Association for Public Policy Analysis & Management

Grade retention, the practice of requiring a student to remain in the same grade the following year (Jackson, 1975), disproportionally affects socio-demographically at-risk students facing academic challenges (Warren et al., 2014). Each year, the U.S. spends $20 billion on retention and two million children are retained (Eide & Goldhaber, 2005). However, the impact of retention remains controversial and not well-understood. Early studies characterize retention as a detrimental intervention (Holmes, 1989; Jimerson, 2001), but recent studies have found mixed effects on academic outcomes (Hong & Raudenbush, 2006; Hong & Yu, 2008; Im et al., 2013). Additional research using methodologically rigorous approaches with national data is needed to clarify the effects of grade retention on academic outcomes in both the short and long term.

The current study uses propensity score methods and national data (Early Childhood Longitudinal Study, Kindergarten Class 1998-1999) to answer two research questions about retention occurring in first or second grade. First, how does retention affect math and reading test scores over time? Second, acknowledging the importance of middle grade outcomes on later school engagement, what is the effect of retention on multidimensional academic indicators six or seven years later? The overall goal of the study is to determine whether retention is an effective intervention for students struggling to meet academic standards in early elementary school.

We conducted a nearest neighbor without replacement match to create a logical counterfactual group. The main analytic sample consists of 287 students retained once in either the first or second grade (treatment group) and 287 matched control group students, similar on observed characteristics but were promoted. Propensity score models were fit to obtain balance between the retained sample and the comparison group of promoted students on 43 covariates. The average treatment effect on the treated was calculated using probability weights and covariates in a regression-adjusted matched estimate. To address potential bias and determine consistency of findings, another propensity score approach, caliper matching with replacement, was used as a robustness check.

Key results are as follows. In spring 2002 (i.e., when promoted students were in third grade and retained students were one grade lower), we found that retention had a statistically significant negative impact on both reading and math test scores (reading: b = -.09, p < .001; math: b = -.06, p < .01). Over time, the negative effects on math test scores waned, but retention had a persistent, negative effect on reading (spring 2004: b = -.04, p < .05). In spring 2007, when promoted students were in eighth grade and retained students were in a lower grade, findings show a statistically significant negative effect on reading achievement (b = -.06, p < .01) and null effects on other academic measures (math achievement, student- and teacher-reported academic competence). The robustness check indicated that our findings were consistent. Thus students would have fared similarly on most outcomes and even better on reading achievement, had they been promoted. Implications for education policy about the costs and benefits associated with grade retention will be discussed.