Content uploaded by Arthur H. Goldsmith
Author content
All content in this area was uploaded by Arthur H. Goldsmith on Aug 30, 2016
Content may be subject to copyright.
Upjohn Press Book Chapters Upjohn Research home page
2012
Causality in the Relationship between Mental
Health and Unemployment
Timothy M. Die!e
Washington and Lee University
Arthur H. Goldsmith
Washington and Lee University
Darrick Hamilton
e New School
William Darity Jr.
Duke University
>is title is brought to you by the Upjohn Institute. For more information, please contact ir@upjohn.org.
Citation
Die?e, Timothy M., Arthur H. Goldsmith, Darrick Hamilton, and William Darity Jr. 2012. "Causality in the Relationship between
Mental Health and Unemployment." In Reconnecting to Work: Policies to Mitigate Long-Term Unemployment and Its
Consequences, Lauren D. Appelbaum, ed. Kalamazoo, MI: W.E. Upjohn Institute for Employment Research, pp. 63-94.
h?p://dx.doi.org/10.17848/9780880994095.ch4
63
4
Causality in the Relationship
between Mental Health
and Unemployment
Timothy M. Diette
Washington and Lee University
Arthur H. Goldsmith
Washington and Lee University
Darrick Hamilton
The New School
William Darity Jr.
Duke University
Unemployment is costly to society and individuals. Fifty years ago
economist Arthur Okun (1962) demonstrated that for the United States
in the postwar period, a 1 percent increase in the unemployment rate
is associated with a 3 percent decline in gross national product. Sub-
sequent work (Moosa 1997) revealed that this rule of thumb, known
as Okun’s Law, closely characterizes most developed economies. At
the individual level, unemployed persons who are laid off experience
fi nancial losses in the form of a drop in income, even if they are covered
by UI. Moreover, when reemployed, their wages typically fall short of
their previous level for a number of reasons, one of which is that work-
ers’ skills are not fully portable across fi rms, occupations, and indus-
tries (Goldsmith and Veum 2002).
Social scientists also assert that unemployment lasting more than
a few weeks is damaging to mental health. For instance, two meta-
analytic studies (McKee-Ryan et al. 2005; Paul and Moser 2009) report
that unemployed persons have substantially poorer psychological well-
up12lartw0ch4.indd 63up12lartw0ch4.indd 63 10/12/2012 12:52:41 PM10/12/2012 12:52:41 PM
64 Diette et al.
being after controlling for a wide range of factors expected to infl uence
emotional health. However, a largely unresolved issue is whether the
poor mental health status associated with the unemployed is caused by
their involuntary joblessness. The purpose of this chapter is to move
toward resolution of that question. First, we offer a new method for
identifying whether there is a causal link between exposure to unem-
ployment and emotional well-being. Second, by using this identifi -
cation strategy, and by drawing upon data from two large nationally
representative data sources—the National Comorbidity Survey Rep-
lication (NCS-R) and the National Latino and Asian American Study
(NLAAS)—we estimate the impact of both short-term and long-term
unemployment on a broad measure of emotional health.
UNEMPLOYMENT, PSYCHOLOGICAL HEALTH,
AND CAUSALITY
Social psychologists have proposed a number of pathways whereby
involuntary joblessness potentially diminishes emotional well-being.
Jahoda (1982) contends that unemployment is psychologically destruc-
tive primarily because it deprives an individual of the latent by-
products of work, including a structured day, shared experiences, sta-
tus, and opportunities for creativity and mastery.1 Erikson (1959), in
his life-span development theory, asserts that healthy emotional well-
being as an adult is contingent upon the realization of occupational suc-
cess for those intent on being breadwinners; therefore, unemployment
is harmful to mental health. Attribution theory (Heider 1958; Weiner
1974) suggests that individuals seek an explanation for developments
in their lives. Those who blame themselves for undesirable happen-
ings such as involuntary joblessness are likely to experience feelings of
“helplessness” (Seligman 1975), which damages mood (i.e., depression,
anxiety) and self-perception.2 Thus, for these persons, unemployment is
expected to foster psychological distress. A number of psychologists and
epidemiologists have asserted that the deleterious effects of unemploy-
ment increase as unemployment duration advances (Jackson and Warr
1984). They support the idea that stress accumulates, so there is reason
to believe that each additional week of joblessness is even more emo-
up12lartw0ch4.indd 64up12lartw0ch4.indd 64 10/12/2012 12:52:43 PM10/12/2012 12:52:43 PM
The Relationship between Mental Health and Unemployment 65
tionally damaging than prior weeks (Eisenberg and Lazarsfeld 1938;
Harrison 1976). This suggests that long-term unemployment is more
harmful to psychological well-being than short-term unemployment.
There is an extensive empirical literature dating to the Great
Depression that documents a negative association between unemploy-
ment and psychological health.3 Ethnographic studies conducted by
Jahoda, Lazarsfeld, and Zeisel (1933) and Eisenberg and Lazarsfeld
(1938) found that the unemployed exhibited both poor emotional well-
being and an inferior view of themselves. Subsequently psychologists
have developed inventories of questions designed to measure various
dimensions of psychological health, including depression (Beck et
al. 1961); anxiety (Spielberger et al. 1983); mastery or self-effi cacy
(Pearlin et al. 1981; Rotter 1966); self-esteem (Rosenberg 1965); and
general psychiatric status (Goldberg and Blackwell 1970). Using these
measures, numerous researchers conducting quantitative studies using
cross-sectional survey data report that unemployed groups have lower
levels of psychological well-being than employed groups. Unemployed
persons have been found to exhibit higher levels of depression (Fryer
and Payne 1986) and anxiety (Kessler, Turner, and House 1989), as
well as lower levels of self-esteem (Feather 1982; Goldsmith, Veum,
and Darity 1997) and self-effi cacy (Goldsmith, Veum, and Darity 1995)
compared to the employed.4 However, because unemployment can be
the consequence of poor mental health, it is not appropriate to interpret
these results as conclusive evidence that unemployment causes deterio-
ration in emotional well-being.
A common strategy to address the issue of reverse causality is to
use longitudinal or panel data and examine whether changes in men-
tal health coincide with changes in workforce status. The fundamental
idea is that if involuntary joblessness leads to psychological distress,
then persons moving from an employed to an unemployed state will
exhibit a decline in mental health, and those switching over time from
an unemployed to a working state will experience an improvement in
psychological well-being. Numerous researchers report evidence con-
sistent with this perspective. Their fi ndings, although compelling, are
not defi nitive evidence in favor of the hypothesis that unemployment
causes deterioration in mental health.5 The problem is that it is still
possible that an individual’s emotional well-being changed, for some
reason, prior to the alteration in workforce status. We attempt to shed
up12lartw0ch4.indd 65up12lartw0ch4.indd 65 10/12/2012 12:52:43 PM10/12/2012 12:52:43 PM
66 Diette et al.
further light on the question of causality by examining whether psycho-
logically resilient persons (i.e., individuals who have always exhibited
sound emotional well-being) exposed to unemployment in the past year
are more likely to experience their fi rst spell of poor emotional well-
being than persons employed throughout the past year.
DATA AND A STRATEGY FOR DETERMINING IF
UNEMPLOYMENT CAUSES POOR MENTAL HEALTH
Data and Methodology
The NCS-R and the NLAAS were designed to collect informa-
tion on potential determinants of mental disorders in the United States
through face-to-face interviews with respondents conducted in the pri-
vacy of their homes. The NCS-R was carried out on a nationally rep-
resentative group of 9,282 racially and ethnically diverse respondents
between February 2001 and April 2003. The NLAAS contains infor-
mation on a nationally representative group of 4,649 Latino or Asian
respondents collected between May 2002 and November 2003. These
data sets, which we merge together, are ideal to use in our investigation
of whether a causal link exists between unemployment and emotional
health because of the way that the survey collects respondent informa-
tion on emotional well-being.
