Accepted Manuscript

Title: Genetic and Environmental Influences on HouseholdFinancial Distress

Authors: Yilan Xu, Daniel A. Briley, Jeffrey R. Brown,William G. Karnes, Brent W. Roberts

PII: S0167-2681(17)30225-1DOI: http://dx.doi.org/doi:10.1016/j.jebo.2017.08.001Reference: JEBO 4119

To appear in: Journal of Economic Behavior & Organization

Received date: 26-1-2017Revised date: 5-4-2017Accepted date: 1-8-2017

Please cite this article as: Xu, Yilan, Briley, Daniel A., Brown, Jeffrey R.,Karnes, William G., Roberts, Brent W., Genetic and Environmental Influenceson Household Financial Distress.Journal of Economic Behavior and Organizationhttp://dx.doi.org/10.1016/j.jebo.2017.08.001

This is a PDF file of an unedited manuscript that has been accepted for publication.As a service to our customers we are providing this early version of the manuscript.The manuscript will undergo copyediting, typesetting, and review of the resulting proofbefore it is published in its final form. Please note that during the production processerrors may be discovered which could affect the content, and all legal disclaimers thatapply to the journal pertain.

1

Genetic and Environmental Influences on Household Financial Distress

Yilan Xu*, Assistant professor,

Department of Agricultural and Consumer Economics, University of Illinois at Urbana-Champaign.

Daniel A. Briley, Assistant professor,

Department of Psychology, University of Illinois at Urbana-Champaign.

Jeffrey R. Brown, William G. Karnes Professor,

School of Business, University of Illinois at Urbana-Champaign.

Brent W. Roberts, Professor,

Department of Psychology, University of Illinois at Urbana-Champaign.

Highlights

Financial behaviors are genetically influenced especially at the extremes of SES.

Personality and cognition are linked to financial distress genetically.

Within-family factors also link personality and cognition to financial distress.

Neuroticism is a more important predictor of financial distress at low SES.

Cognitive ability is a more important predictor of financial distress at high SES.

Abstract

Heterogeneity of household financial outcomes emerges from various individual and

environmental factors, including personality, cognitive ability, and socioeconomic status (SES),

among others. Using a genetically informative data set, we decompose the variation in financial

management behavior into genetic, shared environmental and non-shared environmental factors.

We find that about half of the variation in financial distress is genetically influenced, and

personality and cognitive ability are associated with financial distress through genetic and

* Yilan Xu, corresponding author. Address: 1301 W. Gregory Dr., 309 Mumford Hall, Urbana, IL 61801. Phone:

217-300-0465. Fax: 217-333-5538. Email: yilanxu@illinois.edu. The research project was partially funded by the

National Institute of Food and Agriculture (NIFA) at the United States Department of Agricultural (#ILLU-470-367).

The authors thank Jing Luo for excellent research assistance.

2

within-family pathways. Moreover, the genetic influences of financial distress are highest at the

extremes of SES, which in part can be explained by neuroticism and cognitive ability being more

important predictors of financial distress at low and high levels of SES, respectively. (JEL code:

D14, D31, G31)

Keywords: household finance, personality traits; cognitive ability, socioeconomic status,

behavior genetics.

3

1. Introduction

Recent studies have shown that individual differences in financial decisions, such as saving rates

or portfolio allocations, are considerably heritable. For instance, between a quarter and almost

half of the variability in financial behavior can be explained by variability of genetic endowment

across individuals (Barnea, Cronqvist, & Siegel, 2010; Cesarini, Johannesson, Lichtenstein,

Sandewall, & Wallace, 2010; Cronqvist & Siegel, 2014; 2015). An implication of this finding is

that earlier work on the determinants of household financial decisions may have overlooked

possible endogeneity arising from shared genetic influences across risk factors. It has been long-

established, for example, that family socioeconomic status (SES) has protective effects against

adverse financial outcomes, yet it has never been tested whether such effects minimize or

magnify genetic effects on financial outcomes. Moreover, the studies that have found a genetic

basis for financial behavior have not identified plausible mechanisms that may explain genetic

influences on financial decisions. One potential reason why financial decisions show genetic

influences is that financial decisions reflect the influences of other variables that are themselves

genetically influenced. For example, cognitive and non-cognitive abilities predict earnings and

wealth (Duckworth & Weir, 2010) and depression predicts risk-taking behaviors (Calvet &

Sodini, 2014), all of these factors have been shown to be genetically influenced in the

psychology literature (Bouchard & McGue, 2003; McGue & Christensen, 2003). Nevertheless,

such a hypothesis has not been tested in a genetically informative, multivariate analysis. Our

study advances this literature by combining genetically informative data and data on financial

behaviors to further investigate these relationships with special attention to the roles of SES,

personality, and cognitive ability.

Our outcome of interest is the competence of household financial management, as measured by

an estimate of a latent factor common to various indicators of a household’s difficulty managing

basic finances. We first estimate the heritability of financial distress, i.e., the portion of the

variance in the latent variable that can be explained by genetic variance. We then study the

genetic and environmental nature of the relation of the latent financial distress with the Big Five

personality traits and cognitive ability. The Big Five personality traits reflect patterns of thoughts,

feelings, and behavior that are relatively stable across time and context. Both personality and

cognitive ability may be linked with financial distress through genetic and environmental

4

pathways. Finally, we investigate the extent to which genetic and environmental influences of

financial distress are similar across different environmental contexts.

Using a sample from the National Longitudinal Survey of Adolescent to Young Adult (Add

Health) that includes both genetic and financial information, we apply a behavior genetic

decomposition to analyze financial behaviors. We identify additive genetic influences by using

variation in the observed similarity of groups with different degrees of genetic relatedness:

identical twins (100% genetically related), fraternal twins and full siblings (50% genetically

related among segregating genetic material on average), and half siblings (25% related with a

similar caveat). Beyond genetic influences, our analyses also estimate the variances due to the

shared and the non-shared environment. Shared environmental influences result from between-

family effects that make siblings living in the same home behave similarly, and non-shared

environmental influences result from within-family effects that make siblings less similar,

including idiosyncratic variances.

The empirical results suggest that 43-55% of the variance in latent financial distress is due to

genetic influences, and this result remains robust when the effects of age, sex, race/ethnicity, and

family background are included. We also investigate personality and cognitive ability as

potential mediators for the genetic component of financial distress, where the Big Five

personality traits is 18 - 41% heritable and cognitive ability is 32 - 45% heritable. The remaining

variance in financial distress and personality are minimally associated with shared environmental

effects, in contrast to substantial shared environmental effects on cognitive ability. The pathway

analysis finds that conscientiousness is associated with financial distress mainly through a non-

shared environmental pathway -- 68% of the correlation between conscientiousness and financial

distress is attributable to the non-shared environment. Neuroticism is associated with financial

distress equally through the genetic pathway (52%) and non-shared environmental pathway

(48%). Agreeableness is associated with financial distress solely through a genetic pathway.

Finally, cognitive ability is associated with financial distress through both genetic and non-

shared environmental pathways. In total, the three personality traits and cognitive ability are able

to account for 21.32% of the variance in financial distress through genetic pathways and 10.01%

of the variance through non-shared environmental pathways, leaving significant residual genetic

(33.52%) and non-shared environmental (35.16%) variance.

5

Although we find minimal evidence of shared environmental influences on financial distress in

our baseline model, this result could reflect the interactions between genes and shared

environments. We find that the heritability of financial distress is not constant across

environmental context: genetic influences account for a greater proportion of variance at

extremes of the SES distribution. When this trend is not explicitly modeled, the variance

contributes to estimates of genetic influences (described below and in Purcell, 2002). Evidence

suggests that financial distress is associated with different risk factors at the extremes of the SES

distribution, with neuroticism playing a key role at the low end and cognitive ability exerting

influences at the high end. After we account for the effects of these two risk factors, the genetic

influences of financial distress become similar across the SES.

Our study makes three advances in our understanding of genetics and financial behavior. First,

we employ a multivariate framework to examine multiple financial behaviors under a uniform

structure. Using a latent variable approach, we identify variance common to various household

financial management behaviors, which limits the effects of idiosyncratic shocks and

measurement error specific to single indicators. Although day-to-day household financial

management behaviors have not been the primary focus in the finance literature, the method can

be applied to find a common factor among other financial behaviors such as investment and

retirement saving. Second, we relate financial distress to cognitive and non-cognitive abilities

through genetic and environmental pathways. Multivariate analysis suggests that the genetic

components of three personality traits and cognitive ability overlap with almost half (48%) of the

genetic component of the financial distress. This is evidence that these psychological

characteristics could mediate the genetic influences on financial distress. As shown in an earlier

study, personality can predict financial distress in addition to the effects of household income,

health, and childhood experience (Xu, Beller, Roberts, & Brown, 2015). Hence, the genetic

component of personality is likely to reflect the predisposition to make financial decisions in a

certain pattern beyond the influences of liquidity shocks and early-life experience. Finally, we

show that genetic influences are not fixed for financial distress. We highlight the moderating

effects of SES in explaining the heritability of financial distress, and we identify different risk

factors for populations at different distributions of the SES.

6

2. Relevant Literature

2.1 Personality, Cognitive abilities, and Household Finance

Research has shown that both personality and cognitive abilities contribute to important life

outcomes including financial outcomes (Borghans, Duckworth, Heckman, & Weel, 2008;

Duckworth, Weir, Tsukayama, & Kwok, 2012; Kautz, Heckman, Diris, Weel, & Borghans,

2014). For instance, cognitive ability and personality traits equally predict lifetime earnings for

the household sample covered by the Health and Retirement Survey (Duckworth & Weir, 2010),

and both cognitive ability and personality traits explain strategic behaviors even when investor

risk aversion is controlled for (Rustichini, DeYoung, Anderson, & Burks, 2016). In particular,

higher conscientiousness is associated with higher earnings (Nyhus & Pons, 2005), less

borrowing and more saving (Nyhus & Webley, 2001), less spending out of income during the

recession (Duckworth & Weir, 2011), less financial distress in young adulthood (Donnellan,

Conger, McAdams, & Neppl, 2009; Xu et al., 2015), and more assets in both young adulthood

(Letkiewicz & Fox, 2014) and old age (Duckworth & Weir, 2010). Moreover, high IQ predicts

higher likelihood of stock market participation and better investment performance (Grinblatt,

Keloharju, & Linnainmaa, 2011). High numeracy, a component of cognitive ability, predicts

high retirement savings (Banks, O Dea, & Oldfield, 2010), low mortgage default (Gerardi,

Goette, & Meier, 2013), and lower chance of financial mistakes (Agarwal & Mazumder, 2013).

One important aspect of household finance is managing cash flow to make ends meet. Sound

management of cash flow can help households sustain economic shocks such as unemployment

and housing market collapse. Households that fail to do so are referred to as financially

vulnerable, financially fragile, having economic hardship, or processing low financial capability.

Financial distress associated with failure in cash management can be persistent throughout one’s

lifetime (S. Brown, Ghosh, & Taylor, 2012). Conceptually, financial distress can result from an

array of reasons other than income and wealth, such as unsustainable borrowing, poor

management, adverse life events, and absence of financial instruments (Anderloni, Bacchiocchi,

& Vandone, 2012; Donnellan et al., 2009). To a certain extent, these financial management

behaviors reflect characteristics beyond financial knowledge and earning abilities, such as

7

“managing money”, “planning ahead”, “making choices” and “getting help” (Atkinson, McKay,

Kempson, & Collard, 2006). Such characteristics may be captured by certain personality traits.

For instance, conscientiousness is negatively and neuroticism is positively associated with young

adults’ financial distress even after controlling for early-life background, math skills, health, and

household income (Xu et al., 2015).

A set of similar questions has been used in several national and regional surveys to gauge the

financial distress of households. For instance, the National Survey of America’s Family (NSAF)

included survey questions about difficulty paying bills, skipping meals due to lack of money,

going without phone service for at least one month, and postponing medical care, which can be

used to measure households’ financial standing (Melzer, 2011). The National Financial

Capability Study (NFCS) asked whether the household could make ends meet, which could be an

indicator of financial capability (Mottola, 2014). The Iowa Youth and Families Project asked

survey questions about unmet material needs, unmet financial obligations, and financial cutbacks

that were used to construct measures of economic pressure (Donnellan et al., 2009). Several

surveys in other countries and regions also contained questions about financial distress. For

instance, the British Household Panel Survey asked about housing payment problems, financial

problems that required borrowing, financial problems that required cutbacks, and whether

household has been at least 2 months in arrears in last 12 months (M. Taylor, 2011; M. P. Taylor,

Jenkins, & Sacker, 2011). Questions in the same survey about housing payment difficulties,

evictions, and repossessions were used to measure financial hardship (S. Brown et al., 2012).

The financial vulnerability of Italian households was measured by behind payments for utility

bills, rent, mortgage, and other bills, as reported in a national survey in Italy (Anderloni et al.,

2012). We take a similar approach as found in this past literature to investigate financial distress

by analyzing several items centering on the ability of a survey respondent to make ends meet.

