General Education versus Vocational Training: Evidence from an Economy in Transition

Not obtained at secondary depth; 1992 and 2002 Romanian censuses plus 1995-2000 household surveys

Romanian cohorts around the 1973 education reform, observed 1992-2002 — well into adulthood and after a wholesale economic transition.

Regression discontinuity on the 1973 reform, which moved a large share of students from vocational to general education while holding average years of schooling fixed — an unusually clean test because it varies education TYPE without varying QUANTITY. It is also the best available setting for the "general skills adapt better to technological change" hypothesis, since Romania's transition from communism is about the largest institutional shock a cohort can experience. Weaknesses: cohort-level identification, so compositional shifts within cohorts are possible, and the setting is one country's very specific transition. Read at secondary depth.

Men in reform-affected cohorts were significantly less likely to work in manual or craft occupations. But — in the authors' own framing, in contrast to cross-sectional findings — there was NO difference in labour-force participation or earnings between affected and unaffected cohorts. They conclude that differences in labour-market returns between vocational and general graduates are largely driven by selection. The adaptability hypothesis was tested in the environment most favourable to it and did not hold.

Low for the causal estimate — cohort-level exposure to a national reform. The paper's contribution is precisely to show how large the selection confound is in the cross-sectional literature it contradicts.

Its central conclusion — that the observed general-over-vocational earnings advantage is selection rather than causation — is independently reproduced by Žilić (Croatia, 2018) and by Silliman & Virtanen (Finland, 2022).