The NCS-R and the NLAAS respondents provided retrospective
information on whether they were sad, empty, discouraged, depressed,
or disinterested most of the day nearly every day for at least two weeks
or every month in the past year, which we use to construct a broad
measure of psychological distress.6 An unusual and desirable feature of
the survey is that respondents who had suffered psychological distress
were asked to provide the year during which they fi rst suffered a bout
of poor emotional health. We take advantage of this unique aspect of
the NCS-R and the NLAAS to develop a new strategy for assessing the
link between unemployment and psychological health. Using informa-
tion on the year of fi rst onset of poor psychological health, we stratify
our data into two separate subsamples or data sets. We construct a data
set composed of psychologically resilient persons (resilient)—those
up12lartw0ch4.indd 66up12lartw0ch4.indd 66 10/12/2012 12:52:43 PM10/12/2012 12:52:43 PM
The Relationship between Mental Health and Unemployment 67
who have either never experienced a signifi cant bout of poor emotional
well-being or had their fi rst spell in the past year—and a second data set
of psychologically vulnerable persons (vulnerable)—those who have
experienced psychological distress in prior years.
The resilient subsample allows us to focus on those individuals
without previous bouts of poor mental health. We suspect that persons
who report never experiencing sustained psychological distress over
the course of their life cycle and who are in the workforce will con-
tinue to be emotionally healthy. The resilient subsample allows us to
analyze those least likely to have a bout of poor mental health leading
to unemployment. Therefore, the fi ndings of this subsample represent a
signifi cant step forward in resolving the problem of identifying a causal
relationship between unemployment and poor mental health. However,
there are conditions where the resilient subsample could still suffer
from reverse causality.
For example, it is possible that some individuals in the resilient
subsample are misclassifi ed and should rightfully be in the vulnerable
subsample. These individuals would need to represent a substantial por-
tion of the resilient subsample to undermine the identifi cation strategy.
This would occur if there are many individuals who fail to report their
prior poor mental health status because of poor recall, fail to recog-
nize that they have mental health problems but their employers observe
the problems, or the survey questions fail to identify those with mental
health problems that employers observe. These individuals would be
more likely to have a bout of poor mental health in the current year
that causes unemployment. People may struggle to remember highly
specifi c events, but the questionnaire is designed to identify general
features of distress, such as being sad or feeling empty or discouraged.
Therefore, we suspect that misclassifi cation bias from failure to recall,
poor recognition of their mental state, or inadequate questions is lim-
ited. A separate challenge to our identifi cation strategy arises if a sub-
stantial group of individuals have mental health issues that are latent or
dormant, these issues manifest themselves in the current year, or these
individuals experience unemployment in the past 12 months as well.
These individuals would be misclassifi ed in our resilient subsample,
belonging instead in the vulnerable subsample.
The data also contain information on the number of weeks during
the past year that the respondent spent employed; unemployed; legiti-
up12lartw0ch4.indd 67up12lartw0ch4.indd 67 10/12/2012 12:52:43 PM10/12/2012 12:52:43 PM
68 Diette et al.
mately out of the labor force (i.e., disabled, retired, in school, or taking
care of a family member); and discouraged or out of the labor force
but not for justifi able reasons. We treat the latter category as time spent
unemployed. Following the literature we classify those who spent 26 or
more weeks unemployed during the past year as having suffered from
long-term unemployment, while those who spent less time unemployed
are designated as having experienced short-term unemployment.
Our primary interest is in examining the effect of involuntary unem-
ployment on mental health. Therefore, persons who are out of the labor
force for acceptable or genuine reasons are excluded from the data.7
Thus, we focus our investigation on whether those who experience
either short- or long-term unemployment in the past year had a higher
probability of experiencing their fi rst lifetime bout of emotional distress
than those who spent the past year fully employed while holding con-
stant other economic and social determinants of mental health.
Descriptive Statistics
Our analysis is conducted separately on the subsample of resilient
persons, those who have either never experienced a spell of prolonged
psychological distress or have in the past year had their fi rst bout of
poor emotional health, and on the subsample of vulnerable individuals
who have experienced sustained psychological distress prior to the past
12 months. Table 4.1 reveals that there are 5,485 persons in the resilient
subsample, 5,421 of whom have never been “sad” or experienced a
substantial period of poor mental health, while 64 individuals (slightly
more than 1 percent of the subsample) were sad this past year for the
fi rst time. There are 2,109 respondents who have proven to be vulner-
able to bouts of poor emotional well-being prior to the current year.
Forty percent (845) of these persons also were saddled with psychologi-
cal distress this past year, while 1,264 avoided poor mental health over
the course of the previous 12 months.
Table 4.1 also presents information on labor force status for those
who experienced psychological distress in the past year and for those
who were emotionally healthy throughout the past 12 months, for both
the resilient and vulnerable subsamples. Of interest is whether a dispro-
portionate share of the individuals who are in distress this year experi-
enced unemployment—especially long-term unemployment—over the
past year.
up12lartw0ch4.indd 68up12lartw0ch4.indd 68 10/12/2012 12:52:44 PM10/12/2012 12:52:44 PM
The Relationship between Mental Health and Unemployment 69
A large share (30 percent) of the persons in the resilient subsam-
ple who express being sad or distressed this year—for the fi rst time in
their lives—were exposed to unemployment during the past 12 months.
Among those who experienced no psychological distress in the past
year, only 18 percent spent some weeks unemployed. The same pat-
tern exists for the vulnerable subsample. There is a higher proportion
unemployed among those suffering poor emotional well-being in the
Table 4.1 History of Psychological Distress and Workforce Status
Summary Statistics for Resilient and Vulnerable Subsamples
Panel A: Workforce status—resilient subsample (n = 5,485)
Psychological distress
this past year
(n = 64 = 1%)
No psychological distress
this past year
(n = 5,421 = 99%)
Employed 45
(70%) 4,425
(82%)
Short-term unemployment 5
(8%) 383
(7%)
Long-term unemployment 14
(22%) 613
(11%)
Panel B: Workforce status—vulnerable subsample (n = 2,109)
Psychological distress
this past year
(n = 845 = 40%)
No psychological distress
this past year
(n = 1,264 = 60%)
Employed 619
(73%) 1,051
(83%)
Short-term unemployment 96
(12%) 86
(7%)
Long-term unemployment 130
(15%) 127
(10%)
NOTE: Resilient persons have either never experienced psychological distress—a
sustained period over at least one month in the past year of sadness/discouragement/
disinterest—or had their fi rst spell of distress in the past year. Vulnerable persons have
experienced psychological distress prior to the past 12 months and may also have
experienced a spell of distress in the past year. People who were unemployed in the
past year and spent, in total, less than 26 weeks unemployed are identifi ed as having
experienced a bout of short-term unemployment. The long-term unemployed spent 26
or more weeks in the past year unemployed.
SOURCE: Data are drawn from the NCS-R and the NLAAS.
up12lartw0ch4.indd 69up12lartw0ch4.indd 69 10/12/2012 12:52:44 PM10/12/2012 12:52:44 PM
70 Diette et al.
past year (27 percent) relative to those with good emotional health in
the most recent year (17 percent). Thus, it appears that involuntary job-
lessness is associated with psychological distress, although caution is in
order since we are not controlling for other determinants of emotional
health that could be correlated with unemployment.
