Value of Latent Information: Alternative Event Study Methods

Abstract

This paper presents an econometric model to value latent information underlying corporate events. This model computes the market's inference of the value of latent information from the probability of an event, conditional on firm-specific, preevent information. It provides a convenient framework for testing significance of preevent information variables, such as accounting attributes and lagged stock return. Simulations show that this model, when applied to both event and preevent period data, can decrease the incidence of bias in event studies, If restricted to only event period data, this model reduces to a truncated regTession and does not perform as well as standard procedures. THIS PAPER PRESENTS AN econometric approach to measure the value of latent information underlying corporate events and conducts simulation experiments to analyze the performance of alternative event study methodologies. The original study by Fama, Fisher, Jensen, and Roll (1969) interprets the mean abnormal common stock return over the event period as the value of information released. This and numerous other event studies that followed recognize that events are not complete surprises and estimate the value of information by implicitly or explicitly accounting for the partially anticipated nature of events.1 But, the (standard) methodology used in these studies does not account for the market's inference of the value of information released by a probable event, given publicly available information. Most corporate events result from some decision process although the standard methodology presumes that events are exogenous. Decisions triggering events usually involve latent information not completely available to the market. For example, when a firm announces its bid to acquire a target * Board of Governors of the Federal Reserve System. Opinions expressed in this paper do not necessarily represent the official position of the Board of Governors of the Federal Reserve System or its staff. The author is thankful to Stephen Brown, Gregory Connor, Doug Diamond,