Social integration in the context of increasing immigration is a challenge faced by many industrialized countries. The institutional set-up of an education system is a natural candidate for scrutiny from a policy perspective, since education is at the forefront of social integration and institutional structures are malleable in nature. In this study, we examine immigrant social integration through school segregation and how it relates to the institutional structure of an educational system, in particular the existence of (early between-school) tracking. As a premise for integration, school segregation is crucial since it determines how much interaction between immigrants and non-immigrants occurs. Tracking on the other hand is as controversial as it is momentous since students are placed into different types of schools from an age as young as ten (e.g. Terrin & Triventi 2022, van de Werfhorst & Mijs 2010). This is in stark contrast to the practice of integrated education systems, in which students may be grouped by ability for certain topics or classes, but are only separated as they approach maturity. 

We therefore examine whether tracked education systems show higher levels of school segregation of immigrants as education systems that delay between-school grouping. While sociologists have long documented the negative effects of tracking with regards to equality of opportunity (e.g. Terrin & Triventi 2022, van de Werfhorst & Mijs 2010), we argue that there may be counteracting mechanisms at work in tracked systems when it comes to ethnic segregation between schools. 

We follow theories on educational inequality to understand school segregation and tracking. These theories relate differences in family resources to differences in educational attainment or achievement (i.e. Boudon 1974, Bourdieu 1987, Lareau 2011). Resources in this context can comprise economic capital, strategic knowledge, social contacts and familiarity with modes of behavior in the education system. In this context, we expect that immigrant students are disadvantaged, as many of these resources cannot easily be translated from the home country to the receiving country. We can therefore expect that they will show, on average, lower achievement at the end of primary school (i.e. primary effects; Boudon 1974). These finding has been shown by numerous studies (e.g. Heath et al. 2008). 

Since observed achievement is a major indicator of track placement, primary effects of ethnic and social origin increase the likelihood for immigrant students to be sorted into lower secondary school tracks. However, parental decision making (i.e. secondary effects; Boudon 1974) is another determinant of track placement and it is well-known that immigrant parents tend to choose more ambitious educational pathways (Esser 2016; Gresch et al. 2012) which could compensate for low track placement based on ability. Lastly, school segregation likely exists in non-tracked systems as well. First, because home-to-school-distances are a main factor in selecting a school, residential segregation, which is a common phenomenon in many countries, is reflected in school segregation. Further, school choice behavior of non-immigrant families may contribute to ethnic school segregation, as particularly high status families tend to avoid schools with larger numbers of immigrants (“white flight”; Amor 1980). They do so, because they use immigrant concentration as a proxy for (lower) school quality. This tendency might be lower in tracked-systems, as track level is an accessible indicator of school quality. Non-immigrant families therefore do not need to avoid schools with larger numbers of immigrants (c.p. Meier & Schütz 2007). 

In sum, there may be counteracting mechanisms with regard to school segregation and the age of first tracking. We therefore argue that it remains an empirical question to determine which mechanism outweighs the other.

Previous research on the effects of tracking on ethnic segregation point towards mixed effects of tracking (e.g. Kruse 2019). However, most previous findings look at data from single countries or cities. Moreover, they face the challenge of cross-sectional analyses that might be biased by unobserved heterogeneity. We therefore aim at generating more generalizable findings on the impact of tracking on segregation by combining all data from PISA, TIMSS and PIRLS cycles between 1995 and 2018 for a total of 45 countries. In order to combine the data, we harmonized the relevant information, most importantly information on immigrant background. We define immigrant background by the place of birth of the student (abroad). Based on this information, we calculated measures of segregation (index of dissimilarity D, Duncan & Duncan 1955) for each study-year and each country. Crucial for our analytical approach is the fact that some of the studies are implemented in primary school - when no education system is tracked - and others are administered in secondary school (in grade 8 or at age 15), i.e. after tracking has been exercised. According to our definition (tracking takes place before grade 8) this is the case for nine countries in our sample. Our analysis is based on a difference-in-differences approach that compares the difference in ethnic segregation between primary and secondary school and between tracked and untracked countries. This approach enables us to account for all other time-stable differences between countries. We still included control variables that can change over time: the gross domestic product and the population density and the privatization of the education system. Such decisions (e.g. including control variables or excluding probable outliers) however might have substantial impact on the obtained estimates. We therefore do not conduct a single analysis, instead we follow the approach of multiverse analyses (Simonsohn et al. 2020). The term "multiverse analysis" refers to a type of analysis that accounts for the problem of multiple “forking paths” (Gelman & Loken 2013), because a research design has to be operationalized with variables, samples and estimation techniques. By systematically varying these decisions across all possible paths, we will “expand” a multiverse that incorporates all possible paths. In other words, it is a systematic way of doing robustness checks.

