Association for Public Policy Analysis & Management

We document the extent and origins of racial bias in teacher assessments. Understanding how these biases form and persist in the classroom setting is important given evidence that teacher expectations shape grading patterns alongside students’ beliefs and behaviors (Rosenthal and Jacobson, 1968; Jussim and Harber, 2005; Lavy, 2008; Burgess and Greaves, 2013; Botelho et al., 2015; Lindahl, 2016; Papageorge et al., 2016; Hill and Jones, 2017). We make two contributions to this literature. The first distinguishes between favoritism and discrimination in teacher evaluations using blind-scored and more subjective assessments of student math and reading skills. This permits us to determine whether observed racial disparities in assessments are driven by teacher optimism preferring white students or pessimism focused on black students.

The second contribution develops and tests the hypothesis that early classroom conditions matter for subsequent teacher assessments. While researchers have examined the extent of racial bias in teacher expectations in the US and international contexts, less is known about the formation of these biases. We hypothesize that the context of novice teachers’ first classrooms (“first impressions”), in particular the distribution of academic abilities by student racial group, influence how teachers assess students of the same racial groups later in their professional careers.

We use matched student-teacher administrative data from the North Carolina Education Research Data Center (NCERDC), which provides both blind-scored math and reading results and non-blind teacher assessments of the same underlying student skills. We find evidence for teacher favoritism in both math and reading. Teachers are more likely to over-assess white students even after accounting for raw standardized test scores, behavioral attributes, and teacher-level factors such as the tendency to be lenient or strict across all students. On the flip side, there is no evidence for teacher discrimination in math, suggesting that racial differences in math assessments are solely driven by favoritism. We furthermore document discrimination in reading at a magnitude comparable to favoritism in the same subject.

Next we investigate the formation of these biases by characterizing the early classroom-level scholastic ability distributions faced by novice teachers. In doing so we rely on course membership data linking students and teachers. We find evidence that early classroom compositions matter for future teacher favoritism, with the effects most robust for math performance evaluations. A larger math performance gap favoring white students in those original classrooms results in greater teacher favoritism of subsequent cohorts of white students when assessing math skills. We furthermore evaluate whether teacher assessments are more or less sensitive to outliers in their initial classrooms’ ability distributions. Preliminary evidence suggests that when a novice teacher has a black student as the lowest scorer in either subject, they are more likely to exhibit optimism towards white students in later cohorts. Taken together, these findings shed light on the ways early assignment of teachers’ to schools and classrooms with disadvantaged minorities generates a biased perception of members from those demographic throughout a teacher’s careers. Our findings can inform classroom assignment and professional development policies for novice teachers.