At the Intersection of Health, Health Care and Policy

Cite this article as:

Jonathan Cylus and Mauricio Avendano

Downloaded from http://content.healthaffairs.org/ by Health Affairs on February 7, 2017 by HW Team

Receiving Unemployment Benefits May Have Positive Effects On The Health Of
The Unemployed
Health Affairs 36, no.2 (2017):289-296
doi: 10.1377/hlthaff.2016.1040

The online version of this article, along with updated information and services, is
available at:
http://content.healthaffairs.org/content/36/2/289

For Reprints, Links &

Permissions : http://content.healthaffairs.org/1340_reprints.php
Email Alertings : http://content.healthaffairs.org/subscriptions/etoc.dtl

To Subscribe : https://fulfillment.healthaffairs.org

Health Affairs is published monthly by Project HOPE at 7500 Old Georgetown Road,
Suite 600, Bethesda, MD 20814-6133. Copyright ©
by Project HOPE - The People-to-People Health Foundation. As provided by United
States copyright law (Title 17, U.S. Code), no part of
may be reproduced, displayed, or transmitted in any form or by any means, electronic
or mechanical, including photocopying or by information storage or retrieval systems,
without prior written permission from the Publisher. All rights reserved.

Not for commercial use or unauthorized distribution

Unemployment Insurance & Health

By Jonathan Cylus and Mauricio Avendano

doi: 10.1377/hlthaff.2016.1040

Receiving Unemployment Benefits

HEALTH AFFAIRS 36,
NO. 2 (2017): 289–296
©2017 Project HOPE—
The People-to-People Health

May Have Positive Effects On The Foundation, Inc.

Health Of The Unemployed

Downloaded from http://content.healthaffairs.org/ by Health Affairs on February 7, 2017 by HW Team

Jonathan Cylus (J.D.Cylus@
ABSTRACT Research suggests that job loss can cause illness and premature lse.ac.uk) is a research fellow
at the London School of
death. This raises the question of whether unemployment benefit Economics and Political
programs, which are intended to alleviate the financial stress of job loss, Science and at the European
Observatory on Health
can protect the health of the unemployed. To investigate this question, Systems and Policies, in
we analyzed data for the period 1984–2009 from the Panel Study of England.

Income Dynamics. We found that receiving unemployment benefits Mauricio Avendano is an

significantly reduced the probability of reporting poor health in the year associate professor in the
Department of Global Health
after job loss, by around 5 percentage points. The health-promoting and Social Medicine at King’s
effects of receiving the benefits were robust across multiple model College London and an adjunct
associate professor in the
specifications and after we accounted for preexisting differences between Department of Social and
benefit recipients and nonrecipients. Our results add to the growing body Behavioral Sciences at the
Harvard T. H. Chan School of
of literature that suggests that social policies can have unanticipated Public Health, in Boston,
health effects. Massachusetts.

A
n increasing amount of evidence
suggests that job loss can lead to
increased probability of illness
and premature death.1–7 Unemploy-
ment benefit programs, which are
intended to alleviate the financial stress of job
loss, could have unintended positive effects on
health. If the detrimental health effects of unem-
ployment are in part due to income loss and
financial insecurity, unemployment benefits
may be a mechanism to prevent or reduce some
of the negative health effects of job loss. Al-
though unemployment benefit programs are
not explicitly designed to improve health, a num-
ber of recent studies have demonstrated that
social policies not motivated by health con-
cerns—such as the Earned Income Tax Credit,
welfare reform, and the Supplemental Nutrition
Assistance Program (SNAP, formerly known as
food stamps)—have both positive and negative
consequences for health.8–11
The US Federal-State Unemployment Insur-
ance Program provides temporary wage replace-
ment for eligible workers who become unem-
ployed through no fault of their own. Each
state operates its own unemployment program
but must follow certain general rules established
by the federal government that relate to coverage
and eligibility. Most research on this program
has focused on its impacts on earnings, con-
sumption, and unemployment duration.12–14 Few
studies have examined the potential health con-
sequences of receiving unemployment benefits.
Identifying the effect of unemployment bene-
fits on health is challenging for a number of
reasons. First, people in poor health are more
likely than healthier workers to experience an
unemployment spell.15,16 Second, because there
are strict eligibility requirements to qualify for
benefits, recipients of the benefits are likely to
differ from unemployed nonrecipients in a num-
ber of key characteristics associated with health,
such as income and education.17,18
While some studies have suggested that unem-
ployment benefits may ameliorate some of the
negative health effects of job loss,19–21 they did
not fully account for preexisting differences be-
tween benefit recipients and nonrecipients. Re-
cipients are often a priori in comparatively better
health than their nonrecipient counterparts, a

