The Long-Term Effects of Early Track Choice
Dustmann C, Puhani PA, Schönberg U · 2017
grade Bnatural-experimentindependentunreplicatednumbers spot-checked
Sample
West German birth cohorts 1961-1976 (foreign citizens excluded); social-security records 1975-2006 (≥84% coverage of the cohorts), plus German Microcensus 1976-1987, a Bavaria/Hesse School Census, and the 1987 Census. Wage analyses restricted to men aged 30+.
Population
West German students tracked at ~age 10 into high / medium / low tracks, followed into the adult labour market to age 42+.
Design
Exploits the 1 July school-entry cut-off: children born just after it are older in class and more likely to enter a more advanced track. The paper reports reduced-form effects of birth date on adult outcomes, and converts them to Two-Sample-2SLS by dividing by the first stage. First stage is strong but its size depends on the data source — 3.9 pp in the Microcensus (coarse birth-month data, acknowledged as an underestimate) vs 9.4 pp in the School Census (5.2 pp more likely high track, 4.2 pp less likely low track). Because the TS-2SLS scaling uses the smaller 3.9 pp figure, the reported point estimates and CIs are, if anything, exaggerated — the true effect is even closer to zero. Estimates a LATE for students at the margin between two tracks. The archived DOI is the 2017 Economic Journal version; the full text read is the November 2015 working paper.
Key findings
Where tracks are permeable, being placed on the other side of the track line at age 10 has no long-run consequence for wages, employment, unemployment, occupational choice or final qualification — despite a strong first stage and despite German tracks differing sharply in curriculum, peers and teacher quality. The paper documents the mechanism directly: marginal students up- and downgrade after grade 9/10, and the specialised high-track schools that let medium-track graduates reach university erase the initial gap completely. The load-bearing counterweight to the detracking-reform literature — the damaging ingredient in early tracking is irreversibility, not differentiated instruction. The authors go further: marginal students can make up for 5-6 years in a less challenging environment, provided they later gain access to high-quality schooling.
Genetic confound
Birth-date/school-entry-cut-off design; ability is held constant at the margin by design. Microcensus comparison of July- vs June-born children finds no significant differences in parental characteristics. The authors are explicit that the large observed gaps between tracks in the German Socio-Economic Panel are ability-based selection, not treatment.
Replication notes
The essential scope limit on the other tracking results: harm comes from rigidity, not from grouping as such. `unreplicated` asserts no direct replication of this birth-date design in another rigid-tracking system; related reform studies (Finland, Sweden, France) test different variation.
DOI / URL
Effects
| Outcome | Metric | Value | Measure | Timing | Vs | Horizon | Class |
|---|---|---|---|---|---|---|---|
| Adult log wages (men 30+), marginal track assignment | reduced form / TS-2SLS | reduced form experience-adjusted -0.1% to +0.03% across six specifications, all insignificant; TS-2SLS -2.5% to +0.9%, 95% CIs [-0.074, 0.024] and [-0.050, 0.068]. Raw (unadjusted) wages are lower for July-born, but that is a labour-market-experience artefact that disappears by the late 30s; restricting to men 40+ where wage-experience profiles are flat gives small insignificant differentials on both raw and adjusted wages. | administrative | adult labour market, ages 30-42+ | business-as-usual | adulthood | attainment |
| Share of days unemployed, share of days worked, blue- vs white-collar occupation | reduced form / TS-2SLS | basically zero in all specifications. TS-2SLS for share of days worked since labour-market entry ranges +0.016 (SE 0.016) to -0.005 (0.019). Effect on probability of a white-collar occupation close to zero — despite the low track historically feeding blue-collar and the medium track white-collar apprenticeships. Similar results including women. | administrative | adult labour market, ages 30+ | business-as-usual | adulthood | attainment |
| Highest degree completed (no post-secondary / apprenticeship / college or university) | reduced form | no discontinuity at the school-entry cut-off on any of the three; if anything being born in July (older, more likely high track at 10) has a slightly NEGATIVE effect on college or university completion, consistent with the documented downgrading | administrative | final attainment | business-as-usual | adulthood | attainment |
| The mechanism — later track mobility | percentage points | the initial gap is undone in stages. July-born are 5.2 pp more likely to attend the high track at 14 but only 2.0 pp more likely to graduate from the general high track at 22; once specialised high-track schools are included the difference disappears ENTIRELY. Low-to-medium: 1.9 pp more likely to attend at least the medium track, only 0.3 pp more likely to graduate from at least a medium track. Downgrading also occurs after high-track graduation — 0.7 pp more likely not to enrol in college. Roughly 36% attend the high track in middle school but ~52% eventually graduate from a general or specialised high track. | administrative | ages 15-22 | business-as-usual | over-2yr | attainment |
Cited by
- Ability grouping and tracking — four practices, four verdictsmoderate supportconf: highgc: low