2.2 Genetic Influences on Financial Behaviors, Personality and Cognitive Ability

The behavior genetics literature has shown that genetic variation explains a considerable share of

individual differences in traits, abilities, and behaviors (Briley & Tucker-Drob, 2014; Tucker-

Drob & Briley, 2014; Turkheimer, 2000). Recently, a growing literature has shown that a

considerable share of the variance in financial decisions and economic preferences can be

8

explained by genetic effects. For example, genetic variation explains about one-third of the

variation in the stock market investment decisions (Barnea et al., 2010) and saving behavior

(Cronqvist & Siegel, 2015), and about a quarter of investors’ portfolio choice (Cesarini et al.,

2010) and housing locational choices (Cronqvist, Münkel, & Siegel, 2012). Investment biases,

such as lack of diversification, excessive trading, and the disposition effect, are 20-25% heritable

(Cronqvist & Siegel, 2014), so are behavioral biases such as the conjunction fallacy, default bias,

and loss aversion (Cesarini, Johannesson, Magnusson, & Wallace, 2012). Many economic

preferences are heritable. For instance, time preference (Anokhin, Golosheykin, Grant, & Heath,

2011), risk preference (Cesarini, Dawes, Johannesson, Lichtenstein, & Wallace, 2009a; Zhong et

al., 2012; Zyphur, Narayanan, Arvey, & Alexander, 2009), fairness preference (Wallace,

Cesarini, Lichtenstein, & Johannesson, 2007), cooperativeness (Cesarini et al., 2008),

overconfidence (Cesarini, Lichtenstein, Johannesson, & Wallace, 2009b), and giving (Cesarini et

al., 2009a).

Ample literature has documented that a considerable portion of the variance in personality traits

are due to genetic factors (South, Reichborn-Kjennerud, Eaton, & Krueger, 2012). For instance,

evidence from twin samples of different nationalities suggests that the genetic variation explains

38%-53% of the variation in conscientiousness and 41-52% of the variation in neuroticism

(Bouchard & Loehlin, 2001). Such genetic influences are found in both cross-sectional and

longitudinal settings (Kandler, 2012; McGue, Bacon, & Lykken, 1993; McGue, Elkins, Walden,

& Iacono, 2005) and have been identified using molecular genetic approaches in addition to twin

and family approaches (e.g., Okbay, Beauchamp, Fontana, Lee, & Pers, 2016).

Cognitive ability refers to a suite of skills required to solve complex problems or perform mental

operations. Genetic influences on cognitive ability are also well-established, with estimates

ranging from approximately 40% of the variance up to 80% (Bouchard, 2014). Molecular genetic

evidence is also strong for cognitive ability (e.g., Benyamin et al., 2013; Okbay et al., 2016).

Heritability tends to increase with age (Haworth, Wright, et al., 2009b), primarily due to stable

genetic influences accounting for increasing proportions of variance (Briley & Tucker-Drob,

2013a). This result is consistent with individuals exerting increasing influence over the types of

environments and experiences that are selected which reinforce psychological characteristics

(Scarr & McCartney, 1983; Tucker-Drob & Briley, 2014).

9

2.3 Pathway Analysis

The genetic influences on the financial behaviors may be attributable to underlying

characteristics that are themselves influenced by genetic factors. For instance, saving and

investment behaviors are associated with time and risk preferences, which have been shown to

be genetically influenced (Cesarini et al., 2009a; Zhong et al., 2012; Zyphur et al., 2009).

However, such pathways have rarely been tested to determine the relative importance of genetic

and environmental influences in mediating the correlation between any two constructs. One

study shows that investment biases are genetically correlated with education (Cronqvist & Siegel,

2014). This indicates that there are common genetic influences on both financial behavior and

educational success. Another study shows that saving behaviors and obesity are genetically

correlated, leading to the conjecture that the genetic influences on saving are mediated by self-

control (Cronqvist & Siegel, 2015). However, education and obesity are also correlated with

earnings and thus obscures the interpretation. Another study shows that occupational choices are

genetically influenced, and the genetic factors that influence entrepreneurship also influence the

tendency to be self-employed (Nicolaou & Shane, 2010). In psychology, direct measures of the

underlying characteristics are usually used as the mediator for the genetic influences. For

instance, cognitive abilities are genetically influenced, and genetic influences on cognitive ability

explain part of the variance in scholastic achievement (Thompson, Detterman, & Plomin, 1991).

The high heritability of academic achievement reflects genetic influences through not just

intelligence but also personality, self-efficacy, and behavior problems (Krapohl et al., 2014).

Similarly, personality predicts disordered gambling (Slutske, Cho, & Piasecki, 2013) and

subjective Well-Being (SWB) (Keyes, Kendler, Myers, & Martin, 2015) through genetic

pathways, and mastery also predicts alcohol dependence through the genetic pathway (Kiecolt,

Aggen, & Kendler, 2013). In this paper, we apply similar bivariate and multivariate methods to

decompose the genetic and environmental pathways between psychological factors and financial

distress.

2.4 The Interaction between Genes and Environments

10

The economics literature has been discussing the relative importance of “nature versus nurture”,

which has been mostly informed by studies on adoptees. For instance, some studies use adoptees’

adoptive parents as a proxy for the nurturing environments and their biological parents as a

proxy for the genetic endowment (Black, Devereux, Lundborg, & Majlesi, 2015; Fagereng,

Mogstad, & Rønning, 2015; Sacerdote, 2002; 2007). However, this method does not account for

interactions between genes and environments. Environmental context can moderate genetic

influences, a phenomenon referred to as Gene × Environment interactions. As Purcell (2002, p.

555 for mathematical proofs) notes, unmodeled Gene × Environment interaction exerts a

predictable influence on behavior genetic decompositions. When the interaction occurs with a

shared environment, this process results in additive genetic variance due to the fact that more

genetically related individuals respond to the environment in a similar manner, but less

genetically related individuals respond differently. Put differently, Gene × Shared Environment

interaction could mask the apparent influence of family-level environmental effects when not

explicitly modeled. When the interaction occurs with a non-shared environment, this process

results in non-shared environmental variance due to the fact that this process magnifies

differences between even genetically identical individuals.

The Gene × Environment interaction can take several different forms, which have implications

for the possible developmental mechanisms that generate the effect (Roisman et al., 2012).

Previous work has found evidence for larger estimates of heritability in advantaged environments

(e.g., cognitive ability; Tucker-Drob & Bates, 2016), larger estimates in disadvantaged

environments (e.g., psychopathology; Dick et al., 2007), and still others find heightened

heritability at both extremes of the environment (e.g., general health; South & Krueger, 2013).

One possible implication of the Gene × Environment interaction is that different characteristics

play a larger role at specific regions of the distribution of the environment. The previous

literature provides some evidence for differential associations across socioeconomic status.

Specifically, within the Add Health study, prior research has found support for the resource

substitution hypothesis which states that resources will have more beneficial effects among

people with fewer alternative resources (Mirowsky & Ross, 2003). Personality traits are stronger

predictors of educational attainment at low ends of family socioeconomic status (Shanahan,

Bauldry, Roberts, Macmillan, & Russo, 2014). For instance, openness to experience has a

positive effect and extraversion has a negative effect on college graduation for less-advantaged

11

men but not for men from the high-educated households (Lundberg, 2013). Similar support for

the resource substitution effect for personality has been found at the low end of family

socioeconomic status when predicting adult educational attainment and occupational outcomes

(Damian, Su, Shanahan, Trautwein, & Roberts, 2015). In contrast, the Matthew effect (i.e., the

rich get richer) has been demonstrated for the association between cognitive ability and

educational and occupational attainment (Damian et al., 2015). The effect of cognitive ability

tends to be largest for individuals that come from wealthy families. If the genetic and

environmental composition of the association between a trait and an outcome remains fairly

constant across socioeconomic status, the stronger association at the extremes of the SES may

imply shifts in the heritability of the outcome across levels of family SES.

In the particular case of financial distress, socioeconomics advantage may buffer one from

experiencing economic hardship, reducing heritability at high levels of SES. On the other hand,

individuals from advantaged backgrounds may have more freedom to pursue economic pathways

in line with their genetically influenced dispositions, increasing heritability at high levels of SES.

Turning toward economic adversity, this disadvantaged environment may be strong enough to

overshadow genetically influenced individual differences, leading to low heritability at low

levels of SES; or alternatively, some individuals may possess personal characteristics (e.g.,

conscientiousness) that buffer against adversity, resulting in high heritability at low SES,

assuming the link between personal characteristics and economic characteristics occurs through a

genetic pathway. Because we are aware of no previous research that has explicitly tested these

hypotheses, we conduct exploratory analyses examining whether heritability may differ across

environmental context.

3. Methodology

3.1 Measuring Financial Distress: A Common Factor Analysis

In the literature, indicators of financial hardship experienced in the past 12 months are often used

to measure financial distress; however, each individual binary measure usually has considerable

measurement error. Several studies have used the common factor analysis for financial

12

management (M. Taylor, 2011; M. P. Taylor et al., 2011) and mental well-being (Keyes, Myers,

& Kendler, 2010) to minimize the measurement error. For instance, financial capability was

measured using principal component analysis from the following indicators in the British

Household Panel Survey: current financial situation, financial situation worsened since last year,

whether the household saves, whether the household has housing payment problems, problems

required borrowing, problems required cutbacks, and been at least 2 months in arrears in last 12

months (M. Taylor, 2011; M. P. Taylor et al., 2011). The common factor analysis has several key

strengths. First, gradients of financial hardship can be identified by combining information from

multiple items. Second, aggregating over items potentially reduces the influences of

idiosyncratic shocks that may only affect one aspect of household finance. Finally, our research

hypotheses relate to the ability of families to make ends meet, which is not specific to any one

aspect of finance. Admittedly, the difficulties to make ends meet are more likely to be

experienced by low-income families, yet the common factor leading to theses difficulties can

reflect general financial management abilities pertaining to other financial behaviors, above and

beyond financial knowledge and earning abilities. For instance, it may reflect qualities such as

“managing money”, “planning ahead”, “making choices” and “getting help” (Atkinson, McKay,

Kempson, & Collard, 2006), all of which can be important to other financial behaviors such as

stock investment and retirement saving.

We therefore use confirmatory factor analysis to estimate a latent factor from four indicators of

financial distress: missing utility payments, going without phone service for financial reasons,

being past due on mortgage or rent, and worrying about food depletion. The latent factor reflects

the variance common to the four financial distress items, indicating a latent propensity to

experience financial distress. We then test whether the common variance of the four financial

distress indicators can be explained by the genetic effects and the unique variances can be

explained by non-shared environmental factors, i.e., the generalist genes hypothesis (Haworth,

Kovas, et al., 2009a).

3.2 A Behavior Genetics Method

We apply a behavior genetic approach that infers the relative importance of the genes and

environments from the variance-covariance between siblings with different degrees of genetic

13

relatedness. The identical (MZ) twins share 100% of their genes, the fraternal (DZ) twins and

full siblings (FS) share 50% of their segregating genetic material on average, and half siblings

(HS) share 25% of their segregating genetic material on average. Thus, the variation of a

construct can be decomposed into an additive genetic component (A), the shared or common

environment (C), and the non-shared environment (E). More generally, the shared environment

represents any between-family effects that lead to phenotypic similarity of siblings raised in the

same home, and the non-shared environment represents any within-family effects that lead to

phenotypic dissimilarity. In a univariate case, we can compute the share of co-twin variance in a

characteristic that can be attributable to the genetic component, i.e., the heritability index.

Similarly, the influences of the shared environment, and the influences of the non-shared

environment can be quantified. In a bivariate or multivariate case, i.e., a pathway analysis, we

can compute the genetic correlation and bivariate heritability. The former reflects the extent to

which genetic influences on one characteristic are shared with another characteristic, while the

latter indicates the extent to which the observed association is due to a genetic pathway.

Interpretation of environmental correlations and bivariate environmentality are similar.

Appendix B provides more detailed information about the quantitative genetic method.

3.3 Testing Gene × Environment interaction

As we are aware of few previous studies to examine the varying heritability of financial behavior

across SES, we do not have strong a priori predictions. Because model misspecification can

dramatically alter interpretation of Gene × Environment interaction results, we initially use the

nonparametric LOSEM technique (Briley, Harden, Bates, & Tucker-Drob, 2015) to guide model

specification. Based on the LOSEM results, we use parametric models of Gene × Environment

interaction for a latent factor with binary indicators (see Bauer, 2016 for a similar approach to

measurement invariance). Parametric models tend to be slightly more powerful tests of Gene ×

Environment interaction, but only when the functional form is accurate. Whereas LOSEM

estimates local heritability and environmentality across the moderator, parametric approaches

specify genetic and environmental influences to vary as a function of the moderator (Purcell,

2002).

To capture the trends found with LOSEM, we regress the latent financial distress variable 𝐹 on a

quadratic form of SES as the following.

14

𝐹 = 𝑏1 × 𝑆𝐸𝑆 + 𝑏2 × 𝑆𝐸𝑆2 + 𝜖 (1)

Then we specify the genetic and environmental influences on the residual latent financial distress,

𝜖, to the following function:

𝐴 = 𝑎 + 𝑎′ × 𝑆𝐸𝑆 + 𝑎′′ × 𝑆𝐸𝑆2

𝐶 = 𝑐 + 𝑐′ × 𝑆𝐸𝑆 + 𝑐′′ × 𝑆𝐸𝑆2 (2)

𝐸 = 𝑒 + 𝑒′ × 𝑆𝐸𝑆 + 𝑒′′ × 𝑆𝐸𝑆2

We then estimate the parameters 𝑏1, 𝑏2, 𝑎, 𝑎′, 𝑎′′, 𝑐, 𝑐′, 𝑐′′, 𝑒, 𝑒′, 𝑒′′. Based on these estimates,

we plot trends in proportions of variance due to genetic and non-shared environmental factors

across SES.

