We investigate the ability of a representative agent model with time-separable utility to explain the first and second moments of the risk-free rate and the return to equity. We generalize the standard calibration methodology by accounting for the uncertainty in both the sample moments to be explained and the estimated parameters to which the model is calibrated. We find that the first moments of the data can be matched for a wide range of preference parameter values but the model is unable to generate both first and second moments of returns that are statistically close to those in the sample.