Short-run Effects of Job Loss on Health Conditions, Health

NBER WORKING PAPER SERIES

SHORT-RUN EFFECTS OF JOB LOSS ON HEALTH CONDITIONS, HEALTH INSURANCE,

AND HEALTH CARE UTILIZATION

Jessamyn Schaller

Ann Huff Stevens

Working Paper 19884

http://www.nber.org/papers/w19884

NATIONAL BUREAU OF ECONOMIC RESEARCH

1050 Massachusetts Avenue

Cambridge, MA 02138

February 2014

The views expressed herein are those of the authors and do not necessarily reflect the views of the

National Bureau of Economic Research.

At least one co-author has disclosed a financial relationship of potential relevance for this research.

Further information is available online at http://www.nber.org/papers/w19884.ack

NBER working papers are circulated for discussion and comment purposes. They have not been peer-

reviewed or been subject to the review by the NBER Board of Directors that accompanies official

NBER publications.

to exceed two paragraphs, may be quoted without explicit permission provided that full credit, including

Short-run Effects of Job Loss on Health Conditions, Health Insurance, and Health Care Utilization

Jessamyn Schaller and Ann Huff Stevens

NBER Working Paper No. 19884

February 2014

JEL No. I1,J63

ABSTRACT

Job loss in the United States is associated with long-term reductions in income and long-term increases

in mortality rates. This paper examines the short- to medium-term changes in health, health care access,

and health care utilization after job loss that lead to these long-term effects. Using a sample with more

than 9800 individual job losses and longitudinal data on a wide variety of health-related measures

and outcomes, we show that job loss results in worse self-reported health, including mental health,

but is not associated with statistically significant increases in a variety of specific chronic conditions.

Among the full sample of workers, we see reductions in insurance coverage, but little evidence of

reductions in health care utilization after job loss. Among the subset of displaced workers for whom

the lost job was their primary source of insurance we do see reductions in doctor’s visits and prescription

drug usage. These results suggest that access to health insurance and care may be an important part of

the health effects of job loss for some workers. The pattern of results is also consistent with a significant

role for stress in generating long-term health consequences after job loss.

Jessamyn Schaller

Department of Economics

Eller College of Management

University of Arizona

McClelland Hall, Suite 401

1130 E Helen St., Tucson, AZ 85721

[email protected]

Ann Huff Stevens

Department of Economics

One Shields Avenue

University of California, Davis

Davis, CA 95616

and NBER

[email protected]

Over the course of the Great Recession more than 8 million Americans lost their jobs. A growing

body of research in economics points to adverse and long-lasting consequences of worker displacement.

These include significant decreases in lifetime earnings, reduced job stability, increased likelihood of

divorce, lower fertility, and negative impacts on health, education, and labor market outcomes for the

children of the displaced.

A particularly grim addition to this literature is the finding that job loss

results in sharply increased mortality for US workers. Sullivan and von Wachter (2009) estimate that

displacement is associated with a 10-15% increase in a worker’s annual death hazard – an implied loss in

life expectancy of 1-1.5 years for an individual displaced at age 40.

Despite the dramatic magnitude of the mortality effects of job loss, little is known about why job

loss leads to increased mortality. Many potential pathways exist, including reduced income and access

to health insurance, reductions in continuity of care due to changes in insurance coverage or geographic

mobility, increased prices of health care as a result of insurance loss or changes, and the substantial

stress associated with reduced and variable earnings following job displacement. Persuasively

identifying how job loss leads to reduced health is made difficult by the fact that poor health may lead

to selection for displacement, or to difficulty becoming re-employed, both of which can raise doubts

about causality in cross-sectional comparisons, even with large sample sizes and many controls. These

factors make it critical to utilize methods—often based on longitudinal data—that can more clearly

establish the direction of causality between job loss and health outcomes. The multiple potential

See, among others, Ruhm (1991), Jacobson, Lalonde and Sullivan (1993), Stevens (1997), Charles and Stephens

(2004), Lindo (2009), Lindo (2011), Stevens and Schaller (2011), Oreopoulos, Page, and Stevens (2008).

pathways by which job loss may affect health also make it important to consider a variety of outcomes,

including health conditions, utilization of medical care, and insurance coverage, in the same study and

estimation framework.

This paper combines short panels from the Medical Expenditure Panel Survey (MEPS) covering

the years 1996 through 2011 to study the effects of job loss on health conditions, health insurance

coverage, and health care utilization and expenditures. The MEPS is uniquely suited for analysis of the

health effects of displacement and allows us to improve on existing research in two ways. First, by

combining many panels of the MEPS data we generate larger sample sizes of displaced workers than are

typically available from non-administrative data, with almost 10,000 involuntary job losses. At the same

time, the MEPS contains detailed longitudinal information on a wide set of health outcomes, health

insurance coverage and its source, and detailed (self-) reports of expenditures on and utilization of

health care. This allows us to push the literature on job loss and health forward by presenting estimates

that more convincingly isolate the causal impact of job loss on health outcomes, comparing how these

estimates are affected by specification and treatment of the control group, and comparing effects on

conditions, insurance, and utilization for a common sample.

Using models that include both individual fixed effects and baseline health measures interacted

with trends, we find that involuntary job loss is associated with increased likelihood of self-reported fair

or poor health in the short-run, as well as significant declines in self-reported mental health, depression,

and anxiety. We see no robust, significant effects on the incidence of specific chronic physical conditions

(diabetes, arthritis, hypertension or high cholesterol) within the first two years following job loss.

Turning to estimates of the effects of displacement on insurance status and health care

utilization, we find that job loss is associated with a 9.9 percentage point decrease in the probability of

being covered by any health insurance and a 12.8 percentage point increase in the probability of

experiencing a spell of non-insurance in the current interview round. While we find that job loss has no

statistically significant effect on health care utilization and expenditures for the sample as a whole, we

do find that job loss negatively impacts utilization among the sample of workers who held insurance

through their employer prior to the job loss. Among this sample with larger insurance and utilization

effects, as in the full sample, we again find significant negative effects of job loss on self-reported health

status and mental health outcomes, but no significant effects on the likelihood of reporting any specific

chronic conditions. There is no evidence that greater incidence of uninsurance in this subsample leads to

larger effects on health outcomes.

