Economics and Genetics
- Jason M. FletcherJason M. FletcherSchool of Public Affairs, University of Wisconsin–Madison
Two interrelated advances in genetics have occurred which have ushered in the growing field of genoeconomics. The first is a rapid expansion of so-called big data featuring genetic information collected from large population–based samples. The second is enhancements to computational and predictive power to aggregate small genetic effects across the genome into single summary measures called polygenic scores (PGSs). Together, these advances will be incorporated broadly with economic research, with strong possibilities for new insights and methodological techniques.
This article outlines some of the history of the small but growing literature that incorporates genetic concepts, methods, and data into economic analysis, which was coined genoeconomics around 2007 (Benjamin et al., 2008). There have been several other very useful overviews of this topic (e.g., Beauchamp et al., 2011; Benjamin et al., 2012a), although the field is advancing rapidly enough that even recent overviews are fast becoming out of date. This discussion provides a brief background before describing some of the more recent advances; additional treatments are available elsewhere as well (Conley & Fletcher, 2017; Lehrer & Ding, 2017). The article also focuses on the integration of genetics within microeconomics rather than macroeconomics, which has been outlined elsewhere (Ashraf & Galor, 2017; Conley & Fletcher, 2017), with many interesting examples (Cook, 2014, 2015). There is much to be excited about in this integration, although there are important limitations.
Economics research has incorporated genetic concepts for decades, often treating genetics as a nuisance—an alternative explanation of causal processes that should be controlled in empirical work. This can be seen in the use of monozygotic (identical) twins to control for genetic background when exploring associations between such elements as educational attainment and wages (Taubman, 1976b). Similar ideas have been used when exploring sibling differences as a way to control for family background factors (Fletcher, 2010, 2011a, 2014).
While genetics is often treated as a nuisance in social science research, there are at least three waves of more tangible integration of genetics within economics. The first wave occurred before molecular genetic data existed, instead using family relationships as a way to understand underlying genetic effects. A second wave emerged with the candidate gene paradigm in genetics, using a small number of genetic variants that became available in social science and medical data sets in the early 2000s. A third wave has now capitalized on a shift in genetics away from candidate genes and toward genomewide measures.
Before proceeding, some additional background may be useful for many readers who have less familiarity with genetics. The human genome sequencing was completed in 2000 by the Human Genome Project, which aimed to map nucleotides, the rungs in the spiral double helix of deoxyribonucleic acid (DNA), which come in four varieties (adenine, guanine, cytosine, and thymine, represented by the letters A, C, G, and T), as well as other features of human genetic difference. There are approximately 22,000 genes in humans (much fewer than was originally estimated, and many fewer than in some grains and bugs; Pertea & Salzberg, 2010) and approximately 3 billion pairs of letters located within both these genes and the so-called dark matter of human DNA, where geneticists have been unable to map out genetic functionality.
Economists most typically interact with genetic data collected as a component of larger social survey or medical data. This data most often come in two varieties, depending on how the genetic material was assessed and how many of the 3 billion pairs of letters were measured. The first measure, a single nucleotide polymorphism (SNP, pronounced “snip,” which assesses a single-letter variation), is a measure of the letter at each locus in the genome assessed in the data. Because all humans have the same letters in over 99% of all loci, the number of SNPs in a data set often is in the range of thousands or hundreds of thousands. Also recall that humans have two chromosomes (one maternally and one paternally inherited), so at each locus, an individual has two letters (AT, TA, TT), and the letters A and T go together and C and G go together. When quantifying SNPs, then, a researcher often counts the number of letters that are less frequent in humans (called minor alleles) that the respondent has at a given genetic locus (this range is 0, 1, or 2).1 Human genetic similarity also allows large-scale imputation to be used to fill in the blanks of the letters that were not assessed directly. In addition, some data sets contain measures of structural variation in the human genome, such as the variable number of tandem repeats (VNTR), which measures the number of times that a short sequence of letters (ACGACGACG) is repeated in the genome. This second type of measurement was frequently used in the candidate gene phase of genetics, which is discussed later in this article.
