A major analysis drawing on census and tax records concludes that inherited genetic factors account for around a third of the variation in earnings among steadily employed Americans in their late 20s and 30s. The study also suggests that traditional twin studies may overstate the genetic contribution to career success, partly because same-sex twins are more likely to work at the same company. The findings were published in Proceedings of the National Academy of Sciences (PNAS).
Heritability is a statistical concept that measures how much of the differences in a trait across a population can be traced to inherited genetic variation rather than environment. In economics, researchers have long sought to separate genetic influences from family background when explaining wage differences across individuals.
Historically, economists have relied on the classical twin design, which compares identical twins, who share all their genes, with fraternal twins, who share about half, to tease apart genetic and environmental effects. This approach rests on the equal-environments assumption—the idea that identical and fraternal twins experience similar home environments.
Previous work has shown that family background explains a substantial portion of adult earnings, but ordinary siblings do not allow clean separation of DNA from parental wealth or parenting styles. Twins have been seen as a useful method to disentangle these channels, though challenges remain in fully distinguishing true genetic effects from social influences.
The study combined population-scale records to examine both an individual’s earning power and the premiums paid by employers. It treated wages as consisting of a portable human capital component and an employer pay premium, with the latter reflecting how much a given company adds to earnings for its staff.
The researchers linked the 2000 United States Census household data to quarterly wage records from state unemployment insurance filings, using data from 2018 through the first quarter of 2023 and omitting the final three quarters of 2020 to avoid pandemic disruptions. They defined twins as children in the same household born within two days of each other, and non-twin siblings as those born at least nine months apart.
Among the full wage sample, the study analysed 87,000 twin pairs (60,000 same-sex and 27,000 opposite-sex) and 2,168,000 non-twin sibling pairs (1,072,000 same-sex and 1,096,000 opposite-sex).
For the detailed wage analysis, the authors focused on pairs in which both siblings held stable full-time work, defined as eight quarters in which they earned more than $3,800 from a single employer, excluding the first and last quarters of each job spell. This produced 44,500 twin pairs (31,500 same-sex and 13,000 opposite-sex) and 1,007,000 non-twin sibling pairs (507,000 same-sex and 500,000 opposite-sex).
Because census records do not indicate whether twins are identical or fraternal, the researchers used biological sex as a proxy. All opposite-sex twins are fraternal, while same-sex twins comprise a mix of fraternal and identical twins, with identicals estimated to account for about 58.7 percent of same-sex pairs in the sample.
To account for potential differences in how families treat sons and daughters, the team employed a difference-in-differences approach, subtracting the similarity gap between same-sex and opposite-sex regular siblings from the gap between same-sex and opposite-sex twins.
The researchers analysed average quarterly earnings on a logarithmic scale and used a wage decomposition model to separate each worker’s pay into the portable earning capacity and the average employer pay premium, controlling for work experience and calendar time.
Within the employer pay premium, the authors further broke it down into geographic location premiums, industry premiums and a remaining workplace-specific premium.
Across the full wage sample, the correlation in earnings was 0.543 for same-sex twins, 0.317 for opposite-sex twins, 0.386 for same-sex non-twin siblings, and 0.265 for opposite-sex non-twin siblings, highlighting stronger similarity among same-sex twins.
The team also examined non-twin siblings separated by one to three years of age difference, finding that earnings similarity barely faded with age gaps and that the gap between same-sex and opposite-sex siblings remained steady, supporting the notion that shared family and gender influences do not drastically change for closely spaced siblings.
From their extended twin model, the researchers initially estimated that inherited genetic factors accounted for 36 percent of the variance in overall earnings. They also estimated that genetics explained 23 percent of the variance in having any employment and 32 percent of the variance in the number of quarters employed.
When earnings were decomposed into portable earning power and the employer pay premium, the portable component had an estimated heritability of 40 percent, while the employer pay premium had an estimated heritability of 21 percent.
Within the workplace component, the industry pay premium showed a heritability of 26 percent and the remaining workplace-specific premium 20 percent, while the geographic location premium (18 percent) was not statistically significant because most siblings remained in the region where they grew up.
However, the authors then re-estimated the model after removing all twin and sibling pairs who ever worked at the same establishment. The heritability of the employer pay premium fell from 21 percent to 7 percent and was no longer statistically significant. The industry premium dropped from 26 percent to 12 percent, and the remaining workplace-specific premium fell from 20 percent to essentially zero.
After excluding such pairs, the link between genetics and employer pay premiums aligned more closely with what would be expected from high-paying firms hiring high-skilled workers. Excluding coworkers also lowered the estimated heritability of portable earning power from 40 percent to 35 percent, and the overall earnings heritability from 36 percent to 31 percent, which the authors regard as their best estimates.
Even with this downward adjustment, the finding that genetics account for roughly a third of individual earning differences is consistent with a contemporaneous German study that linked higher cognitive ability at age 23 to higher socioeconomic status at 27 largely through shared genetic factors. Both studies point to a substantial genetic influence on socioeconomic outcomes.
Speaking about the findings, Felix Bittmann, a postdoctoral researcher at the Leibniz Institute for Educational Trajectories and the University of Bamberg who was not involved in the research, noted that the results show heritability is present, but the effect depends heavily on whether twins work at the same firms. “It’s an interesting aspect that deserves more attention. Clearly, the paper shows that even well-done genetically informed studies cannot partial out ‘pure’ genetic effects as social interactions will always be present and can influence the findings strongly.”
Bittmann also said the study’s broader implication is that the social sciences must grapple with how success is determined, a question of nature versus nurture that twin designs help address but cannot fully resolve. “Twin designs are usually an elegant way to attempt to account for both factors, which classical research designs (surveys) cannot capture at all, as genetic transmission pathways are usually not measured,” he added.
Sociologists, he suggested, often try to separate primary and secondary effects—how family background shapes academic performance and how families influence educational choices—but genetic pathways are frequently ignored, which could bias results. “This research can help to estimate how large such a bias could probably be.”
In assessing the study’s design, Bittmann praised the use of a large US database of official statistics, noting that twins are not directly recorded in these databases but can be inferred from birth dates. He cautioned that certain assumptions are necessary, but he regarded them as reasonable within the study’s framework.
A caveat remains that a 31 percent heritability estimate describes variation across a population and does not imply that any one person’s earnings are genetically fixed. The wage analysis also excluded self-employed workers, federal employees and those with minimal attachment to the labour market, leaving room for other pathways to earnings not captured by the study.
The authors acknowledge that model assumptions could push estimates in opposite directions. For example, assortative mating—where partners share genetic traits related to education and earning potential—could inflate sibling-based estimates by about 16 percent. Conversely, unmeasured social interactions could still inflate the portable earning power estimate even after removing same-establishment pairs. The researchers view these caveats as part of ongoing debates about the so-called missing heritability problem in genetics.
Nevertheless, the study, titled Heritability and social interaction in labor market outcomes: Evidence from the population of twins in the United States, was authored by David Card, Jesse Rothstein and Moises Yi.
