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

213 NTU Management Review Vol. 29 No. 1 Apr. 2019 Table 1 Sample Selection Descriptions N Firms listed on TEJ during 2004-2010 12,427 Less: Missing financial data (4,149) Missing data of ex-dividend stock return (342) Firms’ data unavailability for other control variables (3) Impairment loss with error sign (1) Final empirical observations 7,932 Non-impairment loss samples 6,989 Impairment loss without reversals samples 875 Impairment loss with reversals samples 68 Table 2 presents the sample year and industry distributions used in this study. Table 2 shows that the sample in each year reveals an increasing pattern during the observation period. There were 236 observations with asset impairment in 2005 (25.03%), followed by 161 observations in 2008 (17.07%) and 145 observations in 2006 (15.38%). Accordingly, we control the year effect in the following empirical models. Approximately 53.49% (4,243/7,932) of observations come from the electronics industry. This sample structure supports the findings of Wang, Lee, and Huang (2003) that electronic-related industries dominate so-called traditional industries in Taiwan. 7 The remaining observations are spread across other industries. The final sample includes 943 instances of asset impairment, 68 of which were reversed in the following accounting year. Almost 76.99% of impaired observations occur within five industries: Electronics & Telecommunications (code 23, 461/943 = 48.89%), Construction (code 25, 96/943 = 10.18%), Comprehensive (code 99, 60/943 = 6.36%), Spin & Fiber (code 14, 58/943 = 6.15%), and Electric Machinery (code 15, 51/943 = 5.41%). According to Table 2, specific industries with high rates of impairment include Automotive (code 22, 10/34 = 29.41%), Construction (code 25, 96/460 = 20.87%), and Software (code 32, 15/81 = 18.52%). Finally, industries with high rates of impairment reversal include: Construction (code 25, 12/96 = 12.5%), Merchandize & Trade (code 29, 4/22 = 18.18%), and Electronics & Telecommunications (code 23, 34/461 = 7.38%). 7 We rerun the empirical models using cross-sectional data with both year and industry effects. The results are not qualitatively different from the findings.