School social segregation and social inequalities in political engagement among 16 to 20 year olds in fourteen countries

ABSTRACT

This paper assesses the explanatory power of a perspective arguing that school social segregation enhances social inequalities in political engagement because of the distinct effects that concentrations of adolescents of disadvantaged backgrounds in educational settings generate. It tests this argument with data of the 2000 Civic Education Study among Upper Secondary students and uses intentions to participate and political competences as outcomes to represent political engagement. Social inequalities in political competences turn out to be greater in states with the most segregated systems, but social disparities in intentions to participate are unrelated to the level of segregation. The paper argues that policy makers should consider creating a more integrated school system if they seek to reduce social inequalities in informed political participation.

KEYWORDS:

• Social segregation
• political engagement
• upper secondary
• IEA civic education study
• multilevel analysis
• social composition

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

1. I focus on classrooms as it represents the lowest level of aggregation. Segregation at that level is difficult to imagine. In contrast, at higher levels of aggregation, such as that of the school, it is quite possible that a heterogeneous make-up hides considerable homogeneity at a lower level – in which case the school could be characterised as internally segregated. This is likely to happen, for instance, in schools that offer various tracks and that admit students to these tracks on the basis of ability (which, as already mentioned, always has social sorting effects). Thus, school that are equally heterogeneous in their social make-up can differ greatly in levels of internal segregation. This complicates comparisons between schools and is the reason why I measure social composition at the level of classrooms.

2. Although the last three variables are ordinal, I included them as continuous variables in the analyses rather than as a series of dummy variables to save space (see Table 2). Analyses with these variables included as dummies produced identical results (obtainable upon request).

3. Unfortunately classroom ethnic diversity could not be measured more directly because the survey did not query ethnic identity in the majority of the participating countries. However, for the countries where ethnic identity was available (Cyprus, Latvia, Sweden and Switzerland), I could calculate the ethnic fractionalisation index (for the calculation of the EFI, see Posner 2004). This index showed a significant negative correlation at the classroom-level with classroom ethnic homogeneity (−.24* in Cyprus; −.34*** in Latvia; −.68*** in Sweden and.56*** in Switzerland), suggesting that the latter is a good (inverse) proxy of ethnic diversity.

4. The attentive reader will have noticed that I aggregated from the individual-level variable parental education to construct a measure for both classroom social status (at the classroom level) and social segregation (at the country level). There are no other sources available to develop measures for these concepts (OECD PISA data, for instance, could not be used as this data concerns lower secondary). To the knowledge of the author, drawing on the same data source to develop these higher level variables is not producing any specific bias. To the contrary, doing so ensures that the social context variables thus constructed are more valid measures of educational environments because they are directly linked to the respondents. If I had used a measure of school social composition based on, say, the percentage entitled to free school meals from some external data source (if that had been available), then the link with the study’s respondents would have been much more tenuous in view of the one classroom per school sample and the possibility that this classroom would not be representative of the school. Perhaps for this reason, many studies aggregate from individual-level measures of social background to develop measures of school social status (e.g. Kahne and Middaugh 2008; Palardy 2008; Liu et al. 2015) or country-level social segregation (e.g. Jenkins, Micklewright, and Schnepf 2008).

5. I did not impute missing values on the dependent variables as this can add unnecessary random variation into the imputed values (Allison 2012).

6. Economic prosperity was measured with statistics on GDP per capita 2000 from the World Bank while democratic tradition was tapped with the number of years of uninterrupted democracy since the introduction of universal suffrage. The b coefficient of social segregation was.34 and significant at the.05 level. Economic prosperity and democratic tradition were not significantly related to the outcome (full results of this model can be obtained from the author upon request).

7. The coefficients of economic prosperity and democratic tradition are not significant (full results obtainable upon request).

Additional information

Funding

This work was supported by the Economic and Social Research Council [ES/J019135/1].

Notes on contributors

Jan Germen Janmaat

Jan Germen Janmaat is Professor in Comparative Social Science at UCL Institute of Education. His research interests include the effect of education broadly conceived on civic values and the ways in which education can help mitigate inequalities in those values. He has published on these topics in a wide variety of journals in education, sociology, and political science. His latest book, co-authored with Prof Bryony Hoskins, is Education, Democracy and Inequality: Political Engagement and Citizenship Education in Europe (Palgrave: 2019).