J Fam Econ Iss

DOI 10.1007/s10834-015-9443-6

ORIGINAL PAPER

Household Debt and Adult Depressive Symptoms in the United

States
Lawrence M. Berger1 • J. Michael Collins2 • Laura Cuesta3

Ó Springer Science+Business Media New York 2015

Abstract This study used data from Waves 1 (1987–1989) married throughout the observation period. These findings
and 2 (1992–1994) of the National Survey of Families and suggest that short-term debt may have an adverse influence
Households in the United States and a series of regression on psychological wellbeing, particularly for those who are
models, some of which included individual-specific fixed less educated, approaching retirement age, or unmarried.
effects, to estimate associations of particular types and levels
of debt with adult depressive symptoms. Results suggest that Keywords Debt  Depressive symptoms  National
household debt is positively associated with greater de- Survey of Families and Households
pressive symptoms. However, this association appears to be
driven by short-term (unsecured) debt; we found little evi-
dence of associations with depressive symptoms for mid- or Introduction
long-term debt. The link between short-term debt and de-
pressive symptoms persisted with alternative estimation Household debt increased dramatically in the United States
strategies, including defining debt in absolute and relative during the last four decades. Between 1962 and 2008, the
terms. Furthermore, this association was particularly con- median ratio of household debt to income rose from 0.1 to
centrated among 51–64 year-old adults, those with a high 0.6 and aggregate household debt rose from about 60 % to
school education or less, and those who were not stably about 120 % of aggregate household income (Dynan 2009).
Much of this trend was spurred by increased homeownership
financed by mortgages, but it also reflects increased accu-
& Laura Cuesta mulation of unsecured, revolving credit card debt (Durkin
laura.cuesta@gmail.com 2000; Xiao and Yao 2011a, 2011b). Household debt levels
Lawrence M. Berger have declined since 2008, as credit has become less avail-
lmberger@wisc.edu able, households have reduced spending, and debt has been
J. Michael Collins forgiven through bankruptcy filings, but it remains at his-
jmcollins@wisc.edu torically high levels and is subject to much discussion by
1
Institute for Research on Poverty, School of Social Work,
policymakers (US Senate 2011).
University of Wisconsin-Madison, 3420 William H. Sewell Whereas the ability to borrow has clear benefits—al-
Social Sciences Building, 1180 Observatory Drive, Madison, lowing individuals and households to smooth consumption
WI 53706, USA and invest in homeownership, human capital acquisition,
2
Center for Financial Security, La Follette School of Public and other large ticket items that they cannot fully pay for in
Affairs and The School of Human Ecology, University of the present (Dynan and Kohn 2007; Hyman 2011)—it may
Wisconsin-Madison, 4208 Nancy Nicholas Hall, 1300 Linden
also result in increased financial pressure given that debt
Drive, Madison, WI 53706, USA
3
must eventually be repaid. To date, the small body of re-
Institute for Research on Poverty, School of Social Work,
search that has specifically focused on household debt has
University of Wisconsin-Madison, 3415 William H. Sewell
Social Sciences Building, 1180 Observatory Drive, Madison, identified adverse associations with financial and other
WI 53706-1320, USA stress (Norvilitis and MacLean 2010; Watson et al. 2014;

123
 J Fam Econ Iss

Worthington 2006), adult physical health (Drentea and by rapid expansion of (particularly short-term, unsecured)
Lavrakas 2000; Keese and Schmitz 2014; Lenton and debt. We estimated associations between debt and depres-
Mosley 2008), medical care (Kalousova and Burgard 2013, sive symptoms through a series of standard ordinary least
2014), college completion (Dwyer et al. 2012), and marital squares (OLS) regressions with extensive controls and OLS
quality (Dew 2007, 2008). Yet, findings regarding asso- regressions with individual-specific fixed effects. We ex-
ciations between debt and psychological wellbeing have amined these associations in terms of whether a household
not been completely consistent. Studies have found debt to had accrued any debt as well as the amount of debt it had
be associated with poorer psychological functioning accrued, focusing not only on total debt, but also on par-
(Brown et al. 2005), increased probability of mental dis- ticular types (short-, mid-, and long-term) of debt. Addi-
order (Jenkins et al. 2008), and higher levels of anger tionally, we tested the robustness of our results when debt
(Drentea and Reynolds 2012), anxiety (Drentea 2000; was modeled in terms of absolute level and when it was
Drentea and Reynolds 2012), and depression (Bridges and modeled as a proportion of annual income and total assets.
Disney 2010; Drentea and Reynolds 2012; Gathergood Finally, we conducted subgroup analyses to examine whe-
2012). At the same time, however, Dew (2007) has found ther associations of particular types and amounts of debt
debt to be associated with less depression among married with adult depressive symptoms varied by age, educational
couples, and Dwyer et al. (2011) have found positive as- attainment, and whether an individual was stably married
sociations between debt and both self-esteem and mastery. throughout the observation period. We found that accu-
Much remains unknown about the nature of the asso- mulation of short-term debt was associated with greater
ciations between debt and mental health. First, as noted depressive symptoms. Furthermore, this association was
above, results from prior studies are not conclusive. Se- particularly concentrated among 51–64 year-old adults,
cond, social selection and reverse causality pose formid- those with a high school education or less, and those who
able challenges to estimating unbiased associations. With were not stably married throughout the observation period.
respect to social selection, other individual and household These findings suggest that short-term debt may have an
characteristics may influence both debt accumulation and adverse influence on psychological wellbeing, particularly
psychological wellbeing such that correlations between the for those who are less educated, approaching retirement
two are spurious. For example, factors such as unemploy- age, or unmarried. Financial professionals, educators and
ment or limited access to economic resources may both counselors should be aware of the coincidence of depres-
induce one to borrow, and also adversely influence one’s sion and high levels of short-term debt, particularly for less-
psychological wellbeing. It is also possible that psycho- advantaged adults and those approaching retirement age.
logical wellbeing influences debt accumulation as much as,
or more than, debt accumulation influences psychological
wellbeing. Most existing analyses have been unable to Conceptual and Theoretical Framework
adequately adjust for these possibilities.
In addition, prior work has not fully considered potential Social Stress Theory (Pearlin 1989) and the Family Stress
heterogeneity in associations between debt and psycho- Model (Conger and Elder 1994; Conger et al. 1990)—
logical wellbeing, even though links between debt and hereafter ‘‘stress theory’’— suggest that, whereas debt may
psychological wellbeing may differ depending on whether help alleviate economic stress in the short-term (Dwyer
debt was accumulated for investment in education or et al. 2011), over the long-term, debt burdens may lead to
homeownership versus immediate consumption, the debt- increased economic stress and, thereby, decreased psy-
or’s age or proximity to retirement, and the debtor’s in- chological wellbeing. Specifically, borrowing provides
come or asset levels. Particular types of debt may be expanded opportunities to purchase goods and services that
differentially associated with psychological wellbeing could not otherwise be purchased. As such, debt may be
based on common factors that determine both economic positively associated with wellbeing by allowing indi-
status and mental health. viduals and households to maintain or increase consump-
This study adds to a small but growing literature on as- tion as well as to make long-term investments. At the same
sociations between debt and psychological wellbeing. We time, however, debt burden may be inversely associated
used data on roughly 8500 individuals, who were inter- with (particularly longer-term) wellbeing both directly,
viewed between 1987 and 1989 and again between 1992 because resources must be allocated to debt repayment, and
and 1994 as part of the National Survey of Families and indirectly, as a result of increased financial or other stress.
Households (NSFH), to examine associations of particular For example, if high levels of debt require that a significant
types and amounts of debt with adult depressive symptoms. portion of income be allocated to debt repayment, then the
The NSFH included a nationally representative sample of potential benefits of borrowing may be offset by financial
US households. It was fielded during a period characterized pressure or distress (Conger and Elder 1994; Conger et al.