Taken as a whole, our results suggest that job loss has significant negative effects on the health

of displaced workers in the short run, with the largest effects on mental health and self-reported health

status, and no measurable effects on the incidence of specific chronic conditions like arthritis, diabetes,

cholesterol, and hypertension. Loss of insurance seems to drive reductions in health care utilization, but

does not seem to result in more pronounced negative effects on self-reported health or mental health

problems.

I. Related Literature

The background for this study and many other recent studies comes from early work that

establishes the large and permanent effects of job loss on earnings (Jacobson, Lalonde, and Sullivan

(1993), Couch and Placzek (2010)) and, more recently, the large effects of job loss on mortality.

Specifically, Sullivan and von Wachter (2009) show that displaced workers are 10-15% more likely to die

in the two decades following job loss.

These results are concentrated among workers losing jobs before

Smaller mortality effects have been found using European administrative data (Rege, Telle, and Votruba (2009),

Eliason and Storrie (2009), Browning and Heinesen (2012)).

age 50, with older displaced workers facing much smaller mortality effects in the aftermath of job loss.

Sullivan and von Wachter also show that the extent of the mortality effects seem to be correlated with

the size of earnings losses and the volatility of earnings after job loss. Unfortunately, the administrative

data utilized by Sullivan and von Wachter do not contain any information on the cause of death, or the

types of underlying health changes that could lead to increased mortality.

These findings have sparked several recent investigations into short- and medium-run health

changes that result from displacement. Findings using data from the United States (where evidence on

negative earnings and mortality effects from job loss is strongest) produce mixed results. Salm (2009)

uses the Health and Retirement Study (HRS) and a sample of workers ages 50 and over and finds little

evidence of negative effects of job loss on mental or physical health outcomes. This lack of effect for

older workers (average age in Salm’s sample is 55 years) is consistent with earlier work by Sullivan and

von Wachter (2009) who argue that older workers may be close enough to eligibility for retirement

benefits that the earnings losses and uncertainty that follow displacement are mitigated and with them,

many of the negative health effects. Salm also has a relatively small sample (around 370) of displaced

workers (defined as those losing jobs due to firm closings) and includes only two observations per

worker—one pre- and one post-displacement. Other work using the HRS (Gallo, et al. (2000), for

example) faces similar sample size and age limitations.

Using data from the Panel Study of Income Dynamics, Strully (2009) finds that job loss is

associated with a higher likelihood of self-reported fair or poor health and increased onset of new

health conditions. Her sample includes a broader set of ages, but includes only 200 workers losing jobs

due to a firm closing. She finds some evidence of negative effects on self-reported health and increases

in the onset of new conditions after job loss. These results include controls for baseline health reports,

but look only at a single point in time after the job loss.

Several studies have utilized large administrative data sets from European countries that contain

large numbers of displaced workers, but typically lack repeated, longitudinal measures of health

outcomes or other health measures. Black, Devereux, and Salvanes (2012), for example, have

longitudinal data on earnings and employment for all of Norway, but observe health outcomes for

individuals only at a single point in time. They use a rich set of individual characteristics to control for

differences between displaced and not-displaced individuals, but are unable to control for individual

fixed-effects in health measures or use longitudinal data on health. Black, Devereux, and Salvanes find

that job loss results in small increases in smoking and reductions in an index of cardiac health. They also

show, consistent with previous work on job loss in Norway, that the earnings effects of job loss are

much smaller in Norway and that Norwegian workers have access to generous, long-term

unemployment insurance. Thus it is unclear whether we would expect very large associated health

effects in this setting. They do not have information on health insurance coverage or healthcare

utilization. Other papers have also pointed to relatively modest health effects after job loss using

European data (see Browning et al. (2006), Browning and Heinesen (2012)).

The findings of relatively small health effects in Europe could reflect that job loss may be less of

an economic shock in Europe than in the United States, given more generous social insurance and

employment stability policies. It could also reflect broader availability of health insurance not tied to

employment. Overall, the limitations of existing studies leave open the question of how, in the United

States at least, the economic shock associated with job loss is translated into increased mortality over

several decades.

An exception to the pattern of limited health effects using European data comes from Bergemann,

Gronqvist, and Gudbjornsdottir (2011) who use data from Sweden and find increased rates of diabetes

after job loss among certain groups.

Another limitation of the existing literature is that alternative health-related outcomes have

been analyzed in isolation. The above studies focus on health conditions or self-reported health, while

separate studies have looked at effects on insurance coverage (Gruber and Madrian (1997), Lin (2005)).

Job loss generates reductions in insurance coverage, and may change access and costs of care for many,

and these could be related to short- and long-term changes in health. Only by analyzing health

conditions, insurance, and health care utilization in a unified study are we likely to make progress on

understanding the full health effects of job loss.

Given this backdrop, our paper makes an important contribution by combining U.S. data on job

losses, which we know are associated with both large earnings losses and large mortality effects, with a

variety of health-related outcome measures. In addition, because we have longitudinal data and

repeated observations (up to five observations per worker) of our health measures, we can use more

robust models to control for unobserved characteristics that may be correlated with job loss. Specifically

we include individual fixed-effects and baseline health measures interacted with trends to isolate the

effects of job loss on several key health outcomes. We can also compare results across displaced

workers with different baseline insurance sources and status, to gain clues about the likely drivers of the

health effects of job loss.

