Evaluating the degree to which statistical models can explain or predict social outcomes such as educational attainment, fertility or well-being strongly contributes to theory building, the investigation of social stability, scientific discovery and defines space for intervention. However, research relies on computational "black box" methodology, ignores confounding by non-social factors and is unsuccessful in their prediction. It has been overlooked that in the past 15 years, quantitative geneticists developed a transparent methodological pipeline to tackle this challenge. They expected to explain 80 percent of individual differences in height (heritability) using measured molecular data but failed. This "missing heritability" puzzle has triggered massive advances in theory, methodology, and the recognition of gene-environment interaction. Today, geneticists explain 70 percent of individual differences in height using inference statistical models and measured genes. This provides a roadmap for game changing innovations in the social sciences. In this paper, I infuses social sciences with knowledge from genetic methods to find this "missing environmentality". I will be able to fit all social predictors in a joint model in one degree of freedom and to disentangle genetic from social factors and take gene-environment interaction into account. The paper will provide interpretable statistical model specifications which include higher order interactions to model social complexity as well as classic, parsimonious sociological explanations. It will quantify the relative contributions of various social and non-social domains to the distribution of traits in a population, spotlighting targets for subsequent causal analyses and space for interventions.
Felix C. Tropf, University College London & Purdue University