Audit Quality of Second-Tier Auditors: Are All Created Equally? R. Mithu Dey and Lucy S. Lim The BRC Academy Journal of Business 4, no. 1 (2014): 1-26. http://dx.doi.org/10.15239/j.brcacadjb.2014.04.01.ja01 Web Appendix DOI: http://dx.doi.org/10.15239/j.brcacadjb.2014.04.01.wa01
Appendix 1 Equation 1 SAR = a 0 + a 1 E + a 2 LagE + a 3 BDO + a 4 MC + a 5 CR + a 6 E*BDO + a 7 E*MC + a 8 E*CR + a 9 LagE*BDO + a 10 LagE*MC + a 11 LagE*CR + a 12 BETA + a 13 E*BETA + a 14 LagE*BETA + a 15 G + a 16 E*G + a 17 LagE*G + a 18 LOSS + a 19 E*LOSS + a 20 lage*loss + a 21 PERSIST + a 22 E*PERSIST + a 23 LagE*PERSIST + a 24 TENURE + a 25 E*TENURE + a 26 LagE*TENURE + a 27 LnTA + a 28 E*LnTA + a 29 LagE*LnTA + year dummies + industry dummies + ε a ; Equation 2 TotAcc = b 0 + b 1 ((1+k)Ch_Sales-Ch_Rec) + b 2 PPE + b 3 LagTotAcc + b 4 Gr_Sales + b 5 ROA + ε b where k is estimated from: Ch_Rec t = c + k Ch_Sales t + ε k. Equation 3 Ch_Rec t = c + k Ch_Sales t + ε k. Equation 4 ABSDA = c 0 + c 1 BDO + c 2 MC + c 3 CR + c 4 ABSAC + c 5 DA + c 6 MB + c 7 LOSS + c 8 LogTA + c 9 CHNI + c 10 CFLOW + ε c 2
Exhibit 1 Variable Definitions (Compustat data codes are indicated in italics in parentheses.) Earnings Response Coefficient (ERC) Analysis SAR E is the size-adjusted return over the period from the beginning of the fourth month of the fiscal year until the end of the third month of the following fiscal year, where SAR is defined as R i R p, R i is the firm s buy-and-hold stock return, and R p is the buy-and-hold return on a size-matched portfolio, both measured over the aforementioned period; is the current fiscal year earnings before extraordinary items and discontinued operations (IB), scaled by beginning of year market value of common equity (PRCC*CSHO); LagE is earnings before extraordinary items and discontinued operations for the previous fiscal year scaled by beginning of current year market value of equity; BETA is a measure of systematic risk, estimated as the slope coefficient from a regression of daily firm stock returns on daily equally-weighted CRSP NYSE-AMEX-NASD market index returns over the year from the beginning of the fourth month of the current fiscal year to the end of the third month of the following fiscal year, using a minimum of 100 observations, G is a measure of growth opportunities estimated as the ratio of the market value of common equity to the book value of common equity (CEQ) at the beginning of the fiscal year, LOSS is a dummy variable coded as one (zero) in loss (profitable) years, PERSIST is a dummy variable coded as one for firm-years with absolute values of earnings changes deflated by beginning-of-period market value of equity equal to or below the sample median and zero otherwise, and TENURE is the number of consecutive years an auditor has been auditing the company (AUOP), LnTA is the natural logarithm of the total assets (AT), BDO is a dummy variable coded as one if a firm is a BDO client and zero otherwise, MC CR is a dummy variable coded as one if a firm is a McGladrey client and zero otherwise, is a dummy variable coded as one if a firm is a Crowe Horwath client and zero otherwise. 3
