The heritability of self-control: A meta-analysis

31 twin studies from 17 independent samples, more than 100,000 individuals (40,513 monozygotic and 67,863 dizygotic counting independent samples only), ages 1.2 to 33

Mostly United States (26 of 31 studies), plus China, Germany, Italy, Japan and the Netherlands.

A three-level multilevel meta-analysis of twin CORRELATIONS rather than of published ACE estimates - a deliberate choice to avoid the best-fitting-model bias that inflates genetic estimates. 108 monozygotic and 104 dizygotic correlations, with heritability derived via Falconer's formula.

Monozygotic r = 0.58 and dizygotic r = 0.28, giving heritability of 60% with shared environment approximately zero - the monozygotic correlation is twice the dizygotic one, which is the signature of additive genetic variance with little or no common environment. 76% of included studies reported no or very little shared-environment influence. Non-shared environment including measurement error accounts for 40%. No moderation by sex, and age drops out once informant is controlled. THE INFORMANT MODERATOR IS THE IMPORTANT CAVEAT - heritability is 75% by parent report, 53% by self-report and 41% by OBSERVATION, so the estimate is largest exactly where the measure is weakest and a single rater supplies both twins' scores.

This is the genetic evidence, subject to the equal-environments assumption.

Synthesises 31 independent twin studies, so it is itself a replication record. Converges with Rimfeld et al.'s finding of negligible shared environment for grit.