Content area
Full Text
1 Introduction
Experimental evidence in psychology shows that behavioral biases arise in situations that require more judgement. In particular, people exhibit a higher level of overconfidence when they are involved in non-mechanical tasks and when predictability is low and evidence ambiguous ([29] Lichtenstein and Fischhoff, 1977; [21] Griffin and Tversky, 1992). "People are typically overconfident about their knowledge when the issues at hand are difficult" ([33] Shefrin, 2008, p. 58). When uncertainty is high, people tend to construct scenarios and are overconfident in the probability of their success ([26] Kahneman and Tversky, 1979).
In the context of financial decision, [12] Daniel and Titman (1999), [25] Hirshleifer (2001) and [10], [11] Daniel et al. (1998, 2001), posit that uncertainty intensifies psychological biases[1] . They underline the role of overconfidence in the mispricing of hard-to-value stocks and refer this precisely to "R&D-intensive firms comprised largely of intangible assets" ([11] Daniel et al. , 2001, p. 935). [10] Daniel et al. (1998) produces a theoretical model that explains mispricing by over- and underreaction to information caused by overconfidence. The model shows that overconfident investors overreact to private information, and then underreact when information becomes public[2] . In line with [10], [11] Daniel et al. (1998, 2001), this paper focuses on analysts' response to private and public information. We consider earnings announcements and two subsamples - high-tech and low-tech firms - and we examine whether analyst forecasts reflect over- or underreaction to information.
Analysts' overconfidence receives relatively little attention from researchers, compared to that of investors and to the large body of research devoted to analysts' optimism. Many papers document the fact that analysts inefficiently incorporate information, mainly by analyzing how a current earnings forecast for a given period is influenced by earnings for the previous period. They show a serial correlation between current and past errors in forecasting ([31] Mendenhall, 1991; [1] Abarbanell and Bernard, 1992; [2] Ali et al. , 1992). These findings suggest that analysts underreact to new public information (the earnings release), while the pioneer study from [15] De Bondt and Thaler (1990) documented an overreaction. [16] Easterwood and Nutt (1999) show that both misreactions can be observed, and also document that analysts overreact to positive news and underreact to negative news, producing a...