Information Content of Accounting Announcements.

The efficient market hypothesis (EMH) has spawned a large and impressive body of "event studies." Conditional on the maintained hypothesis of market efficiency, these studies test the "information content" of events such as accounting earnings releases (Fama 1991). However, the notion of "information content" of accounting earnings has not been formally defined, except implicitly--via the way it is calibrated in event studies--as percent price change (risk and size adjusted) associated with earnings releases. In the assumption of belief homogeneity underlying the original formulation of EMH, where no competitive non-null trading is expected in equilibrium, price changes will reflect fully the public announcement (see, e.g., Grossman and Stiglitz 1980). Hence, price changes will, conditional on the endowment and preferences configuration and on prior homogeneously held beliefs, constitute a one-to-one mapping with the belief change induced by the earnings release, and could be adopted as an operational definition of information content (IC). However, admitting the possibility of belief-change heterogeneity into the analysis brings into question the theoretical validity and plausibility of this operational definition. Competitive trading that becomes possible as a result of disagreement would perturb the one-to-one mapping between price change and belief change induced by the earnings release. In any case we no longer have a well-defined single construct reflecting heterogeneous belief changes. Under belief homogeneity the market acts as if it were a single decision-maker whose belief change theoretically reflects and calibrates the value or "content" of information. Once we allow for multiple decision-makers, or equivalently, for belief-heterogeneity, however, this simple definition of IC is no longer applicable. How is one to aggregate the belief changes of different traders, and what weights should be attached to such changes in the process of aggregation? Moreover, even if an appropriate operational definition of IC were to be discovered under such circumstances, what would be the empirical manifestations? The derived "information content" measure includes volume reaction, price reaction, and the predisclosure expected belief dispersion. The results imply that neither price reaction nor volume reaction alone will provide a complete characterization of information content; both must be jointly observed along with some aspects of the existing information environment to gauge information content fully.