You are on page 1of 15

1186072

research-article2023
SRDXXX10.1177/23780231231186072Socius: Sociological Research for a Dynamic WorldRauf

Original Article Socius: Sociological Research for


a Dynamic World
Volume 9: 1­–15
Mental Health Effects of Income over the © The Author(s) 2023
Article reuse guidelines:

Adult Life Course sagepub.com/journals-permissions


DOI: 10.1177/23780231231186072
https://doi.org/10.1177/23780231231186072
srd.sagepub.com

Tamkinat Rauf1

Abstract
While a gradient between income and depression is well documented, associational estimates are subject to bias
due to measurement errors, reverse causation, and other sources of confounding. Causal studies are few and
report small and divergent estimates, even in their direction. But prior research has important limitations for
causal inference, such as relying on comparisons between individuals sensitive to each other’s relative income or
capturing only a subset of pathways through which income affects well-being. This study leverages longitudinal and
genomic data to assess how much some known biases affect the income-depression gradient and to what extent
the gradient might be reflective of a causal effect. Findings from three U.S. samples—representing early midlife,
late midlife, and older adulthood—are suggestive of a beneficial effect of income on mental health, especially in
late midlife and beyond. The results have implications for interpreting the nature of socioeconomic disparities in
mental health.

Keywords
depression, mental health, income effect, income-depression gradient, income-health gradient, life course

An inverse association between income and various indica- 2020; Raschke 2019).1 Other types of random shocks also
tors of mental health—including clinical depression, depres- seem to have mixed effects: Recent meta-analyses of ran-
sive symptoms, and other indices measuring psychological domized experiments and government welfare programs
distress and negative affective well-being—is widely docu- report wide-ranging findings, with the average income
mented (Eaton et al. 2001; Heflin and Iceland 2009; Kessler effect being quite small (Ridley et al. 2020; Thomson et al.
1982; Martikainen et al. 2003; Ross and Huber 1985; 2022). One recent experiment entailing cash transfers finds
Stewart-Brown et al. 2015). However, the extent to which no meaningful effects (Courtin et al. 2018), while another
these estimates reflect a causal relationship is unclear. finds that such transfers were detrimental to psychological
Estimating effects by controlling only for observed con- well-being (Jaroszewicz et al. 2022). On the other hand,
founders (or no controls at all) in cross-sectional data is vul-
nerable to reverse causation and other sources of bias that
would overestimate the true casual effect. On the other hand, 1
Effect sizes are not reported in comparable units across all stud-
the presence of certain other confounders can result in under- ies. This article reports estimates in units of standard deviation/unit
estimation, and estimates based on current income may be change in log income. In comparable units, Lindqvist, Östling, and
further attenuated due to measurement error and temporary Cesarini (2020) found an effect of approximately .18 (p > .05) on
income fluctuations. the General Health Questionnaire-12 index (higher values indicate
While innovative causal designs have been employed to improved well-being).
address these issues, findings are divergent. In high-
income countries, the mental well-being effects of winning 1
Stanford University, Stanford, CA, USA
the lottery—a dramatic, exogenous shock to material cir-
Corresponding Author:
cumstances—range from positive to negative (Apouey and Tamkinat Rauf, Department of Sociology, University of Wisconsin-
Clark 2015; Brickman, Coates, and Janoff-Bulman 1978; Madison, 1180 Observatory Drive, Madison, WI 53706, USA.
Gardner and Oswald 2007; Lindqvist, Östling, and Cesarini Email: trauf@wisc.edu

Creative Commons Non Commercial CC BY-NC: This article is distributed under the terms of the Creative Commons Attribution-
NonCommercial 4.0 License (https://creativecommons.org/licenses/by-nc/4.0/) which permits non-commercial use, reproduction and
distribution of the work without further permission provided the original work is attributed as specified on the SAGE and Open Access pages
(https://us.sagepub.com/en-us/nam/open-access-at-sage).
2 Socius: Sociological Research for a Dynamic World 

there are specific cases where U.S. policies do appear to 2017; Ludwig and Brüderl 2018), and improves the stability
have a positive mental health effect among socioeconomi- and quality of romantic partnerships (Karney 2021). Second,
cally disadvantaged populations, such as among beneficia- richer individuals are at lower risk of experiencing adverse
ries of poverty-alleviation programs (Boyd-Swan et al. life events that can “provoke” depression, such as the death of
2016; Shields-Zeeman et al. 2021) and among less edu- a loved one or exposure to violence (Businelle et al. 2014;
cated older adults benefiting from changes to retirement Levitt 1999; McLeod and Kessler 1990). Third, affluence
income policies (Golberstein 2015). However, we cannot increases people’s sense of control over their environment,
extrapolate whether these effects generalize to the broader which in turn improves well-being (Kraus, Piff, and Keltner
population. In addition, given the varying strength of the 2009; Ross and Mirowsky 1992). Fourth, economic segrega-
association between income and mental well-being over tion can widen the gap in the availability and quality of shared
the life course (Cheung and Lucas 2015; Miech and public resources in rich and poor neighborhoods, amplifying
Shanahan 2000), it is possible that the causal effect of income-based disparities in quality of life and hence in men-
income is heterogeneous across stages of adulthood. tal well-being (King and Ogle 2014; Marmot 2002). A final
Studies comparing siblings also report divergent findings: potential mechanism pertains to relative income: Having a
One U.S. study found no income effects on psychological low income rank (i.e., having less income relative to others
distress within monozygotic twins or siblings (Schnittker within a region or social group) can decrease well-being via
2008), whereas twin studies in China (Li et al. 2014) and processes of stressful upward comparisons (Alderson and
Australia (Lam et al. 2019) have reported positive effects Katz-Gerro 2016; Liao 2021; Van Deurzen, Van Ingen, and
on affective well-being and decline in psychological dis- Van Oorschot 2015).
tress, respectively. In general, prior estimates from designs
intended to identify causal effects have been smaller than Sources of Bias in the Income–Well-Being
measures of association, suggesting that the latter are
inflated.
Association
This article advances our understanding of the relation- While a causal effect of income is highly plausible, treating
ship between income and mental well-being by addressing correlation or estimates from simple regression models as
some key limitations of previous associational and causal causal effects can be subject to several biases. Some sources
studies. First, I use longitudinal measures to assess the extent of confounding can lead to overestimation, perhaps even
of bias in the income-depression gradient due to measure- inducing spurious association where no causal relationship
ment errors in income. Second, I leverage genetic indices to exists. One source of spuriousness is reverse causation:
address reverse causation and other sources of confound- Previous depression dampens labor market prospects
ing—an approach that, while not free of limitations, does not (Dooley, Prause, and Ham-Rowbottom 2000; Fletcher 2013),
suffer from the same kinds of biases as previous causal stud- while higher mental well-being predicts greater productivity
ies. Specifically, unlike government policies or windfall and earnings (De Neve and Oswald 2012; Oswald, Proto,
gains that may abruptly affect life chances in adulthood, and Sgroi 2015). Similarly, conditions that independently
genes shape the risk of “exposure” throughout the life course. harm both income and mental well-being can inflate ordi-
And in contrast to family studies, the use of genetic indices nary least squares (OLS) estimates. For example, poor physi-
does not rely on comparing socially related individuals. cal health negatively impacts earning capability (Case,
Finally, to assess if the effects of income vary with age, I Fertig, and Paxson 2005; Haas 2006) and mental well-being
examine this relationship in samples representing three dis- (Steptoe 2019).
tinct stages of adulthood. However, confounders that positively affect income but
negatively affect well-being, or vice versa, can lead to an
underestimation of a true causal effect. One possibility is
Background religiosity: Some suggest a negative effect of religiosity on
Potential Casual Mechanisms Linking Income to income (Gundlach and Opfinger 2013; Herzer and Strulik
2017), while others find that religiosity protects the well-
Mental Well-Being being of the poor through various proposed means: promot-
Pathways through which socioeconomic conditions impact ing anti-wealth norms, facilitating feelings of existential
mental health have been extensively discussed in the litera- certainty, and psychologically buffering against adversities
ture (e.g., McLeod 2013; Muntaner, Borrell, and Chung 2007; (Berkessel et al. 2021; Ellison 1991; Hastings and Roeser
Patel et al. 2018). Here, I concisely review the key theorized 2020). Similarly, “other-oriented feelings” are typically
mechanisms. First, affluence can generate positive life condi- associated with both lower social class and higher mental
tions that mitigate negative feelings, such as stress, anxiety, well-being (Piff and Moskowitz 2018). Conversely, longer
loneliness, and sadness. For example, income reduces the risk work hours (Sato, Kuroda, and Owan 2020) and basing self-
of illness and mortality (Chetty et al. 2016; Strully 2009), worth on financial success (Park, Ward, and Naragon-Gainey
increases selection into marriage (Killewald and Lundberg 2017) may result in higher income but worse mental health.
Rauf 3

