Variance & Distributional Analysis

Mean gaps tell only part of the story. This extension examines the full distribution of earnings by gender — variance ratios, tail concentration, and how dispersion varies across reproductive stages and job contexts.

Sample window note: This page pools ACS 2013–2024 (9.7M observations) to support distributional breakdowns by reproductive stage and job context. The headline trend series covers 2015–2023. Results here are a separate pooled extension, not directly comparable to the year-by-year adjusted gap estimates.

Headline Findings

ACS Pooled 2013–2024 · 9.7M obs

1.7× top-5% overrepresentation

Men hold 6.3% of total hourly earnings in the top 5% vs. women’s 3.6%. The residual gap is concentrated in the upper tail.

12.2% vs. 7.6% top-decile share

Men capture 12.2% of total hourly earnings in the top decile vs. 7.6% for women — a 4.7 pp gap that persists after controls.

1.04× residual dispersion

After full controls, the overall male/female variance ratio is only 1.04. Most of the raw 10% dispersion gap is absorbed by observables.

Motherhood compresses, rigidity flips

Mothers’ residual variance compresses (ratio 0.82–0.86). In the most rigid jobs, women are more dispersed (0.87).

Raw & Residual Dispersion

Log Hourly Wage

Variance Ratio and Tail Metrics by Gender (Pooled)

Variance ratio = male variance / female variance

All metrics computed on log real hourly wages. Residual variance is from a full-controls specification (demographics, geography, job sorting, schedule, family).

Metric	Male	Female	Ratio / Gap
Raw variance ratio	—	—	1.101
Residual variance ratio	—	—	1.039
P90/P10 ratio	1.823	1.747	+0.076
P95/P50 ratio	1.233	1.167	+0.066
Top-10% earnings share	12.2%	7.6%	+4.7 pp
Top-5% earnings share	6.3%	3.6%	+2.7 pp

Men are more spread out at the top of the hourly-wage distribution. The P90/P10 gap is modest (0.08 log points), but top-earner shares diverge sharply: men capture 12.2% of total earnings in the top decile vs. 7.6% for women.

Quantile Profile

ACS 2023 Quantile Regression

Quantile	Female coefficient	Implied gap %	N
P10	−0.1701	15.64	915,284
P25	−0.1453	13.53	915,284
P50	−0.1473	13.69	915,284
P75	−0.1526	14.16	915,284
P90	−0.1863	17.00	915,284

The mean gap understates how unevenly the residual sits across the distribution. In ACS 2023, the adjusted gap is 13.7% at the median but widens to 17.0% at the 90th percentile, with a smaller rise at the bottom tail (15.6% at the 10th percentile). This is consistent with the tail-share result: residual inequality is concentrated toward the top, not spread evenly across the wage distribution.

Selection-Corrected Variance

IPW Adjustment

Raw vs. Selection-Corrected Residual Variance Ratio by Year

IPW reweights for employment probability; dashed = selection-corrected

Selection correction nudges the residual variance ratio from 1.039 to 1.047 overall — a shift of less than 1 pp. Employment selection does not account for the variance gap between male and female earners, paralleling the mean-gap finding that IPW adds only ~1 pp to the hourly gap.

Dispersion by Reproductive Stage

Within-Stage Comparisons

Residual Variance Ratio by Reproductive Stage

Ratio > 1 = male more dispersed; < 1 = female more dispersed

Stage	Raw Ratio	Residual Ratio	N
Childless recently married	1.146	1.092	2,956,274
Childless unpartnered	1.125	1.174	2,216,122
Childless other partnered	1.144	1.000	22,902
Mother (mixed/other)	1.243	0.857	2,318,307
Mother (child under 6)	0.953	0.816	662,428

The pattern inverts for mothers. Among childless workers, men are modestly more dispersed (ratio 1.09–1.17). Among mothers, women’s residual variance compresses relative to their partners (ratio 0.82–0.86) — consistent with motherhood narrowing the range of viable work arrangements rather than simply lowering the mean.

Dispersion by Job Rigidity

O*NET Context Quartiles

Variance Ratio by Job-Rigidity Quartile

Q1 = most flexible jobs, Q4 = most rigid

Rigidity Quartile	Raw Ratio	Residual Ratio	N
Q1 (most flexible)	1.103	1.048	2,415,808
Q2	1.035	1.001	2,415,807
Q3	1.367	1.104	2,415,807
Q4 (most rigid)	0.894	0.868	2,415,807

In the most rigid occupations (Q4), the variance ratio flips: women are more dispersed than men (0.87). In moderately rigid jobs (Q3), men are most dispersed (ratio 1.37 raw, 1.10 residual). This non-monotonic pattern suggests that job rigidity interacts with gender in ways that affect the shape of the distribution, not just its center.

Interpretation Limits

Variance ratios and tail shares are descriptive summaries of the earnings distribution, not causal estimates. Quantile regression shows where the conditional gap is larger or smaller across the wage distribution, but it is still descriptive of conditional location shifts rather than a formal structural decomposition. A higher male variance could reflect greater returns to risk-taking, occupational sorting into high-variance fields, or measurement differences. The reproductive-stage and job-rigidity breakdowns show where dispersion patterns shift, but cannot identify whether the shift reflects employer behavior, worker preferences, or structural constraints.