Résumé
This article examines the link between uncertainty and investors' reaction to goodwill write-offs (GWWOs) for a sample of French firms during the period 2001- 2004. Our theoretical setting is derived from Daniel, Hirshleifer and Subrahmanyam (1998, hereafter DHS98) who posit that overconfidence leads to an overreaction to private information, followed by too little adjustments when the information becomes public and then a long adjustment which reduce slowly the mispricing in the long run. We consider three proxies for uncertainty - stock return volatility, analyst coverage and dispersion in analyst forecasts - and sort two samples of GWWOs according to the level of uncertainty. Our results confirm DHS98 model and, indirectly, that overconfidence is boosted by uncertainty. We identify a particular corporate event - here a bad signal: goodwill write-offs - and a particular context - high uncertainty - that fit DHS98 model, allowing private information prospecting, overconfidence in this information and arbitrage obstacles. Our tests confirm the overconfidence effect on investors' reaction: the high-uncertainty sample is characterized by strongly negative abnormal returns during the period preceding GWWOs announcement, associated with high volatility. At the announcement date, negative abnormal returns are observed in line with the self-attribution bias effect (the overreaction is strengthened by a confirming signal). The overreaction to private information is corrected in the long run, where we observe positive abnormal returns, creating a reversal. No abnormal returns are observed for the low-uncertainty sample. This study offers interesting insights in two ways: (i) in the area of financial markets and efficiency, it provides a test of a major over- and under-reaction model, (ii) in the area of corporate finance and accounting, it helps to explain investors' reaction to corporate financial disclosure according to a theoretical approach of information process and inference