Research on School Discipline
Made by me with ggplot2 in R. Data source.
In the United States ...
African American students are much more likely to be suspended than White students. Being removed from your classroom reduces students' opportunity to learn, and suspension may be a "red flag" on college application forms. I decided to investigate the relationship between suspension and college attendance using a longitudinal data set. My research question was formed based on my observations in schools, interviews of students, and review of existing related research in education, psychology, sociology, and economics.
Much of the past research in education on the effects of suspension has not controlled for pre-existing differences in academic achievement between suspended and non-suspended students. This means that the negative relationships between suspension and educational outcomes that have found may be due to confounding – the pre-existing differences between suspended and non-suspended students rather than suspension itself.
Made by me with ggplot2 in R. Data source.
Indeed, in the data set I used, suspended students (who are depicted with blue dots) tend to have lower math and reading achievement and come from households with lower socioeconomic status than non-suspended students (who are depicted with red dots). Socioeconomic status includes parental occupation, parental education, and household income. These variables can affect both whether students are suspended and whether they go to college. For example, schools may be more reluctant to suspend a student if they think the parent might threaten to sue the school, and parents with more education and more income are more likely to be perceived by administrators as likely to do that. Additionally, students whose parents can pay for them to attend college are more likely to attend.
Made by me with ggplot2 in R. Data source.
Without accounting for these differences, I found that the college attendance rate of suspended students was 27 percentage points lower than that of the non-suspended students. I used multiple methods to control for the pre-existing differences between suspended and non-suspended students and make sure my findings were robust, included mixed effects models, fixed effects logistic regression, propensity score matching, genetic matching, and coarsened exact matching.
I accounted for the differences in academic achievement and socioeconomic status, along with differences in gender, race/ethnicity, sports participation, special education status, ever having been retained in a grade, and living with a single parent. After matching students based on these characteristics, I found that the college attendance rate of suspended students was about 10 percentage points less than that of non-suspended students. This is a large effect for education. These findings suggest that schools should consider alternative punishments, especially as suspensions are frequently given for non-dangerous behaviors like disrespect. Removing questions on college application forms about students' history of suspension may be another low-cost, efficient way of reducing the negative impact of suspension and an opportunity to reduce inequality in higher education.