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

Asymmetric Valuation Adjustments in Accumulated Other Comprehensive Income 152 Model (3) of Table 2 includes both market- and accounting-based information inputs to reflect variations in economic conditions. As predicted, and consistent with the previous two models, the estimated coefficients are 0.093 for and 0.262 for ( t -statistics = 40.05 and 89.83, respectively) when information inputs identify an improvement in the market value of equity and in sales revenue, respectively. On the other hand, both and are negative (-0.09 and -0.112, respectively) and significant at the 1% level ( t -statistics = -28.03 and -30.5, respectively). The downward OCI adjustment is only 0.003 % (i.e., 0.093 – 0.09) per 1% decrease in the market value of equity. It is 0.15% (i.e., 0.262 – 0.112) when sales revenue decreased by 1%. We interpret the results as evidence supporting H1, because of the significant difference in OCI adjustments between improving and deteriorating economic conditions. The adjusted R -squares in Table 2 indicate an interesting difference between market- based and accounting-based information inputs. The accounting-based information input in Model (2) tends to have higher explanatory power for OCI adjustments than market- based inputs in Model (1). The adjusted R -squares increase from 4.97% in Model (1) to 15.24% in Model (2). This difference is consistent with findings from Barth et al. (1998), Eccher et al. (1996) and Nelson (1996) in which compared with accounting disclosure, the market value of equity contains more diversified information other than changes in fair values. However, there is a concern in Models (1) to (3) of Table 2 when we examine the effects of variations in economic conditions on the OCI adjustments. Both improving and deteriorating conditions share one common intercept. Nevertheless, it is also possible that OCI adjustments are larger for firm-years with deteriorating than for firm-years with improving economic conditions despite the intercept identifies some effects of these large adjustments. To address the above concern, Model (4) of Table 2 reports results from regressions that include two independent dummy variables, D M and D S , to allow different intercepts for firm-years with deteriorating economic conditions. We still have significantly positive estimated coefficients, = 0.096 ( t -statistic = 34.66) and = 0.263 ( t -statistic = 78.7). Both indicate a positive association between upward OCI adjustments and improving market-based conditions. In contrast, the estimated coefficients, = -0.09 and = -0.111, are negative and highly significant (i.e., t -statistics are well above 27.91 in absolute value). Both indicate that there are significant differences between upward and downward OCI adjustments when the size of changes in economic conditions is identical.