Empirical Implications of Analyst Forecast Dispersion to the Information Dynamics of Valuation Models: A Two Stage Approach

Abstract

Ohlson (1995) models firm value as a function of abnormal earnings, net book value and other unspecified information. Ohlson (2001) proposes consensus analyst forecasts as a proxy for the previously unspecified other information in his model, which we test using a two stage approach. The first stage identifies information in analyst forecasts that is reflected in current earnings and net book value, and the second stage regresses the first-stage residuals as the proxy for other new information. Our initial results using price-levels regressions concur with Dechow et al.'s (1999) findings that short-run consensus analyst forecasts are effective proxies for other information, and that the proposed model is no more descriptive than capitalizing short-run forecasts in perpetuity. We find that with high forecast dispersion, however, the effectiveness of analyst forecasts as well as the association between earnings and market values are diminished. Overall, we find that the descriptive ability of both the Ohlson model and the capitalized forecast model is dampened with high forecast dispersion, but the dampening is more severe for the capitalized forecast model, suggesting that the descriptive ability of Ohlson's valuation framework is strongest, relative to capitalized analyst forecasts, when uncertainty and information asymmetry are most severe. In contrast to our (and Dechow et al.'s) price-levels regression results, we find with returns regressions that Ohlson's model is consistently and significantly more descriptive than a model that simply capitalizes changes in analyst forecasts.