The project has produced two papers published on the preprint server bioRxiv. These papers are especially important to other researchers who use genetic data to study socioeconomic and health outcomes. In the first paper, we find that genetic information cannot be used to predict outcomes for individuals’ socioeconomic traits, such as educational attainment. Simulations show that this is not only impossible now, but also in the near future, when genetic discovery studies grow and lead to increasing precision in determining individuals’ genetic predispositions. While the former has been shown in other studies previously, we were the first to simulate the boundaries of individual prediction given future increases in statistical power. In the paper, we use genetic information to construct so-called “polygenic scores” (or “polygenic indices”, PGS or PGI), which present an index of genetic predisposition for an individual. PGS can be constructed with a wide range of methodological techniques. We were the first to show that the rank correlations between differently constructed PGS vary enormously, which can have profound implications for how individual genetic data can be used by researchers, and in the future by clinicians as well. We show a concrete example of such an implication for a medical decision-making setting: the prescription of cardiovascular drugs according to genetic profiles. We conclude the paper by making specific recommendations for researchers who use genetic data to make predictions for individuals.
The other published preprint paper uses a well-established econometric technique in an innovative way for removing statistical attenuation bias when studying the effects of polygenic scores (PGS). Statistical attenuation bias is one of the major methodological challenges in the study of polygenic scores, and we show that genetic ORIV outperforms other approaches. Moreover, genetic ORIV can be applied in a way that is convenient and easy to implement in common research designs.
Finally, an unpublished working paper studies the genetic contributions to social inequalities, and models its environmental influences. We show that Stochastic Frontier Analysis (SFA), a traditional econometric technique used to study production maximisation, provides novel insights in the study of social inequalities. Our first specific innovation is to show that SFA can estimate the extent of inequality of opportunity in a dataset, and to separate the effects of accountable effort (a “fair” contributor to inequality of outcomes) from luck (an “unfair” contributor to inequality of outcomes). In an application to data from the United States, we show that poor family socioeconomic conditions pose the biggest threat to equality of opportunity. Other factors, such as childhood health and parental education, are less important. In addition, we conclude that if every person in the dataset had reached their full “genetic potential” unhindered by environmental circumstances, the average person would have been 1.9 years more educated.