Before the Human Genome Project, no measured (i.e., molecular) genetic data were available in social science data sets. Thus, the initial integration in economics with genetic ideas used tools from behavioral genetics to assess the overall importance of genetic factors (known as heritability) for measures of interest to economists, such as earnings (Taubman, 1976a). These approaches use family-based (often twin) samples combined with the basic genetic stylized facts that identical twins share 100% of their DNA and fraternal twins share 50% in order to make these assessments. Goldberger (1979) and others pointed out important (and Herculean) assumptions underlying the approach, which appear to have largely limited its use until a more recent resurgence of interest and newer methods. This recent work often incorporates molecular genetic data (rather than inferring it from familial relationships) and has been used to understand measures of economic preferences, including risk-taking and financial decision-making (Benjamin et al., 2012b; Cesarini, Dawes, Johannesson, Lichtenstein, & Wallace, 2009; Cesarini, Johannesson, Lichtenstein, Sandewall, & Wallace, 2010).
While the literature using twins and siblings appealed to genetics as a confounding factor in efforts to assess causal economic processes, within a larger concern about the issues of family background, the first use (and second wave of integration) of molecular genetic data in economics appeared in a working paper (Ding, Lehrer, Rosenquist, & Audrain-McGovern, 2006), with the first publications appearing in 2008 and 2009 (Ding, Lehrer, Rosenquist, & Audrain-McGovern, 2009; Fletcher & Lehrer, 2009; Norton & Han, 2008) in the context of using genetic variables as instrumental variables to analyze the impacts of health on human capital outcomes. These efforts to use genetic variants as instruments followed similar work in epidemiology, which has been labeled Mendelian randomization (Smith & Ebrahim, 2005). A related innovation was to combine genetic instruments within a sibling-fixed-effects approach (the “genetic lottery”) in an effort to capture the notion of an exogenous shock, as well as eliminate the issue of population stratification common to genetic studies (Fletcher & Lehrer, 2011).2
These initial efforts, as well as a small set of related studies published in the next few years (Hinke Kessler Scholder, Wehby, Lewis, & Zuccolo, 2014; Wehby et al., 2011), often used a single genetic variant or a small set of genetic variants as instruments, justified by the candidate gene approach from the field of genetics, which was typical of the early 2000s. Many of these applications faced issues of weak first-stage F-statistics, failures of accounting for population stratification, the likelihood of failing the exclusion restriction in instrumental variable specifications, and relatively small sample sizes, among others (Fletcher, 2011b).3 Indeed, in hindsight, one might expect that none of the results reported in papers using candidate gene approaches are robust.
During this second wave of integration, a parallel use of genetics in economics, as well as other social sciences, was to examine the possibility of gene–environment interactions (GxEs). As in the instrumental variables case, much of the early work used a single genetic variant or a small set of genetic variants. Often, especially outside of economics, there was a limited focus on examining exogenous environments, further limiting the ability to make causal claims of interaction (Fletcher & Conley, 2013).
While not typical in the gene–environment literature outside of economics, several approaches have been used to attempt to uncover causal estimates. Leveraging the larger tobacco tax literature that uses across-state variations in rates to estimate their impact, Fletcher (2012) examined the interaction of a single nicotinic receptor and state-level tobacco taxes to predict tobacco use and found evidence of interaction. Cook and Fletcher (2015b) used three genetic variants to create a “neuroplasticity score,” showing that this score interacted with birth weight to predict adult IQ and wages using a sibling-difference design (i.e., the genetic lottery approach). Cook and Fletcher (2015a) argued that the APOE gene variant, which increases the risk for cognitive decline in later life, is unknown at the time of educational decisions. They used a sibling-difference design to estimate the interactive effects of educational attainment and the genetic variant in predicting cognition in later life and showed evidence that education can “rescue” genetic vulnerabilities. Thompson (2014) used a sibling-difference design to explore interactions between family income and genetic liabilities to predict educational attainments. Albert et al. (2015) showed evidence of interaction between an educational randomized controlled trial (RCT) and a genetic variant in predicting externalizing behaviors.
