Decentralized matching markets experience high rates of instability due to information frictions. This paper explores the role of these frictions in one of the most unstable markets in the United States, the labor market for first-year school teachers. We develop and estimate a dynamic model of labor mobility that considers non-pecuniary information frictions directly. We find that teachers overestimate the value of hidden amenities and their own preferences for teaching. Improving access to information improves stability by 12% and reduces between-school switching by 18%, but reduces teacher labor supply by over 5%. Compared to each tested alternative, including targeted wage premiums at hard-to-staff schools, bonuses that incentivize retention, and lowered tenure requirements, information revelation improves match quality most.