J Real Estate Finan Econ (2014) 48:79–111 DOI 10.1007/s11146-012-9373-0

Genetics, Homeownership, and Home Location Choice Henrik Cronqvist · Florian Münkel · Stephan Siegel

Published online: 30 May 2012 © Springer Science+Business Media, LLC 2012

Abstract We find that a significant proportion of the cross-sectional variation in the choice to own or rent is attributable to a genetic factor, while parental influence is not found to affect this choice. We also find evidence of gene-environment interactions: The environment moderates genetic effects on homeownership in that growing up in a wealthier family results in a stronger expression of genetic predispositions, while idiosyncratic life experiences appear to explain a larger portion of the variation in homeownership among those who grew up in a less wealthy family environment. Furthermore, we find that home location choices, for example, a familiar home location close to one’s birthplace and an urban versus a rural home location, are explained by both genetic factors and parental influence. Because we control for an extensive set of individual characteristics analyzed in existing research, an interpretation of our evidence is that an individual’s preferences with respect to homeownership and home location are partly genetic. The findings contribute to a deeper understanding of the factors that explain individual behavior with respect to the housing market, and add to an expanding literature on the biological and genetic factors that influence individuals’ economic and financial decisions.

H. Cronqvist Robert Day School of Economics and Finance, Claremont McKenna College, Bauer Center 319, 500 E. Ninth Street, Claremont, CA 91711, USA e-mail: [email protected] F. Münkel · S. Siegel (B) Foster School of Business, University of Washington, Box 353226, Seattle, WA 98195-3226, USA e-mail: [email protected] F. Münkel e-mail: [email protected]

80

H. Cronqvist et al.

Keywords Homeownership · Home location choice · Real estate · Housing · Genetics

Introduction Variation across individuals in homeownership and home location choices may be explained by three different factors: (i) a genetic factor, i.e., individuals are born with a predisposition to a particular behavior, (ii) parental influence, or “nurture,” and (iii) individual-specific environmental factors that an individual is subject to during his or her life. Our study is the first in the real estate literature to report empirical evidence on the extent to which individual homeownership and home location choices are explained by each of these factors. Our analysis focuses on several key choices that any individual must make. First, we study the choice to own one’s home rather than to rent it (“tenure choice”). We also analyze the amount of housing consumption, conditional on being a homeowner. Second, we study the home location choice, i.e., where to live. Specifically, we analyze whether or not an individual chooses a home location close to where he or she was born, and we also analyze the choice to live in an urban versus a rural location. There are at least two reasons why a study of genetics, homeownership, and home location choice is important. First, real estate researchers and economists who model or empirically examine the housing market want to understand the deeper origins of individual choices (see, e.g., Benjamin et al. 2008, p. 16). A recent and rapidly expanding set of papers shows that individuals’ financial decisions are significantly related to genetic and neural factors (e.g., Cesarini et al. 2009; Kuhnen and Chiao 2009; Sapienza et al. 2009; Barnea et al. 2010), but we are not aware of any related work in real estate economics. Second, it may be important from a public policy perspective to understand the extent to which individuals’ choices are influenced by genetic predispositions rather than environmental factors, the latter presumably being more sensitive to policy intervention (e.g., Bernheim 2009). An empirical decomposition of individual behavior into genetic, parenting, and individual-specific environmental factors requires data on the choices of individuals with known genetic differences, e.g., twins.1 The intuition of this approach is as follows: If twins that are genetically identical are significantly more correlated with respect to their homeownership and home location choices than fraternal twins, then those choices can, at least in part, be attributed to a genetic factor. For example, in our data set, we find that the

1 The

use of twin data has a very long tradition in applied microeconomic research (e.g., Behrman and Taubman 1976, 1989; Taubman 1976; Ashenfelter and Krueger 1994; Behrman et al. 1994; Ashenfelter and Rouse 1998).

Genetics, Homeownership, and Home Location Choice

81

correlation of being a homeowner is 0.64 for identical twins versus only 0.40 for fraternal twins. Such correlation analysis suggests that genes influence individual choices in the housing market, but it is only a starting point. We also estimate models used in quantitative behavioral genetics (see, e.g., Neale and Maes 2004 for an overview). We construct a large and comprehensive data set of the homeownership and home location choices of more than 25,000 twins. These data are from the Swedish Twin Registry (STR), which maintains and manages the world’s largest research database of twins. The data have been matched with socioeconomic characteristics and home location data from Statistics Sweden and the Swedish Tax Authority, real estate data from the Swedish Mapping, Cadastral and Land Registration Authority, and distance data from Google Maps. Our evidence may be summarized as follows. Controlling for individual socioeconomic characteristics, we find that the proportion of the residual crosssectional variation in homeownership attributable to a genetic factor is 45% for tenure choice and 23% for housing consumption. That is, the choice to be a homeowner is explained by a significant genetic factor. Because we control for an extensive set of individual characteristics analyzed in the existing real estate literature, an interpretation of the evidence is that individuals’ preferences with respect to homeownership are partly genetic. We also find evidence of gene-environment interactions: The environment moderates genetic effects on homeownership in that growing up in a wealthier family results in stronger expression of genetic predispositions, while idiosyncratic life experiences appear to explain a larger portion of the variation in homeownership among those who grew up in a less wealthy family environment. Finally, we find that home location preferences are partly genetic. Importantly, we also find, differently from the evidence on time, risk, and homeownership preferences (see, e.g., Cesarini et al. 2010; Cronqvist and Siegel 2011), that the family environment explains a significant portion of the variation (32–38%), suggesting that parenting affects home location choices. Our study is related to several existing strands of research. First, our paper is related to an extensive set of papers on tenure choice (see, e.g., the references in Section “Related Research”), but is the first to report empirical evidence on the extent to which homeownership is explained by genetic and environmental factors. Second, while existing research reports evidence of intergenerational transmission of homeownership and home location choices (see, e.g., Mulder and Smits 1999; Charles and Hurst 2003; Helderman and Mulder 2007; Blaauboer 2011), our study is the first to analyze whether such parent-child similarities are because of genes or parenting. Finally, our study is part of “nature versus nurture” research in social science. On the one hand, work in psychology shows that most personality traits have a significant genetic component (e.g., Bouchard et al. 1990) and that parents have little long-lasting effects on their children’s personality (e.g., Harris 1995). Consistent with this evidence, economists have recently shown that genes explain, e.g., an individual’s savings-consumption choice (Cronqvist and Siegel 2011), portfolio choices (e.g., Barnea et al. 2010; Cesarini et al.

82

H. Cronqvist et al.

2010), and several other economic behaviors (Cesarini et al. 2012).2 On the other hand, other work suggests that parenting may influence some choices (e.g., Bisin and Verdier 2000; Sacerdote 2002, 2007; Bisin et al. 2004; Fernandez et al. 2004). The rest of the paper is organized as follows. The next section contains an overview of related research. Section “Data and Methodology” describes our data sources, defines our variables, reports summary statistics, and explains our methodology. Section “Results” reports our results and robustness checks. The last section concludes.

Related Research Homeownership In an economy with complete markets, infinitely divisible assets, no transactions costs, and non-distorting tax systems, individuals’ housing choices may be interpreted as revealed preferences. Some models in real estate economics assume or explicitly model such a preference for homeownership (e.g., Hu 2005; Cauley et al. 2007). In reality, the housing market is characterized by significant market imperfections such that homeownership choices are also affected by determinants other than preferences.3 In our review of related research, we focus on individual characteristics that existing studies have found to (i) affect an individual’s homeownership choice and (ii) have a significant genetic component.4 One important set of factors is an individual’s wealth and income, including borrowing constraints. Several papers have studied the effects of wealth and income on tenure choice (e.g., Struyk and Marshall 1973, 1975; Jones 1989, 1990). Others have examined detailed data directly related to borrowing constraints, e.g., credit quality and other criteria for obtaining a home mortgage (e.g., Linneman and Wachter 1989; Zorn 1989; Haurin et al. 1997, Barakova

2 The

potential importance of research in the intersection of economics and biology has long been recognized by economists (e.g., Becker 1976; Hirshleifer 1977). For an overview of research at the intersection of genetics, neurobiology, and economics, we refer to Camerer et al. (2005) and Benjamin et al. (2008). 3 For models of individual housing tenure choices, see, e.g., Hendershott and Shilling (1982), Henderson and Ioannides (1983), Linneman (1985), and Mills (1990). 4 Our review focuses on the tenure choice literature. In the empirical analysis we also examine the housing consumption choice, i.e., the amount of housing chosen, conditional on being a homeowner. Existing work has identified housing consumption factors that include the same variables that have been shown to also affect the own versus rent decision (e.g., Goodman 1988; Henderson and Ioannides 1989; Davidoff 2006).

