ACS definitions

  • 2024 ACS 1-year PUMS. 50 states and DC, ages 25–54. 1,217,903 observations, representing 133.0 million adults.
  • Main population includes institutional group quarters and armed forces. Employment: ESR 1, 2, 4 or 5.
  • Earnings: positive PERNP among people reporting work in the past year. Adjusted to 2024 dollars with ADJINC.
  • Person weights throughout. 95% intervals using 80-replicate standard errors for main gap specifications.
  • Hispanic of any race. Other categories are non-Hispanic; AIAN combines American Indian and Alaska Native categories.

Return to the talk · Census microdata · Data dictionary

Adjusted employment gaps

Unadjusted and adjusted employment gaps relative to non-Hispanic White adults for all seven other race and ethnicity categories, shown for men and women.

2024 ACS. Adjusted: categorical age, education, state and nativity. Separate models by sex; 95% replicate-weight intervals.
All adults 25–54, including institutions and armed forces. AIAN: American Indian / Alaska Native. NHPI: Native Hawaiian / Pacific Islander.

Adjusted earnings gaps

Unadjusted and adjusted geometric-mean annual earnings gaps relative to non-Hispanic White adults for all seven other race and ethnicity categories, shown for men and women.

Positive annual earners, ages 25–54. Weighted log-earnings models; plotted gaps equal 100 × (exp(coefficient) − 1).
Adjusted: categorical age, education, state and nativity. 95% replicate-weight intervals.

Population and earnings definitions matter

Black–White gap, after age, education, state and nativity controls

Outcome and sample Men Women
Employment, all adults −11.4 pp +2.1 pp
Employment, civilian noninstitutional −8.5 pp +2.1 pp
Positive annual earnings −29.7% −6.6%
Full-time, full-year positive earnings −22.9% −11.3%
Approximate hourly earnings −20.8% −8.3%

Full-time, full-year: at least 35 usual hours/week and 50 weeks/year. Approximate hourly earnings = annual labor earnings ÷ (weeks × usual hours). Point estimates.

Prediction is a different question

Held-out variation explained by each characteristic on its own

Predictor Employment Log annual earnings
Education 5.22% 12.16%
Disability 6.47% 1.34%
Race / ethnicity 0.59% 2.89%
Sex 1.02% 2.13%
Age 0.05% 2.11%
State 0.29% 1.05%
Nativity 0.03% 0.09%

Weighted held-out R². One deterministic 80/20 household split, pooled sexes; additive categorical models.

Sources for the argument