The Evidence on Teaching

School Starting Age and Cognitive Development (with Bassok & Reardon on redshirting prevalence)

Dhuey, E., Figlio, D., Karbownik, K., & Roth, J. (2019); Bassok, D., & Reardon, S. F. (2013) · 2019

grade Bnatural-experimentindependentunreplicated
Sample
Florida population birth/school records, cohorts ~1994-2000; Bassok-Reardon: two nationally representative US cohorts
Population
US children around entry cutoffs, plus descriptive data on academic redshirting.
Design
Population-level RD through college and juvenile-justice outcomes. The Bassok-Reardon component is descriptive (grade D) and covers who actually redshirts.
Key findings
Redshirting is rarer than assumed (4-5.5%), is chosen by advantaged families largely independent of child ability, and is therefore not the remediation it is marketed as. Being older-for-grade does predict better long-run outcomes, but for the average family the decision is confounded with the SES that drives it.
Counterfactual
September vs August births (regression discontinuity on the cutoff).
Genetic confound
RD on exact birth date; but the redshirting-selection finding is descriptive and confounded by SES.
Replication notes
Establishes that redshirting is a high-SES positional strategy, not remediation for struggling children.
DOI / URL
10.1002/pam.22135
Cite
BibTeXRIS

Effects

OutcomeMetricValueMeasureTimingVsHorizonClass
Redshirting prevalence%4-5.5% of children — far lower than popular accountsmixedend-of-treatmentnonenot-applicableattainment
Who redshirtsselectiondriven by male/White/high-SES status; NO evidence low-ability 4-year-olds are redshirted moremixedend-of-treatmentnonenot-applicableattainment
September-vs-August test-score effectSD~0.2, remarkably stable across SES/race/birthweight subgroupsstandardizedgrade-3business-as-usualover-2yrdomain-skill
Older-for-grade long-run outcomesdirectionhigher college attainment, lower juvenile-incarceration riskmixedadultbusiness-as-usualadulthoodattainment

Cited by