Does Soft Information Matter? An Application of Gravity Models to Financial Analysts' Forecasts
21 Pages Posted: 12 May 2011
Date Written: May 1, 2011
We analyze the determinants of financial analysts' forecast accuracy. The empirical literature has enlightened variables related to analysts, to firms or both, in explaining the magnitude of forecast accuracy. But this literature does not explain in a common framework two opposite theoretical results about 'biases' in analysts' forecasts: 1) Firms may manage 'good earnings surprises' by disclosing earnings higher than analysts' forecasts (negative forecast errors) 2) Analyst may preserve 'management access' to 'soft', qualitative, unobservable information by systematically releasing optimistic forecasts (positive forecast errors). We test to what extent forecast accuracy would differ according to specific 'analyst-firm' pairs. The sample consists of 102 876 one-year-ahead Earnings Per Share forecasts about 243 French firms (1997-2007) from 1 966 analysts. We apply a panel regression model with fixed effect decomposition, derived from 'gravity models' used in international economics, in order to assess the weight of 'analyst-firm pairs' effect on forecast accuracy. The results are able to reconcile the two theories: controlling for all usual variables, when the pair-effect is weak, forecasts errors are negative (those pairs manage good surprises) while when the pair-effect is high, forecasts errors are positive (analysts preserve management' access).
Suggested Citation: Suggested Citation