While this wave of research into GxEs moved the field forward by utilizing strong research designs for the E in GxE, the work shares some of the overarching limitations with the use of candidate genes (G), including small effect sizes, the possibility of exploring multiple hypotheses, the difficulty of replicating findings, and the likelihood of false-positive findings.
A key issue facing the work in this area is that the paradigm underlying genetic analysis has changed in the last 10 years. Basically, genetic effects related to a specific trait (i.e., a phenotype) previously were hypothesized within a one-gene/one-disease (OGOD) paradigm, which has largely switched to a many-genes-of-small-effect paradigm. This switch reflects a larger failure of much of genetic analysis, which relied on small samples and had a growing problem of failed replication of analyses (Chabris et al., 2012). The switch in paradigm also has required a shift in how genetic analysis is done. To uncover effects that are small and spread across the genome (i.e., the determinants of outcomes are polygenic), efforts to collect and pool together data to create much larger samples have been a high priority, allowing new variants to be detected. While this practice has focused on health outcomes for over a decade, teams led by economists recently have used the approach to examine a socioeconomic outcome of interest—namely, educational attainment.
Indeed, a landmark study (Rietveld et al., 2013b) uncovered several specific genetic loci tied to educational attainment, using a framework called a genomewide association study (GWAS). In this framework, hundreds of thousands (or millions) of regressions are estimated to link each variant to the outcome (education). Economists are often discomforted by the use of data mining with no a priori hypotheses. Researchers using GWAS approaches attempt to minimize problems with data mining by focusing on estimates with very small p-values (genomewide significant findings) and look to replicate these estimates in a second data source. This GWAS study on education should be read closely by economists interested in the integration of genetics into economics to understand how the genetic effects are estimated and get a sense of the auxiliary mechanistic analysis typically done after this discover stage. There also have been extensions into other socioeconomic and related domains (Barban et al., 2016; Rietveld et al., 2013a).
While work using GWAS methods has recently focused on social science outcomes, a key finding is that most genetic discoveries are of very small effects (even smaller than findings for health outcomes). Because individual genetic variants have very small effect sizes, in order to incorporate these variants of interest into economic analyses, investigators construct summary measures using hundreds (and often thousands or millions) of estimated links between genetic variants and outcomes, called polygenic scores (PGSs). For example, a recent PGS for educational attainment can predict 7%–10% of the variation in the outcome out-of-sample, such as in the Health and Retirement Study (HRS; Okbay et al., 2016a). We should expect the power of the PGS to get larger (likely to 12%–14%), which could make the measure the single largest predictor of educational attainment, replacing maternal education (i.e., family history).
The use of PGS across many outcomes has begun to appear in a variety of economics papers: the third wave of integration between genetics and economics. Benjamin et al. (2012a) outlined several uses of PGS, including as control variables that can serve to reduce omitted variable bias, strengthening the power of experiments with small samples, or both, as well as the possibility of using them as targets of medical interventions. Additional extensions include the use of PGS within an instrumental variable framework or a GxE framework (Bearman, 2013).
