The Relative Risk Aversion Riddle

Purpose: Motivated by the risk aversion puzzle, this article develops a theoretical asset-pricing model that incorporates loss aversion into a framework of state-dependent recursive preferences. The objective is to examine whether allowing preferences to vary across states of nature can help reconcile empirically observed levels of risk aversion with standard asset-pricing implications. Design/Methodology/Approach: The model features a representative agent who is constrained to be risk-averse in all states of nature. Loss aversion is embedded directly into the recursive utility function, allowing attitudes toward gains and losses to differ across states. The model is calibrated using data from a set of selected developed economies in order to assess its quantitative performance and empirical relevance. Findings: The model is able to generate a coefficient of relative risk aversion below ten, which is broadly consistent with empirical estimates and improves upon standard benchmarks. However, despite this improvement, the implied stochastic discount factor fails to match key empirical moments, limiting the model’s ability to fully resolve asset-pricing anomalies. Practical Implications: Furthermore, extending the framework to include a representative agent with S-shaped loss-averse preferences does not lead to a well-behaved solution, suggesting that loss aversion alone is insufficient to address the observed discrepancies in asset pricing. Originality/Value: Overall, the results highlight a previously unexplored limitation of incorporating loss aversion into state-dependent recursive preferences and uncover a new dimension of the risk aversion puzzle. The findings suggest that additional mechanisms beyond loss aversion may be required to jointly explain observed risk attitudes and asset-pricing behavior.