Hardening Soft Information: Analyst Conservative Bias

Abstract

In this paper, we examine whether sell-side financial analysts show a bias when translating their soft information into a hard format. Sell-side analysts produce both soft research output, in the form of a textual report, and hard research output, including earnings forecasts, target prices, and stock recommendations. In our study, we find evidence that analysts’ hard outputs undershoot the neutral implication of their own soft output. Furthermore, our cross-sectional results show that our observed conservative bias increases when the underlying information signals are of poorer quality, which we measure by the forecast horizon, linguistic cues in the report, and characteristics of the firms’ information environments. Consistent with the well-known analyst optimism, we find that hard outputs assimilate analysts’ soft output more conservatively when their soft output conveys bad news. Our findings suggest that the fundamental distinctions between soft and hard information lead to a predicable bias when analysts harden their soft information.

Keywords: Analyst, Soft Information, Textual Analysis, Tone, Forecast Bias, Underreaction

JEL Classification: G23, G24, G14, G10, G20, M41, D82, D83, C11