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

99 NTU Management Review Vol. 30 No. 2 Aug. 2020 BIG it = a dummy variable equal to 1 if the auditor is a Big 8 audit firm (or their successors), and 0 otherwise; EARN it = firm i ’s earnings before extraordinary items, deflated by total assets at the end of t -1; LTGN it = a dummy variable equal to 1 if firm i ’s SIC code is 2833-2836, 8731-8734, 7370-7379, 3570-3577, 3600-3674, and 0 otherwise; IND_Q it = the industry-adjusted Tobin’s Q , which is the log difference between Q and the median Q , measured at the end of t ; for firm i ’s primary four-digit SIC classification. Q is the ratio of market value of assets to the book value of assets, where the market value of assets is the book value of assets less the book value of equity plus the market value of equity; TA it = the natural logarithm of firm i ’s total assets at the end of t ; IA it = the sum of firm i ’s annual change in inventories and the annual change in gross property, plant, and equipment scaled by lagged total assets at the end of t -1; SEG it = the number of firm i ’s business segments; CS it = firm i ’s net cash flows less cash dividends and capital expenditures, scaled by lagged total assets at the end of t -1; CAP it = firm i ’s capital expenditures to sales, measured over year t ; and INDROA it = firm i ’s industry-adjusted return on assets at the end of t , defined as earnings before interest and taxes less its median industry ROA, as classified by two- digit SIC code. Table 3 presents Pearson correlations among the variables in each set of empirical tests. Panels A and B report univariate correlations among the endogenous and other independent variables in our simultaneous equations for the U.S. GAAP and IFRS samples, respectively. We find that DA it is correlated positively with all of the proxies for real activities manipulation, namely RM_SUM it , RM_PROD it , and RM_DISX it . This correlation can be explained by firms engaging in accrual-based earnings management and real activities manipulations simultaneously, as discussed in Cohen and Zarowin (2010) and Roychowdhury (2006). Moreover, the correlations are consistent with our expectations in that most variables (i.e., BBATH it , SM_SUM it , SM_PROD it , SM_DISX it , DEBT it , and ST_DEBT it ) that capture executives’ incentives to alter the magnitudes of abnormal accruals and real activities are significantly correlated with the earnings management proxies (i.e., DA it , RM_SUM it , RM_PROD it , and RM_DISX it ) in the hypothesized directions for U.S. GAAP firms. We also find that the variables that proxy for executives’ reporting incentives, except ST_DEBT it , are negatively correlated with the earnings management proxies for IFRS firms, although some are less significant. The identifying variables (i.e., DA it-1 and LTGN it ) are generally not significantly correlated with the dependent variables of the equations they identify for both categories of sample firms.