Abnormal Accruals Analysis TotAcc is the income before extraordinary items (IBC) minus cash flow from operations (OANCF) scaled by average total assets (AT) for the year, Ch_Sales is net sales (SALE) for the current fiscal year minus net sales for the previous fiscal year scaled by average total assets for the current year, Ch_Rec is the change in net accounts receivable for the year as reported in the cash flow statement (RECCH) scaled by average total assets, PPE is the net property, plant and equipment at the end of the year (PPEGT) scaled by average total assets, LagTotAcc is TotAcc for the prior period, GrSales is percent sales (SALE) growth for the following year relative to the current year, ROA is income before extraordinary items (IBC) divided by average total assets, ABSDA is absolute value of discretionary accruals estimated by the modified Jones model (Dechow et al., 2003), ABSAC is absolute value of total accrual deflated by total asset at t-1, where total accrual as in modified Jones (Dechow et al., 2003) is [Income before Extraordinary Items (IBC) Cash Flow from Operations (OCF)], DA MB is the ratio of long-term debt (DLTT) to total assets (AT), is the ratio of market value of equity (CSHO*PRCC) to book value of equity (CEQ), LOSS is 1 if the firm reported as loss in year t, and 0 otherwise, LogTA is the natural logarithm of total assets (AT), CHNI is 1 if the absolute value of change in net income before extraordinary items and discontinued operations (IB) is in the top two deciles, and 0 otherwise, where change in net income equals net income in year t minus net income in year t-1, CFLOW is the ratio of cash flow from operations (OANCF) to total assets (AT), BDO is defined above, MC CR is defined above, is defined above. 4
Table 1 Sample Selection Panel A: differences in earnings response coefficients (ERCs) Observations with necessary data on Compustat, excluding of non-2 nd tier auditors Less observations with missing CRSP data or representing multiple classes of the same security Less observations deleted as potential extreme values due to: 6,713-2,173 Beginning-of-year price being below $1, and - 213 the magnitude of market-value-scaled earnings changes exceeding 150% - 40 Final sample for tests of Hypothesis 1 4,287 Panel B: differences in abnormal accruals Observations with necessary data on Compustat for the abnormal accrual analysis, and excluding financial firms, regulated firms and of non-2nd-tier auditors Less observations deleted as potential extreme values (values below one percentile and above 99 percentile of the sample distribution) for the calculation of Dechow et al. (2003) model 4,447-446 Less observations deleted when the number of observations in each 2-digit SIC code and year is less than 10 in the Dechow et al. (2003) model -1,279 Less observations deleted as potential extreme values (values more than four standard deviation from the mean of each continuous variable) for the multivariate analysis of abnormal accruals -90 Final sample for tests of Hypothesis 2 2,632 5