In the context of instrumental variable analyses, recent research has extended the framework, termed genetic-instrumental variables (GIVs; DiPrete, Burik, & Koellinger, 2017) There are at least two key insights into the extended framework. The first is that controlling for the PGS of the outcome of interest could allow the score for the endogenous variable to be more likely to be excluded from the main equation. For example, in an analysis of obesity on mental health outcomes, a typical analysis would use a PGS for obesity in order to instrument obesity (Willage, 2017) and hope that the obesity PGS does not directly affect mental health outcomes through other channels. However, as PGS measures become more complex and use information across the whole genome (now often millions of genetic variants), the plausibility that the score is not related directly to other outcomes becomes more tenuous. The GIV framework suggests using an obesity-PGS as an instrument for obesity and also control for a PGS for mental health in the analysis. A second key insight of the paper (Diprete et al., 2017) is the use of two PGS measures from two separate GWAS analysis as a way to reduce measurement error in the main PGS by using the secondary PGS to instrument for the first one. This approach has the advantage of strengthening the first-stage relationship by reducing measurement error in the PGS measure, and it also takes advantage of the way that GWAS is often done, from pooling many data sets together and through replication of findings across multiple data sets.
A second use of PGS is to incorporate these measures as the G in the GxE literature. For example, using the HRS, two recent analyses explore interactions between the PGS for educational attainment constructed by Okbay et al. (2016b) and early family socioeconomic status (SES) to predict old-age economic outcomes, such as earnings and financial decision-making (Barth, Papageorge, & Thom, 2017; Papageorge & Thom, 2017). A disadvantage of this work is the lack of clarity of the research design in terms of assessing causal effects. The PGS for education is inherited from parents and is thus correlated with early family SES, making statistical interactions between the two variables (child PGS and parental SES) difficult to interpret.
Several new papers have attempted to strengthen GxE work that uses PGS as the measure of G by using environmental measures (E) that are more likely to be exogenous. For example, Fletcher (2017, 2019) used state-level measures of educational mobility to show interactions with the PGS for educational attainment in the HRS, and used exposure to the 1918 influenza epidemic as an alternative environmental exposure plausibly unrelated to genotype.
An even smaller body of literature has sought to incorporate molecular genetic measures into structural economic models. For example, Biroli (2015) extended a standard health production function model (Grossman, 1972) to allow measured genetic variants to differentially affect both the health production function technology and individual preferences related to the incentives for health investment faced by individuals. Using data from both the Framingham Heart Study and Avon Longitudinal Study of Parents and Children, Biroli found some evidence that genetic variants related to weight (e.g., FTO) change both the production function of body mass index and the level of health investment. Further work in this direction needs to face the issue that many single genetic variants (FTO is an exception) have very weak effects on outcomes of interest, as discussed previously. The use of PGSs, with their lower level of clarity in interpretation, within structural models is an important direction for future research.
In addition to the many theoretical and methodological uses from further integration of genetics with economics, there are several areas of policy and policy design that economists could opine upon. While the use of PGS measures is gaining momentum in the health and social sciences, these measures also will likely have broad uses outside of academic journals, including clinical and business practices. Indeed, as argued by Conley and Fletcher (2017), PGS will allow a new variety of opportunities for discrimination in a broad set of market and nonmarket interactions. One key piece of legislation that limits their use is the Genetic Information Nondiscrimination Act (GINA) of 2008, which makes it illegal for health insurers and employers to use genetic information in their decisions to insure or hire. However, the narrowness of the law does allow life insurance (and other companies) to discriminate based on genetic risk. While the usefulness of these measures was quite small as recently as a few years ago, advances in predictive algorithms applied to growing data sets (for instance, the website 23andme.com has more than 3 million clients) will substantially increase the usefulness of genetic profiling for many types of companies. A second domain of interest to economists will be how dating and marriage markets are affected by this increase in information. Economists might consider, for example, how precise predictions of future Alzheimer’s disease risk may be used by potential (or current) spouses. A third, and more far-reaching, set of policies will center on the extent to which individuals (and their partners) will be able to use genetic predictions from embryos (preimplantation genetic diagnosis, or PGD) to select the traits of their future children. Economic models of the dating/marriage market and insurance markets, as well as focusing on the economics of the household, will be useful in considering these and other policy issues.