II. Statistical models of the health effects of job loss

Our approach to understanding the health effects of job loss is motivated by the well-

established empirical models of job loss and earnings, but also informed by a standard health

production function. Specifically, many previous authors (Jacobson, Lalonde, Sullivan 2003, Stevens

1997, Schmieder, von Wachter and Bender, 2012) model earnings effects of job loss using specifications

such as:

(1a) 



 



 















 







 



(1b) 



 



 



  















 







 



The log of earnings (E

) for person i in year t is modeled as a function of individual fixed-effects (α

) and a

vector of dummy variables indicating years before and after the job loss (D

). In some specifications, an

individual-specific trend is also included, as shown in equation (1b). Additional control variables (X

) may

be included, such as higher order terms in age or experience, along with a set of calendar year dummies

). The individual fixed effects and trends are necessary to isolate the effects of displacement from pre-

existing levels or trends in earnings that may be correlated with job loss. Such specifications use

observed earnings prior to job loss to establish the counterfactual pattern of earnings if the job loss had

not occurred. A control group of never-displaced workers helps to estimate the common (to displaced

and non-displaced workers) age and calendar year terms and identify the counterfactual pattern of

earnings.

Using this framework to estimate the effects of job loss on health requires consideration of a

simple health production function, as originally suggested by Grossman (1972). Specifically, let an

individual’s health stock in some period be expressed as a function of their lagged health, investment in

health this period (I), and depreciation (at rate θ) of health this period.

(2) 



   







 



(3)    , E)

Investment in health is in itself a function of market-based health care inputs (M), an individual’s time

devoted to improving health (T), and an individual’s human capital (E), which may affect the efficiency

with which purchases of medical care or time investments produce next period’s stock of good health.

Considering these inputs into health production jointly with the direct effects of job loss on

employment, earnings, and health insurance coverage, it is easy to see how job loss can alter health.

Earnings losses and employment changes may reduce spending on health care inputs (M). Loss of

insurance may increase the price of health care inputs and so further reduce the quantity of health

inputs. Time allocation towards health (T) may change after displacement, although the direction of this

change is not clear. Less time may be spent in market work initially, but job changes, stress and

uncertainty may result in less time being allocated toward health after displacement. Combining these

factors and replacing earnings with health on the left-hand-side of equation (1a) or (1b), a dummy

variable indicating displacement is likely be associated with reduced health as the result of

displacement’s effects on health expenditures and (possibly) time investments.

An important element of previous studies of displacement that we extend to our estimates of

the health production function is the use of a fixed-effects specification, or similar methods, to control

for unobservable individual characteristics. A primary concern in this case is that human capital (E),

which is a key determinant of health in our production function, is likely to have both observed and

unobserved components that are correlated with job loss probabilities. Thus, in order to control for

individual characteristics that might be correlated with earnings, human capital and ability, it is

important to include individual fixed effects, or otherwise utilize pre-displacement values of health

outcomes. Given this, our main specifications used in the analysis are given by:

(4a) 



 



 



 







 



(4b) 



 



 



  



 







 



(4c) 



 



 



  



 







 



Equation (4a) is a standard fixed-effects specification. Note that, because we have only five observations

per person, we collapse all post-displacement indicators into a single indicator for the post-displacement

periods (PD

). In addition, in most specifications we do not include a dummy for the period prior to

displacement. In specification testing we show that including the pre-displacement controls in this case

do not substantially change our results. Equation (4b) adds the individual-specific trend, allowing for the

possibility that underlying health may have a trend that is correlated with displacement. In practice,

with a relatively small number of time periods per person, we may not have sufficient power in the

longitudinal data to identify the person-specific trend. As an alternative, we estimate models in which

we specify the individual effect, 



, as a linear function of baseline (round 1) self-reported health (H

), so

that the individual-specific trend in (4b) is replaced with a summary measure of an individual’s health

status (baseline health) prior to any displacements. This specification, summarized by (4c) suggests that

health outcomes should have an individual-specific average and trend, and that the effects of job

displacement will be measured as deviations from that trend.

III. Data

The data for this analysis are from the Household Component of the Medical Expenditure Panel

Survey (MEPS), maintained by the Agency for Healthcare Research and Quality. The MEPS collects data

from a nationally representative subsample of households that participated in the prior year’s National

Health Interview Survey (conducted by the National Center for Health Statistics). The MEPS interviews

respondents five times over a period covering two calendar years, collecting data on demographic

characteristics, health status, health care utilization and expenditures, health insurance coverage,

income, and employment status. Our data include fifteen waves of the MEPS, covering the period 1996-

2011. We limit our analysis to individuals between the ages of 21 and 65 that are employed in the first

round of the survey.

The MEPS is ideally suited to this analysis for a few reasons. First, the set of medical conditions

identified in the MEPS is far broader than those used in previous studies, allowing us to study not only

general self-reported health and mental health, but also a variety of specific conditions including

depression and anxiety, diabetes, arthritis, hypertension, high cholesterol, and cancer, among others.

Second, the MEPS allows us to go beyond the health consequences of displacement and study the

mechanisms behind this relationship – changes in insurance coverage, the frequency of medical care,

and total expenditures on care. Finally, we are not limited by small sample sizes, as previous panel data

studies have been. By combining MEPS data from 1996 through 2011 we have over 460,000 person-

round observations and observe over 9800 job losses.

Our indicator for job displacement is constructed from a survey question asking the respondent

to identify the main reason why he or she changed jobs since the last round of interviews. We define

involuntary job loss to include three responses: (1) job ended; (2) business dissolved or sold; and (3) laid

off. By this definition, we identify 9,892 individuals that experience involuntary job displacement after

round 1 of the survey, and 91,459 individuals that are never displaced.

