A Structural Model of Analyst Forecasts: Applications to Forecast Informativeness and Dispersion

Abstract

We modify Morris and Shin (2002) to develop a structural model of analyst earnings forecasts. The model allows for analysts to herd due to informational effects and non-informational incentives. The benefits of our model are twofold: (1) we can decompose earnings forecasts into informational and bias components, and measure the stock price response to each component, and (2) we can estimate the impact of bias on the dispersion in analyst forecasts. In a pair of empirical exercises, we find a strong relation between the informational component of analyst forecasts and announcement period stock returns. We also find that analyst biases do not have an impact on forecast dispersion.