• Keine Ergebnisse gefunden

Data, Selection, and Descriptive Evidence

school grade in the same year. However, this is only a problem for grade repetition in a specific track and does not apply for grade repetition in general. The grade repetition rate measurement also may contain errors if students who repeat a grade do so in a different federal state.

The advantage of the data from the German Federal Statistical Office is that it covers every student in Germany, because schools have to provide the German Federal Statistical Office with the information about the number of students and the number of students who repeated a grade. The disadvantage of the data is that the only available socio-demographic information is the students’ gender. The data set is well-suited to give answers about average effects or effects separated according to gender, but effect heterogeneity with respect to the family background cannot be studied.

3.4.2 Exclusion of years, federal states, and school tracks

The introduction of the Oberschule in 2011 in Lower Saxony makes it necessary to constrain the estimation sample. The impact of the introduction of theOberschule on educational paths is not clear. Although other federal states had similar reforms in the last several years, it seems to be impossible to control for this reform in a sensible way.

Because the reform was already discussed in the media by 2010, this study does not consider the years after 2009 (i.e., the school starting cohort 2005 is the latest school starting cohort in the sample).14

I exclude the federal states Brandenburg, Bremen, Mecklenburg-West Pomerania, and Schleswig-Holstein, because these federal states combined the low and medium track. Although from an organizational perspective the combination of the two lower school tracks did not affect the academic track, the reform nevertheless was likely to have influenced the decision to enroll in an academic track. Students who would have picked the medium track in a three-track system have now a higher incentive to enroll in the academic track in these federal states, because the alternative to the academic track is to be in a class with the low-achievement students. I also exclude Saxony-Anhalt, because Saxony-Anhalt simultaneously abolished the OS and made school recommendations binding for the students. Thus, the only two federal states

14The robustness section will show that the results are not sensitive to the inclusion of school cohort 2006 in the analysis.

that changed their tracking system and are still in the sample are Hamburg and Lower Saxony. However, the track change in Hamburg first affected the school starting cohort 2006, and the estimation sample only includes federal states up to the school starting cohort 2005.

This study shows how the abolishment of the OS impacted on the share of indi-viduals in the academic track but not how the shares in the medium and low tracks are affected. The first reason to study only the effect on the share of individuals in the academic track is that the difference between the academic track and the other two tracks is large compared to the difference between the medium and low track (see Subsection 3.3.1). The second reason is that some federal states either do not have a low track or had a reform that affected only the two lowest school tracks (see Pi-opiunik, 2014). As a result, analyzing the effect of the abolishment of the OS on the share of individuals in the medium or low track would require the exclusion of some federal states from the estimations. Since some federal states have already had to be excluded for different reasons, a further reduction would lead to a very small sample of control states. The last reason is that, in recent years, federal states have started to mitigate the difference between the low track and medium track. Extreme examples are that Brandenburg (in 2005), Bremen (in 2005), Mecklenburg-West Pomerania (in 2006), and Schleswig-Holstein (in 2007) combined the low and medium tracks into one school track. Since it is difficult to control for this mitigation between the low and medium tracks, I only study the effect of the abolishment on the share of individuals in the academic track.

3.4.3 Descriptive evidence

Figure 3.3 shows the share of individuals in the academic track after the transition;

the data is presented separately for students from Lower Saxony and the control states for the school starting cohorts 1994 to 2005. It shows that the share of individuals in the academic track is increasing over the school starting cohorts for both groups. In the pre-treatment period, the share is always higher in the control group as compared to Lower Saxony. When Lower Saxony abolishes the OS, the share in Lower Saxony overtakes the share in the control states. The share stays higher in Lower Saxony

until the school starting cohort 2004. Therefore, the graph suggests that the reform increased the likelihood that students would enroll in the academic track. The figure also supports the common trend assumption of the DD approach, because the shares move almost parallel in the pre-treatment period.

The top panel of Figure 3.4 displays the grade repetition rate for grade five in Lower Saxony and the control states for the school starting cohorts from 1990 to 2005.

For the pre-treatment cohorts, the grade repetition rate in grade five is on average about 0.9 percentage points lower in Lower Saxony than in the control states, and the gap increases over the cohorts in the pre-treatment period. For the first treated cohort, the share increases from about 0.5 percent to 2.5 percent in Lower Saxony, although it decreases for the control states. For the subsequent cohorts, the share of repeaters is almost constant in Lower Saxony and still decreases for the control states.

The lower panel describes grade repetition for grade six. Once again, the share in the pre-treatment period is smaller in Lower Saxony and increases strongly for the treated cohorts. In both cases (i.e., grade repetition in grades five and six), the pre-treatment trends of Lower Saxony and the control states deviate. Overall, the graph suggests that the reform increased the risk of grade repetition in grades five and six.

Figure 3.5 displays the grade repetition rate for grades seven, eight, and nine for individuals in the academic track; the data is presented separately for Lower Saxony and the control states. For all three grades, the grade repetition rate in the pre-treatment period is lower in Lower Saxony than in the control states, but overtakes the control states in the first treatment period. The magnitude of the increase between treatment and post-treatment cohorts seems to decrease by grade level. The pre-treatment trends between Lower Saxony and control states are rather similar for grade eight and nine. However, the trends vary for grade seven, when the share of repeaters increases significantly for the school starting cohort 1996. Here as well, the figure provides evidence that the reform increased the likelihood of grade repetition.