4. Data and Descriptive Analysis

4.1 The Add Health Data

The genetically informative data for this study are from the National Longitudinal Survey of

Adolescence Health (Add Health). We construct a sibling sample using the Add Health sibling

relationships file. The sample includes 306 MZ twin sibling pairs, 437 DZ twin sibling pairs,

1162 full sibling pairs, and 327 half sibling pairs. In the initial wave in 1994/95, the Add Health

surveyed a nationally representative sample of adolescents in 7th to 12th grades. The

consequences surveys followed in 1995/96 (Wave II), 2001/02 (Wave III), and 2008/09 (Wave

IV). We construct four distinct measures of financial distress from the Wave IV survey (January

2008–February 2009), including missing utility payments, going without phone service for

financial reasons, being past due on mortgage or rent, and worrying about food depletion (Xu et

al., 2015).

The Add Health contains measures for non-cognitive and cognitive abilities. The “Big Five”

personality traits, i.e., conscientiousness, extraversion, neuroticism, agreeableness, and openness

to experience, were measured in Wave IV using a 20-item short-form version of the International

Personality Item Pool-Five-Factor Model (i.e., the Mini-IPIP) (Donnellan, Oswald, Baird, &

Lucas, 2006). The survey questions used to define the financial distress indicators and the

personality traits are listed in Table 1. Cognitive ability was measured in both Wave I (April

15

1995–December 1995) and Wave III (July 2001–April 2002) using a computerized, abridged

version of the Peabody Picture Vocabulary Test—Revised (i.e., the Add Health Picture

Vocabulary (PVT)). We measure cognitive ability by the cross-sectional standardized PVT

scores at Waves I and III. A total of 74 participants received PVT scores more than 3 standard

deviations below the mean. These observations are treated as missing to reduce the influence of

outliers.

We construct a measure of family socioeconomic status (SES) based on a composite of

household income at wave I, mother’s years of education, and father’s years of education where

available. Household income was log-transformed and standardized. Mother and father years of

education are averaged and standardized. The Add Health data also include demographic

information such as sex, age, and race. We restrict the sample to siblings whose ages are no more

than three years apart in the baseline sample. We drop full siblings who reported dates of birth

less than nine months apart. Some observations may have missing values for some of the

variables, and there are cases where one sibling of the pair has the value for a variable while the

other member of the pair has a missing value. We use the Full Information Maximum Likelihood

method to make use of all available information in the estimation.

4.2 Determination of Twin Type (Zygosity)

Zygosity is a measure of the genetic relation between twins. The zygosity of Add Health twins is

corrected by the DNA information collected in Wave III. We determine the zygosity based on

three sets of information. First, the twins self-reported their zygosity in the Wave I in-home

survey. This self-report zygosity can be inconsistent within a twin pair or missing. Second, twins

reported the confusability of their appearance based on a series of questions such as “When you

were young children, did you and {NAME} look very much alike, like two peas in a pod, or did

you just look like members of the same family?” where NAME is the name of the twin sibling,

and “Are strangers ever confused about which of you is which?” “Are your teachers ever

confused?” “Are family members ever confused?” A similarity scale was constructed as the

average of the self-reported confusability of appearance. Third, the DNA diagnoses of zygosity

are available for a subsample of twins in Wave III (Harris, Halpern, Smolen, & Haberstick,

2006).

16

The Add Health identified the zygosity (MZ, DZ, UD) based on the scale of similarity and self-

reported zygosity. If the self-reported appearance was missing, the classification was based on

mother’s report of confusability. The similarity scale was cross-checked with the self-reported

zygosity. If both twins reported DZ type while the similarity scales are high, Add Health

assigned those as undetermined (UD) for undecided. In addition, we obtain the DNA data and

use the DNA diagnoses to supersede the Add Health classification whenever DNA diagnoses are

available. We drop the pairs who remain undetermined after the correction based on DNA.

4.3 Descriptive Analysis of the Add Health Sibling Sample

Table 2 reports the summary statistics for the sample by sibling type. On average, about 10% MZ

twins missed their utility bills, 7% had no phone service, and 5% missed their rent or mortgage

payment, and 8% worried about food depletion. Half siblings (HS) show the highest likelihood

of having these indicators of financial distress, followed by full siblings (FS) and DZ twins. The

four sibling groups show similar levels of personality traits, with the average scores between 14

and 15 except neuroticism, which has an average score around 10. The four groups also have

similar sex and age, but racial differences exist. The full siblings have disproportionally more

Asian and the half siblings have disproportionally more blacks. However, these differences are

fairly modest, and did not substantially impact model fit for subsequent behavior genetic

analyses.

Table 3 shows the pairwise correlation for each sibling type. The four sibling types differ in their

genetic relatedness: MZ twins share 100% of their genes, DZ twins and FS share 50% of their

segregating genetic material on average, and HS share 25% of their segregating genetic material

on average. If the pairwise correlation of a construct decreases as the genetic relatedness

decreases, it is evidence that genetic differences play a role in explaining the variation of the

construct. In terms of the financial distress indicators and personality traits, the MZ twins show

the highest pairwise correlations in general, and the correlation coefficients are significant at 1%

level except for no phone service (p = 0.054) and food depletion (p = 0.027). The DZ twins show

statistically significant pairwise coefficients only for selected variables, and the correlations are

lower than for MZ twins. The FS show lower statistically significant pairwise correlations than

17

for the MZ twins in general. The pairwise correlations for the HS are mostly insignificant except

for agreeableness. In terms of cognitive ability, the pairwise coefficients decrease with the

genetic relatedness, with the MZ twins having about 0.75 correlations as the highest and the HS

having 0.50 correlations as the lowest. The pairwise correlations across sibling type are higher

for the PVT score measured in Wave I than in Wave III, suggesting the importance of the shared

environmental component decreases as the adolescents grow into young adulthood. In sum, the

patterns in the pairwise correlations suggest genetic component in the variation of financial

distress, personality traits, and cognitive ability.

5. The Heritability of Financial Distress

5.1 Heritability and Common Factor of Financial Distress

We begin by decomposing the heritability and environmentality of the four indicators of

financial distress. Table 4 reports the results from the univariate ACE model. The estimates and

the confidence intervals for the relative importance of the additive genetic component (𝐴

𝐴+𝐶+𝐸),

the shared environment (𝐶

𝐴+𝐶+𝐸), and the non-shared environment (

𝐸

𝐴+𝐶+𝐸) are reported in the

table. The results show that the indicators of financial distress are considerably genetically

influenced. Missing utility bill is 28.6% heritable, no phone service is 24.8% heritable, missing

mortgage/rent payment is 37.2% heritable, and worrying about food depletion is 49.2% heritable.

The shared environment explains a small share of the variance in the financial distress indicators,

ranging from less than 1% for to 17.9%. More than half of the total variance of the financial

distress indicators is explained by the non-shared environments. The results are in line with the

findings in the literature that financial outcomes are a quarter to one third heritable.

Next, we show that there exists a single, heritable latent factor that statistically accounts for

common variance of the indicators. Figure 1a illustrates our baseline structural model. A latent

factor captures the variance common to each indicator of financial distress, and this common

variance is decomposed into ACE components. Figure 1b demonstrates the result from a

trimmed model where we drop the insignificant pathways for the individual financial distress

18

indicators. The latent factor that explains variance common to the financial distress indicators is

fairly strong – the factor loadings are 0.90, 0.81, 0.86, and 0.81, respectively for missing utility

payments, going without phone service for financial reasons, being past due on mortgage or rent,

and worrying about food depletion.

Variance in latent financial distress is attributable 43% to genetic effects and 53% to non-shared

environmental effects (both p’s < 0.001). The shared environment accounts for the remaining 4%

of the variance, but this effect is not statistically significant. The heritability is higher than that of

individual indicators of financial distress potentially because measurement error is largely

contained in the item-specific environmental variance. After taking genetic influences common

to each indicator into account at the latent variable level, no residual genetic effects on the

indicators are found except for a small and insignificant genetic effect specific to food depletion.

Trimming insignificant pathways from the model does not produce a significant decrement in

model fit (Δχ2[9] = 7.83, p = .55; ΔCFI = .002). The more parsimonious trimmed model (Figure

1b) implies that all genetic influences on financial distress occur at the level of common variance

with only non-shared environmental variance remaining at the level of the specific indicators.

This is evidence that the latent variable captures the underlying genetic construct that regulates

the financial distress.

5.2 Robustness Checks

Before turning to predictors of genetic influences on financial distress, we are interested in

testing the robustness of the heritability of financial distress. Specifically, we examine whether

genetic and environmental effects on latent financial distress manifest differently for subgroups

in our data. First, we test whether including full and half-siblings in the analyses alters results.

Siblings, unlike twins, differ in age, and half-siblings may come from disproportionately

disadvantaged backgrounds, which could bias results. Further, the inclusion of opposite-sex

siblings may alter results. Using only MZ and same-sex DZ twins, we estimated the model in

Figure 1b. This model fit the data well (RMSEA = .025, CFI = .986) and produced similar

variance component estimates (41% genetic effects and 59% non-shared environmental effects,

both p < .001). To further evaluate possible sex-differences, we estimated the model for males

and females separately (ignoring all opposite-sex pairs). In males, the model fit the data well

19

(RMSEA = .029, CFI = .986) and produced similar variance component estimates (41% genetic

effects and 59% non-shared environmental effects, both p < .001). The result was similar for

females, but with slightly larger estimate of genetic effects (RMSEA = .067, CFI = .953; 70%

genetic effects and 30% non-shared environmental effect, both p < .001). Next, we examine

whether racial/ethnic effects may bias our result and estimate the model for only White

participants. This model fit the data well (RMSEA = .049, CFI = .964) and produces similar

variance component estimates (47% genetic effects and 53% non-shared environmental effects,

both p < .001). Finally, we test whether results differ when examining only participants that

resided with their biological parents as an indicator of potential economic disadvantage. This

model fit the data well (RMSEA = .038, CFI = .971) and produces similar variance component

estimates (52% genetic effects and 48% non-shared environmental effects, both p < .001). Thus,

our primary result holds in the most restrictive sibling subset, among both men and women,

among majority group members, and among participants living with their biological parents.

The shared environment appears to have a trivial influence on financial distress in all robustness

cases discussed above. These results raise an interesting question: given that parental wealth

should be protective against experiencing financial distress, why do we consistently find

negligible shared environmental estimates? There are a few possibilities. Although parental

wealth is a between-family variable, the effects of family wealth may differentially influence

siblings. Thus, factors that are objectively shared across siblings may result in within-family

differences that manifest as non-shared environmental effects (Turkheimer & Waldron, 2000).

Alternatively, children may respond differently to a shared environment on the basis of

genetically influenced characteristics. For example, children with genetically influenced

“resiliency” traits may be less sensitive to environmental adversity. This example represents

Gene × Environment interaction. When such an interaction occurs between genetic influences

and a shared environment, this variance appears as a genetic effect in quantitative genetic models

(Purcell, 2002). We now turn to considering economic disadvantage in more detail.

6 Pathways Between Personality, Cognitive Ability, and Financial Distress

6.1 Heritability of Personality and Cognitive Ability

20

In this section, we estimate the heritability and environmentality for each of the Big Five

personality traits and the cognitive ability measured at two waves. The results suggest that

conscientiousness is 22.5% heritable, neuroticism is 18% heritable, extraversion is 28.3%,

agreeableness is 41.4%, and openness to experience is 38.4% (Table 4). The shared environment

does not explain variance in the personality traits except for neuroticism (1.1%). The non-shared

environment explains most of the variance, ranging from 58.6% for agreeableness to 80.9% for

neuroticism. The heritability estimates are lower than shown in the psychology literature. This is

likely because of the relatively short 20-item version of the Mini-IPIP.

Cognitive ability has a high heritability: the PVT score measured in Wave I is 32.4% heritable,

and the PVT score measured in Wave III is 44.8% heritable (Table 4). This is consistent with the

findings in the literature that the heritability of the cognitive ability increases as individuals leave

home and increasingly select cognitively-relevant environments aligned with genetically

influenced preferences (Briley & Tucker-Drob, 2013b; Scarr & McCartney, 1983). Unlike

financial distress indicators and personality traits, cognitive ability is considerably influenced by

the shared environment, 41.5% for the Wave I PVT score and 27.5% for the Wave III PVT score.

Among the characteristics we have examined, cognitive ability is least influenced by the non-

shared environmental component, 26.0% for Wave I PVT scores and 27.7% for the Wave III

PVT scores.

6.2 Bivariate Pathway Analysis

In this subsection, we examine the extent to which genetic and environmental influences mediate

the correlations between psychological characteristics and financial distress. Because the

univariate results (Table 4) do not indicate any shared environmental effects on the personality

variables, we use a reduced AE model. Table 5 reports the results. The pathway coefficients

(Columns 1-2) can be interpreted analogously to standardized regression parameters. In the

current context, the independent variable is the genetic and environmental influences on

psychological characteristics and the dependent variable is variance in latent financial distress.

