We use a regression discontinuity strategy to produce causal estimates for the effect of remediation under Florida’s test-based promotion policy on multiple outcomes for up to five years after the intervention. Students subjected to the policy were retained in the third grade, were required to be assigned to a high-quality teacher during the retained year, and were required to attend summer school. Exposure to these interventions has a statistically significant and substantial positive effect on student achievement in math, reading, and science in the years immediately following the treatment. But the effect of the treatment dissipates over time. Nonetheless, we find that the effect of remediation under the policy on academic achievement is statistically significant and of a meaningful magnitude several years after the student is exposed to the intervention. Though we cannot completely separate the differential effects of the treatments attached to the policy, we provide some evidence that assignment to a higher-quality teacher in the retained year is not the primary driver of the policy’s effect.

In 2002, Florida adopted a test-based promotion policy in the third grade in an attempt to end social promotion. Similar policies are currently operating in Texas, New York City, and Chicago and affect at least 17 percent of public school students nationwide. Using individual-level data on the universe of public school students in Florida, we analyze the impact of grade retention on student proficiency in reading one and two years after the retention decision. We use an instrumental variable (IV) approach made available by the relatively objective nature of Florida's policy. Our findings suggest that retained students slightly outperformed socially promoted students in reading in the first year after retention, and these gains increased substantially in the second year. Results were robust across two distinct IV comparisons: an across-year approach comparing students who were essentially separated by the year in which they happened to have been born, and a regression discontinuity design.