In principle, one could address confounding by including Raschke 2019)—a determinant of mental well-being—even
covariates. But in practice, this is challenging because con- as upwardly mobile individuals may achieve mental and
founders are not always observed, and those that are often have physical health levels comparable to those with stable high
complex relationships with income and mental well-being. For incomes (Frech and Damaske 2019; Luo and Waite 2005;
example, marital status can sometimes influence income (e.g., Pudrovska and Anikputa 2014). Broadly, this critique
by influencing duration of unemployment; Jacob and Kleinert implies that random shocks to adult income will necessarily
2014), but income also affects selection into marriage (Killewald produce smaller effects than would result from extended,
and Lundberg 2017). Similarly, while physical health may con- cumulative, and anticipated income exposure. On the other
fer an earning advantage, income also affects health. hand, unanticipated affluence can have consequences that
Conditioning on such variables can bias the estimated income do not necessarily accompany stably high or upward income
effect (Angrist and Pischke 2008:47–51). As such, without trajectories. For instance, windfalls can have socioemo-
exogenous income variation, it is not possible to satisfactorily tional fallouts, including overwhelming financial demands
address confounding or even assess the net direction of bias. from nonhousehold members and feeling lack of accom-
A different source of underestimation is the reliance of plishment in everyday tasks (Jaroszewicz et al. 2022;
prior research on current income, which is more likely to be Raschke 2019), and may also affect relationship outcomes
measured with error compared to average long-term income, in unexpected ways, such as decreasing the risk of marriage
or permanent income. Nonsystematic measurement error in among women and that of separation among married men
the explanatory variable can bias OLS estimates toward (Boertien 2012; Hankins and Hoekstra 2011).
attenuation (Griliches 1977), and some research suggests Other studies have exploited genetic similarities between
that measurement errors might shrink the OLS coefficient of twins to examine how a difference in income among other-
income by over 50 percent (Hanandita and Tampubolon wise comparable individuals impacts psychological well-
2014; Powdthavee 2010). Individuals’ current incomes may being. Because the source of causal identification is the
also depart from their stable trajectories due to temporary income difference between two related individuals, it is worth
shocks, such as unemployment or windfalls. Given that indi- considering how relatedness might matter. In general, well-
viduals adjust behaviors to temporally distribute the impact being is particularly sensitive to relative status with respect to
of shocks (Diebold and Rudebusch 1991; Hendren 2017), one’s reference group (Alderson and Katz-Gerro 2016; Liao
random shocks may be construed as a source of noise with 2021). Social comparison between twins often tends to be
respect to the “true” effect of income. horizontal, that is, motivated by feelings of solidarity and
communion rather than status competition (Huguet et al.
2017; Segal 1988; Segal and Hershberger 1999). Consistent
Limitations of Prior Causal Designs with aversion for within-twin inequality, Li et al. (2014)
Given the foregoing issues with using simple observational found that income differences within twin pairs have average
data to estimate the effect of income on mental well-being, negative effects on the well-being of both siblings. In addi-
other approaches have been pursued. Two of the most influ- tion, greater resource transfers from parents toward the chil-
ential are looking at people who have had random shocks to dren with more financial need may mitigate the impact of
their income (e.g., lottery winners or beneficiaries of new income inequalities within siblings (McGarry 2016;
government policies) and using study designs that incorpo- Zissimopoulos and Smith 2011). As such, this particular
rate family relatedness (e.g., twins). While these designs research question may not wield itself to a family study design
have provided valuable contributions in other domains, if the siblings are not raised in independent yet otherwise
extending them to investigate income effects on mental well- comparable environments. In addition, inferences based on
being runs into particular challenges. twin studies may not always be meaningful because of little
Estimates based on random shocks to adult income often outcome variation, small samples, and unobserved environ-
capture only a narrow set of ways in which income affects mental differences (DiPrete, Burik, and Koellinger 2018).
mental health. The theorized effects of income at least par-
tially operate through some mechanisms related to class, Advancing Understanding of the Income-
including conditioning of dispositions through sustained
exposure to certain living conditions and social environ-
Depression Link with New Data
ments (Bourdieu 1979:101; Freese 2017) and having a par- This article makes three improvements to previous associa-
ticular set of available choices at given points in the life tional and causal studies. First, I use longitudinal data to com-
course that in turn influence future life events (Elder 1994, pute the income gradient with respect to permanent income,
1998). Income shocks may not readily influence mecha- which is averaged over multiple time points and is less affected
nisms that operate through social networks or cumulative by measurement errors or temporary fluctuations from indi-
exposures that accompany anticipated income trajectories. viduals’ stable income trajectories. Previous studies of other
As an example, winning the lottery does not improve physi- subjective well-being indicators show that the association with
cal health (Apouey and Clark 2015; Cesarini et al. 2016; current income tends to be significantly weaker than with
4 Socius: Sociological Research for a Dynamic World 

permanent income (Bayer and Juessen 2015; D’Ambrosio, hours, diminishing the net effect of money on well-being.
Jäntti, and Lepinteur 2020; Luhmann, Schimmack, and Eid However, extended exposure to a high income during the adult
2011). As such, we may expect that permanent income will life span may lead to accumulated advantages of money in
also be more strongly linked with depressive symptoms. older age, such as better physical health and positive relation-
Second, to address reverse causation, I control for an index ships. Affluence also contributes more to financial satisfaction
predicting the genetic risk of depression (Freese 2018). in older age, but it may not equally ameliorate financial wor-
Because genes are assigned at birth, predate life events, and ries earlier in life (Plagnol 2011). Indeed, previous research
are constant over the life course, causal arrows can only point does suggest that the income-depression association increases
outward from genes to traits. In addition, conditional on our with age (Miech and Shanahan 2000). However, it is also pos-
parents’ genes, our own genes are randomly assigned at con- sible for the causal effect of money to diminish with age. For
ception. And while social environments are undoubtedly cen- example, income is more strongly associated with life satisfac-
tral in shaping human lives, almost all socially important tion in middle adulthood than later in life (Cheung and Lucas
outcomes are also at least “partially heritable” (Freese 2008). 2015), which could suggest that people perhaps become less
Because of these properties, genetic measures can be usefully concerned with material status as they grow older.
leveraged to partition putatively random variation in predic- To test how the income effect changes over the adult life
tors and outcomes to help unpack how one might affect the span, I examine three adult life stages: early midlife (early 30s
other. Previous genetic studies find high levels of heritability to mid-40s), late midlife (50–64), and older adulthood (65–85).3
in depression and mental well-being (Bartels 2015; Howard These age groups roughly map on to stages of adulthood defined
et al. 2019), which could potentially result in a reverse causal in previous research (Cherlin 2010; Erikson 1968) and mark
effect of depression on income. To the extent that reverse cau- key differences in life course stages. The youngest age group
sation might be resolved by accounting for genetic influences represents primary working years (Gangl 2005) and a time
on the former, one can attempt to control for it using a predic- when many people are undergoing important family transitions,
tor of the genetic risk of depression. such as first marriage and parenthood (Arnett 2012).4 The oldest
Third, to address other sources of confounding, I use a group is lower bounded by the normative retirement age; maxi-
genetic predictor of educational attainment as an instrumen- mum age is bounded to limit survival bias in this group.
tal variable to estimate the causal effect of income (Bollen
2012; Imbens and Angrist 1994). In case of genetic mea-
sures, the methodology is often referred to as “Mendelian
Data and Measures
randomization” (Davey Smith and Ebrahim 2003).2 While I use data from two large, nationally representative panel data
this approach is not free of limitations (as I discuss exten- sets, the National Longitudinal Study of Adolescent to Adult
sively in the following), it does not suffer from the same Health (Add Health) and the Health and Retirement Study
sources of bias as previous causal designs. Specifically, (HRS; 2021). Add Health began as a sample of U.S. adoles-
unlike studies of unanticipated income shocks or policies cents in Grades 7 to 12 in 1994–1995 and includes supple-
that affect people abruptly in adulthood, genetic influences mental oversamples for underrepresented groups (Harris and
impact income via sustained exposures over the life span. Udry 2018). To date, five interview waves have been col-
And in contrast to family studies, analyses using genetic lected, with the last two waves being roughly 10 years apart;
influences are not constrained to comparisons of related indi- the five-wave design encompasses 12,300 individuals. The
viduals. Previous research has fruitfully applied similar ana- HRS represents U.S. adults age 50 years and older and has
lytical approaches to apprehend causal mechanisms been conducted biennially since 1992, with new birth cohorts
underlying other social phenomena, such as educational out- being added every 6 years. In 2018, there were about 21,000
comes (Hughes et al. 2021), income (Wang et al. 2020), and respondents. Both surveys have collected salivary DNA sam-
self-harm behavior (Lim et al. 2020). ples and genotyped most of their respondents. Available
genetic indices are derived from genome-wide association
studies (GWASs) of European-ancestry individuals and may
The Income–Well-Being Link over the Life Course not transfer well to other ancestries (Martin et al. 2019). As
Over the adult life course, the strength of the income-depres- such, these analyses are restricted to individuals with
sion link may vary as money may “buy” well-being for differ-
ent reasons. One possibility is that income effects may grow 3
I do not examine adults under age 30 because during this period
with age. During middle adulthood, higher earnings may of “emerging adulthood,” reported incomes may not necessarily
accompany work-related anxiety and stress and longer work reflect actual living conditions due to high prevalence of financial
dependence on parents (Arnett 2012).
4
Arnett (2012) proposed the term “young adulthood” for 30s to
2
The term alludes to Mendel’s second law of random assortment: mid-40s. Others, however, argue that this label should strictly
The inheritance of one trait does not depend on that of another be used for ages 18 to 25 (Society for Adolescent Health and
(Davey Smith and Ebrahim 2003). Medicine 2017).
Rauf 5