The genetic and non-shared environmental correlations (Columns 3-4) report the strength of

association between the genes and environments that influence psychological dispositions and

21

financial distress. Finally, bivariate heritability and environmentality (Columns 5-6) reports the

proportion of the total covariance that is due to genetic or non-shared environmental effects. The

pathway coefficients reveal that three personality traits, namely, conscientiousness, neuroticism,

and agreeableness, are correlated with financial distress through common genetic and/or non-

shared environmental pathways.

Conscientiousness shares variance with financial distress through the non-shared environment.

The non-shared environmental pathway coefficient of -0.13 (p < .01) indicates that a one-unit

change in the environmental component of conscientiousness is correlated with 0.13 unit

decrease in latent financial distress. However, the non-shared environmental effects on

conscientiousness and financial distress are only modestly correlated (rE = -.18). The phenotypic

association is primarily due to the non-shared environment (68%) with the remaining 32% due to

genetic effects.

Neuroticism shares variance with financial distress through both genetic and non-shared

environmental pathways. The coefficients are 0.34 (p < .01) and 0.15 (p < .01), respectively. The

genetic component of latent financial distress substantially correlates with that of neuroticism

(rA = .46), and 52.2% of the phenotypic correlation between neuroticism and the latent financial

distress can be explained by common genetic effects. The non-shared environments explain the

remaining 47.8% of the phenotypic association with a slightly lower environmental correlation

(rE = .20). These results imply that the association between neuroticism and financial distress

occurs almost equally through genetic and non-shared environmental pathways.

Agreeableness is associated with financial distress through a genetic pathway. A one unit

increase in the genetic effects on agreeableness is associated with 0.15 unit decrease in latent

financial distress. The genetic and non-shared environmental factors are weakly correlated (rA =

-.21; rE = .09). Because the genetic and non-shared environmental associations are in opposite

directions, the proportion of the phenotypic association due to genetic effects is 1.92, meaning

the genetic component of the shared variance is almost twice the magnitude of the non-shared

environmental component.

22

Cognitive ability is correlated with financial distress across both waves (p < .001). Increases in

cognitive ability tend to protect against financial distress. However, when the covariance is

broken down into genetic, shared environmental, and non-shared environmental pathways, the

three coefficients are not precisely estimated. This is most likely due to there being little if any

shared environmental variance in financial distress, leading to difficulty estimating the shared

environmental association. When this pathway was constrained to zero, model fit was not

impacted for PVT1 (Δχ2[1] = 1.70, p = .191; ΔCFI = .001) or PVT3 (Δχ2[1] = .76, p = .385;

ΔCFI = .001). We therefore focus on this reduced model. For both PVT1 and PVT3, genetic

influences on ability were protective against financial distress with moderate effect sizes. On the

other hand, non-shared environmental effects on ability were positively associated with

experiencing financial distress, although this effect size was much more modest. The genetic

correlation was approximately .5, and the non-shared environmental correlation was

approximately -.3 across waves. The results were remarkably consistent across measurement

occasions.

6.3 Multivariate Pathway Analysis

To test whether the bivariate associations represent unique effects rather than effects shared with

other psychological dimensions, we fit a multivariate model in which each characteristic that

predicted a significant amount of variance in financial distress was included. We enter the

variable with the smallest genetic association first and enter subsequent characteristics in

ascending order (due to the fact that the order the variables are entered affects the results,

Loehlin, 1996). Thus, we include conscientiousness, agreeableness, neuroticism, and PVT3 as

predictors of latent financial distress in a Cholesky decomposition. We omit PVT1 as we expect

the effects to be shared across PVT1 and PVT3. The results are largely unchanged. The genetic

association between agreeableness and financial distress is reduced to nonsignificance (p = .08)

because conscientiousness shares a significant portion of genetic variance with agreeableness (ba

= .15, p = .03). This result indicates that we are unable to determine if agreeableness has a

unique genetic association with financial distress apart from variance shared with

conscientiousness. The remaining genetic and non-shared environmental associations remained

statistically significant and showed little change in the magnitude of the effect size. This implies

the identified associations are largely independent. In total, the four predictors account for 21.32%

23

of the variance in financial distress through genetic pathways and 10.01% of the variance

through non-shared environmental pathways with significant residual genetic (33.52%) and non-

shared environmental (35.16%) variance.

7. Gene × Socioeconomic Status Interaction

In section 5, we find that variation in financial distress is primarily associated with genetic and

non-shared environmental variance. Shared environmental factors, which conceptually include

family socioeconomic status, parent education, and race/ethnicity, do not explain significant

portions of variance in financial distress through direct pathways. However, as has been pointed

out frequently in the behavior genetic literature (e.g., Bleidorn, Kandler, & Caspi, 2014; Johnson,

Penke, & Spinath, 2011; Johnson, Turkheimer, Gottesman, & Bouchard, 2009; Tucker-Drob,

Briley, & Harden, 2013), large estimates of heritability should not be taken to reflect an absence

of important gene-environment interplay or the influence of family-level environments. One

interpretation of this extremely common if somewhat perplexing result is that these background

factors are experienced differently by members of a sibling pair, which would result in non-

shared environmental variance. Alternatively, the experience of socioeconomic disadvantage

may interact with genetic dispositions whereby some individuals are resilient to disadvantage but

others are substantially hindered in economic maturation. Such an effect would result in variance

in financial distress becoming associated with genetic effects in behavior genetic analyses that do

not include gene-environment interplay. We test this Gene × SES effect by hypothesizing that

genetic influences of financial distress can vary across the range of family SES.

7.1 Gene × SES Interaction for Financial Distress

We first use the nonparametric LOSEM to examine whether the heritability or environmentality

of latent financial distress vary across SES. Figure 2A presents our LOSEM results. As can be

seen, heritability tends to be highest at either extreme of SES. At approximately 1.5 standard

deviations below the mean, heritability is approximately 71%, and at 1.5 standard deviations

above the mean, heritability is estimated to be 68%. However, heritability is substantially lower

at average levels of SES, accounting for only 31% of the variance at the midpoint. Because we

24

found no evidence of shared environmental effects, the trend for the non-shared environment

mirrors the trend for heritability, necessarily (as the only other estimated variance component).

These results imply that the non-shared environment exerts the largest influence on financial

distress for individuals from relatively average family background. For individuals from either

disadvantaged or advantaged family backgrounds, genetic influences predominate. This result is

in line with the hypothesis that genetically influenced characteristics matter to different degrees

in different environments.

Next we replicate this finding using a parametric approach as described in Equations (1) and (2).

Because the shared environmental effects are minimal, we only model the genetic and non-

shared environmental effects in Equation (2). Hence we estimate 𝑏1 and 𝑏2 from Equation (1)

and 𝑎, 𝑎′, 𝑎′′, 𝑒, 𝑒′, 𝑒′′from Equation (2). The unstandardized results are presented in Columns

1-3 in Table 6. Several results are noteworthy. The effect of SES on liability to financial distress

is nonlinear, with the protective effect increasing more rapidly at high levels of SES, as

evidenced by a significant and negative value of 𝑏2. Total variance in financial distress increases

across SES by approximately 30%, indicating greater individual differences at higher levels of

SES. However, this trend is relatively equivalent across genetic and environmental effects, and

therefore we primarily interpret standardized variance components. The statistically significant

values of a′′ and e′′ suggest that the magnitude of the A and E effects systematically vary across

SES. Based on the estimates of the parameters, Figure 2B presents trends in proportions of

variance due to genetic and non-shared environmental factors across SES. The results largely

match those found with LOSEM. Genetic influences on financial distress tend to be largest at

either extreme of SES. The parametric model is slightly more conservative than the

nonparametric model at the midpoint. The parametric model implies roughly equivalent genetic

and environmental effects at the midpoint, whereas the nonparametric model implies a

predominance of non-shared environmental effects. Apart from this minor difference, the two

approaches converge on the same result of larger genetic effects at extremes of SES.

Our result differs from the finding that saving is more genetically influenced for those growing

up in wealthier families in Sweden (Cronqvist & Siegel, 2015). Part of the difference may come

from the fact that the SES distribution of Sweden is more homogeneous than that for the U.S. or

that these regions differ in terms of social systems in place to aid individuals experiencing

25

economic hardship. Our study is based on the sibling sample constructed from the Add Health

data, which is nationally representative. Our sibling sample is .03 SD lower than the full sample

on SES, using the full sample SD as the metric. Thus, our sample includes many individuals that

grew up economically disadvantaged. Additionally, there may be important differences between

the behaviors associated with savings and those associated with financial distress as measured in

the current study.

7.2 Plausible Mechanisms

In this section, we attempt to investigate possible mechanisms that might explain the higher

heritability of financial distress at extremes of SES. Based on the previous findings, we

hypothesize that financial distress is differentially related to heritable risk factors at different

ranges of SES, which explains the higher heritability at the extremes. To test this hypothesis, we

first test whether the association between genetically influenced psychological characteristics

and financial distress is moderated by SES. The current behavior genetic sample is

underpowered to detect such an effect, but the full Add Health sample is sufficiently powered to

detect these interactions. As the majority of the association between financial distress and the

psychological characteristics is mediated by genetic factors, we would expect these

characteristics to have larger effects at the extremes of socioeconomic status if the characteristics

are plausible mechanisms of the interaction. Of course, this reasoning assumes that the

composition of the association in terms of genetic and environmental effects remains fairly static

across SES. It may be the case that any phenotypic moderation results from magnifying a non-

shared environmental pathway, leaving the genetic association largely unchanged and thus not

acting as a mechanism of the identified Gene × SES interaction.

Using the full Add Health sample with available data (N = 15,317), we estimate a model

regressing latent financial distress individually on each psychological characteristic,

socioeconomic status, and the interactive term. In separate models, an interactive effect was

found for agreeableness, neuroticism, and cognitive ability (both PVT1 and PVT3; we focus on

PVT3, but all conclusions are identical across measurement occasions). To determine whether

these interactive effects are independent, we include each characteristic and interaction term in a

26

joint model. Each interactive effect remains, except for agreeableness (p = .069). For this reason,

we focus on neuroticism and cognitive ability. Socioeconomic status and cognitive ability were

each protective against financial distress (β = -.116 and -.104, respectively, both p < .001), and

neuroticism was a risk factor (β = .241, p < .001). However, these effects are qualified by an

interaction between ability and socioeconomic status (β = -.079, p < .001) and between

neuroticism and socioeconomic status (β = -.051, p < .001). In contrast, the main and interactive

effects of agreeableness were much more modest (β = -.008 and -.025, respectively). The model

implied trends for cognitive ability and neuroticism are displayed in Figure 3, holding the other

characteristics constant at their mean. As can be seen, the effect of neuroticism is greater at low

levels of SES compared to high levels (i.e., the expected difference between those 1.5 SD above

compared to below is greater at low SES). The opposite is true for PVT3; at low levels of SES,

cognitive ability is essentially not associated with financial distress positively or negatively, but

at high levels of SES, a protective effect emerges. Together, these results are consistent with the

explanation that larger estimates of genetic influences on financial distress at extremes of SES

emerging from separate psychological characteristics, with personality characteristics playing a

larger role at low levels of SES and ability playing a larger role at high levels of SES.

Next, we test whether the two risk factors are plausible mechanisms for the observed Gene ×

SES interaction. We include the main and interactive effects of neuroticism and cognitive ability

into the model. Thus, we estimate Gene × SES interaction on variance in financial distress

residualized for the linear effects of SES, neuroticism, and cognitive ability, as well as the

interactive effect of each of these variables with SES. The standardized parameter estimates are

reported in Columns 4-6 of Table 6. As can be seen, the inclusion of psychological

characteristics reduces the magnitude of both the a′′and the e′′parameters. In this model, none of

the interactive terms for the genetic or environmental variance components are statistically

different from zero. The model parameters imply a similar pattern of increased heritability at

extremes of SES. However, the difference between heritability at 1.5 SD above or 1.5 SD below

the mean of SES (57% and 56%, respectively) only differed by approximately 4 percentage

points compared to the heritability at the average SES (53%). In contrast, heritability shifts by

approximately 38 percentage points when neuroticism and cognitive ability are not included in

the model. The finding that the moderation of SES on the genetic influences of financial distress

27

is largely attenuated when the effects of these two risk factors are accounted is suggestive

evidence that the two risk factors explain the variable heritability across the SES.

8 Conclusion and Discussion

Our analysis of genetically informative data from the Add Health reveals genetic and

environmental influences on household financial behaviors and the roles of personality, cognitive

ability, and socioeconomics status. About half of the variability of the latent financial distress is

genetically influenced. This heritability estimate is slightly larger than the heritability of the

individual indicators of financial management failures (25%-49%) and other financial behaviors

studied in the finance literature (22%-45%). This is because the latent variable of financial

distress effectively removes the idiosyncratic factors that contribute to financial distress, which

would typically be attributed to the non-shared environment. Compared to other latent

psychological characteristics, the latent financial distress has a relatively lower heritability. For

instance, the heritability of personality is about 70% (Kandler, Riemann, Spinath, & Angleitner,

2010; Kandler, Bleidorn, Riemann, Spinath, Thiel, & Angleitner, 2010; Tucker-Drob, Briley,

Engelhardt, Mann, & Harden, 2016), cognitive ability about 90% (Tucker-Drob, Reynolds,

Finkel, & Pedersen, 2014), and executive functions almost 100% (Engelhardt, Briley, Mann,

Harden, & Tucker-Drob, 2015; Engelhardt, Mann, Briley, Church, Harden, & Tucker-Drob,

2016; Friedman, Miyake, Young, DeFries, Corley, & Hewitt, 2008).