After three waves of integration of genetics within economic analysis, there is much to be optimistic about in the near future. The advantages of continuing this integration are many. Many microeconomic theoretical and empirical models explicitly focus attention on the role that genetics plays in the process under study. This is most obvious in models of intergenerational transmission, the economics of the family, and health economics, but more generally, there is evidence that many (if not all) economic preferences have some genetic basis. The ability to enhance our empirical models is also compelling, especially because genetic data are becoming a core feature of many social science data sets in the United States, including the HRS, the National Longitudinal Study of Adolescent to Adult Health, the Fragile Families study, and many others (Conley, 2015). This is also true in other countries.
The third wave of integration is occurring so quickly that economists without training in genetics may have difficulty understanding the key issues to focus on. Indeed, there are emerging conventions for analysis to note. One question is how to create PGS measures, as many options are available. Key considerations include whether to weigh some genetic variants more than others (based on their effect size, for example), as well as how many genetic variants to include in the PGS. Current conventions are to include all genetic variants (i.e., millions of variables) and to weigh the variants based on the GWAS (Ware et al., 2017), though these conventions might indeed change.
Other key areas include how to pool data from groups of people with different continental ancestries. For a variety of technical reasons (Conley & Fletcher, 2017), PGS measures created using samples of European ancestry do not predict as well in samples of non-European ancestries. Presently, this often has the consequence that much research focuses only on samples of individuals with European ancestry. This issue will likely be resolved over time, as larger samples of other populations are collected and new GWAS are performed. In the meantime, researchers often appeal to the use of a large number of genetic principal components as control variables, even in research that focuses on individuals with European ancestry.
Interested readers would benefit from looking outside of economics, especially at papers from the Integrating Genetics and Social Sciences conference, which have appeared in several special sections of journals (e.g., Boardman & Fletcher, 2011, 2015; Fletcher & Boardman, 2013) and also attempt to collaborate with interested readers with expertise in this third wave of integration.
While there are many advantages of integrating genetics into economics, there are also some issues to consider when moving forward. First, many economists do not have any training in the biological sciences or in genetic analysis. This suggests that there will be some misuses of genetic data in economics. A remedy is to attempt to work with researchers outside of economics, though it has been my experience that few geneticists would be interested in the type of work that economists do. There is a growing set of researchers who have formally or informally acquired some of the tools to integrate these fields, so working with these bridging individuals, as well as considering participation in short courses, now offered regularly by the Russell Sage Foundation and the University of Michigan, among other institutions, are both worthwhile pursuits.
For those considering making these investments, there is the possibility of large rewards, given the current situation that no work using molecular genetics in a microeconomic context has appeared in any of the top five economics journals. A final consideration, and one that applies to a broad set of economic analyses, is an emerging direction that focuses on predictive analysis rather than causal understanding. One promising component of the second wave of integration, which used candidate gene approaches, was the focus on understanding of biological mechanisms within economic analysis. While the paradigm has been revealed to be largely incorrect, with the shift to the use of PGSs and the associated limited interpretation of what these scores mean, part of the original excitement and benefit of integrating genetic findings into economics—of understanding mechanisms—may be short-lived.
The author gratefully acknowledges support from the William T. Grant Foundation, Robert Wood Johnson Foundation Health & Society Program, National Institutes of Health. This article draws from a range of conversations and research, much of it presented over the last eight years at the Integrating Genetics and Social Sciences conference at the University of Colorado‒Boulder, organized by Jason Boardman. I especially thank Jason Boardman, Dalton Conley, Justin Cook, Ben Domingue, Dan Belsky, Dan Benjamin, and Steven Lehrer for valuable conversations that helped shape my thinking on these topics, although any errors are mine. The author gratefully acknowledges the use of the facilities of the Center for Demography of Health and Aging at the University of Wisconsin‒Madison, funded by NIA Center Grant P30 AG017266.
- Domingue, B. W., Belsky, D. W., Fletcher, J. M., Conley, D., Boardman, J. D., & Harris, K. M. (2018). The social genome of friends and schoolmates in the National Longitudinal Study of Adolescent to Adult Health. Proceedings of the National Academy of Sciences, 201711803.
