Adapt or Abandon: Demographic Shocks and Principal Turnover

ABSTRACT

Given that principals stand at the precipice between policy, the student body, and the community, this article examines the relationship between demographic shocks and principal turnover. In observing more than seven thousand schools in Texas over a seventeen-year period, hazard model results demonstrate that short-term demographic changes are significantly associated with increased principal turnover risk, particularly for changes in the proportion of Students of Color. Results further imply that White and suburban principals were the most sensitive to these changes, suggesting the need for additional training and support to ensure stable leadership for schools undergoing shocks in student composition.

Notes

1. To maintain consistency between the statistical sample and discussion, demographic terms of African American, Hispanic, White, Limited English Proficient, and Economically Disadvantaged are adopted and used from the glossary of the Texas Education Agency (Texas Education Texas Education Agency, 2017a). Racial category definitions used by the TEA are consistent with those used by the U.S. Census bureau.

2. As one principal in a school undergoing rapid demographic change noted: “I don’t know that I could be a good mentor. Maybe students need to see a person of their own ethnic background that’s made it..I don’t know if I can relate. Certainly I have good relations with students but don’t know if it is the same kind of relationship” (Evans, 2007, p. 176).

3. Categorizations of American Indian, Asian, and Pacific Islander differed during the sample period and as such are not included directly (Texas Education Texas Education Agency, 2017a).

4. A few additional details of the data set require clarification. First, principals who exited the profession after reaching the “rule of 80 (years of employment + age) were censored from turnover calculations, given retirements can be considered less relevant policy wise. Second, school closures were censored from the dataset in their last year so as not to be counted as a turnover event. Third, principals with temporary leaves (left one year and returned the next), and those with < 50% FTE status were censored given these represent non-typical turnover situations. To ensure that the censoring was nonsystematic, we ran Little’s Missing Completely at Random (MCAR) test, to find the censoring to be nonsystematic (MCAR p < 0.05) (Li, 2013).

5. Prior literature emphasizes the differences in types of turnover, including principal switches and exits (Farley-Ripple, Solano et al., 2012). However, given that this exploratory analysis emphasizes on how shocks may initiate exit scripts of any kind, we chose to focus on turnover in general.

6. While school accountability ratings are partially based on student achievement (alternative assessments are additionally included), they also incorporate measures of school progress and gap reduction (Texas Education Agency, 2018). While achievement and accountability are indeed related, they do represent distinct metrics. As a result, we included both in the models. They did not display meaningful levels of variance inflation (VIF < 2).

7. Following prior literature, age and experience were accompanied by an additional squared regressor to account for their curvilinear relationship with turnover (Gates et al., 2003; Papa et al., 2002).

8. It should be noted that while the proportion of variance explained (Pseudo R²) is relatively low across models (between 0.027–0.047) this similar to other estimates of principal turnover from administrative data (e.g., Grissom & Bartanen, 2018), and relates to research noting that turnover is not necessarily voluntary (Farley-Ripple, Raffel et al., 2012) and can be based on subjective elements as well (Boyce & Bowers, 2016).

9. Subsequently, also we tested if directional effects are significantly different from one another by using a binary indicator for positive versus negative change. We collapsed change direction to a binary indicator to assess whether the statistical effect of demographic increases was significantly different from decreases in relation to turnover. Positive changes in % Students of Color, % African American, and % White significantly increased turnover risk as compared to negative changes, while positive changes in % Economically Disadvantaged and % Enrollment significantly reduced turnover risk as compared to negative changes. Results available in Appendix 1.