One of the most relevant evidences of experimental psychology is that individuals typically overreact to facts. This paper studies the overreaction of a group of recognized influence on market expectations - security analysts - in earning forecasts of listed companies. Besides observing this regularity in the Brazilian market, it also studies how the pattern of estimates changes as the disclosure of balance sheets approaches. The way the bias is corrected - smoothly or sharply, before revealing the results or exactly at that moment - probably influences the level of difficulty for agents to detect it.