F eb r u a ry 2 0 1 7 36:2 H e a lt h A f fai r s 289

Unemployment Insurance & Health
factor that is hard to control for in an analysis.
In this study we used twenty survey waves of
the Panel Study of Income Dynamics for the pe-
riod 1984–2009 to investigate the impact of re-
ceiving unemployment benefits on the probabil-
ity of reporting poor health after job loss. We
tested this hypothesis under various model spec-
ifications that attempted to adjust for the bias
introduced by preexisting differences between
benefit recipients and nonrecipients.While none
of the approaches could definitively establish
that unemployment benefits had a causal effect
on health, these methodological approaches
partly addressed selection concerns that were
present in earlier studies, and the results are
useful for policy makers and health practitioners
considering reforms to unemployment benefit
programs and similar social protection policies.
Background
Unemployment benefit programs may affect the
health of people who lose their jobs, through
several mechanisms. In the short term, benefits
compensate for the loss of earnings resulting
from job loss and help maintain consumption
levels during unemployment spells.12 This may
enable workers to purchase health-promoting
goods and services such as healthy food and
health insurance, as well as reduce some of the
psychosocial stress associated with financial
losses. However, unemployment benefits may
reduce the marginal incentive to search for a
job, increasing the incidence and duration of
nonemployment.14,22–24 This could lead to skill
depreciation and negative career effects, which
may be detrimental for health in the long run.
A few studies have examined the impact of
unemployment benefits on health, and most of
them have found a positive association. Eunice
Rodriguez19 analyzed data from the United
States, Germany, and Britain and found that un-
employed workers who received different types
of government entitlement benefits (including
unemployment benefits) reported health status
that was similar to that of full-time-employed
workers, which suggests that such support pro-
grams can buffer the health effects of job loss.
Other studies have reported protective effects of
government entitlement benefits for depression
symptoms among unemployed women,25 poor
self-rated health among workers having jobs that
require minimum and medium skills,21 and poor
mental health among unemployed workers in
Spain.26 However, as noted above, these studies
did not fully account for preexisting differences
between benefit recipients and nonrecipients.
Such preexisting differences occur because un-
employed people do not automatically qualify to
290 H e a lt h A f fai r s Fe br uary 201 7 36 :2
receive unemployment benefits: They must meet
several monetary and nonmonetary eligibility
criteria.27 People who have lost their jobs must
also file claims with state unemployment benefit
agencies to receive benefits. Because of these
various steps, not all eligible unemployed people
actually claim benefits. In fact, unemployment
benefit programs in the United States have his-
torically had low take-up rates, with 34.8 percent
of the unemployed applying for benefits in 2005
Downloaded from http://content.healthaffairs.org/ by Health Affairs on February 7, 2017 by HW Team
and only 23.9 percent actually receiving benefits,
according to data from the Current Population
Survey.18 Of the unemployed who did not apply
for unemployment benefits, 51.9 percent re-
ported that they did not apply because they be-
lieved themselves to be ineligible; 17.8 percent
did not apply because of reasons related to atti-
tude, lack of understanding, or other barriers;
and 5.3 percent did not apply because they were
retired, sick, or disabled.
Because of eligibility rules and the need to
apply for benefits, several important differences
arise between unemployed people who receive
benefits and those who do not. Compared to non-
recipients, recipients are more likely to be edu-
cated, to have been earning more money, and to
have previously received benefits.17 This makes it
particularly challenging to establish whether un-
employment benefits alone affect health, and it
introduces the concern that benefit recipients
may have been in better health than their non-
recipient counterparts before receiving benefits.
Recent studies have circumvented this prob-
lem by exploiting variations in the designs of
state unemployment benefit programs.28,29 How-
ever, these studies did not incorporate informa-
tion on receipt of benefits at the individual level,
which makes it difficult to discern whether re-
ceiving unemployment benefits plays a critical
role in the causal pathway linking job loss and
health. The present study was intended to shed
light on this question and address some of the
limitations from previous studies by applying
multiple modeling strategies and using a longi-
tudinal sample of the US population.
Study Data And Methods
Data We used data from the Panel Study of In-
come Dynamics, the longest-running longitudi-
nal household survey in the world, which collects
data on employment status, demographic char-
acteristics, and health.30 Data were collected an-
nually until 1997, after which the study shifted to
a biennial design. Our analysis was based on the
sample of unemployment spells experienced by
working-age (ages 18–65) heads of household
from the 1984–2009 survey waves (health mea-
sures were introduced in the survey in 1984).
Observations with missing data were excluded
from the analysis, which gave us a final sample of
