Association for Public Policy Analysis & Management

Ability tracking, or the practice of sorting students into classrooms based on ability, is regularly used around the world. In the US, around 75% of the schools surveyed by the National Assessment of Education Process (NAEP) reported using tracking in 8th grade mathematics. The OECD also reported that over 95% of students in the United Kingdom, Australia, Singapore, and Malaysia attended schools where students were grouped by ability across classrooms.

However, despite the popularity of tracking, there is still a debate over merits of the tracking system, especially whether or not tracking could harm students who are tracked into lower ability classroom through adverse peer effects. Unfortunately, relatively few papers identify the impact of being tracked into classrooms with higher or lower peer ability. This is, in part, likely due to the fact that while peer quality is a major difference between classrooms, other factors, such as the intensity of curriculum and teacher quality, also often change as well and therefore make it difficult to convincingly estimate the impact of peer quality on student achievement in the tracking system.

The purpose of this paper is therefore to identify the impact of being tracked into higher-ability classroom in a setting where only peer quality changes, thereby separating the effects of peers from other confounding factors. I do so by using data from public middle schools (7th-9th grade) in Thailand, where students are tracked into classrooms based on a preliminary exam taken before 7th grade and all teachers, curriculum, and textbooks are identical throughout classrooms. To distinguish the impact of peers from confounding factors due to selection, I apply a regression discontinuity design (RDD) that compares the academic outcomes of students just above and just below the threshold.

Results indicate that significant increases in peer quality do not lead to statistically higher student GPA. Importantly, the estimates are sufficiently precise as to rule out effects larger than 0.1 standard deviations. My results thus complement Duflo, Dupas, and Kremer (2011), where the authors found no statistically significant impact of peer quality in the tracking system but could not rule out effects smaller than 0.2 standard deviations. Furthermore, they also suggest that the big impact of gifted and talented programs found by Card and Giuliano (2016) and Booij, Haan, and Plug (2017), were likely due to other factors, such as the intensity of the curriculum and teacher quality, changing rather than increases in peer quality.

These results therefore have several important policy implications. First, they suggest that tracking, at least in Asian contexts, does not hurt students through adverse peer effects, as many have worried. Second, they suggest that more focus and resources should be put on other education inputs, such as teacher quality and curriculum, rather than peer quality.