Our outcome variables fall into three categories: (1) health outcomes; (2) health insurance

status; and (3) health care utilization. Data on health outcomes are compiled from the main MEPS

household survey file and the Medical Conditions file. In the household survey, the respondent was

asked to rate the health and mental health status of each person in the family according to the following

categories: excellent, very good, good, fair, and poor. From these variables, we create indicators equal

Additional analysis, available upon request, shows that our results are very similar if we instead restrict the

sample to include only individuals that worked full time in the first round of the survey, or to individuals with one

or more years of job tenure in round 1.

to one if health and mental health were recorded to be fair or poor. The Medical Conditions file includes

specific medical conditions that were reported directly from the respondent, recorded verbatim by the

interviewer, and then coded by professional coders to fully-specified ICD-9-CM codes. Medical

conditions can appear in a given round in one of several ways. First, conditions categorized as “priority

conditions” are asked about in a condition enumeration section of the main survey and coded as

present for a specific round if they are described as “current” in the survey for that round. Second,

health conditions associated with specific events occurring in each round, such as hospital stays, doctor

visits, disability days, and prescription drug purchases, are added to the file. Finally, respondents may

report that a specific condition is “bothering them during a survey round. To preserve confidentiality,

condition codes are collapsed to 3-digit code categories. We use these 3-digit codes to identify a set of

six specific conditions: arthritis, diabetes, high cholesterol, hypertension, cancer (potentially a placebo

test for direct health effects of displacement), and anxiety or depression.

Health insurance indicators are constructed from variables identifying whether a survey

respondent was covered by a particular type of insurance at any time during each month of the panel.

The variables are equal to one if the respondent was insured at any point during the month of the

interview. In order to capture spells of uninsurance, we also generate a variable equal to one if the

respondent was without insurance for any month of the round.

Finally, data on health expenditures and health care utilization come from three separate event

file components of the MEPS – the Prescribed Medicines file, the Emergency Room Visits file, and the

Office-Based Medical Providers file. These files contain data on the number of office-based visits to a

As each of the conditions that we consider is considered a priority condition, the reporting of the conditions

should not be mechanically dependent on the utilization of care (i.e. the number of health “events” in a specific

round). However, to the extent that individuals are more likely to describe as “current” a condition for which they

have recently seen the doctor or filled a prescription, utilization of medical care and likelihood of reporting medical

conditions may be correlated. It is also possible that frequent care makes it more likely that individuals will be

diagnosed and thus become aware that they have a specific medical condition.

doctor and number of visits to the emergency room, as well as the amount paid by family, insurance,

and other sources for each medical event that occurred in the round, including the purchase of

prescription drugs. We collapse the data to the person-round level to generate total number of visits

and expenditures by category for each individual.

IV. RESULTS

Sample means, presented in Table 1, make clear the importance of controlling for observable

and unobservable characteristics that may be correlated both with the likelihood of involuntary job

displacement and with health-related outcomes. We compare characteristics for two groups: workers

that experience an involuntary displacement sometime after the first round of data collection and

workers that are never displaced.

Displaced workers are more likely to be male, more likely to be black

or Hispanic, and slightly younger than their continually-employed counterparts. They are also less-

educated: 16% of the displaced sample has less than a high school degree, compared with 10% of the

non-displaced group, and only half have attended any college, compared with 59% of the non-displaced

group.

Differences between the displaced and non-displaced groups are also apparent when

considering health-related outcomes. Notably, while displaced workers are more likely to have their

general health categorized as “Fair” or “Poor” and are more likely to experience poor mental health,

they are less likely to report several of the specific health conditions, including diabetes, high cholesterol

and hypertension. While this discrepancy may be due to differences in the prevalence of these

conditions between these two groups, it also may be due to the reduced rate of health insurance

coverage among workers that will eventually be displaced: only 68.9 percent of the displaced sample is

Observation numbers in Table 1 show that there is sample attrition between Rounds 1 and 5. We have also

conducted analysis on a balanced panel – a sample in which every observation has complete data in each round of

the panel – and find that our results are robust to this sample change.

covered by any insurance in the initial round of the sample, while 84.2 percent of the never-displaced

sample is covered. The gap in health insurance coverage may also explain differences in health care

utilization between the two groups even prior to any displacements: displaced workers are less likely to

visit the doctor’s office and less likely to use prescription drugs, and more likely to visit the emergency

room, where treatment can often be obtained without health insurance.

These large pre-existing differences in insurance status suggest that it may be critical to check

sensitivity of these results to controlling for pre-existing differences in insurance coverage between

displaced and not displaced workers. If less financially stable firms, for example, are less likely to offer

health insurance policies for their workers, there could be substantial selection of workers into firms

that are most likely to close or lay off their workers, leading to bias in estimates of the effects of job loss

that cannot control for insurance status. This could mean that those who are displaced would have

been less healthy even in the absence of the job loss, given their lower levels of access to health

coverage. In contrast, if healthier workers disproportionately select into firms that do not offer health

insurance as part of their compensation package, the resulting bias could go in the other direction. Our

inclusion of both fixed-effects and trends interacted with baseline health, which has not often been

done in previous work, reduces the likelihood that results are driven by pre-displacement differences

across workers. We also conduct some of our analysis on the subset of workers who have insurance

through their employers as of round 1, and prior to any of the displacements. This results in a better

balance of displaced and non-displaced workers and allows us to draw some suggestive conclusions

about the role of insurance in driving effects on health outcomes and utilization for this subset of

workers.

Table 2 shows results from our baseline set of health regressions. We begin by looking at the

effects of job displacement on health outcomes, defining displacement broadly to include those who

report that they were laid off, their job ended, or their business was dissolved or closed. To account for

both observable and unobservable differences between displaced and non-displaced workers, our

starting point is a standard specification, described in equation (4a) above that includes individual fixed

effects; these results are summarized in column (3) of Table 2. For comparison, column (1) of that Table

shows mean levels of each health outcome as of round 1 and column (2) shows a specification without

individual fixed-effects that instead includes controls for baseline self-reported health and mental health

and dummies for gender, race, and educational attainment categories in addition to age and calendar

year fixed effects – an approach that has been common in the previous literature. Finally, in column (4)

we show our preferred specification (based on equation (4c)) that includes both individual fixed-effects

and an interaction between round 1 self-reported health and a trend term, meant to proxy for

individual-specific trends that are correlated with underlying health status.