Our preliminary results suggest that the presence of early between-school grouping (as compared to late between school grouping) has no discernible impact on immigrant school segregation. While segregation increases in both types of education systems there are heterogenous effects across model specifications with respect to the effect of tracking. By varying the choice of fixed effects, control variables (GDP, private school density and population density) and sample restrictions (different GDP cut-offs and different cut-offs for minimum or maximum share of immigrant students in a country) we obtain about 4000 model specifications of which 60% show a small negative (but overwhelmingly insignificant) effect and 40% show a small positive (but overwhelmingly insignificant) effect on school segregation. In our next steps, we will examine the effects of selectivity on segregation. We expect that higher selectivity will limit the ambitious school choices of immigrant families and therefore lead to higher levels of school segregation.

Armor, D. J. (1980). White flight and the future of school desegregation. School desegregation: Past, present, and future, 187-226. Bourdieu, P. (1987). Die feinen Unterschiede. Suhrkamp. Duncan, O. D., & Duncan, B. (1955). A Methodological Analysis of Segregation Indexes. American Sociological Review, 20(2), 210–217. Retrieved from http://www.jstor.org/stable/2088328 Esser, H. (2016). Bildungssysteme und ethnische Bildungsungleichheiten. Ethnische Ungleichheiten im Bildungsverlauf: Mechanismen, Befunde, Debatten, 331-396. [English: “Education systems and ethnic educational inequalities” in “Ethnic inequality along the educational pathway: Mechanisms, Results, Debates”] Gelman, A., & Loken, E. (2013). The garden of forking paths: Why multiple comparisons can be a problem, even when there is no “fishing expedition” or “p-hacking” and the research hypothesis was posited ahead of time. Department of Statistics, Columbia University, 348, 1-17. Gresch, C., Maaz, K., Becker, M., & McElvany, N. (2012). Zur hohen Bildungsaspiration von Migranten beim Übergang von der Grundschule in die Sekundarstufe: Fakt oder Artefakt. Soziale Ungleichheit in der Einwanderungsgesellschaft. Kategorien, Konzepte, Einflussfaktoren, 56-67. [English: “The case of high educational aspirations among migrants when transitioning from primary school to secondary school: fact or artifact?”] Heath, A. F., Rothon, C., & Kilpi, E. (2008). The Second Generation in Western Europe: Education, Unemployment, and Occupational Attainment. Annual Review of Sociology, 34(1), 211–235. https://doi.org/10.1146/annurev.soc.34.040507.134728 Kruse, H. (2019). Between-school ability tracking and ethnic segregation in secondary schooling. Social Forces, 98(1), 119-146. Lareau, A. (2011). Unequal Childhoods: Class, Race, and Family Life. Univ of California Press. Meier, V., & Schütz, G. (2007). The economics of tracking and non-tracking (No. 50). Ifo working paper. Simonsohn, U., Simmons, J. P., & Nelson, L. D. (2020). Specification curve analysis. Nature Human Behaviour, 4(11), 1208-1214. Terrin, E., & Triventi, M. (2022). The effect of school tracking on student achievement and inequality: A meta-analysis. Review of Educational Research, 00346543221100850. Van de Werfhorst, H. G., & Mijs, J. J. (2010). Achievement inequality and the institutional structure of educational systems: A comparative perspective. Annual review of sociology, 36, 407-428.