The test score analysis was based on standardized achievement tests that the district normally conducts.8 While district-administered test scores may not cover every domain of student achievement that induction might affect, they do capture the content that school districts or states deem most important and worthy of assessing. We aggregated test scores across districts and grades by standardizing each test to a common metric called a z-score, which has a mean of zero and a standard deviation of one. The impact estimates are regression-adjusted using covariates that include the normalized student pre-test score, student characteristics, teacher personal characteristics, teacher professional characteristics, and district-by-grade fixed effects.
The findings, summarized in Tables 2 and 3, show the grade-specific impact estimates to be negative and statistically significant for grade 2 for reading (effect size = -0.22) and for grades 2 and 3 for math (effect size = -0.38 and -0.26, respectively), but the average impacts across all grades were not significantly different from zero for math or reading. The findings were robust to different analysis methods, such as regression with an omitted pre-test or regression with alternative weights or different sets of control variables.