4,247 unemployment spells; for 875 of these
spells, the unemployed person received unem-
ployment benefits (an analysis of a sample of all
unemployment spells yielded similar results).
The Panel Study of Income Dynamics mea-
sures health using the self-rated health item, a
subjective indicator that captures individuals’
perceptions of their health using Likert scales.
Respondents are asked to rate their own health
on a scale ranging from 1 to 5. We collapsed the
scale into a binary variable, in which the first
three categories (“excellent,” “very good,” and
“good”) indicated good health and the last two
categories (“fair” and “poor”) indicated poor
health. This binary indicator has been shown
to be a strong predictor of objective measures
of health, including the risk of death.31–33
We extracted data on employment status from
each survey wave. Based on available informa-
tion, we constructed binary variables that indi-
cated whether an unemployment spell had oc-
curred at some point in the previous year, and
whether the person had received unemployment
benefits during that spell.
Other variables included age, sex, race (white,
black, or other), education (high school or less,
college, or postgraduate), marital status (mar-
ried, single, separated, divorced, or widowed),
and household size. Two other individual-level
variables were lagged by two years: the binary
indicator of poor health and the natural log of
household income. Income was lagged to avoid
simultaneity with job loss. Both variables were
lagged by two years to keep the models consis-
tent when the survey changed from an annual to
a biennial design. Lagging income and health
was important in our attempt to account for
some preexisting individual characteristics that
predict both health and receipt of unemploy-
ment benefits. To control for state-specific labor-
market conditions that may affect employment
and health,34 we also used the state unemploy-
ment rate for the working-age population, calcu-
lated from the Current Population Survey.
Analysis We used linear probability models
(results were similar when we used logistic re-
gression models) to estimate the effects of receiv-
ing unemployment benefits on self-reported
health among the pool of unemployed work-
ing-age respondents, controlling for individual
characteristics (including health status and
household income before job loss) and state
characteristics. To test the robustness of our re-
sults, we estimated alternative models that were
intended to further account for preexisting dif-
ferences between recipients and nonrecipients
of unemployment benefits.
In the first alternative modeling approach,
we implemented a one-to-one nearest-neighbor
propensity score matching model.35 Propensity
score matching is a statistical matching tech-
nique that creates treatment and control groups
that comprise individuals who have comparable
observable characteristics. We matched each un-
employment benefits recipient in the Panel
Study of Income Dynamics sample of unemploy-
ment spells to an unemployed nonrecipient who
Downloaded from http://content.healthaffairs.org/ by Health Affairs on February 7, 2017 by HW Team
had similar individual-level characteristics (de-
scribed above) in the year before job loss and was
unemployed under comparable state labor-
market conditions (for more details on the
matching approach, see the online Appendix).36
Yet even matched estimates may be biased by
unobserved individual-level differences. To fur-
ther test the robustness of our results, in the
second alternative modeling approach, we esti-
mated two-stage least squares (instrumental var-
iable) models that exploited variation in the like-
lihood of receiving unemployment benefits
based on whether or not job loss resulted from
a business closure. The rationale was that busi-
ness closures are generally unrelated to the char-
acteristics of an individual worker. Since unem-
ployment insurance program rules require
benefit recipients to have lost their jobs through
no fault of their own, people who experience job
loss resulting from a business closure are more
likely to receive unemployment benefits than
people who lost their jobs for other reasons.
We therefore estimated the health effects of
receiving unemployment benefits among a sub-
sample of unemployed people who had a greater
probability of receiving unemployment benefits
for reasons that were presumably unrelated to
their previous health. We employed a two-stage
least squares modeling approach in which we
estimated the likelihood of receiving unemploy-
ment benefits by using information on whether
or not job loss was due to a business closure as an
instrumental variable. We used this approach
first on the full sample of unemployment spells
experienced by heads of household in the Panel
Study of Income Dynamics during our study pe-
riod, and then on the propensity score matched
sample (for more details on the two-stage least
squares approach, see the Appendix).36
Limitations There are a number of limitations
in our study. First, our estimated effects of un-
employment benefits are generalizable only to
the sample of heads of household in the analysis.
Second, while the propensity score matching
and two-stage least squares analyses provided
additional evidence on whether receiving bene-
fits played a role in the causal pathway linking
job loss to health, neither method could estab-
lish causality. In the case of the propensity score
F eb r u a ry 2 0 1 7 36:2 H e a lt h A f fai r s 291
Unemployment Insurance & Health