123
J Fam Econ Iss

1990), which may precipitate declines in psychological borrow, and the types of debt one is likely qualify for and/
wellbeing. or utilize. Such factors may include SES, life stage, and
Three aspects of debt are particularly important to un- marital status. In regards to SES, stress theory predicts that
derstand its association with depressive symptoms: (1) households with limited resources are more vulnerable to
agency in borrowing (the degree to which one’s economic potentially stress-inducing factors—such as debt burden—
choice set is constrained), (2) magnitude (the amount of than those with greater access to resources (Pearlin 1989).
debt, often relative to income or assets), and (3), cost (total In addition, debt incurred in response to traumatic events is
charges and fees incurred during the full period over which thought to be particularly stress-inducing (McCloud and
debt is repaid). Cost is closely linked to the default risk of Dwyer 2011; Sullivan et al. 2000). Disadvantaged house-
each type of debt. Most notably, unsecured (short-term) holds are especially likely to incur debt under such cir-
debt is more expensive than secured (long-term) debt since cumstances. Indeed, they tend to incur debt with less
the lender has no collateral and default rates are relatively agency, of greater relative (though not necessarily abso-
high. Cost and magnitude are also linked. Long-term debt, lute) magnitude, and at a higher cost. Most low-SES
such as a home or education loan, often involves a large households lack the liquid assets necessary to support
principal amount, but is amortized such that some of the consumption at the federal poverty level for 3 months
initial principal is repaid each month. These loans also without income (McKernan and Ratcliffe 2009). Research
have lower annual interest rates and are frequently viewed further suggests that having insufficient funds to meet basic
as an investment by the debtor. In contrast, short-term needs encourages borrowing, even at high-cost (Shah et al.
unsecured debt costs considerably more than long-term 2012). That disadvantaged households tend to use debt to
debt, particularly if not paid off quickly and balances re- meet basic needs, whereas affluent households do so as a
volve such that little of the initial balance is paid off. convenience or investment strategy (Dwyer et al. 2011,
Existing studies have predominantly focused either on 2012; Sullivan et al. 2000), implies that, on average,
total debt or on a specific type of debt (home, education, auto, wealthier households exercise greater agency in borrowing.
or unsecured). Few have simultaneously considered the full This is also consistent with studies that have shown cor-
range of types and amounts of debt that households may relations between self-efficacy and overall levels of wealth
accumulate. This is problematic for several reasons. First, (e.g., Lown et al. 2014).
different types of debt have different associated costs (in- The influence of debt on psychological wellbeing may
terest rates and fees), which affect the amount that can be vary by life stage as well. Lifecycle theory (Attanasio and
used for consumption or investment. Notably, (typically Weber 2010; Modigliani 1986; Modigliani and Brumberg
low-cost and amortized) long-term debt is generally used for 1954; Shefrin and Thaler 1988) suggests that individuals
asset or human capital investment, which may be positively and families borrow to meet consumption needs at par-
associated with psychological wellbeing. In contrast, (higher ticular life stages and save (or repay debt) at others. For
cost and revolving) unsecured debt is more often used for example, education debt tends to accumulate in early-
immediate consumption and may be negatively associated adulthood and home loans in mid-adulthood. These large
with psychological wellbeing, particularly over time or if but low cost loans are typically paid off over a relatively
incurred in a context of limited agency. Second, there is long period, but are not expected to persist into older-
likely to be social selection into particular types of debt based adulthood; in reality, however, they often do (Mann 2011;
on factors such as socioeconomic status (SES) and associated Federal Reserve Bank of New York 2012). By contrast,
financial literacy (ability to fully comprehend the implica- unsecured debt may occur throughout the life course,
tions of loan terms). This may reflect borrower decisions vis- sometimes in a context of limited agency, and it may be
a-vis types of debt to pursue as well as a household’s ability most problematic if held in mid- to late-adulthood when
to qualify for particular types of loans (Lusardi and Tufano income streams attenuate (Yilmazer and Devaney 2005;
2009). Third, specific types of debt may be fungible (e.g., Xiao and Yao 2011b). As such, the influence of debt on
home equity loans may be taken to repay unsecured debt; psychological wellbeing may vary by the life course stage
larger education loans may substitute for unsecured debt in at which an individual experiences debt. Specifically, there
order to cover living expenses during schooling) and indi- are likely to be stronger associations between debt and
viduals may substitute or move between various types of debt psychological wellbeing among older working-age adults
for the same purposes (consumption, investment) given than among younger ones. Yet, despite considerable re-
differences in costs (interest rates and fees). Thus, it is im- search that has linked debt accumulation and repayment to
portant to account for the full range of types and amounts of life stages (Azicorbe et al. 2003; Baek and Hong 2004;
debt affecting a household at any point in time. Brown and Taylor 2008; Xiao and Yao 2011b), little has
Individual and household characteristics may influence focused on how associations between debt and wellbeing
one’s ability to qualify for credit, one’s propensity to may vary depending on the life stage at which they are

123
 J Fam Econ Iss

experienced, nor by the type of debt held during a given account for earlier depression status or changes in debt
life stage. between observation points.
Family structure may also influence debt levels. Whereas Using two waves of data from the NSFH (the data
married households and households with children have source used in the current study), Dew (2007) examined
greater average amounts of debt (Baek and Hong 2004; associations of short-term (unsecured or consumer) debt
Yilmazer and Devaney 2005) than other households, co- with marital conflict using a sample of consistently married
habiting and single-parent households are disproportion- individuals and structural equation modeling. Most rele-
ately likely to experience debt-related hardship and vant to this study, he found that, controlling for assets and a
delinquency (Xiao and Yao 2011a).These hardships may composite measure of ‘‘economic pressure,’’ constructed
exacerbate adverse outcomes associated with social disad- from measures of frequency of worrying about paying bills
vantage, including compromised psychological wellbeing. and satisfaction with finances, short-term debt was asso-
ciated with fewer later depressive symptoms. However, it
is important to note that the analyses estimated only be-
Literature Review tween-individual differences in short-term debt levels at
baseline (Wave 1) and between-individual levels of de-
Only a handful of studies have examined associations be- pressive symptoms at follow-up (Wave 2). Changes in in-
tween debt and various forms of psychological wellbeing. dividuals’ own debt levels were not used to estimate
Whereas most have utilized longitudinal data, few have changes in depressive symptoms over time. Notably, the
employed rigorous methods for dealing with social selec- study focused only on short-term debt rather than all types
tion and reverse causality. Of the 11 studies of which we of debt that an individual may have held.
are aware, 6 used UK data, 4 used US data, and 1 used Dwyer et al. (2011) used longitudinal data from the
German data. Two studies used cross-sectional data. 1997 National Longitudinal Survey of Youth in the United
Drentea (2000) used a sample of adults in Ohio from 1997 States and found positive associations of both credit card
and found links between credit card debt and anxiety, with and education debt with both self-esteem and mastery
stress playing a major mediating role in these associations. among young adults after adjusting for earlier measures of
Furthermore, she found higher levels of anxiety among these outcomes. The authors posited that the direction of
younger adults, which she concluded reflected higher levels these associations suggested that there may be psycho-
of stress about credit card debt for these individuals relative logical benefits to borrowing for young adults who accu-
to their older counterparts. Jenkins et al. (2008) analyzed mulate debt as a form of human capital investment.
UK data and found a positive association between the Furthermore, these associations were concentrated among
number of debts held in the past year and the probability of lower-SES and younger young adults, implying that debt
a mental disorder. However, the possibility that these may have a more positive influence for less advantaged
findings were driven by social selection and/or reverse young adults who are investing in their future than for their
causality cannot be ruled out given that both studies relied more advantaged counterparts, as well as that such in-
on point-in-time data. vestments may have a more positive influence on psycho-
The nine additional studies of which we are aware all logical wellbeing when made at an earlier rather than later
utilized longitudinal data. Yet, there is considerable var- period of young adulthood.
iation in the methods employed to address social selection The most rigorous studies to date have employed em-
and reverse causality. Despite having longitudinal data, pirical strategies that have taken full advantage of longi-
several studies did not analyze repeated measures of the tudinal data to estimate within-individual associations
predictors and outcomes. For example, Drentea and Rey- between changes in debt and changes in psychological
nolds (2012) used two waves of data on a sample of US wellbeing over time. Three prior studies exemplify this
adults in the Miami area. They found that having any debt strategy. Keese and Schmitz (2014) used German panel
was associated with higher levels of depression, anxiety, data and both lagged dependent variable and fixed-effects
and anger. However, both debt and the outcome variables regressions and found inverse associations of consumer and
were measured at the same point in time, the follow-up housing debt with both physical health satisfaction and
interview, with only the control variables measured at overall mental health. Lenton and Mosley (2008) used UK
baseline. Thus, the causal direction of association between data and a simultaneous-equation generalized probit model
debt and psychological functioning could not be deter- to estimate relationships between debt burden (size) and
mined. Reading and Reynolds (2001) used two waves of structure (high, medium, or low interest rate) with physical
data, measured 6 months apart. They found that worrying and psychological health. They found evidence that rela-
about debt at baseline was associated with postnatal de- tionships between these factors were bi-directional, and
pression at follow-up. However, their models did not also that repayment structure mattered, such that higher

