The Complexity of Choice and Loss Aversion

Abstract. We present a “satisficing” choice model, in which the consumer looks for a feasible bundle of goods that induces a payoff above a certain threshold. We demonstrate that this problem is computationally difficult (NP-Complete), and therefore solving it entails a significant computation cost. We show that this cost can induce loss aversion, and that the induced loss aversion is non-monotonic with respect to the size of the loss. We present a simple quantitative model and test it on existing experimental results. Our model fits well the experimental results, and in some relevant domains, it has better predictive power than prospect theory.