| dataset_desc |
This investigation critically examines the heritab This investigation critically examines the heritability of human longevity, challenging prior estimates that have ranged between 15–30% by demonstrating that these figures are substantially inflated due to assortative mating—the nonrandom pairing of mates with respect to longevity-associated traits. Using an unprecedentedly large dataset derived from Ancestry public family trees, encompassing hundreds of millions of historical individuals primarily of European descent living in North America and Europe during the 19th and early 20th centuries, the authors applied advanced structural equation modeling to disentangle genetic, sociocultural, and assortative mating effects on lifespan correlations.
The study concludes that the true transferable variance (t²)—an upper bound on heritability (h²) that includes both genetic and sociocultural inherited factors—is well below 10% for birth cohorts across the 1800s and early 1900s. This suggests that earlier heritability estimates of longevity have been substantially overestimated because they did not adequately correct for assortative mating effects.
Key Concepts and Definitions
Term Definition
Heritability (h²) The fraction of phenotypic variance attributable to genetic variance.
Transferable variance (t²) Phenotypic variance due to all inherited factors, encompassing both genetic (h²) and sociocultural (b²) components, plus their covariance.
Sociocultural inheritance (b²) Non-genetic factors that influence phenotype and are transmitted through families (e.g., socioeconomic status).
Assortative mating (a) The correlation between latent genetic and sociocultural states of spouses that influences phenotypic correlations beyond genetic inheritance.
Nominal heritability Heritability estimated without correction for assortative mating or shared environment, typically based on correlation and additive relatedness.
Methodology Overview
Data Source: Aggregated and anonymized pedigrees (SAP) were created by collapsing 54 million publicly available Ancestry subscriber-generated family trees, resulting in over 831 million unique historical individuals linked by parent–child and spousal edges.
Data Quality Controls:
Removed self-edges and gender-incongruent parent-child edges.
Added missing spousal edges between parents.
Focused on individuals with known birth and death years who had offspring, limiting analysis primarily to birth cohorts from the early 1800s to 1920.
Addressed data artifacts such as birth year rounding.
Analysis Approach:
Estimated phenotypic correlations of lifespan between various relatives (siblings, cousins, spouses, in-laws).
Calculated nominal heritability using standard regression methods correcting for variance differences.
Developed and applied a structural equation model incorporating three key parameters:
Transferable variance (t²),
Inheritance coefficient (b),
Assortative mating coefficient (a).
Utilized correlations among siblings-in-law and cosiblings-in-law to solve for these parameters.
Applied an assortment-correction method using remote relative pairs and their in-law equivalents to validate estimates.
Timeline Table: Analytical Focus and Data Coverage
Period Data Characteristics and Focus
Pre-1700 Mostly European births; sparse data quality Not specified
1700–1800 Increasing data quality; European and North American births
1800–1920 Primary focus; high data quality; large sample sizes in millions
Post-1920 Decline in death-year data; excluded from lifespan analysis
Major Findings
1. Nominal Heritability Estimates Confirm Prior Literature but Are Inflated
Nominal heritability estimates for lifespan correlated with previous findings (15–30%).
Lifespan correlations among blood relatives were similar to past studies.
However, spouses and in-law relatives also showed substantial lifespan correlations, sometimes comparable to or exceeding those of blood relatives.
This indicated that shared environments and assortative mating inflate these estimates.
2. Assortative Mating Significantly Inflates Heritability Estimates
Assortative mating coefficient (a) was consistently high across all analyses, often exceeding 0.8, indicating strong nonrandom mating based on lifespan-influencing factors.
The presence of assortative mating causes phenotypic correlations between relatives to deviate from the linear relationship expected under pure additive genetics.
Correlations between in-law relatives (who do not share genetics) were substantial, confirming the importance of assortative mating rather than shared genetics alone.
3. Structural Equation Modeling Reveals True Transferable Variance (t²) Is <10%
Using sibling-in-law and cosibling-in-law correlations, the model estimated transferable variance (t²) consistently below 7% for all gender combinations and birth cohorts.
This t² value represents an upper bound on heritability (h²) because it includes both genetic and sociocultural transmitted factors.
The inheritance coefficient (b) was estimated between 0.40–0.45, slightly less than the genetic expectation of 0.5, reflecting combined genetic and sociocultural inheritance.
Shared household environmental effects were also quantified and found to be substantial but separate from transferable variance.
4. Independent Validation Using Remote Relatives Supports Low Heritability
Assortment-correction method applied to remote relatives (piblings, first cousins, first cousins once removed) and their in-law equivalents consistently estimated assortative mating coefficients (a) close to or above 0.5.
Transferable variance estimates from these analyses also remained below 10%, validating the sibling-in-law modeling approach.
5. Transferable Variance Decreases with Increasing Birth-Cohort Disparity Among Relatives
Lifespan correlation and transferable variance (t²) were higher when relatives were born closer in time; as the birth-year gap increased, t² declined significantly.
Assortative mating coefficient (a) remained stable across birth-year offsets, suggesting that the decline in transferable variance was not due to mating patterns.
This suggests that genetic and sociocultural factors affecting lifespan vary with historical context, likely reflecting changing environmental hazards and causes of death over time.
Quantitative Summary Table: Structural Equation Model Estimates by Birth Cohort
Birth Cohort Period Transferable Variance (t²) Assortative Mating Coefficient (a) Inheritance Coefficient (b) Shared Childhood Environment (csib) Shared Adult Environment (csp)
1800s–1830s ~5.9–6.5% (across relatives) ~0.68–0.88 ~0.40–0.44 ~4.3% (siblings) ~6.6% (spouses)
1840s–1870s ~4.0–5.5% ~0.53–0.88 ~0.40 ~5.1% ~5.0%
1880s–1910s ~4.0–7.2% ~0.43–0.89 ~0.40 ~6.0% ~4.4%
Values represent means across gender pairs with standard deviations; b fixed at 0.5 for some estimates; all data derived from sibling-in-law and remote relative analyses.
Core Insights
Previous heritability estimates of human longevity (~15–30%) are substantially inflated due to assortative mating.
True heritability (h²) is likely below 10%, and possibly considerably lower after accounting for sociocultural inheritance.
Assortative mating for lifespan-related factors is strong, with a coefficient often >0.8, indicating mates tend to share longevity-related traits, both genetic and environmental.
Sociocultural factors (e.g., socioeconomic status) are a significant inherited component influencing longevity, evidenced by lifespan correlations among in-law relatives and supported by sociological literature.
Transferable variance (t²) decreases as birth cohorts diverge, implying that historical environmental changes modulate the impact of inherited factors on longevity.
Fundamental biological aging processes (e.g., rate of hazard doubling) appear consistent historically, but lifespan-affecting factors mostly modify susceptibility to historically transient environmental hazards, not aging rate itself.
Implications
Genetic studies of longevity should account for assortative mating and sociocultural inheritance to avoid overestimating genetic contributions.
Interventions targeting environmental and sociocultural factors could have a larger impact on lifespan extension than currently assumed genetic predispositions.
Historical and birth cohort context is critical when interpreting heritability and lifespan data.
The biological basis of aging remains consistent, but its interaction with environment and social factors is dynamic and complex.
References to Relevant Literature Mentioned
Smart Summary
... |