Table 2 Descriptive Statistics Panel A: Earnings response coefficient analysis (N= 4,287) Mean Median Std. 25% 75% Deviation SAR 0.0062-0.0982 0.5979-0.3425 0.2069 E -0.0278 0.0292 0.2220-0.0660 0.0719 lage -0.0323 0.0310 0.2308-0.0515 0.0664 BETA 0.9397 0.8497 0.7216 0.3800 1.4247 G 2.7505 1.6477 3.8441 1.0148 2.8543 TENURE 4.9100 4.0000 4.0227 2.0000 6.0000 LnTA 5.0686 4.9319 1.6655 3.8548 6.2168 Categorical variables Coded 1 Frequency/percentage Coded 0 Frequency/percentage LOSS 1,652/38.56% 2,634/61.44% PERSIST 2,147/50.08% 2,140/49.92% Panel B: Abnormal accrual analysis (N=2,632) Mean Median Std. 25% 75% Deviation ABSDA 0.0612 0.0450 0.0556 0.0202 0.0859 ABSAC 0.1063 0.0745 0.1057 0.0352 0.1414 DA 0.1196 0.0224 0.1853 0.0000 0.1775 MB 2.4926 1.7583 7.6973 0.9727 3.1020 LogTA 4.2039 4.1844 1.5042 3.1071 5.2015 CFLOW 0.0088 0.0514 0.1992-0.0367 0.1176 Categorical variables Coded 1 Frequency/percentage Coded 0 Frequency/percentage LOSS 1,249/47.45% 1,383/52.55% CHNI 156/5.93% 2,476/94.07% Note: Variable definitions are included in Exhibit 1. 6
Table 3 Correlations Panel A: ERC Analysis (N= 4,287) SAR E lage BETA G LOSS PERSIST TENURE LnTA SAR 1 0.22067-0.02912-0.00663-0.05228-0.18963-0.06427 0.02046-0.05468 <0.0001 0.0566 0.6645 0.0006 <0.0001 <0.0001 0.1804 0.0003 E 0.34247 1 0.52106-0.02226 0.03578-0.62413 0.23078 0.04549 0.11801 <0.0001 <0.0001 0.1451 0.0191 <0.0001 <0.0001 0.0029 <0.0001 lage 0.08035 0.55147 1-0.02658 0.01698-0.35963 0.23623 0.04186 0.1397 <0.0001 <0.0001 0.0819 0.2663 <0.0001 <0.0001 0.0061 <0.0001 BETA -0.05723-0.03286-0.05986 1 0.14769 0.03136 0.02327-0.0233 0.26548 0.0002 0.0314 <0.0001 <0.0001 0.0401 0.1276 0.1273 <0.0001 G -0.07186 0.02211-0.03198 0.2225 1 0.06638 0.13408-0.04062-0.19615 <0.0001 0.1477 0.0362 <0.0001 <0.0001 <0.0001 0.0078 <0.0001 LOSS -0.26966-0.84305-0.49386 0.0151-0.09279 1-0.31041-0.04041-0.25705 <0.0001 <0.0001 <0.0001 0.3231 <0.0001 <0.0001 0.0081 <0.0001 PERSIST 0.01953 0.21682 0.22538 0.03568 0.26139-0.31041 1 0.06081 0.11551 0.2011 <0.0001 <0.0001 0.0195 <0.0001 <0.0001 <0.0001 <0.0001 TENURE 0.01638 0.05804 0.07618-0.00825-0.0356-0.03429 0.06023 1 0.04012 0.2835 0.0001 <0.0001 0.5893 0.0197 0.0247 <0.0001 0.0086 LnTA -0.0109 0.21916 0.25409 0.29594-0.15297-0.26155 0.121 0.03173 1 0.4755 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 0.0377 Panel B: Abnormal Accrual Analysis (N=2,632) ABSDA ABSAC DA MB LogTA CFLOW LOSS CHNI ABSDA 1 0.3833-0.0292 0.0525-0.2027-0.1987 0.18836 0.0192 <0.0001 0.1349 0.0071 <0.0001 <0.0001 <0.0001 0.3256 ABSAC 0.2582 1 0.1168-0.0527-0.2381-0.1558 0.3940 0.1072 <0.0001 <0.0001 0.0068 <0.0001 <0.0001 <0.0001 <0.0001 DA -0.0500 0.0698 1-0.0529 0.1422-0.0058 0.0660 0.0973 0.0103 0.0003 0.0067 <0.0001 0.7675 0.0007 <0.0001 MB 0.0283-0.0693-0.1214 1-0.0021-0.0744 0.0001-0.0094 0.1469 0.0004 <0.0001 0.9152 0.0001 0.9730 0.6304 LogTA -0.1652-0.2134 0.1490 0.0211 1 0.3596-0.2853 0.3662 <0.0001 <0.0001 <0.0001 