Genetics, Homeownership, and Home Location Choice

83

et al. 2003).5 The conclusion from these studies is that wealth, income, and borrowing constraints can explain significant tenure choice differences in the cross-section of individuals even in developed capital markets. At the same time, several studies have shown that wealth and income are partly genetic. Charles and Hurst (2003) find a positive and significant ageadjusted elasticity of child wealth with respect to parental wealth (controlling for past parental gifts and expected future bequests), and other studies show that such intergenerational transmission of wealth and income has a significant genetic component (e.g., Behrman and Taubman 1976; Taubman 1976; Cronqvist and Siegel 2011). Another set of factors affecting tenure choice is an individual’s education, family size and structure (e.g., being married and having children), and income uncertainty (e.g., Gyourko and Linneman 1996, 1997). For example, individuals with higher education have been found to be more likely to be homeowners. Being married and having children also increases the likelihood of being a homeowner. Other studies find that income uncertainty, controlling for the level of income, reduces the likelihood of individuals owning a home (e.g., Haurin and Gill 1987; Haurin 1991; Haurin et al. 1997; Robst et al. 1999). Several of these factors, e.g., education, have a significant genetic component (e.g., Taubman 1976; Behrman and Taubman 1989; Ashenfelter and Krueger 1994; Behrman et al. 1994; Ashenfelter and Rouse 1998). Behavioral genetics researchers have also found that family structure, e.g., being married, is partly explained by a genetic factor (e.g., McGue and Lykken 1992; Jocklin et al. 1996). Nicolaou and Shane (2010) report that occupational choice has a significant genetic component, and research has also found that the choice to be self-employed is genetic (e.g., Nicolaou et al. 2008; Zhang et al. 2009), indicating that income uncertainty may be partly genetic.6 Finally, individual risk preferences can affect an individual’s tenure choice, though the direction of the effect is ambiguous. On the one hand, the standard view in real estate economics has been that owning a home is risky because house prices are volatile and homeowners on average allocate a very significant portion of their net worth to their house. As a result, more risk averse

5 Most

papers in the tenure choice literature do not have data on credit history, FICO scores, or similar measures, a notable exception being the pseudo credit scores constructed by Barakova et al. (2003). Sweden has a system with a betalningsanmärkning, or “payment default score” (i.e., the Swedish system is one with only two credit scores: good or bad credit). Fewer than 10% in the overall population have a payment default score indicating bad credit, so even if we had access to such data, we believe the usefulness would be limited in our study because of the very skewed distribution. Barakova et al. (2003) conclude that wealth is the most important constraint, though it has diminished in importance in recent years, while credit quality has increased in importance as a constraint affecting homeownership. See, e.g., Brueckner (1986) for a model of the downpayment constraint on housing tenure choice. 6 Shore (2011) finds that income volatility is transmitted from one generation to the next, but without observing the channel of such transmission.

84

H. Cronqvist et al.

individuals may choose to rent. On the other hand, Ortalo-Magné and Rady (2002) propose a model in which fluctuating housing rents constitute risk and Sinai and Souleles (2005) empirically show that homeownership is positively related to the variability of housing rents, providing support for the hypothesis that homeownership can act as hedge against changes in rental costs. Recent research shows that individuals’ risk preferences are partly genetic. Barnea et al. (2010) and Cesarini et al. (2010) examine individuals’ financial portfolio choices and find that a genetic factor explains about one third of the variation in portfolio choices, which they attribute to innate differences in risk preferences.7 Researchers have also examined the extent to which the presence of specific gene(s) may explain differences in risk aversion in the cross-section of individuals. For example, Kuhnen and Chiao (2009), Dreber et al. (2009), and Zhong et al. (2009) elicit financial risk preferences using experiments and find that those with specific genes have significantly different financial risktaking behavior compared to those who do not have those genes. If we find a significant genetic component of homeownership when we do not control for individual characteristics, then it is possible that the entire genetic component is attributable to genetic variation in wealth, education, risk aversion, etc. If we find that a significant genetic component remains even after we control for these characteristics, then we may attribute the effect to a genetic homeownership preference. Home Location Choice In an economy with no frictions, individuals may simply trade off housing costs and commuting costs to the work location (e.g., Alonso 1964; Freedman and Kern 1997). However, in reality, the choice of where to live is determined by market imperfections combined with preferences (e.g., McFadden 1977). Familiar Home Location While economists have shown that individuals respond to increased pay in other locations (e.g., Topel 1986), others have found that the sensitivity to labor market opportunities in other locations is small (e.g., Davies et al. 2001). Indeed, Dahl and Sorenson (2010) show that the distance from one’s birthplace is the most important factor explaining home location choice. This apparent preference for a familiar location has been long recognized (e.g., Mincer 1978). There are several economic benefits from such a home location choice. For example, when searching for opportunities in the labor market, local information, and location-specific capital may reduce search

7 Cesarini

et al. (2009) and Zyphur et al. (2009) also report that an individual’s risk preferences have a significant genetic component. Using experiments involving lottery choices and questionnaires, they elicit subjects’ risk preferences and then use twin research methodology to estimate genetic effects.

Genetics, Homeownership, and Home Location Choice

85

costs. There are also other benefits, such as closeness to family and a familiar culture. It is also possible that individuals are born with preferences for the familiar. A recent gene association study by Chew et al. (2012) identifies the genes that affect familiarity preferences. In addition, the neuroimaging study by Hsu et al. (2005) shows that certain parts of the brain were predictably more active under the condition of ambiguity than under the condition of familiarity. We hypothesize that the preference for a familiar home location is partly genetic. Urban or Rural Home Location While an urban or rural home location choice may vary over the life-cycle because of, e.g., age and family structure, existing research also suggests that a persistent factor affects this choice. For example, a previous experience of living in a city increases the probability of return migration to that city (e.g., Feijten et al. 2008), and having lived in a rural location increases the probability of choosing another rural home location (e.g., Van Dam et al. 2002). There is also evidence that the parents’ home location choice is a more important determinant of a individual’s choice compared to an individual’s own previous experiences (e.g., Blaauboer 2011). This evidence raises the question of whether such parent-child similarity in home location choices is innate because some individuals are born with a genetic propensity to choose an urban location over a rural one, or vice versa. The closest study to ours is Whitfield et al. (2005), who suggests that a genetic factor affects this choice. In the empirical analysis, we control for whether the individual lives where he or she was born as well as other socioeconomic characteristics so that we do not confound preferences for a familiar home location with preferences for an urban versus rural home location.

Data and Methodology Data Sources The data set used in this paper was constructed by matching a large number of identical and fraternal twins from the STR, the world’s largest twin registry, with data from several other databases. In Sweden, twins are registered at birth, and the STR collects additional data through in-depth interviews.8 We

8 STR’s databases are organized by birth cohort. The Screening Across Lifespan Twin, or “SALT,”

database contains data on twins born 1886–1958. The Swedish Twin Studies of Adults: Genes and Environment database, or “STAGE,” contains data on twins born 1959–1985. In addition to twin pairs, twin identifiers, and zygosity status, the databases contain variables based on STR’s telephone interviews (for SALT), completed 1998–2002, and combined telephone interviews and Internet surveys (for STAGE), completed 2005–2006. For further details about STR, we refer to Lichtenstein et al. (2006).

86

H. Cronqvist et al.

have data on the zygosity of each twin pair: Identical, or “monozygotic” (MZ), twins are genetically identical, while fraternal, or “dizygotic” (DZ), twins are genetically different.9 The STR is recognized worldwide for the quality of its data, which have been used in a large number of studies. Data on individuals’ homes are from the Real Property Register, managed and maintained by the Swedish Mapping, Cadastral and Land Registration Authority. Socioeconomic data on the individuals and the locations of their homes are from Statistics Sweden. Tax filings data, such as disposable income, are from the Swedish Tax Authority. Home values and socioeconomic and tax data refer to 2002. Data on distances are from Google Maps. Sample Selection and Summary Statistics Our sample selection involves the following steps. First, our methodology requires us to include only twin pairs with complete data for both twins. In addition, an individual must be over 18 years old to be included in our sample. Finally, we exclude twins with no financial data or negative disposable household income. Our final sample consists of 26,876 twins with data on their housing choices at the end of 2002. Panel A of Table 1 reports summary statistics for our data set of twins. 30% percent are identical twins; the rest are fraternal twins. Opposite-sex twins are the most common (38%); identical male twins are the least common (14%). The distribution of twins in the table is consistent with what would be expected from large populations (e.g., Bortolus et al. 1999). Panel B reports summary statistics for the homeownership and home location variables. Tenure Choice is an indicator variable that is one if an individual or his/her spouse owns their primary home, and zero otherwise. We find that 60% of identical twins and 69% of fraternal twins own their primary residence in 2002. As a comparison, the 2002 U.S. Census reports that about 68% of Americans own their home, suggesting that the homeownership rate, as well as variation in homeownership, in Sweden are similar relative to the U.S. For homeowners, we make the common assumption that housing services consumed are proportional to the value of the owner-occupied house, i.e., Housing Consumption is measured as the natural logarithm of the home value divided by the average house price in the same municipality.10 We measure familiarity of home location with: (i) Live Where Born, an indicator variable that is one if an individual lives in the same state (län) in

9 Zygosity

is based on questions about intrapair similarities in childhood. One of STR’s survey questions was: Were you and your twin partner during childhood “as alike as two peas in a pod” or were you “no more alike than siblings in general” with regard to appearance? This method has been validated with DNA as having 98 percent or higher accuracy. For twin pairs for which DNA sampling has been conducted, zygosity status based on DNA analysis is used. 10 Home values are based on assessed tax values from the Swedish Tax Authority, which are computed based on actual transaction values for similar properties in the area of the home.

All twins

26,876 100% All twins N

3,314 12%

Tenure Choice Housing Consumption Mean house value in municipality (USD) Live Where Born Distance to Birthplace (km) Urban Suburban Rural Distance to Urban Center (km) Population Density (per km2 )

26,876 13,718 26,876 26,810 26,390 26,876 26,876 26,876 26,633 26,876

Panel B: Homeownership and home location variables

Variable

Number of twins (N) Fraction (%)

7,984 30%

Total

0.60 −0.35 139,253 0.76 96.11 0.44 0.49 0.07 27.02 562.96

1.00 −0.10 114,656 1.00 0.00 0.00 0.00 0.00 14.10 78.70

Identical twins Mean Median

4,670 17%

Identical twins Male Female

Panel A: Number of twins by genetic similarity and gender

Table 1 Summary statistics

0.49 0.93 83,787 0.43 187.51 0.50 0.50 0.25 38.35 1,144.80

Std. dev.

