Games of Economic Survival: Evidence from the Lab∗

Subjects in a controlled laboratory experiment choose a withdraw policy from a cash flow evolving according to an arithmetic Brownian motion. Earnings are based on total withdraws but aggressive withdraw policies expose the subject to risk of ruin. The policy that maximizes expected earnings does not, in general, maximize the odds of survival. We find that in treatments in which wealth maximization is consistent with a high rate of survival, the median subject quickly adjusts to the optimum. In treatments in which survival and wealth maximization are sharply at odds subjects exhibit a bias towards survival, persistently holding excess cash. The experimental design ensures that this survival bias is not a simple outgrowth of standard risk aversion.