0.2793 <0.0001 <0.0001 <0.0001 CFLOW -0.1258-0.0049-0.0328 0.0411 0.3064 1-0.5077 0.0534 <0.0001 0.8003 0.0921 0.0349 <0.0001 <0.0001 0.0061 LOSS 0.1608 0.3749 0.0507-0.1105-0.2281-0.5868 1 0.0321 <0.0001 <0.0001 0.0094 <0.0001 <0.0001 <0.0001 0.0994 CHNI 0.0317 0.0629 0.0981-0.0102 0.3237 0.0627 0.0321 1 0.1036 0.0012 <0.0001 0.6014 <0.0001 0.0013 0.0994 Note: Variable definitions are in Exhibit 1. Pearson correlations are reported above the diagonal and Spearman correlations below the diagonal. 7
Table 4 Univariate Comparisons Panel A: ERC Analysis Sub-Panel I: Grant Thornton versus other (N for GT=2,029 and N for other =2,258) GT 8 GT SAR 0.0377-0.0221 3.27* -0.0782-0.1132 3.53* E -0.0181-0.0366 2.74* 0.0335 0.0240 2.32* lage -0.0275-0.0365 1.29 0.0338 0.0272 1.77 Beta 0.9686 0.9137 2.49* 0.8889 0.8133 2.93* G 2.7748 2.7286 0.39 1.7009 1.6087 2.24* TENURE 4.9566 4.8680 0.72 4.0000 4.0000 0.96 LnTA 4.9624 5.1641-4.00* 4.8363 5.0828-4.26* Categorical variables LOSS PERSIST GT 1 0 1 0 733/ 1,296/ 920/ 1,338/ 36.13% 63.87% 40.74% 59.26% 1,023/ 1,006/ 1,124/ 1,134/ 50.42% 49.58% 49.78% 50.22% Sub-Panel II: BDO versus other (N for BDO =1,452 and N for other =2,835) BDO BDO differences in proportions -3.10* 0.42 SAR 0.0046 0.0070-0.13-0.1030-0.0968-1.10 E -0.0396-0.0218-2.49* 0.0165 0.0346-3.95* lage -0.0424-0.0270-2.07* 0.0164 0.0363-5.48* BETA 1.0021 0.9078 4.01* 0.9062 0.8149 3.83* G 3.1344 2.5538 4.25* 1.7894 1.5860 3.77* TENURE 5.4470 4.6349 5.97* 4.0000 4.0000 5.21* LnTA 4.9465 5.1312-3.38* 4.8413 4.9858-3.16* Categorical variables LOSS PERSIST BDO 1 0 1 0 637/ 815/ 1,016/ 1,819/ 43.87% 56.13% 35.84% 64.16% 683/ 769/ 1,464/ 1,371/ 47.04% 52.96% 51.64% 48.36% differences in proportions 5.11* -2.85*
Sub-Panel III: McGladrey versus other (N for MC=454 and N for other =3,833) MC MC SAR -0.0116 0.0083-0.78-0.0991-0.0982 0.17 E -0.0179-0.0290 1.01 0.0260 0.0294 0.09 lage -0.0343-0.0320-0.18 0.0271 0.0314-0.22 BETA 0.7771 0.9590-5.09* 0.6472 0.8715-5.47* G 2.3399 2.7991-3.47* 1.5490 1.6643-1.25 TENURE 4.2489 4.9883-4.22* 3.0000 4.0000-3.76* LnTA 4.7874 5.1020-3.53* 4.6380 4.9631-4.26* Categorical variables LOSS PERSIST MC 1 0 1 0 182/ 272/ 1,471/ 2,362/ 40.09% 59.91% 38.38% 61.62% 229/ 225/ 1,918/ 1,915/ 50.44% 49.56% 50.04% 49.96% Sub-Panel IV: Crowe Horwath versus other (N for CR=352 and N for other =3,935 CR 2 nd tier CR differences in proportions 0.71 0.17 SAR -0.1458 0.0198-5.72* -0.1555-0.0912-4.72* E -0.0482-0.0260-1.40 0.0479 0.0269 2.50* lage -0.0150-0.0338 1.28 0.0557 0.0278 6.47 BETA 0.7253 0.9589-6.27* 0.6001 0.8681-5.81* G 1.5558 2.8573-10.30* 1.2379 1.7196-9.17* TENURE 3.2784 5.0559-15.75* 3.0000 4.0000-6.52* LnTA 6.5478 4.9363 22.86* 6.7576 4.8191 17.97* Categorical variables LOSS PERSIST CR 1 0 1 0 101/ 251/ 1,552/ 2,383/ 28.69% 71.31% 39.44% 60.56% 212/ 140/ 1,935/ 2,000/ 60.23% 39.77% 49.17% 50.83% differences in proportions -3.97* 3.97* 9
