In the UN Convention on the Rights of the Children as well as in the Swedish national curriculum the children’s right to an environment free from discrimination, victimization and bullying is clearly expressed. For many years Swedish schools have had action programs against bullying. In 2006 the demand came for each school to have an action program not only considering bullying but all sorts of discrimination. This made the question “what is the most effective program against bullying and victimization” even more timely than before. In 2007 the ministry of education gave commission to a group of researches to evaluate the effects of programs used in Swedish schools with the aim to prevent victimization and bullying. In the perspective of educational research during the last two decades, it is obvious that recent demands on methods supporting “evidence-based” knowledge may create problems. In this paper I will discuss some problems when carrying through an evidence-based study on the effects of programs to prevent victimization and bullying. Four systematic reviews and meta-analysis of the effectiveness of anti-bullying programs give a view of common ways of looking at the design to measure effectiveness and how to create evidence about what works in schools struggle against bullying and victimization (Merell et al., 2008; Rigby, 2002; Smith et al., 2004; Ttofi et al., 2008). All four had the same criteria for programs included in the analysis. To prove possible effects they had to have an experimental or quasi-experimental group design. The study had to report data in a statistical format necessary to calculate an effect size and had to be built on self–reported questionnaires.

In our study eight programs are to be evaluated. Schools apply for to participate and they themselves report what program they work with. Each program is represented by four schools. Two of them are just implementing the program. Two have worked with the program for years. The study is built on both quantitative and qualitative data. Between 2008 and 2010 the pupils are going to answer three self-reported questionnaires. We also starts with mapping each school by interviewing principals, teachers and pupils about their experience of bullying, victimization and the work that is done to prevent these efforts. There is a strong push for traditional evidence-based research, meaning an experimental design with a pretest-post-test design, in the field of education (Biesta, 2007). The study of schools interventions against bullying is an example. I want to focus two problems in relation to our study:

- The experience of visiting the schools in our study show the difficulties to measure the effects of implementing anti-bullying programs in Swedish schools only through an experimental design and with self–reported questionnaires. When investigating the implementation of a program means to study an intervention in a complex reality. In none of the schools in our research you can say that it is the program x that gives the effect. Studying effects of anti-bullying programs essentially means to study school-improvement. An example: A school says they are implementing program x. The program is a whole-school antibullying program. The teachers are opponents. The school also works with program y and z. - Knowledge about what is effective is not enough for action. Action in education depends on what is desirable. The question for schools is not simply what program is most effective but rather “what is appropriate for the pupils in these circumstances”