Table 1. Summary Statistics of Analytic Samples.

Early Midlife Late Midlife Older Adults


Age (average years) 37.27 60.48 71.93
Women (proportion) 0.50 0.54 0.56
Years of education (average) 14.51 13.93 13.33
Household income (in thousands, 2020 $, average) 98.81 129.72 84.44
Depressive symptoms (rescaled 1–10, average) 2.12 2.45 2.39
Birth year (range) 1974–1983 1935–1959 1916–1949
Survey year (range) 2015–2017 2000–2019 2000–2019
N (cases) 3,832 6,393 7,727

Sources: National Longitudinal Study of Adolescent to Adult Health and Health and Retirement Study.
Note: Weighted averages reported. Age and survey year correspond to the wave when depressive symptoms were recorded. All samples are restricted to
individuals with European ancestry.

genotypes indicative of European ancestry (5,731 Add Health income at $500. Current income corresponds to Wave 5 in
and 12,090 HRS respondents). Add Health and the last wave in the HRS analytical sample.
The early midlife sample includes Add Health respon- Permanent income in Add Health is averaged over Waves 4
dents in Wave 5 (ages 33–44), the late midlife sample and 5 (about 10 years), and in HRS, it is averaged over three
includes HRS respondents ages 50 to 64, and the older adults consecutive waves (roughly 5 years)—this corresponds to
sample includes HRS respondents age 65+.5 Table 1 approximations used in previous research (e.g., D’Ambrosio
describes the estimation samples. Sampling weights corre- et al. 2020; Gangl 2005).6 Income measures are adjusted for
sponding to the respondents’ last period of observation are inflation using implicit price deflator indexed for 2020.
used to obtain parameters representative of underlying popu- The relationship between income and various indicators
lations. Because the HRS includes partners of married of mental well-being tends to follow a log-linear pattern
respondents, standard errors are clustered by household. (e.g., Stevenson and Wolfers 2013; Zimmerman and Katon
2005), indicating that the marginal effect of an absolute
increase in income diminishes with the level of income. This
Depressive Symptoms pattern is also reflected in the samples used in this study (dis-
Measures of depressive symptoms (DS) are based on abbrevi- cussed further in the results section). As such, I use logged
ated versions of the Center for Epidemiologic Studies measures of both current and permanent income. The gradi-
Depression (CES-D) scale (Radloff 1977). Abbreviated ver- ent with respect to log income reflects the average effect of a
sions of this scale were administered in Add Health (a five- percentage change in income.
item version) and HRS (an eight-item, dichotomous-response
version). These indices have been validated in prior studies
Polygenic Predictors
and share psychometric properties with the full CES-D
(Perreira et al. 2005; Turvey, Wallace, and Herzog 1999). Both Most traits are understood to be polygenic, that is, affected by
indices have good internal consistency in the analytic samples multiple genetic variants (McCarthy et al. 2008). To identify the
(α = 0.81 in Add Health and α = 0.78 in HRS). Outcomes are causal genetic variants, GWASs are conducted in large samples,
standardized to create z scores comparable across samples. which entail estimating regressions of the trait of interest on
each variant, typically a single nucleotide polymorphism (SNP;
for a succinct primer, see Conley 2016). GWASs are estimated
Income within ancestrally homogeneous populations and control for
Both data sets reported pretax household income. In Add gender, age, and principal components of genetic ancestry that
Health, income excludes government transfers and is mea- adjust for population stratification within ancestral groups.7 The
sured on a categorical scale. Income was transformed into a
continuous measure using midpoint approximation; for the
highest income bin, income is approximated at 33 percent 6
Posttax and transfer household income averaged over 2 to 5 years
above the lower bound. Income in HRS includes government explains majority of the variation in permanent income measured
transfers (Bugliari et al. 2021). I bottom-coded household over 20+ years (Brady et al. 2018).
7
Even within racially homogeneous groups, there may be system-
atic historical differences in exposures due to assortative mating,
5
Nursing home residents are excluded because the HRS target pop- residential sorting, or social distance based on ethnicity or lan-
ulation is noninstitutionalized adults, and so, while these partici- guage, and so on. Including the principal components of genetic
pants are interviewed, they are not assigned a sample weight. ancestry controls for such sources of spurious correlation.
6 Socius: Sociological Research for a Dynamic World 

GWAS yields summary statistics, comprising of a separate the coefficient of the PGI and inflate that of income. Ancillary
regression coefficient for each SNP, which are used as weights analyses tested the sensitivity of these models to measure-
to construct polygenic indices (PGIs), as follows: ment error and are discussed later.
k While the aformentioned models address biases due to
PGI = ∑ β SNP , measurement error and reverse causation, they do not address
i j ij
j =1 omitted variable bias. As mentioned earlier, controlling for
where i indexes individuals and j indexes the frequency of the omitted variables in this case is complicated by the fact that
less frequent (“minor”) allele on the jth SNP. βj is the coeffi- some key sources of confounding (e.g., marital status and
cient of the effect of SNPj on the trait. To avoid overfitting, health) also mediate the effects of income on mental health.
the sample for which a PGI is constructed is not part of the As such, a different strategy is needed to assess the effects of
data set on which the GWAS is based. The PGIs are standard- income differences between individuals. I use two-stage
ized to have mean of 0 and variance of 1 within each sample. least squares (2SLS) regressions to estimate this effect
PGIs used in this study are derived from analyses by Becker (Wooldridge 2010). In the first-stage equation, I use the
et al. (2021). These PGIs are based on multitrait analysis of EA-PGI as an instrumental variable for income to predict the
GWAS, which leverages shared genetic variance between phe- proportion of income that, net of covariates, is not caused by
notypes to enhance the predictive power of any given PGI factors that confound the income-depression relationship.
(Turley et al. 2018). I use a PGI of DS as a genetic control and The predicted values of income from this model are then
a PGI of educational attainment (EA) as an instrument for used to estimate the income effect in the second stage. The
income. The EA-PGI can affect income through a number of 2SLS model is specified as follows:
socially mediated pathways, including, most directly, via edu- Incomei = γ.PGIeai + Xi ∅ + w i
cational attainment and through potential indirect pathways, � (2).
including occupational aspirations, financial savvy, and selec- DSi = β2 SLS .Incomei + Xi δ + εi
tion of a spouse with high socioeconomic status (SES; Barban
These regressions are estimated using the ivreg2 program in
et al. 2021; Barth, Papageorge, and Thom 2019; Belsky et al.
Stata (Baum, Schaffer, and Stillman 2010).
2016, 2018). Assumptions entailed in using the PGI as an
Two important assumptions are built into the baseline
instrumental variable are discussed in the following section.
2SLS model. First, the EA-PGI needs to be a significant and
strong predictor of income. The violation of this assumption
Analytic Strategy can result in “weak instrument bias,” which could exaggerate
β estimates (Staiger and Stock 1997). As such, I first
The analytic strategy is twofold. First, I estimate the baseline 2SLS
OLS, which is consistent with previous associational studies test that the instrument is relevant and strong by comparing
(Model 1). I then make two adjustments to the baseline F statistics from the first-stage regression with critical values
model. In Model 2, I use permanent instead of current income from the Stock-Yogo test for a true rejection rate of 5 percent
to assess the extent of attenuation bias induced by noise in and a maximum tolerated rejection rate of 10 percent (Stock
income measurement. Model 3 includes a control for the and Yogo 2005).
depression-PGI, which will help assess the extent to which The second assumption is that there is no path from the
reverse causation (attributable to genetic risk of depression) EA-PGI to DS that is unmediated by income, also known as
is responsible for inflating the income-depression gradient. the “exclusion restriction” (Wooldridge 2010:91–92). It fol-
The OLS models are specified as follows: lows from this assumption that conditional on income and
covariates, the EA-PGI should be randomly distributed with
DS = β .Income + X δ + ε (1), respect to DS. One potential source of violation of this condi-
i OLS i i i tion is “horizontal pleiotropy,” which is induced when the
where i indexes individuals, X is a vector of covariates, and same parts of the genome predict both the outcome and
the coefficient β estimates the income effect conditional on X. explanatory variable (Paaby and Rockman 2013). Indeed,
The standard set of covariates across models includes the first the PGIs of EA and depression are significantly negatively
10 principal components of genetic ancestry, gender, and age correlated in these estimation samples (in Add Health,
(Burgess et al. 2019). HRS analyses also include period fixed ρ = −.35, p < .001; in HRS, ρ = −.29, p < .001). To address
effects to adjust for variation in timing of data collection. For this, I estimate an adjusted 2SLS model controlling for the
reasons discussed earlier, I do not control for variables that depression-PGI (DiPrete et al. 2018).8 A second source of
could be affected by income (e.g., health or marital status).
It should be noted that estimates of the income coefficient
in Model 3 may be upwardly biased due to classical mea- 8
The method proposed by DiPrete, Burik, and Koellinger (2018)
surement error in the depression-PGI, which gets incorpo- entails obtaining a PGI of the outcome conditional on the exposure.
rated in the computation of PGIs from GWAS summary In the absence of the conditional PGI, the authors suggest control-
statistics (Becker et al. 2021). In turn, this noise can attenuate ling for a PGI predicting the outcome.
Rauf 7

