When is the Price of Dispersion Risk Positive?

We provide robust evidence that the price of analysts' dispersion risk in the cross-section of stock returns changes sign over time, and in particular, turns positive (negative) in periods of high (low) analysts' dispersion. Our result holds for a set of portfolios that are double-sorted on their betas and their coefficients on aggregate dispersion, as well as 36 different sets of test portfolios created by Kenneth French. We construct a general equilibrium model in which analysts of different types have heterogeneous beliefs about aggregate earnings growth. The consumer does not trust either analyst fully, and dynamically adjusts the weight given to each analyst, given the history of their past forecast performance. In equilibrium, each asset's risk premium depends on its exposure to three factors: (i) the market portfolio, (ii) the macroeconomic factor, and, (iii) a ``flight-to-safety'' factor. The first term decreases with dispersion, while the third term increases. These changes occurs because investors shift into assets with lower cash flow betas during periods of high dispersion. We find strong support for such a flight-to-safety in the data.