The Effect of Grade Retention on High School Completion

8,573 (grade 6) and 5,402 (grade 8), cohorts 1997-1999, dropout measured to autumn 2005

Chicago Public Schools sixth and eighth graders near the ITBS promotion cutoffs

American Economic Journal: Applied Economics 1(3), 33-58; full text read from NBER working paper 13514 (October 2007). Same Chicago fuzzy RD as the 2004 paper, extended to high school completion eight years out. The value of the design here is that it tests differential attrition directly - effects on changing district or transferring to private school are all null (grade 6: -0.03, SE 0.03; grade 8: -0.02, SE 0.02) - so the dropout result is not a measurement artefact of students disappearing from the data.

The grade at which a child is held back matters far more than whether. Retention in EIGHTH grade raises the probability of dropping out by 8 percentage points (SE 0.03), 'or 14 percent' against a control mean of 0.57. Retention in SIXTH grade does not: the estimate is -0.04 (SE 0.04) with a confidence interval from -0.12 to +0.04, and the authors state they 'can reject that grade retention increases the drop out rate by more than 8 percent'. Note that OLS gets both wrong in the same direction, giving +0.02 for grade 6 and +0.04 for grade 8. THE MECHANISM IS TIME, NOT STIGMA: sixth-grade retention cuts the probability of eighth-grade retention by 9.0 percentage points (SE 0.029) and sixth-grade retainees 'began high school only about one fifth of a year after their promoted peers' (+0.21, SE 0.05), while eighth-grade retainees started over 0.4 years late. 'Because students retained in earlier grades have more opportunities to catch up with their peers, retention in these grades has less severe long-term impacts.' Eighth-grade harm is concentrated in girls (+0.16, SE 0.05, versus +0.01 for boys, equality rejected at p = 0.02), Black students (+0.11) and free-lunch students (+0.08).

Low. Identification is off a test-score threshold applied to the whole district, and the outcome is an administrative record rather than a test, so neither arm is differentially incentivised. The attrition checks rule out the main remaining threat.

The late-retention harm replicates strongly. New York City finds retention in grades 7 and 8 raises dropout by about 10.8 percentage points, more than 30% of the control complier mean, and cuts credits earned by 7.8, while grades 3 and 5 show nulls. Louisiana finds eighth-grade retention raises dropout by 7.2 percentage points and adult violent-crime conviction by 58%. Uruguay finds grade failure in grades 7-9 costs 0.76-0.91 years of schooling. The early-grade null is the contested half: Louisiana finds grade-4 retention raises dropout 5.1 points, and Texas finds third-grade retention cuts ever graduating by 9.1 points.