Explaining satisficing through risk aversion

Abstract

This paper extends the analysis of the data from the experiment of Hey et al. (2017), which was designed to test Proposition 2 of the theory of Manski (2017). I focus on how the subjects select the aspiration levels when they choose to satisfice, and try to find a better explanation for that story than that of Manski. I assume that the subjects are Expected Utility (EU) (rather than MiniMax regret) agents and that they think of the payoffs as having a uniform risky (rather than an ambiguous) distribution. I consider two special cases of the EU preferences: CRRA and CARA; and I combine these with two different stories for the stochastic specification of errors: beta and normal. To give a fair comparison in finding a better explanation of the individual behaviour, I also fit the data using Manski’s optimal strategy under both stochastic specifications. I estimate using maximum log-likelihood. The estimation is done subject by subject. The results tell us that assuming that the subjects are EU agents and that they see the payoffs as uniformly distributed produces a better statistical explanation than that of Manski. That is the actual aspiration levels are statistically closer to the optimal aspiration levels assuming CRRA and CARA than those of Manski’s prediction. Interestingly, the subjects in the Hey et al. (2017) experiment appear to be risk loving when selecting their aspiration levels.

Keywords: expected utility; risk aversion; maximum log-likelihood

Revise and resubmit to Theory and Decision