Panel B: Abnormal Accrual Analysis Dechow et al. (2003) accrual model Sub-Panel I: Grant Thornton versus other (N for GT =1,441 and N for other =1,322) GT GT TotAcc -0.0863-0.0989 2.05* -0.0644-0.0597 0.32 Ch_Sales 0.0592 0.0423 1.53 0.0573 0.0465 1.06 Ch_Rec 0.0084 0.0096-0.50 0.0049 0.0064-0.38 PPE 0.4741 0.4292 3.08* 0.3323 0.3207 2.84* LagTotAcc -0.0835-0.0957 2.02* -0.0626-0.0620 0.96 GrSales 0.1005 0.0977 0.19 0.0551 0.0571-0.70 ROA -0.0700-0.1132 3.90* 0.0169-0.0144 5.59* Sub-Panel II: BDO versus other (N for BDO=1,012 and N for other = 1,751) BDO BDO TotAcc -0.1095-0.0825-4.00* -0.0632-0.0617-2.32* Ch_Sales 0.0311 0.0628-2.77* 0.0446 0.0566-1.80* Ch_Rec 0.0085 0.0092-0.28 0.0057 0.0054-0.59 PPE 0.4234 0.4695-3.03* 0.3118 0.3360-3.30* LagTotAcc -0.1060-0.0797-4.06-0.0672-0.0608-3.00* GrSales 0.1044 0.0961 0.53 0.0574 0.0554 1.05 ROA -0.1339-0.0656-5.72* -0.0256 0.0141-7.02* Sub-Panel III: McGladrey versus other- (N for MC=255 and N for other =2,508) MC 2 nd tier MC TotAcc -0.0686-0.0948 3.06* -0.0578-0.0625 1.93* Ch_Sales 0.0834 0.0479 1.86 0.0661 0.0520 1.66 Ch_Rec 0.0129 0.0086 1.09 0.0069 0.0050 1.33 PPE 0.4285 0.4551-1.05 0.3226 0.3285-0.51 LagTotAcc -0.0690-0.0914 2.51* -0.0503-0.0632 1.86 GrSales 0.0926 0.0998-0.33 0.0644 0.0550 0.51 ROA -0.0580-0.0940 2.35 0.0032 0.0039 0.85 10
Sub-Panel IV: Crowe Horwath versus other (N for CR= 55 and N for other = 2,708) CR 2 nd tier CR TotAcc -0.0454-0.0933 3.40* -0.0317-0.0631 2.86* Ch_Sales 0.0595 0.0510 0.21-0.0036 0.0541-1.01 Ch_Rec 0.0130 0.0089 0.61 0.0070 0.0055 0.67 PPE 0.5378 0.4509 1.66 0.4696 0.3267 2.29* LagTotAcc -0.0286-0.0906 4.94* -0.0317-0.0631 3.05* GrSales -0.0018 0.1012-3.28* -0.0028 0.0568-2.18* ROA 0.0131-0.0927 4.48* 0.0229 0.0030 2.46* Panel C: Multivariate test of differences in discretionary accruals Sub-Panel I: Grant Thornton versus other (N for GT=1,381 and N for other =1,251) GT GT ABSDA 0.0609 0.0616-0.32 0.0437 0.0466-0.95 ABSAC 0.1048 0.1078-0.72 0.0748 0.0738 0.11 DA 0.1099 0.1302-2.79* 0.0182 0.0338-2.27* MB 2.3764 2.6210-0.81 1.7389 1.7806-0.27 LogTA 4.2915 4.1070 3.14* 4.3249 4.0489 3.84* CFLOW 0.0243-0.0085 4.23* 0.0656 0.0352 5.93* Categorical variables LOSS CHNI GT 1 0 1 0 590 791 659 592 42.72% 57.28% 52.68% 47.32% 79 1,302 77 1,174 5.72% 94.28% 6.16% 93.84% differences in proportions -5.11* -0.80 11
Sub-Panel II: BDO versus other (N for BDO=952 and N for other =1,680) BDO BDO ABSDA 0.0620 0.0607 0.57 0.0463 0.0445 0.79 ABSAC 0.1133 0.1023 2.48* 0.0764 0.0740 1.38 DA 0.1342 0.1113 2.92* 0.0315 0.0210 1.78 MB 2.6455 2.4060 0.71 1.9084 1.6973 1.22 LogTA 4.1731 4.2213-0.77 4.1443 4.2146-1.51 CFLOW -0.0163 0.0230-4.81* 0.0298 0.0640-6.29* Categorical variables LOSS CHNI BDO 1 0 1 0 521 431 728 952 54.73% 45.27% 43.33% 56.67% 69 883 87 1,593 7.25% 92.75% 5.18% 94.82% Sub-Panel III: McGladrey versus other (N for MC=248 and N for other = 2,384) MC MC differences in proportions 5.62* 2.16* ABSDA 0.0610 0.0612-0.06 0.0477 0.0446 0.67 ABSAC 0.0929 0.1076-2.40* 0.0742 0.0745-1.42 DA 0.1048 0.1211-1.53 0.0371 0.0222 0.08 MB 2.7047 2.4706 0.59 1.6637 1.7680-0.23 LogTA 3.7338 4.2527-5.68* 3.5429 4.2574-5.35 CFLOW 0.0091 0.0087 0.03 0.0477 0.0522-0.24 Categorical variables LOSS CHNI MC 1 0 1 0 118 130 1,131 1,253 47.58% 52.42% 47.44% 52.56% 6 242 150 2,234 2.42% 97.58% 6.29% 93.71% differences in proportions 0.04-2.46* 12
