Social Emotional Learning, Student Attendance, and Chronic Absenteeism in Pre- and Post-Pandemic Periods

We develop and implement a quasi-experimental panel data model to address the relationships between social emotional learning competencies (SEL) and annual student attendance in the pre- and post-pandemic periods. Although panel data models tend to focus on changes over time in outcomes and predictors, we develop a model that decomposes the effects of SEL into stable and transitory components. The model addresses two types of error in SEL variables: standard error due to unreliability in the measured variables and error in student average SEL as a measure of the stable component of SEL. We find that the stable measures of self-efficacy and self-management are strong positive predictors of student attendance. We construct a composite SEL measure that optimally combines four SEL measures. Effect size estimates of the stable SEL composite increase from 0.138 to 0.223 between the pre- and post-pandemic periods. Effect size estimates are even higher for students with predicted attendance level at the 10% quantile level: 0.227 and 0.405 in the two periods, respectively. Although the effect sizes of SEL in both the pre- and post-pandemic periods are relatively large they explain only 5 to 10 percent of the decline in attendance. Our results establish the predictive validity of SEL measures with respect to student attendance. The estimated effects are strong enough to suggest that the causal effects of SEL may be large, especially for students with low expected attendance.