In this study, we analyse the predictive power of home and school environment-related factors for determining pupils' aggression. The multiple regression analyses are performed for fourth- and eighth-grade pupils based on the Trends in Mathematics and Science Study (TIMSS) 2007 (N = 8394) and TIMSS 2011 (N = 9415) databases for Slovenia. At the national level, the "Lestvica agresivnosti" aggression scale was administered in both TIMSS cycles. For home environment variables, we included those related to socio-economic status, pupils' educational aspirations, parental activities with their children and pupils' free time activities. The results show that the variables related to socio-economic status, spare time activities and parental activities are significant predictors. The results differ in both analysed data-sets. For school environment variables, we include those related to the school climate, pupils' attitudes towards school and school subjects and pupils' achievement in mathematics. We find that the variables related to school climate and students' self-confidence are significant predictors. These results are stable in both years. The predictive power of the school characteristics model (including only the school environment variables) is larger (based on the proportion of explained variance) compared with the home characteristics model. The hierarchical linear model of data from 2007 to 2011 shows small differences in aggression between schools. The inclusion of two data cycles collected in two time- periods allows us to observe changes in aggression predictors over time. Practical implications are finally included.