What a New Massive Mexican Family Study Tells Us About the Effects of Ancestry on Different Traits
How does genetic ancestry shape human traits? A new study by Wang and colleagues (2025), drawing on data from over 140,000 adults in Mexico City, set out to answer this question. By leveraging family-based designs in a highly admixed population, the researchers examined 15 complex traits, but highlighted three in particular - height, type 2 diabetes (T2D), and educational attainment (EA) - to show how genetic and environmental influences can play out very differently.
The results reveal powerful genetic effects on height and T2D, and almost entirely “environmental” (below I explain why the quotes) explanations for others.
The cohort itself is highly admixed, reflecting Mexico’s population history. On average, participants had about two-thirds Indigenous American ancestry, nearly a third European ancestry, and small contributions from African (~3%) and East Asian (<1%) sources. This rich variation in ancestry proportions across individuals and families made it possible to ask whether differences in traits reflect direct genetic effects of ancestry or environmental factors correlated with ancestry.
A powerful design: within-family ancestry effects
The study exploited a special feature of Mexican genetics: most individuals have mixed proportions of Indigenous American (IAM), European (EUR), and smaller amounts of African (AFR) and East Asian (EAS) ancestry.
At the population level, ancestry proportions correlate with both genes and environment (e.g. diet, healthcare, school access).
Within families, however, siblings differ in ancestry proportions due to random segregation at meiosis. Those differences act as a kind of natural randomization: if siblings with more IAM ancestry are consistently shorter, that points to a direct genetic effect.
By comparing between-family vs within-family effects, the authors could separate environmental correlations from genuine causal ancestry effects.
Height: clear evidence of a genetic effect
Population-level: More IAM ancestry was strongly associated with shorter stature (≈ −2 SD, about 13 cm difference relative to 100% EUR).
Within-family: The effect persisted, at −1.5 SD, confirming a direct genetic ancestry effect.
Genetic follow-ups showed IAM ancestry correlated with fewer “height-increasing” alleles identified in GWAS, consistent with biological mechanisms.
The authors even found signs of natural selection acting on height-associated loci in Indigenous populations, echoing prior findings in Peru and Greenland.
Type 2 Diabetes: a strong genetic signal
Population-level: IAM ancestry was associated with a much higher risk of T2D.
Within-family: The effect remained striking, with a log-odds ratio of 5.13 - evidence that ancestry-linked alleles increase susceptibility to diabetes.
Polygenic analyses confirmed that IAM ancestry carried a greater burden of T2D-increasing alleles.
Takeaway: The high prevalence of T2D in Mexico is not solely due to diet and healthcare access - there’s a genuine, causal genetic contribution linked to Indigenous ancestry.
Education: all about environment/assortative mating (at least here)
Education told a completely different story.
Population-level: IAM ancestry was strongly associated with lower education (over 2 SD differences).
Within-family: The direct genetic effect vanished. Siblings who differed in ancestry had the same educational outcomes.
Variance decomposition: All sibling similarity in EA was attributed to the shared environment, with heritability estimates effectively zero (even slightly negative, but within error margins).
This contrasts sharply with European studies, where EA is typically 20–40% heritable. Why?
Measurement: They used a 4-level categorical education variable, less precise than “years of education.”
Context: Participants (aged 35+ in 1998–2004) grew up in a Mexico City where access to schooling was strongly shaped by social and economic conditions.
Assortative mating: The study found very strong mate correlation by ancestry (≈ 0.5 for IAM/EUR). In these models, that extra genetic resemblance gets misclassified as “shared environment” (More on this below).
Statistical power: Within-family ancestry variation is small. In MCPS, the between-family SD in IAM ancestry was ~0.167, but the within-family SD was only ~0.020. The SE for EA heritability was ±0.11, leaving plenty of room for hidden genetic effects.
Takeaway: In this cohort, education differences were driven by environment, not direct genetic ancestry. But that doesn’t mean education is never heritable - just that here, environmental constraints overwhelmed genetic contributions.
Why assortative mating can make genes look like “environment”
One subtlety in the Mexico City study is assortative mating - the tendency of people to choose partners who are similar to themselves. The researchers found very strong assortative mating on ancestry: They found:
Indigenous American (IAM): spouse correlation ≈ 0.53
European (EUR): spouse correlation ≈ 0.52
African (AFR): spouse correlation ≈ 0.41
East Asian (EAS): spouse correlation ≈ 0.02.
