Identifying Beliefs from Asset Prices

We propose a novel procedure to identify the marginal stock market investor's beliefs from observed asset prices. Our approach recovers price-consistent beliefs, i.e. the distribution of macro and financial variables that satisfy the conditional Euler equations, given a cross-section of assets, a pricing kernel, and a conditioning set. Our procedure is nonparametric, not requiring any assumptions about the data generating process or investor rationality. The recovered beliefs about consumption growth exhibit non-Gaussianity, strong procyclicality in the mean, flat volatility, and bimodality in recessions. Beliefs about the stock market are countercyclical and correlate with survey data on institutionalinvestors' expectations