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

Earnings Informativeness of Long-Lived Assets Impairment Recognized and Reversals 242 We also run additional tests to address issues related to potential self-selection bias. Our regressions include the inverse Mills ratio estimated from the Reg. (3) of the determinant of asset write-off recognition model (Heckman, 1979). We then add the inverse Mills ratio (IMR) generated from the estimation to the regressions. The additional results are reported in the “Self-selection Correction” model in Table 10, which shows that the coefficients of IM t *X t and IM t *X t3 are respectively -12.238 ( t = -5.75) and 1.590 ( t = 1.75), both statistically significant. The coefficients of IM t *X t and IM t *X t3 are respectively -13.407 ( t = -5.63) and 2.479 ( t = 2.73) in the REV model, both statistically significant. We also find that the coefficients of REV t *IM t *X t and REV t *IM t *X t3 are respectively 33.817 ( t = 1.62) and -14.970 ( t = -1.41), both statistically insignificant. The combined coefficient of IM t *X t and REV t *IM t *X t ( β 7 + β 12 ) is 20.409 ( t = 0.95), meanwhile, the combined coefficient of IM t *X t3 and REV t *IM t *X t3 ( β 8 + β 13 ) is -12.491 ( t = -1.15), which is statistically insignificant. The results from this robustness test provide some corroborative evidence to exclude self-selection bias in our analysis. 5.6 Voluntarily Early Adoption of SFAS No.35 and The Occurrence of Global Financial Crisis Consideration Note that SFAS No. 35 was promulgated in July, 2004 and began enforcement in January 2005. However, SFAS No.35 allowed for voluntary adoption prior to the effective mandatory date. Thus, early adoption in 2004 allowed managers to convey private information regarding asset valuations and/or alter measures of financial performance used in debt agreements and compensation contracts. To understand whether the voluntary adoption of SFAS No. 35 had an effect on our analysis (Hsieh and Wu, 2005), this study excludes the year 2004 and runs the regressions using the remaining years. In addition, the 2008 global financial crisis represents a relatively exogenous shock that increased the demand for quality accounting reporting. It is expected that the global crisis would have increased demand for higher quality accounting reporting. We thus further exclude observations from the year 2008 and examine whether the global financial crisis confounded the relationship between asset impairment recognition and the informativeness of earnings in the initial analysis. Empirical results are reported in Table 11.