13

Drawing on an evaluation test of the mathematics performances of Moroccan secondary school students, the article intends to identify the main significant variables explaining the performance in mathematics. The data gathered from TIMSS 2003 program covers a sample of 1 140 students attending 68 different schools across the country.Using a multilevel procedure, we find evidence of the significant role played by gender related variable as well as wealth indicators (asset ownership) at the level of households. Just few variable related to the school context (school localization, age of the teachers and experience) have a significant effect.The results bring implications at two levels. On one side we are able to provide lessons for improving household strategies and inform public decision makers on how to improve school performance in mathematics. From a methodological view point the study also confirms the appropriateness of using multilevel models to process data that have a hierarchical structure.