Figure 1. Relationship between depressive symptoms and income.


Sources: National Longitudinal Study of Adolescent to Adult Health and Health and Retirement Study.
Note: Lines are fitted using locally weighted scatterplot smoothing (Lowess).

violation is “genetic nurture”—that is, the indirect influence of the method and analysis, and the results of the 2SLS mod-
of parents’ genes on a person’s mental well-being that is els are discussed in light of the decomposition analysis.
transmitted via the rearing environment (Kong et al. 2018)—
which can induce a correlation between EA-PGI and depres-
Results
sion. This concern is partly mitigated by controlling for
parental education (Hart, Little, and van Bergen 2021).9,10 Using Permanent Income Instead of Current
Sample sizes for the adjusted 2SLS models are relatively Income
smaller due to missing parental education data.
Another assumption about the exclusion restriction is that I begin by discussing patterns of association between DS and
education and depressive symptoms are conditionally inde- logged income measures. Figure 1 depicts the income-depres-
pendent. This assumption is supported by previous causal sion gradient with respect to log of current and permanent
research (Avendano, de Coulon, and Nafilyan 2020; income in the three samples. Consistent with expectations
Dahmann and Schnitzlein 2019; Sperandei et al. 2023). based on prior research, the gradient with respect to perma-
However, in ancillary analyses, I also test this assumption by nent income is steeper across all three samples.
decomposing the causal education effect into income and Results of OLS models are summarized in Figure 2. In
nonincome effects (Dippel et al. 2020). While this decompo- the baseline OLS model, current income gradients range
sition method itself entails assumptions that may not be fully between −.2 and −.3 (full results in Table S1).11 The gra-
satisfied, it provides at least some useful benchmarks for the dients with respect to permanent income are larger, by .05
extent and direction of bias in β induced by a nonin- SD on average, suggesting that fluctuations or measure-
2SLS ment errors in current income are a potential source of
come effect of education. Supplement C provides full details
attenuation bias. For both current and permanent income,
coefficient estimates for older adults are significantly
9
smaller compared to estimates for adults in early and late
The ideal strategy to block this pathway is to control for parental
midlife.
EA-PGIs (Barcellos, Carvalho, and Turley 2021). Parental PGIs
are not available in these data. Another potential alternative is fam-
ily fixed effects, however those estimates can be too conservative
(Trejo and Domingue 2018) and can be biased because siblings
11
influence each others’ environments. One cannot compare these results directly to previous studies
10
Parental education is defined here as the average of the years of because of differences in populations and outcomes. Stevenson
education of both parents or the available parent. Including controls and Wolfers’s (2013) estimated average effect for positive affective
for both parents separately does not meaningfully change the results well-being, .67 SD/log income, provides the closest benchmark in
but reduces sample sizes due to missing data (Table S5). comparable units.
8 Socius: Sociological Research for a Dynamic World 

Figure 2. Ordinary least squares estimates of the income-depression gradient across adult life stages.
Sources: National Longitudinal Study of Adolescent to Adult Health and Health and Retirement Study.
Note: Nonoverlapping confidence intervals indicate significant differences between estimates at 5 percent level.

Adjusting for Genetic Risk of Depression 16.38 (Stock and Yogo 2005), indicating that the instrument
relevance condition is satisfied (Table S2).12
Across all three age groups, average DS between individ-
All estimates from the baseline 2SLS models are statisti-
uals significantly vary by the depression-PGI (Figure 3).
cally significant (right panel in Figure 4 and Table S3) and
This variation in mental health is reflecting influences
are comparable in magnitude to the gradient estimated from
that rather than being caused by income, may conversely
OLS regressions. Because the confidence intervals of 2SLS
influence income attainment. Indeed, when a control for
estimates are wider, differences between estimates for late
the depression-PGI is included in Model 3, estimates of
midlife and older adulthood are no longer distinguishable,
the standard OLS model shrink by about .01 SD (Figure 2
but early midlife estimates are significantly smaller than esti-
and Table S1). As mentioned earlier, these analyses may
mates for older adulthood. As discussed earlier, baseline esti-
underestimate reverse causation due to measurement error
mates could be biased due to horizontal pleiotropy and
in the DS-PGI (Becker et al. 2021). In ancillary analyses,
genetic nurture. The adjusted 2SLS model assesses the sensi-
estimates adjusted for measurement error contract the
tivity of estimates to these biases. In the adjusted model, esti-
estimates of income effect by.02 SD on average (Table
mates remain significant for older adulthood and are
S6). Again, estimates for older adulthood are substantially
marginally significant for late midlife (p = .06) but are insig-
smaller compared to earlier in the life course.
nificant for early midlife. In terms of magnitude, the adjusted
2SLS estimates for the two older age groups are comparable
Two-Stage Least Squares Estimation of Causal to OLS coefficients, but estimates for early midlife are virtu-
Income Effect ally zero. The consistency of estimates across baseline and
adjusted models suggests that income does have a causal
The 2SLS models assess the extent of bias in the income- effect on depression in late midlife and older adulthood.
depression gradient resulting from factors in addition to The 2SLS models also assume that net of income, education
those discussed previously. The first-stage regression pre- does not directly affect DS. To assess how much the violation
dicts the proportion of income that is in turn predicted by the of this assumption might bias 2SLS estimates, I conducted a
EA-PGI. Income differences between individuals signifi- decomposition analysis assessing the effect of education on DS
cantly vary by the EA-PGI across all three samples in both via income and nonincome pathways (Figure 5 and Table S4).
the baseline models and models adjusted for depression-PGI
and parental SES (left panel in Figure 4). Across all first-
stage models, the EA-PGI has a significant coefficient, and 12
I report Kleibergen-Paap rk Wald F statistic, which adjusts for
the first-stage F statistics are larger than critical value of clustered standard errors (Baum, Schaffer, and Stillman 2010).
Rauf 9

Figure 3. Distribution of depressive symptoms by depression PGI.


Sources: National Longitudinal Study of Adolescent to Adult Health and Health and Retirement Study.
Note: Bars show weighted averages; 95 percent confidence intervals shown.

Figure 4. Two-stage least squares estimates of income effect on depressive symptoms.


Sources: National Longitudinal Study of Adolescent to Adult Health and Health and Retirement Study.
Note: Nonoverlapping confidence intervals indicate significant difference between estimates at 5 percent level.

