Statement
Generalized disappointment aversion preferences combined with persistent long-run volatility risk (without a predictable component in expected consumption growth) suffice to explain the equity premium, return volatility, low risk-free rate, and return predictability patterns observed in U.S. data. This is in sharp contrast to Kreps-Porteus preferences, where the long-run risk in mean channel is essential.
Evidence summary
Bonomo, Garcia, Meddahi, and Tedongap (2011) show using closed-form Markov-switching solutions:
- Random walk consumption + GDA: equity premium 7.21%, sigma(R) 19.33%, E[Rf] 0.93%, return predictability R^2 increasing with horizon.
- Adding LRR in mean (full Bansal-Yaron model) with GDA: statistics nearly identical.
- With Kreps-Porteus preferences: removing LRR in mean collapses the equity premium.
- Mechanism: persistent volatility increases the probability of disappointing outcomes under GDA, making the SDF more volatile and generating counter-cyclical effective risk aversion.
Conditions and scope
- Depends on specific GDA calibration (gamma=2.5, alpha=0.3, kappa=0.989).
- Risk-free rate volatility is undermatched (2.34% vs 4.1% in data).
- The 4-state Markov switching approximation may introduce discretization bias.
- Results hold for EIS below 1 as well, but with higher risk-free rate level and volatility.
Counter-evidence
- The GDA parameters are not directly estimated from micro data; identification is indirect.
- higher-order-effects-asset-pricing-models shows that Markov-switching approximations of persistent processes may introduce their own errors.
Linked ideas
Open questions
- Can GDA preferences be identified from micro-level consumption and portfolio data?
- How do GDA preferences interact with regime-switching macro fundamentals (not just endowment)?