Gap Heterogeneity
The gender earnings gap varies substantially across occupation, race/ethnicity, education, and region. All figures are raw hourly gaps from 2023 ACS.
By Occupation
ACS 2023Raw Gap by Broad Occupation (ACS 2023)
Horizontal bars sorted by gap magnitude
Gap = (male median hourly − female median hourly) / male median hourly × 100.
| Occupation | Raw Gap % |
|---|---|
| Healthcare | 30.59 |
| Sales / Office | 28.63 |
| Management / Professional | 25.27 |
| Service | 20.06 |
| Military | 5.47 |
By Race / Ethnicity
ACS 2023Raw Gap by Race / Ethnicity
Horizontal bars sorted by gap magnitude
| Race / Ethnicity | Raw Gap % |
|---|---|
| Asian | 20.39 |
| White non-Hispanic | 19.30 |
| Hispanic | 13.04 |
| AIAN | 10.06 |
| Black | 5.64 |
The gap is widest among Asian (20.4%) and White non-Hispanic (19.3%) workers. Among Black workers, the raw gap is only 5.6%—reflecting both lower male earnings and relatively higher female earnings in that group.
By Education
ACS 2023Raw Gap by Education Level
Bar chart by highest degree attained
| Education Level | Raw Gap % |
|---|---|
| Less than HS | 18.10 |
| HS Diploma | 19.30 |
| Some College | 22.09 |
| Associate’s | 19.01 |
| Bachelor’s | 23.15 |
| Master’s | 27.42 |
| Professional | 20.55 |
| Doctorate | 17.69 |
The gap is widest among workers with a Master’s degree (27.4%) and narrowest among doctorate holders (17.7%). The gap generally widens with education through the Master’s level.
By Region
ACS 2023Raw Gap by Census Region
Four Census regions
| Region | Raw Gap % |
|---|---|
| South | 17.49 |
| Midwest | 16.90 |
| West | 16.06 |
| Northeast | 15.93 |
Method Comparison (2023)
ACS · CPS · SIPP| Method | ACS | CPS | SIPP |
|---|---|---|---|
| Raw gap % | 16.81 | 15.77 | 15.09 |
| OLS adjusted % | 13.23 | 16.96 | 10.91 |
| DML adjusted % | 17.88 | 19.61 | 12.89 |
| Oaxaca unexplained % (supplemental) | 87.81 | 90.88 | 85.60 |
DML (double/debiased machine learning with elastic net) produces larger adjusted gaps than OLS in all three datasets, suggesting OLS may understate the residual gap. The Oaxaca row is shown as a decomposition cross-check rather than a headline gap estimate.