Abstract: We are an interdisciplinary team of ecologists and education researchers using techniques from ecology to understand patterns of attrition and retention of high school science and math teachers with an eye towards equitable outcomes for teachers and students. Teacher attrition is a major problem: it disrupts student learning, is costly for schools, and is especially prevalent among STEM teachers. Much previous research focuses on personal factors and reveals racial disparities in retention. Here we asked how school environment and individual factors (demographics: age and gender) interact to influence attrition. We also tested whether department policies, such as the presence of senior teachers as content coaches, can improve retention.
We used non-metric multidimensional scaling (NMDS) to combine twelve school-wide metrics of economic need, discipline, spending, student performance, and course offerings into a single school environment variable. Schools with positive NMDS scores were those with high graduation rates, student enrollments, and numbers of Advance Placement courses. Schools with negative NMDS scores had high levels of chronic absenteeism, disciplinary actions, and eligibility for Free or reduced-price lunch (FRPL). We evaluated ten years (2009-2018) of administrative data for public high schools in Connecticut. Using survival analysis (mixed-effects Cox regressions) separately for math and science teachers, we tested how race, gender, school environment (NMDS), and coaches influenced attrition.
Black math teachers had a higher attrition rate than White math teachers, but there was no effect of gender or its interaction with race. Among science teachers, all races had similar attrition rates. Overall, female science teachers had slightly lower attrition rates than males. However, amongst Asian and Latino science teachers, females had higher attrition rates. Higher NMDS scores were significantly correlated with lower attrition rates. The ameliorative effect of content coaches was marginally significant for math, but not science teachers. Our predictors interacted in complex ways. For example, the disadvantage of Black male math teachers was worse at high NMDS, but content coaches helped counteract that influence. Conversely, Asian science teachers had lower attrition at high NMDS.
An ecological approach productively identified a variety of factors contributing to STEM teacher attrition. School environments correlated to teacher attrition, but the influence was not felt equally amongst teachers of different backgrounds. Neither did content coaches improve retention equally for teachers of all backgrounds. This work improves our understanding of the hierarchical nature of the problem of teacher retention and the limitations of proposed solutions.