Psychologists expect a variety of social and economic factors to
cushion the impact of unemployment on emotional health.8 A valuable
aspect of the NCS-R and the NLAAS data is the provision of informa-
tion on a myriad of factors, both economic and social, that are believed
to buffer the impact of unemployment on psychological health. This
makes it possible to account for these features of a person’s environ-
ment when examining the infl uence of unemployment on psychologi-
cal health. The potential buffers that we are able to control for in our
analysis include the number of siblings, the number of adult children,
the extent of their wealth, and if the respondent has a parent who is still
living, is currently married, has friends he speaks to often, and is part of
a close-knit religious community. Table 4A.1 in Appendix 4A provides
detailed defi nitions for all of the variables used in our formal analyses
of psychological health.
The NCS-R and the NLAAS also provide extensive information
on demographic factors that may contribute to psychological health,
including a respondent’s gender, educational attainment, age, and racial/
ethnic heritage. Moreover, information is available on respondents’
family characteristics when they were youths, allowing us to control
for whether they were raised by both of their parents, whether the fam-
ily received public assistance, and parents’ education.
Appendix Table 4A.2 presents summary statistics on all of these
variables used in our empirical analysis for both the resilient and vul-
nerable subsamples. We describe these characteristics below beginning
with the resilient subsample. About half of the subsample is female (49
percent), 67 percent are married, 55 percent completed more than high
school or are highly educated, 72 percent are more than 30 years old, 34
percent have young children in their homes, 44 percent are foreign born
(unsurprising, since much of the data come from the NLAAS), and the
average individual has accumulated $65,000 of net worth. The resilient
subsample we analyze is very diverse with respect to race/ethnicity:
7 percent are African American, 34 percent are of Hispanic origin, 27
up12lartw0ch4.indd 70up12lartw0ch4.indd 70 10/12/2012 12:52:44 PM10/12/2012 12:52:44 PM
The Relationship between Mental Health and Unemployment 71
percent are Asian, and 32 percent are white. Most people were raised
by both parents (79 percent), around half have highly educated moth-
ers (49 percent) and fathers (47 percent), and only 4 percent grew up in
poor families.
A third of the respondents in the resilient subsample had a mother
who was still alive, and a quarter reported that their dad was still living.
The typical person has 1.5 siblings and 1.3 adult children. Moreover,
45 percent say they speak to friends regularly and are frequent partici-
pants in a religious community. The characteristics of the vulnerable
subsample are similar to those of the resilient subsample on a number
of dimensions. However, the vulnerable group, relative to the resilient
group, are only half as likely to be born outside the United States, more
likely to be female (63 percent), more likely to have young children,
less likely to be Asian, twice as likely to have grown up in a family on
welfare, and have amassed substantially less wealth.
Empirical Procedures
In order to investigate the impact on emotional well-being of expo-
sure to short- or long-term unemployment during the past year relative
to employment throughout the past 12 months, we use Equation (4.1) to
estimate the following model of psychological distress:
(4.1) PsyDistress = α + β(ShortTermUnem) + ψ(LongTermUnem)
+ δ(Buffer) + λ(X) + ε .
PsyDistress takes on a value of 1 if the respondent reports being sad,
empty, discouraged, depressed, or disinterested most of the day nearly
every day for either at least two weeks or every month in the past year,
otherwise it is 0. Two bivariate indicators are used to capture the extent
of a person’s unemployment experience over the past year. Those indi-
viduals who experienced some unemployment in the past year and the
total number of weeks, whether or not they were concurrent, fall short
of 26 weeks and are identifi ed as having experienced short-term unem-
ployment, in which case ShortTermUnem = 1. The variable LongTerm-
Unem = 1 if an individual spent more than 25 weeks unemployed in the
past year. Buffer is a vector containing social and economic support
up12lartw0ch4.indd 71up12lartw0ch4.indd 71 10/12/2012 12:52:44 PM10/12/2012 12:52:44 PM
72 Diette et al.
variables expected to mitigate or exacerbate the impact of involuntary
joblessness on emotional health. X is a vector of demographic and fam-
ily control variables.
We estimate Equation (4.1) using a logistic regression to estimate
the impact of unemployment and other factors on the odds that a person
has suffered psychological distress in the past year. We report the odds
ratios from the logistic regression. The odds ratios represent the effect
of a unit increase in a continuous independent variable or a value of 1
for a bivariate variable on the odds of experiencing psychological dis-
tress in the past 12 months, relative to the odds when that same variable
takes on a value of 0. A coeffi cient greater than 1 indicates an increase
in the odds of suffering psychological distress (i.e., a coeffi cient esti-
mate of 1.2 means a 20 percent increase in odds relative to when the
bivariate variable is 0). A coeffi cient estimate of 1 suggests no change
in the odds of poor emotional health occurring and a value less than 1
means the probability of poor emotional well-being in the past year is
reduced (i.e., an estimate of 0.8 means the odds are 20 percent smaller
relative to when the bivariate variable is zero).
For individuals in the resilient data set, the estimation of Equation
(4.1) tests whether unemployment in the past year enhances the odds that
a person will experience their fi rst ever bout of sustained psychological
distress in the past year. It is a commonly held belief that unemploy-
ment causes a decline in emotional well-being. The advantage of esti-
mating Equation (4.1) with these data is that if unemployment is found
to be associated with a greater likelihood of poor emotional health, the
impact can be interpreted as causal with a high degree of confi dence.
Since these are resilient individuals who have only experienced their
fi rst bout of poor emotional health in the past year, it seems question-
able that this bout of poor emotional health led to their current stretch of
involuntary joblessness. A more likely story is that unemployment over
the past year led to a deterioration of psychological well-being among
persons with a history of sound psychological health.
In addition, to explore whether social and economic support medi-
ates the impact of unemployment on contemporaneous emotional
health, we stratify our subsamples by the presence (or not) of each buf-
fer and reestimate the model.
up12lartw0ch4.indd 72up12lartw0ch4.indd 72 10/12/2012 12:52:44 PM10/12/2012 12:52:44 PM
The Relationship between Mental Health and Unemployment 73
RESULTS
Unemployment and Psychological Distress
Table 4.2 is a summary table that presents our estimates of the
impact of both short- and long-term unemployment on the chances of
experiencing psychological distress in the past year for the resilient
subsample (Panel A) and the vulnerable subsample (Panel B). How-
ever, in our view reverse causality may mar the accuracy of the fi ndings
using the vulnerable population, while estimation of Equation (4.1) on
a subsample of resilient persons may well purge the estimates of the
endogeneity generated by reverse causality. Thus, the use of the resil-
ient subsample can produce estimates that are capable of illuminating
whether unemployment causes deterioration in emotional well-being.
Model 1 is a sparse specifi cation of Equation (4.1), where psychologi-
cal distress is stipulated to depend solely on workforce status. Model
2 adds controls for a host of social and economic buffers. Model 3,
the most complete specifi cation, further augments the model to account
for individual characteristics and family features when growing up.
Full results for the resilient subsample are presented in Table 4A.3 in
Appendix 4A, and Table 4A.4 reports our complete set of fi ndings for
the vulnerable subsample.
Panel A in Table 4.2 reveals that in all three models exposure to
long-term unemployment in the past year signifi cantly increases the
odds that a resilient person will experience their fi rst ever bout of poor
emotional well-being in the current year relative to resilient individu-
als who were employed throughout the past year. The estimates range
from a 125 percent increase in likelihood in Model 1 to a 218 percent
increase in Model 2. However, those resilient persons who are subject
to short-term unemployment during the past year have the same like-
lihood of experiencing their fi rst bout of poor mental health as persons
who were employed throughout the past year. Thus, our fi ndings sug-
gest that long-term unemployment has a larger detrimental impact on
emotional health than bouts of short-term unemployment.
