Accounting Ethics

THE IMPACT OF SARBANES-OXLEY

ACT ON COSMETIC EARNINGS

MANAGEMENT

June Y. Aono and Liming Guan

ABSTRACT

This study examines the mitigating effect of Sarbanes-Oxley Act on

cosmetic earnings management, referred by Kinnunen and Koskela

(2003) as earnings manipulative behavior to round earnings such that

they result in an upward bias. This behavior reports income numbers to

achieve key cognitive reference points represented by N� 10k. Using

Benford’s law, our analysis compares the distribution of second digits in

reported annual net income for publicly listed US companies between a

2-year periods before and after the year 2002 when Sarbanes-Oxley Act

went into effect. Our empirical results suggest that, in the 2-year period

prior to the Act, there was evidence of cosmetic earnings management.

However, such behavior in manipulating net income has noticeably

decreased in the period after the Act. This finding is consistent with the

notion that Sarbanes-Oxley Act has a deterring impact on corporate

America’s manipulative behavior to report earnings that achieve certain

key reference points.

Research in Accounting Regulation, Volume 20, 205–215

Copyright r 2008 by Elsevier Ltd.

All rights of reproduction in any form reserved

ISSN: 1052-0457/doi:10.1016/S1052-0457(07)00212-3

205

INTRODUCTION

Since earnings have been regarded as one of the most important items in the financial reports to investors, analysts, boards, and senior executives, standard setters are very concerned with how earnings numbers are derived (Beaver, 1998). Following the debacle of Enron, earnings management has attracted extensive attention by regulators, accounting academics, and the investment community. A primary purpose of the Sarbanes-Oxley Act (hereafter SOX) is to protect investors by improving the accuracy and reliability of corporate disclosures and to restore investors’ confidence in the integrity of companies’ financial reporting (Lobo & Zhou, 2006). However, due to the recentness of SOX, there are only a handful of empirical studies examining its impact on mitigating managers’ behavior in manipulating earnings.

The purpose of this study is to examine the impact of SOX on deterring firms from reporting earnings that was rounded upward. Prior to SOX, research had reported that many companies reported earnings that tended to be rounded upward (Carslaw, 1988; Thomas, 1989; Van Caneghem, 2002; Kinnunen & Koskela, 2003; Skousen, Guan, & Wetzel, 2004; Guan, Skousen, & Wetzel, 2005). Kinnunen and Koskela (2003) referred to such earnings manipulative behavior as the cosmetic earnings management. Our study complements other recent studies that examined the mitigating effect of SOX on earnings management (see Cohen, Dey, & Lys, 2005; Lobo & Zhou, 2006). These studies used various accrual models to estimate manage- ment discretion over accounting choices and found that earnings management (measured by discretionary accruals) decreased in the post-SOX period as compared to the pre-SOX period.1 Our study differs from these studies in that we examine the rounding upward of income in an attempt to address cosmetic earnings management. One advantage of the method is that we do not have to estimate the potentially noisy abnormal accruals (Healy & Wahlen, 1999). Another appealing feature is that we can identify a large set of potential earnings manipulators without invoking specific assumptions about earnings management motivation or methods (Burgstahler & Dichev, 1997).

LITERATURE REVIEW AND DEVELOPMENT

OF HYPOTHESIS

SOX was passed by Congress as a result of the corporate scandals of companies such as Enron and WorldCom. SOX was designed to improve the reliability of financial reporting by requiring CEOs and CFOs of public

JUNE Y. AONO AND LIMING GUAN206

companies to certify the accuracy and completeness of the financial statements. SOX also imposes severe criminal penalties, fines, and other penalties on CEO/CFOs for issuing false statements and for securities fraud. Thus, SOX is expected to result in less biased and more conservative accounting practices.

Several studies have investigated the effects of the passage of SOX on financial reporting. Lobo and Zhou (2006) found a decrease in discretionary accruals after SOX and that companies incorporate losses more quickly than gains into income after SOX than in the years preceding SOX. Hence, this suggests aggressive accounting practices declined after SOX and resulted in the improvement of the quality of earnings. Using an accrual model to measure the extent of earnings management, Cohen et al. (2005) found that earnings management increased steadily during the period preceding SOX, but declined significantly after SOX.

