This paper will investigate the development of segregation in Swedish compulsory schools with respect to educational achievement, socioeconomic status and ethnicity, and the causal links among the three aspects. Since the early 1990s the Swedish school system has undergone some fundamental changes. The shift of control and responsibilities of education from central government to different levels of local authorities, the increasing number of independent schools and the introduction of a voucher system with free school choice have thoroughly transformed the Swedish school system into a highly decentralized and deregulated one (Björklund, et al., 2005). While the goal that educational quality should be uniform over Swedish schools is still maintained, trends towards increased socioeconomic and ethnic segregation of students have been observed in recent studies (Gustafsson, 2006). It may be assumed that these trends are at least partially due to the increased frequency of school choice (Levin, 1998). There also are tendencies towards increased differences in level of achievement between schools (Skolverket, 2006; Gustafsson & Yang-Hansen, 2009), which may be related to differences between municipalities in the amount of support given to education. It may also be related to the increased socioeconomic and ethnic segregation, since research findings indicate that schools with a greater percentage of students from high SES family background provide more favorable learning environments, which leads to a higher school average achievement (Hoxby, 2000; Thrupp et al., 2002). It may thus be hypothesized that segregation has become intensified in Swedish compulsory schools; the changes in the between-school achievement differences may be related to the changes in between-school differences in SES and ethnicity. Even though the assumption of the causal links between achievement segregation and the socioeconomic and ethnic segregation sounds reasonable, there are, however, very few studies that have examined the direct causality between these factors. To test the hypothesis, two questions are to be answered: • Whether or not the differences in educational achievement, socioeconomic status of school intakes, proportion students with ethnic background have changed between Swedish compulsory schools? • Are there any causal links among these three aspects? The trend towards decentralization and deregulation of schooling with free choice of school has been observed in a European and a global context. The consequences of such a trend on educational stratification and equality attract great public and political attention worldwide. The empirical evidences from Sweden may offer insights to the impact of the global school reforms.

The Gothenburg Educational Longitudinal Database contains ample amount of educational and demographic information on all individuals being born year 1972-1992. Each cohort is comprised of approximately 100 000 students and 1000 schools. There are also repeated measures of, among other things, students’ school marks, their parents’ education level and their ethnicity. These three variables of the 9th graders between 1998 and 2007 will be focused upon in this study. Applying the Growth modeling technique (Hedeker, 2004), the trend in school segregation in school achievement, SES and ethnicity can be studied. Growth models with random intercept and random slope will be fitted respectively for the three aspects of segregation. The estimates of random effects of intercept and slope in these growth models capture the overall pattern of changes in these aforementioned aspects, which will be linked together to examine the causality between them in a later step.

The current hypothesis can hopefully be confirmed by empirical results. An intensive growth of between-school differences in the merit score can be expected. The growths of between-school differences may also be found in both SES and ethnicity, but to a less extend, compared to that of school achievement. It can also be expected that the changes in between-school variation in school achievement can be explained by the changes in the differences of school SES and of proportion students with immigrant background. Also the variance in the intercept of school marks is expected to be related with the variance in the intercepts of school SES and ethnicity.