ABSTRACT
This study examines the role of educational tracking in shaping school segregation among first and second-generation migrants, a pressing issue for societal integration. Drawing on the pooled data from international student assessments (PISA, PIRLS, & TIMSS) from 1995 to 2019, we examine a total of 75 countries and assess the extent to which early tracking exacerbates segregation compared to more integrated school systems. Our analysis, based on a Difference-in-Differences approach, reveals no significant differences in migrant school segregation between tracked and integrated systems. Robustness checks, employing multiverse analyses, confirm the stability of our findings across alternative model specifications. However, considerable variation in migrant school segregation exists within both types of education systems. This underscores the importance of investigating additional factors beyond early tracking contributing to variation in migrant school segregation. Our study highlights the complex interplay between educational policies and migrant integration in school settings.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Data availability statement
We used data from the PISA, PIRLS, and TIMSS studies, which are available at https://www.oecd.org/pisa/data/ and https://timssandpirls.bc.edu/databases-landing.html. respectively. The R scripts for data preparation, aggregation and analyses are available at the following link: https://osf.io/9vfz7/?view_only=e11dcb7ea14542268c8ddc15b8b092b8.
Supplementary material
Supplemental data for this article can be accessed online at https://doi.org/10.1080/03057925.2024.2370294
Notes
1. Some surveys could be matched with several others (e.g. TIMSS 2011 4th-grade students could be matched with both PISA 2012 students and TIMSS 2011 8th-grade students), leading to multiple observations for certain studies within the dataset. To address this issue, we combined the senate weights with duplicate weights using the inverse of the student observations (1/n). This approach ensured that each respondent’s influence on the analyses remained equal, regardless of the number of observations.