Teachers’ versus parental choice and the tracking distribution of students: a natural experiment

Abstract

This article studies the secondary school track choice and considers to what extent parents’ and teachers’ assessment of students diverge. We take advantage of a reform in the German state of North Rhine-Westphalia (NRW) in 2006. The reform replaced parents’ choice about their children’s secondary school type by a binding teacher recommendation. Our data comprise class-level information on all public primary schools in the state. We find that teachers tend to recommend higher school types than parents. However, more precise analysis shows that this effect can be limited to districts with above average proportion of immigrants.

Keywords:

• educational tracking
• selective school systems
• educational inequalities

JEL Classification:

• I21
• I24
• I28
• J24

Notes

¹ To examine closely whether parents’ or teachers’ assessment of the student’s potential is more appropriate, one needs information on student performance after the tracking decision. We study this question in a subsequent paper.

² For an overview see Meier and Schütz (2008).

³ Ochsen (2011) argues that parental deviations from teacher recommendations have a sizeable and significant positive effect on children’s educational attainment in Germany.

⁴ Dollmann (2011) also refers to the reform of 2006 in NRW and considers one cohort before and one after the reform. In contrast to our analysis, he does not evaluate effects on the total tracking distribution. Furthermore, his two-year sample is restricted to the city of Cologne and immigrants are excluded from the dataset. Therefore, he does not give evidence about reform effects on this population group.

⁵ Kleine et al. (2010) compare two German federal states with differing regulations.

⁶ Schnepf (2002) argues that although immigrants are predominantly selected to lower academic school tracks, they do not face educational inequalities if their socio-economic background and measured ability is similar to that of German nationals.

⁷ Lüdemann and Schwerdt (2013) do not find evidence for ethnic discrimination by teachers per se. They state that inequalities at the transition to secondary school are due to the socio-economic background in general that affects track recommendations, even conditional on student achievement.

⁸ Other tracking countries include e.g. Austria, Hungary and Slovakia (c.f. Hanushek and Wössmann (2006)).

⁹ A more detailed description of the German education system can be found in Winkelmann (1996).

¹⁰ c.f. Second Law of the Reform of School Law of the State of North Rhine-Westphalia, 27 June 2006: $\S$ 11, Section IV.

¹¹ We chose those years because IT-NRW started selecting data explicitly on teacher recommendations in 2009. For the pre-reform period, we only have information on parental track choice, i.e. the actual school track.

¹² Although the reform was enacted in 2006, the cohort starting secondary school in August 2007 was the first one to be affected.

¹³ Unfortunately, our data only reveal information about the school track recommendations of comprehensive school students for 2008/09. We thus cannot determine the reform effect on the composition of students attending comprehensive schools.

¹⁴ In Germany, grade retention occurs when the student’s performance is not sufficient to be promoted to the next grade. It can be mandatory or voluntarily chosen depending on the student’s school grades.

¹⁵ In fact, Schneider et al. (2012) analyse the introduction of free primary school choice in NRW in 2008. If the choice is practised primarily by socio-economically advantaged families, this change in primary school law may cause biases regarding the tracking decision. However, they do not find any effects regarding the amount of school choice and segregation. Besides the effects of the reform not being significant, we can rely on the fact that we analyse students on the verge of secondary school: the primary school reform in 2008 is more likely to affect the initial primary school choice (first grade) than to provoke a transfer to another school one year before secondary school transition (fourth grade).

¹⁶ If we were to use the number of rows as number of observations, we would not receive the proper number of students but the number of primary schools times three or four (# of school types). Instead, by using frequency weights with the actual number of switching students as weights variable, the proper number of students is used in the regression.

¹⁷ Since it causes problems of multicollinearity in combination with district fixed effects, unemployment rates are left out.

¹⁸ The question of the effects of school inputs on examination results has been debated for several decades with mixed results. However, many of the early studies suffer from endogeneity problems of the variable school inputs.

¹⁹ Not shown in the table.

²⁰ The estimations are available on request.

²¹ These efficiency questions are left for future research.

²² In fact, Ai and Norton (2003) point out that 72 articles published between 1980 and 2000 used interaction terms in nonlinear models; but none of them interpreted the coefficient on the interaction term correctly.

²³ Results are available on request.

²⁴ i.e. fifth grade students attending private high schools over the total number of fifth grade students in high school; source: state-level data by the Bureau of Statistics of NRW.

²⁵ c.f. Jackson et al. (2007).