matching models, either the treatment or con- Study Results

trol groups could have been biased by unmea-
sured factors that were correlated with both
health and receiving unemployment benefits.
Similarly, the two-stage least squares analysis
estimated the effect of receiving benefits specifi-
cally among those whose probability of receiving
them was altered by having lost a job because of a
business closure. The estimates therefore re-
flected the so-called local average treatment ef-
We found important differences between recipi-
ents and nonrecipients of unemployment bene-
fits. Compared to nonrecipients, recipients were
more likely to be married, white, and male and to
have had higher household incomes (Exhibit 1)—
all of which is consistent with evidence from
official government sources.17 By contrast, non-
recipients were more likely than recipients to be
single and black. Unemployed people were more

Downloaded from http://content.healthaffairs.org/ by Health Affairs on February 7, 2017 by HW Team

fect among this particular group and may not be
generalizable to the broader unemployed popu-
lation. Nevertheless, we believe that both alter-
native approaches serve as important tests of the
relationship between benefit receipt and health.
Third, although self-rated health has been
shown to be a strong predictor of objective mea-
sures of health, including the risk of death,31–33
data on other indicators of health would have
provided a more nuanced analysis of the poten-
tial mechanisms linking receiving benefits to
health. Unfortunately, the Panel Study of Income
Dynamics did not collect detailed information on
the incidence and timing of other health out-
comes for a sufficiently long period.
likely to receive benefits if they were jobless in
states and years with high unemployment rates.
Nonrecipients were also more likely than re-
cipients to report poor health, both in the year
before job loss (21.2 percent compared to
15.3 percent, t-value = 3.99) and in the year after
job loss (25.8 percent compared to 18.4 percent,
t-value = 4.72) (Exhibit 2). Compared to recipi-
ents, a slightly higher percentage of nonreci-
pients who did not report poor health in the year
before job loss reported it in the year after job
loss (12.0 percent versus 10.9 percent; t-value:
0.94) (data not shown).
The results from simple unadjusted linear
probability models that controlled only for poor

Exhibit 1

Selected characteristics of heads of household, ages 18–65, who experienced an unemployment spell in 1984–2009, by
whether or not they received unemployment benefits
Person received
unemployment Person did not receive All unemployment
benefitsa unemployment benefitsb spells
Characteristic Mean SD Mean SD Mean SD
Male 69.0% 0.5 56.2% 0.5 58.8% 0.5
Age (years) 40.4 11.2 39.5 13.2 39.7 12.8
Marital status
Married 44.1% 0.5 31.4% 0.5 34.0% 0.5
Single 27.6 0.4 38.4 0.5 36.2 0.5
Widowed 3.4 0.2 5.1 0.2 4.8 0.2
Divorced 17.6 0.4 16.8 0.4 17.0 0.4
Separated 7.3 0.3 8.2 0.3 8.0 0.3
Race
White 51.6% 0.5 39.4% 0.5 41.9% 0.5
Black 41.1 0.5 56.2 0.5 53.1 0.5
Other 6.9 0.3 3.7 0.2 4.3 0.2
Education
High school or less 72.8% 0.4 76.9% 0.4 76.1% 0.4
College 26.3 0.4 21.5 0.4 22.5 0.4
Postgraduate 0.8 0.1 1.6 0.1 1.4 0.1
Household size 2.9 1.6 2.7 1.7 2.8 1.7
Household income in year before
unemployment spell $38,149 $31,121 $30,133 $43,496 $31,783 $41,377
Working-age state unemployment
rate in year of unemployment spell 5.1% 1.6 4.7% 1.6 4.8% 1.6

SOURCE Authors’ analysis of data for 1984–2009 from the sample of unemployment spells in the Panel Study of Income Dynamics and
from the Current Population Survey. NOTES All income amounts are in 1999 dollars. SD is standard deviation. a20.6% of the
unemployment spell sample. b79.4% of the unemployment spell sample.

