Testing for Causal Clustering in K-12 Student Discipline
Sean Mulherin
MASDS, 2025
SCHOENBERG, FREDERIC R.
The degree to which causal contagion explains the event of student misconduct is investigated. A test introduced by Kresin (2023) and McGovern (2024) is applied, wherein likelihood-ratio tests are performed using information gain statistics to compare the fit of a Neyman-Scott model to that of a Hawkes model. Data records the number of disciplinary actions for K-12 students grouped by day, school, and school year spanning from 2016 to 2023. Evidence of causal clustering appears to vary across grade levels and school years. Out of the twelve different school/year combinations tested, seven exhibited statistically significant evidence of causal triggering. Interestingly, both frequency and magnitude were found to hold strong governance over the conclusion of hypothesis tests.
2025

