Table 1.
Literature overview: causal effects of compulsory schooling on mortality.
| Authors | Country/Data Source | Year/Content of the Reform | Identification Strategy | Main Results |
|---|---|---|---|---|
| Albouy and Lequien [18] | France/ Longitudinal data: Echantillon Demographique Permanent Census data (1968, 1975, 1982, 1990, 1999) Register Data of Deaths from 1968–2005 |
1936 (Zay Reform)/6→7 1967 (Berthoin Reform)/7→9 |
Regression Discontinuity Design on birth cohorts |
Zay Reform: survival till 82 for those surviving until 1968 increased by 6% (Wald-estimate). Berthoin Reform: survival until 52 for those surviving until 1968 increased by 1% (Wald-estimate) Effects statistically insignificant |
| Gathmann et al. [17] | Various European Countries/ Human Mortality Database European Social Service International Social Survey Programme Survey of Health Ageing and Retirement |
19 different Reforms | Regression Discontinuity Design on birth cohorts Meta analysis (for pooled estimate over the 19 reforms) |
Substantial heterogeneity in time and space: Effects probably larger for reforms implemented earlier in the 20th century. Gender differences: no effects for women; reduction of 2.8% in 20-year male mortality from age 18 (reduced form) |
| Van Kippersluis et al. [19] | Netherlands/ Dutch Cross-sectional General Household Survey (1997–2005) Tax Records (1998) Cause-of-Death register (1998–2005) Dutch Municipality Register |
1928/6→7 | Regression Discontinuity Design on date of birth (individual data) |
2%–3% decrease in mortality until the age of 89 for those surviving until the age of 81 (reduced form). Reduced form similar to two-stage least squares estimates as a rise in education between 0.6–1, depending on specification |
| Clark and Royer [10] | England and Wales/ Mortality Data from the Office for National Statistics: All deaths for the years 1970 to 2007 |
1947/8→9 1972/9→10 |
Regression Discontinuity Design | Hardly any evidence for a reduction of mortality; some estimates even with a positive sign |
| Meghir et al. [14] | Sweden Swedish population censuses; all individuals born in Sweden between 1946 and 1957 |
Implemented by municipalities between 1949 and 1962. From 1962 nationwide/ (7 or 8)→9 |
Reduced Form Difference in Difference/IV |
Short-lived gain in expected male years of life from a shift in mortality from ages 45–50 to ages 50–55. Overall life expectancy not significantly affected Heterogeneity with respect to social background |
| Lleras-Muney [13] | U.S./ Census (1960, 1970, 1980) National Health and Nutrition Examination Survey |
1915–1939 Various U.S. states with different extensions |
Difference in Difference/IV Regression Discontinuity Design |
Extension of one year of education decreases 10 year mortality for those surviving until 1960 by 3.6% (instrumental variable (IV)) relative to a baseline mortality of 10%. Estimates challenged by Mazumder [16]: Sensitive to state-specific time trends; effects mainly due to earliest cohorts |
| Lager and Torssander [15] | Sweden/ Swedish population censuses All individuals born in Sweden between 1943 and 1955 |
Implemented by municipalities between 1949 and 1962 From 1962, nationwide/ (7 or 8)→9 |
Reduced Form Difference in Difference/IV |
Overall, all-cause mortality not significantly affected Lower mortality from causes related to education (e.g., cancer and accidents). Socioeconomic heterogeneity with lower mortality among the least educated |