Correlated Errors - Why a Monotone Relationship between Forecast Precisionand Trading Profitability May Not Hold

Abstract:     
This paper argues that the relation between financial analysts' earnings forecast accuracy and their recommendation profitability has to be augmented by the extent of commonality in their forecast errors. We show that while accuracy is positively related to expected performance, the correlation in forecasting errors has a negative impact. This implies that a monotonic relationship between ex ante identifiable forecast accuracy and ex post recommendation profitability does not need to hold. Thus, agents may be better off by making comparatively large but less correlated errors, than making precise but highly correlated forecasts.

Number of Pages in PDF File: 40

Keywords: Forecast accuracy, analysts recommendation profitability, learning, Kalman filter

JEL Classification: M4, G14

working papers series