Classrooms are becoming more diverse, with students of all abilities being mainstreamed into general education classrooms. Teachers are now tasked with differentiating instruction for a wide range of learning levels (Tomlinson, 2000). Even good teachers are not necessarily good at teaching all students all things. Unfortunately, little differentiation occurs in most classrooms (see Reis & Renzulli, 2010 for a review), and recent studies suggest that teachers often focus most of their attention on average students or students who are close to meeting a proficiency benchmark (Kenney et al., 2024; Neal & Schanzenbach, 2010; Rambo-Hernandez & McCoach, 2014). Correspondingly, the highest performing students may grow more slowly than others, suggesting that these students may not be reaching their potential.

Although these studies shed light on what happens in the average classroom, they do not necessarily address teachers’ relative effectiveness across different student groups. Using 8 years of rich administrative data from a large California district, we investigated the differential effectiveness of teachers across student populations, with a focus on gifted and talented (GT) students. Using data from Los Angeles, we estimated teacher effectiveness (proxied by value-added measures [VAMs] of teachers’ contributions to student achievement growth) for GT students and examined how they are matched to teachers. We additionally estimated VAMs for high- and low-testing students to explore whether GT students are assigned to teachers who are relatively more effective for GT students, or to teachers with broad effectiveness across student groups. Many teachers exhibited similar effectiveness across all subgroups, with particularly strong alignment between GT and high-testing VAMs. This suggests a strong overlap in instructional strengths for high-achieving students, regardless of GT classification. At the same time, a substantial share of teachers showed meaningful differences in their relative effectiveness with GT students compared to low-testing students. GT students were consistently assigned to teachers with higher VAMs across all subgroups; however, this alignment was strongest with teachers who were most effective for GT and high-testing students. To a modest extent, GT students were also more likely to be matched to teachers with a relative advantage in teaching GT students compared to their effectiveness with low-testing students.

For some highlights of the results, first we were able to detect differences in teacher quality measured via GT specific value-added (VA). The measured gap between VA for GT and low-testing students for each teacher showed considerably more variation than would be expected if we randomly assigned students to GT and low-testing categories, indicating a sizable signal.

In general, teacher’s observable characteristics did not tend to relate strongly with their GT or high testing value-added, although late career math teachers somewhat surprisingly seemed to have worse GT value-added than early career teachers.

In LAUSD, GT students tend to get higher VA teachers along all three dimensions – GT, high-testing, and low-testing VA, and disproportionately gained access to teachers who were better specifically at GT instruction.

GT students were also more likely to be assigned to teachers who had a relative advantage in instructing GT students (e.g., they have a higher VA in GT than low-testing students regardless of their overall VA). We learned that there was sufficient information in value-added models to distinguish between teachers who were more and less effective at instructing GT students specifically. We also found that, at least in LAUSD, GT students were largely matched with teachers who were more able to improve GT performance. Overall, we hope the study can help schools and districts better target instruction for GT students.