The higher heritability of financial distress relative to that of psychological characteristics

reflects the closer connection between finances and the environment compared to psychological

dispositions. Importantly, these environmental influences do not operate to make siblings

growing up in the same home more similar in their financial behavior, but rather differentiate

family members. This result is common in behavior genetics (Plomin & Daniels, 1987). Two

explanations are likely relevant in the current context. First, two siblings could objectively

experience the same level of parental education or income, but the effect of those experiences

might differentiate siblings. A substantial portion of the non-shared environmental influences on

financial distress may reflect this type of infleucne. Second, unmodeled Gene × Environment

Interaction could mask shared environmental influences (Purcell, 2002). We demonstrated this

empirically by using family SES, an objective shared environment. In addition to possibly

having differential within-family effects, family SES moderated the genetic and environmental

28

influences on financial distress in the current sample. For these reasons, it would be incorrect to

assume based on the small estimate of the shared environment that family-level variables are

unimportant. It is also worth noting that that a relatively high heritability of a characteristic does

not imply that the characteristic is unchangeable. For instance, while cognitive abilities are as

much as 80% heritable at age 18 (Boomsma, Busjahn, & Peltonen, 2002), there is evidence that

cognitive skills continue to evolve over lifetime (Tucker-Drob, 2009). Behavior genetic models

estimate variance components at the population-level based on the actual experiences of the

sample under study. Policy interventions can impact financial behavior, even if moderately

heritable, similar to how eyeglasses can correct vision even though heritability of uncorrected

vision is high (Manski, 2011).

We find that three personality traits and cognitive ability explain nearly half of genetic variance

in financial distress. The pathway analyses provide nuanced insight into the associations between

financial outcomes and cognitive and non-cognitive abilities. They suggest that cognitive and

non-cognitive abilities mediate the genetic influences on financial outcomes. Future genetically

informative, longitudinal research will be required to establish when in the lifespan these

associations manifest, which would provide critical information for crafting interventions or

policy recommendations. Our results also suggest that personality traits and cognitive ability can

be used as markers for financially at-risk groups because those high on neuroticism and low on

ability experience greater financial hardship. One cannot conclude, however, that modifying

personality or cognitive ability will directly change financial status because alternative causal

pathways are possible even when shared genetic factors regulate both financial distress and

personality/cognitive ability. For instance, the shared genetic factors can regulate financial

distress through personality, or the shared genetic factors can influence personality and financial

distress independently. In the first case, experimentally manipulating personality would be

expected to alter financial behavior, but in the second case, manipulating personality would not

alter financial behavior because the underlying causal agent (i.e., genetic variation acting as a

common cause) would not be manipulated. A better understanding of these pathways would be a

useful direction for future research.

We show that the heritability of financial distress is the highest at both the low and high ends of

family of origin SES. One interpretation of this finding is that genetically influenced individual

29

differences are more important at either end of SES (South & Krueger, 2013). Our results using

the full Add Health dataset imply that neuroticism plays a relatively more important role among

individuals from low SES backgrounds and cognitive ability plays a relatively more important

role among individuals from high SES backgrounds. In combination, this pair of psychological

characteristics could explain our finding of higher heritability of financial distress at extremes of

SES. We find support for this hypothesis by showing that variation in the heritability of financial

distress across SES largely disappears after controlling for the effects of neuroticism and

cognitive ability. The current results imply that individuals with different socioeconomic

backgrounds may experience financial distress through qualitatively different pathways. This

knowledge can help address the intergenerational persistence of financial inequality and the

stratification by wealth. Our results provide a foundation for future work on targeted

interventions that aim at improving financial well-being. Prior to implementing such work, it will

first be necessary to replicate the findings as the current results were exploratory in nature.

We foresee future applications of behavior genetic methods to household finance along two lines.

The first is further investigation into the intergenerational transmission of financial distress and

the dynamics of genetic and environmental factors of financial distress over the lifecycle. The

second is to study the gene-environment correlation that explains the gene-mediated sorting into

adverse environments that lead to financial distress. We advocate for diverse methods with

unique strengths and weakness to extend this line of research. For example, the current research

uses a twin and sibling method to infer the genetic influences on a trait. This method has

substantial statistical power by aggregating genetic effects across the entire genome, which also

allows for estimating aggregate Gene × Environment Interaction (rather than testing individual

candidate genes which are unlikely to replicate; Duncan & Keller, 2011). With the growing

availability of molecular genetic data, other methods can be employed for future studies. For

instance, a Genome-wide Complex Trait Analysis (GCTA) relaxes the assumptions for twin and

family studies and instead compares genetic similarity of unrelated strangers to their measured

similarity on a trait to measure heritability. However, this technique requires vastly larger sample

sizes (tens of thousands) to achieve a similar level of statistical power as twin and sibling studies

because the difference in genetic relatedness across individuals is miniscule compared to the

difference between identical and fraternal twins. Polygenic risk scores offer an additional

molecular tool (Dudbridge, 2013). These risk scores rely on genome-wide association studies

30

(GWAS) which link specific genetic variants with an outcome of interest. The risk score is

created by aggregating each individual’s genotype with the effect size for each variant. Both

GCTA and polygenic risk scores are largely similar to twin and family studies in that they assess

genetic variance in some manner but are incapable of identifying the specific variants. Only

GWAS offers this additional piece of information. GWAS is limited because for most or all

complex psychological phenotypes the effect size of each variant is extremely small (Chabris et

al., 2015). Due to these combined methodological strengths and weaknesses, we foresee progress

in genetic approaches to economic behavior continuing to flow from each study design.

31

References

Agarwal, S., & Mazumder, B. (2013). Cognitive abilities and household financial decision making. American

Economic Journal: Applied Economics, 5(1), 193–207.

Anderloni, L., Bacchiocchi, E., & Vandone, D. (2012). Household financial vulnerability: An empirical analysis.

Research in Economics, 66(3), 284–296. http://doi.org/10.1016/j.rie.2012.03.001

Anokhin, A. P., Golosheykin, S., Grant, J. D., & Heath, A. C. (2011). Heritability of delay discounting in

adolescence: a longitudinal twin study. Behavior Genetics, 41(2), 175–183.

Atkinson, A., McKay, S., Kempson, E., & Collard, S. (2006). Levels of Financial Capability in the UK: results of a

baseline survey. UK: financial Services Authority.

Barnea, A., Cronqvist, H., & Siegel, S. (2010). Nature or nurture: What determines investor behavior? Journal of

Financial Economics, 98(3), 583–604. http://doi.org/10.1016/j.jfineco.2010.08.001

Bauer, D. J. (2016). A More General Model for Testing Measurement Invariance and Differential Item Functioning.

Psychological Methods.

Benyamin, B., Pourcain, B., Davis, O. S., Davies, G., Hansell, N. K., Brion, M.-J., et al. (2013). Childhood

intelligence is heritable, highly polygenic and associated with FNBP1L, 19(2), 256–261.

http://doi.org/10.1038/mp.2012.184

Black, S. E., Devereux, P. J., Lundborg, P., & Majlesi, K. (2015). Poor Little Rich Kids? The Determinants of the

Intergenerational Transmission of Wealth. NBER Working Paper.

Bleidorn, W., Kandler, C., & Caspi, A. (2014). The behavioural genetics of personality development in adulthood—

Classic, contemporary, and future trends. European Journal of Personality, 28(3), 244-255.

Boomsma, D., Busjahn, A., & Peltonen, L. (2002). Classical twin studies and beyond. Nature Reviews Genetics,

3(11), 872–882. http://doi.org/10.1038/nrg932

Borghans, L., Duckworth, A. L., Heckman, J. J., & Weel, ter, B. (2008). The Economics and Psychology of

Personality Traits. The Journal of Human Resources, 43(4), 972–1059.

http://doi.org/10.2307/40057376?ref=search-gateway:b4269d8698e33736e66007849514253e

Bouchard, T. J., & McGue, M. (2003). Genetic and environmental influences on human psychological differences.

Journal of Neurobiology, 54(1), 4–41.

Bouchard, T. J., Jr. (2014). Genes, evolution and intelligence. Behavior Genetics, 44(6), 549–577.

http://doi.org/10.1007/s10519-014-9646-x

Bouchard, T. J., Jr, & Loehlin, J. C. (2001). Genes, Evolution, and Personality. Behavior Genetics, 31(3), 243–273.

http://doi.org/10.1023/A:1012294324713

Briley, D. A., & Tucker-Drob, E. M. (2013a). Explaining the Increasing Heritability of Cognitive Ability Across

Development: A Meta-Analysis of Longitudinal Twin and Adoption Studies. Psychological Science, 24(9),

1704–1713. http://doi.org/10.1177/0956797613478618

Briley, D. A., & Tucker-Drob, E. M. (2013b). Explaining the Increasing Heritability of Cognitive Ability Across

Development: A Meta-Analysis of Longitudinal Twin and Adoption Studies. Psychological Science, 24(9),

1704–1713. http://doi.org/10.1177/0956797613478618

32

Briley, D. A., & Tucker-Drob, E. M. (2014). Genetic and environmental continuity in personality development: A

meta-analysis. Psychological Bulletin, 140(5), 1303–1331. http://doi.org/10.1037/a0037091

Briley, D. A., Harden, K. P., Bates, T. C., & Tucker-Drob, E. M. (2015). Nonparametric Estimates of

Gene × Environment Interaction Using Local Structural Equation Modeling. Behavior Genetics, 45(5), 581–

596. http://doi.org/10.1007/s10519-015-9732-8

Brown, S., Ghosh, P., & Taylor, K. (2012). The existence and persistence of household financial hardship. Working

Paper, University of Sheffield.

Calvet, L. E., & Sodini, P. (2014). Twin Picks: Disentangling the Determinants of Risk-Taking in Household

Portfolios. The Journal of Finance, 69(2), 867–906. http://doi.org/10.1111/jofi.12125

Cesarini, D., Dawes, C. T., Fowler, J. H., Johannesson, M., Lichtenstein, P., & Wallace, B. (2008). Heritability of

cooperative behavior in the trust game. Proceedings of the National Academy of Sciences, 105(10), 3721–3726.

http://doi.org/10.1073/pnas.0710069105

Cesarini, D., Dawes, C. T., Johannesson, M., Lichtenstein, P., & Wallace, B. (2009a). Genetic Variation in

Preferences for Giving and Risk Taking. Quarterly Journal of Economics, 124(2), 809–842.

http://doi.org/10.1162/qjec.2009.124.2.809

Cesarini, D., Johannesson, M., Lichtenstein, P., Sandewall, Ö., & Wallace, B. (2010). Genetic Variation in Financial

Decision-Making. The Journal of Finance, 65(5), 1725–1754. http://doi.org/10.1111/j.1540-6261.2010.01592.x

Cesarini, D., Johannesson, M., Magnusson, P. K. E., & Wallace, B. (2012). The Behavioral Genetics of Behavioral

Anomalies. Management Science, 58(1), 21–34. http://doi.org/10.1287/mnsc.1110.1329

Cesarini, D., Lichtenstein, P., Johannesson, M., & Wallace, B. (2009b). Heritability of overconfidence. Journal of

the European Economic Association, 7(2‐3), 617–627.

Chabris, C. F., Lee, J. J., Cesarini, D., Benjamin, D. J., & Laibson, D. I. (2015). The fourth law of behavior

genetics. Current Directions in Psychological Science, 24(4), 304-312.

Conley, D., Rauscher, E., Dawes, C., Magnusson, P. K. E., & Siegal, M. L. (2013). Heritability and the Equal

Environments Assumption: Evidence from Multiple Samples of Misclassified Twins. Behavior Genetics, 43(5),

415–426. http://doi.org/10.1007/s10519-013-9602-1

Cronqvist, H., & Siegel, S. (2014). The genetics of investment biases. Journal of Financial Economics, 113(2), 215–

234.

Cronqvist, H., & Siegel, S. (2015). The Origins of Savings Behavior. The Journal of Political Economy, 123(1),

123–169.

Cronqvist, H., Münkel, F., & Siegel, S. (2012). Genetics, Homeownership, and Home Location Choice. The Journal

of Real Estate Finance and Economics, 48(1), 79–111. http://doi.org/10.1007/s11146-012-9373-0

Damian, R. I., Su, R., Shanahan, M., Trautwein, U., & Roberts, B. W. (2015). Can personality traits and intelligence

compensate for background disadvantage? Predicting status attainment in adulthood. Journal of Personality

and Social Psychology, 109(3), 473–489.

