Risk Aversion in the Field : Evidence on Prevalence and Motives from an Employee Rewards Program

Given its importance for theory, welfare, and policy, economists have long sought to understand the prevalence of and motives for risk aversion in the field. In practice, this inquiry is often confounded by the potential for biased beliefs (e.g., betting, investing), imperfect understanding (e.g., insurance), or limited generalizability (e.g., game shows). We overcome these challenges with rare data detailing the choices, productivity, and beliefs of 20,133 employees across 18 large North American firms who participated in a simple, all-or-nothing, goal-rewards program with $9.4 million in incentives. We estimate nearly one-half of employees selected a goal lower than the EV-maximizing benchmark, assuming rational expectations, resulting in a 46 percent average loss of potential rewards. This conservative goal choice persisted across diverse financial stakes ($69 to $4,500) and employee experience. We additionally show that conservative choice cannot be explained by a standard expected utility (EU) model with plausible risk preferences or through common departures from EU such as biased-beliefs (employees exhibit substantial overconfidence about productivity), non-linear decision weights, or gain-loss utility. We replicate the pattern of conservative choice, corroborate limits of EU-based explanations, and rule out potential confounds through an incentive-compatible online goal-reward paradigm. We propose—and experimentally validate—a novel decision-heuristic in which risk averse choice emerges from an inferential bias due to contingency neglect in the context of pairwise comparisons. We conclude with experimental evidence suggesting that this heuristic offers a potential explanation for empirical puzzles in other risky-choice settings of economic interest such as deductible-based insurance and portfolio allocation