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

Predicting Future Performance Using Fair Value versus Historical Cost: Evidence from Investment Property 322 by total assets, SUM_FV , is 0.002. 3.5% of firms have their investment properties monitored by an external appraiser ( APP ), suggesting that few companies hire external appraisers to monitor the fair value of investment properties. Table 2 presents the Pearson correlations and the Spearman rank correlations between the test variables. Based on Pearson correlations, IN t has a positive and significant correlation with IN t+1 (0.523), IN t+2 (0.440), and IN t+3 (0.275). IP*D_FV has a positive and significant correlation with IN t+1 (0.230), IN t+2 (0.213), and IN t+3 (0.316). The correlations are as predicted for H1. SUM_FV has a positive and significant correlation with IN t+1 (0.516), IN t+2 (0.463), and IN t+3 (0.455). The correlations are as predicted for H2. In general, under the Spearman rank correlations, the signs are the same as the Pearson correlations, but IP*D_FV and SUM_FV do not have significant correlations with IN t+1 , IN t+2 , and IN t+3 . 5.2 Regression Results Table 3 presents the regression results from estimates of equation (1), which measure the incremental predictability of the fair value model on future earnings. Columns (1), (2), and (3) of Table 3 show findings on the firm’s future performance defined as one-year- ahead, two-years-ahead, and three-years-ahead of income, respectively. Table 3 shows that the coefficient on incomes measured under the fair value model, IN*D_FV , is significantly positive for the one-year-ahead income, two-years-ahead income, and three-years-ahead income. These results indicate that incomes in firms with investment property measured by the fair value model generate higher predictability than firms measured with the cost method. These findings, therefore, support H1. Table 4 reports that the coefficient on accumulated changes in the fair value of investment properties, SUM_FV*D_FV , is significantly associated with one-year-ahead income, two-years-ahead income, and three-years-ahead income. These results corroborate our earlier findings that investment property measured by the fair value model generates a higher predictability of future income than investment properties measured by the cost model, and recognition that changes in the fair value of investment properties can positively affect the predictability of current earnings for future incomes.