Recall that we classify people who have experienced poor mental
health prior to the current year, regardless of the source of their poor
emotional states, as vulnerable. Among these persons, exposure to
up12lartw0ch4.indd 73up12lartw0ch4.indd 73 10/12/2012 12:52:44 PM10/12/2012 12:52:44 PM
74 Diette et al.
Table 4.2 Logit Estimates of the Impact of Short-Term and Long-Term
Unemployment on the Odds of Currently Experiencing
Psychological Distress for Resilient and Vulnerable
Subsamples—Summary Table
Model 1 Model 2 Model 3
Variables Odds ratio Odds ratio Odds ratio
Panel A: Resilient subsample
Workforce status
Short-term unemployment 1.28 1.10 1.04
(0.61) (0.53) (0.52)
Long-term unemployment 2.25*** 3.18*** 2.85***
(0.69) (0.99) (0.96)
Observations 5,485 5,485 5,485
Panel B: Vulnerable subsample
Workforce status
Short-term unemployment 1.90*** 1.85*** 1.80***
(0.30) (0.29) (0.29)
Long-term unemployment 1.74*** 1.69*** 1.58***
(0.23) (0.24) (0.22)
Observations 2,109 2,109 2,109
Controls
Buffers No Yes Yes
Demographics & family factors No No Yes
NOTE: *** p < 0.01. Reference group for unemployment is employed throughout the
previous year, those out of the labor force are excluded from the data, and discouraged
workers are counted as unemployed. The set of buffer variables includes measures
of assets, marital status, parents living, number of living siblings, number of adult
children, having close friends, being part of a religious community, and the lack of
young children in the home (see Table 4A.1 for detailed defi nitions of all variables
included in the estimated models). Demographic controls include indicators for for-
eign born, gender, education level, age cohort, and racial and ethnic heritage. Family
characteristics as a youth contain indicators that reveal who raised the respondent,
their parents’ education level, and the fi nancial status of the family when the respon-
dent was a youth. In addition, Models 2 and 3 include indicators for missing data on
assets, number of siblings, talking on the phone with friends, and regular attendance
at religious services.
SOURCE: Data are drawn from the NCS-R and the NLAAS.
up12lartw0ch4.indd 74up12lartw0ch4.indd 74 10/12/2012 12:52:44 PM10/12/2012 12:52:44 PM
The Relationship between Mental Health and Unemployment 75
either short- or long-term unemployment over the past year leads to a
signifi cant increase in their reporting to have experienced poor emo-
tional health in the past year relative to similar persons who worked
throughout the past year. For instance, vulnerable individuals who
were subject to long-term unemployment were 58 percent more likely
(Model 3) to experience psychological distress compared to those vul-
nerable persons in the labor force who worked the entire past year.
Consistent with our theory, we fi nd that a number of buffers—being
married, having adult children, having friends with whom you are in
regular contact, and being part of a religious community—signifi cantly
reduce the odds of experiencing psychological distress over the past
year, regardless of exposure to unemployment, for vulnerable persons
(see Appendix Table 4A.4). However, emotional health does not appear
to be directly related to such buffers for resilient persons.
Do Buffers Mediate the Link between Unemployment and
Psychological Distress?
An interesting question is whether social characteristics or features
of a person’s life act to insulate them from the adverse impact of unem-
ployment on their psychological health. We explore this question by
evaluating the link between unemployment and emotional well-being
when a potential social buffer is present and when it is absent across
both of our subsamples. Our fi ndings for seven social buffers (i.e.,
being married or having a mother who is alive) are presented in Table
4.3. Table 4A.4 presents evidence on the prevalence of the various buf-
fers in our data sets and on the size of the subsamples used to estimate
the impact of unemployment on psychological health when a potential
buffer is present and when it is absent.
Among resilient persons (the left side of Table 4.3), long-term
unemployment is positively associated with the odds of experiencing
psychological distress (i.e., an estimated coeffi cient > 1) in all seven
cases when the buffer is not present (on 4 occasions the estimate is
statistically signifi cant), but also for 6 of the seven scenarios when
the buffer is present (again, 4 of the estimated impacts are statistically
signifi cant). Moreover, the odds of poor emotional health due to long-
term unemployment exposure are elevated to a greater extent when the
buffer is not present relative to when it is present on three occasions
up12lartw0ch4.indd 75up12lartw0ch4.indd 75 10/12/2012 12:52:44 PM10/12/2012 12:52:44 PM
76 Diette et al.
Table 4.3 The Impact of Social and Economic Buffers on the Effect of
Short-Term and Long-Term Unemployment on the Odds of
Currently Experiencing Psychological Distress
Resilient subsample Vulnerable subsample
Panel A: Marriage stratifi cations
Not married
(n = 1,732) Married
(n = 3,649) Not married
(n = 939) Married
(n = 1,170)
Short-term
unemployment 1.63
(1.03) 0.45
(0.48) 2.61***
(0.63) 1.30
(0.30)
Long-term
unemployment 4.03***
(2.00) 1.92
(0.93) 1.84***
(0.39) 1.41*
(0.28)
Panel B: Mother stratifi cations
Mom not alive
(n = 3,531) Mom alive
(n = 1,731) Mom not alive
(n = 975) Mom alive
(n = 1,134)
Short-term
unemployment 3.10*
(1.92) 0.49
(0.36) 1.47
(0.41) 1.99***
(0.40)
Long-term
unemployment 4.366***
(2.21) 2.03
(1.01) 1.45**
(0.26) 1.83**
(0.44)
Panel C: Father stratifi cations
Dad not alive
(n = 763) Dad alive
(n = 1,376) Dad not alive
(n = 607) Dad alive
(n = 851)
Short-term
unemployment 1.75
(2.52) 1.24
(0.72) 1.79**
(0.53) 1.51*
(0.37)
Long-term
unemployment 11.14***
(8.10) 0.57
(0.60) 1.98**
(0.54) 1.12
(0.34)
Panel D: Adult children stratifi cations
No adult
children
(n = 2,256) Adult children
(n = 2,845)
No adult
children
(n = 1,042) Adult children
(n = 1,067)
Short-term
unemployment 0.84
(0.52) 1.730
(1.38) 1.59**
(0.32) 2.29***
(0.62)
Long-term
unemployment 2.34*
(1.17) 3.69**
(1.92) 1.19
(0.28) 1.80***
(0.33)
up12lartw0ch4.indd 76up12lartw0ch4.indd 76 10/12/2012 12:52:45 PM10/12/2012 12:52:45 PM
The Relationship between Mental Health and Unemployment 77
Table 4.3 (continued)
Resilient subsample Vulnerable subsample
Panel E: Talk to friends stratifi cations
Talk rarely
(n = 2,824) Talk often
(n = 2,473) Talk rarely
(n = 1,022) Talk often
(n = 1,027)
Short-term
unemployment 1.12
(0.90) 1.30
(0.81) 1.56*
(0.38) 1.77**
(0.39)
Long-term
unemployment 1.88
(1.06) 4.26***
(1.87) 1.72***
(0.35) 1.40
(0.29)
Panel F: Attend religious services stratifi cations
Attend rarely
(n = 2,472)
Attend
regularly
(n = 2,292) Attend rarely
(n = 1,011)
Attend
regularly
(n = 864)
Short-term
unemployment 1.27
(0.75) 0.71
(0.75) 2.31***
(0.54) 1.26
(0.38)
Long-term
unemployment 1.52
(0.88) 4.44***
(2.34) 1.91***
(0.38) 1.30
(0.31)
Panel G: Young children in the home stratifi cations
Children
(n = 1,054) No children
(426) Children
(n = 626) No children
(1,483)
Short-term
unemployment 1.02
(0.81) 0.99
(0.65) 1.82*
(0.56) 1.79***
(0.36)
Long-term
unemployment 1.27
(1.05) 3.53***
(1.43) 1.03
(0.36) 1.73***
(0.28)
NOTE: *** p < 0.01, ** p < 0.05, * p < 0.1.