Sub-Panel IV: Crowe Horwath versus other (N for CR=51 and N for other 2 nd tier = 2,581) CR CR ABSDA 0.0560 0.0613-0.68 0.0457 0.0449-0.71 ABSAC 0.0787 0.1068-2.40* 0.0475 0.0751-2.18* DA 0.1800 0.1184 2.35* 0.1446 0.0220 1.84 MB 1.7559 2.5072-2.60* 1.0782 1.7697-2.80* LogTA 4.6881 4.1943 2.32* 5.0752 4.1765 2.69* CFLOW 0.0532 0.0079 2.37* 0.0568 0.0513 0.94 Categorical variables LOSS CHNI CR 1 0 1 0 20 31 1,229 1,352 39.22% 60.78% 47.62% 52.38% 2 49 154 2,427 3.92% 96.08% 5.97% 94.03% differences in proportions -1.19-0.61 13
Table 5 Regression Analysis to Compare ERCs of Clients among Auditors SAR = a 0 + a 1 E + a 2 LagE + a 3 BDO + a 4 MC + a 5 CR + a 6 E*BDO + a 7 E*MC + a 8 E*CR + a 9 LagE*BDO + a 10 LagE*MC + a 11 LagE*CR + a 12 BETA + a 13 E*BETA + a 14 LagE*BETA + a 15 G + a 16 E*G + a 17 LagE*G + a 18 LOSS + a 19 E*LOSS + a 20 lage*loss + a 21 PERSIST + a 22 E*PERSIST + a 23 LagE*PERSIST + a 24 TENURE + a 25 E*TENURE + a 26 LagE*TENURE + a 27 LnTA + a 28 E*LnTA + a 29 LagE*LnTA + year dummies + industry dummies + ε a Variable Predicted Sign Intercept /a 0? E / a 1 + LagE / a 2? BDO / a 3? MC / a 4? CR / a 5? E*BDO / a 6? E*MC / a 7? E*CR / a 8? LagE*BDO / a 9? LagE*MC / a 10? LagE*CR / a 11? BETA / a 12 + E*BETA / a 13 - LagE*BETA / a 14? G / a 15 - Pooled Cluster-year regressions Regression 2000-2002 - 2005-2008 - 2001 2004 2007 2010-0.078-0.573 0.482-0.110-0.150 (-0.98) (-2.54)* (3.52)* (-1.26) (-1.32) 2.768 4.311 4.731 2.961 0.961 (5.78)* (2.37)* (5.42)* (4.02)* (1.63) -0.339-0.657-1.115-0.397-0.156 (-1.02) (-0.36) (-2.23)* (-0.62) (-0.44) -0.019-0.053 0.021-0.020-0.002 (-0.77) (-0.65) (0.32) (-0.92) (-0.06) -0.015-0.022-0.011 0.010 0.001 (-0.68) (-0.26) (-0.12) (0.34) (0.03) 0.015-0.278 0.314 0.048 0.026 (0.32) (-0.75) (1.54) (1.82) (0.58) -0.134-1.261-0.166-0.380-0.042 (-0.65) (-2.40)* (-0.66) (-1.16) (-0.15) -0.040-0.695-0.416-0.970 0.392 (-0.19) (-0.61) (-0.75) (-1.79) (1.20) 0.400-1.309 3.579-0.695 0.506 (1.58) (-2.06)* (1.42) (-2.20)* (1.66) 0.124 0.662 0.693 0.398-0.261 (0.63) (1.37) (4.05)* (1.23) (-0.74) 0.106 0.483 0.763 0.612-0.257 (0.61) (0.54) (1.82) (1.34) (-0.96) -0.014 0.237-4.613-0.392-0.169 (-0.05) (0.12) (-2.35)* (-0.60) (-0.52) 0.012-0.092 0.034 0.079 0.037 (0.76) (-1.29) (0.70) (0.41) (1.01) 0.363 0.415 0.419 0.557 0.382 (4.13)* (1.40) (3.07)* (2.73)* (2.62)* -0.245-0.416-0.222-0.691-0.248 (-2.31)* (-1.54) (-2.22)* (-3.14)* (-1.53) -0.006-0.003-0.005-0.004-0.006 (-2.47)* (-0.47) (-0.56) (-1.29) (-1.48) 14