3,734 14%

Fraternal twins Same sex: Male

10,208 38%

Opposite sex

0.69 −0.31 129,984 0.77 91.70 0.42 0.51 0.07 29.65 497.63

1.00 −0.06 105,367 1.00 0.00 0.00 1.00 0.00 17.40 64.90

Fraternal twins Mean Median

4,950 18%

Same sex: Female

0.46 0.91 79,306 0.42 182.67 0.49 0.50 0.26 39.50 1,099.75

Std. dev.

18,892 70%

Total

Genetics, Homeownership, and Home Location Choice 87

43.51 0.18 0.46 0.35 0.41 0.99 0.16 82,372 36,224 0.23 0.85 0.47

26,876 26,876

1.00 0.43

45.00 0.00 0.00 0.00 0.00 0.00 0.00 32,647 28,968 0.20

Identical twins Mean Median

26,876 26,876 26,876 26,876 26,876 26,876 26,876 26,876 26,876 26,876

All twins N

0.36 0.38

15.87 0.39 0.50 0.48 0.49 1.32 0.37 175,049 28,971 0.15

Std. dev.

0.84 0.45

49.61 0.25 0.45 0.30 0.51 0.84 0.18 105,371 42,322 0.21

1.00 0.40

52.00 0.00 0.00 0.00 1.00 0.00 0.00 52,066 35,961 0.18

Fraternal twins Mean Median

0.37 0.37

14.04 0.43 0.50 0.46 0.50 1.19 0.39 490,584 53,421 0.14

Std. dev.

Panel A provides information about the number of identical and non-identical twins used in this study. Panel B reports summary statistics for the Homeownership and Home Location variables. Panel C provides summary statistics for individual socioeconomic characteristics. All United States Dollar (USD) values were calculated using the exchange rate at the end of 2002 (SEK/USD 8.6815). All variables are defined in detail in Appendix Table 8

Age Education level 1 Education level 2 Education level 3 Married Number of children in household Poor health Net worth (USD) Disposable household income (USD) St. dev. of log growth rate of disposable household income Stock market participation Share in equities

Panel C: Socioeconomic characteristics

Variable

Table 1 (continued)

88 H. Cronqvist et al.

Genetics, Homeownership, and Home Location Choice

89

which he or she was born, and (ii) Distance to Birthplace, the distance between the individual’s current home location and the birth state.11 The average driving distance to the state of birth is about 96 km for identical and 92 km for fraternal twins. Finally, we construct three measures to capture the choice between living in an urban or rural area: (i) Urban, Suburban, and Rural are indicator variables based on Statistics Sweden’s classification, (ii) Distance to Urban Center is the distance to the closest municipality classified as urban, and (iii) Population Density is the population density of the municipality where the individual lives. The average distance to an urban center is 27 km for identical twins and 30 km for fraternal twins. Importantly, we find that identical and fraternal twins, in aggregate, are very similar with respect to all analyzed housing choices, but significant variation exists across individuals. Panel C reports summary statistics for socioeconomic characteristics that have been shown to influence homeownership and home location choices, such as age, education, family structure, wealth, and income. If an individual is married, we measure these characteristics, if applicable, at the household-level. All variables are defined in Appendix Table 8. Methodology To decompose the cross-sectional variation in homeownership and home location choices into genetic and environmental components, we model each behavior or outcome yij for twin j (1 or 2) of pair i as a possibly nonlinear function of observable characteristics Xij as well as three unobserved effects. In particular, we assume that in addition to Xij, yij reflects an additive genetic effect, aij, an effect of the environment common to both twins (e.g., parenting), ci , and an individual-specific error term, eij, that represents idiosyncratic environmental effects (e.g., life experiences) as well as measurement error, that is,: yij = f (Xij, aij, ci , eij).

(1)

We assume that aij, ci , and eij are uncorrelated with one another and across twin pairs and normally distributed with zero means and variances σa2 , σc2 , and σe2 , respectively, so that the total residual variance σ 2 is the sum of the three variance components. Identifying variation due to aij, ci , and eij separately is possible due to covariance restrictions suggested by genetic theory. Consider two twin pairs

11 As

we do not observe the exact location of an individual’s birthplace within a state, i.e., we do not have data on the municipality where an individual was born, we use the population-weighted average distance to all municipalities in the birth state. We set Distance to Birthplace to zero for individuals that live in the state of their birth.

90

H. Cronqvist et al.

i = 1, 2 with twins j = 1, 2 in each pair, where the first pair is a pair identical twins and the second pair is a pair of fraternal twins. The corresponding genetic effects are: a = (a11 , a12 , a21 , a22 ) . Analogously, the vectors of common and idiosyncratic environmental effects are: c = (c11 , c12 , c21 , c22 ) and e = (e11 , e12 , e21 , e22 ) . Identical and fraternal twin pairs differ in their genetic similarity. Identical twins are genetically identical and the correlation between a11 and a12 is set to one. Fraternal twins, however, share on average only 50% of their genetic make up, such that the correlation between a21 and a22 is set to 0.5. For both types of twin pairs, the same common (parenting) environment is assumed. Thus, we impose the following covariance matrices: ⎡ ⎤ ⎡ ⎤ ⎡ ⎤ 110 0 1100 1000 ⎢1 1 0 0⎥ ⎢ ⎥ ⎢ ⎥ 2 ⎢1 1 0 0⎥ 2 ⎢0 1 0 0⎥ ⎥ Cov(a) = σa2 ⎢ ⎣ 0 0 1 1 ⎦ , Cov(c) = σc ⎣ 0 0 1 1 ⎦ , Cov(e) = σe ⎣ 0 0 1 0 ⎦ . 2 0011 0001 0 0 12 1 Specifically, for continuous outcomes, such as Housing Consumption, we assume that f is a linear function, such that: yij = β0 + βXij + aij + ci + eij ,

(2)

where β0 is an intercept term and β measures the effects of the observable covariates (Xij ). We use maximum likelihood to estimate the model; see, e.g., Neale and Maes (2004). For censored outcomes, such as the Distance to Birthplace and the Distance to Urban Center, we estimate a Tobit model. That is, we assume a linear model for an underlying latent variable, y∗ij, y∗ij = β0 + βXij + aij + ci + eij ,

(3)

where y∗ij equals the observed outcome, yij, if larger than zero, but equals zero in all other cases. For binary and ordered categorical outcomes, such as Tenure Choice or the location choice between Urban, Suburban, and Rural, we assume a linear relationship between a latent response variable y∗ij and the observable as well as unobservable characteristics:12 y∗ij = βXij + aij + ci + eij,

(4)

We estimate a threshold (or in case of ordered categorical outcomes, multiple thresholds), τ , such that yij = 1 when y∗ij is above τ , and yij = 0 otherwise. Assuming that aij, ci , and eij are normally distributed and standardizing the

12 In

quantitative genetics research, the model is known as the threshold model for liabilities. It was introduced by Wright (1934) and first applied to human genetics by Falconer (1965).

Genetics, Homeownership, and Home Location Choice

91

total residual variance σ 2 (= σa2 + σc2 + σe2 ) to one, the model corresponds to a probit model (or in case of categorical outcomes, a generalized probit model), such that: Prob (yij = 1|Xij) = Prob (y∗ij > τ |Xij) = 1 − [τ − βXij],

(5)

where  is the standard normal distribution function. We estimate models of censored (Tobit), binary, and ordered categorical outcomes using the robust mean- and variance-adjusted weighted least squares estimator proposed by Muthén (1984).13

Results Evidence from Correlations We first report separate correlation coefficients for identical and fraternal twin pairs for homeownership and home location choices. If genetically identical twins are significantly more correlated with respect to their homeownership and home location choices than are fraternal twins, then we may conclude that there is evidence that those choices are, at least in part, attributable to genetic factors. That is, the correlation evidence provides a first and intuitive indication of whether the studied choices have a genetic component. For fraternal twins, we report separate correlations for same-sex and oppositesex twins. We also report the correlations between twins and random age-, gender-, and municipality-matched non-twins from the overall population as a reference point. Figure 1 shows these correlations for different outcomes.14 We draw several conclusions from the figure. First, for each variable, we find that the correlation is significantly greater for identical twins than for fraternal twins. For example, for Tenure Choice, the correlation is 0.64 for identical twins, compared to only 0.40 for fraternal twins. Second, the correlations for same-sex fraternal twins are generally only slightly greater than those for opposite-sex twins, e.g., 0.41 versus 0.38 for Tenure Choice. As a result, we include a gender indicator as a control variable when we estimate our main models. Finally, the correlation between twins and age-, gender-, and municipality-matched, but otherwise randomly selected non-twins, is smaller than the correlations among identical or fraternal twin pairs.15 At the same time, the correlation between

13 All

models are estimated using the Mplus software. While maximum likelihood estimation in principle is an alternative, it is computationally more costly as it requires numerical integration. Reported standard errors are bootstrapped with 1,000 resamples. 14 For continuous variables, including censored outcomes, we report Pearson correlations; for the binary variable, we report tetrachoric correlations. 15 For Distance to Urban Center, the randomly selected non-twins are only matched by age and gender, not by municipality.