In the baseline models, the total effect of education is negative, Ancillary analyses show that the direction of income effect in
with nearly all of the beneficial effect being driven via the early midlife specifically flips when parental education is
income effect; the nonincome effect is positive. In the adjusted included in the model. A potential reason may be that parental
models, this pattern continues to hold for late midlife and older SES is more strongly correlated with household income in
adulthood (suggesting that the assumption of a negligible non- early midlife than at later ages (e.g., in Wave 5, about 8 percent
income effect of the EA-PGI is largely valid and results in only of Add Health participants lived with their parents, and 11 per-
a small amount of downward bias in the 2SLS estimates). cent received large financial transfers from parents and other
However, income effect is reversed in the early midlife sample relatives).
when additional covariates are included, indicating that the A final concern is that the EA-PGI only captures income
exclusion restriction is strongly violated for this sample. variation via pathways affected by the PGI. Income can of
10 Socius: Sociological Research for a Dynamic World 

Figure 5. Decomposition of effect of education on depressive symptoms.


Sources: National Longitudinal Study of Adolescent to Adult Health and Health and Retirement Study.

course also vary due to other reasons, and in turn, these alter- While using permanent income is an improvement over prior
native pathways to a higher income can have distinct impli- literature, these estimates are averaged across heterogeneous
cations for mental well-being. More generally, the concern is adult income trajectories (Frech and Damaske 2019; Song
that 2SLS estimates reflect local average treatment effects et al. 2022), which is a residual source of noise. A shortcom-
based on “compliers,” who may differ from the sample aver- ing shared with prior literature is that the income measures
age (Card 2001; Imbens and Angrist 1994). Ancillary analy- are not fully adjusted for fiscal transfers. Fiscally adjusted
ses approximating the profile of compliers (using an approach income distributions tend to be less unequal (Brady et al.
outlined by Marbach and Hangartner 2020) indicate that 2018), which could have implications for these results.
compliers were largely similar to the full sample in terms of Finally, the exclusion restriction for the 2SLS estimator was
class background, gender, birth cohort, and depression PGI not fully satisfied across samples, being strongly violated in
(Figures S1–S3). the early midlife sample.
This article makes two contributions to the understanding
of the income-depression link. First, improvements to the
Discussion associational model allow adjudging the extent to which
Social scientists have long documented a negative gradient two major known sources of bias might be affecting esti-
between income and mental health, but whether this correlation mates of the income–well-being gradient. The fact that this
reflects a causal effect has been an open question. Indeed, sev- gradient only marginally changed due to these corrections—
eral previous causal studies have reported small, null, and even and the biases acted in opposite directions—suggests that
negative mental health effects. This study assessed this prob- these are perhaps not the most worrisome sources of bias.
lem from a different vantage using data and methods different Second, the putative causal effect of income was estimated
from those employed in previous literature. Results suggested using an approach that—while not free of limitations—does
that measurement errors in income and reverse causation due to not suffer from the same shortcomings as previous causal
the genetic risk of depression are meaningful sources of bias, designs. Specifically, it was an improvement over previous
acting in opposite directions. In models estimating putatively studies using natural experiments or policy changes that
causal effects of income, the effect sizes were as large as the typically capture short-term and narrow effects of income
income-depression gradient for adults age 50 and older. For the shocks and family studies where within-family effects do
early midlife sample, the more stringent assumptions of the not readily generalize to unrelated individuals. Pooled
causal model were not satisfied, but estimates from the less across samples and models, the results presented here are
stringent model aligned with those for older adults. Age differ- suggestive of an average beneficial causal effect of income
ences in estimates—while present—were relatively small and on mental well-being.
inconsistent in direction across OLS and 2SLS regressions.
Key limitations of these analyses include the following. Authors’ Note
Because genetic data are not representative of underlying Replication files are available at https://osf.io/grjs9/. Add Health
populations at all levels of data collection, we cannot directly restricted-use data can be requested at: https://data.cpc.unc.edu/
use PGIs derived from European ancestry populations in projects/2/view. HRS data are publicly-accessible at: https://hrs.isr.
non-European ancestry samples with reasonable confidence. umich.edu/data-products.
Rauf 11

Funding GMM and AC/HAC, LIML and k-Class Regression.” http://


ideas.repec.org/c/boc/bocode/s425401.html.
The author(s) disclosed receipt of the following financial support
Bayer, Christian, and Falko Juessen. 2015. “Happiness and the
for the research, authorship, and/or publication of this article: The
Persistence of Income Shocks.” American Economic Journal:
article uses data from the Health and Retirement Study, which is
Macroeconomics 7(4):160–87.
conducted by the University of Michigan and is sponsored by the
Becker, Joel, Casper A. P. Burik, Grant Goldman, Nancy
National Institute on Aging (NIA U01AG009740), and the National
Wang, Hariharan Jayashankar, Michael Bennett, Daniel W.
Longitudinal Study of Adolescent to Adult Health, which is con-
Belsky, et al. 2021. “Resource Profile and User Guide of
ducted by the University of North Carolina at Chapel Hill and is
the Polygenic Index Repository.” Nature Human Behaviour
sponsored by the National Institute on Aging (NIA U01AG071448
5(12):1744–58.
and U01AG071450). This research was supported by the Stanford
Belsky, Daniel W., Benjamin W. Domingue, Robbee Wedow,
Sociology Research Opportunity Grant and a dissertation fellow-
Louise Arseneault, Jason D. Boardman, Avshalom Caspi,
ship by the Stanford Institute for Research in the Social Sciences.
Dalton Conley, et al. 2018. “Genetic Analysis of Social-Class
Mobility in Five Longitudinal Studies.” Proceedings of the
ORCID iD National Academy of Sciences 115(31):E7275–84.
Belsky, Daniel W., Terrie E. Moffitt, David L. Corcoran, Benjamin
Tamkinat Rauf https://orcid.org/0000-0001-8900-3432 Domingue, Hona Lee Harrington, Sean Hogan, Renate Houts,
et al. 2016. “The Genetics of Success: How Single-Nucleotide
Supplemental Material Polymorphisms Associated with Educational Attainment
Relate to Life-Course Development.” Psychological Science
Supplemental material for this article is available online. 27(7):957–72.
Berkessel, Jana B., Jochen E. Gebauer, Mohsen Joshanloo, Wiebke
Bleidorn, Peter J. Rentfrow, Jeff Potter, and Samuel D. Gosling.
References
2021. “National Religiosity Eases the Psychological Burden of
Alderson, Arthur S., and Tally Katz-Gerro. 2016. “Compared to Poverty.” Proceedings of the National Academy of Sciences of
Whom? Inequality, Social Comparison, and Happiness in the the United States of America 118(39):1–6.
United States.” Social Forces 95(1):25–54. Boertien, Diederik. 2012. “Jackpot? Gender Differences in the
Angrist, Joshua, and Jörn-Steffen Pischke. 2008. Mostly Harmless Effects of Lottery Wins on Separation.” Journal of Marriage
Econometrics: An Empiricist’s Companion. Princeton, NJ: and Family 74(5):1038–53.
Princeton University Press. Bollen, Kenneth A. 2012. “Instrumental Variables in Sociology and
Apouey, Benedicte, and Andrew E. Clark. 2015. “Winning Big but the Social Sciences.” Annual Review of Sociology 38:37–72.
Feeling No Better? The Effect of Lottery Prizes on Physical Bourdieu, Pierre. 1979. Distinction: A Social Critique of the
and Mental Health.” Health Economics 24(5):516–38. Judgement of Taste. London: Routledge.
Arnett, Jeffrey Jensen. 2012. “New Horizons in Research on Emerging Boyd-Swan, Casey, Chris M. Herbst, John Ifcher, and Homa
and Young Adulthood.” Pp. 231–44 in Early Adulthood in a Zarghamee. 2016. “The Earned Income Tax Credit, Mental
Family Context, edited by A. Booth, S. L. Brown, N. S. Landale, Health, and Happiness.” Journal of Economic Behavior and
W. D. Manning, and S. M. McHale. New York, NY: Springer. Organization 126:18–38.
Avendano, Mauricio, Augustin de Coulon, and Vahé Nafilyan. Brady, David, Marco Giesselmann, Ulrich Kohler, and Anke
2020. “Does Longer Compulsory Schooling Affect Mental Radenacker. 2018. “How to Measure and Proxy Permanent
Health? Evidence from a British Reform.” Journal of Public Income: Evidence from Germany and the U.S.” Journal of
Economics 183:104137. doi:10.1016/j.jpubeco.2020.104137. Economic Inequality 16(3):321–45.
Barban, Nicola, Elisabetta De Cao, Sonia Oreffice, and Climent Brickman, Philip, Dan Coates, and Ronnie Janoff-Bulman.
Quintana-Domeque. 2021. “The Effect of Education on 1978. “Lottery Winners and Accident Victims: Is Happiness
Spousal Education: A Genetic Approach.” Labour Economics Relative?” Journal of Personality and Social Psychology
71:102023. doi:10.1016/j.labeco.2021.102023. 36(8):917–27.
Barcellos, Silvia H., Leandro Carvalho, and Patrick Turley. 2021. Bugliari, Delia, Joanna Carroll, Orla Hayden, Jessica Hayes,
“The Effect of Education on the Relationship between Genetics, Michael Hurd, Adam Karabatakis, Regan Main, et al. 2021.
Early-Life Disadvantages, and Later-Life SES.” NBER Working “RAND HRS Longitudinal File 2018 (V2) Documentation
Paper No. 28750, National Bureau of Economic Research. http:// (Final Release). RAND Corporation, Santa Monica, CA.”
www.nber.org/papers/w28750. Burgess, Stephen, George Davey Smith, Neil M. Davies, Frank
Bartels, Meike. 2015. “Genetics of Wellbeing and Its Components Dudbridge, Dipender Gill, M. Maria Glymour, Fernando P.
Satisfaction with Life, Happiness, and Quality of Life: A Hartwig, et al. 2019. “Guidelines for Performing Mendelian
Review and Meta-analysis of Heritability Studies.” Behavior Randomization Investigations.” Wellcome Open Research
Genetics 45:137–56. 4(186):1–19.
Barth, Daniel, Nicholas W. Papageorge, and Kevin Thom. 2019. Businelle, Michael S., Britain A. Mills, Karen G. Chartier,
“Genetic Endowments and Wealth Inequality.” Journal of Darla E. Kendzor, Jennifer M. Reingle, and Kerem Shuval.
Political Economy 128(4):1474–522. 2014. “Do Stressful Events Account for the Link between
Baum, Christopher F., Mark E. Schaffer, and Steven Stillman. 2010. Socioeconomic Status and Mental Health?” Journal of Public
“ivreg2: Stata Module for Extended Instrumental Variables/2SLS, Health 36(2):205–12.
12 Socius: Sociological Research for a Dynamic World 

