臺大管理論叢 NTU Management Review VOL.29 NO.1

Earnings Informativeness of Long-Lived Assets Impairment Recognized and Reversals 216 3.2 Variables Measurement Dependent Variable: Stock Return (R t ): High quality information increases the likelihood of correctly forecasting the outcomes of past or present events (SFAC No. 2). If recognition of asset impairment deteriorates (improves) earnings informativeness, stock prices will provide less (more) information about current (future) earnings. This study follows Collins et al. (1994), Lundholm and Myers (2002) and Tucker and Zarowin (2006) in using a firm’s ex-dividend annual stock return in year t (R t ) as the dependent variable and then examines whether recognition of asset impairment embedded in the current stock price reflects information about current and future earnings. 8 Pivotal Explanatory Variables: Magnitude of Assets Impairment Recognized (IM t ): This study uses disclosures of long-lived asset impairment in aggregate, rather than broken down into asset categories, since a separate analysis of the impairments would require assumptions about the distribution of assets across cash generating units. Accordingly, the magnitude of asset recognized as impaired (IM t ) is measured as the disclosed aggregate of long-lived impaired assets for the firm in each year scaled by the lagged total assets (Francis et al., 1996; Riedl, 2004; Young and Wu, 2009). 9 Note that the magnitude of assets recognized as impaired is tiny in some listed firms and thus is likely to have a negligible impact on earnings. Therefore, IMt values below 0.1% are rounded down to zero. While the magnitude of assets recognized as impaired is a continuous 8 Collins et al. (1994) pointed out that annual earnings are announced about several months after the fiscal year-end, resulting in the mismatch between earnings and contemporaneous annual stock returns. Consequently, a portion of year t returns is in response to the previous fiscal year’s (year t -1) earnings and dividend reported in year t . Thus, Collins et al. (1994) and Tucker and Zarowin (2006) suggested the use of lagged earnings in the model and used ex-dividend annual returns to exclude the mismatch of current earnings and annual stock returns. 9 We also use an alternative measure of assets impairment (IM), i.e., a firm’s magnitude of assets impairment in year t deflated by the total long-term assets at the beginning of year t , and rerun the models. The results do not qualitatively change the initial findings.