92

H. Cronqvist et al.

Correlations by Genetic Similarity Identical Twins

Fraternal Twins

Fraternal Twins - Same Sex

Fraternal Twins - Opposite Sex

Random Match

0.80 0.70 0.60 0.50 0.40 0.30 0.20 0.10 0.00 Tenure Choice

Housing Consumption

Distance to Birthplace

Distance to Urban Center

Fig. 1 Correlations by genetic similarity. Correlation coefficients for Tenure Choice, Housing Consumption, Distance to Birthplace, and Distance to Urban Center between twins for different types of twin pairs as well as for twins randomly matched with non-twins (controlling for age, gender, and in case of Tenure Choice, Housing Consumption, and Distance to Birthplace for municipality). For the binary variable Tenure Choice, we report tetrachoric correlations, for the remaining three we report standard Pearson correlations

two randomly matched individuals of the same age and gender who live in the same municipality is not necessarily zero, suggesting that accounting for age, gender, as well as municipality characteristics is important in order to identify the role of genetic and common environmental factors. This correlation evidence indicates that cross-sectional variation in homeownership and home location choices may be explained, in part, by a significant genetic component. However, because the correlation for identical twins is significantly smaller than one, there is also evidence that genetic variation does not completely explain these choices. Below, we use the previously specified models to decompose the variation into genetic and environmental components, controlling for individual characteristics and, in case of tenure choice, for the average house price in the municipality. Evidence from Homeownership Models Main Results Table 2 reports results for Tenure Choice, i.e., the choice to own one’s home rather than to rent it. In particular, we report marginal effects for each variable as well as estimates of the variance components A, C, and E based on the

Genetics, Homeownership, and Home Location Choice

93

Table 2 Tenure Choice models (1)

(2)

(3)

(4)

(5)

(6)

Marginal effects for columns (4)–(6) 0.006 (0.020) 0.059 (0.006) 0.027 (0.005) 0.000 (0.006) −0.072 (0.007) 0.008 (0.004) 0.014 (0.005) 0.081 (0.005) 0.016 (0.002) 0.122 (0.005) 0.230 (0.007) 0.289 (0.009) 0.013 (0.009) −0.051 (0.009) −0.014 (0.004) 0.017 (0.005) −0.013 (0.003) −0.017 (0.006) −0.007 (0.004) −0.045 (0.020)

Male Age 35–44 Age 45–54 Age 55–64 Age ≥ 65 Education level 2 Education level 3 Married Number of children in household Second net worth quartile indicator Third net worth quartile indicator Highest net worth quartile indicator Log of disposable income St. dev. of log growth rate of disposable household income Stock market participation Share in equities Urban Rural Poor health Log of mean house value in municipality

model in Eq. 5. A is the proportion of the total residual variance of an individual’s choice attributable to an additive genetic component: A=

σa2 σa2 + σc2 + σe2

(6)

The proportions attributable to the common environment (C) and individualspecific environmental effects (E) are computed analogously. Columns 1 through 3 report the model with no controls, while columns 4 through 6 report the model including individual characteristics as controls. As

94

H. Cronqvist et al.

Table 2 (continued) (1) Residual variance

1.000

(2)

1.000

0.505 (0.011) 0.495 (0.011)

0.498 (0.046) 0.146 (0.034) 0.356 (0.018)

A component C component

(3)

1.000

(4) 1.000

1.000 (0.000)

(5)

(6)

1.000

1.000

0.319 (0.019) 0.681 (0.019)

0.449 (0.051) 0.006 (0.032) 0.545 (0.029)

E component

1.000 (0.000)

N  χ2 p-value

26,876

26,876 1905.4 0.000

26,876 114.8 0.000

26,876

26,876 298.4 0.000

26,876 37.5 0.000

R2

0.000

0.000

0.000

0.547

0.547

0.547

Marginal effects (evaluated at the sample mean) and standard errors for probit models of Tenure Choice. We estimate the Tenure Choice model by weighted least squares. Standard errors are bootstrapped with 1,000 resamples. We report the total residual variance (normalized to 1.000) and the variance fractions due to a genetic ( A), common environmental (C), and individual-specific environmental (E) component. We use a  χ 2 test (Satorra 2000) to compare the fit of the CE model relative to the E model as well as of the ACE model relative to the CE model. We also report R2 for probit models (see Hagle and Mitchell 1992). All variables are defined in detail in Appendix Table 8

a benchmark for model fit, columns 1 and 4 report an “E model” in which both A and C are set to zero. Columns 2 and 5 report a “CE model,” in which A is constrained to zero. Finally, columns 3 and 6 report the full “ACE model.” To compare the fit across models, we report an adjusted χ 2 difference test together with the corresponding p-values.16 We find that the full ACE model is the preferred model specification. That is, a genetic factor significantly improves the fit of a standard model that attempts to explain cross-sectional variation in individuals’ tenure choice. When not controlling for individual socioeconomic differences, we estimate that the proportion of the cross-sectional variation attributable to a genetic component A is 50%, while the variation due to common (parental) influence is 15%, both estimates being statistically significant at the 1%-level. Since homeownership has been shown to depend not only on preferences, but also individual circumstances and characteristics, we cannot interpret the findings in column 3 as evidence that homeownership preferences are genetic. Instead, it is possible that variation in other determinants is partly genetic. We attempt to account for the determinants in columns 4 through 6. The R2 reveals that about 55% of the variation in the underlying latent homeownership variable can be explained by variation in socioeconomic characteristics.17 This suggests

we report χ 2 differences for the Satorra-Bentler scaled χ 2 (Satorra (2000)), i.e., we test for the difference in χ 2 for an E versus CE model, and a CE versus ACE model. 17 The R2 reported in Table 2 is calculated as the variance of y ˆ ij divided by yˆ ∗ + 1 (Hagle ˆ ∗ij = βX ij and Mitchell 1992). 16 Specifically,

Genetics, Homeownership, and Home Location Choice

95

that preferences account for up to 45% of the variation in homeownership. The genetic component of homeownership preferences is significant at 46%, while the C component drops to 1%, suggesting that homeownership preferences are not affected by parenting. Overall, we conclude that the genetic variation in preferences accounts for up to 21% (= 0.46 × 0.45) of the total cross-sectional variation in homeownership. It is possible that our results are affected by cross-sectional variation in mobility, another possible determinant of tenure choice, that may also be heritable. Using a sub-sample for which we have occupational data, we address this concern following Sinai and Souleles (2005). First, we include occupation indicator variables, since some occupations (e.g., military) are associated with more mobility than other occupations.18 Second, we include a variable, P(ST AY), defined as the proportion in the same age-occupation-marital status cell that did not move in the previous year. In untabulated results, we find that the relative amount of genetic variation, i.e. the A component, does not depend upon whether we include controls for mobility. In Table 3, we report results for Housing Consumption. Columns 1 through 3 report model results without control variables, while we include all controls in columns 4 through 6. To compare the fit across these linear models, we report likelihood ratio tests with the corresponding p-values. We find that the full ACE model is the preferred specification. We estimate that the genetic component is 23–26% depending on whether we include controls. The estimates are statistically significant at the 1%-level.19 The C component, on the other hand, is statistically insignificantly. Overall, we conclude that even if we condition the analysis on only homeowners, genetic factors explain the amount of housing individuals choose. We draw several conclusions from the evidence in Tables 2 and 3. First, our results are consistent with tenure choice and housing consumption being explained, in part, by genetic preferences. Second, we find that the effect of the common environment is close to zero. This evidence suggests that differences in parental influence, or “nurture,” do not significantly explain variation in homeownership preferences. That is, our evidence suggests that significant parent-child similarities with respect to homeownership preferences (e.g., Charles and Hurst 2003) are attributable to a genetic factor, but not parenting. Finally, we find that the idiosyncratic environmental component (E) explains the variation in homeownership across individuals to a significant extent. It is important to emphasize that while E is often the largest component of the estimated models, it captures a large set of potential individual-specific life

18 The

occupation data are based on the International Standard Classification of Occupations (ISCO-88) by the International Labour Organization (ILO). 19 We measure Housing Consumption only for homeowners. We therefore report results from using Heckman’s 1976 two-stage sample selection approach. In the first stage, we estimate a probit model for Tenure Choice. In addition to the controls used in the second stage, the probit specification also includes Log of Mean House Value in Municipality.

96

H. Cronqvist et al.

Table 3 Housing Consumption models (1)

Intercept

(2)

(3)

(6)

0.842

0.547

Age 35–44 Age 45–54 Age 55–64 Age ≥ 65 Education level 2 Education level 3 Married Number of children in household Second net worth quartile indicator Third net worth quartile indicator Highest net worth quartile indicator Log of disposable income St. dev. of log growth rate of disposable household income Stock market participation Share in equities Urban Rural Poor health Inverse mills ratio 0.842

0.842

0.206 (0.016) 0.794 (0.016)

0.310 (0.068) 0.007 (0.037) 0.683 (0.037)

A component C component E component

(5)

Coefficient estimates for columns (4)–(6) 1.825 (0.224) 0.038 (0.014) −0.320 (0.039) −0.197 (0.037) −0.225 (0.037) −0.158 (0.042) −0.023 (0.017) 0.064 (0.019) 0.198 (0.026) 0.029 (0.007) −0.485 (0.051) −0.550 (0.068) −0.386 (0.072) −0.100 (0.015) 0.557 (0.056) −0.018 (0.023) 0.052 (0.022) −0.143 (0.015) 0.188 (0.029) 0.029 (0.017) −1.693 (0.085)

Male

Residual variance

(4)

Coefficient estimates for columns (1)–(3) −0.323 (0.009)

1.000 (0.000)

1.000 (0.000)

0.547

0.547

0.148 (0.016) 0.852 (0.016)

0.234 (0.056) 0.000 (0.029) 0.766 (0.034)

Genetics, Homeownership, and Home Location Choice

97

Table 3 (continued) N LR p-value R2

(1)

(2)

(3)

(4)

(5)

(6)

