This paper investigates how college students update their future earnings beliefs using a unique “information” experiment: We provide college students true information about the population distribution of earnings, and observe how this information causes them to update their future earnings beliefs. We show that college students are substantially misinformed about population earnings, logically revise their self-earnings beliefs, and have larger revisions when the information is more specific and is “good” news. We classify the updating behaviors observed and find that the majority of students are non-Bayesian updaters. While the average welfare gains from our information provision are positive, we show that counterfactually imposing Bayesian processing of information vastly overestimates the gains from the intervention. Finally, we present evidence that our intervention has long-lasting effects on students’ earnings beliefs.