Association for Public Policy Analysis & Management

Schools employ several strategies to foster a stable and safe learning environment, including dedicating additional resources towards hiring staff and counseling services. A school’s discipline policy can also be used to achieve this aim. Removing misbehaving students from the school environment – either temporarily through out-of-school suspension (OSS) or permanently through expulsion – can allow teachers to focus on academic instruction rather than behavior management and minimize the classroom disruptions. Alternatively, while away from school, disciplined students receive no academic instruction, lose access to school resources, and are often in unsupervised environments; all factors that unlikely conducive to learning. Indeed, previous research has demonstrated that OSS is associated with lower levels of academic achievement, continued misbehavior, and increased likelihood of both dropout and crime.

Unfortunately, data limitations and the endogenous nature of the applied punishment have made it challenging to determine the casual effect of OSS. In this paper, we attempt to estimate the effect of OSS suspensions on student outcomes by making use of a historical proclivity for OSS in New Orleans. Prior to Hurricane Katrina, OSS accounted for over 90 percent of suspensions; and, while this rate has dropped slightly following Katrina, OSS still constitute over 80 percent of suspensions. In contrast, OSS account for only 40 percent of suspensions across the state of Louisiana. Our analysis is premised on the assumption that this long-standing tendency towards assigning OSS indicates the practice is less reflective of individual discretion in New Orleans and more reflective of standard norms.

We estimate the effects of OSS in New Orleans relative to in-school suspensions (ISS) on student achievement, future misbehavior, the likelihood of dropping out of school, and the likelihood of graduating within four years of 9^th grade entry using several quasi-experimental designs based on propensity score matching techniques. For example, we control for different school discipline policies by matching students within schools on their likelihood to receive an OSS. Alternatively, to minimize the potential for individual administrator discretion in the OSS versus ISS decision, we match similar students across schools. In all models, matched comparison groups are constructed by matching within school year to capture differences in policies over time. Moreover, we control for lingering cross-group differences in school characteristics in a second stage estimation. Finally, in addition to estimating average treatment effects, we explore if these effects are differentiated by family income, ethnicity, and gender. While no single analysis is likely to produce a bias-free estimation of the relative effects of OSS on student outcomes, taken together, our results will help to hone in on the true effect.

While research has demonstrated that OSS is negatively related to student achievement, behaviors, and later education outcomes, the endogenous nature of the decision to use OSS or ISS makes it hard to determine the casual effect of this intervention. This paper, by attempting to estimate casual effects of OSS suspensions in New Orleans, a city with a historical proclivity towards OSS, can aid education policymakers when considering the future school suspension policies.