- Linnér, R. K., Biroli, P., Kong, E., Meddens, S. F. W., Wedow, R., Fontana, M. A., . . . Beauchamp, J. (2018). Genome-wide study identifies 611 loci associated with risk tolerance and risky behaviors. bioRxiv, 261081.
- Albert, D., Belsky, D. W., Crowley, D. M., Latendresse, S. J., Aliev, F., Riley, B., . . . Dodge, K. A. (2015). Can genetics predict response to complex behavioral interventions? Evidence from a genetic analysis of the fast track randomized control trial. Journal of Policy Analysis and Management, 34(3), 497–518.
- Ashraf, Q. H., & Galor, O. (2017). The macrogenoeconomics of comparative development. Cambridge, MA: National Bureau of Economic Research.
- Barban, N., Jansen, R., De Vlaming, R., Vaez, A., Mandemakers, J. J., Tropf, F. C., . . . Nolte, I. M. (2016). Genome-wide analysis identifies 12 loci influencing human reproductive behavior. Nature Genetics, 48(12), 1462.
- Barth, D. J., Papageorge, N. W., & Thom, K. (2017). Genetic ability, wealth, and financial decision-making.
- Bearman, P. S. (2013). Genes can point to environments that matter to advance public health. American Journal of Public Health, 103(S1), S11–S13.
- Beauchamp, J. P., Cesarini, D., Johannesson, M., van der Loos, M. J. H. M., Koellinger, P. D., Groenen, P. J. F., . . . Christakis, N. A. (2011). Molecular genetics and economics. Journal of Economic Perspectives, 25(4), 57–82.
- Benjamin, D. J., Cesarini, D., Chabris, C. F., Glaeser, E. L., Laibson, D. I., Age, G., . . . Purcell, S. (2012a). The promises and pitfalls of genoeconomics. Annual Review of Economics, 4, 627–662.
- Benjamin, D. J., Cesarini, D., van der Loos, M. J. H. M., Dawes, C. T., Koellinger, P. D., Magnusson, P. K. E., . . . Johannesson, M. (2012). The genetic architecture of economic and political preferences. Proceedings of the National Academy of Sciences, 109(21), 8026–8031.
- Benjamin, D. J., Chabris, C. F., Glaeser, E. L., Gudnason, V., Harris, T. B., Laibson, D. I., . . . Purcell, S. (2008). Genoeconomics. In National Research Council (US) Committee on Advances in Collecting and Utilizing Biological Indicators and Genetic Information in Social Science Surveys; M. Weinstein, J. W. Vaupel, & K. W. Wachter (Eds.), Biosocial surveys. Washington, DC: National Academies Press. 15. https://www.ncbi.nlm.nih.gov/books/NBK62422/
- Biroli, P. (2015). Genetic and economic interaction in the formation of human capital: The case of obesity. Zurich, Switzerland: University of Zurich.
- Boardman, J. D., & Fletcher, J. M. (2011). Introduction to the special issue on integrating genetics and the social sciences. Biodemography and Social Biology, 57(1), 1–2.
- Boardman, J. D., & Fletcher, J. M. (2015). The promise of integrating genetics into policy analysis. Journal of Policy Analysis and Management, 34(3), 493–496.
- Cesarini, D., Dawes, C. T., Johannesson, M., Lichtenstein, P., & Wallace, B. (2009). Genetic variation in preferences for giving and risk taking. Quarterly Journal of Economics, 124(2), 809–842.
- Cesarini, D., Johannesson, M., Lichtenstein, P., Sandewall, O., & Wallace, B. (2010). Genetic variation in financial decision‐making. Journal of Finance, 65(5), 1725–1754.