The results in column 4 of

Table 2 show that displacement increases the probability of having a self-reported physical health status

of fair or poor by approximately 1.5 percentage points, or roughly 18 percent of the baseline probability,

with the effect statistically different than zero at the 1 percent level. The effect is slightly larger than

effects shown in column 2, without fixed effects, and significantly larger than effects shown in column 3

that do not control for a time trend. For virtually all outcomes examined (see later rows of Table 2) we

find somewhat smaller effects of displacement in the specifications that use fixed-effects specifications.

Thus, to be conservative in quantifying the causal portion of negative health outcomes following job

loss, we rely on the specification from column 4 for the remaining results.

The next rows of Table 2 show the effects of displacement on a variety of other health

conditions. We include a number of specific chronic conditions that may be plausibly linked to job loss

via direct or indirect mechanisms. First, we consider the incidence of arthritis and diabetes, as much

We have also estimated a slightly less restricted model in which trends are interacted with five dummies for the

value of self-reported health (rather than just the continuous measure of self-reported health) and obtained very

similar results.

evidence suggests that episodes of acute stress (and more chronic stress) can be associated with onset

and increased symptoms of both diseases (Hasset and Clauw (2010), deBrouwer et al. (2010), Surweit,

Schneider and Feinglos (1991)). We also include cholesterol and hypertension – two conditions that

have not been linked directly to stress, but may reflect changes in health behaviors after displacement.

Specifically, if job loss and the associated stress leads to changes in smoking, drinking, nutrition, or

exercise, conditions likely to be affected by such behaviors may be affected as well, and we include

cholesterol and hypertension to check for this mechanism. Finally, as a placebo test for direct health

effects of displacement, we include cancer since it seems unlikely that any behavioral change (or

changes to health care) from job loss would be associated with cancer in the short (approximately two

years on average) time frame after job loss considered here.

As shown in the next two rows of Table 2, we find positive and significant increases of

approximately 7 to 8 percent in the incidence of both arthritis and diabetes following displacement but

these effects are not robust to inclusion of baseline health interacted with a trend. We find no evidence

of increases in either high cholesterol or hypertension. Finally, as expected, we find no significant

increase in the onset of cancer, which suggests that we are not picking up spurious negative changes in

health that are simply correlated with the likelihood of job loss. We do, however, find a marginally

significant reduction in reported cancer following job loss. While this is only suggestive, one possibility is

that reductions in health care utilization delay diagnosis of some conditions, like cancer, leading to the

appearance of reduced incidence when we rely on self-reports of conditions.

We also study the effects of displacement on two mental health outcomes: a self-reported

measure of mental health that is fair or poor, and a report of experiencing depression or anxiety. Mental

health effects of displacement may play an important role in mediating the effects of job loss not only

on physical health, but also on a number of other outcomes that have been examined in the literature

including marital dissolution and children’s wellbeing. These results are presented in the lower panel of

Table 2. Not surprisingly, we find that job loss leads to substantial increases in both of these measures of

mental health. The probability of self-reporting fair or poor mental health rises by 38 percent after job

loss, and reports of depression or anxiety rise by about 22 percent.

The fact that we find significant increases in the number of workers reporting poor or fair

health, but no increase in specific physical conditions is somewhat surprising. Of course, the increase in

anxiety and depression could lead workers to report worse health. It may also be that other conditions

and minor illnesses are occurring that lead to these reports. Finally, reporting a specific condition

requires that individual have been diagnosed (or at least that they have self-diagnosed) and this may

depend on access to and utilization of health care, which we explore below. Previous work shows that

specific conditions are frequently misreported by individuals (see Baker et al. (2002)), and so self-

reported summaries of health may be more reliable. It is also true that summary measures of self-

reported health have been linked to mortality (Idler and Benyamini (1997)), and so may be particularly

important here, given our interest in understanding the connections between short-term health effects

of displacement and mortality.

We next consider the extent to which displacement alters access to health insurance and health

care utilization, again using specification (4c). In Panel A of Table 3 we show the effects of displacement

on several measures of insurance, including the probability of having any insurance coverage at the

round-specific survey date, the probability of having private or public insurance coverage, and the

probability of having any period (between interviews) uninsured. The first column shows that

individuals who are displaced are approximately 10 percentage points less likely to have any source of

health insurance following job loss. Recall that the “post-displacement” period here ranges from a few

months (for those displaced between rounds 4 and 5) to just under 2 years for those displaced soon

after the round 1 interview. This effect consists of a slightly larger reduction in access to private

insurance (which shows a reduction of approximately 11 percentage points) and a statistically significant

increase of 1.3 percentage points in access to public insurance. Finally, roughly 13 percent of those

displaced experience some period without insurance in the period after displacement. These effects are

consistent with, though somewhat smaller than, previous estimates by Gruber and Madrian (1997) on

the effects of job loss on insurance status. Gruber and Madrian report a reduction in the likelihood of

having insurance after a job separation of approximately 20 percentage points. The fact that they

include only men and consider both voluntary and involuntary changes job changes in their study may

explain much of this difference.

The lower panel of Table 3 shows results summarizing the response of health care utilization

and total expenditures. Specifically, the outcomes we consider having any office-based medical visits

since the last survey round, any ER visits, or receiving any prescription medications, along with total

expenditures on office visits, ER visits and prescription drugs. This shows no evidence of changes in

either utilization or expenditures. Given that only 1 in 10 displaced workers face lasting changes in

insurance status following their job losses, it is perhaps not surprising that we see little evidence of

changes in utilization.

We next consider a subset of workers who are likely to face larger changes in insurance

coverage, and examine whether this group also has larger effects on health outcomes or utilization.

Specifically, we limit our analysis sample to the group of individuals who, as of round 1, are holding

health insurance through their own employer (“round 1 insurance holders”). If the loss of insurance

coverage drives other outcomes, such as health conditions or utilization, we would expect results for

this subsample to differ from those in Tables 2 and 3. This restriction eliminates from both our

treatment and control groups those who are uninsured at round 1, as well as those who are insured, but

whose coverage is through either a spouse’s employer or an individual policy. A potential advantage of

this approach is that it forces both treatment and control groups to be more similar in terms of

observables (such as insurance status) at round 1. While fixed-effects and baseline health trends are

meant to correct for potential baseline differences between treatments and controls, this sample

restriction may make the sample more similar in unobservable ways as well. Of course, the disadvantage

is that the effects of displacement here apply to only a subset of job losers, and that subset, given their

access to employer-based insurance in round 1, may be relatively advantaged.

