The large and persistent achievement gaps separating minority and nonminority students are arguably the most important educational problem in the United States. In particular, reducing or eliminating these gaps by raising the achievement of minority students is widely seen as a critical component of promoting broader social equality with respect to a variety of outcomes like educational attainment and earnings as well as crime, health, and family structure (eg, Christopher Jencks and Meredith Phillips, 1998). The more modest gender gaps in achievement are also viewed as a prominent policy concern, particularly with respect to the fields of science and mathematics (eg, American Association of University Women, 1992). The recent federal legislation, the No Child Left Behind (NCLB) Act of 2001, clearly reflects these concerns, explicitly requiring that these demographic subgroups make “adequate yearly progress” toward proficiency on state tests. NCLB also emphasizes that schools should meet these and other goals by implementing effective reforms grounded in “scientifically based” research. However, the available empirical evidence suggests that the determinants of the demographic achievement gaps are not very well understood. For example, Jencks and Phillips (1998) argue that traditional explanations for the black–white achievement gaps (ie, those based on income inequality, differences in family structure, and school spending) actually have relatively little explanatory power. They also suggest that more successful future explanations of achievement gaps are likely to be based on credible data that assess more nuanced hypotheses about the dynamics within schools, classrooms and families. The notion that the classroom dynamics between teachers and students make a substantive contribution to the demographic gaps in achievement already has a wide currency among educational researchers and commentators. For example, proposals for promoting racial and gender equity often emphasize the need for improvements in teacher training and professional development (eg, American Association of University Women, 1992; Ronald F. Ferguson, 1998). Another frequently recurring proposal for increasing the relative achievement of minority students is to improve the recruitment and retention of minority teachers (eg, Patricia A. Graham, 1987; Joint Center for Political Studies, 1989; National Commission on Teaching and America’s Future, 1996; Beatriz Chu Clewell and Ana Marıa Villegas, 1998). However, the evidence that the demographic interactions between students and teachers matter is surprisingly thin, sometimes contradictory, and usually based on small, localized samples. In this study, I present new evidence on this issue by evaluating whether assignment to a demographically similar teacher influences the teacher’s subjective evaluations of student behavior and performance. This analysis is based on a large, nationally representative survey, the National Education Longitudinal Study of 1988 (NELS: 88). These data allow me to examine whether the effects of demographically similar teachers vary with key student traits or across Census regions. However, the key innovation of this analysis arguably involves how the identification strategy exploits a unique feature of the NELS: 88 survey design to purge the potential biases created by the nonrandom sorting of students across and within schools. Specifically, for each of the 8th-grade students surveyed, NELS: 88 solicited student-specific evaluations from teachers in two distinct academic subjects. This unusual feature of the data makes it pos-* Department of Economics, 500 …