Game Shows and Economic Behavior: Risk-Taking on "Card Sharks"

Abstract

This paper studies individual risk-taking behavior with a unique data set: risky decisions of contestants during the threeyear history of the television game show, Card Sharks. In the bonus round of the game, individuals choose how much to wager on positive expected-value gambles with known and varying probabilities, and large possible payouts. Since no special skills or knowledge are involved, empirical analysis is not complicated by a contestant's superior information about her own abilities. Stakes can reach $16,000 and average $3,200; few experimental budgets permit even one observation of this magnitude. First, I simply estimate how risk-averse contestants behave. Under the assumption of expected-utility maximization, I estimate a lower bound for the coefficient of absolute risk aversion of 0.00031, which when multiplied by per capita U. S. personal income gives an estimate of relative risk aversion of 4.8. This estimate is higher than most econometric estimates of risk aversion. Contestants, however, do not act in a manner consistent with expected-utility maximization, so this number should not be given a structural interpretation. Rather, it is simply a useful way to summarize average willingness to bear risk of individuals on the show. Second, I explore how individual behavior is inconsistent with expected-utility theory. The amount a contestant wishes to wager depends critically on the amount of wealth she has accumulated in the bonus round and therefore, on the size of the stakes. This is at odds with expected-utility theory since the size of the stakes, although far from insignificant, is small relative to a person's wealth. Contestants should then have close to constant absolute risk aversion with respect to accumulated winnings in the game. A more striking result deals with contestants who had previ-