Human populations differ in disease prevalences and in average values of phenotypes, but the extent to which differences are caused by genetic or environmental factors is unknown for most complex traits. Family-based analyses of genetically admixed individuals offer a powerful framework for estimating unconfounded effects of genetic ancestry on phenotypes. We leveraged genetic data from admixed adults in the Mexico City Prospective Study (MCPS) to estimate within-family ancestry effects. We quantified associations between genetic ancestry and complex traits among 53,000 unrelated individuals and in 40,000 relatives across 18,000 families. At the population level, relative to a European ancestry baseline, we estimate an effect of Indigenous American (IAM) ancestry of -2 standard deviations for height and a natural log-odds ratio (lnOR) of 1.7 for type 2 diabetes. We estimated a within-family effect of IAM of -1.5 standard deviations for height and lnOR of 5.1 for risk of T2D. These within-family effects are supported by trait-specific between-ancestry differences in trait-increasing allele counts and evidence of selection at trait-associated loci. We will discuss alternative explanations for the observed within-family genetics.