BACKGROUND: Compulsory schooling laws (CSLs) are commonly used as instrumental variables (IVs) to estimate the causal effect of schooling on later-life health, which avoids confounding by eliminating individual characteristics. A recent systematic review, Hamad et al. (2018), identified 89 articles evaluating the health effects of education using CSL based IVs. IV analyses sacrifice statistical power for reduced confounding bias, so meta-analyses of these results are essential. We conducted a meta-analysis of studies using CSLs to estimate the effects of years of schooling on late-life cognition.
METHODS: From the articles identified by Hamad’s systematic review, we restricted our analysis to 6 studies evaluating the effects education on cognitive domains (12 effect estimates for global cognition, memory, or executive function based on 117,988 cognitive assessments). We extracted and normalized coefficient estimates to conduct a random-effects meta-analysis, combining IV-based estimates of the effect of an additional year of schooling on late life cognition, using a robust variance estimation to account for multiple effect estimates in the same studies. We evaluated whether effect estimates differed by outcome domain or gender.
RESULTS: The studies were from the U.S., Europe, and China. The meta-analyzed effect estimate for each additional year of schooling on cognition (pooling all domains) was 0.21 (95% CI: 0.13, 0.28) (Figure). There was substantial between-study heterogeneity (I2=72%). Effects for memory (0.28; 95% CI: 0.18, .38) were larger than for executive function (0.09, 95% CI: 0.01, 0.16). We found no evidence of differences by gender or when using a fixed effects meta-analysis model.
CONCLUSION: Although individual IV study estimates are often imprecise, our meta-analysis of results from CSL studies indicates a substantial effect of additional educational attainment on subsequent cognitive outcomes.
