Turnover in the teacher workforce imposes significant costs to schools, both in terms of student achievement and the time and expense required to recruit and train new staff. This paper examines the potential for structured ratings of teacher applicants, solicited from their professional references, to inform hiring decisions through the selection of teachers who are less likely to turn over. Specifically, we analyze the predictive validity of reference ratings with respect to retention outcomes among subsequently employed applicants. We find that a summative reference ratings measure is modestly predictive of retention in a teacher's school, with a one-standard deviation change associated with a 3.2 percentage point increase in the probability of school retention. When we account for rater fixed effects, we find substantially stronger relationships between reference ratings and retention, with a one-standard deviation change in our summative ratings measure associated with an increase in the probability of school retention of 8.5 percentage points. These findings suggest that raters themselves are a large source of variation in the distribution of reference ratings. So, while we find predictive validity of professional ratings, their potential to inform good hiring decisions depends on, among other things, the ability of hiring managers to account for rater variation when interpreting references' assessments of applicants.