Variable Predicted Sign E*G / a 16 +/- LagE*G / a 17? LOSS / a 18? E*LOSS / a 19 - Pooled Regression 2000-2001 Cluster-year regressions 2002-2004 2005-2007 2008-2010 0.016 0.003-0.056 0.007 0.007 (0.48) (0.04) (-1.41) (0.24) (0.10) -0.011 0.036 0.042-0.022-0.013 (-0.48) (0.44) (1.04) (-0.69) (-0.23) -0.084-0.065-0.148-0.063-0.081 (-2.37)* (-0.60) (-2.49)* (-1.57) (-1.33) -1.897-3.232-4.260-1.670-0.485 (-4.19)* (-2.34)* (-6.79)* (-3.48)* (-1.25) LagE*LOSS / a 20? -0.124 0.513 0.329 0.169-0.331 (-0.76) (0.58) (1.20) (0.42) (-1.62) PERSIST / a 21? -0.066-0.101-0.136-0.017-0.059 (-2.08)* (-1.27) (-4.46)* (-0.50) (-1.30) E*PERSIST / a 22 + 1.815 2.268-0.310 5.306 1.164 (4.34)* (1.89) (-0.27) (3.73)* (1.94) LagE*PERSIST / -1.939-2.300-0.017-5.779-0.998? a 23 (-4.11)* (-1.79)* (-0.02) (-4.33)* (-1.58) TENURE / a 24? -0.001-0.006 0.002 0.002-0.002 (-0.41) (-0.62) (0.51) (0.54) (-0.53) E*TENURE / a 25 +/- 0.015-0.063-0.040-0.022 0.027 (0.66) (-0.95) (-1.74) (-0.56) (0.83) LagE*TENURE / -0.026 0.008 0.058 0.066-0.081? a 26 (-1.02) (0.16) (2.14)* (1.60) (-3.25)* LnTA / a 27? -0.014 0.043-0.070 0.013-0.004 (-1.47) (1.60) (-4.87)* (0.14) (-0.23) E*LnTA / a 28 + -0.178-0.011-0.087-0.181-0.129 (-2.26)* (-0.04) (-0.89) (-2.02)* (-1.17) LagE*LnTA / a 29? 0.064-0.043-0.009-0.015 0.119 (1.24) (-0.13) (-0.12) (-0.15) (1.91) Year dummies Yes Yes Yes Yes Yes Industry dummies Yes Yes Yes Yes Yes N 4,287 411 812 1,532 1,532 Adjusted R 2 17.37% 27.98% 37.66% 15.38% 12.48% p-value a 6 a 7 0.62 0.61 0.68 0.14 0.18 p-value a 6 a 8 0.06 0.93 0.15 0.18 0.16 p-value a 7 a 8 0.08 0.59 0.12 0.51 0.58 Note: Variable definitions are in Exhibit 1. Huber-White t-statistics for the estimated coefficients clustered by industry are in parentheses. **indicates significance at the 5% level for two-tailed tests. 15
Table 6 Descriptive Statistics for Coefficient Estimates of Dechow et al. (2003) Non-discretionary Accrual Models with Control for Performance, Estimated by two-digit SIC code and year (i=16 industries; y=11 years; n=112 sets of coefficients) TotAcc = b 0 + b 1 ((1+k)Ch_Sales-Ch_Rec) + b 2 PPE + b 3 LagTotAcc + b 4 Gr_Sales + b 5 ROA + ε b Standard 25 th 75 th Coefficient Mean Median deviation percentile percentile Intercept / b 0-0.0303-0.0323 0.0660-0.0730 0.0133 (1+k) Ch_Sales Ch_Rec / -0.0131-0.0134 0.1493-0.1176 0.0679 b 1 PPE / b 2-0.0538-0.0424 0.1089-0.1084 0.0057 LagTotAcc / b 3 0.0480 0.0456 0.3371-0.1415 0.2292 Gr_Sales / b 4-0.0057 0.0031 0.1577-0.0489 0.0821 ROA / b 5 0.3768 0.3427 0.3135 0.1884 0.5650 Adj R 2 0.4904 0.5134 0.2869 0.3132 0.7000 Ch_Rec = c + k Ch_Sales + ε k Coefficient Mean Median Standard 25 th 75 th deviation percentile percentile Intercept / c 0.0026 0.0027 0.0137-0.0036 0.0117 Ch_Sales / k 0.0949 0.0860 0.0693 0.0472 0.1250 Adj R 2 0.2205 0.1582 0.2360 0.0310 0.3644 1 Variable definitions are in Exhibit 1. 16
Table 7 Univariate Comparisons of Absolute Values of Discretionary Accruals Panel A: GT versus other Mean Median Sample size Sample GT 2 nd GT -tier 2 nd Mann -tier t-statistic Whitney U GT statistic Pooled 0.0609 0.0616-0.32 0.0437 0.0466-0.95 1,381 1,251 2000-2001 0.0716 0.0750 0.43 0.0554 0.0535-0.42 114 145 2002-2004 0.0669 0.0602 1.46 0.0452 0.0451 0.73 360 291 2005-2007 0.0583 0.0584 0.02 0.0407 0.0431-0.07 505 461 2008-2010 0.0556 0.0613-1.55 0.0424 0.0493-1.87 402 354 *indicates significance of differences at the 5% level for two-tailed tests. Panel B: BDO versus other Mean Median Sample size Sample BDO 