In the field of educational effectiveness research, school effects are generally studied in the short term (i.e. during the same phase of schooling). The aim of this study is to investigate long-term primary school effects on students' achievement in mathematics at the end of secondary education. We also investigate which primary school characteristics are of importance in the long term. Data from the longitudinal SiBO project, in which a cohort of 6,000 Flemish pupils were intensively followed from kindergarten to grade 7, was used. At the age of 17, the same cohort participated in follow-up data collection. Cross-classified multilevel models showed small continuing effects of primary school on the mathematics achievement of students (i.e. over and above what had been reached at the end of primary education). No long-term effect was found of the proportion of high-risk students at primary school. Students coming from a primary school with a higher effectiveness obtained higher mathematics results at age 17, but when the mathematics achievement of students at the end of primary school was taken into account, this effect disappeared. We also observed that students coming from Catholic primary schools performed better in mathematics at age 17 compared with students coming from public schools. The implications of the findings are discussed.