This fact changes how genetic and environmental influences are disentangled.
In a simple world of random mating, siblings are genetically similar only because of random chance: the proportion of DNA they share “identical by descent” (IBD). Variance models can use this IBD information to separate genetic variance, which reflects the effects of inherited alleles, from shared environment, which includes family factors like household income or school district. But under assortative mating, parents start off genetically similar, and their children inherit more similar sets of alleles than they would under random mating. This boosts sibling resemblance in traits linked to the parents’ similarity. However, that the model doesn’t “know” about assortative mating. It only tracks IBD. As a result, the extra resemblance between siblings - which is genuinely genetic - is not explained by IBD and instead gets lumped into the “shared environment” category.
The result is that genetic effects are systematically under-assigned to the genetic component and over-assigned to the shared environment.
The authors themselves wrote: “Moreover, in these analyses, genetic variance in the population due to assortative mating would appear as common environmental variance”.
More technically, the shared environment is overestimated because the model assumes the parent–child and sibling kinship coefficients are fixed (0.25) under random mating.
With assortative mating, the true genetic similarity is higher, but since the model doesn’t adjust ϕ (kinship coefficient), the unaccounted-for genetic variance is misclassified as environment.
For a trait like education, where parents with similar ancestry and similar schooling levels are more likely to pair up, assortative mating can make the genetic contribution invisible in these variance decompositions. That’s why the study’s estimate of “all environment, no heritability” for education should be read cautiously.
Why within-family estimates are so imprecise
One striking feature of the study is how much noisier the within-family estimates are compared to the population-level ones. At the population level, statistical precision comes from two things: the huge sample size and the large variation in ancestry across families. Within families, though, things look very different. Precision depends not on the total number of participants, but on the number of sibling pairs, and on how much siblings actually differ in ancestry.
In the Mexico City cohort, siblings differ very little in their ancestry proportions - the within-family variance is only about 1.4% of the between-family variance. Combine that with the fact that there are fewer sibling pairs than individuals overall, and the result is more than a 100-fold drop in precision for within-family analyses. That’s why the within-family estimates, especially for binary traits like type 2 diabetes, came with huge confidence intervals. The authors clearly state this in the discussion section of the paper.
Within-family designs are powerful because they eliminate environmental confounding, but they pay a heavy price in statistical efficiency. For traits like education or diabetes, where the genetic signal is already subtle, the Mexico City study shows that even larger samples will be needed to get precise within-family effect estimates.
Takeaway Lessons
This study is one of the most convincing demonstrations that:
Some traits (height, T2D) are strongly shaped by genetic ancestry, via allele frequency differences between populations.
Other traits (education) may show huge population-level associations with ancestry, but those vanish within families, due to assortative mating, shared environment and lack of precision.
Summary of the results and concluding thoughts
Height: Genetic effect of ancestry, partly shaped by natural selection.
Type 2 Diabetes: Direct genetic effect, explaining part of Mexico’s high disease burden.
Education: No detectable genetic ancestry effect; family environment/assortative mating dominates.
How should we interpret the education result? At first glance, it might look as though education in this cohort was entirely determined by environment - access, poverty, and local conditions. But we know from many large studies in European and other populations that education is substantially heritable. The more likely explanation here is that strong assortative mating on ancestry and education in Mexico City pushed genuine genetic effects into the “shared environment” bucket of the model. In other words, the heritability of education is not truly zero in this sample, it’s just misclassified. The authors themselves note that genetic variance created by assortative mating will appear as shared environment, and the large standard errors reinforce the idea that the apparent lack of genetic effects is an artefact of modeling rather than a genuine absence.
References
Wang, S., Berumen, J., Vergara-Lope, A., Baca, P., Barrera, E., Rivas, F., Aguilar-Ramirez, D., Collins, R., Emberson, J. R., Hill, M., Goddard, M. E., Yengo, L., Young, A. S., Alegre-Díaz, J., Kuri-Morales, P., Tapia-Conyer, R., Torres, J., & Visscher, P. M. (2025). Direct effect of genetic ancestry on complex traits in a Mexican population. medRxiv. https://doi.org/10.1101/2025.09.09.25335237