123
J Fam Econ Iss

interest (unsecured) debt was more strongly inversely It is also notable that the 1987–1994 period pre-dated
linked with health than was lower interest debt. Bridges bankruptcy reform in the United States, which shifted
and Disney (2010) used data from a longitudinal household borrowing behavior in the 2000s (Bird et al. 1999). It also
survey in the UK and fixed-effects regressions to estimate pre-dated the rise of subprime mortgage lending (Chom-
associations between debt problems (difficulty repaying sisengphet and Pennington-Cross 2006). Avoiding these
debt) and the probability of depression. They found that confounding periods is a benefit of using the NSFH.
individuals who experienced the onset of debt problems Second, we exploited the longitudinal nature of the
also exhibited an increased probability of depression. NSFH to estimate a series of OLS models with individual-
Two additional studies used instrumental variables specific fixed effects. This strategy adjusted for time-in-
methods to leverage variation in debt accumulation which variant unobserved factors with persistent effects. It was
was determined only by factors that did not directly influence therefore a more rigorous identification strategy than has
psychological wellbeing (i.e., was exogenously determined). previously been employed with US data.
This strategy has the potential to account for social selection Third, we included in our models the three major cate-
and reverse causality when estimating associations between gories of debt held by households: short-term (unsecured),
debt and psychological wellbeing and thereby produce un- mid-term (non-mortgage bank, installment, auto, and per-
biased estimates that justify causal interpretation. Brown sonal), and long-term (home mortgage and education) debt.
et al. (2005), for example, used British Household Panel We also modeled debt in three ways: as a dichotomous
Survey data to estimate effects using a variety of instru- indicator of whether a household had debt and as con-
mental variables. They found that outstanding non-mortgage tinuous measures of absolute and relative (to income and
debt was associated with poorer overall psychological assets) amounts of debt. For each measure, we estimated
wellbeing across multiple model specifications; this was not models focusing on total debt as well as models focusing
true for mortgage debt. However, their instrument for pre- on particular types of debt. Finally, we tested whether as-
dicting debt was whether an individual had a credit card. This sociations between debt and depressive symptoms varied
is unlikely to have been a valid instrument as it is not ex- by age, which is a proxy for life stage, by educational
ogenous to debt accumulation. Finally, Gathergood (2012) attainment, which is an important indicator of SES, and by
used UK panel data and variation in local housing prices to marital stability which, in addition to being a measure of
isolate the exogenous component of what he called ‘‘problem SES, is related to the ability of households to have stable or
mortgage debt.’’ He found such debt to be associated with multiple incomes (and the focus of at least one prior study).
decreased general psychological wellbeing and anxiety-re- Our analyses tested the hypothesis that debt burden
lated illness. However, the ‘‘problem debt’’ measure asses- would be positively associated with adult depressive
sed difficulty paying housing costs and whether consumer symptoms over an approximately 6-year time period. In
credit repayment was a ‘‘heavy burden,’’ both of which may light of prior theory and evidence, we expected that short-
reflect (perceived) material hardship or limited income term debt, which tends to be of high cost and is often
rather than actual debt burden. incurred for immediate consumption and with limited
The current study extends prior literature in three ways. agency, would be associated with increased depressive
First, we utilized a large, nationally representative US symptoms. Our a priori expectations for medium- and long-
dataset in which various types and amounts of debt as well term debt were ambiguous. Furthermore, we expected the
as depressive symptoms were measured at two time points. magnitude of (particularly short-term) debt to be positively
These data were collected during a period in which unse- associated with depressive symptoms. Although we as-
cured debt expanded rapidly among US households. They sumed consumption and economic pressure are the primary
allowed us to estimate associations of changes in debt and mechanisms through which these associations would op-
changes in depressive symptoms across an approximately erate, we estimated only the direct effects of debt on de-
6-year interval. Although this time period was necessitated pressive symptoms, rather than examining potential
by the NSFH data, which were collected between 1987 and mediators. We took this approach for two reasons. First, we
1989 (Wave 1) and 1992 and 1994 (Wave 2), it is a relevant focused on estimating the full association of debt with
period over which to study these associations. For example, depressive symptoms, rather than parsing out the portion of
personal bankruptcy in the United States remains in credit this association which was explained through each of these
bureau and other records for 7–10 years. It also takes mechanisms. That is, we were ultimately interested in
debtors several years to recover from a default and return to whether there is likely to be a causal link between debt and
asset and debt levels experienced before a negative shock depressive symptoms. Second, our data did not include
(Jagtiani and Li 2013). Overall, 6 years is thus an appro- high quality measures of consumption or economic stress,
priate time frame over which to consider substantial chan- making adequate measurement of these mediators
ges in debt and their influences on psychological wellbeing. problematic.

123
 J Fam Econ Iss

Data and Measures each symptom on a 0- to 7-point scale. Following the

standard protocol described in Radolff (1977), individuals
Sample were assigned 0 points for each symptom that they did not
experience during the past week, 1 point for each symptom
Our analyses used data from the first two waves of the experienced for 1–2 days, 2 points for each symptom ex-
NSFH, a longitudinal, nationally representative household perienced for 3–4 days, and 3 points for each symptom
survey that started with a sample of 13,007 respondents in experienced for 5–7 days. An individual’s overall score
Wave 1 (1987–1989), and re-interviewed 10,005 of these (0–36 points) for the scale was then constructed by sum-
respondents in Wave 2 (1992–1994). Households were ming the scores on the individual items.
randomly selected from 1700 selection units that resulted As noted above, prior literature has been inconsistent
from selecting 17 enumeration districts within each of the with regard to how best to model household debt. As such,
100 primary sampling units; the interview response rate we modeled debt in several ways, which allowed us to test
was 74 % (for further detail, see Sweet et al. 1988; Sweet the robustness of our findings to multiple specifications of
and Bumpass 1996). The primary respondent was a person our key predictor. The first specification consisted of a
living in the household who was randomly selected using a simple dichotomous variable indicating (1 = yes) that a
Kish grid; this person was asked to provide detailed in- household had any debt, defined to include credit card debt,
formation about his or her family of origin; current family installment loans, bank loans, loans from friends, (overdue)
structure, composition, and living arrangements; fertility, bills owed for more than 2 months, vehicle debt, home
education and employment history; earnings and income; improvement loans, education debt, and mortgage debt.
health and mental health; and household assets and debt. The second measure consisted of the logarithm of total
household debt, assessed by summing the amounts of each
of these types of debt.
Missing Data
In addition, we estimated associations of specific types
of debt with adult depressive symptoms. Here, we con-
We utilized multiple imputation techniques to impute
sidered three types of debt: short-term, mid-term, and long-
values for all variables with missing data for the full initial
term. For each, we constructed both a dichotomous mea-
NSFH sample (N = 13,007). Specifically, we imputed 25
sure of whether the household had any debt of that type and
complete datasets using Stata’s MI program (StataCorp
also a continuous measure of the (logarithm of the) total
2013). We then limited our sample to observations of in-
amount of that type of debt. Short-term debt included credit
dividuals who were between 21 and 65 years of age in both
card debt and overdue bills (bills owed for more than
survey waves and who did not own their own business in
2 months). The variable was not sensitive to the inclusion
either wave. We excluded individuals younger than 21 and
of unpaid bills. However, because juggling due dates is a
older than 65 to focus on working-age adults. We excluded
common strategy for households to extend liquidity, we
business owners in order to ensure that our debt measures
determined it was important to include overdue bills as a
reflected personal, rather than business, debt. These criteria
form of short-term debt (the costs of which are a late fee or
resulted in the exclusion of just over a third of the initial
other penalty). Mid-term debt consisted of installment
sample. Our final analysis consisted of 16,964–17,059 in-
loans, home improvement loans, vehicle loans, loans from
dividual-wave observations (two observations per respon-
friends, and other bank loans. Long-term debt consisted of
dent) of 8457–8516 individuals per dataset, across the 25
education and mortgage debt. Finally, in order to test the
imputed datasets.
robustness of our results using absolute measures of
amounts of debt to those achieved when relative debt
Measures measures were employed, we also constructed a debt-to-
annual income ratio and a debt-to-assets ratio, for total debt
Our outcome of interest was adult depressive symptoms as well as for each type of debt (short-, mid- and long-
which were measured by the Center for Epidemiologic term).
Studies Depression Scale (CES-D) partial scale (Radolff Each of the debt measures was self-reported by the re-
1977). The CES-D partial scale is a truncated version of the spondent. Specifically, respondents were asked whether
full CES-D, a self-reported measure of depressive symp- they had each type of debt and, if they responded affir-
toms that was designed for research in the general matively, how much they owed. The accuracy of self-re-
population. At each NSFH interview, respondents were ported debt data is a serious concern. Evidence suggests
asked to report the frequency with which they experienced that borrower and lender reports are extremely similar for
a series of 12 depressive symptoms. Individuals were asked all forms of debt except credit card debt, for which
to report how many days in the last week they experienced borrowers report considerably less than lenders. This

