Survival versus Profit Maximization in a Dynamic Stochastic Experiment

Subjects in a laboratory experiment withdraw earnings from a cash reserve evolving according to an arithmetic Brownian motion in near-continuous time. Aggressive withdrawal policies expose subjects to risk of bankruptcy, but the policy that maximizes expected earnings need not maximize the odds of survival. When profit maximization is consistent with high rates of survival (HS parameters), subjects adjust decisively towards the optimum. When survival and profit maximization are sharply at odds (LS parameters), subjects persistently (and sub-optimally) hoard excess cash in an evident effort to improve survival rates. The design ensures that this hoarding is not due to standard risk aversion. Analysis of period-to-period adjustments in strategies suggests instead that hoarding is due to a widespread bias towards survival in the subject population. Robustness treatments varying feedback, parameters, and framing fail to eliminate the bias.

Continue reading at Wiley Online Library.