Publications
Better Together? Heterogeneous Effects of Tracking on Student Achievement. (2021). The Economic Journal, Volume 131, Issue 635, pp. 1269–1307. [Ungated version here.] [Replication package here.]
Abstract: I study the effects of early between-school ability tracking on student achievement. For identification I exploit institutional differences between German federal states: in all states, about 40% of students transition to separate academic-track schools after comprehensive primary school. Depending on the state, the remaining student body is either directly tracked between two additional school types or taught comprehensively for another two years. Comparing these students before and after tracking in a triple-differences framework, I find evidence for positive effects of prolonged comprehensive schooling on mathematics and reading scores. These are almost entirely driven by large effects for low-achievers, whereas for high-achievers effects are null. Early and rigid forms of tracking can thus impair both equity and efficiency of school systems.
TLDR: I summarize my main findings in this Twitter thread.
For this paper I was awarded the NEPS Publication Award 2021.
No Evidence that Strict Educational Tracking Improves Student Performance through Classroom Homogeneity. A Critical Reanalysis of Esser and Seuring (2020). (2021). (joint with Jan Paul Heisig, WZB Berlin & FU Berlin). Zeitschrift für Soziologie, Volume 51, Issue 1, pp. 99–111. [Replication package here.]
Abstract: In a recent contribution to this journal, Esser and Seuring (2020) draw on data from the National Educational Panel Study to attack the widespread view that tracking in lower secondary education exacerbates inequalities in student outcomes without improving average student performance. Exploiting variation in the strictness of tracking across 13 of the 16 German federal states (e.g., whether teacher recommendations are binding), Esser and Seuring claim to demonstrate that stricter tracking after grade 4 results in better performance in grade 7 and that this can be attributed to the greater homogeneity of classrooms under strict tracking. We show these conclusions to be untenable: Esser and Seuring’s measures of classroom composition are highly dubious because the number of observed students is very small for many classrooms. Even when we adopt their classroom composition measures, simple corrections and extensions of their analysis reveal that there is no meaningful evidence for a positive relationship between classroom homogeneity and student achievement—the channel supposed to mediate the alleged positive effect of strict tracking. We go on to show that students from more strictly tracking states perform better already at the start of tracking (grade 5), which casts further doubt on the alleged positive effect of strict tracking on learning progress and leaves selection or anticipation effects as more plausible explanations. On a conceptual level, we emphasize that Esser and Seuring’s analysis is limited to states that implement different forms of early tracking and cannot inform us about the relative performance of comprehensive and tracked systems that is the focus of most prior research.
De-Tracking at the Margin: How Alternative Secondary Education Pathways Affect Student Attainment. (2025). (joint with Camilla Borgna, University of Turin & Collegio Carlo Alberto). Economics of Education Review, Volume 104. [Open Access] [Replication package here.]
Abstract: This paper estimates how marginal increases in the flexibility of between-school tracking affect student attainment by exploiting the addition of non-selective ‘comprehensive schools’ and hybrid ‘vocational high schools’ to Germany’s tracked school system. These schools opened up alternative pathways to the university-entrance certificate, which traditionally could only be obtained at academic-track schools. We use administrative records to compile a county-level panel of school supply and attainment for 13 cohorts between 1995 and 2007. Cross-sectionally, the supplies of all three school types awarding the university-entrance certificate correlate positively with its attainment. However, for academic-track and comprehensive schools this association is not robust to the inclusion of regional controls, suggesting that it reflects regional differences in educational demand rather than supply-side effects. For vocational high schools, in contrast, we find robust evidence for positive attainment effects not only in cross-sectional and two-way fixed-effects panel regressions, but also in an event-study design that exploits the quasi-random timing of new school openings. Likely reasons for their success are that they lower the (perceived) costs of educational upgrading for late-bloomers, and their hybrid curriculum, which may retain students in general schooling who would otherwise enter vocational training.
TLDR: I summarize my main findings in this Bluesky thread.