292 H e a lt h A f fai r s Fe br uary 201 7 36 :2

 Exhibit 2
Percentages of recipients and nonrecipients of
unemployment benefits who reported poor health before
and after job loss
SOURCE Authors’ analysis of data for 1984–2009 from the sam-
ple of unemployment spells in the Panel Study of Income Dynam-
ics.
health in the year before job loss suggested that
receiving unemployment benefits was associated
with a significant reduction of 4.6 percentage
points in the probability of reporting poor health
(data not shown). Those results were consistent
with the results from model 1—a linear probabil-
ity model that controlled for poor health in the
year before job loss and other factors (marital
status, race, education, age, sex, household size,
and household income in the year before job
loss; state unemployment rates; and state and
year fixed effects). The results from model 1 in-
dicated that receiving unemployment benefits
was associated with a significant 4.7-percent-
age-point reduction in the probability of report-
Exhibit 3
Estimated effects of receiving unemployment benefits on the probability of poor health
Linear probability model 1
All unemployment
spells
Received unemployment benefits −0.0466***
Poor health in the year before job loss 0.437***
Observations 4,247
R-squared 0.237
SOURCE Authors’ analysis of data for 1984–2009 from the sample of unemployment spells in the Panel Study of Income Dynamics and
from the Current Population Survey. NOTES The models included marital status, race, education, age, sex, number in household, and
logged real household income; state unemployment rates; and state and year fixed effects. The matched sample is explained in the text,
along with details about the linear probability model. SE is robust standard error. *p < 0:10 ***p < 0:01
ing poor health (95% confidence interval: −7.5,
−1.8) (Exhibit 3) (full results from all models can
be found in Appendix Exhibit A1).36
A potential concern is that ex ante differences
between recipients and nonrecipients of unem-
ployment benefits could have biased the results,
even after observable individual- and state-level
characteristics were controlled for. We therefore
estimated propensity score matching models.
This left us with a matched sample of unemploy-
Downloaded from http://content.healthaffairs.org/ by Health Affairs on February 7, 2017 by HW Team
ment spells that did not reveal significant differ-
ences between recipients and nonrecipients in
observable individual characteristics in the year
before job loss (Appendix Exhibit A2).36 The
standardized bias was reduced considerably
across the sample and across all covariates (Ap-
pendix Exhibits A3 and A4).36
When we estimated the effects of receiving
unemployment benefits based on the propensity
score matched sample, the fully adjusted linear
probability model (model 2) indicated that re-
ceiving unemployment benefits reduced the
probability of reporting poor health by 3.0 per-
centage points (95% CI: −6.6, 0.5) (Exhibit 3).
We found no significant difference in the esti-
mated effects of receiving benefits between the
two models shown in Exhibit 3, since the 95 per-
cent confidence intervals estimated from the
propensity score matched sample overlapped
with those estimated from the full sample.
As an additional robustness check, we estimat-
ed two-stage least square models that examined
the effect of receiving unemployment benefits
among those whose likelihood of receiving ben-
efits was influenced by the fact that they had lost
their jobs because of a business closure. Results
from the first stage indicated that workers who
lost their jobs because of a business closure were
significantly more likely than others to receive
benefits. Among the sample of all unemploy-
ment spells, when we controlled for individual
characteristics, we found that losing a job be-
cause of a business closure increased the proba-
Linear probability model 2
Propensity score
SE matched sample SE
(0.0147) −0.0304* (0.018)
(0.0184) 0.383*** (0.0327)
1,750
0.197
F eb r u a ry 2 0 1 7 36:2 H e a lt h A f fai r s 293
Unemployment Insurance & Health

bility of receiving unemployment benefits nonrecipients are themselves of concern, as they

significantly—by 15.8 percentage points (Appen-
dix Exhibit A5).36 However, workers who lost
their jobs because of a business closure did not
differ markedly from workers who lost their jobs
for other reasons, in terms of health before job
loss and other observable characteristics (Ap-
pendix Exhibit A6).36
In line with our original models, receiving un-
employment benefits significantly reduced the
suggest that unemployment benefits dispropor-
tionately favor socioeconomically advantaged
workers more than vulnerable workers with low-
er socioeconomic status.
Unemployment benefits may affect health
through income (by helping maintain consump-
tion patterns or reducing financial stress) or
through time (by subsidizing leisure). Income
is a well-known health determinant,38 and there

Downloaded from http://content.healthaffairs.org/ by Health Affairs on February 7, 2017 by HW Team

probability of poor self-reported health in results are a multitude of ways by which it could affect
from the second-stage estimates (Exhibit 4). Al- health. For example, income may allow people to
though the point estimates were large in magni- consume healthy goods and services, such as
tude, the estimates were less precise than and did fruit and vegetables that are often more expen-
not differ significantly from those in our original sive than unhealthy food.39 Income derived from
two models (Exhibit 3). Given the lack of preci- unemployment benefits may also enable unem-
sion, the magnitude of the estimated effect
should be interpreted cautiously, and emphasis
should be placed on the direction of the effect. Exhibit 4
Estimates also might not be generalizable to the
broader unemployed sample, since they reflect Estimated effects of receiving unemployment benefits on
the local average treatment effect among the the probability of reporting poor health
business closure subsample of unemployment
spells. Overall, however, results from two-stage
least squares models are consistent with those
from the other modeling approaches and sug-
gest that receiving unemployment benefits was
associated with better health among workers
experiencing job loss.

