Empirical tests of Euler equations relating security returns and consumption usually appear to reject the model. Using a common specification of aggregate preferences and instrumental variables, this paper examines some potential reasons for rejections. The evidence indicates that maintained stationarity assumptions of previous tests fail for post-war U.S. quarterly and monthly data. Shifts in model parameters are found across policy regimes (pre-1951 and post-1979) and across stages of the business cycle (recession versus non-recession). Controlling for some of these factors, less evidence is found against a simple consumption-based asset pricing model in non-recession periods.

Tests of asset‐pricing models are developed that allow expected risk premiums and market betas to vary over time. These tests exploit the relation between expected excess returns and current market values. Using weekly data for 1963 through 1982 on ten common stock portfolios formed according to equity capitalization, a single‐risk‐premium model is not rejected if the expected premium is time varying and is not constrained to correspond to a market factor. Conditional mean‐variance efficiency of a value‐weighted stock index is rejected, and the rejection is insensitive to how much variability of expected risk premiums is assumed.