
84
-0 .069
0.154
-0 .262
94
538
238
1.143
228
1.349
1.426
394
77
1.742
1.730
229.256
222.563
15 0
Std dev
234
0.315
324
557
248
0.216
0.163
555
0.145
L748
594
9 9
0.418
3.142
95
456.809
379.742
Non-fraud
1
N
Mean
79
68
0.011
-0 .017
0.041
477
0.169
1.034
0.166
1.050
1.143
0.197
0.103
897
1.617
878.376
937.830
irms
: 60,453
Std dev
0.143
0.160
0.139
384
258
234
0.137
442
0.115
0.441
435
867
875
1.115
1.115
3,994.640
4.220.710
Fraud firms
(maximum)
N
Mean
0.112
-0 . 2 1 9
0.331
-0 .482
59
534
245
1.252
226
1.306
1.451
374
8
1.825
1.973
173.958
162.247
56
Std dev
0.174
347
353
575
0.182
0.213
0.181
675
0.152
645
524
1.031
396
3 533
1.125
403.085
295.107
/-value
0.231
- 4 . 3 1 3
4.363
- 4 . 3 4 0
2.114
2.808
4.132
1.937
4.256
1.694
4.708
2.140
-0 .604
2.662
1.730
-3 . 3 8 0
-3 .658
Notes:
Eamings and operating cash flows are scaled by beginning-of-year assets. Other variables are
defined in Table 2 . The prediscovery fraud sample inc ludes t lrm-yeans for the 3 years up
to Ihe year of frand discovery. The m aximum fraud sample includes only the prediscov-
ery firm-year in which the excess of eamings over cash flows was largest, (-values are for
comparisons of nonfraud firm-years with maximum-value firm-years for fraud firms, t
values assume unequal variances.
Significant differences between the fraud and nonfraud samples exist for most
of the variables in Table 3. The table reports i values that compare the maximum -
value fraud firm-yea rs with the nonfraud firm-years. The t values take into account
unequal variances between the samples and arc more conservative than tests that
assume equal variances. Similar results (not reported) occur for the three-year
sample, though the
i
values are slightly lower.
The excess of earnings over cash flows is significantly larger on average for
the fraud than the nonfraud firms. This difference results from the significantly
lower operating cash fiows of the fraud firm s. Earnings for the tw o samples are not
significantly different. The free cash flow measure also is significantly lower for
the fraud firms. Compared with nonfraud firms, the fraud firms, on average, issue
more equity securities, are more highly levered, have larger amounts of receivables
and inventories, have larger sales growth, have larger market retums and larger
market values relative to assets, and are sm aller in terms of assets and sales.
8/9/2019 Earnings and Operating Cash Flow
5/5
The Difference between Eam ings and Operating Cash Flow 769
Table 4 provides descriptive statistics for noncontinuous variables. Percentages
are based on firm-years in tbe fraud up to the discovery year) and nonfraud
sam ples. A larger percentage of firms in tbe fraud sample are OTC firms tban for
those in tbe non-fraud sample. A smaller percentage of fraud firms are AMEX
firms tban for those in tbe nonfraud sample. A larger percentage of fraud firms
experienced auditor changes than for tbe nonfraud sample. A larger percentage of
qualified opinions were issued for the fraud sample. Finally, though distributed
across many industries, fraud firms were grouped in the 3000-39 99 SIC codes to
an extent that was larger than tbe proportions for nonfraud firms. Indicator
variables are used for tbese variables in the empirical ana lysis. Using an indicator
variable for tbe 30 K) SIC code group resulted in stronger results tban using sepa-
rate indicator variables for two-digit SIC code groups within the 3000 group.
Table 5 p rovides a correlation matrix for continuous variables examined in the
study. Data are from tbe com bined sam ple of fraud and nonfraud firm-years
60,688 observations). Most of the correlations in the table are significant at the
0.01 level. Because of the large sample size, only correlations less than 0.02 are
insignificant at tbe 0.05 level. The correlations indicate that several variables are
correlated witb tbe excess of eaming s over casb flows; free cash fiows, inventory,
change in inventory, receivables, change in receivables, sales growth, and sales to
assets. These variables are likely to be proxying, at least partially, for the same
accrual construct as tbe excess of eam ings over cash flows. Consequently, they are
not likely to provide much additional information with respect to signaling fraud.
Most of the correlations in the table are low tbougb significant), other than tbose
among tbe accrual-related variables.
Table
6
reports statistics for
the
earnings and operating cash flow variables for
fraud firms across the five years, beginning tbree years prior to the fraud discovery.
Means and medians reveal tbe same pattem s. Eam ings are higher prior to the year
the frauds were discovered than after, wbile operating cash flows are lower in tbe
prior years. As a result, the earnings -cash flow variable is positive in tbe years up
to the fraud discovery eam ings are higher tban cash fiow). Tbe
test for mean
greater tban zero is significant in prediscovery years and insignificant in
postdiscovery years.
Table 7 com pares the distribution of tbe eam ings-cash flow variable for fraud
firms relative to the distribution for nonfraud firm-years. The table lists selected
percentiles of the distribution of the eamings-cash fiow variable for 60,453 firm-
years using all available CO MPU STA T observations for nonfraud firms from 1974
to 1993. These distributions indicate tbat a large percentage of observations for
fraud firms falls within the extreme upper percentiles of tbe distribution of all
COM PUST AT firm-years. Almost 93 percent of the fraud firms exbibit values of
eamings-cash flow above the median value for nonfraud fimis in at least one of tbe
prediscovery years. Over 64 percent of tbe fraud firms exbibit a value greater than
tbe ninetieth percentile, and over 5 percent exbibit a value greater than the ninety-
fifth percentile. These results are consistent with the hypothesis that the difference
between net income and operating cash flow is higb for fraud firms in prediscovery
years relative to tbe difference for a broad sample of firm-years.