We test the robustness of the logistic regression results to distinguishing restatements as either errors or irregularities and also to the effects of the restatement initiator and Blue Ribbon Recommendations.13 We include an irregularity variable consistent with Hennes et al. (2008) who report that a restatement due to accounting irregularity increases the likelihood of a restatement-related securities class action compared to a restatement categorized as an error. When a restatement is associated with an accounting irregularity, it is expected that the class action will be more meritorious and less likely to be dismissed because an inference of scienter or a mental state embracing intent to deceive, manipulate or defraud shareholders is more easily established. Following Hennes et al. (2008), accounting irregularity is defined as admission of irregularity or fraud by a firm, or the investigation of accounting irregularity or fraud by the SEC, Department of Justice or other independent entity.
We also consider the effect of the restatement initiator on the likelihood of restatement-induced shareholder litigation. Disclosure of the restatement initiator could influence probability of restatement-induced class action because the restatement initiator variable is associated with shareholder losses around restatement announcements. Palmrose et al., 2004 Z. Palmrose, V.J. Richardson and S. Scholz, Determinants of market reactions to restatement announcements, Journal of Accounting and Economics 37 (1) (2004), pp. 59–90.Palmrose et al. (2004) provide empirical evidence that company or auditor-initiated restatement is associated with more negative restatement announcement returns. Accordingly, disclosure of the restatement initiator could influence the likelihood of shareholders initiating a securities class action lawsuit because information about the restatement initiator may be associated with a more adverse investor valuation of the restating firm.
Finally, we examine the effect of the Blue Ribbon Recommendations on the probability of restatement-induced class action lawsuits. The Blue Ribbon Recommendations, which were issued in 1999, were aimed at improving the effectiveness of audit committees. The recommendations required the revision of the definition of director independence, audit committee independence and at least three members who must also be financially literate on the audit committee. Under the Blue Ribbon Recommendations, each audit committee must also adopt a formal charter. Adoption of the above recommendations strengthens corporate audit committees and may improve monitoring by audit committees.
Smith (2006) asserts that following the Blue Ribbon Recommendations, audit committees became more active and made more serious efforts to monitor management. Accordingly, the Blue Ribbon Recommendations could result in changes in audit committee behavior. If shareholders perceive boards and audit committees of restating firms as more effective following the Blue Ribbon Recommendations, they would perceive their allegations as less meritorious and would be less likely to initiate restatement-induced class action lawsuits. We expect information on the restatement initiator and Blue Ribbon Recommendations could influence shareholder perception of the effectiveness of the board and audit committee, which in turn would influence the merit of a class action lawsuit and the decision to initiate the class action lawsuit.
Table 8 shows estimated logistic regression coefficients and Wald test p-values for seven models of restatement-induced class action logistic regressions with accounting irregularity (IRREG), restatement initiator variables and a dummy variable for the effect of Blue Ribbon Recommendations. Restatement initiator variables are dummy variables equal to one if a restatement is initiated by the auditor (AudIn), SEC (SECIN), or company (COMPIN), and zero, otherwise. Blue Ribbon Recommendations measure (BRC) is a dummy variable equal to one, if the board and audit committee variables are measured after the Blue Ribbon Recommendations were issued, and zero, otherwise.
--------------------------------------------------------------------------------
Table 8. Robustness tests.
Variable Pred. sign Model
1 2 3 4 5 6 7
INTERCEPT − 7.0102a (0.0001) − 6.6139a (0.0008) − 7.0481a (0.0001) − 7.4765a (0.0002) − 5.4352a (0.0012) − 5.4748a (0.0004) − 8.8516a (0.0004)
CAR(− 1, 1) (−) − 0.5458a (0.0031) − 0.5686a (0.0030) − 0.5398a (0.0033) − 0.5369a (0.0040) − 0.4964a (0.0010) − 0.4923a (0.0010) − 0.5107a (0.0037)
BrdIndep (−) − 4.5974b (0.0102) − 4.4736b (0.0133) − 4.4338b (0.0133) − 4.4773b (0.0143) − 3.6994b (0.0147) − 3.7031b (0.0148)
AudIndep (−) − 0.1422 (0.8395)
AccExp (−) 0.4948 (0.5327) 0.5083 (0.5156)
FinExp (−) − 0.7654 (0.5263) − 0.1961 (0.8319)
MajFinexp (−) − 0.1408 (0.0514) − 0.0595 (0.9246) − 0.3142 (0.5510)
IRREG (+) 3.0604a (< 0.0001) 3.1538a (< 0.0001) 3.0697a (< 0.0001) 2.9581a (< 0.0001) 2.7790a (< 0.0001)