13,718

13,718 296.3 0.000 0.000

13,718 39.3 0.000 0.000

13,718

13,718 149.6 0.000 0.350

13,718 23.9 0.000 0.350

0.000

0.350

Estimates and standard errors for regressions of Housing Consumption onto a constant only (columns 1–3) and socioeconomic characteristics (columns 4–6). We estimate the Housing Consumption model by maximum likelihood and report coefficient estimates, bootstrapped standard errors (with 1,000 resamples) as well as R2 . We report the total residual variance and the variance fractions due to a genetic ( A), common environmental (C), and individual-specific environmental (E) component. We compare the fit of the CE model against the E model as well as of the ACE model against the CE model using a likelihood ratio (LR) test. All variables are defined in detail in Appendix Table 8

experiences and non-genetic factors other than the included control variables, as well as measurement error. A caveat with respect to any study of genetic versus environmental influences is that an estimated genetic component is not a universal biological constant, but an estimate relative to the amount of genetic and environmental variation in the examined sample. The variance decomposition we perform and therefore our estimates of the relative importance of genetic variation are from a specific country, i.e., Sweden, at a specific point in time, i.e., 2002. While the homeownership rate in Sweden and that in, for example, the U.S. were similar in 2002 (66 versus 68%), it is still possible that the relative contribution of genetic and environmental variation differs between the two countries. Results by Age Group In Table 4, we study whether the A, C, and E components for homeownership preferences, i.e. after controlling for socioeconomic differences, vary systemTable 4 Decomposition by age group Age group

N

Age < 35

6,026

Age 35–54

10,730

Age ≥ 55

10,120

Variance components A C

E

0.241 (0.140) 0.383 (0.079) 0.204 (0.059)

0.415 (0.051) 0.617 (0.052) 0.796 (0.053)

0.343 (0.109) 0.000 (0.045) 0.000 (0.019)

Results from weighted least squares estimations of probit models of Tenure Choice for separate age groups. Tenure Choice is modeled as a nonlinear function of up to three unobserved effects representing an additive genetic effect ( A), a common environmental effect (C), as well as an individual-specific environmental effect (E). We include the same socioeconomic characteristics (excluding age indicators) as controls as in Table 2. For each estimated model, we report the fraction of the residual variance explained by each unobserved effect ( A - for the additive genetic effect, C - for common environmental effect, E - for the individual-specific environmental effect) as well as the corresponding bootstrapped standard errors (1,000 resamples). When the nonnegativity constraint for a variance parameter is binding, we report a zero

98

H. Cronqvist et al.

Table 5 Gene-environment interaction Parental net worth

N

Below median parental net worth

7,008

Above median parental net worth

7,026

Variance components A C

E

0.285 (0.141) 0.563 (0.087)

0.626 (0.068) 0.437 (0.051)

0.089 (0.094) 0.000 (0.058)

Results from weighted least squares estimations of probit models of Tenure Choice for separate groups based on parental net worth. Tenure Choice is modeled as a nonlinear function of up to three unobserved effects representing an additive genetic effect ( A), a common environmental effect (C), as well as an individual-specific environmental effect (E). We include the same socioeconomic characteristics as controls as in Table 2. For each estimated model, we report the fraction of the residual variance explained by each unobserved effect ( A - for the additive genetic effect, C - for common environmental effect, E - for the individual-specific environmental effect) as well as the corresponding bootstrapped standard errors (1,000 resamples)

atically with an individual’s age. Specifically, we estimate separate models for the youngest (Age < 35), the oldest (Age ≥ 55), and those in-between, i.e., 35–54 year-old individuals. The table shows two important results. First, the A component does not decay to zero for older twins, but in fact remains statistically significant. That is, genetic variation explains homeownership across individuals also among the oldest individuals in our sample, i.e., for those 55 and older, in spite of these individuals having had exposure to significant idiosyncratic factors during their life spans. Second, while we find no overall effect of parenting on homeownership, the C component is 34% for those younger than 35 years, i.e., parental influence appears important for the homeownership choice of the youngest individuals. These results are broadly consistent with research in behavioral genetics which has found a significant effect of the common family environment in early ages, but also shown that such effects decay to zero in adulthood (e.g., Bouchard 1998). Gene-environment Interactions In Table 5, we study gene-environment interactions (e.g., Rutter 2006; Cunha and Heckman 2010). While an individual’s genes may provide an innate predisposition to a specific behavior, environmental conditions may determine the extent to which the behavior is actually expressed. One hypothesis, based on the bioecological theory (e.g., Bronfenbrenner and Ceci 1994), is that more supportive environments, broadly speaking, may result in a stronger expression of a genetic predisposition. In our context, one approach to measuring “supportive environment” is the wealth of the individual’s parents.20

20 We are not able to measure parents’

wealth when the twins grew up so we use parents’ net worth in the first year that we have data on them.

Genetics, Homeownership, and Home Location Choice

99

We find that genetic variation explains 56% of the cross-sectional variation in tenure choice preferences among those who grew up in wealthy families (using median parental wealth as the cutoff), compared to only 29% among those who grew up in poor families. This evidence is consistent with the environment moderating genetic effects on homeownership, and in particular a more privileged upbringing resulting in a stronger expression of genetic predispositions with respect to homeownership. Those who grow up in a relatively wealthy family may have been exposed to a broader set of choices with respect to homeownership, and this exposure when growing up may affect subsequent genetic expressions with respect to homeownership. In contrast, idiosyncratic life experiences, that individuals were exposed to by chance, appear to explain a larger portion of the variation in homeownership among those with a less privileged upbringing. Evidence from Home Location Choice Models Panel A of Table 6 reports estimates of variance components A, C, and E for a familiar home location choice, i.e., the choice of a home location close to where the individual was born. Panel B reports evidence for an urban versus a rural home location. In particular, we use an ordered categorical model with outcomes Urban, Suburban, and Rural. We include a broad set

Table 6 Decomposition of home location choices N Panel A: Familiar home location Live where born Distance to Birthplace

26,760 26,072

Panel B: Urban or rural home location Urban 26,760 Distance to Urban Center

26,386

Population Density

26,760

Variance components A C

E

0.374 (0.052) 0.402 (0.077)

0.303 (0.039) 0.322 (0.050)

0.324 (0.020) 0.276 (0.030)

0.338 (0.042) 0.294 (0.087) 0.185 (0.053)

0.319 (0.031) 0.383 (0.058) 0.330 (0.038)

0.343 (0.017) 0.323 (0.032) 0.485 (0.022)

Results from home location choice models. For binary (Live Where Born) and ordered categorical (Urban) home location variables, we use weighted least squares estimation of (generalized) probit models. For continuous variables (Distance to Birthplace, Distance to Urban Center, and Population Density), we use weighted least squares estimation of Tobit models. For each estimated model, the home location choice variable is modeled as a function of up to three unobserved effects representing an additive genetic effect ( A), a common environmental effect (C), as well as an individual-specific environmental effect (E). We include various socioeconomic characteristics as controls (see Appendix Table 9 for details). For each estimated model, we report the fraction of the residual variance explained by each unobserved effect ( A - for the additive genetic effect, C - for common environmental effect, E - for the individual-specific environmental effect) as well as the corresponding bootstrapped standard errors (1,000 resamples)

100

H. Cronqvist et al.

of socioeconomic characteristics when we estimate these models. Appendix Table 9 reports the full models with all controls for Distance to Birthplace and Distance to Urban Center. We draw two conclusions from the evidence in the table. First, both the preference for a familiar home location and an urban versus rural location have significant genetic components. A is 37–40% for familiar home location, and 19–34% for urban versus rural location, the estimates being statistically significant at the 1%-level. Most of the coefficient estimates on the control variables have the expected signs, and several are statistically significant. This evidence indicates that genetic variation across individuals explains variation in home location choices to a significant extent. Second, we find that the common environment (C) explains a significant portion of the cross-sectional variation in behavior: 30–32% for familiar home location and 32–38% for urban versus rural location. This evidence is in contrast with the previously reported evidence for homeownership as well as investment and savings decisions (see, e.g., Cesarini et al. 2010; Cronqvist and Siegel 2011), and suggests that home location choices reflect commonality between twins that is unrelated to their genetic similarity. In conclusion, our evidence suggests that significant parent-child similarities with respect to home location choices (e.g., Whitfield et al. 2005; Helderman and Mulder 2007; Blaauboer 2011) are attributable to both a genetic factor and a common environment, such as parental influence. Robustness Equal Environments One concern with our empirical approach is the equal environments assumption, i.e., the assumption that the common environment is not more important for identical twins than for fraternal twins. If parents treat identical twins more similarly with respect to the behavior being examined than they treat fraternal twins, then our models would attribute such environmental influence to the genetic component, biasing our estimates of A. Existing evidence from studies of behavioral genetics suggests that findings of significant genetic effects on behaviors and personality traits are robust. More specifically, research which has used twins reared apart, i.e., twins separated at birth or early in life, for which there is no common parental environment by definition, report similar conclusions as studies with twins reared together (e.g., Bouchard et al. 1990; Bouchard 1998). This should reduce concerns that genetic and environmental effects are confounded in our analysis.21

21 See,

e.g., Goldberger (1979), Taubman (1981) and Bouchard and McGue (2003) for further discussion of some of the common concerns with respect to analysis of data on twins.

Genetics, Homeownership, and Home Location Choice

101

Measurement Error Measurement error in yij is captured by eij in the models in Eq. 2 through Eq. 4. As a result, the A component may be underestimated if there is significant measurement error in data. Because our data set comes from Statistics Sweden and the Swedish Tax Agency, which receive the data directly from financial institutions, we consider measurement error to be relatively rare, at least compared to data sets based on questionnaires or surveys. Despite the quality of our data, we recognize some potential sources of measurement error. For example, some individuals, i.e., recent movers, may not be in equilibrium with respect to homeownership in a specific year (e.g., Edin and Englund 1991). We therefore re-estimate the ACE model for homeownership defining an individual as a homeowner if the individual owned a house in any year from 1999–2002. We still find a highly significant A component of 47% and an insignificant C component for Tenure Choice. Random Mating Researchers have examined non-random mating based on, e.g., education and the extent to which individuals marry to diversify their labor income risk versus marry for other reasons (e.g., Pencavel 1998; Hess 2004). We are not aware of any studies on mating based on homeownership or home location choices, but positive assortative mating would not be surprising (e.g., individuals with preferences for an urban location marry individuals with similar preferences, and vice versa). Such non-random mating between the twins’ parents would make fraternal twins more similar relative to identical twins and would bias the estimate of the genetic component downwards (e.g., Neale and Maes 2004). Does the Evidence Change Interpretations of Existing Research? The choice to own one’s home rather than to rent it has been studied extensively in previous research. In what ways does our evidence change interpretations of such existing research? To understand the implications of controlling for genetic effects, we can consider a pair i of identical twins ( j = 1, 2) who were reared by the same parents.22 Let us assume that yij is linear in observable fixed effects and unobservable genetic (ai ) and common environmental (ci ) effects (e.g., parenting), such that: yij = β0 + βXij + ai + ci + eij

(7)

We can eliminate genetic or common environmental effects by considering the difference between the twins: yi1 − yi2 = β(Xi1 − Xi2 ) + ei1 − ei2 .

