Are Students Harmed by Being Held Back in Elementary School?

Zhong (2024) asks whether third-grade test-based retention improves or harms students’ long-term outcomes. He examines administrative data from Texas public schools linked to college enrollment records and state earnings data, following students from third grade into their mid-20s. Using a regression discontinuity design around the reading test promotion cutoff, he finds that retention briefly raises test scores but increases absenteeism, violent behavior, and juvenile crime. Retained students are 14% less likely to graduate from high school and earn about $3,500 less per year at age 26, a roughly 19% reduction compared with similar peers who were promoted.

Why This Article Was Selected for The Policy Scientist

This article addresses a central policy question with implications far beyond third grade: whether accountability-driven retention policies meaningfully improve life chances for low-achieving students or unintentionally shape long-run inequality. That question is especially timely as states reconsider early literacy mandates and reintroduce test-based promotion rules after pandemic learning losses. Zhong’s work is an important contribution because it extends a large retention literature—long focused on short-term test scores—into early adulthood outcomes that matter for economic mobility. The administrative data are unusually strong, linking schooling, postsecondary enrollment, and earnings over two decades. The regression discontinuity design provides credible causal estimates, aligning well with best practices in modern policy analysis. While drawn from Texas, the institutional features closely resemble policies used elsewhere, supporting cautious generalization.

Full Citation and Link to Article

Zhong, J. (2025). Early grade retention harms adult earnings. American Economic Journal: Applied Economics. https://doi.org/10.1257/app.20230121.

Central Research Question

This article asks whether test-based third-grade retention improves or harms students’ long-run outcomes. Specifically, it examines whether retaining students who narrowly fail a third-grade reading exam affects their educational attainment, behavioral outcomes, and labor market performance into early adulthood. The central focus is not limited to short-term academic remediation but instead evaluates whether early grade retention alters life trajectories in measurable and persistent ways. By linking early schooling decisions to earnings in the mid-20s, the study addresses a major gap in education policy research: whether policies designed to raise academic standards at an early age translate into improved economic outcomes or instead impose lasting costs on marginal students.

Previous Literature

The literature on grade retention is large but divided. Early work and policy debates often justified retention as a corrective tool, arguing that additional instructional time could help struggling students master foundational skills. Several influential studies on third-grade retention policies—particularly in Florida—found short-term gains in reading and math scores following retention, especially when paired with intensive remediation and instructional supports. However, these gains often faded over time and did not consistently translate into higher graduation rates or college enrollment.

In contrast, research on retention in later grades has produced more uniformly negative findings. Studies examining eighth-grade or high school retention show increased dropout rates, higher involvement in crime, and worse labor market outcomes. Yet until recently, evidence linking early grade retention to adult earnings was largely absent. This article builds on widely cited work by Jacob and Lefgren, Schwerdt et al., Figlio and Özek, and others by extending the outcome window well beyond schooling. It contributes to the literature by isolating the effect of retention itself—rather than bundled remediation programs—and by providing causal evidence on long-term economic outcomes, an area where prior studies were constrained by data limitations.

Data

The study uses comprehensive administrative data from Texas that track students from elementary school into adulthood. The core education data come from the Texas Education Agency and include standardized test scores, grade progression, absenteeism, disciplinary incidents, and high school completion. These records are linked to postsecondary enrollment data from both Texas institutions and the National Student Clearinghouse, allowing observation of college enrollment and degree completion nationwide.

Crucially, the analysis also links students to earnings records from the Texas Workforce Commission, which capture annual wages reported by employers within the state. This linkage allows the author to observe labor market outcomes from ages 23 to 26. The data span cohorts exposed to Texas’s test-based third-grade reading retention policy between the early 2000s and the policy’s discontinuation. The resulting dataset is unusually rich, longitudinal, and well-suited for evaluating long-run outcomes of early education policies, with minimal attrition due to the administrative nature of the records.

Methods

The study employs a fuzzy regression discontinuity design to estimate the causal effects of third-grade retention. Under Texas policy, students were required to pass a reading exam cutoff on their third attempt to be promoted to fourth grade. While exemptions were possible through appeals, the probability of retention increased sharply for students scoring just below the cutoff.

By comparing students whose scores fell narrowly above and below this threshold, the design exploits quasi-random variation in retention status among otherwise similar students. The fuzzy design accounts for imperfect compliance with the rule by using failure to meet the cutoff as an instrument for actual retention. Local linear regressions with optimal bandwidth selection are used to estimate treatment effects, following best practices in modern causal inference.

The study conducts extensive balance tests to verify that observable characteristics are smooth around the cutoff and that there is no manipulation of test scores. The strong first-stage relationship between failing the test and being retained supports the validity of the design. This approach aligns closely with the author’s stated preference for causal inference methods over correlational multivariate regression.

Findings/Size Effects

The results show that third-grade retention produces short-term academic gains but substantial negative long-term effects. Retained students experience temporary improvements in reading and math test scores, particularly in the first year after retention. These gains fade within a few years and disappear by middle school, consistent with prior findings in the literature.

In contrast, behavioral outcomes worsen. Retention increases absenteeism in the years immediately following third grade and significantly raises the incidence of violent behavior and juvenile crime over subsequent years. These effects are sizable, with violent incidents increasing by more than 60 percent relative to students who narrowly passed the exam.

Educational attainment also declines. Retained students are about 14 percent less likely to graduate from high school and are substantially less likely to graduate on time. While retention slightly increases community college enrollment, it reduces enrollment in public four-year universities, and overall college completion effects are weak and imprecisely estimated.

The most consequential findings emerge in the labor market. By age 26, students retained in third grade earn approximately $3,500 less per year than comparable peers, a reduction of roughly 19 percent relative to the mean earnings of students just above the cutoff. Average earnings across ages 23 to 26 show similar declines. Retention also reduces the probability of having any positive earnings, suggesting weaker labor force attachment, and lowers the likelihood of reaching higher earnings percentiles. These results indicate that early retention affects not only wages but also employment stability and access to higher-paying jobs.

Conclusion

This article provides some of the strongest evidence to date that early grade retention carries long-term costs that outweigh its short-term academic benefits. By linking third-grade retention to adult earnings using a credible causal design and high-quality administrative data, the study moves the policy debate beyond test scores and graduation rates to outcomes directly tied to economic mobility.

The findings suggest that retention functions less as an effective remediation strategy and more as a structural disruption that alters students’ educational and labor market trajectories. While the analysis is grounded in Texas, the institutional features of the policy closely resemble early literacy mandates used in many states, supporting cautious generalization. Overall, the study represents a significant contribution to the education policy literature by demonstrating that early academic interventions must be evaluated not only by short-run achievement gains but by their long-run consequences for human capital and earnings.