Thomas (1989) proposed two general reasons why managers may engage in cosmetic earnings management. One reason relates to earnings numbers as key cognitive reference points in the eyes of financial statement users. The pricing phenomenon of ‘‘$1.99’’ in marketing suggests that consumers view a product priced at $1.99 to be significantly cheaper than a product priced at $2.00. Similarly, earnings of $698,000 may be perceived by investors to be much lower than $700,000. Therefore, if current income is perceived as being lower and changes the investors’ expectation of future earnings, then managers may have incentives to report income which is rounded upward. The use of budgeting, lending, and bonus and options contracts provides another reason why managers occasionally round earnings numbers upward. Due to uncertainty related to managers’ productive efforts, these contracts tend to be based on ex ante estimates and rounded to rough figures that emphasize the first digit in the contractual number (Carslaw, 1988). Thus, small changes in such contractual parameters may have a large cash flow effect (Thomas, 1989).

This study investigates the cosmetic earnings management of publicly listed US companies with the expectation that earnings are less likely to be overstated following SOX. We compare the rounding upward of earnings for the pre-SOX period compared to the post-SOX period by analyzing the proportion of zeros in the second digit of reported net income. Since an excess in the number of zeros as the second digit of reported earnings indicates an upward bias in the rounding of earnings, we expect that the magnitude of the upward bias diminishes after SOX. Thus, the primary hypothesis to be tested in this study is stated as follows:

Hypothesis. The degree of cosmetic earnings management is significantly lower after the Sarbanes-Oxley Act.

Impact of Sarbanes-Oxley Act on Cosmetic Earnings Management 207

SAMPLE AND RESULTS

The sample used in this study was obtained from Standard & Poor’s Research Insight database. The original sample included positive annual net incomes of publicly listed US firms from 2000 to 2004. We excluded net incomes with less than three digits because any truncation method used by Research Insight in reporting data to the nearest thousands of dollars could have unpredictable impact on our analysis of the distribution of the second digits. Our empirical analysis involved a comparison of the distribution of zero as the second digit of net incomes between the pre-SOX period (2000 and 2001) and the post-SOX period (2003 and 2004).2 Observations in year 2002 (the year SOX went into effect) were excluded because this was a transition period. In order to reduce the possibility of spurious statistical inference due to different sample sizes of the pre-SOX and post- SOX periods, we limited the pre-SOX years to 2000 and 2001.3 Our final sample consisted of 10,413 observations for the pre-SOX period and 9,809 observations for the post-SOX period.4

If managers manipulate net income upward so that the numbers achieve certain key reference points, denoted by N� 10k, we would expect to observe more zeros in the second digit or reported income. Benford (1938) developed formulas (see Appendix) for the distribution of naturally occurring numbers. This series of formulas for digit distribu- tion of naturally occurring numbers is known as Bendord’s law. Benford’s law applies to many types of data such as market values, net incomes, and daily trading on the NYSE. Nigrini (1994, 1996, 1997) also applied these formulas to population growth, taxes, and fraud detection.

Table 1 reports the results of distribution of digits 0–9 in the second place of earnings numbers for both the pre-SOX and post-SOX periods. In the pre-SOX period (years 2000 and 2001), the proportion of zeros as the second digit, expected to be 11.97 percent of the sample, is actually higher by 0.79 percent (Z-statistic ¼ 2.45, p ¼ 0.014 for two-tailed test and 0.007 for one- tailed test). The results also indicate a systematic lack of nines as the second digit of earnings. The proportion of nines, expected to be 8.5 percent of the sample, is actually lower by 0.55 percent (Z-statistic ¼ 1.99, p ¼ 0.047 for two-tailed test and 0.023 for one-tailed test). This result confirms the findings of Thomas (1989) using a sample of more recent time period. While the direction of the deviation of zeros and nines in the second place of earnings is the same as in Thomas (1989), the magnitude of the deviation is smaller. In particular, Thomas (1989, p. 776) documented an excess of

JUNE Y. AONO AND LIMING GUAN208

T a

b le

1 .

