Educational Reform, Ability, and Family Background
Meghir C, Palme M · 2005
grade Bnatural-experimentindependentreplicatednumbers spot-checked
Sample
19,550 individuals in the pooled attainment analysis (1948 cohort n = 10,638; 1953 cohort n = 9,105), drawn from the Gothenburg Individual Statistics surveys; earnings observed annually 1985-1996
Population
Swedish children born 1948 and 1953, surveyed in grade 6 with three IQ tests and subject grades, linked to the 1990 education register and 1985-96 tax records. 83% had "unskilled" fathers (compulsory schooling only); 17% had fathers with more than compulsory schooling.
Design
The Swedish comprehensive-school reform (a) raised compulsory schooling to 9 years, (b) abolished grade-based selection into academic vs non-academic streams after grade 6 (age 12-13), and (c) imposed a national curriculum. It was phased in municipality by municipality from 1949, so the 1948 and 1953 cohorts were assigned to different systems in "switching" municipalities. Difference-in-differences across cohorts and municipalities, with municipality fixed effects, grade-6 test scores and grades as covariates, standard errors clustered by municipality. Municipalities were NOT chosen randomly — the National School Board picked from applicants to form a set it judged representative — so this is a strong quasi-experiment, not a lottery. Robust to using either comparison group alone, to excluding the three big cities, to excluding movers, and to IV on birth municipality. The archived DOI is the 2005 AER version; the full text read is IFS WP04/10 (April 2004).
Key findings
Abolishing early selection raised attainment (+0.298 years overall, +0.406 for children of unskilled fathers) and raised adult earnings by 3.4% for children of unskilled fathers, while children of educated fathers lost 5.5% with no attainment change — the authors attribute the loss to a dilution of quality at the top once selective junior secondary schools were abolished. The ability interaction is about WHERE the attainment gain landed, not how big the earnings gain was: low-ability children of unskilled fathers were pulled up to the new compulsory level, high-ability children of unskilled fathers were pushed beyond it, and the earnings difference between the two ability groups (4.14% vs 2.81%) is not statistically significant. Critically, the reform was NOT a pure redistribution — the paper's own conclusion is that "the reform improved on average both the educational attainment and the earnings of a large part of the population" and that the overall effect was positive.
Genetic confound
Reform timing is plausibly exogenous to family genotype (municipalities selected administratively, not by parents), and ability is measured in grade 6 before the reform bit, so the ability-by-background heterogeneity is estimated rather than assumed. The finding that able children of unskilled fathers were the ones who moved BEYOND compulsory schooling is what a hereditarian lens predicts if early selection was mis-sorting on background rather than ability. Note the counterweight — pre-reform, IQ differences between switching and non-switching municipalities were small and insignificant.
Replication notes
Same direction as the Finnish (Pekkarinen et al. 2009) and French (Canaan 2020) detracking reforms. Not established by this paper.
Effects
| Outcome | Metric | Value | Measure | Timing | Vs | Horizon | Class |
|---|---|---|---|---|---|---|---|
| Educational attainment, whole population | years of education / percentage points | +0.298 years (SE 0.075); +8.57 pp (1.68) reaching the new compulsory level and +2.63 pp (1.14) going BEYOND it. All highly significant. Baseline in non-reform areas: 11.18 years. | administrative | final attainment | business-as-usual | adulthood | attainment |
| Educational attainment, children of low-educated (unskilled) fathers | years of education / percentage points | +0.406 years (0.070) — "the entire effect is due to" this group. The MARGIN differs by ability: low-ability children of unskilled fathers gained +16.82 pp (2.73) at the new compulsory level but only +1.41 pp (ns) beyond it, while high-ability children of unskilled fathers gained +6.19 pp (2.33) BEYOND the compulsory level. In total years the low-ability group actually gained slightly more (0.427 vs 0.358). | administrative | final attainment | business-as-usual | adulthood | attainment |
| Educational attainment, children of educated fathers | years of education | -0.132 years (0.123) — not significant; no meaningful change in attainment for this group | administrative | final attainment | business-as-usual | adulthood | attainment |
| Adult earnings, whole population | percent effect on annual earnings | +1.45% (SE 0.85), significant only at the 8.8% level. Positive but marginal — the paper's own summary is "The overall effect of the reform was however positive", so this is NOT a pure redistribution. | administrative | annual earnings 1985-1996 (ages ~32-48) | business-as-usual | adulthood | attainment |
| Adult earnings, children of unskilled fathers | percent effect on annual earnings | +3.39% (0.90), highly significant (men +3.09, women +3.80). By ability: high ability +4.14 (1.25) vs low ability +2.81 (1.25) — the high-ability point estimate is larger but the DIFFERENCE IS NOT STATISTICALLY SIGNIFICANT, and for men the two are indistinguishable. Implied upper-bound returns to education if the whole gain ran through years of schooling: 12.3% (high ability), 7.3% (low ability), 8.5% overall. | administrative | annual earnings 1985-1996 | business-as-usual | adulthood | attainment |
| Adult earnings, children of educated fathers (17% of the population) | percent effect on annual earnings | -5.54% (1.87), significant, and negative for both men (-7.53) and women (-3.95). Robust across comparison groups (-5.88, -5.70, -5.98) though attenuated to -3.15 (2.70) excluding movers and wildly imprecise under IV (-19.55, 13.03). Occurs DESPITE no significant change in this group's attainment — the authors read it as a quality dilution from abolishing selective junior secondary schools. | administrative | annual earnings 1985-1996 | business-as-usual | adulthood | attainment |
Cited by
- Ability grouping and tracking — four practices, four verdictsmoderate supportconf: highgc: low