Discussion
Estimating the health effects of receiving unem-
ployment benefits is challenging because benefit
recipients are often a priori better off than non-
recipients. Inferring causal effects by comparing
the health of recipients to that of nonrecipients
requires great care. In this study we used a variety
of modeling strategies to examine the impact of
receiving unemployment benefits on the health
of the unemployed. Although we still cannot con-
firm that there is a causal link between receiving
benefits and health, our estimates consistently
indicated that, compared to unemployed people
who did not receive benefits during their unem-
ployment spell, those who did receive benefits
were at lower risk of reporting poor health in the
year following job loss.
Our objective was to examine whether receiv-
ing unemployment benefits might influence the
health of the unemployed. Yet the preexisting
differences between benefit recipients and non- SOURCE Authors’ analysis of data for 1984–2009 from the sam-
recipients in health, wealth, and education are ple of unemployment spells in the Panel Study of Income Dynam-
ics and from the Current Population Survey. NOTES The full sam-
policy relevant, as they indicate significant in- ple (all unemployment spells) and propensity score matched
equalities in access to benefits. Unemployment sample are explained in the text, along with details of the linear
benefits help maintain recipients’ consumption probability model and two-stage least squares model. Error bars
indicate 95 percent confidence intervals. The models included
levels and provide an opportunity to search for
marital status, race, education, age, sex, number in household,
new employment.12,37 Therefore, the observed so- and logged real household income; state unemployment rates;
cioeconomic differences between recipients and and state and year fixed effects.

294 Health Affai rs Feb ruary 2017 36 :2

ployed people to obtain health care. In our sam-
ple, most people who experienced job loss were
also likely to lose access to their employer-based
health insurance. Although people who lose
their jobs are able to keep their employer-spon-
sored coverage under the Consolidated Omnibus
Budget Reconciliation Act (COBRA), they are
responsible for paying the full premium, which
makes insurance accessible only to those with
financial liquidity. One study found that only
ing that receiving unemployment benefits im-
proved self-rated health suggests that income
losses and financial uncertainty are potential
mechanisms through which this occurs. In the
absence of benefits, some unemployed people
may feel distressed or be unable to pay for
health-promoting goods and services. Alterna-
tively, receiving unemployment benefits may
help the unemployed cope with some of the
stress associated with financial insecurity.

Downloaded from http://content.healthaffairs.org/ by Health Affairs on February 7, 2017 by HW Team

14 percent of eligible people maintained their
employer-based insurance coverage in 2010,
while 57 percent became uninsured.40
Alternatively, income-related health effects of
receiving unemployment benefits may occur
through a pathway that is not related to con-
sumption but is still a result of the short-term
income subsidy provided by benefits. For exam-
ple, it is possible that unemployment benefits
may have an independent psychological effect
by providing comfort and security to job losers.
In addition to income, there are alternative
explanations for the impact of receiving unem-
ployment benefits on health. Michael Gross-
man’s canonical model of demand for health
posits that demand for time-intensive health-
promoting activities will increase as the price
of engaging in these activities decreases.41 Time
spent working increases income, which allows
people to purchase health inputs such as healthy
food. But at the same time, working reduces the
time available to invest in health-promoting ac-
tivities such as exercise, and it may even harm
health as a result of exposure to adverse working
conditions. However, people who are not work-
ing may have more leisure time available that can
be used for health-promoting, time-consuming
activities such as exercise. Unemployment ben-
efits may therefore protect health by subsidizing
time out of work and providing the unemployed
with additional time to engage in health-promot-
ing leisure activities. This notion is consistent
with research suggesting that unemployment
benefits may lengthen unemployment duration
by underwriting leisure time.22,23
Our results also offer insight into the potential
mechanisms linking job loss to health. The find-
Conclusion
We found evidence that receiving unemployment
benefits may have positive effects on the health
of unemployed people. Policy makers should use
these results to increase the uptake of unemploy-
ment benefits. One recent policy reform intro-
duced in many states aimed to increase access to
the benefits by altering the base period used to
calculate eligibility, but that reform had limited
impacts on take-up of benefits.42 This may be
partly because the stigma associated with claim-
ing unemployment benefits makes many eligible
people choose not to apply—a choice that high-
lights the need to change attitudes about bene-
fits. In addition, about half of unemployed peo-
ple are unaware of their eligibility for benefits.18
Increasing awareness of benefit rules would be
crucial to ensuring that the program reaches
those in greatest need.
As unemployment rates rose during the Great
Recession in the last decade, the US government
responded with an unprecedented extension of
the standard twenty-six-week duration of un-
employment benefits to a maximum of ninety-
nine weeks.43 This Emergency Unemployment
Compensation program expired at the end of
2013. There was considerable debate in Congress
around the time of expiration about whether to
continue the extension, but there is no evidence
that the health effects of maintaining unemploy-
ment benefits were taken into account.44 Our
findings suggest that policy makers should con-
sider the potential health consequences of future
extensions of unemployment benefits and of
benefit cuts and program reforms. ▪

