Introduction The concept of social cohesion can be used at school level to denote the degree of connectivity between individuals or groups of individuals (cf. Beauvais & Jenson, 2002; Farrington, 1993; Peschar, 2005). Qualitative examples are presented in a report entitled ‘Encouraging tolerance and social cohesion through school education’ that was written for the Australian Department of Education, Science and Training (Carbines et al., 2006). According to these examples, environmental variables like degree of urbanisation and degree of poverty of the region where the school is located, are supposed to be relevant in the social cohesion and exclusion processes at school level. Moreover, school size is shown to be important (Mooij, 1994, 2001), in addition to school variables reflecting attention for the pupils’ involvement, attention for teaching qualities of teachers, instruction and learning progress of pupils, in-school regulation of social behaviour, and collaboration with external institutions to supervise the social behaviour of pupils in particular (Mooij & Smeets, 2006, in press). At pupil level, variables like gender, age, level of education of mother, level of education of father, level of educational attainment, (not) feeling at home in the Netherlands, and being religious, are expected to be relevant. Moreover, the individual pupil’s problem behaviour with respect to playing truant and possession of drugs and weapons, and the pupil’s perception of whether or not the school undertakes actions against such problem behaviour, will influence the feelings of safety of the pupil with respect to school. Systematic quantitative research into these issues is relatively scarce, however (cf. Loeber et al., 2008; Mooij, 2005). One of the reasons for this lack of information is that, using paper questionnaires in traditional survey or monitoring research, clarification requires the involvement of large numbers of different persons in rather extended research designs. Nowadays, however, possibilities to carry out such investigations are made possible by the use of the Internet as a means to collect data in comprehensive ways with multiple users (cf. Mooij, 2006, 2007). Using the Internet in fine-tuned procedures allows the flexible and large-scale use of differentiated ways to measure individual pupil and school characteristics in related ways. Research question Which variables at school level, and which variables at pupil level, do contribute to the explanation of the feelings of safety of individual pupils at school?

Data were collected with the aid of Dutch nationwide surveys carried out via the Internet in secondary schools, including schools for pupils with special educational needs. Questionnaires were completed in 2006 and in 2008 by about 80.000 pupils, 6.500 teachers and educational support staff, and 600 school leaders, from about 210 school locations. Data were checked for reliability, scale homogeneity, and representativeness. Social cohesion is indicated by aggregating scale and item scores of school leaders, teachers and support staff, and pupils, to the school level. Explorative two-level analysis using the 2006 data is carried out to construct a fitting explanation of the individual pupils’ feelings of safety. The resulting model is tested statistically by using the 2008 data. Moreover, school level data of 2006 is used to check relationships with individual level outcomes in 2008. Analyses are carried out by multilevel analysis and multiple regression analysis.

The analyses are currently being carried out. Results and conclusions will be presented at the conference.