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

87 NTU Management Review Vol. 30 No. 2 Aug. 2020 managers may take the opportunity to report other discretionary bad news (i.e., takeing a “big bath”) (Yin and Cheng, 2004). Accordingly, to proxy for the “big bath” incentive, we include BBATH it , which is set to 1 when the firm’s ROE ranks in the bottom 20% of its industry, and 0 otherwise. H 2 predicts the “big bath” variable to be negatively associated with discretionary accruals as well as real activities manipulation. Conversely, the income smoothing literature posits that managers have incentives to reduce earnings volatility. Specifically, when earnings are unusually high, firms are likely to make income-reducing accounting decisions; when earnings are unusually low, they are likely to make income-increasing accounting decisions. To measure earnings performance for a given year, we must specify a benchmark to which to compare earnings. Assuming that annual earnings follow a random walk, firms’ reported earnings in the prior year are used as the benchmark. That is, earnings performance in the following year is measured by next year’s earnings changes. To proxy for the smoothing incentive, SMOOTH it is defined as the difference between firm i ’s current earnings (measured before discretionary accruals and real activities manipulation) and prior-year earnings, divided by beginning total assets, when this change is either above the median of nonzero positive values or below the median of nonzero negative values of this variable, and 0 otherwise. 9 Extant research documents a significant association between “income smoothing” behavior and accrual management without taking into account real activities manipulation (e.g., Chen, Chang, and Weng, 2017; Subramanyam, 1996; Tucker and Zarowin, 2006). H 2 asserts that the smoothing proxy is negatively associated with discretionary accruals as well as real activities manipulation. 9 In order to capture the effect of the smoothing incentive on individual versus overall real earnings management activities, we use three measures to proxy for “income smoothing” reporting ( SMOOTH it ), depending on whether RM_PROD it and RM_DISX it , or the comprehensive metric of real activities manipulation, RM_SUM it , is tested. Specifically, SM_PROD it is the difference between firm i ’s earnings (measured before discretionary accruals and abnormal production costs) for year t and earnings for year t -1, divided by total assets at the end of t -1, when this change is either above the median of nonzero positive values or below the median of nonzero negative values of this variable, and 0 otherwise; SM_DISX it is equal to the difference between firm i ’s earnings (measured before discretionary accruals and abnormal discretionary expenditures) for year t and earnings for year t -1, divided by total assets at the end of t -1, when this change is either above the median of nonzero positive values or below the median of nonzero negative values of this variable, and 0 otherwise; SM_ SUM it is the difference between firm i ’s earnings (measured before discretionary accruals, abnormal production costs, and abnormal discretionary expenditures) for year t and earnings for year t -1, divided by total assets at the end of t -1, when this change is either above the median of nonzero positive values or below the median of nonzero negative values of this variable, and 0 otherwise.