Card, David. 2001. “Estimating the Return to Schooling: Progress Eaton, William W., Carles Muntaner, Gregory Bovasso, and
on Some Persistent Econometric Problems.” Econometrica Corey Smith. 2001. “Socioeconomic Status and Depressive
69(5):1127–60. Syndrome: The Role of Inter- and Intra-generational Mobility,
Case, Anne, Angela Fertig, and Christina Paxson. 2005. “The Government Assistance, and Work Environment.” Journal of
Lasting Impact of Childhood Health and Circumstance.” Health and Social Behavior 42(3):277–94.
Journal of Health Economics 24(2):365–89. Elder, Glen H. 1994. “Time, Human Agency, and Social Change:
Cesarini, David, Erik Lindqvist, Robert Ostling, and Bjorn Wallace. Perspectives on the Life Course.” Social Psychology Quarterly
2016. “Wealth, Health, and Child Development: Evidence 57(1):4–15.
from Administrative Data on Swedish Lottery Players.” The Elder, Glen H. 1998. “The Life Course as Developmental Theory.”
Quarterly Journal of Economics 131(2):687–738. Child Development 69(1):1–12.
Cherlin, Andrew J. 2010. “Demographic Trends in the United Ellison, Christopher G. 1991. “Religious Involvement and
States: A Review of Research in the 2000s.” Journal of Subjective Well-Being.” Journal of Health and Social
Marriage and Family 72(3):403–19. Behavior 32(1):80–99.
Chetty, Raj, Michael Stepner, Sarah Abraham, Shelby Lin, Erikson, Erik H. 1968. Identity: Youth and Crisis. New York, NY:
Benjamin Scuderi, Nicholas Turner, Augustin Bergeron, et al. W. W. Norton & Company.
2016. “The Association between Income and Life Expectancy Fletcher, Jason. 2013. “Adolescent Depression and Adult
in the United States, 2001-2014.” Journal of the American Labor Market Outcomes.” Southern Economic Journal
Medical Association 315(16):1750–66. 80(1):26–49.
Cheung, Felix, and Richard E. Lucas. 2015. “When Does Money Frech, Adrianne, and Sarah Damaske. 2019. “Men’s Income
Matter Most? Examining the Association between Income and Trajectories and Physical and Mental Health at Midlife.”
Life Satisfaction over the Life Course.” Psychology and Aging American Journal of Sociology 124(5):1372–412.
30(1):120–35. Freese, Jeremy. 2008. “Genetics and the Social Science Explanation
Conley, Dalton. 2016. “Socio-genomic Research Using Genome-Wide of Individual Outcomes.” American Journal of Sociology
Molecular Data.” Annual Review of Sociology 42(1):275–99. 114(Suppl.):S1–35.
Courtin, Emilie, Peter Muennig, Nandita Verma, James A. Freese, Jeremy. 2017. “What about the Behavioral Constellation
Riccio, Mylene Lagarde, Paolo Vineis, Ichiro Kawachi, of Advantage?” Behavioral and Brain Sciences 40:e326.
et al. 2018. “Conditional Cash Transfers and Health of Low- doi:10.1017/S0140525X17000966.
Income Families in the US: Evaluating the Family Rewards Freese, Jeremy. 2018. “The Arrival of Social Science Genomics.”
Experiment.” Health Affairs 37(3):438–46. Contemporary Sociology 47(5):524–36.
Dahmann, Sarah C., and Daniel D. Schnitzlein. 2019. “No Evidence Gangl, Markus. 2005. “Income Inequality, Permanent Incomes, and
for a Protective Effect of Education on Mental Health.” Income Dynamics: Comparing Europe to the United States.”
Social Science and Medicine 241:112584. doi:10.1016/j.soc- Work and Occupations 32(2):140–62.
scimed.2019.112584. Gardner, Jonathan, and Andrew J. Oswald. 2007. “Money and
D’Ambrosio, Conchita, Markus Jäntti, and Anthony Lepinteur. Mental Wellbeing: A Longitudinal Study of Medium-Sized
2020. “Money and Happiness: Income, Wealth and Subjective Lottery Wins.” Journal of Health Economics 26(1):49–60.
Well-Being.” Social Indicators Research 148(1):47–66. Golberstein, Ezra. 2015. “The Effects of Income on Mental Health:
Davey Smith, George, and Shah Ebrahim. 2003. “‘Mendelian Evidence from the Social Security Notch.” The Journal of
Randomization’: Can Genetic Epidemiology Contribute to Mental Health Policy and Economics 18(1):27–37.
Understanding Environmental Determinants of Disease?” Griliches, Zvi. 1977. “Estimating the Returns to Schooling: Some
International Journal of Epidemiology 32(1):1–22. Econometric Problems.” Econometrica 45(1):1–22.
De Neve, Jan Emmanuel, and Andrew J. Oswald. 2012. “Estimating Gundlach, Erich, and Matthias Opfinger. 2013. “Religiosity as a
the Influence of Life Satisfaction and Positive Affect on Later Determinant of Happiness.” Review of Development Economics
Income Using Sibling Fixed Effects.” Proceedings of the 17(3):523–39.
National Academy of Sciences of the United States of America Haas, Steven A. 2006. “Health Selection and the Process of
109(49):19953–58. Social Stratification: The Effect of Childhood Health on
Diebold, Francis X., and Glenn D. Rudebusch. 1991. “Is Socioeconomic Attainment.” Journal of Health and Social
Consumption Too Smooth? Long Memory and the Deaton Behavior 47(4):339–54.
Paradox.” The Review of Economics and Statistics 73(1):1–9. Hanandita, Wulung, and Gindo Tampubolon. 2014. “Does Poverty
Dippel, Christian, Robert Gold, Stephan Heblich, and Rodrigo Reduce Mental Health? An Instrumental Variable Analysis.”
Pinto. 2020. “Mediation Analysis in IV Settings with a Single Social Science and Medicine 113:59–67.
Instrument.” https://christiandippel.com/IVmediate_.pdf. Hankins, Scott, and Mark Hoekstra. 2011. “Lucky in Life,
DiPrete, Thomas A., Casper A. P. Burik, and Philipp D. Koellinger. Unlucky in Love?: The Effect of Random Income Shocks
2018. “Genetic Instrumental Variable Regression: Explaining on Marriage and Divorce.” Journal of Human Resources
Socioeconomic and Health Outcomes in Nonexperimental 46(2):403–26.
Data.” Proceedings of the National Academy of Sciences of the Harris, Kathleen Mullan, and J. Richard Udry. 2018. “The National
United States of America 115(22):E4970–79. Longitudinal Study of Adolescent to Adult Health (Add Health),
Dooley, David, Joann Prause, and Kathleen Ham-Rowbottom. 2000. Waves I-V” [Machine-readable data file and documentation].
“Underemployment and Depression: Longitudinal Relationships.” Chapel Hill: Carolina Population Center, University of North
Journal of Health and Social Behavior 41(4):421–36. Carolina.
Rauf 13

