Looking Beyond Socioeconomic Status: Using Quantitative Measures of Student Misconduct to Identify "At-Risk" Schools

Abstract

The research project used regression analysis to study the relationship between students' misconduct and their schools' corresponding level of academic achievement at the middle and high school levels. This was a non-experimental, ex post facto study conducted using data from a large, Southeastern school district collected over three academic years, 2001 to 2004. The operational variables for this analysis were defined as: 1) socioeconomic status (SES) as measured by the schools' free or reduced lunch percentages, 2) student misconduct (DAR) as measured by the number of suspensions per 100 students in a given school year, and 3) academic achievement (ABC-PC) as measured by the schools' overall academic performance composites calculated via North Carolina's accountability formulas. A data transformation was conducted on the DAR variable by calculating its logarithm (logDAR) to improve the normality of the variable's distribution. Regression analyses were run using the SAS 9.1 analytic software package to determine the nature of the relationships between: 1) SES and ABC-PC, 2) logDAR and ABC-PC, and 3) logDAR and ABC-PC while controlling for SES. It was determined that: 1) SES has a significant relationship to ABC-PC at the middle and high school levels, 2) logDAR has a significant relationship to ABC-PC at the middle and high school levels, 3) logDAR does not have a significant relationship to ABC-PC at the middle school level when SES is entered into the regression equation as a control variable, and 4) logDAR does have a significant relationship to ABC-PC at the high school level, even when SES is entered into the regression equation as a control variable. Regression models using both SES and logDAR as independent variables had greater explanatory power than regression models using either SES or logDAR in isolation.

It was concluded that quantitative measures of student misconduct, such as the logDAR covariate, can be useful in identifying schools at the greatest risk of academic failure — particularly at the high school level.