Latent Utility Shocks in a Structural Empirical Asset Pricing Model

We consider a random utility extension of the fundamental Lucas (1978) equilibrium asset pricing model. The resulting structural model leads naturally to a likelihood function. We estimate the model using U.S. asset market data from 1871 to 2000, using both dividends and earnings as state variables. We find that current dividends do not forecast future utility shocks, whereas current utility shocks do forecast future dividends. The estimated structural model produces a sequence of predicted utility shocks which provide better forecasts of future long-horizon stock market returns than the classical dividend-price ratio.