22 See

Taubman (1976) for an application of this empirical methodology.

(8)

102

H. Cronqvist et al.

Table 7 Controlling for genes and the common environment Tenure Choice Level Intercept Male Age group 35–44 Age group 45–54 Age group 55–64 Age >= 65 Education level 2 Education level 3 Married Number of children in household Second net worth quartile indicator Third net worth quartile indicator Highest net worth quartile indicator Log of disposable income St. dev. of log growth rate of disposable household income Stock market participation Share in equities Urban Rural Poor health Log of mean house value in municipality N R2

Twin-twin difference −0.001 (0.007)

0.405 (0.141) 0.015 (0.008) 0.136 (0.016) 0.073 (0.015) 0.030 (0.016) −0.076 (0.021) 0.033 (0.011) 0.044 (0.013) 0.138 (0.012) 0.010 (0.004) 0.286 (0.013) 0.525 (0.014) 0.579 (0.015) 0.077 (0.007) −0.115 (0.034) −0.050 (0.015) 0.052 (0.013) −0.022 (0.009) −0.022 (0.019) −0.006 (0.012) −0.084 (0.009)

0.028 (0.023) 0.074 (0.028) 0.158 (0.019) 0.038 (0.007) 0.334 (0.019) 0.536 (0.022) 0.604 (0.025) −0.021 (0.014) −0.011 (0.047) −0.026 (0.021) 0.048 (0.019) −0.012 (0.017) −0.036 (0.039) 0.014 (0.018) −0.084 (0.019)

7,984 0.455

3,992 0.263

Estimates and heteroskedasticity-adjusted standard errors for a linear regression of Tenure Choice onto an individual’s socioeconomic characteristics. Only identical twins are included. The first column reports results for the model in levels, while the second column reports results for the model in differences between identical twins within each pair. See Appendix Table 8 for a detailed description of the variables

Genetics, Homeownership, and Home Location Choice

103

The “Level” column of Table 7 reports coefficient estimates for homeownership based on Eq. 7, and the “Twin-Twin Difference” column reports coefficient estimates for the differenced model in Eq. 8. Comparing the estimates in the columns, we find that, for example, the effect of income drops from 0.077 to being statistically insignificant. That is, controlling for a genetic effect, we conclude that the effect of income on homeownership is significantly lower than we would previously have concluded. Similar conclusions apply to several other tenure choice determinants, including income volatility and stock market participation. At the same time, the importance of net worth as a determinant for tenure choice remains unaffected. In conclusion, some, but not all key determinants of the choice to own versus rent become insignificant once we control for genetic effect, i.e., there is no independent exogenous effect of those variables once the genetic effect on those variables is accounted for.

Conclusion The objective of this study is to analyze the extent to which variation across individuals in homeownership and home location choices is explained by: (i) a genetic factor, i.e., individuals are born with a predisposition to a particular behavior, (ii) parental influence, or “nurture”, and (iii) individual-specific environmental factors that an individual is subject to during his or her life. Evidence related to these questions contributes to a deeper understanding of the factors that explain individual behavior with respect to the housing market, and adds to an expanding literature on the biological and genetic factors that influence an individual’s economic and financial decisions. Several conclusions may be drawn based on our evidence. First, a genetic factor explains a significant proportion of the variation in homeownership and home location choices. Controlling for an extensive set of standard determinants (e.g., wealth, education, risk aversion) we also find that a significant proportion of the residual variation of these choices is explained by a genetic factor. An interpretation of this evidence is that individual housing preferences are partly genetic, which seems consistent with models that assume a preference for homeownership (e.g., Cauley et al. 2007). Genetic predispositions are found to be moderated by the environment. For example, growing up in a wealthier family results in stronger expression of genetic predispositions, while idiosyncratic life experiences appear to explain a larger portion of the variation in homeownership among those who grew up in a less wealthy family environment. It is important to emphasize that our evidence does not imply that there exists a “homeownership gene” or a “home location gene.” Complex behaviors, such as those we have studied, are generally a function of several genes. Economists, in collaboration with genetics researchers, have recently performed studies, which have identified some specific genes related to, e.g., financial risk taking (e.g., Kuhnen and Chiao 2009), and others have attempted to identify

104

H. Cronqvist et al.

the biochemical mechanisms by which genes affect behavior (e.g., Sapienza et al. 2009). While it is beyond the scope of this study to identify specific genes or biochemical mechanisms affecting individuals’ housing choices, subsequent studies may attempt to make progress in such a direction. A second conclusion based on our evidence is that the family environment, i.e., parenting, does not explain variation in homeownership choices, other than among the very youngest individuals we study. That is, while the genetic effect on homeownership remains throughout an individual’s life, the effect of parenting decays to zero already for those in their 30’s. While the family environment does not affect the homeownership choice, it does explain a significant proportion of the variation in the choice where to live. Finally, a very large proportion of the variation in preferences over homeownership and home location choices is explained by individual-specific and idiosyncratic life-experiences. This evidence raises the question of what those factors are. For example, which non-genetic, individual-specific life experiences increase the probability of being a homeowner? Acknowledgements We are thankful for comments from seminar participants at the ARES 2012 Annual Meeting, Claremont Graduate University (School of Politics & Economics), Claremont McKenna College, and University of Washington, and Steve Cauley, Jack Goldberg, Greg Hess, David Hirshleifer, Steven Laposa, Peter Linneman, Clara Mulder, Andrey Pavlov, Dustin Read, Ed Rice, Tony Sanders, Nancy Segal, and Stephen Shore. We acknowledge research funding from the Financial Economics Institute and the Lowe Institute of Political Economy at Claremont McKenna College, and the Global Business Center and the CFO Forum at the University of Washington. This project was pursued in part when Cronqvist was Olof Stenhammar Visiting Professor at the Institute for Financial Research (SIFR), which he thanks for its support, and while Siegel was visiting Arizona State University, which he thanks for their hospitality. Statistics Sweden and the Swedish Twin Registry (STR) provided the data for this study. STR is supported by grants from the Swedish Research Council, the Ministry of Higher Education, AstraZeneca, and the National Institute of Health (grants AG08724, DK066134, and CA085739). Any errors or omissions are our own.

Appendix Table 8 Variable definitions Variable Types of twins Identical twins

Description

Twins that are genetically identical, also called monozygotic twins. Zygosity is determined by the Swedish Twin Registry based on questions about intrapair similarities in childhood. Non-identical Twins that share on average 50% of their genes, also called dizygotic twins or fraternal twins. Non-identical twins can be of the same sex or of opposite sex. Zygosity is determined by the Swedish Twin Registry based on questions about intrapair similarities in childhood. Homeownership and home location Tenure Choice An indicator that is one if the individual’s or, if married, her spouse’s tax data show a positive value in the category residence, condominium, or home on a farm at the end of 2002 and zero otherwise.

Genetics, Homeownership, and Home Location Choice

105

Table 8 (continued) Variable Housing Consumption

Description

Housing Consumption is defined as the natural logarithm of the house value divided by the mean house value in the same municipality. The house value is the maximum value reported in the categories residence, condominium, and home on a farm at the end of 2002 for an individual. For married individuals we add the corresponding values across both spouses. House values are based on actual market transactions. The data are obtained from Statistics Sweden. Mean house We calculate the mean house value for each municipality for 2002 value using tax values reported by Statistics Sweden for a large random in municipality sample of Swedish residents. Tax values are based on actual market transactions. Live Where Born An indicator that is one if the individual lives in the state in which she was born and zero otherwise. Distance to The driving distance in kilometers to the state of birth. We define Birthplace this distance to be the average distance to the center of all municipalities within the state of birth weighted by their population. If the twin lives in her state of birth we set the distance to zero. The distance is obtained from Google Maps. The population numbers are obtained from Statistics Sweden. Urban An indicator that is one if the individual’s current municipality is classified as “Large City” or “City” and zero otherwise. The data are obtained from Statistics Sweden. Suburban An indicator that is one if the individual’s current municipality is not classified as either “Large City”, “City”, “Rural” or “less than 12,500 inhabitants” and zero otherwise. The data are obtained from Statistics Sweden. Rural An indicator that is one if the individual’s current municipality is classified as “Rural” or “less than 12,500 inhabitants” and zero otherwise. The data are obtained from Statistics Sweden. Distance to The driving distance in kilometers to the closest municipality Urban Center defined as “urban.” Distances are obtained from Google Maps. Population Density Number of individuals per square kilometer. The data are available for all twins at the municipality level (there are 290 municipalities in our data set). The data are obtained from Statistics Sweden. Socioeconomic characteristics Male An indicator variable that equals one if an individual is male and zero otherwise. Gender is obtained from Statistics Sweden. Age An individual’s age on 31 December 2002 as reported by Statistics Sweden. Education level 1 An indicator variable that equals one if an individual received 9 years of education or less and zero otherwise. Education level 2 An indicator variable that equals one if an individual has at least a highschool degree and at most 2 years of university education and zero otherwise. Education level 3 An indicator variable that equals one if an individual received at least 4 years of university education and zero otherwise. Married An indicator variable that equals one if an individual is married in year 2002 and zero otherwise. It is obtained from the Statistics Sweden. Number of children The number of children living in the household in year 2002. in household The number is obtained from Statistics Sweden. Poor health An indicator variable that is one if an individual receives payments due to illness, injury, or disability in year 2002 and zero otherwise. The data are obtained from Statistics Sweden.