AudIn 1.0504 (0.4558) 0.8983 (0.4850)
COMPIN 0.4081 (0.5806) 0.4876 (0.5024)
SECIN 0.9341 (0.4160) 1.2768 (0.2477)
BRC − 0.0813 (0.8766) − 0.0568 (0.9137) − 0.9244 (0.2325)
AUMeet (−) − 0.1387 (0.2988) − 0.1104 (0.4184) − 0.1307 (0.3208) − 0.1629 (0.2643) − 0.2022 (0.0905) − 0.2015 (0.0887) − 0.2258 (0.1208)
BDMeet (−) 0.1855b (0.0292) 0.1785b (0.0330) 0.1769b (0.0360) 0.1735 (0.0501) 0.1279 (0.0682) 0.1211 (0.0870) 0.1990b (0.0276)
RevRest (+) 1.5424b (0.0174) 1.5695b (0.0157) 1.5550b (0.0166) 1.7133b (0.0123) 1.5129a (0.0041) 1.5082a (0.0042) 1.7502a (0.0091)
AcctsNo (+) − 0.1342 (0.7127) − 0.0927 (0.8017) − 0.0997 (0.7859) − 0.0562 (0.8807) − 0.0390 (0.8979) − 0.0418 (0.8906) − 0.0663 (0.8618)
YrsRest (+) 0.4233 (0.1753) 0.4228 (0.1785)
NiEffect (+) 0.0556 (0.4096) 0.0511 (0.4632) 0.0542 (0.4161) 0.0557 (0.4307) 0.0239 (0.6955) 0.0227 (0.7065) 0.0408 (0.4063)
LEVERAGE (+) − 3.7753 (0.1265) − 3.8907 (0.1268) − 3.5378 (0.1439) − 3.2245 (0.1816) − 3.9335b (0.0457) − 3.9717b (0.0413) − 2.4167 (0.2887)
SIZE (+) 1.1209a (< 0.0001) 1.1597a (< 0.0001) 1.1269a (< 0.0001) 1.1271a (0.0001) 1.0612a (< 0.0001) 1.0724a (< 0.0001) 0.9505a (< 0.0001)
GROWTH (+) − 0.0148 (0.8595) − 0.0420 (0.6464) − 0.0249 (0.7711) − 0.0280 (0.7548) − 0.0286 (0.6965) − 0.0311 (0.6616) − 0.0037 (0.9611)
Sample size 143 143 143 143 143 143 143
L-ratio Chisq (p-value) 90.9254a (< 0.0001) 90.9284a (< 0.0001) 90.5880a (< 0.0001) 91.6539a (< 0.0001) 63.8928a (< 0.0001) 64.2054a (< 0.0001) 86.8423a (< 0.0001)
Hosmer and Lemeshow Chisq (p-value) 1.9499 (0.9825) 1.8545 (0.9852) 3.7773 (0.8766) 4.0388 (0.8536) 3.5529 (0.8950) 8.1625 (0.4178) 3.1968 (0.9214)
Full-size table
Table 8 presents logistic regressions for the robustness tests. Dependent variable is the probability of restatement-induced securities class action (ClassAc), which equals 1 if a firm is subject to a securities class action lawsuit following the restatement, 0, otherwise. CAR(− 1, 1) equals cumulative abnormal returns over the three-day (− 1, 1) interval beginning on the day prior to the restatement announcement date. BrdIndep equals proportion of independent directors on the board. AudIndep equals 1 if the audit committee is fully independent, 0, otherwise. AccExp equals 1 if an independent accounting expert serves on the audit committee, 0, otherwise. FinExp equals 1 if restating firm has at least one independent audit committee financial expert, 0, otherwise. MajFinexp equals 1 if majority of audit committee members are independent audit committee financial experts, 0, otherwise. IRREG equals 1 if the restatement involves an accounting irregularity, 0, otherwise. AUMeet equals number of audit committee meetings in the fiscal year preceding the end of the misstatement. BDMeet equals the number of board of director meetings in the fiscal year preceding the end of the misstatement. RevRest equals 1 if a firm restates its revenues, 0, otherwise. AcctsNo equals number of accounts affected by the restatement. YrsRest equals number of years restated with restatement of a quarter equal to 0.25. NiEffect is the size of the restatement, which equals initially reported net income less restated net income/market value of equity (expressed as a percentage). AudIn equals 1 if a firm's auditor initiates the restatement, 0, otherwise. COMPIN equals 1 if a firm initiates the restatement, 0, otherwise. SECIN equals 1 if the SEC initiates the restatement, 0, otherwise. BRC equals 1 if the governance data is measured subsequent to the issuance date of the Blue Ribbon Committee (BRC) recommendations, 0, otherwise. LEVERAGE equals long-term debt divided by total assets. SIZE equals log of market value of equity. GROWTH equals ratio of market-to-book value of equity. p-Values are in parentheses below the coefficient estimates for each variable. Statistical significance at the 1 and 5% levels is denoted by (respectively) a and b. Likelihood ratio test statistics for all the models correspond to p-values less than 0.0001. Hosmer and Lemeshow goodness-of-fit (Pr > Chisq) p-values for all the models are greater than 0.1. Variance Inflation Factors for the regressions are between 1.0862 and 1.6539 indicating that multicollinearity is not a problem.
View Within Article
We find that the results are unchanged when we include accounting irregularity, restatement initiator and Blue Ribbon Committee variables. Coefficients of restatement cumulative abnormal return CAR(− 1, 1), board independence (BrdIndep) and revenue restatement (RevRest) are significant as before. We also find a significantly positive coefficient for accounting irregularity (IRREG). Likelihood ratio (p < 0.01) and Hosmer and Lemeshow's goodness of fit (p > 0.1) test results indicate that all the regression models fit the data at an acceptable level and significantly predict the probability of restatement-induced class action lawsuits. Variance Inflation Factors of all the regressions for the robustness tests are between 1.0862 and 1.6539 indicating that multicollinearity is not a problem.
