Market Reactions to Accounting Policy Deliberations: The Inflation Accounting Case Revisited - A Reply.

Abstract The Basu and N&S methodologies differ on three dimensions: choice of test statistic, method of computing significance levels, and method for sampling from among the possible observable reactions. In our opinion, neither methodology dominates the other, and we could easily envisage future researchers drawing from both approaches. To a large extent, the three basic differences between the methodologies involve independent choices, so that it is possible to choose from the best features of each methodology. For example, a researcher might decide to use correlations as the test statistic (N&S), compute significance levels relative to an empirical distribution (N&S), and test for a significant correlation between the abnormal returns for one "partitioning" event and the abnormal returns for each of the other possible events (Basu).