Signal Strength, Conflicting Signals, and Belief Updating : Evidence from Sell-Side Analysts’ Forecasts

I empirically study how confirmatory bias works for strong and contradictory signals using sell-side analysts’ data. I first model an agent who is prone to confirmatory bias and whose task is to value a stock based on a signal, and then introduce the effects of signal strength by relaxing Rabin and Schrag's (1999) assumption of a constant bias severity. I then use target prices to measure forecast bias and the growth in Earnings Per Share as signals, and regress analysts' forecast bias over different deciles of favorable signals interacted with prior negative forecast bias in a dynamic panel data model. I find that analysts do not react positively to favorable signals when the prior is pessimistic, except for sufficiently strong signals, which cause analysts to issue more optimistic target prices