Abel's (1998) intertemporal asset pricing model implies that the autocorrelation pattern in expected returns reflects that observed in output growth rates. Consequently, by using the observed autocorrelation properties of macroeconomic data, we are able to provide univariate tests with power to detect deviations from the stationary random walk model over the post-World War II sample period. After regressing excess returns against industrial production's cyclical component, these univariate tests provide little evidence of serial correlation in the resultant residuals, confirming the presence of a business cycle effect in excess returns. However, our multivariate analysis concludes that while the business cycle contributes to these deviationsf rom the stationary random walk model, predictable long term swings in expected returns arising from variable trends in macroeconomic data still remains.