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

Earnings Informativeness of Long-Lived Assets Impairment Recognized and Reversals 212 Excluding reporting incentives for recognition of write-offs, impairment reversals may only reflect managers’ inaccurate valuation processes. In this case, ex post reversals to some extent suggest that managers’ initial estimations of asset write-offs are insufficient. If investors are unable to ravel measurement errors in managers’ write-off judgment, the relatively “inaccurate” recognized impairment embedded in current earnings weakens the linkage between earnings and future cash flows which, in turn, may deteriorate the relationship between current returns and future earnings. Accordingly, we suggest that the informativeness of future earnings for firms with impairment reversals is mitigated as compared to firms without impairment reversals. We establish the third hypothesis as follows: H3. Ceteris paribus , the informativeness of future earnings is mitigated for firms with assets impairment reversals. 3. Research Design 3.1 Data and Sample Selection The years 2004-2010 are chosen as the observation period because we need the earning and stock return data for the subsequent consecutive three years to examine the earnings informativeness in the CKSS model. 2004 is chosen as the starting year because SFAS No. 35 was first enforced in Taiwan that year. Empirical data are retrieved from the Taiwan Economic Journal (TEJ) Database. Table 1 reports the sample selection process in the study. The sample firms used in this study are publicly traded companies that listed on Taiwan Stock Exchange (TSE) or Over-The-Counter (OTC). The observations on the TEJ database from 2004 to 2010 include 12,427 firm-years (excluding finance-related institutions (codes No. 28 and 30 in TEJ) as they are subject to different disclosing requirements). This study deleted 4,149 observations for financial data and 342 observations for firms for which stock return data was unavailable. We also deleted three observations due to data unavailability for other control variables and one observation for its unreasonable value in impairment recognition. This selection procedure yielded a final sample of 7,932 observations for empirical analysis.