International research on PISA data has shown that early tracking has a negative effect on the equity in achievement without having any obvious effect on efficiency (e.g. Hanushek & Wössmann, 2005). The Flemish PISA20003 results both confirmed and contradicted these findings. The tracked Flemish education system indeed produces a very high level of inequality (with a gap of 347 score points on the mathematics performance scale between the highest and lowest performing students), but at the same time the Flemish mean performance in mathematical literacy in 2003 was the highest of all participating countries. Furthermore the group of highest performing students was very large (34,3% of the Flemish PISA students performed at the two highest levels for mathematical literacy). This implied that in Flanders for mathematics the gain from tracking by the more able was higher that the loss by the less able, resulting in a high efficiency for this domain. PISA2006 provides the first opportunity to look at trends in math performance of countries. For Flanders this first indication isn’t very promising: the average mathematics performance in PISA2006 is significantly lower than that of 2003, essentially because of a lower performance at the higher end of the distribution. Our better performing students did worse in PISA2006 while the performance among the lower end of the scale remained broadly unchanged. In this paper we will try to explain this shift in student performance through secondary analysis on the PISA2003 and PISA2006 datasets. First we’ll have an in-depth look at the distribution of student’s math performances in both cycles. As a second step we’ll bring the features of the tracked Flemish education system (types of education, fields of study) into account. Finally we will try to explain the Flemish decline in mathematical performance by some policy decisions on the (re)grouping of students within tracks. Research question: How can the trend in Flemish mean performance in mathematical literacy be explained? (What effects do policy decisions concerning the (re)grouping of students within tracks have on the mathematical performance of Flemish students?)