I show that time-varying risk aversion can generate a term structure of equity risk premia that is upward sloping in bad times and downward sloping in good times. I derive three conditions that jointly generate this result. First, risk aversion is negatively correlated with consumption growth. Second, risk aversion is mean reverting and, third, preferences can become risk seeking in good states. Moreover, I propose a highly parsimonious model with one single state variable and a representative agent with utility over both consumption and wealth fluctuations that allows for states of risk-seeking preferences in equilibrium and, hence, can match the recent empirical findings on the equity term structure. In my model, the unconditional term structure of equity risk premia can be increasing, flat, or decreasing depending on the relative frequency of good and bad times. I show that time-varying risk aversion provides the first simultaneous explanation for the cyclicality of the equity term premia, the positive equity term risks, and, the predictability of stock returns without needing to assume predictability in cashflows. More broadly, my results show the relevance of time-varying risk aversion for explaining maturity-dependent risk premia.

We study the effects of limited attention on belief formation in a dynamic environment. We propose a simple theoretical model of inattentive learning and derive closed-form expressions for the limiting distortion of the true underlying distribution. When attention is drawn towards extreme observations, the limiting probability distortion takes the form of an S-shape. These belief distortions yield behavior that is similar to that implied by the preference-based distortions of prospect theory. We thus find that part of the empirical success of prospect theory might be due to abstracting away the impact of limited attention on learning. We show that our model of inattentive learning explains several aspects of investors' behavior, for instance, a preference for skewed assets; overextrapolation of past returns; stock market momentum; and the simultaneous demand for lotteries and insurance. When attention has cognitive costs, limited attention thus provides a rational unification of several biases in investor behavior.

We document a new phenomenon of IPOs, namely, that the trading volume in the first two years after an initial public offering (IPO) is u-shaped. Our findings are strongest for small stocks. While there are explanations for the high initial turnover using models with heterogeneous priors, the upward slope of the trading volume has neither been reported empirically, nor - as we show in the paper - can standard models account for this empirical feature. We propose a general equilibrium model with heterogeneous agents who differ both in their prior beliefs and in their degree of ambiguity aversion. The demand for the IPO asset of the ambiguity averse investors increases over time as the ambiguity of the expected return decreases when more returns realize. We show that the combination of heterogeneous priors with ambiguity aversion, where ambiguity is resolved over time, can explain the u-shaped trading pattern observed in the data.