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

Asymmetric Valuation Adjustments in Accumulated Other Comprehensive Income 148 decrease more when activity falls by an equivalent amount. They argue that this approach provides several advantages. First, the ratio form and log specification improve the comparability of the variables across firms and alleviate potential heteroskedasticity when estimations are cross-sectional for a wide variety of industries and large differences in terms of firm size. 9 Second, the log specification also accommodates our research questions and economic interpretation of the estimated coefficients. Specifically, D M is 0 when the market value of equity suggests an improvement in business conditions from period t –1 to t . In this case, we expect a positive coefficient β M , which measures the percentage increase in net assets through OCI for firm i that has a 1% improvement in its business conditions. Similarly, D S is 0 when sales revenue increases. In this case, we expect a positive coefficient β S , which measures the percentage increase in net assets through OCI for firm i with a 1% deterioration in its business conditions. On the other hand, the value of D M is 1 when the market value of equity reveals that business conditions of firm i deteriorate from period t –1 to t . In this case, the sum of the coefficients, β M + β M , measures the total changes (i.e., percentage decrease) in net assets with a 1% deterioration in business conditions. If managers’ OCI adjustments to variations in business conditions are asymmetric, their upward adjustments of fair value with respect to an improving business condition should be greater than their downward OCI adjustments of that with respect to a deteriorating business condition of equivalent size. For this hypothesis of asymmetric adjustments to hold, β M needs to be negative and makes β M + β M less than β M . Similarly, when sales revenue for firm i decreases from period t –1 to t , D S equals 1 to proxy for a deteriorating business condition. In this case, a negative β S that makes β S + β S less than β S suggests that downward OCI adjustments due to a deteriorating condition are less than upward OCI adjustments due to an improving condition of equivalent size. Prior studies, for example Ohlson (1995), usually regress the market value of equity on reported book value. Nevertheless, in this study our use of equity value as an independent variable is a reverse regression. We follow Beaver, Lambert, and Morse (1980) and Basu (1997) to reverse the equity-book value regressions for two reasons. 9 Empirically, the Davidson and MacKinnon (1981) test rejects the linear form in favor of the log linear model. However, the results are qualitatively similar to those presented for all of our models when we estimate them using linear specifications.