123
J Fam Econ Iss

under-reporting likely reflects that individuals are often neither debt, psychological wellbeing, nor any associations
uniformed about their debt (Brown et al. 2011). If such between the two could have influenced these (fixed) char-
under-reporting in our data was systematic in ways that are acteristics. Furthermore, these baseline characteristics
associated with depressive symptoms, it will have biased could not have been determined by the same factors or
our estimates. It is less worrisome for our general conclu- processes that determined debt or psychological wellbeing.
sions if respondents’ self reports of their debt placed them As discussed below, because these characteristics are time-
in the same relative position in the debt distribution (de- invariant (fixed) they cannot be directly modeled in fixed-
spite potential underestimation of their absolute amount of effects regressions, which estimate average associations
debt) as would be the case if actual (administrative) debt between changes in predictors and changes in outcomes for
data were used. There are no publicly available data linking individuals. Although parameters are not directly estimated
credit reports to self-reported debt at the micro level. Thus, for these measures, the estimates produced by fixed-effects
all existing studies of debt and wellbeing must rely on self- regression are adjusted for these (and all other observed or
report data. However, Brown et al. 2011 compared aggre- unobserved) time-invariant individual characteristics.
gate debt levels based on self-reported and administrative We included the following time-varying socioeconomic
data and found that time trends in consumer debt are factors as controls in the standard OLS and fixed-effects
similar (though at different absolute levels) across data regressions: respondent age; respondent marital status
sources, and that age and regional patterns are also similar. (indicators for currently married and never married, with
At the same time, they also found that single-person currently single as the reference category); measures of the
households tend to report debt more accurately than larger proportion of household members age 6–17, 18–24, 25–44,
households (which may reflect better information among 45–65, and greater than 65 (with the proportion of house-
the former). Although we cannot rule out that our estimates hold members under age 6 as the reference category); re-
were biased by systematic underreporting, we were further spondent education (less than a high school education and
encouraged that Brown et al. found no evidence that the high school degree or GED, with greater than a high school
standard correction—universally multiplying credit card degree as the reference category); the logarithm of total
debt by a common factor (constant), which assumes that household income; whether the respondent was working at
underreporting is consistent across groups and over time, is the time of the interview; total household assets (the sum of
inappropriate in empirical analyses. Unfortunately, poten- the self-reported value of all real estate holdings, vehicles,
tial under-reporting of (particularly short-term) debt was a investments, and savings); and whether the respondent re-
necessary limitation of our analyses. ported being in excellent, good, or fair health (versus poor
In our standard OLS (but not fixed-effects) regressions, health). It is important to note that, with the exception of
we included as control variables time-invariant baseline age, these factors may be jointly determined with debt and
measures of whether the respondent was male, the re- psychological wellbeing. Again, as discussed further be-
spondent’s race/ethnicity (black, Hispanic, and other race/ low, by estimating associations between changes in the
ethnicity, with white as the reference category), and indi- predictors and changes in the outcome, the fixed-effects
cators for the highest level of educational attainment by the regressions functioned to reduce potential bias that may
respondent’s most educated parent (less than a high school result from such jointly determined—or endogenous—
education and high school degree or GED, with greater relationships.
than a high school degree as the reference category). Finally, we included an indicator for wave of observa-
Ideally, we would have allowed the respondent’s parents’ tion (a wave fixed effect) in all of the models. All dollar
educational attainment to vary over time; unfortunately, amounts were converted to constant, year 2010 dollars.
however, this information was collected only in Wave 1
(and not in Wave 2) of the NSFH. Given that respondents
in our sample were 21–65 years old at the first (and sub- Empirical Strategy
sequent) observation, it is unlikely that a large share of
their parents increased their educational attainment over The primary analytic challenge in estimating unbiased as-
the observation period. Thus, this limitation is unlikely to sociations between debt and depressive symptoms was
have had much of an influence on our results. With the reducing bias due to social selection and ruling out reverse
possible exception of the respondent’s parents’ educational causality. With regard to social selection, we were par-
attainment, these variables are independent of (could not ticularly concerned that individual and household charac-
have been jointly determined with) debt and psychological teristics that we could not observe in our data were
wellbeing. In other words, whereas each of the covariates correlated with both debt accumulation and depressive
may have influenced subsequent debt accumulation, psy- symptoms, such that they would obscure the true relation
chological wellbeing, or any associations between the two, between them. We were also concerned that greater

123
 J Fam Econ Iss

depressive symptoms may have driven debt accumulation, DEPit  DEPi ¼ kt  k þ qðDEBTit  DEBTi Þ
rather than vice versa. ð3Þ
þ ðXit  Xi Þb þ ðeit ei Þ
We relied on two empirical estimation strategies to at-
tempt to adjust for social selection. First, we estimated a in which the value of each variable at a given time point was
series of OLS regressions with extensive controls. These modeled as the deviation between an individual’s actual
models allowed us to examine the extent to which the value at that time point and the individual’s average value
unadjusted (bivariate) associations between debt and de- on that variable across both observation points. Differenc-
pressive symptoms were driven by differences in the ing each variable from its mean across the time periods
characteristics and circumstances of debtors and non- effectively removed the individual-specific fixed effect
debtors. The OLS models took the form: from the model. It thereby served to adjust for all observed
DEPit ¼ b0 þ qDEBTit þ bXit þ eit ð1Þ and unobserved time-invariant characteristics of the indi-
vidual and, thus, to eliminate bias due to such factors from
where DEPit was the depressive symptoms measure for the estimated association between debt and depressive
individual i at time t; DEBTit was one or more debt mea- symptoms. However, the estimate was still subject to bias
sures; Xit was a vector of covariates (including a wave due to omitted time-varying characteristics or characteris-
fixed effect); and eit was a disturbance term. We estimated tics that have time-varying associations with the outcome.
three specifications of this model. The first included only In contrast to the standard OLS regressions, the fixed-
the debt measure(s) and an indicator for wave of obser- effects regression estimates reflect intra-individual change
vation as predictors. We added the time-invariant baseline for individuals observed with and without (a particular type
characteristics, including gender, race/ethnicity, and par- of) debt (or with different amounts of debt). As such, they
ents’ educational attainment, to the second specification, as were interpreted as the average difference in depressive
well as a time-varying (but exogenous) measure of re- symptoms when the same individual was observed with
spondent age. In the final specification, we added the time- debt and without debt (or with different amounts of debt).
varying (and potentially endogenous) socioeconomic fac- We estimated two specifications—controlling only for
tors including, family structure and composition, educa- wave of observation and controlling for the full set of time-
tional attainment, income, work status, assets, and health varying socioeconomic factors—of this model.
status. This strategy allowed us to assess the extent to
which the associations of interest varied with the inclusion
of more extensive controls. The standard errors in these Results
models were adjusted for intra-cluster correlation due to
multiple observations of the same individuals. Descriptive Statistics
The coefficients produced by the standard OLS models
were interpreted as average differences in depressive Table 1 presents descriptive statistics for the full sample
symptoms between individuals who had (a particular type and by whether the respondent had any debt. The raw data
of) debt and those who did not (or between those who had (averaged across the two time periods) revealed that those
greater and lesser amounts of debt in the models in which with no debt had a higher mean number of depressive
debt amount was the key predictor). Despite the inclusion symptoms than those with debt. Overall, 79 % of house-
of controls, however, the OLS estimates were subject to holds had some debt, with 49, 53, and 48 % reporting
bias due to unmeasured factors associated with both debt short-, mid-, and long-term debt, respectively. Among
and depressive symptoms. In addition, because debt and households that had debt, 62 % had short-term debt, 67 %
depressive symptoms were measured at the same point in had mid-term debt, and 60 % had long-term debt. On av-
time, these estimates did not account for the possibility of erage, households had about $42,000 of total debt, con-
reverse causality. For these reasons, we also estimated two- sisting of about $2000 of short-term, $6500 of mid-term,
period fixed-effects models, which further adjusted for and $34,000 of long-term debt (measured in constant 2010
time-invariant unobservable characteristics and better ac- dollars). Among those households that had any debt, these
counted for the possibility of reverse causality. The fixed- figures were approximately $53,500, $3000, $8000, and
effects models took the form: $43,000. Long-term debt by far accounted for the largest
DEPit ¼ ai þ kt þ qDEBTit þ bXit þ eit ð2Þ portion of total debt in our data, followed by mid- and
short-term debt. These patterns are mirrored in the Survey
where ai was an individual-specific fixed effect, which was of Consumer Finances (SCF)—a survey widely considered
comprised of all time-invariant (observed or unobserved) as the most reliable benchmark for measures of household
characteristics of individual i, and kt was the wave fixed net worth (Kennickell and Shack-Marquez 1992). As a
effect. Equation (2) was reduced to: comparison, for example, the 1989 SCF by the Federal

123
J Fam Econ Iss

Table 1 Descriptive statistics, full sample and by any debt status

Full sample No debt Any debt

CES-D depression (0–36 points) 9.195 (0.072) 9.993 (0.176) 8.982 (0.077)***
Debt
Any debt 0.790
Any short-term debt 0.492 0.624
Any mid-term debt 0.530 0.671
Any long-term debt 0.476 0.603
Total debt amount $42,257.28 (552.202) $53,509.53 (661.393)
Total short-term debt $2,030.22 (49.753) $2,570.82 (61.665)
Total mid-term debt $6,470.38 (110.994) $8,193.34 (138.221)
Total long-term debt $33,756.69 (517.499) $42,745.38 (628.050)
Total debt/annual income 1.690 (1.018) 2.140 (1.291)
Short-term debt/annual income 0.168 (0.102) 0.213 (0.129)
Mid-term debt/annual income 0.218 (0.122) 0.277 (0.155)
Long-term debt/annual income 1.303 (0.917) 1.651 (1.163)
Total debt/assets 1.541 (2.033) 1.951 (2.573)
Short-term debt/assets 0.326 (0.121) 0.413 (0.153)
Mid-term debt/assets 0.366 (0.410) 0.463 (0.519)
Long-term debt/assets 0.849 (1.745) 1.075 (2.209)
Baseline (exogenous and time-invariant) characteristics
Male 0.397 0.370 0.405**
White 0.674 0.538 0.711***
Black 0.213 0.294 0.191***
Hispanic 0.096 0.152 0.082***
Other race/ethnicity 0.016 0.017 0.016
Parents less than high school 0.409 0.571 0.366***
Parents high school/GED 0.347 0.271 0.368***
Parents greater than HS/GED 0.243 0.158 0.266***
Time-variant socioeconomic factors
Age 39.442 (0.087) 41.596 (0.228) 38.869 (0.094)***
Married 0.544 0.337 0.599***
Never married 0.178 0.279 0.151***
Single (divorced, separated, widowed) 0.278 0.384 0.250***
Proportion of persons less than age 6 0.092 (0.001) 0.084 (0.003) 0.094 (0.002)**
Proportion of persons age 6–17 0.178 (0.002) 0.162 (0.004) 0.183 (0.002)***
Proportion of persons age 18–24 0.064 (0.001) 0.069 (0.003) 0.063 (0.001)*
Proportion of persons age 25–44 0.397 (0.003) 0.319 (0.006) 0.418 (0.003)***
Proportion of persons age 45–65 0.248 (0.003) 0.327 (0.007) 0.227 (0.003)***
Proportion of persons greater than 65 0.020 (0.001) 0.035 (0.002) 0.016 (0.001)***
Less than high school 0.193 0.358 0.149***
High school/GED 0.383 0.389 0.381
Greater than HS/GED 0.425 0.253 0.470***
Logarithm of household’s total income 10.615 (0.010) 10.072 (0.029) 10.759 (0.011)***
Currently working 0.701 0.491 0.757***
Total assets ($10,000s) 13.137 (0.164) 7.042 (0.279) 14.760 (0.194)***
Excellent, good or fair health status 0.946 0.913 0.955***
Wave observed 0.501 0.555 0.486***
Proportion of sample 0.790 0.210

