Previous studies of loss aversion in decisions under risk have led to mixed results. Losses appear to loom larger than gains in some settings, but not in others. The current paper clarifies these results by highlighting six experimental manipulations that tend to increase the likelihood of the behavior predicted by loss aversion. These manipulations include: (1) framing of the safe alternative as the status quo; (2) ensuring that the choice pattern predicted by loss aversion maximizes the probability of positive (rather than zero or negative) outcomes; (3) the use of high nominal (numerical) payoffs; (4) the use of high stakes; (5) the inclusion of highly attractive risky prospects that creates a contrast effect; and (6) the use of long experiments in which no feedback is provided and in which the computation of the expected values is difficult. In addition, the results suggest the possibility of learning in the absence of feedback: The tendency to select simple strategies, like “maximize the worst outcome” which implies “loss aversion”, increases when this behavior is not costly. Theoretical and practical implications are discussed.

We build a stochastic discount factor—SDF—using U.S. domestic financial data only, and provide evidence that it accounts for stylized facts about foreign markets that escape SDFs generated by consumption-based models. When our SDF is interpreted as the projection of the pricing kernel from a fully specified model in the space of returns, our results indicate that a model that accounts for the behavior of domestic assets goes a long way toward accounting for the behavior of foreign asset prices. In our tests, we address predictability, a defining feature of the forward premium puzzle—FPP—by using instruments that are known to forecast excess returns in the moment restrictions associated with Euler equations both in the equity and in the foreign markets.

This paper studies asset returns adopting an alternative strategy to assess a model's goodness of fit. Based on spectral analysis, this approach considers a model as an approximation to the process generating the observed data, and characterizes the dimensions for which the model provides a good approximation and those for which it does not. Our aim is to offer new evidence regarding the size and the location of approximation errors of a set of stochastic growth models considered to be decisive steps in the progress of the asset pricing research program. Our specific objective is to reevaluate the results of Jermann's (1998) model extending the calculations to the spectral domain. Spectral results are relatively satisfactory: the benchmark model needs very few contributions of approximation errors to account for the empirical equity premium. Second, the location of the approximation errors, when they are substantial, seems to be essentially concentrated at high frequencies.

We study the quantitative properties of a dynamic general equilibrium model. Agents face both idiosyncratic and aggregate income risk, state-dependent borrowing constraints that bindoccasionally, and markets that are incomplete. Equilibrium consumption-savings plans and asset pricesare computed under various assumptions about income uncertainty. Then, we investigate whether themodel replicates two empirical observations: the high correlation between individual consumption andindividual income, and the equity premium puzzle. We find that, when the driving processes arecalibrated according to the data from wage incomein different sectors of the U.S. economy, theresults move in the direction of explaining these observations,but we fall short of explaining theobservations quantitatively. If the incomes of agentsare assumed to be independent of each other, theobservations can be explained quantitatively.