Anti-school attitudes, school culture and friendship networks

Abstract

This article explores the relationship between educational outcomes and anti-school attitudes at different levels of social organization in schools. Data were collected in St. Petersburg, Russia (104 schools, 7300 students) and analyzed using multi-level regression models that included three levels: individual, clique of friends and school. A clique is defined as a tight group of friends in a school class; we used social network analysis software Kliquefinder for clique identification. We demonstrate that friends’ attitudes are strongly related to the educational outcomes of a student (net of person’s individual attitudes and socio-demographic characteristics). In contrast, school-level effects disappear in the multi-level model when individual characteristics are included. The results of the study clearly demonstrate that the socio-economic and curricular differentiation of schools does not always lead to the polarization of ‘school academic cultures’. A school social environment is sufficiently heterogeneous, and different value systems in small peer groups may coexist.

Keywords:

• Anti-school culture
• differentiation–polarization
• peer effects
• friendship networks
• cliques
• multi-level models

Acknowledgements

The authors gratefully acknowledge the participation of the schools and the students who made this project possible.

Notes

1. In formal graph theory, clique is defined as a complete subgraph; that is, a set of nodes where each node is connected with all others. In this article we define ‘clique’ as a network community; that is, a group of nodes that have more dense ties within the group than outside the group. Software for network community detection include Kliquefinder, Ucinet, Igraph and others.

2. See Accessed 14 September 2017. https://msu.edu/~kenfrank/social%20network/Introduction%20to%20KliqueFinder.pdf

3. See Accessed 20 November 2017. http://www.harryganzeboom.nl/isco08/index.htm

4. International migration is a relatively new phenomenon in Russia. First waves of migration from new independent states, formerly Soviet Republics, started at the beginning of the 1990s, after the dissolution of the USSR. Traditionally researchers distinguish between first and second generations of migrants, but when applied to children this classification does not work well, as demonstrated by Rumbaut (2004). In addition to the second generation (the US-born children of foreign-born parents), Rumbaut proposed a more fine-tuned classification based on age and life stage at arrival: the 1.75 generation who arrived as pre-school children, ages 0–5; the 1.5 generation, arriving at ages 6–12; and the 1.25 generation, arriving at ages 13–17. According to this classification, immigrant students in our sample mostly belonged to the second, 1.75 and 1.5 generations.

5. ICC = (between mean square – within mean square) / between mean square.

6. In preliminary analysis we used school size as an additional control variable on the school level. The analysis shows that the school size coefficient is insignificant. Moreover, it does not improve the quality of the model (as demonstrated by negligible difference in –2*loglikelihood).

7. In preliminary analysis, school type and school size were included as control variables in the last step. The coefficients were insignificant, and there was no improvement in the model fit (no difference in –2*loglikelihood).

8. The deviance statistic (defined as –2*loglikelihood) is a measure of the overall goodness of fit of a model; the smaller the value, the better the statistical fit. The fit of two nested models can be compared by comparing their deviances.