16,964–17,059 individual-wave observations of 8457–8516 individuals per dataset, across 25 imputed datasets. Means (and standard errors) or
proportions presented. Statistical significance of bivariate tests for mean difference between those with any debt and those with no debt:
* p \ 0.05, ** p \ 0.01, *** p \ 0.001

123
 J Fam Econ Iss

Reserve reported total debts of $61,000, of which $3200 associated with fewer depressive symptoms. This asso-
was short term credit card debt. The levels are larger in the ciation remained negative, but became non-significant in
SCF than NSFH, but the ratios of short-term debt to total Model 1, in which we controlled for age, sex, and race/
debt are similar (4.7 and 5.0 % respectively). The 1992 ethnicity, and the highest educational attainment of the
SCF reported about $65,000 in total debt and $3500 in respondent’s parents. However, in Model 2, in which we
credit card debt. The periods are not synchronized but are further controlled for family structure and composition,
relatively close. Average total debt to annual household education, income, work status, assets, and health status,
income and assets ratios were 1.7 and 1.5 (2.1 and 2.0 this association became large and statistically significant.
among those with debt). Thus, after controlling for these time-varying socioeco-
Our raw data also revealed considerable differences in nomic factors we found debt to be associated with greater
the characteristics of those who had debt and those who did depressive symptoms. This likely reflects that those who
not. Those who had debt were younger, more likely to be have debt tend to be more socioeconomically advantaged
male, and less likely to be black or Hispanic than those than those with no debt and that socioeconomic advantage
who did not. The former had more highly educated parents, is inversely associated with depressive symptoms.
were more highly educated themselves, and were more The Model 2 estimate was confirmed by the fixed-ef-
likely to be married and working. On average, they also fects results, both with and without controls. The estimate
had greater income and assets and were in better health from Model 4, the most restrictive specification, indicates
than those without debt. In short, the raw data indicated that having any debt was, on average, associated with a
that adults who had debt tended to be more socioeco- 0.92 point increase in depressive symptoms. This implies
nomically advantaged than those who did not. These dif- that having debt was associated with having approximately
ferences may, in part, explain why having debt was 9 % more depressive symptoms (given a mean of 10 de-
associated with fewer depressive symptoms in the raw data. pressive symptoms among those with no debt), which we
They highlight the importance of adjusting for individual interpret as constituting a modestly large magnitude of
and household characteristics when estimating associations effect. Among the covariates (results not shown), we found
between debt and depressive symptoms. that being married, higher income, working, and in good
In addition, we examined descriptive statistics by physical health were inversely associated with depressive
specific types of debt (results not shown). We found that symptoms, as would be predicted.
those with short-term debt had greater depressive symptoms Results for total amount debt, as well as for each type of
than those without short-term debt, although this difference debt (any and amount), are presented in Panels B through D
fell short of statistical significance. In contrast, those with in Table 2. Consistent with the results for having any debt
mid- and long-term debt had significantly fewer depressive (Panel A), we found that the logarithm of total debt (Panel
symptoms than those without such debt. We also found that B) was negatively associated with depressive symptoms in
men were more likely than women to have mid- and long- Model 0 (which controlled only for wave of observation)
term debt, but less likely to have short-term debt. For the and Model 1 (which controlled for wave of observation and
most part, patterns for the other covariates were similar time-invariant baseline characteristics). In Model 2, which
across all three debt measures, with the exception that those also controlled for time-variant socioeconomic factors, total
with greater assets were more likely to have mid-, and debt was positively associated with depressive symptoms,
particularly long-term debt than those with fewer assets. although this estimate (for amount of debt) failed to attain
This makes sense given that those with greater assets are statistical significance, unlike the estimate for having any
considerably more likely to own a home and to have taken debt. Both of the fixed-effects regression estimates (Models
on a mortgage for its purchase. These results suggest that 3 and 4) were positive, and the Model 4 estimate was sta-
examining associations between particular types of debt and tistically significant. The coefficient from Model 4 indicates
depressive symptoms separately is warranted. that a 10 % increase in debt was roughly associated with a
1.4 point (about 14 %) increase in depressive symptoms.
Regression Results Turning to specific types of debt (Panels C and D), we
found that the association between debt and depressive
Table 2 shows results from our primary standard OLS symptoms was predominantly driven by short-term debt.
(Models 0, 1, and 2) and fixed-effects (Models 3 and 4) Short-term debt was associated with greater depressive
regression models. Panel A presents results from models in symptoms in all of the models. The Model 4 estimates
which we used a dichotomous indicator for having any debt indicate that having any short-term debt was associated with
as the key predictor. Consistent with the bivariate results, 8 % more depressive symptoms and that a 10 % increase in
the estimate from Model 0, which adjusted only for wave short-term debt was associated with roughly a 24 % in-
of observation, revealed that having any debt was crease in depressive symptoms, given an average of 9.1

123
J Fam Econ Iss

Table 2 OLS and fixed effects regression results

(0) (1) (2) (3) (4)

Panel A: Any debt

Any debt -1.014 (0.199)*** -0.216 (0.203) 0.743 (0.196)*** 0.559 (0.291)? 0.916 (0.293)**
Panel B: Amount of total debt
Logarithm of total debt -0.374 (0.030)*** -0.295 (0.031)*** 0.037 (0.034) 0.024 (0.052) 0.138 (0.053)*
Panel C: Any short-, mid-, and long-term debt
Any short-term debt 0.622 (0.154)*** 0.528 (0.150)*** 0.779 (0.145)*** 0.640 (0.213)** 0.699 (0.207)**
Any mid-term debt -0.289 (0.163)?,a -0.062 (0.163)a 0.289 (0.161)?,a 0.113 (0.219) 0.350 (0.217)
Any long-term debt -1.976 (0.153)***,a,b -1.101 (0.158)***,a,b -0.325 (0.160)*,a,b -0.186 (0.230)a 0.272 (0.231)
Panel D: Any short-, mid-, and long-term debt
Logarithm of short-term debt 0.193 (0.046)*** 0.170 (0.045)*** 0.229 (0.043)*** 0.192 (0.065)** 0.215 (0.063)***
,a ?,a a a
Logarithm of mid-term debt -0.130 (0.036)*** -0.062 (0.036) 0.033 (0.036) 0.010 (0.048) 0.071 (0.048)c
a,b ,a,b ,a,b a
Logarithm of long-term debt -0.351 (0.024) -0.210 (0.026)*** -0.079 (0.027)** -0.054 (0.040) 0.032 (0.041)a
Baseline (exogenous) characteristics No Yes Yes Yes Yes
Socioeconomic factors No No Yes No Yes
Fixed effects No No No Yes Yes

16,964–17,059 individual-wave observations of 8457–8516 individuals per dataset, across 25 imputed datasets. OLS coefficients (and standard
errors) presented. Standard errors from the OLS models were adjusted for intracluster correlation due to multiple observations of each individual.
Covariates are listed in Table 1
?
p \ 0.10; * p \ 0.05, ** p \ 0.01, *** p \ 0.001
a
Differs from short-term debt at p \ 0.05
b
Differs from mid-term debt at p \ 0.05
c
Differs from short-term debt at p \ 0.10

depressive symptoms for those with no short-term debt (see was stably married throughout the observation period.
the note to Table 2). In contrast, the Model 4 estimates for Specifically, we estimated separate regression models for
mid- and long-term debt were small and non-significant. age 21–30, 31–50, and 51–64; those with a high school
Table 3 presents results from the OLS (Model 2) and education or less and those with greater than a high school
fixed-effects (Model 4) regressions with the full set of education; and those who were and were not consistently
controls, in which we tested the robustness of our findings married through the observation period. Results from the
to two specifications of relative debt levels: debt-to-annual fixed-effects regressions indicate that associations between
household income and debt-to-total household assets. short-term debt and greater depressive symptoms were most
Again, we present results for total debt and for the specific heavily concentrated among the older (those age 51–64) and
types of debt. We found no evidence of association be- less educated (those with a high school degree or less) groups
tween either total debt-to-income or total debt-to-assets and of adults in our sample, as well as among those who were not
depressive symptoms. Turning to types of debt, the OLS consistently married throughout the observation period.
and fixed-effect results using debt-to-annual income ratios Finally, because 10 % of the respondents in our sample
were consistent with those using the absolute debt mea- reported having no depressive symptoms, we also esti-
sures: short-term debt was associated with increased de- mated a series of Tobit models with random effects as an
pressive symptoms and the short-term debt estimate also additional robustness check. The Tobit model, which is a
significantly differed from the estimates for mid- and long- censored regression model, was used to estimate the linear
term debt in the fixed-effects models. Whereas the OLS relationship between debt and depressive symptoms while
results using the debt-to-assets measures were also con- adjusting for the fact that depressive symptoms may have
sistent with those using the absolute debt measures, the been left censored, as indicated by the relatively large
fixed-effects results suggest no association of any of the proportion of zeros, or even right censored if actual de-
debt-to-assets ratios with depressive symptoms. pressive symptoms may have extended beyond the max-
The results presented in Table 4 focus on subgroup dif- imum value of the measure. Results (not shown, available
ferences in associations between debt and depressive upon request) were consistent with those of our primary
symptoms by age, education, and whether the respondent models.

