Early Grade Retention Harms Adult Earnings

roughly 17,400-22,070 students in the bandwidth; Texas third-grade cohorts 2002-03 through 2004-05, linked to state earnings, National Student Clearinghouse and juvenile records

Texas third graders at the Student Success Initiative TAKS reading cutoff on their third and final attempt

Forthcoming in American Economic Journal: Applied Economics; no DOI yet. Full text read from the author's job-market paper. Fuzzy RD on Texas's Student Success Initiative, where third graders got THREE attempts to pass the TAKS reading test; the running variable is the third-attempt score minus the annually set cutoff, local linear with a triangular kernel and a bandwidth of about 8 points. Scoring below the cutoff raises retention probability by 35.5 percentage points; McCrary density test p = 0.4382. THE REASON THIS SOURCE MATTERS IS THE CONTRAST WITH FLORIDA, and the author names it: the Florida studies 'estimate the combined effects of grade repetition, a summer program, intensive 90-minute daily reading instruction in the subsequent year', whereas Texas mandates no instructional support after retention, gives three attempts rather than one, and retained roughly 2% of third graders against Florida's 12%. Two consequences follow: the Texas estimate is closer to retention alone, and the Texas retainees are a much more negatively selected group, so stigma is plausibly larger. The year-by-year test-score trajectory appears only in figures and was not obtained numerically.

This is the study that stops the RD literature from being read as a rehabilitation of retention. Achievement follows the familiar arc - 'while third-grade retention initially boosts reading scores, this advantage diminishes by the fifth year', and same-grade gains dissipate by eighth grade. But the long-run outcomes are bad. Ever graduating from high school falls 9.1 percentage points (SE 0.043), a 14.17% decline against a 64.2% above-cutoff mean. Annual earnings at age 26 fall $3,477, a 19% reduction against a mean of $18,557, with estimates of -$4,324, -$2,921, -$3,278 and -$3,477 across ages 23 to 26 and a 5-point fall in the probability of positive wages; the losses are larger at higher earnings percentiles. As in Florida, on-time graduation collapses (-36.6 points) while graduating one year late rises (+27.7) and two, three and four years late are all about zero - retained students catch up and graduate exactly one year behind. Ever dropping out is null (+0.021, SE 0.033) and college enrolment is null on every margin, so the earnings damage is not routed through college. Absenteeism rises 1.25 days (+23%) in year one and 1.38 days (+25%) in year two.

Low at the cutoff. Assignment is a state test threshold on the third attempt, and the density test shows no manipulation. Earnings and criminal outcomes come from administrative registers rather than survey report. The usual local-average caveat applies with force here because only about 2% of Texas third graders were retained, so the compliers are a small and unusual group.

Directly contradicts Schwerdt, West and Winters's precisely estimated zeros on Florida attainment and college-going, on the same intervention at the same grade. The best available reconciliation is the policy bundle - Texas retains without mandated remediation, Florida retains with it - which is the same distinction Goos and colleagues quantify meta-analytically as 'mere rehearsal' g = -0.10 versus 'retention plus' g = +0.20. It is directionally consistent with Louisiana's dropout and crime results and with Uruguay's attainment losses. It is the only study in the literature reaching adult earnings, so the finding is unreplicated on its headline outcome.