This paper investigates the effect of the timing of tracking on completion of higher education by exploiting unique variation from the Dutch education system. At the age of 12 Dutch students can enrol in tracked schools or in comprehensive schools. The comprehensive schools postpone enrolment into tracked classes by one or two years. OLS- and IV-estimates, using regional variation in the supply of schools as instruments, show that early tracking has a detrimental effect on completion of higher education for students at the margin of the Dutch high and low tracks. The negative effects of early tracking are larger for students with relatively high ability or students with a higher socioeconomic background. In addition, we find no negative effects of comprehensive classes on higher ability students. These results suggest that increasing participation in comprehensive classes would increase graduation from higher education. (Contains 10 tables and 1 figure.)