- Chabris, C. F., Hebert, B. M., Benjamin, D. J., Beauchamp, J., Cesarini, D., van der Loos, M., . . . Atwood, C. S. (2012). Most reported genetic associations with general intelligence are probably false positives. Psychological Science, 23(11), 1314–1323.
- Conley, D. (2015). Genotyping a new, national household panel study: White paper prepared for NSF-sponsored conference, May 2014. Journal of Economic and Social Measurement, 40(1–4), 375–395.
- Conley, D., & Fletcher, J. (2017). The genome factor: What the social genomics revolution reveals about ourselves, our history, and the future. Princeton, NJ: Princeton University Press.
- Cook, C. J. (2014). The role of lactase persistence in precolonial development. Journal of Economic Growth, 19(4), 369–406.
- Cook, C. J. (2015). The natural selection of infectious disease resistance and its effect on contemporary health. Review of Economics and Statistics, 97(4), 742–757.
- Cook, C. J., & Fletcher, J. M. (2015a). Can education rescue genetic liability for cognitive decline? Social Science & Medicine, 127, 159–170.
- Cook, C. J., & Fletcher, J. M. (2015b). Understanding heterogeneity in the effects of birth weight on adult cognition and wages. Journal of Health Economics, 41, 107–116.
- Ding, W., Lehrer, S. F., Rosenquist, J. N., & Audrain-McGovern, J. (2006). The impact of poor health on education: New evidence using genetic markers. Cambridge, MA: National Bureau of Economic Research.
- Ding, W., Lehrer, S. F., Rosenquist, J. N., & Audrain-McGovern, J. (2009). The impact of poor health on academic performance: New evidence using genetic markers. Journal of Health Economics, 28(3), 578–597.
- DiPrete, T. A., Burik, C., & Koellinger, P. (2017). Genetic instrumental variable (GIV) regression: Explaining socioeconomic and health outcomes in non-experimental data.
- Fletcher, J. M. (2010). Adolescent depression and educational attainment: Results using sibling fixed effects. Health Economics, 19(7), 855–871.
- Fletcher, J. M. (2011a). The medium term schooling and health effects of low birth weight: Evidence from siblings. Economics of Education Review, 30(3), 517–527.
- Fletcher, J. M. (2011b). The promise and pitfalls of combining genetic and economic research. Health Economics, 20(8), 889–892.
- Fletcher, J. M. (2012). Why have tobacco control policies stalled? Using genetic moderation to examine policy impacts. PloS One, 7(12), e50576.
- Fletcher, J. M. (2014). The effects of childhood ADHD on adult labor market outcomes. Health Economics, 23(2), 159–181.
- Fletcher, J. M. (2017). Genetic Endowments, Educational Attainment and Social Mobility. Mimeo.
- Fletcher, J. M. (2019). Environmental bottlenecks on children’s genetic potential for adult socioeconomic attainments: Evidence from a health shock. Population Studies, 73(1), 139–148.
- Fletcher, J. M., & Boardman, J. D. (2013). Integrating work from genetics and the social sciences: An introduction. Biodemography and Social Biology, 59(1), 1–3.
- Fletcher, J. M., & Conley, D. (2013). The challenge of causal inference in gene–environment interaction research: Leveraging research designs from the social sciences. American Journal of Public Health, 103(S1), S42–S45.
- Fletcher, J. M., & Lehrer, S. F. (2009). The effects of adolescent health on educational outcomes: Causal evidence using genetic lotteries between siblings. Forum for Health Economics & Policy, 12(2).
- Fletcher, J. M., & Lehrer, S. F. (2011). Genetic lotteries within families. Journal of Health Economics, 30(4), 647–659.
- Goldberger, A. S. (1979). Heritability. Economica, 46(184), 327–347.
- Grossman, M. (1972). On the concept of health capital and the demand for health. Journal of Political Economy, 80(2), 223–255.
- Hinke Kessler Scholder, S., Wehby, G. L., Lewis, S., & Zuccolo, L. (2014). Alcohol exposure in utero and child academic achievement. The Economic Journal, 124(576), 634–667.
