Abstract

Some of life’s most important decisions require assessing our ability relative to others. Findings in lab experiments reveal that individuals hold biased beliefs about their relative ability. However, there is little evidence on the relevance of beliefs outside of the lab. To fill this gap, I conduct a field experiment with students who are preparing to take a college admission exam and take weekly practice tests in a test preparation center. Overall, I find that over 50% of students hold biased beliefs, either by overestimating or underestimating their relative performance in math and reading. I observe substantial biases in information processing after students receive a signal of whether they are in the top or bottom half of the distribution: Students update about 67% of what a Bayesian would and update more when receiving a "top" rather than a "bottom" signal. Moreover, the data show evidence of "confirmatory bias", which is the tendency to update more closely to the Bayesian benchmark when receiving a signal confirming one's prior and to ignore information when receiving a disconfirming signal. When receiving full feedback about their relative performance, treated students are 10 pp less likely to overestimate their math performance in the next practice test if they receive a “bottom” signal and are 10 pp less likely to underestimate if they receive a “top” signal. Furthermore, when receiving a “bottom” signal, individuals study less for the next practice test. Interestingly, these effects go in opposite directions by gender: women who get to know their low performance tend to study more while men tend to study less.