When the Winds Shift: How Changes in Racial Bias Relate to Changes in Educational Outcomes

• In Event: Level Matters: Inter- and Intra- District Analyses Exploring Nuances in Racialized Power Across Space and Place Over Time

Abstract

Prior studies have found that county-level rates of implicit and explicit anti-Blackness are predictive of worse educational outcomes for Black students, including exclusionary discipline, academic achievement, and talented and gifted program placement (Author et al., 2020; Author & Author, 2022; Riddle & Sinclair, 2019). Studies have also shown that aggregate measures of anti-Blackness are predictive of larger Black-White racial disparities in academic and disciplinary outcomes (Author et al., 2020; Author, 2022). Moreover, work within schools has demonstrated that racial bias plays a role in driving racial disparities in discipline (e.g., Okonofua & Eberhardt, 2015). Recent empirical work has documented that racial bias changes over time (Charlesworth & Banaji, 2019) and may be responsive to exogenous shocks (Author et al., 2020); and that individuals’ sociodemographic characteristics may moderate the relationship between exposure to exogenous shocks and shifts in bias (Author et al., 2020).

Taken together, prior research thus suggests that aggregate rates of racial bias are related to educational outcomes and disparities therein, that bias can change, and that bias does not change uniformly across populations. Given these findings, we ask: Do county-level rates of racial bias change over time and, if so, do changes in racial bias relate to changes in educational outcomes and disparities? To answer this question, we merge implicit and explicit anti-Blackness data from the Project Implicit datasets with a variety of data regarding student outcomes (including the Department of Education’s Civil Rights Data Collection and Stanford’s Education Opportunity Project). This work is thus the first to empirically ascertain whether counties that see declines (increases) in anti-Blackness also evidence improvements (declines) in Black students’ educational outcomes relative to their peers.

Our key regression models seek to identify how changes in bias over time are related to changes in educational outcomes over time, netting out all time-invariant endogeneity. To empower these regressions, for each county in each year, we estimate mean outcomes (e.g., Black student test scores and Black student discipline rates) and we estimate mean levels of our exposures (implicit and explicit bias). We then combine these county-year estimates to create a county-level panel dataset that features both our exposures and outcomes of interest. Finally, we leverage this panel dataset to regress changes in outcomes against changes in exposure.

While we enter this analytic process with the intention of implementing a fixed effects strategy, we will also conduct appropriate robustness tests and sensitivity analyses. These may include investigating how changes in anti-Blackness relate to changes in outcomes for non-Black students; investigating how lagged measures of changes in educational outcomes are related to measures of prior changes in racial bias; reviewing district-level analyses to determine whether the geographic unit of analysis drives the results we observe; and investigating whether changes in teachers’ racial biases are related to changes in educational outcomes.

By providing novel evidence on the importance of geographic variation in measures of anti-Blackness, we hope to both provide further rigorous evidence on the need to address anti-Blackness in schools and to further understand the tractability of these biases.

Authors

• Mark Chin, Vanderbilt University

• Sean Darling-Hammond, University of California - Berkeley

• Francis A. Pearman, Stanford University