A growing number of American states and school districts require students to meet basic performance standards in core academic subjects at key transition points in order to be promoted to the next grade. We exploit a discontinuity in the probability of third grade retention under a Florida test-based promotion policy to study the causal effect of retention on student outcomes over time. Regression discontinuity estimates indicate large short-term gains in achievement among retained students and a sharp reduction in the probability of retention in subsequent years. The achievement gains from retention fade out gradually over time, however, and are statistically insignificant after six years. Despite this fade out, our results suggest that previous evidence that early retention leads to adverse academic outcomes is misleading due to unobserved differences between retained and promoted students. They also imply that the educational and opportunity costs associated with retaining a student in the early grades are substantially less than a full year of per pupil spending and foregone earnings. (Contains 10 figures, 11 tables, and 19 footnotes.)