SOURCE: Data are drawn from the NCS-R and the NLAAS.
(marriage, mother alive, father alive), but for the other four social buf-
fers the deleterious impact of long-term unemployment on emotional
well-being is larger when the buffer is present. Thus, the evidence is
mixed on whether social factors considered buffers reduce the impact
of long-term unemployment on mental health for resilient persons. Fur-
thermore, the results exhibit the same mixed pattern for the vulnerable
population.
Short-term unemployment is essentially unrelated to psychological
health regardless of whether social buffers are present or not for resilient
individuals. Experiencing short-term unemployment only signifi cantly
up12lartw0ch4.indd 77up12lartw0ch4.indd 77 10/12/2012 12:52:45 PM10/12/2012 12:52:45 PM
78 Diette et al.
damages emotional well-being for those without a mother who is alive
in our resilient subsample. However, the situation is very different for
the vulnerable who, prior to the current year, reported having suffered
through bouts of poor emotional health. For them, whenever social buf-
fers are not present, short-term unemployment leads to elevated odds of
psychological distress, and in 6 out of 7 cases, the impact is statistically
signifi cant. The same pattern holds when the social barrier is present,
which suggests that for vulnerable people the presence of what could
well be a buffer does not mitigate the deleterious impact of short-term
unemployment on mental health status. Thus, for persons with a prior
history of poor emotional well-being, short-term unemployment exhib-
its the same negative pattern of effects on psychological health as long-
term unemployment.
Do Demographic Factors and Education Mediate the Link
between Unemployment and Psychological Distress?
It is possible that the connection between psychological well-being
and unemployment is infl uenced by demographic factors such as age
and gender, as well as skill level or educational investment. To explore
this possibility we stratifi ed our data sets by gender, education level
(more than high school, high school or less), and age (30 years of age
or older, less than 30 years old). The results, reported in Table 4.4,
offer three key insights. First, for the resilient individuals, short-term
unemployment is unrelated to emotional well-being, regardless of gen-
der, education level, or age cohort. Second, the results for the vulner-
able individuals are consistent with the fi ndings in Table 4.2, Panel B:
both short- and long-term unemployment signifi cantly damage men-
tal health, regardless of gender, educational attainment, or age cohort.
Finally, among the resilient population, those most negatively affected
by long-term unemployment are males, highly educated, and older indi-
viduals—groups typically associated with being primary breadwinners.
CONCLUSION
A longstanding belief among social scientists is that unemploy-
ment, especially long bouts, has deleterious effects on emotional health.
up12lartw0ch4.indd 78up12lartw0ch4.indd 78 10/12/2012 12:52:45 PM10/12/2012 12:52:45 PM
The Relationship between Mental Health and Unemployment 79
Table 4.4 The Impact of Select Demographic Factors on the Effect of
Short-Term and Long-Term Unemployment on the Odds of
Currently Experiencing Psychological Distress
Panel A: Gender stratifi cations
Resilient subsample Vulnerable subsample
Male
(n = 2,683) Female
(n = 2,349) Male
(n = 790) Female
(n = 1,319)
Short-term
unemployment 0.59
(0.64) 1.33
(0.82) 1.94***
(0.49) 1.78***
(0.38)
Long-term
unemployment 5.62***
(2.93) 2.15*
(0.98) 1.93**
(0.53) 1.43**
(0.24)
Panel B: Education level stratifi cations
More than
high school
(n = 2,933)
High school
or less
(n = 2,468)
More than
high school
(n = 1,234)
High school
or less
(n = 875)
Short-term
unemployment 1.32
(0.93) 0.75
(0.61) 1.85***
(0.39) 1.82**
(0.47)
Long-term
unemployment 5.74***
(2.55) 1.53
(0.73) 1.42*
(0.29) 1.80***
(0.36)
Panel C: Age stratifi cations
More than 29
(n = 3,934) Less than 30
(n = 1,443) More than 29
(n = 1,565) Less than 30
(n = 544)
Short-term
unemployment 2.39
(1.33) 0.26
(0.27) 1.87***
(0.38) 1.87**
(0.52)
Long-term
unemployment 4.03***
(1.81) 1.96
(1.09) 1.63***
(0.26) 1.21
(0.38)
NOTE: *** p < 0.01, ** p < 0.05, * p < 0.1.
SOURCE: Data are drawn from the NCS-R and the NLAAS.
There is extensive evidence of a direct link between mental health and
involuntary joblessness; however, the possibility that poor emotional
well-being leads to long periods of unemployment has left the question
of causality unresolved. This chapter introduces a new approach to the
assembly of data that allows estimation of the link between emotional
health and unemployment that may address concerns about the direc-
tion of causality. Our estimates are conducted using a subsample of
resilient persons—those who until the current year have never experi-
up12lartw0ch4.indd 79up12lartw0ch4.indd 79 10/12/2012 12:52:45 PM10/12/2012 12:52:45 PM
80 Diette et al.
enced poor mental health. If resilient individuals are exposed to unem-
ployment and exhibit poor mental health, it seems most likely that the
joblessness harmed their psychological health. We fi nd that long-term
unemployment—but not short-term unemployment—promotes psy-
chological distress among resilient persons. Moreover, the negative
psychological consequences of long-term unemployment are present
even when buffers exist, suggesting that policymakers consider both the
monetary and nonpecuniary costs of unemployment when formulating
policy to address economic downturns. Our fi ndings suggest that the
Great Recession and subsequent slow recovery have likely generated
extraordinary negative psychological consequences: at the peak of this
recession, about 45 percent of the unemployed had been out of work six
months or longer, and one-third of the unemployed were jobless for at
least a year.
Notes
1. Warr’s (1987) vitamin model is similar to Jahoda’s (1982) functionality frame-
work, in that desired features of work—like vitamins—contribute to psychologi-
cal health, and when they are withheld or withdrawn through unemployment,
emotional well-being is impaired.
2. Similarly, the Life Event model advanced by Brenner (1976) and Catalano and
Dooley (1977) argues that any alterations in life circumstances, especially those
deemed important to personal identity and status such as joblessness, are stressful
and thus may hamper psychological health.
3. Poorer mental health status for the unemployed relative to the employed has been
found for both men (Ensminger and Celentano 1990; Rowley and Feather 1987),
and women (Dew, Bromet, and Penkower 1992), and long-term unemployment is
especially damaging (Warr and Jackson 1985).
4. For a meta-analysis review of cross-sectional studies of the link between various
forms of emotional health and unemployment, see Paul and Moser (2009).
5. For a meta-analytic review of longitudinal studies fi nding improvements in emo-
tional health for unemployed who fi nd work, see McKee-Ryan et al. (2005).
6. Kessler et al. (2003) combined respondents’ self-reports on a similar set of feel-
ings and emotions to construct a nonspecifi c psychological distress score to assess
mental health.
7. Examples of acceptable reasons included those who are retired, homemakers, in
school, and physically or mentally unable to work.