Hart, Sara A., Callie Little, and Elsje van Bergen. 2021. “Nurture Lam, Janine R., Jessica Tyler, Katrina J. Scurrah, Nicola J.
Might Be Nature: Cautionary Tales and Proposed Solutions.” Reavley, and Gillian S. Dite. 2019. “The Association between
Nature Science of Learning 6(1):1–12. Socioeconomic Status and Psychological Distress: A within
Hastings, Orestes P., and Kassandra K. Roeser. 2020. “Happiness and between Twin Study.” Twin Research and Human Genetics
in Hard Times: Does Religion Buffer the Negative Effect of 22(5):312–20.
Unemployment on Happiness?” Social Forces 99(2):447–73. Levitt, Steven D. 1999. “The Changing Relationship between Income
Health and Retirement Study. 2021. “Health and Retirement Study, and Crime Victimization.” Economic Policy Review 5(3):87–98.
Public Use Dataset.” https://hrs.isr.umich.edu/data-products. Li, Hongbin, Pak Wai Liu, Maoliang Ye, and Junsen Zhang. 2014.
Heflin, Colleen M., and John Iceland. 2009. “Poverty, Material Hardship, “Does Money Buy Happiness? Evidence from Twins in Urban
and Depression.” Social Science Quarterly 90(5):1051–71. China.” https://www.semanticscholar.org/paper/Does-Money-
Hendren, Nathaniel. 2017. “Knowledge of Future Job Loss and Buy-Happiness-Evidence-from-Twins-in-Li-Liu/98f621335d9
Implications for Unemployment Insurance.” American Economic b48f227daa41cbc4d21d355300716
Review 107(7):1778–823. Liao, Tim Futing. 2021. “Income Inequality, Social Comparison,
Herzer, Dierk, and Holger Strulik. 2017. “Religiosity and Income: and Happiness in the United States.” Socius 7. doi:10.
A Panel Cointegration and Causality Analysis.” Applied 1177/2378023120985648.
Economics 49(30):2922–38. Lim, Kai Xiang, Frühling Rijsdijk, Saskia P. Hagenaars, Adam
Howard, David M., Mark J. Adams, Toni Kim Clarke, Jonathan Socrates, Shing Wan Choi, Jonathan R. I. Coleman, Kylie
D. Hafferty, Jude Gibson, Masoud Shirali, Jonathan R. I. P. Glanville, et al. 2020. “Studying Individual Risk Factors
Coleman, et al. 2019. “Genome-Wide Meta-analysis of for Self-Harm in the UK Biobank: A Polygenic Scoring
Depression Identifies 102 Independent Variants and Highlights and Mendelian Randomisation Study.” PLoS Medicine
the Importance of the Prefrontal Brain Regions.” Nature 17(6):e1003137. doi:10.1371/journal.pmed.1003137.
Neuroscience 22(3):343–52. Lindqvist, Erik, Robert Östling, and David Cesarini. 2020. “Long-
Hughes, Amanda, Kaitlin H. Wade, Matt Dickson, Frances Rice, Run Effects of Lottery Wealth on Psychological Well-Being.”
Alisha Davies, Neil M. Davies, and Laura D. Howe. 2021. Review of Economic Studies 87(6):2703–26.
“Common Health Conditions in Childhood and Adolescence, Ludwig, Volker, and Josef Brüderl. 2018. “Is There a Male Marital
School Absence, and Educational Attainment: Mendelian Wage Premium? New Evidence from the United States.”
Randomization Study.” NPJ Science of Learning 6(1):1–9. American Sociological Review 83(4):744–70.
Huguet, Pascal, Michèle Carlier, Conor V. Dolan, Eco J. De Geus, Luhmann, Maike, Ulrich Schimmack, and Michael Eid. 2011.
and Dorret I. Boomsma. 2017. “Social Comparison Orientation “Stability and Variability in the Relationship between Subjective
in Monozygotic and Dizygotic Twins.” Twin Research and Well-Being and Income.” Journal of Research in Personality
Human Genetics 20(6):550–57. 45(2):186–97.
Imbens, Guido W., and Joshua D. Angrist. 1994. “Identification and Luo, Ye, and Linda J. Waite. 2005. “The Impact of Childhood
Estimation of Local Average Treatment Effects.” Econometrica and Adult SES on Physical, Mental, and Cognitive Well-
62(2):467–75. Being in Later Life.” The Journals of Gerontology. Series B,
Jacob, Marita, and Corinna Kleinert. 2014. “Marriage, Gender, and Psychological Sciences and Social Sciences 60(2):S93–101.
Class: The Effects of Partner Resources on Unemployment Marbach, Moritz, and Dominik Hangartner. 2020. “Profiling
Exit in Germany.” Social Forces 92(3):839–71. Compliers and Noncompliers for Instrumental-Variable
Jaroszewicz, Ania, Jon M. Jachimowicz, Oliver P. Hauser, and Julian Analysis.” Political Analysis 28(3):435–44.
Jamison. 2022. “How Effective Is (More) Money? Randomizing Marmot, Michael. 2002. “The Influence of Income on Health:
Unconditional Cash Transfer Amounts in the US.” SSRN. Views of an Epidemiologist.” Health Affairs 21(2):31–46.
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4154000. Martikainen, Pekka, Jerome Adda, Jane E. Ferrie, George Davey
Karney, Benjamin R. 2021. “Socioeconomic Status and Intimate Smith, and Michael Marmot. 2003. “Effects of Income and
Relationships.” Annual Review of Psychology 72:2.1–2.4. Wealth on GHQ Depression and Poor Self Rated Health in
Kessler, Ronald C. 1982. “A Disaggregation of the Relationship White Collar Women and Men in the Whitehall II Study.”
between Socioeconomic Status and Psychological Distress.” Journal of Epidemiology and Community Health 57(9):718–23.
American Sociological Review 47(6):752–64. Martin, Alicia R., Masahiro Kanai, Yoichiro Kamatani, Yukinori
Killewald, Alexandra, and Ian Lundberg. 2017. “New Evidence Okada, Benjamin M. Neale, and Mark J. Daly. 2019. “Clinical
against a Causal Marriage Wage Premium.” Demography Use of Current Polygenic Risk Scores May Exacerbate Health
54(3):1007–28. Disparities.” Nature Genetics 51(4):584–91.
King, Katherine, and Christin Ogle. 2014. “Negative Life Events McCarthy, Mark I., Gonçalo R. Abecasis, Lon R. Cardon, David
Vary by Neighborhood and Mediate the Relation between B. Goldstein, Julian Little, John P. A. Ioannidis, and Joel N.
Neighborhood Context and Psychological Well-Being.” PLoS Hirschhorn. 2008. “Genome-Wide Association Studies for
One 9(4):e93539. doi:10.1371/journal.pone.0093539. Complex Traits: Consensus, Uncertainty and Challenges.”
Kong, Augustine, Gudmar Thorleifsson, Michael L. Frigge, Bjarni Nature Reviews Genetics 9(5):356–69.
J. Vilhjalmsson, Alexander I. Young, Thorgeir E. Thorgeirsson, McGarry, Kathleen. 2016. “Dynamic Aspects of Family Transfers.”
Stefania Benonisdottir, et al. 2018. “The Nature of Nurture: Journal of Public Economics 137:1–13.
Effects of Parental Genotypes.” Science 359(6374):424–28. McLeod, Jane D. 2013. “Social Stratification and Inequality.” Pp.
Kraus, Michael W., Paul K. Piff, and Dacher Keltner. 2009. “Social 229–53 in Handbook of the Sociology of Mental Health, edited
Class, Sense of Control, and Social Explanation.” Journal of by C. S. Aneshensel, J. C. Phelan, and A. Bierman. Dordrecht,
Personality and Social Psychology 97(6):992–1004. Netherlands: Springer.
14 Socius: Sociological Research for a Dynamic World 