123
 J Fam Econ Iss

Table 3 Summary of
(2) (4)
robustness checks for ols and
fixed effects regression results: Panel A: Amount of total debt/annual income
debt as a proportion of income
and assets Total debt/annual income 0.000 (0.004) -0.002 (0.003)
Panel B: Amount of short-, mid-, and long-term debt/annual income
Short-term debt/annual income 0.023 (0.007)** 0.019 (0.010)?
Mid-term debt/annual income 0.032 (0.038) 0.002 (0.042)a
Long-term debt/annual income -0.004 (0.002)?,a -0.005 (0.004)a
Panel C: Amount of total debt/assets
Total debt/assets 0.009 (0.007) -0.001 (0.011)
Panel D: Amount of short-, mid-, and long-term debt/assets
Short-term debt/assets 0.015 (0.003)*** -0.027 (0.043)
Mid-term debt/assets 0.037 (0.037) 0.038 (0.038)
Long-term debt/assets -0.009 (0.014) -0.007 (0.019)
16,964–17,059 individual-wave observations of 8457–8516 individuals per dataset, across 25 imputed
datasets. OLS coefficients (and standard errors) presented. Standard errors from the OLS models were
adjusted for intracluster correlation due to multiple observations of each individual. Models adjust for all of
the covariates listed in Table 1
?
p \ 0.10; * p \ 0.05, ** p \ 0.01, *** p \ 0.001
a
Differs from short-term debt at p \ 0.05

Table 4 Fixed effects regression results: subgroup analyses by age, education attainment, and consistently married for amount of debt by type
Age Education Consistently married
21–30 31–50 51–64 BHS [HS No Yes
(4) (4) (4) (4) (4) (4) (4)

Logarithm of short-term 0.198 (0.192) 0.097 0.393 0.281 0.144 0.299 0.109 (0.084)
debt (0.089) (0.192)* (0.093)** (0.079)? (0.088)**
Logarithm of mid-term 0.202 (0.149) 0.077 0.061 (0.138) 0.052 (0.067)a 0.080 (0.062) -0.029 0.110
debt (0.065) (0.071)a (0.060)?
Logarithm of long-term -0.034 0.019 0.180 (0.137) 0.028 (0.062)a 0.056 (0.057) -0.049 0.018 (0.055)
debt (0.142) (0.063) (0.062)a
Fixed effects Yes Yes Yes Yes Yes Yes Yes
Proportion of sample 0.251 0.559 0.189 0.575 0.425 0.452 0.548
16,964–17,059 individual-wave observations of 8457–8516 individuals per dataset, across 25 imputed datasets. OLS coefficients (and standard
errors) presented. Standard errors from the OLS models were adjusted for intracluster correlation due to multiple observations of each individual.
Models adjust for all of the covariates listed in Table 1
?
p \ 0.10; * p \ 0.05, ** p \ 0.01, *** p \ 0.001
a
Differs from short-term debt at p \ 0.05

Discussion and Conclusions individuals or households may use home equity loans to
pay off short-term debt (thus, effectively converting short-
Several limitations should be considered when interpreting term debt to mid- or long-term debt). Individuals may also
the results of this study. First, the fixed-effects regressions boost current (short-term) consumption by taking student
are based on the assumption that associations of types and loans of a larger size than is needed to cover direct
levels of debt with depressive symptoms are time invariant. educational expenses. Our analyses were unable to account
As such, our estimates may have been biased by omitted for these possibilities. Rather, we observed only the
time-varying factors or by time-invariant factors that have amounts and types of debt owed at two given points in
time-varying influences on debt or depressive symptoms. time. Third, there may be considerable heterogeneity in
This may be a particularly important concern given the associations between debt and depressive symptoms by
length of time between observations. Second, specific types factors beyond age, education, and marital stability, in-
of debt are, to some extent fungible. For example, cluding other measures of SES, gender, and race/ethnicity;

123
J Fam Econ Iss

our analyses were silent in this regard. Fourth, it is im- source, found short-term debt at Wave 1 to be associated
portant to note that depressive symptoms, as measured by with fewer depressive symptoms at Wave 2 (the study did
this application of the CES-D, do not constitute clinical not examine mid- or long-term debt). The difference be-
depression. As such, our results cannot be interpreted as tween results from that study and our results reflects that
estimates of associations between debt and clinical de- our estimates were adjusted for initial levels of depressive
pression, and it cannot be assumed that debt has a similar symptoms. Accounting for initial differences between
association with diagnosable depression. Finally, although debtors and non-debtors addresses selection into borrow-
our fixed-effects regressions account for within-individual ing, as does focusing on changes in debt and changes in
changes in debt and estimate associations between changes depressive symptoms over time. To better compare our
in debt and changes in depressive symptoms, reverse results to Dew, we also estimated associations between
causality continues to be a concern. Specifically, because debt and depressive symptoms separately for stably mar-
we observe individuals at only two distinct points in time, ried individuals and other individuals. These results
roughly 6 years apart, we cannot determine the order in showed that the association between debt and depressive
which debt levels and depressive symptoms changed symptoms was considerably stronger among individuals
(which preceded which) during the intermittent time peri- who were not consistently married. The association was
od. To gain further insight into the potential direction of smaller and non-significant, but positive, when the model
these associations, we estimated cross-lagged path (struc- was estimated using only the subsample of consistently
tural equation) models in which we simultaneously re- married families. This further suggests that the difference
gressed Wave 2 debt on Wave 1 depressive symptoms and between our estimates and Dew’s (2007) likely reflects
Wave 2 depressive symptoms on Wave 1 debt, while also differences in model specification more so than differences
accounting for the correlation between debt and depressive in sample definition.
symptoms within each wave. Results (not shown) revealed On the whole, our results suggest that short-term or
that, in all cases, associations between Wave 1 debt and unsecured debt is associated with increased depressive
Wave 2 depressive symptoms were substantially larger symptoms, whereas longer-term debt does not exhibit such
than associations between Wave 1 depressive symptoms an association. Short-term debt is often used for immediate
and Wave 2 debt. This strongly suggests that the (causal) consumption and entails higher interest rates and fees than
direction of association runs from debt to depressive long-term debt; short-term debt may also be taken on with
symptoms rather than vice versa. relatively less agency than longer-term types of debt. For
Notwithstanding these limitations, the results from this example, people may perceive mortgages as a lifetime
study suggest that, after adjusting for background charac- asset, or education loans as an investment in human capital.
teristics, debt accumulation is associated with greater de- Both would be incurred with considerable agency. These
pressive symptoms among US adults. This finding was sorts of investments may not have the same associations
robust to both standard OLS and fixed-effects specifica- with depressive symptoms as short-term borrowing for
tions of our regression models. Furthermore, our analyses ongoing expenses. That is, individuals may experience
of specific types of debt revealed that, rather than this as- optimism with regard to debt that will ‘‘pay off’’ in the
sociation being driven by overall amount of debt, it was future through increased earnings or wealth acquisition (in
much more closely aligned with the type of debt incurred. the form of home equity), but experience psychological
Specifically, we found that the association between debt burden with regard to debt that does not contribute to asset
and depressive symptoms was primarily driven by short- or skill accumulation.
term debt, whether modeled dichotomously or by amount, This interpretation is consistent with Dwyer et al. (2011)
and whether considered in absolute or relative terms. On interpretation of their finding that educational debt was
the whole, short-term debt was significantly associated positively associated with self-esteem and mastery among
with depressive symptoms in the majority of our models, young adults, which they argued is likely to reflect that such
and the magnitude of this association was moderately large. debt has positive psychological influences because it is
By contrast, mid-term and long-term debt never sig- accrued for human capital investment. At the same time, we
nificantly differed from zero in the fixed-effects models. In found an association between short-term debt and greater
addition, the estimate for short-term debt significantly depressive symptoms, whereas they found an association
differed from that for long-term debt in most models, and between credit card debt and higher levels of self-esteem
significantly differed from that of mid-term debt in several; and mastery. They also interpreted this finding as reflecting
the estimates for mid-term and long-term debt were rarely positive influences associated with human capital invest-
significantly different from one another. ment. Two factors should be considered when weighing our
Our findings with regard to short-term debt provide results and theirs. First, the two studies measure psycho-
nuance to those of Dew (2007) who, using the same data logical wellbeing in very different domains. Second, in our