106

H. Cronqvist et al.

Table 8 (continued) Variable Disposable household income

St. dev. of log growth rate of disposable household income Net worth and net worth quartile indicators

Stock market participation Share in equities

Description The average disposable income of the individual and, if married, her spouse in 2002, as defined by Statistics Sweden, that is the sum of income from labor, business, and investment, plus received transfers, less taxes and alimony payments. When reported in United States Dollars (USD), Swedish Krona (SEK) amounts have been converted at SEK/USD 8.6815, the end of year exchange rate for 2002. The data are obtained from Statistics Sweden. The time-series standard deviation of the log growth rate of disposable household income between 2000 and 2007. The variable is missing if four or more of the log growth rates are missing. The log growth rate distribution is trimmed at the top and bottom one percentile of the log growth rate distribution. The average difference between the market value of an individual’s and, if married, her spouse’s assets and liabilities, calculated by Statistics Sweden at the end of 2002. When reported in United States Dollars (USD), Swedish Krona (SEK) amounts have been converted at SEK/USD 8.6815, the end of year exchange rate for 2002. The data are obtained from Statistics Sweden. In our empirical models, we use four indicators indicating the quartile of the net worth a household belongs to. An indicator that is one if an individual or her spouse holds equity in 2002 and zero otherwise. The data are obtained from Statistics Sweden. The average market value of equity of an individual and, if married, her spouse in 2002 scaled by the average market value of all financial assets of the individual (including spouse, if married) in 2002. The data are obtained from Statistics Sweden.

Table 9 Distance to Birthplace and Distance to Urban Center

Intercept Male Age group 35–44 Age group 45–54 Age group 55–64 Age ≥ 65 Education level 2 Education level 3 Married Number of children in household

Distance to Birthplace

Distance to Birthplace

Distance to Urban Center

Distance to Urban Center

−2.559 (0.064)

−9.861 (1.396) −0.505 (0.107) 0.530 (0.175) 0.709 (0.161) 0.989 (0.172) 1.042 (0.231) 0.852 (0.107) 2.703 (0.113) −0.044 (0.097) −0.305 (0.037)

0.098 (0.005)

0.589 (0.136) 0.020 (0.010) 0.090 (0.020) 0.175 (0.019) 0.207 (0.020) 0.203 (0.027) −0.049 (0.010) −0.181 (0.013) 0.074 (0.011) 0.040 (0.004)

Genetics, Homeownership, and Home Location Choice

107

Table 9 (continued) Distance to Birthplace

Distance to Birthplace

Distance to Urban Center

−0.531 (0.103) −0.807 (0.114) −0.733 (0.118) 0.464 (0.077) 1.369 (0.286) −0.030 (0.123) 0.220 (0.114) 0.230 (0.081) −0.872 (0.194)

Second wealth quartile indicator Third wealth quartile indicator Highest wealth quartile indicator Log of disposable income St. dev. of log growth rate of disposable household income Stock market participation Share in equities Urban Rural

Distance to Urban Center 0.037 (0.011) 0.024 (0.012) −0.022 (0.010) −0.046 (0.007) −0.102 (0.030) 0.007 (0.013) 0.009 (0.013)

Live where born

0.086 (0.011)

Residual variance

22.460

19.994

0.363

0.337

A component

0.384 (0.076) 0.357 (0.049) 0.259 (0.029)

0.402 (0.077) 0.322 (0.050) 0.276 (0.030)

0.340 (0.089) 0.377 (0.058) 0.284 (0.034)

0.294 (0.087) 0.383 (0.058) 0.323 (0.032)

C component E component N R2

26,072 0.000

26,072 0.110

26,386 0.000

26,386 0.072

Estimates and standard errors for regressions of Distance to Birthplace and Distance to Urban Center onto (i) a constant only and (ii) socioeconomic characteristics. All dependent variables are expressed in 100’s of kilometers. We estimate Tobit models by weighted least squared and report coefficient estimates, bootstrapped standard errors with 1,000 resamples as well as R2 . We report the total residual variance and the variance fractions due to a genetic (A), common environmental (C), and individual-specific environmental (E) component. All variables are defined in detail in Appendix Table 8

References Alonso, W. (1964). Location and land use: Toward a general theory of land rent. Cambridge: Harvard University Press. Ashenfelter, O., & Krueger, A. B. (1994). Estimates of the economic returns to schooling from a new sample of twins. American Economic Review, 84(5), 1157–1173. Ashenfelter, O., & Rouse, C. (1998). Income, schooling, and ability: Evidence from a new sample of identical twins. Quarterly Journal of Economics, 113(1), 253–284. Barakova, I., Bostic, R., Calem, P., & Wachter, S. (2003). Does credit quality matter for homeownership? Journal of Housing Economics, 12(4), 318–336. Barnea, A., Cronqvist, H., & Siegel, S. (2010). Nature or nurture: What determines investor behavior? Journal of Financial Economics, 98, 583–604. Becker, G. S. (1976). Altruism, egoism, and genetic fitness: Economics and sociobiology. Journal of Economic Literature, 14, 817–826.

108

H. Cronqvist et al.

Behrman, J., & Taubman, P. (1989). Is schooling “mostly in the genes”? Nature-nurture decomposition using data on relatives. Journal of Political Economy, 97, 1425–1446. Behrman, J., Rosenzweig, M., & Taubman, P. (1994). Endowments and the allocation of schooling in the family and in the marriage market: The twins experiment. Journal of Political Economy, 102(6), 1131–1174. Behrman, J. R., & Taubman, P. (1976). Intergenerational transmission of income and wealth. American Economic Review, 66, 436–440. Benjamin, D., Chabris, C., Glaeser, E., Gudnason, V., Harris, T., Laibson, D., et al. (2008). Genoeconomics (Chapter 15, pp 304–335). The National Academies Press. Bernheim, B. (2009). On the potential of neuroeconomics: A critical (but hopeful) appraisal. American Economic Journal: Microeconomics, 1(2), 1–41. Bisin, A., & Verdier, T. (2000). Beyond the melting pot: Cultural transmission, marriage, and the evolution of ethnic and religious traits. Quarterly Journal of Economics, 115(3), 955–988. Bisin, A., Topa, G., & Verdier, T. (2004). Religious intermarriage and socialization in the United States. Journal of Political Economy, 112(3), 615–664. Blaauboer, M. (2011). The impact of childhood experiences and family members outside the household on residential environment choices. Urban Studies, 48, 1635–1650. Bortolus, R., Parazzini, F., Chatenoud, L., Benzi, G., Bianchi, M., & Marini, A. (1999). The epidemiology of multiple births. Human Reproduction Update, 5(2), 179–187. Bouchard, T., & McGue, M. (2003). Genetic and environmental influences on human psychological differences. Journal of Neurobiology, 54(1), 4–45. Bouchard, T. J. (1998). Genetic and environmental influences on adult intelligence and special mental abilities. Human Biology, 70, 257–279. Bouchard, T. J., Lykken, D. T., McGue, M., Segal, N. L., & Tellegen, A. (1990). Sources of human psychological differences: The Minnesota study of twins reared apart. Science, 250, 223–228. Bronfenbrenner, U., & Ceci, S. J. (1994). Nature-nurture in developmental perspective: A bioecological theory. Psychological Review, 101, 568–586. Brueckner, J. (1986). The downpayment constraint and housing tenure choice: A simplified exposition. Regional Science and Urban Economics, 16(4), 519–525. Camerer, C., Loewenstein, G., & Prelec, D. (2005). Neouroeconomics: How neuroscience can inform economics. Journal of Economic Literature, 43, 9–64. Cauley, S., Pavlov, A., & Schwartz, E. (2007). Homeownership as a constraint on asset allocation. Journal of Real Estate Finance and Economics, 34, 283–311. Cesarini, D., Dawes, C. T., Johannesson, M., Lichtenstein, P., & Wallace, B. (2009). Genetic variation in preferences for giving and risk taking. Quarterly Journal of Economics, 124, 809– 842. Cesarini, D., Johannesson, M., Lichtenstein, P., Sandewall, O., & Wallace, B. (2010). Genetic variation in financial decision making. Journal of Finance, 65, 1725–1754. Cesarini, D., Johannesson, M., Magnusson, P. K. E., & Wallace, B. (2012). The behavioral genetics of behavioral anomalies. Management Science, 58(1), 21–34. Charles, K. K., & Hurst, E. (2003). The correlation of wealth across generations. Journal of Political Economy, 111(6), 1155–1182. Chew, S. H., Ebstein, R. P., & Zhong, S. (2012). Ambiguity aversion and familiarity bias: Evidence from behavioral and gene association studies. Journal of Risk and Uncertainty, 44, 1–18. Cronqvist, H., & Siegel, S. (2011). The origins of savings behavior. Working paper, University of Washington, Michael G. Foster School of Business. Cunha, F., & Heckman, J. J. (2010). Investing in our young people. Working Paper 16201, National Bureau of Economic Research. Dahl, M., & Sorenson, O. (2010). The social attachment to place. Social Forces, 89(2), 633– 658. Davidoff, T. (2006). Labor income, housing prices, and homeownership. Journal of Urban Economics, 59(2), 209–235. Davies, P., Greenwood, M., & Li, H. (2001). A conditional logit approach to U.S. state-to-state migration. Journal of Regional Science, 41(2), 337–360. Dreber, A., Apicella, C. L., Eisenberg, D. T. A., Garcia, J. R., & Zamore, R. S. (2009). The 7R polymorphism in the dopamine receptor D4 gene (DRD4) is associated with financial risktaking in men. Evolution and Human Behavior, 30, 85–92.