- Lehrer, S. F., & Ding, W. (2017). Are genetic markers of interest for economic research? IZA Journal of Labor Policy, 6(1), 2.
- Norton, E. C., & Han, E. (2008). Genetic information, obesity, and labor market outcomes. Health Economics, 17(9), 1089–1104.
- Okbay, A., Baselmans, B. M. L., De Neve, J.-E., Turley, P., Nivard, M. G., Fontana, M. A., . . . Derringer, J. (2016a). Genetic variants associated with subjective well-being, depressive symptoms, and neuroticism identified through genome-wide analyses. Nature Genetics, 48(6), 624–633.
- Okbay, A., Beauchamp, J. P., Fontana, M. A., Lee, J. J., Pers, T. H., Rietveld, C. A., . . . Meddens, S. F. W. (2016b). Genome-wide association study identifies 74 loci associated with educational attainment. Nature, 533(7604), 539.
- Papageorge, N. W., & Thom, K. (2016, September 7). Genes, education, and labor market outcomes: Evidence from the Health and Retirement Study.
- Pertea, M., & Salzberg, S. L. (2010). Between a chicken and a grape: Estimating the number of human genes. Genome Biology, 11(5), 206.
- Rietveld, C. A., Cesarini, D., Benjamin, D. J., Koellinger, P. D., De Neve, J.-E., Tiemeier, H., . . . Krueger, R. F. (2013a). Molecular genetics and subjective well-being. Proceedings of the National Academy of Sciences, 110(24), 9692–9697.
- Rietveld, C. A., Medland, S. E., Derringer, J., Yang, J., Esko, T., Martin, N. W., . . . Agrawal, A. (2013b). GWAS of 126,559 individuals identifies genetic variants associated with educational attainment. Science, 340(6139), 1467–1471.
- Smith, G. D., & Ebrahim, S. (2005). What can Mendelian randomisation tell us about modifiable behavioural and environmental exposures? British Medical Journal, 330(7499), 1076.
- Taubman, P. (1976a). The determinants of earnings: Genetics, family, and other environments: A study of white male twins. American Economic Review, 66(5), 858–870.
- Taubman, P. (1976b). Earnings, education, genetics, and environment. Journal of Human Resources, 11(4), 447–461.
- Thompson, O. (2014). Economic background and educational attainment: The role of gene-environment interactions. Journal of Human Resources, 49(2), 263–294.
- Ware, E. B., Schmitz, L. L., Faul, J. D., Gard, A., Mitchell, C., Smith, J. A., . . . Kardia, S. L. R. (2017). Heterogeneity in polygenic scores for common human traits. bioRxiv:106062.
- Wehby, G. L., Fletcher, J. M., Lehrer, S. F., Moreno, L. M., Murray, J. C., Wilcox, A., & Lie, R. T. (2011). A genetic instrumental variables analysis of the effects of prenatal smoking on birth weight: Evidence from two samples. Biodemography and Social Biology, 57(1), 3–32.
- Willage, B. (2017). The effect of weight on mental health: New evidence using genetic IVs. Journal of Health Economics, 57, 113–130.
1. If guanine (G) is the most common nucleotide at a specified location, and an individual with genotype GG would have zero minor alleles, a person with a genotype of either GC or CG would have one minor allele, and a person with genotype CC would have two minor alleles.
2. Population stratification refers to cases when the frequencies of genetic variants differ between populations (based on racial group, continental ancestry, or both). For example, if the likelihood of nucleotides TT at a genetic locus are higher in black populations in the United States than alternative nucleotide pairs (TA/AA/AT) at that locus, and if black populations in the United States have higher rates of stroke, a naïve genetic analysis could “find” a “gene for stroke” if population differences in the frequency of TT between blacks and whites are not adjusted in the analysis.
3. The term from genetics is pleiotropy, where genetic variants affect multiple outcomes.