Reproductive-Burden Extension

Family → Reproductive stage: Adding reproductive-stage and couple-type controls reduces the gap by ~3 pp (13.7% → 10.8%). Job context → Interactions: Interaction terms absorb the remaining main female effect — the gap does not disappear but gets channeled through specific reproductive × job-context pathways.

Job rigidity is the largest single channel: women in rigid-schedule occupations face a 12.1 pp additional penalty. Having a young child in a rigid job compounds this by another 4.6 pp. The positive coefficients for recent birth and young children reflect survivor selection — women who remain employed through these events tend to be in higher-paying positions.

Reproductive burden is the second-largest channel. Its log-point contribution is stable across 2015–2023 (mean −0.036, SD 0.003), while its share of explained variation ranges from 27.0% to 36.4% because the total explained amount moves by year. Job sorting remains dominant (~70%), but the reproductive channel is the most important single block beyond sorting. The schedule block works in the opposite direction: controlling for hours reveals a larger female penalty, consistent with women working fewer hours in higher-paying arrangements. Generic family status contributes nothing beyond the reproductive-stage controls.

These household variables work as sensitivity checks, not new explanatory wins. Extra-adult household structure barely moves the M7 residual, and partner-resource controls increase the partnered-sample residual because they are partly post-market. The current ACS extract in this repo does not expose MULTG, so the fitted composition row uses other_adults_present only. relative_earnings is excluded because it is constructed from the respondent’s own wage.

Among childless women who have not yet had children, those with the highest predicted fertility risk earn $12.62 less per hour than those with the lowest risk. This gradient exists before any children are born and likely reflects both sorting and anticipatory labor-market behavior.

Interpretation note: Age is one of the predictors in the fertility-risk score, so lower-risk quartiles skew older within the 25–44 window. Part of the raw wage gradient therefore reflects age–experience returns rather than fertility risk alone. The older-placebo groups below help bound this confound.

Lesbian married women earn 12% more than otherwise-similar heterosexual married women even after full controls. This is consistent with same-sex couples facing different household specialization incentives — if the motherhood penalty is partly driven by gendered household division, couples without a male earner may specialize differently.

Mothers of children under 6 work 149 fewer paid minutes per day than comparable men — more than double the overall gender gap. The childcare and housework burden increases sharply with young children, consistent with the schedule-rigidity penalty found in the wage regressions.

SIPP confirms the ACS pattern: the gap is smallest among childless unpartnered workers (−1.2%) and largest among mothers with school-age children (−26.4%). The SIPP reproductive controls are noisier than ACS but directionally consistent.