Genetics, Homeownership, and Home Location Choice

109

Edin, P., & Englund, P. (1991). Moving costs and housing demand: Are recent movers really in equilibrium? Journal of Public Economics, 44(3), 299–320. Falconer, D. S. (1965). The inheritance of liability to certain diseases, estimated from the incidence among relatives. Annals of Human Genetics, 29, 51–76. Feijten, P., Hooimeijer, P., & Mulder, C. (2008). Residential experience and residential environment choice over the life-course. Urban Studies, 45(1), 141–162. Fernandez, R., Fogli, A., & Olivetti, C. (2004). Mothers and sons: Preference formation and female labor force dynamics. Quarterly Journal of Economics, 119(4), 1249–1299. Freedman, O., & Kern, C. (1997). A model of workplace and residence choice in two-worker households. Regional Science and Urban Economics, 27(3), 241–260. Goldberger, A. (1979). Heritability. Economica, 46(184), 327–347. Goodman, A. C. (1988). An econometric model of housing price, permanent income, tenure choice, and housing demand. Journal of Urban Economics, 23(3), 327–353. Gyourko, J., & Linneman, P. (1996). Analysis of the changing influences on traditional households’ ownership patterns. Journal of Urban Economics, 39(3), 318–341. Gyourko, J., & Linneman, P. (1997). The changing influences of education, income, family structure, and race on homeownership by age over time. Journal of Housing Research, 8, 1–26. Hagle, T. M., & Mitchell, G. E. (1992). Goodness-of-fit measures for probit and logit. American Journal of Political Science, 36(3), 762–784. Harris, J. R. (1995). Where is the child’s environment? A group socialization theory of development. Psychological Review, 102(3), 458–489. Haurin, D. (1991). Income variability, homeownership, and housing demand. Journal of Housing Economics, 1(1), 60–74. Haurin, D., & Gill, H. (1987). Effects of income variability on the demand for owner-occupied housing. Journal of Urban Economics, 22(2), 136–150. Haurin, D., Hendershott, P., & Wachter, S. (1997). Borrowing constraints and the tenure choice of American youth. Journal of Housing Research, 8, 137–154. Heckman, J. J. (1976). The common structure of statistical models of truncation, sample selection and limited dependent variables and a simple estimator for such models. Annals of Economic and Social Measurement, 5, 475–492. Helderman, A., & Mulder, C. (2007). Intergenerational transmission of homeownership: The roles of gifts and continuities in housing market characteristics. Urban Studies, 44(2), 231. Hendershott, P., & Shilling, J. (1982). The economics of tenure choice, 1955–1979. Research in Real Estate, 1, 105–133. Henderson, J., & Ioannides, Y. (1983). A model of housing tenure choice. American Economic Review, 73(1), 98–113. Henderson, J. V., & Ioannides, Y. M. (1989). Dynamic aspects of consumer decisions in housing markets. Journal of Urban Economics, 26(2), 212–230. Hess, G. (2004). Marriage and consumption insurance: What’s love got to do with it? Journal of Political Economy, 112(2), 290–318. Hirshleifer, J. (1977). Economics from a biological viewpoint. Journal of Law and Economics, 20, 1–52. Hsu, M., Bhatt, M., Adolphs, R., Tranel, D., & Camerer, C. (2005). Neural systems responding to degrees of uncertainty in human decision-making. Science, 310(5754), 1680–1683. Hu, X. (2005). Portfolio choices for homeowners. Journal of Urban Economics, 58(1), 114–136. Jocklin, V., McGue, M., & Lykken, D. T. (1996). Personality and divorce: A genetic analysis. Journal of Personality and Social Psychology, 71, 288–299. Jones, L. (1989). Current wealth and tenure choice. Real Estate Economics, 17(1), 17–40. Jones, L. (1990). Current wealth constraints on the housing demand of young owners. Review of Economics and Statistics, 72(3), 424–432. Kuhnen, C. M., & Chiao, J. (2009). Genetic determinants of financial risk taking. PLoS ONE, 4(2):e4362. doi:10.1371/journal.pone.0004362. Lichtenstein, P., Sullivan, P. F., Cnattingius, S., Gatz, M., Johansson, S., Carlström, E., et al. (2006). The Swedish twin registry in the 3rd millennium: An update. Twin Research and Human Genetics, 9, 875–882. Linneman, P. (1985). An economic analysis of the homeownership decision. Journal of Urban Economics, 2, 230–246.

110

H. Cronqvist et al.

Linneman, P., & Wachter, S. (1989). The impacts of borrowing constraints on homeownership. Real Estate Economics, 17(4), 389–402. McFadden, D. (1977). Modelling the choice of residential location. Cowles Foundation for Research in Economics at Yale University. McGue, M., & Lykken, D. T. (1992). Genetic influence on risk of divorce. Psychological Science, 3, 368–373. Mills, E. (1990). Housing tenure choice. Journal of Real Estate Finance and Economics, 3(4), 323– 331. Mincer, J. (1978). Family migration decisions. Journal of Political Economy, 86(5), 749–773. Mulder, C., & Smits, J. (1999). First-time home-ownership of couples: The effect of intergenerational transmission. European Sociological Review, 15(3), 323–337. Muthén, B. (1984). A general structural equation model with dichotomous, ordered categorical, and continuous latent variable indicators. Psychometrika, 49, 115–132. Neale, M. C., & Maes, H. H. M. (2004). Methodology for genetic studies of twins and families. NATO ASI Series D: Behavioural and Social Science. Dordrecht, The Netherlands: Kluwer Academic Publishers B.V. Nicolaou, N., & Shane, S. (2010). Entrepreneurship and occupational choice: Genetic and environmental influences. Journal of Economic Behavior & Organization, 76(1), 3–14. Nicolaou, N., Shane, S., Cherkas, L., Hunkin, J., & Spector, T. D. (2008). Is the tendency to engage in entrepreneurship genetic? Management Science, 54, 167–179. Ortalo-Magné, F., & Rady, S. (2002). Tenure choice and the riskiness of non-housing consumption. Journal of Housing Economics, 11(3), 266–279. Pencavel, J. (1998). Assortative mating by schooling and the work behavior of wives and husbands. American Economic Review, 88, 326–329. Robst, J., Deitz, R., & McGoldrick, K. (1999). Income variability, uncertainty and housing tenure choice. Regional Science and Urban Economics, 29(2), 219–229. Rutter, M. (2006). Genes and behavior: Nature/nurture interplay explained. Oxford, UK: Blackwell Publishers. Sacerdote, B. (2002). The nature and nurture of economic outcomes. American Economic Review, 92, 344–348. Sacerdote, B. (2007). How large are the effects from changes in family environment? A study of Korean American adoptees. Quarterly Journal of Economics, 122, 119–157. Sapienza, P., Zingales, L., & Maestripieri, D. (2009). Gender differences in financial risk aversion and career choices are affected by testosterone. Proceedings of the National Academy of Sciences, 106(36), 15,268–15,273. Satorra, A. (2000). Scaled and adjusted restricted tests in multi-sample analysis of moment structures. In R. Heijmans, D. Pollock, & A. Satorra, (Eds.), Innovations in multivariate statistical analysis (pp. 233–247). Kluwer Academic Publishers. Shore, S. H. (2011). The intergenerational transmission of income volatility: Is riskiness inherited? Journal of Business and Economic Statistics, 29(3), 372–381. Sinai, T., & Souleles, N. (2005). Owner-occupied housing as a hedge against rent risk. Quarterly Journal of Economics, 120(2), 763–789. Struyk, R., & Marshall, S. (1973). The determinants of household home ownership. Urban Studies, 11(3), 289–99. Struyk, R., & Marshall, S. (1975). Income and urban home ownership. Review of Economics and Statistics, 57(1), 19–26. Taubman, P. (1976). The determinants of earnings: Genetics, family, and other environments: A study of white male twins. American Economic Review, 66(5), 858–870. Taubman, P. (1981). On heritability. Economica, 48(192), 417–420. Topel, R. (1986). Local labor markets. Journal of Political Economy, 94(3), 111–143. Van Dam, F., Heins, S., & Elbersen, B. (2002). Lay discourses of the rural and stated and revealed preferences for rural living: Some evidence of the existence of a rural idyll in the Netherlands. Journal of Rural Studies, 18(4), 461–476. Whitfield, J., Zhu, G., Heath, A., & Martin, N. (2005). Choice of residential location: Chance, family influences, or genes? Twin Research and Human Genetics, 8(1), 22–26. Wright, S. (1934). An analysis of variability in number of digits in an inbred strain of guinea pigs. Genetics, 19, 506–536.

Genetics, Homeownership, and Home Location Choice

111

Zhang, Z., Zyphur, M. J., Narayanan, J., Arvey, R. D., Chaturvedi, S., Avolio, B. J., et al. (2009). The genetic basis of entrepreneurship: Effects of gender and personality. Organizational Behavior and Human Decision Processes, 110, 93–107. Zhong, S., Israel, S., Xue, H., Ebstein, R. P., & Chew, S. H. (2009). Monoamine Oxidase A gene (MAOA) associated with attitude towards longshot risks. PLoS ONE, 4(12), 1–4. Zorn, P. (1989). Mobility-tenure decisions and financial credit: Do mortgage qualification requirements constrain homeownership? Real Estate Economics, 17(1), 1–16. Zyphur, M. J., Narayanan, J., Arvey, R. D., & Alexander, G. J. (2009). The genetics of economic risk preferences. Journal of Behavioral Decision Making, 22, 367–377.