Do Expectations Reflect Information Reliability? Evidence From Odds on Tennis Matches

Abstract

We examine whether people form expectations by placing a larger weight on more reliable signals. To test this notion, we analyze subjective probabilities inferred from odds offered on the outcomes of men's tennis matches, exploiting exogenous variation in information reliability related to whether a tennis match is played in a long or short format. The premise of our tests is that higher-skilled players, who are more likely to win any single point, will win more often in longer matches, where more points are generally played. This notion, which is confirmed in the data, suggests that skill-related signals are relatively more reliable in longer matches, and should thus affect odds in those matches more strongly. However, we find that the likelihood of higher-ranked players winning in longer matches is under-estimated. This result is robust to inferring expectations from odds offered by professional bookmakers, or odds achieved on a person-to-person betting exchange. The resulting biases in expectations are costly. Results from various robustness tests, including a laboratory experiment and a placebo test using women’s tennis data where all matches are played in the same length, support our conclusions. Overall, our analysis suggests that information reliability neglect influences expectations and outcomes in real-world markets.