8. Numerous studies report that social support buffers the psychological distress asso-
ciated with unemployment. See, for instance, Atkinson, Liem, and Liem (1986).
up12lartw0ch4.indd 80up12lartw0ch4.indd 80 10/12/2012 12:52:45 PM10/12/2012 12:52:45 PM
Appendix 4A
up12lartw0ch4.indd 81up12lartw0ch4.indd 81 10/12/2012 12:52:45 PM10/12/2012 12:52:45 PM
82 Diette et al.
Table 4A.1 Defi nition of Variables Used in Logit Estimation of the
Infl uence of Unemployment on Psychological Distress
Variable name Variable defi nition
Data sets
Resilient 1 if respondent has never experienced psychological
distress (see outcome defi nition below) or had their fi rst
bout in the past year, 0 otherwise
Vulnerable 1 if respondent has experienced psychological distress
prior to the current year, 0 otherwise
Outcome
PsyDistress 1 if respondent reports being sad, empty, discouraged,
depressed, or disinterested most of the day nearly every
day in the past year for either at least two weeks or
every month, 0 otherwise
Work force status
Short-term
unemployment 1 if experienced unemployment during the past year
and the total weeks summed to 25 or fewer weeks, 0
otherwise
Long-term
unemployment 1 if experienced unemployment during the past year
and the total weeks summed to 26 or more weeks, 0
otherwise
Employed 1 if employed throughout the past year at least 40
weeks and experienced no unemployment in past 12
months
Economic &
social buffers
Assets Respondent’s estimated value of assets less debts in
thousands
Married 1 if respondent is currently married or cohabitating, 0
otherwise
Mother living 1 if respondent’s biological mother is still alive, 0
otherwise
Father living 1 if respondent’s biological father is still alive, 0
otherwise
Siblings Number of siblings respondent had while growing up,
top coded at 8
up12lartw0ch4.indd 82up12lartw0ch4.indd 82 10/12/2012 12:52:46 PM10/12/2012 12:52:46 PM
The Relationship between Mental Health and Unemployment 83
Variable name Variable defi nition
Economic &
social buffers
Adult children Total number of adult children respondent has that are
living—both biological and nonbiological, 0 otherwise.
Friends 1 if respondent often talks on phone or gets together
with friends most every day or a few times a week, 0 if
less often.
Religious
community 1 if respondent attends religious services at least 3
times per month, 0 otherwise.
Young children Total number of living biological and nonbiological
children under 17 years of age living in respondent’s
home.
Demographics
Foreign born 1 if respondent reports being born outside the United
States, 0 otherwise.
Female 1 if respondent is female, 0 otherwise.
Highly educated 1 if respondent reports having completed more than 12
years of formal education, 0 otherwise.
Young 1 if respondent is less than 31 years of age, 0 otherwise.
African American 1 if respondent reports being African Caribbean or
African American, 0 otherwise.
Hispanic 1 if respondent reports being Hispanic, 0 otherwise.
Asian 1 if respondent reports being Asian, 0 otherwise.
Family characteristics
Both parents 1 if respondent reports being raised by both their
biological father and biological mother, 0 otherwise.
Mother highly
educated 1 if respondent reports their mother completed 12 or
more years of formal education, 0 otherwise.
Father highly
educated 1 if respondent reports their father completed 12 or
more years of formal education, 0 otherwise.
Welfare 1 if respondent reports their family was on welfare at
some time during their youth, 0 otherwise.
Table 4A.1 (continued)
up12lartw0ch4.indd 83up12lartw0ch4.indd 83 10/12/2012 12:52:46 PM10/12/2012 12:52:46 PM
84
Table 4A.2 Summary Statistics for All Variables Used in Logit Estimates for Resilient and Vulnerable Samples
Variable Resilient
(n = 5,485) Vulnerable
(n = 2,109) Variable Resilient
(n = 5,485) Vulnerable
(n = 2,109)
PsyDistress 0.01
(0.11) 0.40
(0.49) Young children 0.34
(0.81) 0.50
(0.95)
Short-term unemployment 0.07
(0.26) 0.09
(0.28) Foreign born 0.44
(0.50) 0.21
(0.41)
Long-term unemployment 0.11
(0.32) 0.12
(0.33) Female 0.49
(0.50) 0.63
(0.48)
Assets 65.05
(163.43) 75.25
(179.56) Highly educated 0.55
(0.50) 0.59
(0.49)
Assets—missing 0.38
(0.49) 0.31
(0.46) Young 0.28
(0.45) 0.26
(0.44)
Married 0.67
(0.47) 0.56
(0.50) African American 0.07
(0.25) 0.07
(0.26)
Mother living 0.35
(0.48) 0.54
(0.50) Hispanic 0.34
(0.48) 0.26
(0.44)
Father living 0.26
(0.44) 0.40
(0.49) Asian 0.27
(0.45) 0.11
(0.31)
Father living—missing 0.57
(0.51) 0.31
(0.46) Both parents 0.79
(0.41) 0.76
(0.43)
Siblings 1.51
(2.29) 2.37
(2.44) Mother highly educated 0.49
(0.50) 0.59
(0.49)
Siblings—missing 0.57
(0.50) 0.30
(0.46) Mother highly educated—missing 0.11
(0.31) 0.09
(0.29)
up12lartw0ch4.indd 84up12lartw0ch4.indd 84 10/12/2012 12:52:46 PM10/12/2012 12:52:46 PM
85
Adult children 1.31
(1.48) 1.04
(1.37) Father highly educated 0.47
(0.50) 0.52
(0.50)
Friends 0.45
(0.50) 0.49
(0.50) Father highly educated—missing 0.20
(0.40) 0.19
(0.39)
Friends—missing 0.03
(0.18) 0.03
(0.17) Welfare 0.04
(0.20) 0.08
(0.28)
Religious community 0.45
(0.50) 0.41
(0.49) Welfare-missing 0.57
(0.50) 0.31
(0.46)
Religious community—missing 0.10
(0.30) 0.11
(0.31)
SOURCE: Data drawn from the NCS-R and the NLAAS. Means are reported with standard errors in parentheses. Indicator variables are
constructed that take on a value of 1 if the individual does not answer a question and therefore have a missing value and a value of zero
for a valid response. We use the name construct of “variable name—missing” for each of these indicators. These indicators allow the
observation to be included in the sample but not infl uence the coeffi cient of the related variable.
up12lartw0ch4.indd 85up12lartw0ch4.indd 85 10/12/2012 12:52:46 PM10/12/2012 12:52:46 PM
86 Diette et al.
Table 4A.3 The Impact of Short-Term and Long-Term Unemployment
on the Odds of Currently Experiencing Psychological
Distress for Resilient Individuals—Full Results
Model 1 Model 2 Model 3
Variables Odds ratio Odds ratio Odds ratio
Workforce status
Short-term unemployment 1.28 1.10 1.04
(0.61) (0.53) (0.52)
Long-term unemployment 2.25*** 3.18*** 2.85***
(0.70) (0.99) (0.96)
Buffers
Assets 1.00 1.00
(0.00) (0.00)
Assets—missing 1.07 1.04
(0.32) (0.30)
Married 0.71 0.80
(0.193) (0.22)
Mother living 1.20 1.12
(0.47) (0.47)
Father living 1.30 1.12
(0.46) (0.42)
Father living—missing 0.10** 0.03**
(0.10) (0.04)
Siblings 1.03 1.02
(0.07) (0.07)
Siblings—missing 4.19 5.02
(43.00) (5.07)
Adult children 0.98 1.02
(0.09) (0.10)
Friends 1.01 0.99
(0.26) (0.26)
Friends—missing 0.50 0.46
(0.38) (0.36)
Religious community 0.69 0.64
(0.19) (0.18)
Religious community—missing 0.77 0.82
(0.34 (0.37)
Young children 1.15 1.16
(0.11) (0.13)
Born in foreign country 0.93 0.86
(0.40) (0.37)
up12lartw0ch4.indd 86up12lartw0ch4.indd 86 10/12/2012 12:52:46 PM10/12/2012 12:52:46 PM
The Relationship between Mental Health and Unemployment 87
Demographics
Female 1.96**
(0.55)
Highly educated 0.87
(0.24)
Young 1.42
(0.43)
African American 1.45
(0.64)
Hispanic 1.91*
(0.76)
Asian 1.64
(0.91)
Family characteristics
Both parents 1.08
(0.34)
Mother highly educated 1.07
(0.39)
Mother highly educated—missing 0.92
(0.46)
Father highly educated 1.34
(0.47)
Father highly educated—missing 0.95
(0.37)
Welfare 0.69
(0.38)
Welfare—missing 1.91
(2.00)
Constant 0.01*** 0.02*** 0.01***
(0.01) (0.01) (0.01)
Observations 5,485 5,485 5,485
NOTE: *** p < 0.01, ** p < 0.05, * p < 0.1. Resilient persons have either never expe-
rienced psychological distress—a sustained period over at least one month in the past
year of sadness/discouragement/disinterest—or had their fi rst spell of distress in the
past year. Reference group for unemployment is employed throughout the previous
year, those out of the labor force are excluded from the data, and discouraged workers
are counted as unemployed. Indicator variables are constructed that take on a value
of 1 if the individual does not answer a question and therefore have a missing value
and a value of zero for a valid response. We use the name construct of “variable
name—missing” for each of these indicators. These indicators allow the observation
to be included in the sample but not infl uence the coeffi cient of the related variable.
