Are forecasts of corporate profits rational? A note and further evidence

Abstract

This paper revisits the claim by Keane and Runkle [J. Polit. Econ. 106 (1998) 768] that analyst forecasts of corporate profits may be rational, despite claims to the contrary by De Bondt and Thaler [Am. Econ. Rev. 80 (1990) 52] and others. We replicate the Keane and Runkle's [J. Polit. Econ. 106 (1998) 768] research method and test some of the underlying assumptions in their modelling strategy. Specifically, we examine two key assumptions, the independence of forecast errors across time and the failure of forecast errors to spillover industrial boundaries. We find strong evidence of autocorrelation and interindustry forecast error spillovers. Our findings prompt us to extend Keane and Runkle's tests to a revised General Method of Moments (GMM) estimate of model parameters as well as standard errors. Our revised GMM estimates employ two instruments designed to capture the impact of autocorrelation and spillover effects. These revised estimates suggest analyst forecasts of earnings are not rational with respect to publicly available information.