Zusammenfassung:
I generalize the long-run risks (LRR) model of Bansal and Yaron (2004) by incorporating recursive smooth ambiguity aversion preferences from Klibanoff et al. (2005, 2009) and time-varying ambiguity. Relative to the Bansal-Yaron model, the generalized LRR model is as tractable but more flexible due to its separation of ambiguity aversion from both risk aversion and the intertemporal elasticity of substitution. This three-way separation allows the model to further account for the variance premium puzzle besides the puzzles of the equity premium, the risk-free rate, and the return predictability. Specifically, the model matches reasonably well key asset-pricing moments with risk aversion under 5. Model calibration shows that the ambiguity aversion channel accounts for 77 percent of the variance premium and 40 percent of the equity premium.