We structurally estimate an investment-based asset pricing model, where firms’ exposure to macroeconomic risk is unknown. Bayesian beliefs about this parameter are updated from firms’ and industry peers’ comovement between their productivity and consumption growth. The model implies that discount rates rise endogenously with the perceived risk exposure of firms, thereby depressing investment and valuation ratios. We test these predictions in the data and find strong support for them. In particular, we find that cross-sectional learning from peers is crucial in this context and alternative risk estimates, which ignore peer observations, do not predict firm variables.