McLeod, Jane D., and Ronald C. Kessler. 1990. “Socioeconomic Work, and Short Rest Periods.” Social Science and Medicine
Status Differences in Vulnerability to Undesirable Life 246:112774. doi:10.1016/j.socscimed.2019.112774.
Events.” Journal of Health and Social Behavior 31(2):162–72. Schnittker, Jason. 2008. “Happiness and Success: Genes, Families,
Miech, Richard Allen, and Michael J. Shanahan. 2000. and the Psychological Effects of Socioeconomic Position and
“Socioeconomic Status and Depression over the Life Course.” Social Support.” American Journal of Sociology 114(S1):233–59.
Journal of Health and Social Behavior 41(2):162–76. Segal, Nancy L. 1988. “Cooperation, Competition, and Altruism
Muntaner, Carles, Carme Borrell, and Haejoo Chung. 2007. “Class in Human Twinships: A Sociobiological Approach.” Pp. 168–
Relations, Economic Inequality and Mental Health: Why Social 206 in Sociobiological Perspectives on Human Development,
Class Matters to the Sociology of Mental Health.” Pp. 127–41 edited by K. M. MacDonald. New York, NY: Springer.
in Mental Health, Social Mirror, edited by W. R. Avison, J. D. Segal, Nancy L., and Scott L. Hershberger. 1999. “Cooperation and
McLeod, and B. A. Pescosolido. Boston, MA: Springer. Competition between Twins.” Evolution and Human Behavior
Oswald, Andrew J., Eugenio Proto, and Daniel Sgroi. 2015. 20(1):29–51.
“Happiness and Productivity.” Journal of Labor Economics Shields-Zeeman, Laura, Daniel F. Collin, Akansha Batra, and Rita
33(4):789–822. Hamad. 2021. “How Does Income Affect Mental Health and
Paaby, Annalise B., and Matthew V. Rockman. 2013. “The Many Health Behaviours? A Quasi-experimental Study of the Earned
Faces of Pleiotropy.” Trends in Genetics 29(2):66–73. Income Tax Credit.” Journal of Epidemiology and Community
Park, Lora E., Deborah E. Ward, and Kristin Naragon-Gainey. Health 75(10):929–35.
2017. “It’s All about the Money (for Some): Consequences of Society for Adolescent Health and Medicine. 2017. “Young Adult
Financially Contingent Self-Worth.” Personality and Social Health and Well-Being: A Position Statement of the Society
Psychology Bulletin 43(5):601–22. for Adolescent Health and Medicine.” Journal of Adolescent
Patel, Vikram, Jonathan K. Burns, Monisha Dhingra, Leslie Tarver, Health 60(6):758–59.
Brandon A. Kohrt, and Crick Lund. 2018. “Income Inequality Song, Xi, Emma Zang, Kenneth C. Land, and Boyan Zheng.
and Depression: A Systematic Review and Meta-analysis of 2022. “Intergenerational Income Mobility Table Revisited: A
the Association and a Scoping Review of Mechanisms.” World Trajectory Group Perspective.” Research in Social Stratification
Psychiatry 17(1):76–89. and Mobility 80:100713. doi:10.1016/j.rssm.2022.100713.
Perreira, Krista M., Natalia Deeb-Sossa, Kathleen Mullan Harris, Sperandei, Sandro, Andrew Page, Matthew J. Spittal, and Jane
and Kathleen Mullan Harris. 2005. “What Are We Measuring ? Pirkis. 2023. “Low Education and Mental Health among Older
An Evaluation of the CES-D Across Race / Ethnicity and Adults: The Mediating Role of Employment and Income.”
Immigrant Generation.” Social Forces 83(4):1567–601. Social Psychiatry and Psychiatric Epidemiology 58(5):823–31.
Piff, Paul K., and Jake P. Moskowitz. 2018. “Wealth, Poverty, Staiger, B. Y. Douglas, and James H. Stock. 1997. “Instrumental
and Happiness: Social Class Is Differentially Associated with Variables Regression with Weak Instruments.” Econometrica
Positive Emotions.” Emotion 18(6):902–905. 65(3):557–86.
Plagnol, Anke C. 2011. “Financial Satisfaction over the Life Steptoe, Andrew. 2019. “Happiness and Health.” Annual Review of
Course: The Influence of Assets and Liabilities.” Journal of Public Health 40(1):339–59.
Economic Psychology 32(1):45–64. Stevenson, Betsey, and Justin Wolfers. 2013. “Subjective Well-
Powdthavee, Nattavudh. 2010. “How Much Does Money Being and Income: Is There Any Evidence of Satiation?”
Really Matter? Estimating the Causal Effects of Income on American Economic Review 103(3):598–604.
Happiness.” Empirical Economics 39(1):77–92. Stewart-Brown, Sarah, Preshila Chandimali Samaraweera, Frances
Pudrovska, Tetyana, and Benedicta Anikputa. 2014. “Early- Taggart, Ngianga Bakwin Kandala, and Saverio Stranges. 2015.
Life Socioeconomic Status and Mortality in Later Life: An “Socioeconomic Gradients and Mental Health: Implications for
Integration of Four Life-Course Mechanisms.” Journals of Public Health.” British Journal of Psychiatry 206(6):461–65.
Gerontology. Series B, Psychological Sciences and Social Stock, James H., and Motohiro Yogo. 2005. “Testing for Weak
Sciences 69(3):451–60. Instruments in Linear IV Regression.” Pp. 80–108 in
Radloff, Lenore S. 1977. “A Self-Report Depression Scale for Identification and Inference for Econometric Models, edited
Research in the General Population.” Applied Psychological by D. W. K. Andrews and J. H. Stock. Cambridge, UK:
Measurement 1(3):385–401. Cambridge University Press.
Raschke, Christian. 2019. “Unexpected Windfalls, Education, and Strully, Kate W. 2009. “Job Loss and Health in the U.S. Labor
Mental Health: Evidence from Lottery Winners in Germany.” Market.” Demography 46(2):221–46.
Applied Economics 51(2):207–18. Thomson, Rachel M., Erik Igelström, Amrit Kaur Purba, Michal
Ridley, Matthew, Gautam Rao, Frank Schilbach, and Vikram Patel. Shimonovich, Hilary Thomson, Gerry McCartney, Aaron
2020. “Poverty, Depression, and Anxiety: Causal Evidence and Reeves, et al. 2022. “How Do Income Changes Impact on
Mechanisms.” Science 370(6522):eaay0214. doi:10.1126/sci- Mental Health and Wellbeing for Working-Age Adults? A
ence.aay0214. Systematic Review and Meta-analysis.” The Lancet Public
Ross, Catherine E., and Joan Huber. 1985. “Hardship and Depression.” Health 7(6):e515–28.
Journal of Health and Social Behavior 26(4):312–27. Trejo, Sam, and Benjamin W. Domingue. 2018. “Genetic Nature
Ross, Catherine E., and John Mirowsky. 1992. “Households, or Genetic Nurture? Introducing Social Genetic Parameters to
Employment, and the Sense of Control.” Social Psychology Quantify Bias in Polygenic Score Analyses.” Biodemography
Quarterly 55(3):217–35. and Social Biology 64(3–4):187–215.
Sato, Kaori, Sachiko Kuroda, and Hideo Owan. 2020. “Mental Turley, Patrick, Raymond K. Walters, Omeed Maghzian, Aysu Okbay,
Health Effects of Long Work Hours, Night and Weekend James J. Lee, Mark Alan Fontana, Tuan Anh Nguyen-Viet, et al.
Rauf 15

2018. “Multi-trait Analysis of Genome-Wide Association Summary Zimmerman, Frederick J., and Wayne Katon. 2005. “Socioeconomic
Statistics Using MTAG.” Nature Genetics 50(2):229–37. Status, Depression Disparities, and Financial Strain: What
Turvey, Carolyn L., Robert B. Wallace, and Regula Herzog. 1999. Lies behind the Income-Depression Relationship?” Health
“A Revised CES-D Measure of Depressive Symptoms and a Economics 14(12):1197–215.
DSM-Based Measure of Major Depressive Episodes in the Zissimopoulos, Julie M., and James P. Smith. 2011. “Unequal
Elderly.” International Psychogeriatrics 11(2):139–48. Giving: Monetary Gifts to Children across Countries and over
Van Deurzen, Ioana, Erik Van Ingen, and Wim J. H. Van Oorschot. Time.” Pp. 289–328 in Persistence, Privilege, and Parenting:
2015. “Income Inequality and Depression: The Role of Social The Comparative Study of Intergenerational Mobility, edited
Comparisons and Coping Resources.” European Sociological by T. M. Smeeding, R. Erikson, and M. Jäntti. New York, NY:
Review 31(4):477–89. Russell Sage Foundation.
Wang, Jun, Qihui Chen, Gang Chen, Yingxiang Li, Guoshu Kong,
and Chen Zhu. 2020. “What Is Creating the Height Premium?
Author Biography
New Evidence from a Mendelian Randomization Analysis
in China.” PLoS One 15(4):13–14. doi:10.1371/journal. Tamkinat Rauf is a post-doctoral fellow and an incoming assistant
pone.0230555. professor in the Department of Sociology at the University of
Wooldridge, Jeffrey M. 2010. Econometric Analysis of Cross Wisconsin-Madison. Her research examines how social, economic,
Section and Panel Data. 2nd ed. Cambridge, MA: MIT political, and biological factors jointly influence various aspects of
Press. human well-being, including mental and physical health.

You might also like