Does uncertainty boost overconfidence? The case of financial analysts' forecasts

Abstract

Purpose – This article aims to examine the link between uncertainty and analysts' reaction to earnings announcements for a sample of European firms during the period 1997-2007. In the same way as Daniel et al ., the authors posit that overconfidence leads to an overreaction to private information followed by an underreaction when the information becomes public. Design/methodology/approach – In this study, the authors test analysts' overconfidence through the overreaction preceding a public announcement followed by an underreaction after the announcement. If overconfidence occurs, over- and underreactions should be, respectively, observed before and after the public announcement. If uncertainty boosts overconfidence, the authors predict that these two combined misreactions should be stronger when uncertainty is higher. Uncertainty is defined according to technology intensity, and separate two types of firms: high-tech or low-tech. The authors use a sample of European firms during the period 1997-2007. Findings – The results support the overconfidence hypothesis. The authors jointly observe the two phenomena of under- and overreaction. Overreaction occurs when the information has not yet been made public and disappears just after public release. The results also show that both effects are more important for the high-tech subsample. For robustness, the authors sort the sample using analyst forecast dispersion as a proxy for uncertainty and obtain similar results. The authors also document that the high-tech stocks crash in 2000-2001 moderated the overconfidence of analysts, which then strongly declined during the post-crash period. Originality/value – This study offers interesting insights in two ways. First, in the area of financial markets, it provides a test of a major over- and underreaction model and implements it to analysts' reactions through their revisions (versus investors' reactions through stock returns). Second, in a broader way, it deals with the link between uncertainty and biases. The results are consistent with the experimental evidence and extend it to a cross-sectional analysis that reinforces it as pointed out by Kumar.