Since the launch of the voucher system with free school choice in Sweden in the early 1990s, individual students were offered opportunities to attend schools of their preferences. And the independent school reform made establishing independent schools easier. The growing number of independent schools offers different alternatives for students to choose from. Even though the free school choice and the independent school reform can possibly break the residential segregation and thus reduce the educational inequality, empirical evidence has shown that the social background effect on academic achievement has been strengthened between schools and municipalities in the past decades (Yang Hansen & Gustafsson, 2019). Similar reforms have also observed in Norway, albeit to a different extent and in different ways.   

One of the key factors behind strengthened educational inequality is believed to be peer effects. Education policy changes, such as free school choice, often make substantial changes to the composition of a student’s peer group, especially when such a policy is mostly taken advantage of by the well-educated middle-class parents (Levin, 2002). It may be expected that the learning outcomes may be affected by the school/classroom mix. When students in schools or classrooms are sorted by social, ethnic background and/or prior achievement, both students’ cognitive and non-cognitive outcomes will be affected by the characteristics of their peers (Hanushek et al., 2003, Hoxby, 2000, Marsh et al., 2008).

Other school organisational factors, such as tracking and ability grouping, also affect students composition in classrooms or schools. However, informal ability grouping for instruction has become more common in recent years in Sweden (Mullis et al., 2012). Ammermüller & Pischke (2009) studied peer effects in six European countries using IES PIRLS 2001 data. Although the average peer effect estimated for Sweden was somewhat similar to those in the rest of the countries, they did find evidence that students in the Swedish sample may not have been randomly assigned to their classes.

Though research has demonstrated the importance of peer effects, the estimated size was rather small. This is partly due to the methodological challenges in estimating peer effect since students composition in schools and classrooms is often not randomly formed. The endogeneity hence will bias the estimation of peer effects. One of the commonly used approaches is using the within-school variance to identify peer effect in a school fixed-effect model. However, this approach assumes that the students in the classroom are randomly assigned. If students' characteristics in the classrooms are related to the model's unmeasured components, the estimation of variance between classrooms does not represent the peer characteristics.

However, Hoxby (2000) demonstrated that such selection biases could be tackled by analysing adjacent cohorts simultaneously. She assumed that “there is some variation in the adjacent cohorts’ peer composition within a grade within a school that is idiosyncratic and beyond the easy management of parents and schools (p.3). However, such adjacent cohorts data is rarely available. Therefore, the proposed study is to explore the measurement of peer effects and estimate the size of peer effects in Sweden and Norway over the past decade.

There are method challenges in investigating peer effects. A first challenge is the selection bias caused by the parental choice of school based on its population of peers and by parents’ and schools’ manipulation of the assignment of students to classes within their grades. Secondly, factors that student selection is based upon often interact with other unobservable characteristics. Thirdly, the implementation of policy changes is accompanied by other characteristics of schools and their student intake, which most likely mediate with peers and linearly or non-linearly affect student’s learning outcome. The study is thus to apply multilevel techniques with fixed and random effects (Lüdtke et al. 2008), the Instrument Variable approach used by Ammermüller and Pischke (2009), and the analytical strategies using adjacent cohorts data proposed by Hoxby (2000). The IEA TIMSS data between 2003 and 2015 from Sweden and Norway will be used. Variables like sex, age, number of books at home, educational recourses and language use at home, parental education, the country that the parents were born in, will be used to measure peer characters. The results from different approaches will be compared, and to see how the estimates of peer effects differ under different assumptions of students assignment.

The analysis is still ongoing, and it is still too early to make any generalisations of the results. Ammermüller and Pischke (2009) found that a standard deviation change in peer composition measure leads to .17 standard deviation change in reading test score. Hoxby’s estimates (2000) were between .15 and .40 depending on the specification. We expect our estimates to be around those ranges. However, our preliminary results showed no significant peer effects when controlling for school- and classroom differences. Multilevel techniques and IV-approaches applied in the current analyses differed in details, but yielded the same main results. Descriptively, peer effects were weak. Further check and analyses need to be carried out.

Ammermüller, A. & Pischke, J-S. (2009). Peer Effects in European Primary Schools: Evidence from PIRLS. Journal of Labor Economics, 27(3), 315-348. Hanushek, E. A., Kain, J. F., Markman, J. M., & Rivkin, S. G. (2003). Does Peer Ability Affect Student Achievement? Journal of Applied Econometrics, 18(5), 527-544. Hoxby, C. (2000). Peer Effects in the Classroom: Learning from Gender and Race Variation. NBER working paper no. 7867. National Bureau of Economic Research. Hattie, J. (2009). Visible learning. A synthesis of over 800 meta-analyses relating to achievement. London: Routledge. Levin, H. M. (2002). A Comprehensive Framework for Evaluating Educational Vouchers. Educational Evaluation and Policy Analysis, 24, 159. Lüdtke, O., Marsh, H. W., Robitzsch, A., Trautwein, U., Asparouhov, T., & Muthén, B. (2008). The multilevel latent covariate model: A new, more reliable approach to group-level effects in contextual studies Psychological Methods 13, 203-229. Marsh, H. W., Seaton, M., Trautwein, U., Lüdtke, O., Hau, K. T., O’Mara, A. J., & Craven, R. G. (2008). The big-fish-little-pond-effect stands up to critical scrutiny: Implications for theory, methodology, and future research. Educational Psychology Review, 20, 319-350. Sallis, J. F., Prochaska, J. J., & Taylor, W. C. (2000). A review of correlates of physical activity of children and adolescents. Medicine & Science in sports & Exercise, 32, 963-975. Sund, K. (2009). Estimating peer effects in Swedish high school using school, teacher, and student fixed effect. Economics of Education Review, 28(3), 329-336. Yang Hansen, K., & Gustafsson, J. E. (2019). Identifying the key source of deteriorating educational equity in Sweden between 1998 and 2014. International Journal of Educational Research, 93, 79-90. https://doi.org/10.1016/j.ijer.2018.09.012