Optimal Thresholds in Accounting Recognition Standards

Abstract

This paper investigates the design of recognition thresholds in accounting standards. In statistics, a threshold classi.es evidence to balance two types of recognition errors weighted by their respective costs to a decision maker. In accounting recognition standards, a threshold induces firms to respond strategically and thus affects the very distribution of evidence the threshold classi.es. With this strategic effect, the optimal recognition threshold is determined by not only the decision maker’s loss function but also the transaction’s features. We compare the optimal threshold’s properties under the statistical and strategic approaches, provide their respective empirical predictions, and discuss the limitations of using a statistical approach to guide accounting standard setting.