C o m p a ri so n o f D ev ia ti o n s o f D ig it s in

S ec o n d P la ce

o f P o si ti v e E a rn in g s b et w ee n th e P re -S O X

a n d th e P o st -S O X

P er io d .

D ig it

E x p ec te d

F re q u en cy

(% )

P re -S O X

P er io d (n ¼

1 0 ,4 1 3 )

P o st -S O X

P er io d (n ¼

9 ,8 0 9 )

C h a n g e fr o m

P re - to

P o st -

S O X

P er io d

O b se rv ed

fr eq u en cy

(% )

O b se rv ed

d ev ia ti o n

(% )

Z -s ta ti st ic s

O b se rv ed

fr eq u en cy

(% )

O b se rv ed

d ev ia ti o n

(% )

Z -s ta ti st ic s

D if fe re n ce

in d ev ia ti o n

(% )

Z -s ta ti st ic s

0 1 1 .9 7

1 2 .7 6

0 .7 9

2 .4 5 ��

1 1 .8 5

� 0 .1 2

0 .3 6

� 0 .9 1

1 .9 4 �

1 1 1 .3 9

1 1 .0 9

� 0 .3 0

0 .9 4

1 1 .4 8

0 .0 9

0 .2 7

0 .3 9

0 .8 5

2 1 0 .8 8

1 0 .3 3

� 0 .5 5

1 .7 8 �

1 0 .8 5

� 0 .0 3

0 .0 9

0 .5 1

1 .1 6

3 1 0 .4 3

1 0 .5 3

0 .1 0

0 .3 2

1 0 .8 5

0 .4 2

1 .3 6

0 .3 2

0 .7 2

4 1 0 .0 3

1 0 .3 3

0 .3 0

1 .0 1

9 .9 5

� 0 .0 8

0 .2 5

� 0 .3 8

0 .8 8

5 9 .6 7

9 .6 2

� 0 .0 5

0 .1 4

9 .7 0

0 .0 3

0 .0 7

0 .0 7

0 .1 5

6 9 .3 4

9 .2 6

� 0 .0 8

0 .2 6

9 .3 0

� 0 .0 4

0 .1 2

0 .0 4

0 .0 7

7 9 .0 4

9 .2 5

0 .2 1

0 .7 4

9 .3 6

0 .3 2

1 .1 0

0 .1 1

0 .2 5

8 8 .7 6

8 .8 8

0 .1 2

0 .4 0

8 .6 0

� 0 .1 6

0 .5 5

� 0 .2 8

0 .6 8

9 8 .5 0

7 .9 5

� 0 .5 5

1 .9 9 ��

8 .0 7

� 0 .4 3

1 .4 9

0 .1 2

0 .3 0

N o

te : P re -S O X

p er io d in cl u d es

y ea rs

2 0 0 0 a n d 2 0 0 1 , a n d p o st -S O X

p er io d in cl u d es

y ea rs

2 0 0 3 a n d 2 0 0 4 .

�� , � :

si g n ifi ca n t a t 0 .0 5 a n d 0 .1 0 , re sp ec ti v el y .

Impact of Sarbanes-Oxley Act on Cosmetic Earnings Management 209

1.09 percent of zeros and a lack of 0.76 percent of nines in the second place of earnings numbers.

The distribution of digits in the second place of earnings during the post- SOX period exhibits a noticeably different pattern from the pre-SOX period. In the post-SOX period (years 2003 and 2004), none of the digits in the second place of earnings significantly deviates from the proportions predicted by Benford’s law. Thus, the observed pattern of an excess of zeros and lack of nines in the pre-SOX period does not repeat in the post-SOX period. In the post-SOX period, the deviations of zeros and nines in the second places of earnings from the expected proportions are �0.12 and �0.43, respectively, and neither is statistically significant. This result suggests that in the 2-year period immediately after the SOX, cosmetic earnings management does not appear to be an apparent phenomenon.

