The study was designed to examine decision-making under uncertainty in the context of the trivia game show “Cash Cab.” Kelley and Lemke explored the influences of pre-game (e.g., gender, race, age, group size) and in-game( e.g., winnings, accuracy, confidence) factors on the decision to accept or reject a risky double-or-nothing proposition.
With regard to pre-game factors, their study supported the standard findings of the decision literature—namely, men, groups, and younger adults were more likely to gamble than women, individuals, and older adults, respectively.
After controlling for these pre-game effects, they considered their question of central interest: do contestants use aspects of their game play to update their subjective probabilities of success when considering the decision to gamble?
Surprisingly, they found that in-game performance, such as the number correct, did not influence the decision to gamble.
“Although these results seem counterintuitive on the surface, they make more sense when you consider the contestant’s confidence,” Kelley said.
Specifically, they reported that the probability of accepting the gamble increased significantly as the number of highly confident and correct (HC) answers increased, as well as when contestants had long streaks of HCs, ended the game with a long streak of HCs, and even when just the final question was answered correctly with high confidence.
Given that eventual gamblers and non-gamblers were indistinguishable on measures of initial ability and confidence, they argued that contestants paid particular attention to their in-game confidence—and perhaps how well their confidence is calibrated—and used this information to update their subjective probability of success when deciding whether to gamble.
The next step of the research will be to try to replicate the results in a laboratory setting.
“Economists tend to look for natural experiments, such as decisions made on a game show like “Cash Cab,” to investigate social phenomenon,” Lemke said. “With these results in hand, we are now in a better position to design a controlled laboratory experiment so that we can better determine how inherent risk tolerance, self-confidence, and knowledge affect decision-making in an uncertain environment.”