The Gender Wage Gap: What the Data Actually Shows

The most-cited figure—“women earn 77 cents on the dollar”—comes from comparing all male and female workers’ annual earnings, regardless of occupation, hours worked, industry, or experience. This is a real gap, but it’s a measure of aggregate earnings differences, not a measure of pay discrimination within identical jobs.

The “77 cents” figure predates 2015 and uses the all-workers comparison. The current comparable figure is approximately 75–76 cents, or 82 cents when restricted to full-time workers. Neither is wrong as a description of aggregate earnings; both are misleading if used to describe pay discrimination within comparable jobs.

The raw gap combines several factors: differences in occupation and industry, differences in average hours worked (women average ~36 hours/week vs. men’s ~40 among full-time workers), differences in years of work experience and seniority, and differences in employer size and sector. None of these differences are neutral in terms of cause—they themselves may reflect structural inequalities—but conflating them with “same job, different pay” misrepresents what the data shows.

The appropriate question is: how much of the gap remains after controlling for these factors? That is the adjusted gap.

Economists decompose the wage gap into the portion explained by measurable worker characteristics (occupation, education, hours, experience, industry, firm size) and an unexplained residual. The unexplained portion is often called the “adjusted gap.” It captures a mix of potential discrimination and unmeasured factors.

The academic consensus, summarized by Claudia Goldin’s Nobel Prize-winning research (2023), is that the largest single driver of the gender wage gap is not direct discrimination but the “greedy jobs” premium—the disproportionate compensation for long, inflexible hours in certain high-earning professions (law, finance, consulting). Because women disproportionately reduce hours or switch to more flexible arrangements around childbirth, they lose access to this premium at higher rates than men.

The unexplained residual gap does not equal discrimination. It represents: (1) unmeasured worker characteristics (commute willingness, specific skills, performance, negotiation); (2) unmeasured job characteristics (risk, physical demands); (3) actual discrimination in pay-setting; and (4) statistical noise. Separating these components requires methods beyond standard decomposition—primarily audit studies and natural experiments.

Approximately 50% of the raw wage gap is attributable to women and men working in different occupations and industries—a phenomenon called occupational sorting. Women are overrepresented in education, healthcare support, administrative services, and social work; men are overrepresented in engineering, construction, finance, and technology.

The key structural finding from Goldin and others is the “devaluation” pattern: when women enter a field in large numbers, wages in that field tend to decline relative to other fields, and vice versa. Research on the computing industry (once female-dominated) and other fields shows this pattern historically. This means “choice of occupation” and “labor market discrimination” are not cleanly separable—the value placed on occupations is partly endogenous to gender composition.

Paula England and others have documented this systematically: female-dominated occupations pay less than male-dominated occupations with similar skill requirements, educational demands, and working conditions. This gap is not fully explained by the characteristics of the work itself.

The most well-documented driver of the gender wage gap is not raw discrimination but the motherhood penalty: the earnings loss women experience following childbirth, relative to similarly qualified men who have children.

The motherhood penalty is driven by several mechanisms: (1) career interruptions and reduced experience accumulation; (2) reduction in hours worked; (3) switching to lower-paying but more flexible employers or roles; and (4) possible employer discrimination against mothers (the “maternal wall”).

The relative contributions of these mechanisms vary by study. Claudia Goldin’s work emphasizes the role of temporal flexibility: in professions that pay a very high premium for being available at specific times (law, finance, consulting), reducing hours by even 20% reduces pay by far more than 20%. Women who shift to flexible arrangements post-childbirth exit the “greedy job” pay structure entirely.

Audit studies find some direct employer discrimination against mothers: identical résumés with signals of parenthood result in lower callback rates for women and the same or higher rates for men (Correll et al., 2007, American Journal of Sociology). This “maternal wall” effect is separate from the hours/flexibility channel.

The most rigorous direct evidence on gender discrimination in hiring and pay comes from audit studies (randomized resume experiments) and natural experiments. These bypass the self-selection problems in observational data.

Audit studies find discrimination is real but not universal. It varies significantly by occupation (more prevalent in male-dominated high-status fields), seniority level, and context. Studies specifically designed to measure discrimination in pay (rather than hiring) are fewer and harder to conduct. The conclusion from this literature is that discrimination is one component of the wage gap—not the sole explanation, but not zero.

The gender wage gap is a global phenomenon but varies substantially across countries, suggesting policy and institutional factors matter.

Countries with smaller gender wage gaps share several features: universal, affordable childcare (reduces career interruptions); paid parental leave available to both parents (Iceland and Scandinavian models reduce the signal value of maternity); equal pay legislation with enforcement; and lower occupational segregation. However, even Sweden—with generous family policy—maintains a 7.3% gap, suggesting structural factors persist regardless of policy.

The Nordic gap is primarily attributable to occupational segregation (Nordic labor markets are among the most occupationally segregated by gender in the developed world, despite high female employment) rather than within-occupation pay differences.

The motherhood penalty of ~20% over 10 years is documented in high-quality administrative data across multiple countries. It is not primarily explained by women freely choosing less demanding work—it is driven by the structure of high-paying jobs that penalize flexibility disproportionately. Women who take the same leave as men in countries with gender-neutral parental leave systems still experience larger career interruptions because of unequal uptake.

Occupational devaluation is real: wages in female-dominated fields are systematically lower than male-dominated fields with comparable skill demands, and this pattern emerged historically as women entered fields (computing, biology). The “choice of occupation” argument therefore doesn’t fully explain the gap—the market doesn’t value female-coded work equally. Audit studies confirm direct discrimination in specific high-value hiring contexts (STEM, high-status service industries). The residual 8–14% adjusted gap is real and partially reflects employer behavior.

After controlling for occupation, hours, industry, and experience, approximately 60–70% of the gap is explained by measurable differences. Young, childless, college-educated women in the same field as comparably qualified men earn at near parity in many sectors. The gap is primarily a phenomenon that emerges with childbearing, not a uniform experience across the female workforce.

The “greedy jobs” premium is not discrimination per se—it is a market valuing specific time patterns. The solution is changing the structure of high-paying work (reducing the premium for continuous availability) rather than attributing earnings differences entirely to bias. Countries with equal pay laws and high female labor force participation still have significant gaps because of occupational sorting, suggesting legislation alone does not resolve the structural issue.

Men take more dangerous jobs (93% of occupational fatalities are male), work longer hours on average, and are more concentrated in high-variance/high-risk career paths. These factors are compensated by higher average wages and are not captured in simple comparisons.

Both sides of this debate share common ground on several empirical points: the raw gap is real; a significant portion is explained by occupational and hours differences; a residual of 8–14% persists after controls; the motherhood penalty is the dominant long-run driver; and discrimination exists in some contexts but is not uniform. The main dispute is over interpretation—whether the occupational structure itself reflects discrimination (devaluation thesis) or free preferences operating in a neutral market (compensating differentials thesis). Both have empirical support.