The impact of the SOX on cosmetic earnings management is formally tested by a comparison of the degree of deviation of zeros and nines in the second place of earnings between the pre- and post-SOX periods. Table 1 also reports the result of the test. While the lack of nines in the second place of earnings decreases from the pre-SOX period to the post-SOX period (i.e., from 0.55 to 0.43 percent of the sample), the decrease is not statistically significant (Z-statistic ¼ 0.30). However, there is a significant decrease in the deviation of zeros in the second place of earnings from the pre-SOX period to the post-SOX period. While there is an excess of zeros in the pre-SOX period (0.79 percent of the sample), there is a decrease of zeros in the post- SOX period (�0.12 percent of the sample). The decrease of 0.91 percent is statistically significant (Z-statistic ¼ 1.94, pr ¼ 0.052 for two-tailed test and 0.026 for one-tailed test). Overall, our findings suggest that cosmetic earnings management has changed from significantly apparent in the pre- SOX period to statistically less apparent in the post-SOX period, and that the change is mostly due to the decrease of the deviation of zeros in the second place of earnings. This lends support to our hypothesis that SOX has a deterring impact on cosmetic earnings manipulative behavior.

SUMMARY AND CONCLUSIONS

Our empirical analysis involved comparing the distribution of digits in income numbers for all publicly listed US companies between the pre-SOX period (2000 and 2001) and the post-SOX period (2003 and 2004). Consistent with prior studies, we find that cosmetic earnings management is pervasive in the pre-SOX period. Specifically, there are significantly more

JUNE Y. AONO AND LIMING GUAN210

zeros and fewer nines in the second place of the earnings numbers. However, such earnings manipulative behavior has largely disappeared in the post- SOX period. Further analysis shows that the decrease in the magnitude of cosmetic earnings management is largely due to the decrease of zeros in the second place of earnings. We interpret this finding as being consistent with the notion that SOX has a deterring effect on corporate America’s earnings manipulative behavior to report earnings that achieve certain key reference points.

A limitation of the study is that it examines the impact of SOX on a particular type of earnings management: achieving threshold represented by N� 10k. Other types of earnings management to achieve certain thresholds include: (1) avoid losses, (2) avoid earnings decreases, and (3) avoid negative earnings surprises (see Burgstahler & Dichev, 1997; Degeorge, Patel, & Zeckhauser, 1999). While Degeorge et al. (1999) suggested that the earnings threshold hierarchy follows the above sequence, Brown and Caylor (2005) found that the hierarchy has reversed in a more recent time period and that for the period 1996–2002, avoiding negative earnings surprises is the most important earnings management incentive. Our study does not examine how cosmetic earnings management fits into the hierarchy nor does it examine whether SOX has an impact on these other types of threshold-achieving earnings manipulative behavior. It should also be noted that there are other types of earnings management. For example, income smoothing has been documented in many studies (Buckmaster, 2001). If SOX deters the cosmetic earnings management, it is likely that these other types of earnings manipulative behavior are also mitigated after the passage of the SOX. We encourage future studies to investigate these important issues.

NOTES

1. It should be noted that inferences drawn from these studies are a joint test of both incentives to manage earnings and the construct validity of the accrual models used to estimate managers’ accounting discretion. Beneish (1997) provided evidence that the accrual models have low detective ability even among firms whose behavior is extreme enough (i.e., GAAP violators) to warrant the attention of regulators. Thomas and Zhang (2000) found that the accrual models are of low power in detecting earnings management. Thus, to the extent that the models fail to correctly extract the discretionary portion from total accruals, the results from studies using the accrual models should be interpreted with caution. 2. While expanding the time length in the two periods would increase the sample

size, it also introduces the noises of the confounding factors. Restricting to a 2-year period may provide a reasonably clean test of the immediate impact of SOX on

Impact of Sarbanes-Oxley Act on Cosmetic Earnings Management 211

earnings management because no other significant federal regulations specifically targeting the quality of financial reporting were passed during this period. The research design of the long-term effect SOX on earnings management would, at the minimum, need to control for other confounding events and/or the history effect. 3. We also conducted the same analysis for the pre-SOX period for up to 5 years,

and found stronger evidence of the decrease in cosmetic earnings management in the post-SOX period. Specifically, although the magnitude of proportion of zero in the second place exhibited similar decrease from the various pre-SOX periods to the post-SOX period, the Z-statistics of the difference in the proportions are larger. Because larger sample size due to the longer pre-SOX windows would increase the Z-statistic used to measure the significance of changes in the observed proportion of digits between the two periods, the results using 2 years for the pre-SOX period are the most conservative. 4. Similar empirical analysis is also conducted on firms reporting losses. Of the

