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

Asymmetric Valuation Adjustments in Accumulated Other Comprehensive Income 154 5.2 Time Span and Asymmetric OCI Adjustments In this section, we test whether information inadequacy is a constraint and makes OCI adjustments asymmetric. To do so, we aggregate periods to allow managers to have more information. In Table 3, the estimated coefficients for are consistently positive from 0.104, 0.134, to 0.153 when the length of the information period is extended from two to three and then four years. Their t -statistics also remain high and are 33.07, 35.04, and 34.03 from Models (1) to (3), respectively. All of them are significant at the 1% level. In Table 3, as the information window is extended from two to four periods, the estimated coefficients, , remain negative, and are -0.042 in Model (1), -0.051 in Model (2), and -0.029 in Model (3). These negative coefficients run counter to the argument that attributes asymmetric OCI adjustments to inadequate information inputs. The combined values of + remain less than . In Model (1), + equals 0.062 (i.e., 0.104 – 0.042). They are 0.803 (i.e., 0.134 – 0.051) in Model (2) and 0.124 (i.e., 0.153 – 0.029) in Model (3). All of these combined coefficients (i.e., + ) are less than because is negative. Once again, variations in sales revenue (i.e., accounting-based inputs) over adjacent periods have greater explanatory power for OCI adjustments, compared with the market- based information inputs. Specifically, the estimated coefficients, , remain stable, and are 0.396, 0.439, and 0.466 from Models (1) to (3), respectively, when the adjacent information windows are extended from two to three and then four years. Their t -statistics also remain stable, and are 79.36, 72.32, and 64.99 from Models (1) to (3), respectively. Thus, for a 1% increase in sales revenue, OCI adjustments increase net assets by approximately 0.396% to 0.466%. In contrast, ranges from 0.104, 0.134, to 0.153. For the results supporting H2, the estimated coefficients, , from Models (1) to (3), are -0.032 ( t -statistic = -3.5), -0.050 ( t -statistic = -3.36), and -0.107 ( t -statistic = -5.24), respectively. They become more negative when the information windows extend from two years in Model (1) to four years in Model (3). When combining values, + , we have 0.364 (i.e., 0.396 – 0.032), 0.389 (i.e., 0.439 – 0.050), and 0.359 (i.e., 0.466 – 0.107) in Models (1), (2), and (3), respectively. Overall, the magnitude of upward adjustments is always larger than that of downward adjustments. Model (4) of Table 3 uses information inputs that are one-year forward as a robustness test. Results consistently demonstrate that additional information inputs do not mitigate the asymmetry of OCI adjustments. Specifically, both and are negative