Dick, D. M., Viken, R., Purcell, S., Kaprio, J., Pulkkinen, L., & Rose, R. J. (2007). Parental Monitoring Moderates

the Importance of Genetic and Environmental Influences on Adolescent Smoking. Journal of Abnormal

Psychology, 116(1), 213–218. http://doi.org/10.1037/0021-843X.116.1.213

33

Domingue, B. W., Fletcher, J., Conley, D., & Boardman, J. D. (2014). Genetic and educational assortative mating

among US adults. Proceedings of the National Academy of Sciences, 111(22), 7996–8000.

http://doi.org/10.1073/pnas.1321426111

Donnellan, M. B., Conger, K. J., McAdams, K. K., & Neppl, T. K. (2009). Personal Characteristics and Resilience

to Economic Hardship and Its Consequences: Conceptual Issues and Empirical Illustrations. Journal of

Personality, 77(6), 1645–1676. http://doi.org/10.1111/j.1467-6494.2009.00596.x

Donnellan, M. B., Oswald, F. L., Baird, B. M., & Lucas, R. E. (2006). The Mini-IPIP Scales: Tiny-yet-effective

measures of the Big Five Factors of Personality. Psychological Assessment, 18(2), 192–203.

http://doi.org/10.1037/1040-3590.18.2.192

Duckworth, A. L., & Weir, D. R. (2010). Personality, Lifetime Earnings, and Retirement Wealth. University of

Michigan Retirement Research Center WP 2010-235.

Duckworth, A. L., Weir, D., Tsukayama, E., & Kwok, D. (2012). Who does well in life? Conscientious adults excel

in both objective and subjective success. Frontiers in Psychology, 3.

Duckworth, A., & Weir, D. (2011). Personality and Response to the Financial Crisis. SSRN Electronic Journal.

http://doi.org/10.2139/ssrn.2006595

Dudbridge, F. (2013). Power and predictive accuracy of polygenic risk scores. PLoS Genet, 9(3), e1003348.

Duncan, L. E., & Keller, M. C. (2011). A critical review of the first 10 years of candidate gene-by-environment

interaction research in psychiatry. American Journal of Psychiatry, 168(10), 1041-1049.

Eaves, L. J., Heath, A. C., & Martin, N. G. (1984). A note on the generalized effects of assortative mating. Behavior

Genetics, 14(4), 371–376.

Engelhardt, L. E., Briley, D. A., Mann, F. D., Harden, K. P., & Tucker-Drob, E. M. (2015). Genes unite executive

functions in childhood. Psychological science, 26(8), 1151-1163.

Engelhardt, L. E., Mann, F. D., Briley, D. A., Church, J. A., Harden, K. P., & Tucker-Drob, E. M. (2016). Strong

genetic overlap between executive functions and intelligence.

Fagereng, A., Mogstad, M., & Rønning, M. (2015). Why do wealthy parents have wealthy children? Statistics

Norway Research Department Discussion Papers No. 813.

Friedman, N. P., Miyake, A., Young, S. E., DeFries, J. C., Corley, R. P., & Hewitt, J. K. (2008). Individual

differences in executive functions are almost entirely genetic in origin. Journal of Experimental Psychology:

General, 137(2), 201.

Gerardi, K., Goette, L., & Meier, S. (2013). Numerical ability predicts mortgage default. Proceedings of the

National Academy of Sciences, 110(28), 11267–11271. http://doi.org/10.1073/pnas.1220568110

Grinblatt, M., Keloharju, M., & Linnainmaa, J. (2011). IQ and Stock Market Participation. The Journal of Finance,

66(6), 2121–2164. http://doi.org/10.2307/41305186?ref=search-gateway:e8015dd14f9d476351ea09e471fa8fdf

Harris, K. M., Halpern, C. T., Smolen, A., & Haberstick, B. C. (2006). The national longitudinal study of adolescent

health (Add Health) twin data. Twin Research and Human Genetics, 9(6), 988–997.

Haworth, C. M. A., Kovas, Y., Harlaar, N., Hayiou-Thomas, M. E., Petrill, S. A., Dale, P. S., & Plomin, R. (2009a).

Generalist genes and learning disabilities: a multivariate genetic analysis of low performance in reading,

34

mathematics, language and general cognitive ability in a sample of 8000 12-year-old twins. Journal of Child

Psychology and Psychiatry, 50(10), 1318–1325. http://doi.org/10.1111/j.1469-7610.2009.02114.x

Haworth, C. M. A., Wright, M. J., Luciano, M., Martin, N. G., de Geus, E. J. C., van Beijsterveldt, C. E. M., et al.

(2009b). The heritability of general cognitive ability increaseslinearly from childhood to young adulthood.

Molecular Psychiatry, 15(11), 1112–1120. http://doi.org/10.1038/mp.2009.55

Johnson, W. (2007). Genetic and environmental influences on behavior: Capturing all the interplay. Psychological

Review, 114(2), 423–440. http://doi.org/10.1037/0033-295X.114.2.423

Johnson, W., Penke, L., & Spinath, F. M. (2011). Heritability in the era of molecular genetics: Some thoughts for

understanding genetic influences on behavioural traits. European Journal of Personality, 25(4), 254-266.

Johnson, W., Turkheimer, E., Gottesman, I. I., & Bouchard Jr, T. J. (2009). Beyond heritability: Twin studies in

behavioral research. Current directions in psychological science, 18(4), 217-220.

Kandler, C. (2012). Nature and Nurture in Personality Development: The Case of Neuroticism and Extraversion.

Current Directions in Psychological Science, 21(5), 290–296. http://doi.org/10.1177/0963721412452557

Kandler, C., Bleidorn, W., Riemann, R., Spinath, F. M., Thiel, W., & Angleitner, A. (2010). Sources of cumulative

continuity in personality: a longitudinal multiple-rater twin study. Journal of personality and social

psychology, 98(6), 995.

Kandler, C., Riemann, R., Spinath, F. M., & Angleitner, A. (2010). Sources of Variance in Personality Facets: A

Multiple‐Rater Twin Study of Self‐Peer, Peer‐Peer, and Self‐Self (Dis) Agreement. Journal of

personality, 78(5), 1565-1594.

Kautz, T., Heckman, J., Diris, R., Weel, ter, B., & Borghans, L. (2014). Fostering and Measuring Skills: Improving

Cognitive and Non-Cognitive Skills to Promote Lifetime Success. nber.org. NBER Working Paper Series.

Keyes, C. L., Kendler, K. S., Myers, J. M., & Martin, C. C. (2015). The Genetic Overlap and Distinctiveness of

Flourishing and the Big Five Personality Traits. Journal of Happiness Studies, 16(3), 655–668.

Keyes, C. L., Myers, J. M., & Kendler, K. S. (2010). The structure of the genetic and environmental influences on

mental well-being. American Journal of Public Health, 100(12), 2379–2384.

Kiecolt, K. J., Aggen, S. H., & Kendler, K. S. (2013). Genetic and Environmental Influences on the Relationship

Between Mastery and Alcohol Dependence. Alcoholism: Clinical and Experimental Research, 37(6), 905–913.

Krapohl, E., Rimfeld, K., Shakeshaft, N. G., Trzaskowski, M., McMillan, A., Pingault, J.-B., et al. (2014). The high

heritability of educational achievement reflects many genetically influenced traits, not just intelligence.

Proceedings of the National Academy of Sciences, 111(42), 15273–15278.

http://doi.org/10.1073/pnas.1408777111

Letkiewicz, J. C., & Fox, J. J. (2014). Conscientiousness, Financial Literacy, and Asset Accumulation of Young

Adults. Journal of Consumer Affairs, 48(2), 274–300.

Loehlin, J. C. (1996). The Cholesky approach: A cautionary note. Behavior Genetics, 26(1), 65–69.

Lundberg, S. (2013). The College Type: Personality and Educational Inequality. Journal of Labor Economics, 31(3),

421–441. http://doi.org/10.1086/671056

Manski, C. F. (2011). Genes, Eyeglasses, and Social Policy. The Journal of Economic Perspectives, 25(4), 83–94.

http://doi.org/10.1257/jep.25.4.83

35

McGue, M., & Christensen, K. (2003). The heritability of depression symptoms in elderly Danish twins: occasion-

specific versus general effects. Behavior Genetics.

McGue, M., Bacon, S., & Lykken, D. T. (1993). Personality stability and change in early adulthood: A behavioral

genetic analysis. Developmental Psychology, 29(1), 96–109.

McGue, M., Elkins, I., Walden, B., & Iacono, W. G. (2005). The essential role of behavioral genetics in

developmental psychology: Reply to Partridge (2005) and Greenberg (2005). Developmental Psychology, 41(6),

993–997. http://doi.org/10.1037/0012-1649.41.6.993

Melzer, B. T. (2011). The Real Costs of Credit Access: Evidence from the Payday Lending Market. The Quarterly

Journal of Economics, 126(1), 517–555.

Mirowsky, J., & Ross, C. E. (2003). Education, social status, and health. Transaction Publishers.

Mottola, G. R. (2014). The financial capability of young adults—A generational view. Insights: Financial

Capability—March.

Neale, M., & Cardon, L. (1992). Methodology for genetic studies of twins and families. Dordrecht, The Netherlands:

Kluwer Academic Publishers.

Nicolaou, N., & Shane, S. (2010). Entrepreneurship and occupational choice: Genetic and environmental influences.

Journal of Economic Behavior & Organization, 76(1), 3–14.

Nyhus, E. K., & Webley, P. (2001). The role of personality in household saving and borrowing behaviour. European

Journal of Social Psychology, 15(S1), S85–S103. http://doi.org/10.1002/per.422

Okbay, A., Beauchamp, J. P., Fontana, M. A., Lee, J. J., & Pers, T. H. (2016). Genome-wide association study

identifies 74 loci associated with educational attainment. Nature, 533(7604), 539–542.

Plomin, R., & Daniels, D. (1987). Why are children in the same family so different from one another?. Behavioral

and Brain Sciences, 10(01), 1-16.

Purcell, S. (2002). Variance components models for gene–environment interaction in twin analysis. Twin Research,

5(6), 554–571.

Rammstedt, B., Spinath, F. M., Richter, D., & Schupp, J. (2013). Personality and Individual Differences. Personality

and Individual Differences, 54(7), 832–835. http://doi.org/10.1016/j.paid.2012.12.007

Roisman, G. I., Newman, D. A., Fraley, R. C., Haltigan, J. D., Groh, A. M., & Haydon, K. C. (2012). Distinguishing

differential susceptibility from diathesis–stress: Recommendations for evaluating interaction effects.

Development and Psychopathology, 24(02), 389–409. http://doi.org/10.1017/S0954579412000065

Rustichini, A., DeYoung, C. G., Anderson, J. E., & Burks, S. V. (2016). Toward the integration of personality

theory and decision theory in explaining economic behavior: An experimental investigation. Journal of

Behavioral and Experimental Economics. http://doi.org/10.1016/j.socec.2016.04.019

Sacerdote, B. (2002). The Nature and Nurture of Economic Outcomes. American Economic Review, 92(2), 344–348.

Sacerdote, B. (2007). How Large Are the Effects from Changes in Family Environment? A Study of Korean

American Adoptees. The Quarterly Journal of Economics, 122(1), 119–157.

http://doi.org/10.2307/25098839?ref=search-gateway:01382001149534aa91272d6905745d0b

Scarr, S., & McCartney, K. (1983). How people make their own environments: A theory of genotype→ environment

effects. Child Development, 424–435.

36

Shanahan, M. J., Bauldry, S., Roberts, B. W., Macmillan, R., & Russo, R. (2014). Personality and the reproduction

of social class. Social Forces, 93(1), 209–240.

Slutske, W. S., Cho, S. B., & Piasecki, T. M. (2013). Genetic overlap between personality and risk for disordered

gambling: Evidence from a national community-based Australian twin study. Journal of Abnormal Psychology,

122(1), 250.

South, S. C., & Krueger, R. F. (2013). Marital Satisfaction and Physical Health: Evidence for an Orchid Effect.

Psychological Science, 24(3), 373–378. http://doi.org/10.1177/0956797612453116

South, S. C., Reichborn-Kjennerud, T., Eaton, N. R., & Krueger, R. F. (2012). Genetics of personality. Handbook of

Psychology, Volume Five: Personality and Social Psychology. Hoboken, NJ: Wiley.

Taylor, M. (2011). Measuring financial capability and its determinants using survey data. Social Indicators

Research, 102(2), 297–314.

Taylor, M. P., Jenkins, S. P., & Sacker, A. (2011). Financial capability and psychological health. Journal of

Economic Psychology, 32(5), 710–723. http://doi.org/10.1016/j.joep.2011.05.006

Thompson, L. A., Detterman, D. K., & Plomin, R. (1991). Associations between cognitive abilities and scholastic

achievement: Genetic overlap but environmental differences. Psychological Science, 2(3), 158–165.

Tucker-Drob, E. M. (2009). Differentiation of cognitive abilities across the life span. Developmental Psychology,

45(4), 1097–1118.

Tucker-Drob, E. M., & Bates, T. C. (2016). Large Cross-National Differences in Gene x Socioeconomic Status

Interaction on Intelligence. Psychological Science, 27(2), 138–149. http://doi.org/10.1177/0956797615612727

Tucker-Drob, E. M., & Briley, D. A. (2014). Continuity of genetic and environmental influences on cognition across

the life span: A meta-analysis of longitudinal twin and adoption studies. Psychological Bulletin, 140(4), 949–

979. http://doi.org/10.1037/a0035893

Tucker-Drob, E. M., Briley, D. A., Engelhardt, L. E., Mann, F. D., & Harden, K. P. (2016). Genetically-mediated

associations between measures of childhood character and academic achievement. Journal of personality and

social psychology, 111(5), 790.

