Additional analysis:

Severity of restatements

Although regulators express concern over the number of accounting restatements, there is

0.009), and the two-ranked concurring partner having insignificant coefficient (p=0.227). Therefore, we define the top-ranked lead and concurring partners as specialists in our paper.

variation in the severity of the response to restatements. For example, Palmrose et al. (2004) find that the market reaction to restatements announcement is associated with characteristics presumed to be related to the severity of the underlying error. In this section, we focus on misstatements most likely to be of consequence to investors, and further investigate the effect of industry auditor expertise on these characteristics of restatements.

Material restatements

Following Kinney et al. (2004), we define a misstatement as material if its absolute size is greater than or equal to 5% of the absolute value of the net income or loss originally reported. Of the 326 restatements during our sample period, 99 meet this materiality threshold. We rerun equation (1) where the dependent variable is an indicator for whether or not the restatement is material. Column 1 of Table 7 reveals that the coefficients of CPA1 and CPA3 are significant and negative (p < 0.01), and the coefficients CPA2 is insignificant, indicating that companies of industry auditor specialists are less likely to make material misstatement, and the differential restatements likelihood due to industry expertise is primarily attributable to the signing partner specialists (i.e. lead auditor specialists) rather than auditor specialists at the firm level.

[Insert Table 7 here]

Core earnings

In addition to considering the materiality of the restatements, we also consider whether they affect core earnings. We make this distinction because we posit that investors are differentially concerned about restatements that effect core versus non-core earnings. Core earnings are more persistent or sustainable into future periods and are fundamentally linked to the expected future prospects of the company (Penman, 2001). Thus core earnings are more value relevant than are non-core earnings (Bradshaw and Sloan, 2002). Recent studies also document that restatements containing core earnings have a more negative effect on stock returns

than do those involving non-core earnings (Palmrose et al., 2004) and are associated with litigation against restating companies and their auditors (Palmrose and Scholz, 2004).

Following Palmrose et al. (2004), we define core-earnings restatements as those involving revenue, cost of sales or on-going operating expenses. In addition, non-core earnings are defined as those from all other activities. Of the 326 restatements in our sample, 116 contain core earnings. Consistent with our prediction, Column 2 of Table 7 reveal that the lead auditor specialists, either alone or in conjunction with firm-level specialists, are significantly associated with lower likelihood of core earnings restatements, indicating that companies retaining auditor specialists are less likely to misstate core earnings.

Pervasiveness and duration of restatements

Our third characteristic of restatements is related to the pervasiveness of the restatements within the income statement. We follow Palmrose et al. (2004) and count the number of account groups that represent the focus of the restatements: revenue, cost of goods sold, on-going operating expenses, special items/one-time events, merger accounting, non-operating income accounts, and other items. Thus this variable can range from one to seven. We expect this variable to be negatively associated with the industry experts. Finally, we consider a variable for the persistence of the misstatement, DURATION, as measured by the number of years’ financials restated (where a restated quarter = 0.25). We expect this variable to be negatively related with the industry experts.

Column 3 of Table 7 shows that the coefficient of CPA1 is negative and significant, while CPA3 have predicted signs, but is marginally different from zero. Therefore, our results indicate that the auditors that are industry specialists at both the firm-level and the partner-level are negatively associated with the pervasiveness of the restatements. Finally, Column 4 reveals

that the coefficients of CPA1 and CPA3 are marginally significant in the predicted signs, indicating that the empirical results do not support our prediction that industry auditor specialists are negatively associated with the persistence of the restatements.

Initiator of restatements

According to Article 6 of SELER, the TSFB should require companies to restate financial reporting when misstatements of companies meet a “serious” threshold. Thus, the TSFB-initiated restatements are generally perceived to be more serious than those initiated by the others. In this section, we will provide direct evidence on whether or not the TSFB-initiated restatements are more “serious” than are the companies/auditors-initiated restatements.

Table 8, Panel A profiles the restatements sample according to characteristics of accounting restatements as described in previous analyses, tabulated by the TSFB-initiated and voluntary restatements (i.e., companies or auditors-initiated restatements). Panel A shows that the duration for the TSFB-initiated subsample is significantly longer than those for voluntary subsample, but there is no significant difference in the mean (median) pervasiveness between these two subsamples. The results suggest that the TSFB-initiated restatements, on average (at the median), affect more financial reports than do voluntary restatements. Panel A also shows that the TSFB-initiated restatements tend to have more material effect on incomes and are more likely to effect core earnings than do voluntary restatements. The results provide compelling evidence that the TSFB-initiated restatements generally are more serious than are voluntary restatements.

To gain further insights into whether the findings regarding effect of industry specialists on restatements are primarily driven by the TSFB-initiated group, we rerun equation (1) using the TSFB-initiated and voluntary restatements subsamples, respectively. Panel B reveals that for these two groups, the coefficients of CPA1 and CPA3 are significant in the predicted signs (p <

0.01), and the coefficients CPA2 is insignificant, consistent with Table 4. Therefore, although the Pseudo R² and the absolute values of the coefficients on CPA2 and CPA3 for the TSFB-initiated restatements group are greater that those for voluntary group, the results in Table 4 are driven by both types of restatements rather than by the TSFB-initiated restatements group only.

Annual and interim restatements

As mentioned above, restatements sample can be partitioned into restatements affecting annual (audited) reports and those affecting only quarterly reports. Analyses related to both misstatement types are important since auditors, investors and regulators are presumably concerned about both (Myers et al., 2005). In this section, we re-run our regression by using the annual and interim restatements, respectively. Table 9 shows that the coefficients of CPA1 and CAP3 are significant and negative in both partitions, consistent with those in Table 4. These analysts indicate that our results in Table 4 are driven by both annual and interim restatements.

The results are consistent with the findings that the market reactions to restatements announcements do not differ for annual versus interim misstatements (Palmrose et al., 2004).²⁸

However, the coefficient of CPA2 is marginally significant for interim restatements (p=0.145, two-tailed), while it is insignificant for annual restatements, indicating that firm-level auditor specialists alone are weakly associated with the restatements likelihood when performing quarterly reviews. Table 9 also shows that the magnitude of coefficient on CPA1 is greater than that on CPA3 for quarterly reports subsample. These results suggest that firm-level auditor specialists appear to add something over and above the effects of the signing auditor specialists alone when performing quarterly reviews.

[Insert Table 9 here]

28 Our results also are consistent with the study of restatements by Financial Executives International including both types (Moriarty and Livingston, 2001).

4.5 Sensitivity analysis and robustness checks