8,737 reported losses in the period of 2000 and 2001, there were 0.67 percent fewer zeros in the second place than expected (Z-statistic=1.92). Of the 5,935 reported losses in the period of 2003 and 2004, there were 0.77 percent fewer zeros in the second place than expected (Z-statistic=1.87). There was no significant change in the deviation of zeros in the second place of losses between the two periods. Other numbers (1–9) in the second place of losses did not exhibit significant deviation from the expected proportions in either period. This result suggests that firms reporting losses have also engaged in cosmetic earnings management. For example, when a firm’s true loss was �$2.01 million, the management may have rounded the loss down to, say, �$1.98 million so that the loss could be perceived to be much smaller than �$2.01 million (in magnitude). Such behavior did not seem to change significantly from the pre-SOX period to the post-SOX period. Thus, the deterring effect of SOX on cosmetic earnings management is observed only among firms reporting profits.

ACKNOWLEDGMENT

The authors are grateful to the participants of the 2006 conference of the Allied Academics in Reno, Nevada, for their helpful comments on an earlier version of this paper, as well as to the editor and two anonymous reviewers for their benevolent comments and constructive suggestions.

REFERENCES

Beaver, W. (1998). Financial reporting: an accounting revolution (3rd ed.). New Jersey: Prentice-

Hall, Inc.

Beneish, M. (1997). Detecting GAAP violation: Implications for assessing earnings manage-

ment among firms with extreme financial performance. Journal of Accounting and Public

Policy, 16(3), 271–309.

JUNE Y. AONO AND LIMING GUAN212

Benford, F. (1938). The law of anomalous numbers. Proceedings of the American Philosophical

Society, March, pp. 551–572.

Brown, L., & Caylor, M. (2005). A temporal analysis of quarterly earnings thresholds:

Propensity and valuation consequences. The Accounting Review, 80(2), 423–440.

Buckmaster, D. (2001). Development of income smoothing literature 1893–1998: A focus on the

United States. Amsterdam, The Netherland: Elsevier Science, Ltd.

Burgstahler, D., & Dichev, I. (1997). Earnings management to avoid earnings decreases and

losses. Journal of Accounting and Economics, 24(1), 99–126.

Carslaw, C. (1988). Anomalies in income numbers: Evidence of goal oriented behavior.

The Accounting Review, 63(2), 321–327.

Cohen, D., Dey, A., & Lys, T. (2005). Trends in earnings management and informativeness of

earnings announcements in the pre- and post-Sarbanes Oxley periods. Working Paper.

Northwestern University.

Degeorge, F., Patel, J., & Zeckhauser, R. (1999). Earnings management to exceed thresholds.

Journal of Business, 72(1), 1–33.

Fleiss, J. (1981). Statistical methods for rates and proportions (2nd ed.). Hoboken, NJ: Wiley.

Guan, L., Skousen, C., & Wetzel, T. (2005). Unusual patterns in reported earnings: Additional

evidence. Journal of Forensic Accounting, 6(2), 317–332.

Healy, P. M., & Wahlen, J. M. (1999). A review of the earnings management literature and its

implications for standard setting. Accounting Horizons, 13(4), 365–383.

Kinnunen, J., & Koskela, M. (2003). Who is Miss World in cosmetic earnings management?

A cross-national comparison of small upward rounding of net income numbers among

eighteen countries. Journal of International Accounting Research, 2(2), 39–68.

Lobo, G., & Zhou, J. (2006). Did conservatism in financial reporting increase after the

Sarbanes-Oxley act? Initial evidence. Accounting Horizons, 20(1), 57–73.

Nigrini, M. (1994). Using digital frequencies to detect fraud. The White Paper (April), 3–6.

Nigrini, M. (1996). A taxpayer compliance application of Benford’s law. The Journal of the

American Taxation Association, 18(1), 72–91.