Tucker-Drob, E. M., Briley, D. A., & Harden, K. P. (2013). Genetic and environmental influences on cognition

across development and context. Current directions in psychological science, 22(5), 349-355.

Tucker-Drob, E. M., Reynolds, C. A., Finkel, D., & Pedersen, N. L. (2014). Shared and unique genetic and

environmental influences on aging-related changes in multiple cognitive abilities. Developmental

psychology, 50(1), 152.

Turkheimer, E. (2000). Three Laws of Behavior Genetics and What They Mean. Current Directions in

Psychological Science, 9(5), 160–164. http://doi.org/10.1111/1467-8721.00084

Turkheimer, E., & Waldron, M. (2000). Nonshared environment: a theoretical, methodological, and quantitative

review. Psychological Bulletin, 126(1), 78. http://doi.org/10.1037//0033-2909.1261.78

Visscher, P. M., Medland, S. E., Ferreira, M. A., Morley, K. I., Zhu, G., Cornes, B. K., et al. (2006). Assumption-

free estimation of heritability from genome-wide identity-by-descent sharing between full siblings. PLOS

Genet, 2(3), e41.

37

Wallace, B., Cesarini, D., Lichtenstein, P., & Johannesson, M. (2007). Heritability of ultimatum game responder

behavior. Proceedings of the National Academy of Sciences, 104(40), 15631–15634.

http://doi.org/10.1073/pnas.0706642104

Xu, Y., Beller, A. H., Roberts, B. W., & Brown, J. R. (2015). Personality and young adult financial distress. Journal

of Economic Psychology, (51), 90–100.

Yang, J., Lee, S. H., Goddard, M. E., & Visscher, P. M. (2011). GCTA: a tool for genome-wide complex trait

analysis. The American Journal of Human Genetics, 88(1), 76-82.

Zhong, S., Chew, S. H., Set, E., Zhang, J., Xue, H., Sham, P. C., et al. (2012). The Heritability of Attitude Toward

Economic Risk. Twin Research and Human Genetics, 12(01), 103–107. http://doi.org/10.1375/twin.12.1.103

Zyphur, M. J., Narayanan, J., Arvey, R. D., & Alexander, G. J. (2009). The genetics of economic risk preferences.

Journal of Behavioral Decision Making, 22(4), 367–377. http://doi.org/10.1002/bdm.643

38

Figure 1a: Financial Distress Common Factor. Behavior genetic parameters are reported in terms of percentages of variance, and

significant parameters are reported in bold. All factor loadings are significant at p < .001. A = genetic effects. C = shared

environmental effects. E = non-shared environmental effects.

Figure 1b: Financial Distress Common Factor (Trimmed Model). Behavior genetic parameters are reported in terms of

percentages of variance, and significant parameters are reported in bold. All factor loadings are significant at p < .001. A =

genetic effects. E = non-shared environmental effects. Share environmental effects, C, is dropped from the model based on

previous analysis.

39

Figure 2. Nonparametric and parametric estimates of heritability and environmentality of latent financial distress

across the distribution of Socioeconomic Status (SES).

Figure 3. Phenotypic moderation of neuroticism and PVT3 by socioeconomic status.

40

Appendix A: A Discussion of the Assumptions for the ACE model

The ACE model depends on three assumptions (Neale & Cardon, 1992). First, the model

assumes that genetic and environmental influences operate independently. Gene-environment

interplay (i.e., Gene × Environment interaction and gene-environment correlation) influence

ACE models in a predictable manner (Purcell, 2002). For example, active gene-environment

correlation leads to increases in genetic portions of variance (e.g., as seen in increasing

heritability of cognitive ability), and when genes interact with shared environments (e.g., family

socioeconomic status), the proportion of variance associated with genetic effects increases. Such

effects do not invalidate the ACE model, but rather, call for a nuanced interpretation of the ACE

parameters as they can (and likely do) depend on gene-environment interplay (Johnson, 2007). In

fact, we demonstrated fairly substantial Gene × SES interaction for financial distress.

Second, the equal environment assumption (EEA) requires that sibling types are not

systematically exposed to certain environments on the basis of zygosity. Specifically, violation

of the EEA requires that MZ twins are treated differently than other individuals simply because

of their zygosity in a manner that affects their development. For example, it may be the case that

parents tend to dress MZ twins alike more often than other siblings, but as long as similar dress

as a child does not affect adult psychological or economic outcomes, then the EEA holds.

Numerous studies have tested the EEA and found strong support for it, even in the current

sample (Conley, Rauscher, Dawes, Magnusson, & Siegal, 2013).

Third, the random mating assumption requires that humans do not mate based on genetic

similarity. If genetically similar individuals mate, then their offspring will be more genetically

related than would be expected (e.g., fraternal twins and full siblings would share more than 50%

of segregating genetic material). Assortative mating attenuates estimates of heritability (Eaves,

Heath, & Martin, 1984). However, minimal evidence for strong assortative mating on personality

or cognitive ability has been found (e.g., (Domingue, Fletcher, Conley, & Boardman, 2014;

Rammstedt, Spinath, Richter, & Schupp, 2013). Further, (Visscher et al., 2006) measured the

genetic similarity of full siblings and found an average genetic similarity of .498 with a standard

deviation of only .036, which is largely consistent with the specification of full siblings and

fraternal twins as sharing 50% of segregating genetic material.

41

Appendix B: The Genetic Method

Without explicitly identifying the underlying genetic architecture, the quantitative genetic theory

assumes that an observable characteristic of an individual, 𝑦𝑖𝑗 , such as a personality trait or

financial behavior for individual 𝑖 in family 𝑗, can be written as an additive model of three latent

variables,

𝑦𝑖𝑗 = 𝐴𝑖𝑗 + 𝐶𝑖𝑗 + 𝐸𝑖𝑗

𝐴𝑖𝑗 ~𝑁(0, 𝜎𝐴2) is the additive genetic factor, 𝐶𝑖𝑗~𝑁(0, 𝜎𝐶

2) is the shared environments,

including but not limited to shared family background and shared life experiences, and

𝐸𝑖𝑗~𝑁(0, 𝜎𝐸2) is the non-shared environment, including idiosyncratic shocks and measurement

error.

Consider two unrelated sibling pairs 𝑗 = 1,2 with individual 𝑖 = 1,2 in each pair. Denote the

additive genetic component 𝐴 = (𝐴11, 𝐴21, 𝐴12, 𝐴22) , the shared environments 𝐶 =

(𝐶11, 𝐶21, 𝐶12, 𝐶22), non-shared environment 𝐸 = (𝐸11, 𝐸21, 𝐸12, 𝐸22). If pair 𝑗 = 1 are MZ twins

and pair 𝑗 = 2 are DZ twins, the covariate matrices for the three components are:

𝑐𝑜𝑣(𝐴) = 𝜎𝐴2

[ 1 1 0 01 1 0 0

0 0 11

2

0 01

21]

𝑐𝑜𝑣(𝐶) = 𝜎𝐶2 [

1 1 0 01 1 0 00 0 1 10 0 1 1

] 𝑐𝑜𝑣(𝐸) = 𝜎𝐸2 [

1 0 0 00 1 0 00 0 1 00 0 0 1

]

If pair 𝑗 = 1 are full siblings and pair 𝑗 = 2 are half siblings, the covariate matrices for the three

components are:

𝑐𝑜𝑣(𝐴) = 𝜎𝐴2

[ 1

1

20 0

1

21 0 0

0 0 11

4

0 01

41]

𝑐𝑜𝑣(𝐶) = 𝜎𝐶2 [

1 1 0 01 1 0 00 0 1 10 0 1 1

] 𝑐𝑜𝑣(𝐸) = 𝜎𝐸2 [

1 0 0 00 1 0 00 0 1 00 0 0 1

]

The Univariate Model

42

Denote the vector of characteristics for the two pairs as 𝑦 = (𝑦11, 𝑦21, 𝑦12, 𝑦22)′, the covariate

structure is 𝐶𝑜𝑣(𝑦) = 𝐶𝑜𝑣 (𝐴) + 𝐶𝑜𝑣 (𝐶) + 𝐶𝑜𝑣 (𝐸) . The heritability index ℎ =𝜎𝐴

2

𝜎𝐴2+𝜎𝐶

2+𝜎𝐸2 ,

denoted as 𝐴

𝐴+𝐶+𝐸 , measures the share of co-twin variance in a characteristics that can be

attributable to the genetic component. Similarly, the influences of the shared environment are

measured by 𝜎𝐶

2

𝜎𝐴2+𝜎𝐶

2+𝜎𝐸2 (denoted as

𝐶

𝐴+𝐶+𝐸), and the influences of the non-shared environment are

measured by 𝜎𝐸

2

𝜎𝐴2+𝜎𝐶

2+𝜎𝐸2 (denoted as

𝐸

𝐴+𝐶+𝐸). Figure B1 illustrates the path diagram.

The Bivariate and Multivariate Models

In the bivariate analyses, we examine the genetic and environmental pathways between a

psychological characteristic and a financial behavior. In addition to the usual assumptions of a

univariate model, we further assume that the covariance between the two characteristics can be

decomposed into genetic and environmental components. Denote a psychological characteristic

as xij, and financial distress as yij, the latent variable constructs can be written as the following:

𝑥𝑖𝑗 = 𝐴𝑖𝑗𝑥 + 𝐶𝑖𝑗

𝑥 + 𝐸𝑖𝑗𝑥

𝑦𝑖𝑗 = 𝐴𝑖𝑗𝑦

+ 𝐶𝑖𝑗𝑦

+ 𝐸𝑖𝑗𝑦

Figure B2 illustrates the path diagram. The pathway coefficients of interest are 𝑎𝑦𝑥 , which

measures how much the variation in the genetic component of 𝑥𝑖𝑗 explains the variation in 𝑦𝑖𝑗,

𝑐𝑦𝑥 , which measures the influences of the shared environment of 𝑥𝑖𝑗 on 𝑦𝑖𝑗 , and 𝑒𝑦𝑥 , which

measures the influences of the non-shared environment of 𝑥𝑖𝑗 on 𝑦𝑖𝑗. We can also compute the

genetic correlation (i.e., 𝑎𝑦𝑥

√𝑎𝑦𝑥2 +𝑎𝑦𝑦

2) and bivariate heritability (i.e.,

𝑎𝑥𝑥𝑎𝑦𝑥

𝑎𝑥𝑥𝑎𝑦𝑥+𝑐𝑥𝑥𝑐𝑦𝑥+𝑒𝑥𝑥𝑒𝑦𝑥). The

genetic correlation reflects the extent to which genetic influences on one characteristic are shared

with another characteristic. Bivariate heritability indicates the extent to which the observed

association is due to a genetic pathway. Interpretation of environmental correlations and

bivariate environmentality are similar.

Similarly, we can expand the bivariate analysis to the multivariate case to allow the A, C, and E

components of multiple psychological characteristics to influence financial distress. This model

43

is the behavior genetic analog to regression analysis because covariation among the independent

variables is controlled to identify effects specific to the individual predictors. Figure B3

demonstrates the pathway diagram for a multivariate model where the A, C, and E that affect 𝑥𝑖𝑗

and 𝑧𝑖𝑗 are assumed to affect 𝑦𝑖𝑗. The characteristics 𝑥𝑖𝑗 and 𝑧𝑖𝑗 can be a personality trait and/or

cognitive ability, the characteristic 𝑦𝑖𝑗 is latent financial distress. From the multivariate analysis,

we can infer whether the genetic or environmental pathways from personality and cognitive

abilities to financial distress are independent or common. If the pathway coefficients from the

multivariate analysis are similar to those from the bivariate analysis, it is evidence that the

pathways are independent.

Figure B1: Univariate ACE Path Way Diagrams. Each square denotes an observable characteristic 𝑝𝑖𝑗 for individual 𝑗 in pair 𝑖. A

characteristic can be a personality trait (𝑥𝑖𝑗) or a financial behavior (𝑦𝑖𝑗). Each circle denotes an unobservable factor such as

additive genetic component (A), the shared environment (C), and the non-shared environment (E).

Figure B2: Bivariate ACE Path Way Diagram. Each square denotes an observable characteristic 𝑝𝑖𝑗 for individual 𝑗 in pair 𝑖. A

characteristic can be a personality trait (𝑥𝑖𝑗) or a financial behavior (𝑦𝑖𝑗). Each circle denotes an unobservable factor such as

additive genetic component (A), the shared environment (C), and the non-shared environment (E). Assume 𝐶𝑜𝑟(𝐴1𝑗𝑝

, 𝐴2𝑗𝑝

) = 1

44

for MZ twins, 𝐶𝑜𝑟(𝐴1𝑗𝑝

, 𝐴2𝑗𝑝

) = 0.5 for DZ twins and FS, 𝐶𝑜𝑟(𝐴1𝑗𝑝

, 𝐴2𝑗𝑝

) = 0.25 for HS, and 𝐶𝑜𝑟(𝐶1𝑗𝑝, 𝐶2𝑗

𝑝) = 1 and

𝐶𝑜𝑟(𝐸1𝑗𝑝

, 𝐸2𝑗𝑝

) = 0. The bivariate model further assumes the A, C, and E that affects a personality trait also affect financial

behavior.