123
 J Fam Econ Iss

subgroup analyses, we found the association between short- Future research should also further investigate our finding
term debt and depressive symptoms to be most concentrated that the association of short-term debt with depressive
among adults age 51–64. Indeed, this association was not symptoms is particularly concentrated among 51–64 year
significant among younger adults, who are the focus of their old adults, those with a high school education or less, and
study. Thus, the two sets of findings should not be inter- those who are not stably married, which suggests that short-
preted as inconsistent. term debt may have an adverse influence on psychological
Few prior studies have investigated how mid-term debt, wellbeing, particularly for those who are less educated,
which we defined to include installment loans, home im- approaching retirement age, and either single or in unstable
provement loans, vehicle loans, loans from friends, and other relationships. This may be of concern given that these
(non-mortgage or education) bank loans, may influence populations groups may be least likely to experience sub-
psychological wellbeing. Such debt is often characterized by stantial future income growth. If this finding is confirmed, it
relatively low interest rates and amortized principal pay- may have implications for targeting financial capacity
ments. It may also be used to for the purchase of necessity building interventions toward these populations.
items or items that are consumed over a relatively long time Our results also have implications for the financial in-
period, such as a car or home improvements. As such, it may dustry and financial professionals. Financial planners,
share some characteristics with long-term debt and may educators, counselors and advisers may not be fully con-
therefore be similarly associated with psychological func- sidering how mental health issues and credit issues coin-
tioning or wellbeing. At the same time, such debt is not cide. Professionals should recognize signals of depressive
generally accrued as a result of human capital investment or symptoms and, especially, associations between mental
the purchase of an appreciable asset. In this way, it may differ health and short-term debt. There remains the potential that
considerably from long-term debt. Future research should lenders offering high-cost short-term debt will target dis-
explore these possibilities and uncover how mid-term debt advantaged populations by offering ‘‘easy’’ solutions to
may influence individual and household wellbeing. short-term cash flow problems, which serve to exacerbate
Although we found no links between long-term debt and both indebtedness and the client’s mental health problems.
depressive symptoms, it is important to note that our data Debt contracts offered to vulnerable borrowers might
were collected long before the recessions of 2001 or 2008, benefit from provisions that require affirmative disclosures
the housing boom and bust, or the significant increases in of terms and conditions, mandatory financial counseling, or
student debt that marked the last decade. It is quite possible even the right to rescission within a specified time frame.
that mortgage debt, especially in areas with declining home Likewise, mental health professionals may benefit from
values, is perceived differently now than in the past, which taking a closer look at their clients’ borrowing behaviors
may suggest a differential pattern of association with de- and overall financial status. Understanding how depression
pressive symptoms or other forms of psychological strain and borrowing may be reciprocally related remains an
(see Talbot et al. 2014 for a discussion). Likewise, given important issue for practitioners to consider.
shifts in lending for student loans and the experiences of Finally, our results have implications for public policy.
recent cohorts of young adults (see Friedline et al. 2014), The Great Recession and associated housing crisis prompted
the relative agency of investments in human capital may considerable regulatory reforms in federal laws governing
also be eroding. Future research should study associations mortgages and consumer (largely credit card) debt, as well as
of mortgage and education debt with individual and the creation of the Consumer Financial Protection Bureau.
household functioning in the last decade, given the exten- To the extent that such policies serve to limit the accrual of
sive shifts in financial and housing markets. short-term debt, they may also function as protective factors
Our results highlight the importance of considering vis-à-vis depressive symptoms that may increase with such
specific types and amounts of debt, rather than overall debt, debt. Future research should examine whether these recent
when modeling associations between debt and psycho- policy initiatives actually affect short-term debt accrual and,
logical wellbeing. Prior studies (e.g., Bridges and Disney in turn, whether reductions in short-term debt are associated
2010; Brown et al. 2005; Dew 2007; Gathergood 2012) with reductions in depressive symptoms.
have largely failed to simultaneously consider the full range
of specific types of debt that households may hold. As such,
previous findings may have confounded multiple and po-
tentially counterbalancing influences of different types of References
debt. Future research should further explore the influences
Attanasio, O. P., & Weber, G. (2010). Consumption and saving:
of specific types and amounts of debt on a range of domains Models of intertemporal allocation and their implications for
of functioning and wellbeing in order to provide a more public policy. Journal of Economic Literature, 48, 1–32. doi:10.
nuanced understanding of the nature of these relationships. 1257/jel.48.3.693.

123
J Fam Econ Iss

Azicorbe, A. M., Kennickell, A. B., & Moore, K. B. (2003). Recent Dynan, K. E. (2009). Changing household financial opportunities and
changes in US family finances: Evidence from the 1998 and economic security. Journal of Economic Perspectives, 23(4),
2001 Survey of Consumer Finances. Federal Reserve Bulletin 49–68. doi:10.1257/jep.23.4.49.
(January), A1–A32. Retrieved from http://www.federalreserve. Dynan, K. E., & Kohn, D. (2007). The rise in US household
gov/pubs/bulletin/2003/0103lead.pdf. indebtedness: Causes and consequences. Federal Reserve Board
Baek, E., & Hong, G. S. (2004). Effects of family life-cycle stages on Finance and Economics Discussion Series 2007–37. Retrieved
consumer debts. Journal of Family and Economic Issues, 25(3), from http://www.federalreserve.gov/pubs/feds/2007/200737/
359–385. doi:10.1023/B:JEEI.0000039946.59422.5f. 200737pap.pdf.
Bird, E., Hagstrom, P. A., & Wild, R. (1999). Credit card debts of the Federal Reserve Bank of New York. (2012). Student Loan Debt
poor: High and rising. Journal of Policy Analysis and Manage- History. Retrieved from http://www.newyorkfed.org/studentloan
ment, 18, 125–133. doi:10.1002/(SICI)1520-6688(199924). debt/.
Bridges, S., & Disney, R. (2010). Debt and depression. Journal of Friedline, T., Nam, I., & Loke, V. (2014). Household net worth
Health Economics, 29, 388–403. doi:10.1016/j.jhealeco.2010.02. accumulation patterns and young adult financial health: Ripple
003. effects of the great recession? Journal of Family and Economic
Brown, M., Haughwout, A., Lee, D., & van der Klaauw, W. (2011). Issues, 35(3), 390–410. doi:10.1007/s10834-013-9379-7.
Do we know what we owe? A comparison of borrower- and Gathergood, J. (2012). Debt and depression: Causal links and social
lender-reporter consumer debt. Federal Reserve Bank of New norm effects. The Economic Journal, 122, 1094–1114. doi:10.
York. Staff Report no. 523. Retrieved from http://www.new 1111/j.1468-0297.2012.02519.x.
yorkfed.org/research/staff_reports/sr523.pdf. Hyman, L. (2011). Politics and society in twentieth century America.
Brown, S., & Taylor, K. (2008). Household debt and financial assets: In W. Chave, G. Gerstle, L. Gordon, & J. Zelizer (Eds.), Debtor
Evidence from Germany, Great Britain and the USA. Journal of nation: The history of America in red ink. Princeton: Princeton
the Royal Statistical Society, 171(3), 615–643. doi:10.1111/j. University Press.
1467-985X.2007.00531.x. Jagtiani, J., & Li, W. (2013). Credit access and credit performance
Brown, S., Taylor, K., & Price, S. (2005). Debt and distress: after consumer bankruptcy filing: New evidence. Federal
Evaluating the psychological cost of credit. Journal of Economic Reserve Bank of Philadelphia Working Paper Series, 13–24.
Psychology, 26, 642–663. doi:10.1016/j.joep.2005.01.002. Retrieved from http://papers.ssrn.com/sol3/papers.cfm?abstract_
Chomsisengphet, S., & Pennington-Cross, A. (2006). The evolution id=2269621.
of the subprime mortgage market. Federal Reserve Bank of St. Jenkins, R., Bhugra, D., & Bebbington, P. (2008). Debt income and
Louis Review, 88(1), 31. Retrieved from https://research. mental disorder in the general population. Psychological Medi-
stlouisfed.org/publications/review/06/01/ChomPennCross.pdf. cine, 38(10), 1485–1493. doi:10.1017/S0033291707002516.
Conger, R. D., & Elder, G. H, Jr. (1994). Families in troubled times: Kalousova, L., & Burgard, S. A. (2013). Debt and foregone medical
Adapting to change in rural America. New York: Aldine de care. Journal of Health and Social Behavior, 54(2), 204–220.
Gruyter. doi:10.1177/0022146513483772.
Conger, R. D., Elder, G. H., Jr., Lorenz, F., Conger, K., Simons, R., Kalousova, L., & Burgard, S. A. (2014). Tough choices in tough
Whitbeck, L., Huck, S., & Melby, J. (1990). Linking economic times: Debt and medication nonadherence. Health Education &
hardship to marital quality and distress. Journal of Marriage Behavior, 41(2), 155–163. doi:10.1177/1090198113493093.
and the Family, 52, 643–656. Retrieved from http://www.jstor. Keese, M., & Schmitz, H. (2014). Broke, ill, and obese: Is there an
org/stable/352931. effect of household debt on health? Review on Income and
Dew, J. (2007). Two sides of the same coin? The differing roles of Wealth, 60, 525–541. doi:10.1111/roiw.12002.
assets and consumer debt in marriage. Journal of Family and Kennickell, A. B., & Shack-Marquez, J. (1992). Changes in family
Economic Issues, 28, 89–104. doi:10.1007/s10834-006-9051-6. finances from 1983 to 1989: Evidence from the Survey of
Dew, J. (2008). Debt change and marital satisfaction change in Consumer Finances. Federal Reserve Bulletin, 78, 1–18.
recently married couples. Family Relations, 57, 60–71. doi:10. Retrieved from http://www.federalreserve.gov/econresdata/scf/
1111/j.1741-3729.2007.00483.x. files/bull0192.pdf.
Drentea, P. (2000). Age, debt, and anxiety. Journal of Health and Social Lenton, P., & Mosley, P. (2008). Debt and health. Sheffield
Behavior, 41, 437–450. Retrieved from http://www.jstor.org/ Economic Research Paper Series. University of Sheffield.
stable/2676296. Retrieved from http://www.sheffield.ac.uk/polopoly_fs/1.
Drentea, P., & Lavrakas, P. J. (2000). Over the limit: The association 105541!/file/SERP2008004.pdf.
among health, race and debt. Social Science and Medicine, 50, Lown, J., Kim, J., Gutter, M., & Hunt, A.-T. (2014). Self-efficacy and
517–529. doi:10.1016/S0277-9536(99)00298-1. savings among middle and low income households. Journal of
Drentea, P., & Reynolds, J. R. (2012). Neither a borrower nor a lender Family and Economic Issues. doi:10.1007/s10834-014-9419-y.
be: The relative importance of debt and SES for mental health Lusardi, A., & Tufano, P. (2009). Debt literacy, financial experiences, and
among older adults. Journal of Aging and Health, 24, 673–695. over indebtedness (Working Paper No. w14808). National Bureau of
doi:10.1177/0898264311431304. Economic Research. Retrieved from http://www.nber.org/papers/
Durkin, T. A. (2000). Credit cards: Use and consumer attitudes, w14808.
1970–2000. Federal Reserve Bulletin, 86, 623–634. Retrieved Mann, A. (2011). The effect of late-life debt use on retirement
from http://www.federalreserve.gov/pubs/bulletin/2000/0900 decisions. Social Science Research, 40, 1623–1637. doi:10.1016/
lead.pdf. j.ssresearch.2011.05.004.
Dwyer, R. E., McCloud, L., & Hodson, R. (2011). Youth debt, McCloud, L., & Dwyer, R. E. (2011). The fragile American: Hardship
mastery, and self-esteem: Class-stratified effects of indebtedness and financial troubles in the 21st Century. The Sociological
on self-concept. Social Science Research, 40, 727–741. doi:10. Quarterly, 52, 13–35. doi:10.1111/j.1533-8525.2010.01197.x.
1016/j.ssresearch.2011.02.001. McKernan, S. M., & Ratcliffe, C. (2009). Asset Building for Today’s
Dwyer, R. E., McCloud, L., & Hodson, R. (2012). Debt and Stability and Tomorrow’s Security. Washington: Urban Institute.
graduation from American universities. Social Forces, 90, Retrieved from http://www.urban.org/uploadedpdf/1001374_
1133–1155. doi:10.1093/sf/sos072. asset_building.pdf.