Nigrini, M., & Mittermaier, L. (1997). The use of Benford’s law as an aid in analytical

procedures. Auditing: A Journal of Practice and Theory, 16(2), 52–67.

Skousen, C., Guan, L., & Wetzel, T. (2004). Anomalies and unusual patterns in reported

earnings: Japanese managers round earnings. Journal of International Financial

Management and Accounting, 15(3), 212–234.

Thomas, J. (1989). Unusual patterns in reported earnings. The Accounting Review, 64(4),

773–787.

Thomas, J., & Zhang, X. (2000). Identifying unexpected accruals: A comparison of current

approaches. Journal of Accounting and Public Policy, 19(4/5), 347–376.

Van Caneghem, T. (2002). Earnings management induced by cognitive reference points. British

Accounting Review, 34(2), 46–57.

APPENDIX. BENFORD’S LAW AND TEST

OF DEVIATIONS

Benford (1938) demonstrated that the expected distributions of naturally occurring numbers are skewed toward the number one for the first digit

Impact of Sarbanes-Oxley Act on Cosmetic Earnings Management 213

(i.e., left-most digit) and zero for the second digit. He then generalized this finding by formulating the approximated proportions or occurrence of a number as the first digit in a number series as follows:

proportion ða is the first digitÞ ¼ Log10ðaþ 1Þ � Log10ðaÞ (A.1)

Table A1 shows the expected occurrences of each digit in the first and second places.

Further, the expected proportion of a given number a as the first digit and the number b as the second digit can be found in the following relation:

Log10 aþ bþ 1

10

� � � Log10 aþ

b

10

� � (A.2)

Using the above equations and summing over all possible a values for any b value gives an overall expected proportion for b as the second digit. This equation is as follows:

proportion ðb is the second digitÞ ¼ X

Log10 aþ bþ 1

10

� ��

�Log10 aþ b

10

� �� ðA:3Þ

The expected proportion of the numbers in the third, fourth, fifth digit, and so on can be similarly derived. This series of formulas for digit

Table A1. Expected Frequency Occurrences for Each Digit in the First and Second Places.

Digit First Digit Expected Frequency (%) Second Digit Expected Frequency (%)

0 – 11.968

1 30.103 11.389

2 17.609 10.882

3 12.494 10.433

4 9.691 10.031

5 7.918 9.668

6 6.695 9.337

7 5.799 9.035

8 5.115 8.757

9 4.576 8.500

Source: Nigrini and Mittermaier (1997).

JUNE Y. AONO AND LIMING GUAN214

distributions of naturally occurred numbers have since been known as Benford’s law.

To test the null hypothesis of no managerial effort to round earnings, we compared the observed frequency for each number x in the second place of earnings numbers to the expected occurrences of the number as predicted by Benford’s law (Eqs. (A.1)–(A.3)). To perform a significance test of the observed deviations from the expected proportions, a normally distributed Z-statistic has been used:

Z ¼ jp� p0j � ð1=2nÞffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ðp0ð1� p0ÞÞ=n

p (A.4)

where p and p0 are the observed and expected proportions, respectively. The sample size is represented by n. The second term in the numerator is a correction term, and should be applied only when it is smaller than |p�p0| (Thomas, 1989). These Z-statistics would reject the null hypothesis at the 10, 5, and 1 percent level if their values exceed 1.64, 1.96, and 2.57, respectively.

In addition, a Z-statistic is used to test the difference in the deviation between the pre-SOX period and the post-SOX period. The formula used to calculate the Z-statistic is:

Z ¼ jpi � pjj � 1=2ð1=ni þ 1=njÞffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

p̄q̄ð1=ni þ 1=njÞ p (A.5)

where q̄ ¼ 1� p̄, p̄ ¼ ni=ðni þ njÞ, ni is the total observations in quarter i, nj

is the total observations in quarter j, pi=proportion of zero as the second digit in quarter i, and pj=proportion of zero as the second digit in quarter j. The formula is adapted from Fleiss (1981, p. 23).

Impact of Sarbanes-Oxley Act on Cosmetic Earnings Management 215