Table 4 shows the effects of displacement in the sample that conditions on having own-

employer-based insurance in round 1. As expected, there are much larger effects on access to health

insurance, with a 26 percentage point reduction in the likelihood of having any insurance following job

loss. Echoing results in Table 2, the reduction in private insurance coverage from job loss is offset only

very slightly by increased access to public insurance. There is a 1.8 percentage point increase in the

probability of having public insurance as the result of displacement. This likely reflects limited eligibility

for public insurance (Medicaid) for most able-bodied adults under current law, something that could

change substantially as the Affordable Care Act is implemented in the coming years.

Consistent with the much larger reductions in insurance coverage for this subset of our sample,

we find evidence of larger changes in utilization for this group. Panel B of Table 4 shows a reduction of

3.6 percentage points (relative to a baseline mean of 48.6 percent) in the likelihood of having an office

visit since the last round. The probability of visiting the ER is reduced by approximately 1 percentage

point, relative to a mean of 4.0 percent, and the likelihood of using prescription medications during the

round is similarly reduced by 3.3 percentage points, relative to a mean of 49.2 percent. Meanwhile,

Summary statistics by displacement status for the set of Round 1 insurance holders, presented in Appendix Table

A2, show that the restricted sample is less likely to be Hispanic and more likely to be college-educated than that

sample as a whole.

there is evidence that job displacement also results in decreased expenditures on ER visits and

prescription drugs, though some of these effects are only marginally significant.

Taken together, the larger effects on insurance coverage for this sub-sample, and the

emergence of statistically significant reductions in health care utilization provide a rough guide to how

disruptions in insurance status from displacement might affect the probability of receiving medical care.

Suppose that all of the effect of displacement on health care utilization in this sample comes as the

result of reductions in insurance coverage. In particular, this assumes (contrary to the findings in Table

2) that there is no direct effect of job loss on health. Then, the results in Panels A & B of Table 4 suggest

that loss of insurance coverage reduces office visits and use of prescription medications by

approximately 15 percentage points per round. This magnitude is similar to the estimates in Currie and

Gruber (1996), for children, that suggest becoming Medicaid eligible increased the probability of an

office visit by approximately 9 percentage points per year. Of course, if displacement also produces

independent, negative effects on health (perhaps due to stress or lifestyle changes), that would tend to

increase the likelihood of medical visits and so this estimate may be a lower bound of the direct

reduction in utilization from loss of insurance.

In the final panel of Table 4 we show the effects of displacement on health conditions for the

sub-sample of round 1 employer-based insurance holders. For the self-reported health measure, this

sample shows slightly smaller increases in the probability of reporting fair or poor health, and similar

effects (as in the full sample) on fair or poor mental health, anxiety and depression. Also similar to the

full sample, there is no evidence of significant changes in the incidence of the specific chronic physical

conditions. This suggests that, at least in the short-run, it is not access to insurance and health care

utilization that drives observed changes in health conditions immediately after job loss. It may be that

reduced utilization, or reductions in continuity of care over a longer period, contribute to long-term

effects on mortality, but we do not see evidence of specific physical conditions emerging in the years

immediately after job loss.

While the relatively short panels we have available in this data limit the extent to which we can

examine the timing of health changes around displacement, we can distinguish the immediate effects of

displacement from those that occur in the next few months. In addition, we can look at health,

insurance, and utilization effects in the period just prior to job loss, to examine whether there is

evidence of effects that begin before the actual displacement. Much of the literature on earnings losses

from displacement, for example, show that some of the effects appear prior to the displacement event,

as firms begin to struggle and may reduce pay and hours. There may be similar clues to the process that

appear from looking at the timing of these health-related effects.

In Table 5 we repeat our basic analyses, but instead of a single, post-displacement indicator, we

include three variables: one for the round immediately prior to job loss, one for the round of job loss,

and one for the rounds after job loss. In Panel A, we see small reductions (of around 2 percentage

points) in insurance coverage in the round just prior to job loss, but much larger effects at the time of

job loss and after. Focusing on the availability of private insurance, for example, coverage falls by 14

percentage points in the round of the job loss, but recovers slightly to an 11 percentage-point reduction

in later rounds. Importantly, the much larger effects at and following job loss provide additional

evidence that these are direct effects of job loss, and not pre-existing characteristics of those workers

who are most likely to lose jobs.

Panel B of Table 5 examines the timing of changes in health care utilization. Once again, it is

reassuring to see no evidence of effects emerging in the period prior to the displacement. Interestingly,

for these measures, there are some small but statistically significant increases in utilization (measured

by having any office visit or any prescription drug coverage) in the round of the job loss. This may reflect

increased utilization among those who anticipate losing coverage upon displacement. These effects

turn negative, as expected, in the period after displacement.

Finally, when we look at health reports and conditions, we again see little evidence of changes in

health that begin prior to displacement. Effects on being in poor health and on reported anxiety or

depression grow very slightly from the period of displacement to later rounds, but are generally fairly

similar across the two “post-displacement” periods.

We have also estimated models that add an interaction between the displacement variable and

the reason for displacement (firm closing versus layoffs or job endings). This could be important in light

of concerns that layoffs may disproportionately affect workers with low productivity, possibly linked to

poor health. Thus, firm closings may provide a more exogenous employment shock than the full set of

displacements. If this concern is important, we might expect to see smaller effects of displacement on

health-related outcomes among those losing jobs in business closings. As summarized in Appendix

Table 1, we see little evidence that this is the case. In short, we see evidence of slightly larger effects on

self-reported and mental health indicators, and smaller effects on insurance coverage. The effects on

insurance coverage, however, seem mainly to reflect differences in the source of insurance coverage by

those whose firms close relative to those who are laid off. Displacements due to business closings may

also capture workers who lose jobs in particularly difficult local economic environments so are not

perfectly comparable to the broader set of displacements. Our findings, however, do not suggest that

laid off workers (for whom there is a greater possibility of negative, health related selection for layoff)

drive the significant negative effects we report here.

