臺大管理論叢 NTU Management Review VOL.30 NO.2

109 NTU Management Review Vol. 30 No. 2 Aug. 2020 benchmark, and 0 otherwise. 21.22 Following Gunny (2010), we define firms suspected of meeting last-year earnings as firm-year observations with change in net income scaled by lagged total assets between 0 and 0.01, and firms suspected of meeting zero earnings as firm-year observations with net income as a percentage of lagged total assets between 0 and 0.01. The inferences on SMOOTH it as well as the other explanatory variables are consistent with those in the primary analysis. 23 The evidence suggests that SMOOTH it is more likely picking up reporting incentive behaviors, as opposed to the underlying economics of the firm. 7.3 Analyses Controlling Country Effects and Macroeconomic Factors Due to the small sample size, we are unable to control the country effect by incorporating all of the country indicator variables into the models. Instead, we examine the country effect by adding the indicator variable one by one to the models. Panel A of Table 1 indicates that Canada has the largest number of U.S. GAAP firms while the other countries have a single-digit number of either U.S. GAAP or IFRS firms. Accordingly, we re-estimate the simultaneous equations (6) and (7) after including a country dummy variable ( COUNTRY it ) coded 1 for the sample firms that domicile in Canada and 0 otherwise. In untabulated results, we find that altering specifications after including COUNTRY it produce similar results. Specifically, the parameters and significance levels on the reporting incentive variables remain similar to those reported in Table 5. Investigating the effects of countries other than Canada will result in substantially uneven sample size at a greater extent for the two groups compared and may therefore bias the regression results. Accordingly, we do not repeat the analysis for the sample firms in other countries. 21 Given that abnormal production costs and discretionary expenditures have different implications for earnings, we first multiply the latter by -1 and then add it to the former as an aggregate measure before considering their income effect in the measurement of SM_SUM it . 22 We do not examine analysts’ forecasts for two reasons: (1) real activities manipulation has to take place before the fiscal year-end, and managers are unlikely to know the forecasted earnings prior to the earnings announcement; and (2) we do not have sufficient data about analysts’ forecasts of earnings for our sample firms. 23 We also examine an alternative threshold for BBATH it , which is redefined to equal 1 when earnings (adjusted for discretionary accruals and real activities manipulation) are negative and lower than the four-quarters-ago earnings, and 0 otherwise. Similar to those for SMOOTH it , BBATH it is proxied by three measures (i.e., BB_PROD it , BB_DISX it , and BB_SUM it ) depending on whether the abnormal level of production costs, the abnormal level of discretionary expenditures, or the aggregate measure of real activities manipulation is used in these alternative measurements of the “big bath” incentive. The results (not reported) are qualitatively similar and the inferences are unchanged.