Panel Paper: Can Student Test Scores Provide Useful Measures of School Principals' Performance?

Friday, November 3, 2017
Comiskey (Hyatt Regency Chicago)

*Names in bold indicate Presenter

Hanley Chiang, Moira McCullough, Stephen Lipscomb and Brian Gill, Mathematica Policy Research

Principals have the potential to influence student achievement in numerous ways, such as by shaping their school's academic culture, recruiting and retaining effective teachers, and promoting teachers' growth on the job. However, in contrast to the growing value-added literature on measuring teachers' impacts on student achievement, evidence is scant on how to design valid measures of principals' impacts on student achievement. In this study, we examine the validity of several types of principal performance measures that states currently use or could widely implement with student test score data. Using a quasi-experimental approach, we estimate each measure's validity by the extent to which ratings from the measure predict future test-score changes at schools in which higher-rated principals replace, or are replaced by, lower-rated principals. We apply this approach to student and principal data from Pennsylvania over 7 years. We find that performance measures based only on students' end-of-year achievement, without accounting for their past achievement, provide no information for predicting principals' impacts on student achievement in the following year. Value-added measures that account for students' past achievement provide, at most, a small amount of information for predicting principals' impacts in the following year. These findings indicate that value-added measures are the only test-based measures of principal performance with any validity, but even those measures will lead to many errors when identifying effective and ineffective principals.