This work was primarily supported by on Aging (Grant Nos. R01AG037398 and a seed grant from the Robert Wood
the European Research Council (Grant R01AG040248) and the McArthur Johnson Foundation.
No. 263684). Mauricio Avendano was Foundation Research Network on Aging.
also supported by the National Institute Part of this work was also supported by

NOTES
1 Strully KW. Job loss and health in the
U.S. labor market. Demography.
2009;46(2):221–46.
2 Gathergood J. An instrumental var-
iable approach to unemployment,
psychological health and social
norm effects. Health Econ. 2012;
22(6):643–54.
3 Martikainen PT, Valkonen T. Excess
mortality of unemployed men and
women during a period of rapidly
increasing unemployment. Lancet.
1996;348(9032):909–12.
4 Catalano R, Goldman-Mellor S,
Saxton K, Margerison-Zilko C,

F eb r u a ry 2 0 1 7 36:2 H e a lt h A f fai r s 295

Unemployment Insurance & Health
Subbaraman M, LeWinn K, et al. The
health effects of economic decline.
Annu Rev Public Health. 2011;32:
431–50.
5 Huijts T, Reeves A, McKee M,
Stuckler D. The impacts of job loss
and job recovery on self-rated health:
testing the mediating role of finan-
cial strain and income. Eur J Public
Health. 2015;25(5):801–6.
6 Schröder M. Jobless now, sick later?
Investigating the long-term
consequences of involuntary job loss
on health. Adv Life Course Res.
2013;18(1):5–15.
7 Riumallo-Herl C, Basu S, Stuckler D,
Courtin E, Avendano M. Job loss,
wealth and depression during the
Great Recession in the USA and
Europe. Int J Epidemiol. 2014;43(5):
1508–17.
8 Snyder SE, Evans WN. The effect of
income on mortality: evidence from
the Social Security notch. Rev Econ
Stat. 2006;88(3):482–95.
9 Almond D, Hoynes HW,
Schanzenbach DW. Inside the War
on Poverty: the impact of food
stamps on birth outcomes. Rev Econ
Stat. 2011;93(2):387–403.
10 Bitler MP, Gelbach JB, Hoynes HW.
Welfare reform and health. J Hum
Resour. 2005;40(2):309–34.
11 Schmeiser MD. Expanding wallets
and waistlines: the impact of family
income on the BMI of women and
men eligible for the Earned Income
Tax Credit. Health Econ. 2009;
18(11):1277–94.
12 Gruber J. The consumption
smoothing benefits of unemploy-
ment insurance. Am Econ Rev.
1997;87(1):192–205.
13 Card D, Chetty R, Weber A. The spike
at benefit exhaustion: leaving the
unemployment system or starting a
new job? Am Econ Rev. 2007;97(2):
113–8.
14 Katz LF, Meyer BD. The impact of
the potential duration of unemploy-
ment benefits on the duration of
unemployment. J Public Econ. 1990;
41(1):45–72.
15 Böckerman P, Ilmakunnas P. Un-
employment and self-assessed
health: evidence from panel data.
Health Econ. 2009;18(2):161–79.
16 Burgard SA, Brand JE, House JS.
Toward a better estimation of the
effect of job loss on health. J Health
Soc Behav. 2007;48(4):369–84.
17 Government Accountability Office.
Unemployment insurance: factors
associated with benefit receipt [In-
ternet]. Washington (DC): GAO;
2006 Mar [cited 2016 Dec 15]. (Re-
port No. GAO-06-341). Available
from: http://www.gao.gov/new
.items/d06341.pdf
18 Vroman W. Unemployment insur-
ance recipients and nonrecipients in
296 Health A ffairs F eb r u a ry 2 0 1 7 36:2
the CPS. Monthly Labor Review
[serial on the Internet]. 2009 Oct
[cited 2016 Dec 15]. Available from:
https://www.bls.gov/opub/mlr/
2009/10/art4full.pdf
19 Rodriguez E. Keeping the unem-
ployed healthy: the effect of means-
tested and entitlement benefits in
Britain, Germany, and the United
States. Am J Public Health. 2001;
91(9):1403–11.
20 Rodriguez E, Lasch K, Mead JP. The
potential role of unemployment
benefits in shaping the mental
health impact of unemployment. Int
J Health Serv. 1997;27(4):601–23.
21 McLeod CB, Hall PA, Siddiqi A,
Hertzman C. How society shapes the
health gradient: work-related health
inequalities in a comparative per-
spective. Annu Rev Public Health.
2012;33:59–73.
22 Chetty R. Moral hazard versus li-
quidity and optimal unemployment
