Panel Paper: School Organizational Contexts, Teacher Turnover, and Student Achievement: Evidence from Panel Data

Friday, November 7, 2014 : 9:10 AM
Enchantment Ballroom C (Hyatt)

*Names in bold indicate Presenter

Matthew Kraft1, William Marinell2 and Darrick Yee2, (1)Brown University, (2)Harvard University
There is increasing recognition among experts that school contexts play a central role in shaping teachers’ career decisions and students’ learning opportunities.  Expansions of teacher surveys have created new opportunities to measure school context dimensions and examine the relationship of these measures to student achievement and teacher turnover.  Researchers have documented the cross-sectional relationship among contexts, achievement and turnover, but have not produced evidence that improving contextual features in a school raises student achievement or lowers teacher turnover.  Furthermore, many studies employ school context measures that are collinear, making it difficult to advise policymakers about which school context features are most important.

We extend this literature by leveraging panel data to examine the relationship between contexts, achievement and turnover among 200-plus middle schools in a large urban district.  We focus on traditional grade 6-8 middle schools because they are among the most challenging contexts for students and teachers to experience success.  Our data contain teacher survey responses from 2008 to 2012 on a School Climate Survey with annual response rates of over 60%.  We identify four distinct school climate dimensions captured by 33 survey items using principal component analysis: Leadership & Professional Development (Leadership), Academic Expectations (Expectations), Teacher Collegiality & Collaboration (Collegiality) and School Safety & Student Behavior (Safety).  We construct measures of each dimension by conducting factor analysis with an orthogonal rotation and averaging teachers’ factor scores to the school-year level.

We examine two school-year-level outcomes, median student growth percentile (SGP) and teacher turnover percentage using OLS and GLM regression models, respectively.  Our models include a vector of average student and teacher characteristics, as well as fixed effects for years and schools. By including school fixed effects, we are able to remove all time-invariant school characteristics that could account for the observed relationship between context, achievement, and turnover. We also fit models that include all four school context measures simultaneously to examine the relative importance of each dimension.

We find strong evidence that, within schools, improvements in school context measures over time are associated with increased achievement and decreased teacher turnover.  Safety and Expectations are both significant predictors of student growth in mathematics and English Language Arts.  A one standard deviation difference in Expectations is associated with almost a two percentile point increase (p<.01) in median SGP in mathematics.  We also find that Safety, Leadership, and Collegiality are all independent predictors of teacher turnover.  A one standard deviation difference in Collegiality is associated with a 1.4 percentage point decrease (p<.01) in teacher turnover.  These findings are robust to a variety of specification checks and alternative models. Our results provide new empirical evidence of the link between schools’ contexts, student growth, and teachers’ career decisions, and provide specific guidance to practitioners on how to design targeted interventions.