V. CONCLUSION

In this paper, we use data from the Medical Expenditure Panel Survey to examine the short-

term effects of job displacement on health outcomes, and to examine the role of health insurance

shocks and changes medical care utilization in mediating these effects. By combining 16 panels of MEPS

data, we generate a sample that includes almost 10,000 individuals that were involuntarily displaced

from their main jobs between 1996 and 2011 – a much larger sample than other non-administrative

datasets that have been used previously in the literature.

Using an empirical specification that includes both individual fixed effects and baseline health

interacted with a linear time trend to account for unobservable individual characteristics that may be

correlated with both health and the probability of job loss, we find that job displacement is associated

with significant declines in self-rated physical and mental health and increased reports of anxiety or

depression. There is no evidence of increases in specific chronic physical conditions including diabetes,

arthritis, high cholesterol or hypertension. Focusing on a restricted sample that includes only individuals

that held insurance through their employer in Round 1, we find that these individuals are more likely to

lose their insurance as a result of job loss (30.5 percent of the restricted sample experience a spell

without insurance after job loss compared with 12.8 percent of the full sample), and to reduce their

health care utilization (round-1 insurance holders experience a 7.4-percent decrease in the probability

of a doctor’s visit after displacement). However, we do not find larger health effects for the restricted

sample of Round 1 holders, which suggests that the negative health effects for the full sample are not

driven by reductions in medical care in the short-term.

The lack of immediate increases in physical health problems among those likely to lose

insurance after job loss does not rule out the possibility that these reductions in insurance and

utilization of health care over a longer time frame could have negative consequences for the longer-

term health of affected workers. Previous work, for example, has shown that access to Medicaid and

Medicare lowers health care utilization and mortality among children and the elderly (Currie and Gruber

(1996), Card, Dobkin and Maestas (2008, 2009)). For displaced workers, however, only a minority faces

a prolonged period without access to healthcare. The fact that we do see substantial reductions in

health care utilization among the subset of displaced workers most likely to lose health insurance

suggests that access to insurance, or perhaps continuity of care, could play an important role in

generating long-term effects on mortality. Given the changes in coverage in the U.S. likely to come

about as the Affordable Care Act is implemented, this suggests that an important area for study is the

relationship between insurance status and utilization in the aftermath of job loss.

This brings us to our second main finding. Perhaps the most robust effects of job loss found

here involve effects on mental health. A large literature establishes a strong association between stress

and mortality, though a causal link from stress to long-term mortality has not been established (See, for

example, Russ, et al. (2012) and citations therein). This is also consistent with arguments made by

Sullivan and von Wachter (2009) who document a relationship between the extent of post-displacement

income volatility (which may indicate prolonged uncertainty and stress) and mortality effects. While we

do not see short-term increases in physical conditions that may be influenced by stress (such as diabetes

or arthritis), prolonged mental health effects could lead to longer-term deterioration in physical health.

Job loss in the U.S. brings with it long-lasting reductions in earnings power, highly variable

earnings, and a host of related effects on the health and well-being of individuals and their families.

While the causal nature of this association has been well-established in the literature focusing on

earnings effects of job loss, prior work on job loss and health outcomes has often lacked the longitudinal

data and methods to establish causality. By controlling for individual fixed effects and health related

trends, we show that job loss also results in decreased self-reported health, small increases in some

chronic physical conditions, and marked declines in mental health. Future work should further explore

the long-term effects of the stress of job loss on mental and physical health of workers, following up on

the short-term indications found here that this could be an important link in the relationship between

job loss and long-term health.

References

Baker, M., Stabile, M. & Deri, C. (2004). What do Self-Reported, Objective Measures of Health Measure?

The Journal of Human Resources, 39(4): 1067-1093.

Black, S. E., Devereux, P. J., & Salvanes, K. G. (2012). Losing heart? The effect of job displacement on

health. National Bureau of Economic Research Working Paper No. w18660.

Bergemann, A., Grönqvist, E., & Gudbjörnsdottir, S. (2011) The effects of job displacement on the onset

and progression of diabetes. Netspar Discussion Paper 03/2011-025.

Browning, M., Moller Dano, A., & Heinesen, E. (2006) Job displacement and stress‐related health

outcomes. Health Economics, 15(10), 1061-1075.

Browning, M., & Heinesen, E. (2012). Effect of job loss due to plant closure on mortality and

hospitalization. Journal of Health Economics.

Card, David, Carlos Dobkin, and Nicole Maestas (2008). The impact of nearly universal insurance

coverage on health care utilization and health: evidence from Medicare. American Economic Review,

98(5): 2242-2258.

Card, David, Carlos Dobkin, and Nicole Maestas (2009). Does Medicare save lives?" The Quarterly Journal

of Economics 124, no. 2 (2009): 597-636.

Charles, K. K., & Stephens Jr, M. (2004). Disability, Job Displacement and Divorce. Journal of Labor

Economics, 22(2): 489-522.

Couch, K. A., & Placzek, D. W. (2010). Earnings losses of displaced workers revisited. The American

Economic Review, 100(1), 572-589.

Currie, Janet, and Jonathan Gruber (1996) "Health insurance eligibility, utilization of medical care, and

child health." The Quarterly Journal of Economics, 111(2): 431-466.

Eliason, M., & Storrie, D. (2009). “Does job loss shorten life?” Journal of Human Resources, 44(2), 277-

302.

Gallo, W. T., Bradley, E. H., Siegel, M., & Kasl, S. V. (2000). Health effects of involuntary job loss among

older workers: Findings from the health and retirement survey. The Journals of Gerontology Series B:

Psychological Sciences and Social Sciences, 55(3), S131-S140.