SOURCE: Data are drawn from the NCS-R and the NLAAS.
Table 4A.3 (continued)
up12lartw0ch4.indd 87up12lartw0ch4.indd 87 10/12/2012 12:52:47 PM10/12/2012 12:52:47 PM
88 Diette et al.
Table 4A.4 The Impact of Short-Term and Long-Term Unemployment
on the Odds of Currently Experiencing Psychological
Distress for Vulnerable Individuals—Full Results
Model 1 Model 2 Model 3
Variables Odds ratio Odds ratio Odds ratio
Workforce status
Short-term unemployment 1.90*** 1.85*** 1.80***
(0.30) (0.30) (0.29)
Long-term unemployment 1.74*** 1.69*** 1.58***
(0.23) (0.24) (0.22)
Buffers
Assets 1.00*** 1.00***
(0.00) (0.00)
Assets—missing 0.89 0.89
(0.09) (0.10)
Married 0.61*** 0.63***
(0.06) (0.06)
Mother living 0.99 0.97
(0.13) (0.13)
Father living 0.97 0.99
(0.12) (0.13)
Father living—missing 1.86* 1.71
(0.61) (0.59)
Siblings 1.01 0.99
(0.03) (0.03)
Siblings—missing 0.66 0.79
(0.22) (0.31)
Adult children 0.93** 0.93**
(0.03) (0.03)
Friends 0.77*** 0.77***
(0.07) (0.07)
Friends—missing 0.73 0.69
(0.21) (0.20)
Religious community 0.82** 0.84*
(0.08) (0.08)
Religious community—missing 0.87 0.87
(0.13) (0.13)
Young children 1.03 1.01
(0.05) (0.05)
Born in foreign country 0.99 1.05
(0.14) (0.15)
up12lartw0ch4.indd 88up12lartw0ch4.indd 88 10/12/2012 12:52:47 PM10/12/2012 12:52:47 PM
The Relationship between Mental Health and Unemployment 89
Demographics
Female 1.06
(0.10)
Highly educated 0.88
(0.09)
Young 1.21
(0.15)
African American 1.11
(0.22)
Hispanic 0.80
(0.15)
Asian 0.71
(0.17)
Family characteristics
Both parents 0.92
(0.11)
Mother highly educated 0.93
(0.11)
Mother highly educated—missing 1.06
(0.20)
Father highly educated 0.95
(0.12)
Father highly educated—missing 1.06
(0.17)
Welfare 1.46**
(0.26)
Welfare—missing 1.02
(0.33)
Constant 0.59*** 1.11 1.31
(0.03) (0.22) (0.35)
Observations 2,109 2,109 2,109
NOTE: *** p < 0.01, ** p < 0.05, * p < 0.1. Vulnerable persons have experienced
psychological distress—a sustained period over at least one month in the past year
of sadness/discouragment/disinterest—or had their fi rst spell of distress in the past
year, prior to the past 12 months and may also have experienced a spell of distress in
the past year. Reference group for unemployment is employed throughout the previ-
ous year, those out of the labor force are excluded from the data, and discouraged
workers are counted as unemployed. Indicator variables are constructed that take on
a value of 1 if the individual does not answer a question and therefore have a missing
value and a value of zero for a valid response. We use the name construct of “variable
name—missing” for each of these indicators. These indicators allow the observation
to be included in the sample but not infl uence the coeffi cient of the related variable.
SOURCE: Data are drawn from the NCS-R and the NLAAS.
Table 4A.4 (continued)
up12lartw0ch4.indd 89up12lartw0ch4.indd 89 10/12/2012 12:52:47 PM10/12/2012 12:52:47 PM
90
Table 4A.5 Sample Size for Buffers and Demographics Used to Stratify the Data to Evaluate If the Impact of
Unemployment on the Odds of Psychological Distress Depends on These Elements
Variable (n prior to stratifi cation)
Resilient subsample
Variable (n prior to stratifi cation)
Vulnerable subsample
Variable status Variable status
Yes No Yes No
Buffers Buffers
Married (n = 5,485) 66.5 33.5 Married (n = 2,109) 55.5 44.5
Mother living (n = 5,485) 34.6 65.4 Mother living (n = 2,109) 53.8 46.2
Father living (n = 2,356) 59.4 40.6 Father living (n = 1,458) 58.4 41.6
Adult children (n = 5,485) 58.9 41.1 Adult children (n = 2,109) 50.6 49.4
Friends (n = 5,297) 46.7 53.3 Friends (n = 2,049) 50.1 49.9
Religious community (n = 4,921) 49.8 50.2 Religious community (n = 1,875) 46.1 53.9
Young children (n = 5,485) 20.5 79.5 Young children (n = 2,109) 29.7 70.3
Demographics Demographics
Female (n = 5,485) 48.9 51.1 Female (n = 2,109) 62.5 37.5
Highly educated (n = 5,485) 55.0 45.0 Highly educated (n = 2,109) 58.5 41.5
Young (n = 5,485) 28.3 71.7 Young (n = 2,109) 25.8 74.2
NOTE: Sample size prior to stratifi cation may be smaller than the full subsamples used in the estimates presented in Tables 4.2 and 4.3. In
the full subsamples, some observations contain missing values for specifi c buffers or demographics. Estimates with the full subsample
include separate indicator variables for missing values for each variable. The stratifi cation analysis eliminates observations with a miss-
ing value for the buffer or demographic variable that is the basis for stratifying the resilient or vulnerable subsamples.
up12lartw0ch4.indd 90up12lartw0ch4.indd 90 10/12/2012 12:52:47 PM10/12/2012 12:52:47 PM
The Relationship between Mental Health and Unemployment 91
References
Atkinson, Thomas, Ramsay Liem, and Joan H. Liem. 1986. “The Social Costs
of Unemployment: Implications for Social Support.” Journal of Health and
Social Behavior 27(4): 317–331.
Beck, Aaron T., Calvin H. Ward, M. Mendelson, J. Mock, and J. Erbaugh.
1961. “An Inventory for Measuring Depression.” Archives of General Psy-
chiatry 4(6): 561–571.
Brenner, M. Harvey. 1976. Estimating the Social Costs of National Economic
Policy: Implications for Mental and Physical Health and Clinical Aggres-
sion. Report to the Joint Economic Committee of the U.S. Congress. Wash-
ington, DC: U.S. Government Printing Offi ce.
Catalano, Ralph, and C. David Dooley. 1977. “Economic Predictors of
Depressed Mood and Stressful Life Events.” Journal of Health and Social
Behavior 18(3): 292–307.
Dew, Mary A., Evelyn J. Bromet, and Lili Penkower. 1992. “Mental Health
Effects of Job Loss in Women.” Psychological Medicine 22(3): 751–764.