insurance. J Polit Econ. 2008;
116(2):173–234.
23 Moffitt R, Nicholson W. The effect of
unemployment insurance on unem-
ployment: the case of federal sup-
plemental benefits. Rev Econ Stat.
1982;64(1):1–11.
24 Krueger AB, Mueller A. Job search
and unemployment insurance: new
evidence from time use data. J Public
Econ. 2010;94(3–4):298–307.
25 Rodriguez E, Frongillo EA, Chandra
P. Do social programmes contribute
to mental well-being? The long-term
impact of unemployment on de-
pression in the United States. Int J
Epidemiol. 2001;30(1):163–70.
26 Artazcoz L, Benach J, Borrell C,
Cortès I. Unemployment and mental
health: understanding the inter-
actions among gender, family roles,
and social class. Am J Public Health.
2004;94(1):82–8.
27 Department of Labor, Employment
and Training Administration. Chap-
ter 3, Monetary Entitlement 2009
[Available from: http://workforce
security.doleta.gov/unemploy/pdf/
uilawcompar/2009/monetary.pdf
28 Cylus J, Glymour MM, Avendano M.
Health effects of unemployment
benefit program generosity. Am J
Public Health. 2015;105(2):317–23.
29 Cylus J, Glymour MM, Avendano M.
Do generous unemployment benefit
programs reduce suicide rates? A
state fixed-effect analysis covering
1968–2008. Am J Epidemiol.
2014;180(1):45–52.
30 Panel Study of Income Dynamics,
public use dataset. Produced and
distributed by the Survey Research
Center, Institute for Social Research,
University of Michigan, Ann Arbor,
MI (2015).
31 Idler EL, Benyamini Y. Self-rated
health and mortality: a review of
twenty-seven community studies. J
Health Soc Behav. 1997;38(1):21–37.
32 Liang J. Self-reported physical health
among aged adults. J Gerontol. 1986;
41(2):248–60.
33 Burström B, Fredlund P. Self rated
health: is it as good a predictor of
subsequent mortality among adults
in lower as well as in higher social
classes? J Epidemiol Community
Health. 2001;55(11):836–40.
34 Ruhm CJ. Are recessions good for
your health? Q J Econ. 2000;
Downloaded from http://content.healthaffairs.org/ by Health Affairs on February 7, 2017 by HW Team
115(2):617–50.
35 Heckman JJ, Ichimura H, Todd PE.
Matching as an econometric evalu-
ation estimator: evidence from eval-
uating a job training programme.
Rev Econ Stud. 1997;64(4):605–54.
36 To access the Appendix, click on the
Appendix link in the box to the right
of the article online.
37 Acemoglu D. Good jobs versus bad
jobs. J Labor Econ. 2001;19(1):1–21.
38 Kawachi I, Adler NE, Dow WH.
Money, schooling, and health:
mechanisms and causal evidence.
Ann N Y Acad Sci. 2010;1186:56–68.
39 Rao M, Afshin A, Singh G,
Mozaffarian D. Do healthier foods
and diet patterns cost more than less
healthy options? A systematic review
and meta-analysis. BMJ Open.
2013;3(12):e004277.
40 Collins SR, Doty MM, Robertson R,
Garber T. Help on the horizon: how
the recession has left millions of
workers without health insurance,
and how health reform will bring
relief: findings from the Common-
wealth Fund Biennial Health Insur-
ance Survey of 2010 [Internet]. New
York (NY): Commonwealth Fund;
2011 Mar [cited 2016 Dec 15].
Available from: http://www
.commonwealthfund.org/~/media/
files/publications/fund-report/
2011/mar/1486_collins_help_on_
the_horizon_2010_biennial_
survey_report_final_v2.pdf
41 Grossman M. On the concept of
health capital and demand for
health. J Polit Econ. 1972;80(2):
223–55.
42 Gould-Werth A, Shaefer HL. Do al-
ternative base periods increase un-
employment insurance benefit re-
ceipt among low-educated
unemployed workers? J Policy Anal
Manage. 2013;32(4):835–52.
43 Executive Office of the President.
Unemployment insurance exten-
sions and reforms in the American
Jobs Act [Internet]. Washington
(DC): White House; 2011 Dec [cited
2016 Dec 15]. Available from:
https://www.whitehouse.gov/sites/
default/files/ui_report_final.pdf
44 Peters JW. Senate fails to pass three-
month extension of jobless aid. New
York Times. 2014 Feb 6.