Association for Public Policy Analysis & Management

Average student performance on standardized math and reading tests has improved since the 1970s (US Department of Education, 2016), but educational opportunities are not evenly distributed across places and student populations (Reardon 2018). Notably, we know little about how national trends in improvement vary by geography and subgroup. This study describes and explains the wide variation in trends in school district improvement across school districts and across student subgroups in the post-recession period, 2009-2016. If the primary forces contributing to improvement act at a national level (e.g. federal policies, national norms for schooling), we might expect to see little variation in trends at the district-level (net of trends induced by changes in demography). However, we find significant variation between school districts, indicating that the primary forces contributing to improvement are local.

To describe the variation in these trends, we leverage the Stanford Education Data Archive, which includes over 330 million third through eighth-grade test scores from 2009 through 2016 from over 11,000 school districts in the US. We use the Common Core of Data (CCD), the Civil Rights Data Collection (CRDC), the American Community Survey (ACS), and data from Stephens-Davidowitz (2014) on racial animus to show how districts’ academic performance trends relate to changes in district-level characteristics, including demographics, socioeconomic status, racial animus, and teacher-student ratios.

We find that the average change in achievement for all school districts is .007 standard deviations (SDs) per year, which is roughly equivalent to a gain of ⅕ of a grade level (GL) over a decade. This indicates substantial long-term improvement, though there is significant variation between districts and student demographic groups. Over 10 years, ⅙ of districts will have improved by more than .35 SDs (over 1 GL), and ⅙ will have declined by more than .21 SD (⅔ GL).

Further, while all student groups experienced improvement on average, the average district’s low-income students improved annually by .009 SD (.026 GL), and middle- and high-income students improved by .015 SD (.044 GL), showing an expansion of the average district’s income achievement disparity. The average district’s white students’ scores improved by .008 SD (.022 GL), and black students’ by .003 SD (.010 GL), suggesting a widening of the white-black achievement disparity. The average white-Hispanic disparities appear to be closing, as the average district’s Hispanic students’ scores improved by .012 SD (.035 GL).

We show that while opportunity is increasing on average across school districts and subgroups, opportunity is not increasing at the same rate for all groups of students. We analyze the variation in these trends to determine how much these differences are driven by changing populations including shifting demographics, gentrification, and growing wealth. Trends in racial and ethnic as well as economic disparities are largely driven by levels and changes in segregation and inequality. Still, the majority of variation in improvement appears to be fueled by local differences in policies and practices.