Chapter in its early stage can be viewed here. For presentation slides, click here.
Abstract: Japan has among the most generous parental leave statutes in the OECD, yet one of the widest motherhood employment gaps. Using an event-study design applied to 662 first births in the JHPS and KHPS panels (2004-2022), I show that the penalty is strongly stratified by labour-market tier. Employment falls 13.2 percentage points at childbirth, with a clean near-lead null at t = −2. Among women employed be- fore birth, non-regular contract status predicts non-employment at t = 0 with an odds ratio of 7.50 (p < 0.001), while small-firm effects are positive but imprecise. Post-birth ”recovery” is largely an extensive-margin illusion - returning mothers shift from full-time to part-time work, and conditional hours, earnings, and implied hourly wages remain durably below pre-birth levels. Government training subsidies designed to support re-entry are effectively unused, with fewer than 3% of moth- ers reporting utilisation at any post-birth horizon. These patterns are consistent with weaker effective access to job-continuity protections in non-regular and small- firm segments; statutory entitlements alone appear insufficient where labour-market dualism channels mothers into employment tiers with limited institutional support.
Chapter in its early stage can be viewed here, and a quick overview of main results here.
Abstract: How noncognitive skills are measured shapes whether they appear as complements or substitutes to cognition in educational production. Using data from the Growing Up in Ireland longitudinal study, I estimate translog production functions for Maths and English achievement among Irish secondary students and show that the estimated degree of skill complementarity depends on the noncognitive instrument. When the noncognitive input is a school-proximal behavioural measure (SDQ Focused Behaviour), elasticity-of-substitution estimates are consistently below one. Cognitive and behavioural skills are hard to trade off against each other. This result is robust to baseline controls, alternative scaling, bootstrap inference, and extension to Leaving Certificate outcomes. When the noncognitive input is instead measured with a broad personality inventory (TIPI Conscientiousness), the same complementarity pattern weakens, precision falls, and one subgroup (girls in Maths) produces a point estimate above unity. The contribution is that measurement choices on the noncognitive margin matter for production-function conclusions: skill complementarity is empirically stable for attentional regulation but not for generic personality proxies.
Chapter in its early stage can be viewed here, and a quick overview of main results here.
Abstract: Research on gender differences in Maths achievement has documented persistent gaps, but most studies examine the gap at a single point in time and say little about whether its composition changes as children develop. This paper asks how the structure of the gender gap in Maths differs between late childhood and early adolescence. Using longitudinal data from Ireland and Oaxaca-Blinder decompositions of the same college-entrance exam outcome with predictors measured in primary school (age 9) and secondary school (age 13), I find that the gap shifts from coefficient-dominant (driven by differences in how similar skills map into outcomes) to endowment-dominant (driven by differences in measured skills themselves) across this transition: in late childhood, boys and girls have similar measured skills but those skills map differently into outcomes; by early adolescence, measured skill differences-principally numeracy-account for most of the gap. An English comparison provides an internal benchmark: the pattern reverses by subject, with girls' advantage remaining coefficient-dominant even when boys hold stronger measured endowments. Distributional decompositions show that average differences mask substantial concentration in the middle of the score distribution. A secondary extension examines father absence under stricter measurement and attrition constraints.