Unnatural experiments: the case of television game shows

Abstract

Economists often use television game shows as ‘natural experiments’ to test theories on risk aversion. This article demonstrates that some of these natural experiments may not be so natural after all. Using data from Jeopardy!, we show that contestants are chosen to maintain a gender balance, with the result that men outperform women across several dimensions. This selection bias suggests that caution should be exercised in generalizing findings based on these shows to the population as a whole.

Notes

¹Technically, the Jeopardy! host, Alex Trebek, reads an answer and the contestant tenders the question. For ease of writing, and to be consistent with most game shows, we will refer to the host's role as providing the question and the contestants' as providing the answer. Questions are occasionally in the form of an audio or video clip.