General Education, Vocational Education, and Labor-Market Outcomes over the Lifecycle

15,218 males with completed secondary or first-stage tertiary education across 18 IALS countries; 11 "vocational countries" in the main sample; earnings sample only 5,885 (10 countries) and 1,964 (3 apprenticeship countries); German Microcensus 85,680

Males aged 16-65 in the International Adult Literacy Survey fielded 1994-1998, plus the German Microcensus 2006 and Austrian administrative plant-closure data.

This is the paper that made the lifecycle-decay claim, and its design cannot deliver it. There is NO exogenous variation in education type: the "difference-in-differences" differences employment across AGE within education type, in cross-sectional data. Age and cohort are perfectly collinear in a single cross-section, so a cohort trend in who chose vocational, or in what vocational meant, reproduces the result exactly. The authors defend against this by showing that selection correlates (literacy, mother's education) do not vary with age and with an Oster coefficient-of-proportionality argument — informative, but not identification. The one genuinely exogenous piece is the Austrian plant-closure analysis, and it proxies "vocational" by BLUE-COLLAR occupation, which is a different variable. Earnings samples are small (n = 1,964 for the apprenticeship countries), and the crossover ages are, in the authors' own words, quite imprecise and varying across specifications.

Pooling 11 vocational countries, individuals with general education are 6.9 percentage points LESS likely to be employed at the normalised age of 16, and the gap closes by 2.1 points per decade, implying a crossover at roughly age 49. In the German Microcensus 2006 the crossover comes at age 43 and by 65 the general-educated are 11.5 points MORE likely to be employed. Income shows the same age pattern with an earlier crossover, flattening around 50. The pattern is sharpest in the three apprenticeship countries (Denmark, Germany, Switzerland). Lifetime-earnings balance favours vocational in slower-growing Switzerland but general in faster-growing Denmark and Germany. In the Austrian plant closures, after displacement blue-collar workers' relative employment advantage falls 0.7 points per year of age at displacement (coefficient -0.0069, SE 0.001) and is significantly negative above 50.

HIGH. This is the archive's premise operating at full strength. The comparison is between people who CHOSE different tracks; track choice correlates with literacy and maternal education in the paper's own appendix; and three separate regression-discontinuity studies conclude the observed vocational-general differential is selection. Any verdict resting on this design must be capped at low confidence for causal claims.

The employment reversal replicates in Hampf & Woessmann (2017), which is a same-author extension to PIAAC rather than an independent test, and in Forster, Bol & van de Werfhorst (2016), an independent team on PIAAC — but Forster et al. put the crossover roughly ten years later (about 55 for men) AND find, contrary to this paper's central mechanism, that the reversal does NOT vary with how occupationally specific a country's vocational system is. Meanwhile every design with exogenous variation in track — Silliman & Virtanen (Finland RD), Malamud & Pop-Eleches (Romania), Žilić (Croatia), Bertrand et al. (Norway) — fails to find the decay.