Abstract
We use panel data to examine the prediction of Merton's intertemporal CAPM that time varying risk premiums arise from the conditional covariances of returns on assets with the return on the market. We find a positive and significant risk-return tradeoff that is driven by the time series variation in the conditional covariances, and the risk-premium on the market remains positive and significant after controlling for additional state-variables. Our estimation method allows us to estimate the risk-return tradeoff in the ICAPM using a large number of test assets.