Association for Public Policy Analysis & Management

An influential paper by Dee (2005) exploits a unique feature of the widely used, nationally representative National Education Longitudinal Study of 1998 (NELS:88) dataset—that two teachers appraise the behavior of each student—to estimate the effect of demographic mismatch between students and teachers on teachers’ perceptions of students’ behaviors using a student fixed effects (FE) identification strategy. Exploiting this within-student variation in teacher perceptions and exposure to teachers from different demographic backgrounds, the author finds that when students are assigned to one demographically mismatched teacher who is of a different race or gender and one same-race or same-gender teacher, the mismatched teacher is significantly more likely to perceive the student as being frequently disruptive, frequently inattentive, and less likely to complete homework than is the teacher of a similar demographic background. Importantly, these findings provide insights into the potential mechanisms through which student-teacher demographic mismatch might affect academic achievement, as rigorous evidence of such effects is mounting (Clotfelter, Ladd, & Vigdor, 2007; Dee, 2004; Dee, 2007; Egalite, Kisida, & Winters, 2015). 

The proposed paper exploits a similar feature of the Educational Longitudinal Study of 2002 (ELS), and employs a similar identification strategy, to estimate the effects of student-teacher demographic mismatch on teachers’ beliefs over students’ ultimate educational attainment. In doing so, we build on existing work by Dee and others by investigating potential sources of biased beliefs and imperfect information that likely affect educational attainment, particularly that of information-poor, disadvantaged students. In addition to considering a different outcome, we extend Dee’s seminal analysis in two ways that both prove to be important. First, we examine more nuanced measures of mismatch that allow for interactive effects of being assigned an “other race” and “other sex” teacher. Second, we investigate heterogeneity by students’ gender and race simultaneously. 

Overall, being assigned to an “other-race” is associated with a three percentage point increase in the likelihood that the teacher expects the student to complete a high school diploma or less, and this effect is marginally statistically significant (p < 0.10). No corresponding effect is found on the likelihood that the teacher expects the student to complete a four-year college degree or more. However, analyses of the full sample mask substantial heterogeneities in the causal relationship between student-teacher demographic mismatch and teacher expectations. Strikingly, when black students are assigned to an other-race teacher, teachers are 12 percentage points more likely to expect a high school diploma or less and 9 percentage points less likely to expect a four-year college degree or more, and both of these effects are strongly statistically significant. From a policy standpoint these effects are arguably practically significant and constitute changes of 39 and 24 percent, respectively. These effects are greater in magnitude for black boys than they are for black girls, and similarly sized effects are observed for low-income students. These results have implications for the optimal assignment of students and teachers to classrooms, teacher training programs, and for the potential importance of (mis)information in the education production function.