Although all European countries allocate their students to different educational tracks at some stage of secondary education, in some countries, the decision about which track to attend has to be made at a relatively early stage of the education process, e.g. at the age of ten in Austria and Germany. Proponents of educational tracking emphasize that all students benefit from homogenous classrooms, which result from the placement of students into differing-ability schools or classes. While proponents stress the positive effects of tracking on overall efficiency, opponents' central arguments are based on efficiency and equity considerations. They claim that early tracking systematically discriminates students with less favorable parental background (Dustmann, 2004), increases educational inequality and tends to reduce mean performance (Hanushek and Wößmann, 2006). Given that peer effects are nonlinear, heterogeneous classrooms may also be more efficient as they lead to higher aggregate achievement. Apart from that, early tracking may be inefficient due to the following reason: Ideally, track choice should be based on the ability of a student. Actually, a student's ability is unobserved and track choice is based on a measure of ability, i.e. educational achievement in primary school. We argue that at the age of ten, prior achievement is not a precise measure of actual ability for all students and therefore, track choice does not only depend on ability but also on other factors that determine achievement in primary school, e.g. the relative age of a student. Bedard and Dhuey (2006) show for a number of OECD countries that relatively younger students perform significantly worse than their older peers in grades 4 and 8. Since school enrollment is based on a certain cutoff date, the relative age of a student within grade is determined by his or her birth month. If track choice is influenced by prior achievement, age-related achievement differences may translate into age-related differences in track choice. Assuming that birth month is exogenous with respect to ability, track choice is indeed at least partly determined by a random factor. In this paper we use register data for a major Austrian city for the period 1984 to 2006 to estimate the causal effect of relative age on track choice after primary school ("Hauptschule" vs. "Gymnasium"). We propose that relatively younger students are less likely to choose the academic track ("Gymnasium") than their older peers in the same grade.

We estimate the causal effect of relative age on track choice using register data for Linz from 1984 to 2006. Since school enrollment is based on a certain cutoff date, the relative age of a student within grade is determined by his or her birth month. However, due to grade retention and the possibility of late or early enrollment, the observed relative age at track choice is endogenous. Therefore, we estimate our binary regression model (Probit) using the difference between birth month and the cutoff date as an instrument (IV) for observed relative age. Our identification strategy is based on the exogenous variation in birth month between students within a grade.

We find a strong and significant positive effect of relative age on track choice for grades 5-8. According to our estimates, being eleven months older increases the probability to attend the academic track by 15-17 percentage points. After grade 8, Austrian students have the possibility to revise their track choice. There are four different tracks available, two of which are academic tracks (higher general and higher vocational) whereas the other two are non-academic tracks (pre-vocational and intermediate vocational). We use additional data from PISA 2000 and 2003 to show that the relative age effect does not disappear due to the possibility of track revision after grade 8. Our results suggest that for students who comply with the cutoff-rule track choice is influenced by birth month and therefore, tracking is not only based on academic ability but also on a random factor.