2 nd - BDO 2 nd Mann -tier tier t-statistic Whitney U BDO statistic Pooled 0.0620 0.0607 0.57 0.0463 0.0445 0.79 952 1,680 2000-2001 0.0769 0.0706 0.81 0.0543 0.0554-0.88 118 141 2002-2004 0.0602 0.0661 1.25 0.0470 0.0446 0.61 240 411 2005-2007 0.0581 0.0585-0.11 0.0420 0.0427-0.38 354 612 2008-2010 0.0622 0.0564 1.46 0.0501 0.0439 1.53 240 516 *indicates significance of differences at the 5% level for two-tailed tests. Panel C: McGladrey versus other Mean Median Sample size Sample MC 2 nd - MC 2 nd Mann -tier tier t-statistic Whitney U MC statistic Pooled 0.0610 0.0612-0.06 0.0477 0.0446 0.67 248 2,384 2000-2001 0.0672 0.0741-0.52 0.0596 0.0550 0.49 24 235 2002-2004 0.0632 0.0640-0.08 0.0518 0.0449 0.26 45 606 2005-2007 0.0614 0.0580 0.55 0.0487 0.0415 1.06 89 877 2008-2010 0.0578 0.0583-0.09 0.4428 0.0462-0.26 90 666 *indicates significance of differences at the 5% level for two-tailed tests. 17
Panel D: Crowe Horwath versus other Mean Median Sample size Sample CR 2 nd - CR 2 nd Mann -tier tier t-statistic Whitney U CR statistic Pooled 0.0560 0.0613 0.68 0.0457 0.0449-0.71 51 2,581 2000-2001 0.0604 0.0736-0.36 0.0145 0.0552-0.80 3 256 2002-2004 0.0359 0.0642-1.16 0.0226 0.0461-1.41 6 645 2005-2007 0.0492 0.0585-0.85 0.0351 0.0425-0.64 18 948 2008-2010 0.0655 0.0580 0.70 0.0586 0.0455 0.79 24 732 *indicates significance of differences at the 5% level for two-tailed tests. 18
Table 8 Multivariate Comparisons of Absolute Values of Abnormal Accruals ABSDA = c 0 + c 1 BDO + c 2 MC + c 3 CR + c 4 ABSAC + c 5 DA + c 6 MB + c 7 LOSS + c 8 LogTA + c 9 CHNI + c 10 CFLOW + year dummies + industry dummies + ε c Variable Predict. Sign Intercept / c 0? BDO / c 1? MC / c 2? CR / c 3? ABSAC / c 4 + DA / c 5 + MB / c 6 + LOSS / c 7? LogTA / c 8? CHNI / c 9 + CFLOW / c 10 - Pooled Regression 2000-2001 By cluster-year regressions 2002-2004 2005-2007 2008-2010 0.057 0.072 0.056 0.053 0.057 (7.29)* (4.30)* (8.79)* (7.23)* (4.30)* -0.002 0.001-0.006-0.004 0.003 (-0.77) (0.13) (-1.44) (-1.57) (1.04) 0.000-0.005-0.004 0.005-0.000 (0.09) (-0.44) (-0.46) (0.61) (-0.00) 0.004 0.001-0.011-0.003 0.013 (0.41) (0.04) (-0.59) (-0.22) (0.94) 0.194 0.198 0.180 0.252 0.148 (9.09)* (5.89)* (4.95)* (9.45)* (4.56)* -0.017-0.027-0.023-0.029-0.001 (-1.84) (-1.36) (-1.64) (-3.01)* (-0.04) 0.000-0.002 0.001 0.000 0.001 (3.43)* (-2.15) (1.78) (2.62)* (3.55)* -0.005-0.019-0.002-0.005-0.004 (-3.55)* (-2.92)* (-0.32) (-1.69) (-1.85) -0.003-0.003-0.002-0.003-0.003 (-2.05) (-1.04) (-2.05) (-1.98) (-1.05) 0.006 0.011 0.005 0.013 0.004 (1.06) (0.38) (0.43) (1.54) (0.78) -0.037-0.040-0.020-0.046-0.052 (-3.17)* (-2.65)* (-0.80) (-2.57)* (-2.70)* Sample size 2,632 259 651 966 756 Adjusted R 2 17.77% 15.21% 14.10% 24.55% 14.08% P value c 1 c 2 0.53 0.26 0.80 0.26 0.63 P value c 1 c 3 0.60 0.99 0.76 0.97 0.46 P value c 2 c 3 0.74 0.86 0.72 0.63 0.32 Note: Variable definitions are in Exhibit 1.Huber-White t-statistics for the estimated coefficients clustered by industry are in parentheses. *indicates significance at the 5% level for two-tailed tests. 19