Abstract This paper conducts a meta-analysis of causal studies examining the impact of Early Childhood Education and Care (ECEC) programs on maternal employment in developed countries. Using 981 effect-size estimates from 35 studies, I harmonize employment outcomes into percentage-point changes and assemble 70 study-level characteristics describing program design, subgroups, contexts, and empirical strategies. I apply a battery of linear and non-linear publication-bias tests and use model-averaging methods to account jointly for publication bias and what I term weak design bias, defined by whether studies satisfy core identification diagnostics for RDD, IV, RCT, and DiD designs. Publication bias is modest but non-negligible: correcting only for selective reporting reduces the descriptive mean effect of around 5 percentage points (p.p.) to roughly 1 p.p., yielding insignificant implied intention-to-treat (ITT) effects and average treatment-on-the-treated (ATT) effects of about 4 p.p. Once I also reweight the literature toward designs that meet identification checks, the implied ITT effect rises to about 8 p.p., and the implied ATT effect stabilizes around 10 p.p. The corrected ATT effects are particularly pronounced for child care programs, delivered by public providers, and implemented in high-employment settings. Overall, the results suggest that modern ECEC expansions generate sizeable employment gains for already attached mothers facing binding care constraints. They also provide bias-corrected benchmarks for evaluating ECEC reforms and their contribution to mitigating child penalties and gender gaps in labour-market outcomes.
Pablo Brugarolas (Thu,) studied this question.