Eisenberg, Philip, and Paul F. Lazarsfeld. 1938. “The Psychological Effects of
Unemployment.” Psychological Bulletin 35(6): 358–390.
Ensminger, Margaret E., and David D. Celentano. 1990. “Gender Differences
in the Effect of Unemployment on Psychological Distress.” Social Science
and Medicine 30(4): 469–477.
Erikson, Erik H. 1959. “Identity and the Life Cycle.” Psychological Issues
1(1): 50–100.
Feather, Norman T. 1982. “Unemployment and Its Psychological Correlates:
A Study of Depressive Symptoms, Self-Esteem, Protestant Ethic Values,
Attributional Style and Apathy.” Australian Journal of Psychology 34(3):
309–323.
Fryer, David M., and Roy L. Payne. 1986. “Being Unemployed: A Review
of the Literature on the Psychological Experience of Unemployment.” In
International Review of Industrial and Organizational Psychology, Cary L.
Cooper and Ivan T. Robertson, eds. Chichester, England: Wiley, pp. 235–
278.
Goldberg, David P., and B. Blackwell. 1970. “Psychiatric Illness in General
Practice: A Detailed Study Using a New Method of Case Identifi cation.”
British Medical Journal 1: 439–443.
Goldsmith, Arthur H., and Jonathan R. Veum. 2002. “Wages and the Composi-
tion of Experience.” Southern Economics Journal 69(2): 429–443.
Goldsmith, Arthur H., Jonathan R. Veum, and William Darity Jr. 1995. “Are
Being Unemployed and Being Out of the Labor Force Distinct States? A
up12lartw0ch4.indd 91up12lartw0ch4.indd 91 10/12/2012 12:52:47 PM10/12/2012 12:52:47 PM
92 Diette et al.
Psychological Approach.” Journal of Economic Psychology 16(2): 275–
295.
———. 1997. “Unemployment, Joblessness, Psychological Well-Being and
Self-Esteem: Theory and Evidence.” Journal of Socio-Economics 26(2):
133–158.
Harrison, Richard. 1976. “The Demoralising Experience of Prolonged Unem-
ployment.” Department of Employment Gazette 84(4): 339–348.
Heider, Fritz. 1958. The Psychology of Interpersonal Relations. New York:
Wiley.
Jackson, Paul R., and Peter B. Warr. 1984. “Unemployment and Ill-Health:
The Moderating Role of Duration and Age.” Psychological Medicine 14:
605–614.
Jahoda, Marie. 1982. Employment and Unemployment: A Social-Psychologi-
cal Analysis. New York: Cambridge University Press.
Jahoda, Marie, Paul F. Lazarsfeld, and Hans Zeisel. 1933. Marienthal: The
Sociography of an Unemployed Community (English translation, 1971).
Chicago, IL: Aldine.
Kessler, Ronald C., Peggy R. Barker, Lisa J. Colpe, Joan F. Epstein, Joseph C.
Gfroerer, Eva Hiripi, Mary J. Howes, Sharon-Lise T. Normand, Ronald W.
Manderscheid, Ellen E. Walters, and Alan M. Zaslavsky. 2003. “Screening
for Serious Mental Illness in the General Population.” Archives of General
Psychiatry 60(2): 184–189.
Kessler, Ronald C., J. Blake Turner, and James S. House. 1989. “Unemploy-
ment, Reemployment, and Emotional Functioning in a Community Sam-
ple.” American Sociological Review 54(4): 648–657.
McKee-Ryan, Frances M., Zhaoli Song, Connie R. Wanberg, and Angelo J.
Kinicki. 2005. “Psychological and Physical Well-Being during Unem-
ployment: A Meta-Analytic Study.” Journal of Applied Psychology 90(1):
53–76.
Moosa, Imad A. 1997. “A Cross-Country Comparison of Okun’s Coeffi cient.”
Journal of Comparative Economics 24(3): 335–356.
Okun, Arthur M. 1962. “Potential GNP: Its Measurement and Signifi cance.”
Cowles Foundation Paper No. 190. New Haven, CT: Yale University,
Cowles Foundation.
Paul, Karsten I., and Klaus Moser. 2009. “Unemployment Impairs Mental
Health: Meta-Analyses.” Journal of Vocational Behavior 74(3): 264–282.
Pearlin, Leonard, Elizabeth G. Meneghan, Morton A. Lieberman, and Joseph
Mullan. 1981. “The Stress Process.” Journal of Health and Social Behavior
22(4): 337–356.
Rosenberg, Morris. 1965. Society and the Adolescent Self-Image. Princeton,
NJ: Princeton University Press.
up12lartw0ch4.indd 92up12lartw0ch4.indd 92 10/12/2012 12:52:47 PM10/12/2012 12:52:47 PM
The Relationship between Mental Health and Unemployment 93
Rotter, Julian B. 1966. “Generalized Expectancies for Internal versus External
Control of Reinforcement.” Psychological Monographs 80(1): 1–28.
Rowley, K. M., and Norman T. Feather. 1987. “The Impact of Unemployment
in Relation to Age and Length of Unemployment.” Journal of Occupational
Psychology 60(4): 323–332.
Seligman, Martin, E. P. 1975. Helplessness: On Depression, Development and
Death. San Francisco, CA: W. H. Freeman.
Spielberger, Charles D., Richard L. Gorsuch, Robert E. Lushene, Peter R.
Vagg, and Gerard A. Jacobs. 1983. “Manual for the State–Trait Anxiety
Inventory.” Palo Alto, CA: Consulting Psychologists Press.
Warr, Peter B. 1987. Work, Unemployment and Mental Health. Oxford, United
Kingdom: Oxford University Press.
Warr, Peter B., and Paul R. Jackson. 1985. “Factors Infl uencing the Psycho-
logical Impact of Prolonged Unemployment and Re-employment.” Psycho-
logical Medicine 15(4): 795–807.
Weiner, Bernard. 1974. Achievement Motivation and Attribution Theory. Mor-
ristown, NJ: General Learning Press.
up12lartw0ch4.indd 93up12lartw0ch4.indd 93 10/12/2012 12:52:47 PM10/12/2012 12:52:47 PM
up12lartw0ch4.indd 94up12lartw0ch4.indd 94 10/12/2012 12:52:47 PM10/12/2012 12:52:47 PM
Reconnecting to Work
Policies to Mitigate
Long-Term Unemployment
and Its Consequences
Lauren D. Appelbaum
Editor
2012
W.E. Upjohn Institute for Employment Research
Kalamazoo, Michigan
Library of Congress Cataloging-in-Publication Data
Reconnecting to work : policies to mitigate long-term unemployment and its
consequences / Lauren D. Appelbaum, editor.
p. cm.
Papers presented at a conference held on Apr. 1–2, 2011.
Includes bibliographical references and index.
ISBN-13: 978-0-88099-406-4 (pbk. : alk. paper)
ISBN-10: 0-88099-406-1 (pbk. : alk. paper)
ISBN-13: 978-0-88099-408-8 (hardcover : alk. paper)
ISBN-10: 0-88099-408-8 (hardcover : alk. paper)
1. Labor policy—United States—Congresses. 2. Unemployment—United
States—Congresses. 3. Full employment policies—United States—Congresses.
4. Recessions—United States—Congresses. I. Appelbaum, Lauren D.
HD5724.R337 2012
331.13'770973—dc23
2012034390
© 2012
W.E. Upjohn Institute for Employment Research
300 S. Westnedge Avenue
Kalamazoo, Michigan 49007-4686
The facts presented in this study and the observations and viewpoints expressed are
the sole responsibility of the authors. They do not necessarily represent positions of
the W.E. Upjohn Institute for Employment Research.
Cover design by Alcorn Publication Design.
Index prepared by Diane Worden.
Printed in the United States of America.
Printed on recycled paper.
