One signal, two opinions : strategic heterogeneity of analysts' forecasts

We present a model of investors acquiring forecasts from a group of investment analysts. Investors may pick an analyst based on his past performance. In the literature it is typically assumed that agents' rewards depend solely on the type they are perceived to be, which leads to typical herding results. In contrast here, analysts' rewards not only depend on their own reputation but also on the number of analysts with a similar reputation. There exist two interesting types of equilibria: in the first type it is optimal for investors to ignore analysts' past performance, even though analysts make predictions according to their best knowledge. In a second type investors do use past performance to select analysts. However this induces analysts to predict strategically, i.e. some analysts knowingly make wrong predictions