Working Papers
On Track to Success? Returns to Vocational Education against different Alternatives. (2025). (joint with Gugliemo Ventura, LSE & University of Barcelona). Submitted.
Previous version published as: LSE-CVER Discussion Paper No. 38.
Abstract: Many countries consider expanding vocational curricula in secondary education to boost skills and labour market outcomes among non-university-bound students. However, critics fear this could divert other students from more profitable academic education. We study labour market returns to vocational education in England, where until recently students chose between a vocational track, an academic track and quitting education at age 16. Identification is challenging because self-selection is strong and because students' next-best alternatives are unknown. Against this backdrop, we leverage multiple instrumental variables to estimate margin-specific treatment effects, i.e., causal returns to vocational education for students at the margin with academic education and, separately, for students at the margin with quitting education. Identification comes from variation in distance to the nearest vocational provider conditional on distance to the nearest academic provider (and vice-versa), while controlling for granular student, school and neighbourhood characteristics. The analysis is based on population-wide administrative education data linked to tax records. We find that the vast majority of marginal vocational students are indifferent between vocational and academic education. For them, vocational enrolment substantially decreases earnings at age 30. This earnings penalty grows with age and is due to wages, not employment. However, consistent with comparative advantage, the penalty is smaller for students with higher revealed preferences for the vocational track. For the few students at the margin with no further education, we find merely tentative evidence of increased employment and earnings from vocational enrolment.
TLDR: I summarize our main findings in this Twitter thread.
For this paper I was awarded the BeNA Innovative Research Award 2020 and it won the ifo Munich's Big Data research competition.
Work in Progress
The mechanisms underlying rank effects in education: self-concept vs. sorting. (joint with Michael Becker, TU Dortmund & DIPF, Richard Murphy, UT Austin, and Felix Weinhardt, EU Viadrina & DIW Berlin).
The role of post-compulsory education for early career wage growth and firm sorting. (joint with Guglielmo Ventura, LSE, UCL & CVER).
Heterogeneous effects of ethnic classroom composition on student achievement. (joint with Pia Schilling, U Konstanz).
Policy Publications
Längeres gemeinsames Lernen macht einen Unterschied. WZBrief Bildung 40. WZB Berlin Social Science Center.
For this policy brief I was awarded the Friends of the WZB Award 2020.
Für besseren Datenzugang im Bereich Bildung. Stellungnahme aus dem Verein für Socialpolitik. (2023). (joint with Hertweck, F., Isphording, I. E., Schneider, K. & Spieß, C. K.) RatSWD Working Paper 282/2023. Rat für Sozial- und Wirtschaftsdaten (RatSWD).
Abstract: Die Stellungnahme der Arbeitsgruppe „Bildungsdaten“ zeigt die gravierenden Datenmängel im Bereich der Verfügbarkeit und Verknüpfbarkeit von Bildungsdaten in Deutschland auf. Sie fordert die Umsetzung eines Bildungsverlaufsregisters über alle Stufen formaler Bildung sowie den regelhaften Zugang der Forschung zu diesen Daten. International würde Deutschland mit einem derartig verknüpfbaren Bildungsverlaufsregister an die Standards anderer europäischer Staaten aufschließen.
Bildungsdaten: Datenlücken durch ein Bildungsverlaufsregister schließen. (2023). (joint with Hertweck, F., Isphording, I. E., Schneider, K. & Spieß, C. K.) Wirtschaftsdienst, 103(11), 733-736.
Abstract: Evidence-based education policy needs comprehensive data. Education economists of the Verein für Socialpolitik call for the implementation of a longitudinal education register in Germany to understand the complex interrelationships of education systems. The education register should cover all levels of formal education and include data on student performance in order to close a serious data gap in Germany. The statement also calls for regular access to this data for independent research. Once implemented, the education register should enable linking with other data sources, for example from the labour market, tax and health records, in order to be able to analyse the long-term effects of education. By establishing a longitudinal education register, Germany would finally catch up with established standards of other European countries, creating a resource where education register data not only exists, but are available for researchers and can be linked with other data.