Do Suspensions Affect Student Outcomes?

Student panel data, School District of Philadelphia

Philadelphia public school students, US.

Student fixed effects plus an instrumental-variable strategy that leverages a district-wide change in suspension policy. The fixed-effects design compares a student to himself before and after suspension, which removes stable heritable and family characteristics — the exact confound this archive treats as the default explanation for correlational discipline findings. Peer spillovers are estimated separately.

A suspension decreases math and reading achievement for the suspended student, and the result is robust to IV estimates exploiting the district-wide policy change. Suspensions matter much more for the student who experiences one than for peers. Exposure to peers' suspensions for more serious misconduct has very small negative spillovers onto peer achievement and does not change peer absences.

Low. Within-student fixed effects difference out time-invariant individual and family characteristics, which is the design the archive asks for before believing a discipline correlation.

Direction agrees with Bacher-Hicks et al. on long-run harm to the suspended student. The peer-spillover result sits against Carrell & Hoekstra and Figlio, who find substantial classroom externalities from disruptive peers; the reconciliation is that removing a disruptive student and suspending a student are not the same intervention.