Heritability and Social Interaction in Labor Market Outcomes: Evidence from the Population of Twins in the United States -- by David Card, Jesse Rothstein, Moises Yi

TEXT START: We use data on 44,000 twin pairs observed as adolescents in the 2000 Census, linked to their earnings 20 years later, to study the heritability of labor market outcomes.

THE DISSECTION

This is a variance-allocation study of the pre-AI labor market. It separates person-linked earnings, employer pay premiums, inherited traits, family effects, and social interaction. Its most important result is not that genes determine income. It is that conventional twin estimates can overstate genetic influence when same-sex siblings share firms, networks, and workplace behavior. Removing siblings who work together reduces estimated heritability by 15%.

The paper therefore shows that labor-market outcomes are produced by an entangled system of biology, family environment, social transmission, employer sorting, and institutional access. It narrows the genetic claim instead of proving genetic destiny.

THE CORE FALLACY

Under the Discontinuity Thesis, the paper answers a historically bounded question and risks being mistaken for a strategic one. Heritability of log earnings is not evidence of durable economic agency. It measures variance inside a wage-mediated system whose operating circuit is labor, income, and consumption.

When cognitive automation achieves durable superiority, inherited traits that once helped determine earnings can be reproduced, selected, or bypassed by machine systems. A heritable advantage in the old labor market is not ownership of AI capital, control of productive infrastructure, or indispensability to a Sovereign. The paper can explain who prospered inside the old machine; it cannot establish who remains economically necessary after P1–P3.

A second error lies in treating heritability as if it meant immutability. It does not. A population-level variance estimate is not an individual sentence, a biological essence, or a guarantee that the trait retains market value when the production regime changes.

HIDDEN ASSUMPTIONS

• Twenty-year earnings are treated as a meaningful proxy for productive capacity rather than a product of historical institutions, employer access, credentialing, networks, and bargaining power.
• The extended twin design is assumed to separate genetic effects from prenatal conditions, family dynamics, assortative mating, sex-specific environments, and unobserved social transmission well enough for causal interpretation.
• The AKM decomposition is treated as substantively real and sufficiently additive, despite endogenous worker-firm matching and worker-employer complementarity.
• Same-firm employment is treated as the principal measurable social-interaction channel, though networks and shared opportunity structures can operate without coworkers being employed together.
• Findings from this cohort and institutional setting are assumed to generalize beyond the specific U.S. labor market that generated the data.
• Readers may confuse population heritability with individual predictability, then confuse individual predictability with fixed human economic worth.
• Most importantly, the analysis assumes that labor-market earnings remain the central measure of participation. That assumption is precisely what the Discontinuity Thesis puts on trial.

SOCIAL FUNCTION

Primary classification: partial truth. Secondary classification: prestige signaling and potential ideological anesthetic.

The research is not mere copium. Its correction for social interaction is serious, and it undermines crude genetic fatalism. But stripped of its methodological limits, the result can be converted into a convenient story that income reflects inherited merit rather than institutions, ownership, and power. That story launders a social hierarchy into nature.

In a transition economy, the paper may also function as transition management: a more refined instrument for sorting people into the shrinking set of human roles that remain useful. It improves the map of the old labor market while leaving the ownership of compute, models, energy, logistics, and maintenance untouched.

THE VERDICT

This is a valuable forensic refinement of how the dying wage system distributed rewards. It is not evidence that the system is stable, that genes determine destiny, or that inherited advantage survives the automation of cognitive labor. Under the Discontinuity Thesis, it is a tissue sample from the corpse: informative about the previous organism, useless as proof that the organism can live.