Time Variations in the Equity Premium : Is It Habit Formation or Loss Aversion?

The large spread between equity returns and risk-free rates (the "equity premium puzzle") has been the subject of intense debate. Two main families of models claim to solve this puzzle: habit-formation models and loss-aversion models. The goal of this paper is to assess empirically which of them best fits the observed excess returns. In order to do this, I first show how to express both models as linear stochastic discount factor models. This gives explicit and testable constraints for the excess return variations. I then compare the theoretical evolution generated by these models with the observed path. I find that the constant relative risk aversion model and the external habit model are not likely to correspond to the observed data. The loss aversion model and the internal habit model could fit the observed excess return evolution