Figure B3: Multivariate Analysis. Each square denotes an observable characteristic and 𝑝𝑖𝑗 for individual 𝑗 in pair 𝑖 . The

characteristics 𝑥𝑖𝑗 and 𝑧𝑖𝑗 can be a personality trait and/or cognitive ability, the characteristic 𝑦𝑖𝑗 is the latent financial distress.

Each circle denotes an unobservable factor such as additive genetic component (A), the shared environment (C), and the non-

shared environment (E). Assume 𝐶𝑜𝑟(𝐴1𝑗𝑝

, 𝐴2𝑗𝑝

) = 1 for MZ twins, 𝐶𝑜𝑟(𝐴1𝑗𝑝

, 𝐴2𝑗𝑝

) = 0.5 for DZ twins and FS, 𝐶𝑜𝑟(𝐴1𝑗𝑝

, 𝐴2𝑗𝑝

) =

0.25 for HS, and 𝐶𝑜𝑟(𝐶1𝑗𝑝, 𝐶2𝑗

𝑝) = 1 and 𝐶𝑜𝑟(𝐸1𝑗

𝑝, 𝐸2𝑗

𝑝) = 0. The multivariate model further assumes that the A, C, and E that

affect 𝑥𝑖𝑗 and 𝑧𝑖𝑗 also affect 𝑦𝑖𝑗.

45

Table 1: Variable Definitions

This table provides the survey questions from a 20-item short-form version of the International Personality Item Pool-Five-Factor

Model (i.e., the Mini-IPIP) used in the Add Health Wave IV survey to measure the Big Five personality traits.

Variable Add Health Wave IV Survey Question

Financial distress

no utility payment In the past 12 months, was there a time when {YOU/YOUR HOUSEHOLD} didn't pay

the full amount of a gas, electricity, or oil bill because you didn't have enough money?

no phone service In the past 12 months, was there a time when {YOU/YOUR HOUSEHOLD} was

without phone service because you didn't have enough money?

no mortgage/rent payment In the past 12 months, was there a time when {YOU/YOUR HOUSEHOLD} didn't pay

the full amount of the rent or mortgage because you didn't have enough money?

past year worried food depleted In the past 12 months, was there a time when {YOU/YOUR HOUSEHOLD

WERE/WAS} worried whether food would run out before you would get money to buy

more?

Personality traits 'How much do you agree with each statement about you as you generally are now, not as

you wish to be in the future?'

Conscientiousness get chores done right away

forget return things properly

I like order

I make a mess of things

Neuroticism have frequent mood swings

relaxed most of the time

I get upset easily

I seldom feel blue

Extraversion I keep in the background

socialize freely at parties

do not talk a lot

I am the life of the party

Agreeableness sympathize w/ others' feelings

disinterest in others' problems

feel others' emotions

uninterested in others

Openness have vivid imagination

46

not interested abstract ideas

abstract concepts hard to get

lack good imagination

47

Table 2: Summary Statistics by Sibling Type

This table reports the means and standard deviations of the baseline sample. The sample includes 306 MZ twin sibling

pairs, 437 DZ twin sibling pairs, 1,162 full sibling pairs, and 327 half sibling pairs.

MZ DZ FS HS

mean s.d. mean s.d. mean s.d. mean s.d.

Financial distress

no utility payment 0.100 0.310 0.160 0.360 0.130 0.340 0.250 0.440

no phone service 0.070 0.260 0.110 0.310 0.100 0.290 0.180 0.380

no mortgage/rent payment 0.050 0.230 0.090 0.290 0.080 0.280 0.180 0.380

past year worried food depleted 0.080 0.280 0.130 0.330 0.110 0.310 0.200 0.400

Personality traits

Conscientiousness 14.850 2.700 14.920 2.600 14.670 2.660 14.440 2.750

Neuroticism 10.410 2.640 10.290 2.820 10.430 2.760 11.000 2.770

Extraversion 13.180 3.100 13.330 3.040 13.050 3.030 13.010 3.080

Agreeableness 15.230 2.370 15.190 2.390 15.140 2.490 14.940 2.400

Openness 14.380 2.330 14.500 2.410 14.240 2.560 14.200 2.440

Cognitive Ability

PVT1STD 98.310 15.580 96.780 15.070 98.420 15.150 92.620 15.840

PVT3STD 96.920 17.130 97.360 15.380 98.730 15.390 92.220 17.730

Demographics

sex 0.500 0.500 0.510 0.500 0.490 0.500 0.490 0.500

age (by 12/31/2008) 29.620 1.550 29.430 1.630 29.650 1.640 29.190 1.760

white 0.540 0.500 0.540 0.500 0.550 0.500 0.400 0.490

black 0.200 0.400 0.280 0.450 0.170 0.380 0.390 0.490

Hispanic 0.160 0.370 0.140 0.340 0.150 0.360 0.160 0.370

Asian 0.060 0.240 0.030 0.160 0.100 0.300 0.020 0.150

other race 0.030 0.180 0.030 0.160 0.030 0.160 0.030 0.160

48

Table 3: Pairwise Correlations by Sibling Type

This table reports the pair-wise correlation coefficients and the p-values. The sample includes 306 MZ twin sibling pairs,

437 DZ twin sibling pairs, 1,162 full sibling pairs, and 327 half sibling pairs.

MZ DZ FS HS

corr. p-value corr. p-value corr. p-value corr. p-value

Financial distress

no utility payment 0.189 0.004 0.090 0.101 0.125 0.000 0.103 0.132

no phone service 0.125 0.054 0.105 0.056 0.165 0.000 0.068 0.323

no mortgage/rent payment 0.214 0.001 0.010 0.850 0.056 0.100 0.026 0.705

past year worried food depleted 0.143 0.027 0.127 0.021 0.154 0.000 -0.029 0.672

Personality traits

Conscientiousness 0.313 0.000 0.080 0.147 0.077 0.025 0.029 0.667

Neuroticism 0.201 0.002 0.050 0.365 0.106 0.002 0.068 0.320

Extraversion 0.417 0.000 0.089 0.105 0.089 0.009 -0.046 0.505

Agreeableness 0.440 0.000 0.163 0.003 0.184 0.000 0.176 0.009

Openness 0.360 0.000 0.188 0.001 0.177 0.000 0.094 0.172

Cognitive Ability

PVT1STD 0.782 0.000 0.516 0.000 0.597 0.000 0.421 0.000

PVT3STD 0.748 0.000 0.451 0.000 0.582 0.000 0.321 0.000

Demographics

sex 1.000 0.000 0.103 0.032 0.137 0.000 0.010 0.856

age (by 12/31/2008) 1.000 0.000 1.000 0.000 0.092 0.002 0.031 0.582

white 0.947 0.000 0.945 0.000 0.944 0.000 0.886 0.000

black 0.980 0.000 0.972 0.000 0.976 0.000 0.923 0.000

Hispanic 0.940 0.000 0.980 0.000 0.935 0.000 0.798 0.000

Asian 0.944 0.000 0.956 0.000 0.967 0.000 0.813 0.000

other race 0.754 0.000 0.349 0.000 0.453 0.000 0.096 0.082

49

Table 4: Heritability of Financial Distress, Personality, and Cognitive Ability

This table reports the estimates and the 95% confidence intervals of the relative importance of the additive genetic component

(𝐴

𝐴+𝐶+𝐸), the shared environment (

𝐶

𝐴+𝐶+𝐸), and the non-shared environment (

𝐸

𝐴+𝐶+𝐸) are reported. The confidence intervals are

based on 1000 bootstrap draws. The sample includes 306 MZ twin sibling pairs, 437 DZ twin sibling pairs, 1,162 full sibling

pairs, and 327 half sibling pairs.

Raw data

A/(A+C+E) C/(A+C+E) E/(A+C+E)

Financial distress indicators

no utility payment .286 [.000, .557] .104 [.000, .302] .610 [.420, .828]

no phone service .248 [.000, .660] .176 [.000, .360] .576 [.323, .794]

no mortgage/rent payment .372 [.000, .572] .000 [.000, .129] .628 [.424, .924]

past year worried food depleted .492 [.000, .662] .009 [.000, .272] .498 [.326, .806]

Personality traits

Conscientiousness .225 [.133, .312] .000 [.000, .000] .775 [.685, .867]

Neuroticism .180 [.000, .284] .011 [.000, .120] .809 [.713, .922]

Extraversion .283 [.186, .372] .000 [.000, .000] .717 [.627, .810]

Agreeableness .414 [.290, .506] .000 [.000, .000] .586 [.493, .693]

Openness .384 [.274, .462] .000 [.000, .055] .616 [.536, .711]

Cognitive Ability

PVT1STD .324 [.223, .431] .415 [.335, .484] .260 [.214, .308]

PVT3STD .448 [.292, .595] .275 [.178, .376] .277 [.211, .349]

50

Table 5: Genetic Correlation and Bivariate Heritability

The first two columns represent standardized regression coefficients for the genetic (ba) and non-shared

environmental (be) effects of the psychological variables on latent financial distress. We did not estimate shared

environmental associations because latent financial distress did not show evidence of shared environmental effects,

and included these cross-pathways for cognitive ability dramatically inflated standard errors of the other parameters.

Constraining the shared environmental association to zero did not lead to model misfit (p’s > .19 and ΔCFI = .001).

The next two columns represent the genetic and non-shared environmental correlation. The final two columns

represent bivariate heritability (i.e., the proportion of the phenotypic association due to common genetic effects)

and bivariate non-shared environmentality (i.e., the proportion of the phenotypic association due to common non-

shared environmental effects). Negative values are possible if the genetic and environmental effects are in opposite

directions. Genetic and non-shared environmental correlations and bivariate heritability and environmentality were

calculated based on results from the Cholesky decomposition (Loehlin, 1996). As such, we only report inferential

statistics for the b’s. The sample includes 306 MZ twin sibling pairs, 437 DZ twin sibling pairs, 1,162 full sibling

pairs, and 327 half sibling pairs.

p < .05; ** p < .01; *** p < .001

Latent Financial Distress

(1) (2) (3) (4) (5) (6)

ba be rA rE Bi-h2 Bi-e2

Personality traits

Conscientiousness

-.115

(.093)

-.127*

(.050)

-.155 -.183 .320 .680

Neuroticism

.339**

(.100)

.154**

(.051)

.458 .200 .522 .478

Extraversion

-.099

(.089)

.018

(.055)

-.134 .026 1.436 -.436

Agreeableness

-.152*

(.071)

.064

(.062)

-.205 .092 1.920 -.920

Openness

-.054

(.079)

-.049

(.061)

-.073 -.072 .453 .547

Cognitive Ability

PVT1STD

-.406

(.099)

***

.182

(.081)

*

-.548 .271 1.475 -.475

PVT3STD -.417

(.084)

.222

(.087) -.563 .330 1.580 -.580

51

*** *

52

Table 6 Gene × SES Interaction Parameters

In Model 1, 𝑏1 and 𝑏2 are estimated from 𝐹 = 𝑏1 × 𝑆𝐸𝑆 + 𝑏2 × 𝑆𝐸𝑆2 + 𝜖, where 𝐹 is the latent financial distress, and the SES

is the socioeconomic status. In Model 2, 𝑏1 and 𝑏2 are estimated from 𝐹 = 𝑏1 × 𝑆𝐸𝑆 + 𝑏2 × 𝑆𝐸𝑆2 + 𝑏3 × Neu + 𝑏4 × Neu ×

𝑆𝐸𝑆 + 𝑏5 × Cog + 𝑏6 × Cog × 𝑆𝐸𝑆 + 𝜖 , where Neu is neuroticism and Cog is cognitive ability. The residual latent financial

distress, 𝜖, from Model 1 and Model 2 respectively, is further decomposed into the genetic effect, 𝐴, and environmental effect, 𝐸,

in the following form :𝐴 = 𝑎 + 𝑎′ × 𝑆𝐸𝑆 + 𝑎′′ × 𝑆𝐸𝑆2 and 𝐸 = 𝑒 + 𝑒′ × 𝑆𝐸𝑆 + 𝑒′′ × 𝑆𝐸𝑆2. The sample includes 274 MZ

twin sibling pairs, 404 DZ twin sibling pairs, 1,078 full sibling pairs, and 290 half sibling pairs where the SES information is

available.

Model 1 Model 2

(1) (2) (3) (4) (5) (6)

Parameters Estimate S.E. p-value Estimate S.E. p-value

𝑏1 -1.249 (.225) < .001 -1.041 (.229) < .001

𝑏2 -.337 (.124) .006 -.347 (.115) .003

𝑎 2.350 (.311) < .001 2.211 (.300) <.001

𝑎′ .184 (.168) .274 .135 (.285) .635

𝑎′′ .309 (.122) .011 .129 (.179) .472

𝑒 2.474 (.328) < .001 2.099 (.299) <.001

𝑒′ .069 (.246) .778 .126 (.360) .727

𝑒′′ -.548 (.167) .001 .038 (.176) .827