This paper investigates the role played by self-confidence in college applications. Using incentivized experiments, we measure the self-confidence of more than 2,000 students applying to colleges in France. The best female students and students from low socioeconomic status (low-SES) significantly underestimate their rank in the grade distribution compared to male and high-SES students. By matching our survey data with administrative data on real college applications and admissions, we show that miscalibrated confidence affects college choice controlling for grades. We then estimate the impact of a randomized intervention that corrects students' under-and overconfidence by informing them of their real rank in the grade distribution. The intervention fully offsets the impact of under- and overconfidence for college applications. Providing feedback also makes the best students, who were initially underconfident, apply to more ambitious programs with stronger effects for female and low-SES students. Among top students, our intervention closes 72% of the gender gap in admissions to elite programs, and 95% of the social gap. We conclude that confidence is an important behavioral consideration for the design of college admission markets.