Job market paper
Exclusionary school discipline techniques, such as out-of-school suspension, are often criticized for their inability to improve students' behavior, their adverse effects on students' achievement outcomes and their disproportionate use on minority students. Using comprehensive data on all North Carolina public school students, I find that harsher disciplinary rules (measured by higher out-of-school suspension likelihood) significantly deter students from committing first offenses, but that they are less effective (or ineffective) for repeat offenses. I also find that their adverse effects on offending students' achievement outcomes, such as end-of-grade test scores and high school dropout probability, are much smaller than the effects documented in the existing literature. In addition, I find that harsher disciplinary rules could significantly improve the academic achievement of middle school students with no offense record. The method I use to identify causal effects combines the instrumental variable method and a machine learning cluster method (k-means) to carefully address endogeneity and selection issues in a big data context. These findings suggest that current policy reform of exclusionary school discipline should carefully balance its benefits and costs for different student populations.