We present a novel methodology for estimating/testing the Bansal-Yaron (2004) and related long-run risks (LRR) models based on the observation that the latent state variables are known functions of observables. The large standard error of the estimated IES explains the controversy on its magnitude. The model requires higher persistence of consumption and dividend growth to explain the cross-section of returns than that observed in the data. The model matches the unconditional moments of

more »

... ion and dividend growth, but implies a higher risk free rate and lower volatility of the price-dividend ratio, risk free rate, and market return than those observed in the data. Contrary to the model implications, the conditional variance of the LRR variable fails to capture the large time-variation in the equity premium.