Wie viele Schulen braucht ein fairer Vergleich? Sozialindex-basierte Adjustierungsstrategien im Vergleich

Large-scale educational competence tests provide teachers and school management with information regarding their students’ performance. For a so-called fair comparison, a school’s mean score can be compared with the mean score of schools with a similar social composition. Similarity is usually determined by a social school index that is based on contextual factors that lie beyond the teachers’ influence. The present study compares two adjustment strategies along with three types of indices with regard to their fairness in terms of explained performance variance (R2). Each combination is modelled with 24 different comparison group sizes in order to find out how many schools are needed for a fair comparison. The analysis is based on two cohorts with 2330 elementary schools and 1261 secondary school, respectively. A socio-cultural index performs better than a socio-economic index based on georeferenced data and also better than a combined socio-cultural/socio-economic index. Using comparison groups based on index quantile splits yields slightly higher R2-values than using schools right above and below the target school on the index. If possible, the number of comparison groups should at least 20 with each group containing at least 20 schools. The results suggest that a reasonably fair comparison can be obtained efficiently while costing little, remaining robust across grades and subjects, and being comprehensible for teachers and school management personal.