THE ECONOMIC VALUE OF FINANCIAL DISTRESS INFORMATION: AN EMPIRICAL ASSESSMENT

SU-JANE HSIEH, Purdue University

Abstract

After accounting numbers are applied to a selected bankruptcy prediction model, new information concerning changes in the firm's financial health can be obtained. The purpose of this project is to investigate whether the market rewards the acquisition of this information. If it does, is this reward realized before or after the annual report release date? Based on the market equilibrium model proposed by Grossman and Stiglitz (1980), any private information should be compensated to preserve the incentive for information collection and to make market equilibrium possible. In this case, the return to acquiring information concerning financial distress should be realized prior to the release of the accounting numbers. Moreover, if one believes in semi-strong market efficiency, one will expect no excess returns associated with the financial distress information after the annual report release date, since this is publicly available information. A few technical problems emerged in this project. They are: the estimation of the systematic risk of potentially bankrupt firms, the determination of the optimal cut-off point for classification, and the type I and type II error cost estimation. A random coefficient model developed by Chen (1981) was used to estimate the systematic risk of the potentially failing firms when necessary. The type I and type II error cost was estimated using samples of actually bankrupt firms and the matched firms by assuming a wrong classification for each firm. The estimated error cost was applied to a Bayesian decision model to obtain the optimal cut-off point which minimizes the total error cost. The results of this project are consistent with the following expectations: a significant negative residual return was detected for the 156 selected firms with predicted bankruptcy status changes for the prior-RRD period, but no significant residual return was found for the post-RRD period. Moreover, different return generating models were applied to see whether the results were sensitive to model specifications. This study indicates that the results are robust to return model specifications.

Degree

Ph.D.

Subject Area

Accounting

COinS