123
 J Fam Econ Iss

Modigliani, F. (1986). Life cycle, individual thrift, and the wealth of October 4, 2011. Washington: US G.P.O. Retrieved from http://
nations. American Economic Review, 76(3), 297–313. Retrieved www.banking.senate.gov/public/index.cfm?FuseAction=Hearings.
from http://www.jstor.org/stable/1813352. Hearing&Hearing_ID=efab1264-2fe0-4518-aa83-1c42117d
Modigliani, F., & Brumberg, R. (1954). Utility analysis and the 40b6.
consumption function: An interpretation of cross-section data. In Watson, S., Barber, B., & Dziurawiec, S. (2014). The role of
K. K. Kurihara (Ed.), Post Keynesian economics (pp. 388–436). economizing and financial strain in Australian university
New Brunswick: Rutgers University Press. students’ psychological well-being. Journal of Family and
Norvilitis, J. M., & MacLean, M. G. (2010). The role of parents in Economic Issues. doi:10.1007/s10834-014-9404-5.
college students’ financial behaviors and attitudes. Journal of Worthington, A. C. (2006). Debt as a source of financial stress in
Economic Psychology, 31, 55–63. doi:10.1016/j.joep.2009.10.003. Australian households. International Journal of Consumer
Pearlin, L. I. (1989). The sociological study of stress. Journal of Science, 30(1), 2–15. doi:10.1111/j.1470-6431.2005.00420.x.
Health and Social Behavior, 30, 241–256. Retrieved from http:// Xiao, J. J., & Yao, R. (2011a). Debt holding and burden by family
www.jstor.org/stable/2136956. structure in 1989–2007. Networks Financial Institute at Indiana
Radolff, L. (1977). The CES-D Scale: A self-report depression scale for State University. Working Paper 2011-WP-04. Retrieved
research in the general population. Applied Psychological Mea- from http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1799
surement, 1(3), 385–401. doi:10.1177/014662167700100306. 362.
Reading, R., & Reynolds, S. (2001). Debt, social disadvantage and Xiao, J. J., & Yao, R. (2011b). Consumer debt delinquency over life
maternal depression. Social Science and Medicine, 53, 441–453. cycle stages. Networks Financial Institute at Indiana State
doi:10.1016/S0277-9536(00)00347-6. University. Working Paper 2011-WP-18. Retrieved from http://
Shah, A. K., Mullainathan, S., & Shafir, E. (2012). Some conse- papers.ssrn.com/sol3/papers.cfm?abstract_id=1903726.
quences of having too little. Science, 338, 682–685. doi:10.1126/ Yilmazer, T., & DeVaney, S. A. (2005). Household debt over the life
science.1222426. cycle. Financial Services Review, 14(4), 285–304. Retrieved
Shefrin, H. M., & Thaler, R. H. (1988). The behavioral life-cycle from http://web.b.ebscohost.com.ezproxy.library.wisc.edu/ehost/
hypothesis. Economic Inquiry, 26(4), 609–643. doi:10.1111/j. pdfviewer/pdfviewer?vid=2&sid=f3b067dd-10c4-4b61-9e63-
1465-7295.1988.tb01520.x. f62a4797933b%40sessionmgr110&hid=110.
StataCorp. (2013). Stata Statistical Software: Release 13. College
Station: StataCorp LP. Lawrence M. Berger is Director of the Institute for Research on
Sullivan, T. A., Warren, E., & Westbrook, J. L. (2000). The fragile Poverty and Professor and Ph.D. Chair in the School of Social Work
middle class: Americans in debt. CT: Yale University Press. at the University of Wisconsin-Madison. His research focuses on the
Sweet, J., Bumpass, L., & Call, V. (1988). The design and content of ways in which economic resources, sociodemographic characteristics,
the National Survey of Households and Families. NSFH and public policies affect parental behaviors and child and family
Working Paper No. 1. Center for Demography and Ecology, wellbeing.
University of Wisconsin–Madison. Retrieved from http://www.
ssc.wisc.edu/cde/nsfhwp/nsfh1.pdf. J. Michael Collins is faculty director of the Center for Financial
Sweet J. A., & Bumpass, L. L. (1996). The National Survey of Families Security and Associate Professor at the LaFollette School of Public
and Households—waves 1 and 2: Data description and documen- Affairs and the School of Human Ecology at the University of
tation. Center for Demography and Ecology, University of Wisconsin-Madison. He studies consumer decision-making in the
Wisconsin–Madison. Retrieved from http://www.ssc.wisc.edu/ financial marketplace, including the role of public policy in
nsfh/home.htm. influencing credit, savings, and investment choices.
Talbot, A., Tobe, E., & Ames, B. (2014). The experience of un-or
underemployment and home foreclosure for mature adults: A Laura Cuesta is a Ph.D. candidate in Social Welfare and a Graduate
phenomenological approach. Journal of Family and Economic Research Fellow of the Institute for Research on Poverty at the
Issues. doi:10.1007/s10834-014-9421-4. University of Wisconsin-Madison. She holds a bachelor’s degree and
US Senate. (2011). Consumer protection and middle-class wealth a master’s degree in Economics from Universidad de los Andes
building in an age of growing household debt. Hearing before (Bogotá, Colombia). Laura’s current research focuses on international
the Subcommittee on Financial Institutions and Consumer approaches to child and family policy, poverty and inequality, and
Protection of the Committee on Banking, Housing, and Urban parental incarceration among disadvantaged families.
Affairs, United States Senate. 112th Congress, 1st session.

123