This paper examines whether two well-known models, Campbell and Cochrane’s habit model (1999) and Bansal and Yaron’s long-run risks model (2004), can produce significant return predictability. We find that a state variable, the surplus consumption ratio in the habit model, explains the countercyclical time-varying risk premium. In the long-run risks model, the main sources of volatility in the price–dividend ratio are a persistent and predictable consumption growth rate and economic uncertainty. Following Kirby’s work in 1998, we test whether the two models can explain the observed risk, ie, return volatility. Both models fail to generate any significant predictive power. However, the habit model exhibits relatively strong volatility, which implies that variation in the expected excess returns is largely attributable to the time-varying risk aversion.