Grossman, M. (1972). On the concept of health capital and the demand for health. The Journal of

Political Economy, 80(2), 223-255.

Gruber, J., & Madrian, B. C. (1997). Employment separation and health insurance coverage. Journal of

Public Economics, 66(3), 349-382.

Idler, E.L., & Benyamini, Y. (1997). Self-Rated Health and Mortality: A Review of Twenty-Seven

Community Studies, Journal of Health and Social Behavior, 38(1): 21-37.

Jacobson, L. S., LaLonde, R. J., & Sullivan, D. G. (1993). Earnings losses of displaced workers. The

American Economic Review, 685-709.

Lin, E. Y. (2005). Health insurance coverage and reemployment outcomes among older displaced

workers. Contemporary Economic Policy, 23(4), 529-544.

Lindo, J. M. (2010). Are Children Really Inferior Goods? Evidence from Displacement-driven Income

Shocks. Journal of Human Resources, 45(2), 301-327.

Lindo, J. M. (2011). Parental job loss and infant health. Journal of Health Economics, 30(5), 869-879.

Oreopoulos, P., Page, M. E., & Stevens, A. H. (2006). The intergenerational effects of compulsory

schooling. Journal of Labor Economics, 24(4), 729-760.

Rege, M., Telle, K., & Votruba, M. (2009). The effect of plant downsizing on disability pension utilization.

Journal of the European Economic Association, 7(4), 754-785.

Ruhm, C. J. (1991). Are workers permanently scarred by job displacements?. The American Economic

Review, 81(1), 319-324.

Russ, Tom, Emmanuel Stamatakis, Mark Hammer, John M. Starr, Mika Kivimaki, and David Batty (2012).

Association between psychological distress and mortality: individual participant pooled analysis of

10 prospective cohort studies. BMJ 342:1-14.

Salm, M. (2009). Does job loss cause ill health?. Health Economics, 18(9), 1075-1089.

Schmieder, J. F., Von Wachter, T., & Bender, S. (2010). The long-term impact of job displacement in

Germany during the 1982 recession o n earnings, income, and employment. IAB discussion paper No

2010, 1.

Stevens, A. H. (1997). Persistent effects of job displacement: The importance of multiple job losses.

Journal of Labor Economics, 165-188.

Stevens, A. H., & Schaller, J. (2011). Short-run effects of parental job loss on children's academic

achievement. Economics of Education Review, 30(2), 289-299.

Strully, K. W. (2009). Job loss and health in the US labor market. Demography, 46(2), 221-246.

Sullivan, D., & Von Wachter, T. (2009). Job displacement and mortality: An analysis using administrative

data. The Quarterly Journal of Economics, 124(3), 1265-1306.

Table 1

Summary Statistics By Displacement Status

Never Displaced

Displaced

Demographic Characteristics

Male

0.53

0.58

Age

41.0

38.3

Black, Non-Hispanic

0.11

0.13

Other Race, Non-Hispanic

0.06

0.05

Hispanic

0.12

0.17

High School Dropout

0.10

0.16

High School Graduate

0.30

0.33

Any College

0.59

0.50

Round

Round 5

Round 1

Round 5

Physical Health

Health Fair/Poor

8.2

7.0

10.7

9.9

Arthritis

8.3

14.8

8.2

15.4

Diabetes

3.6

4.7

3.7

5.1

High Cholesterol

7.3

10.6

5.9

9.0

Hypertension

11.3

15.2

9.9

13.6

Cancer

1.7

4.1

1.2

2.9

Mental Health

Mental Health Fair/Poor

3.2

3.7

5.1

6.4

Depression/Anxiety

7.1

15.1

8.5

19.1

Insurance Coverage

Any Insurance

84.2

83.8

68.9

59.8

Private Insurance

80.7

80.3

62.4

52.0

Public Insurance

5.6

5.2

8.0

9.2

Ever Uninsured

18.5

16.9

34.8

42.5

Health Care Utilization

Any Office-Based Visit

45.3

37.2

40.5

32.5

Expenditures on Office-Based Visits

189.6

200.8

158.6

170.9

Any ER Visit

3.9

3.0

5.1

4.2

Expenditures on ER Visits

31.2

29.1

32.9

44.6

Any Prescription Drug

46.1

40.9

40.6

34.8

Expenditures on Prescription Drugs

142.1

167.9

120.0

157.2

Observations

91459

78184

9892

8927

Notes: The data are from the 1996 through 2011 waves of the Household Component of the Medical

Expenditure Panel Survey. The sample includes individuals aged 21 to 65 that are employed in the

first round of the survey. Observations are weighted using MEPS person-level sampling weights.

Table 2

Effects of Job Displacement on Physical and Mental Health Outcomes

Regression Specification

Round 1

Mean

OLS with

Baseline

Controls

Fixed Effects

With Baseline

Health*Trend

Physical Health

Health Fair/Poor

0.084

0.011***

0.004

0.015***

(0.003)

Arthritis

0.083

0.020***

0.006**

0.002

(0.004)

(0.002)

Diabetes

0.036

0.005

0.003**

0.002

(0.003)

(0.001)

High Cholesterol

0.071

0.001

-0.001

(0.004)

(0.002)

Hypertension

0.112

0.003

0.001

-0.002

(0.004)

(0.002)

Cancer

0.016

0.000

-0.002

-0.003*

(0.002)

(0.001)

Mental Health

Mental Health Fair/Poor

0.034

0.018***

0.008**

0.013***

(0.002)

(0.003)

Depression or Anxiety

0.073

0.053***

0.024***

0.016***

(0.005)

(0.003)

Notes: N=463,243. Standard errors (in parentheses) are clustered at the individual level. * p<.1 ** p<.05

*** p<.01. Estimates are weighted using MEPS person-level sampling weight. OLS regressions (column

(2)) include dummies for Round 1 health and mental health (1 = excellent, 2=very good, 3 = good, 4 =

fair, 5 = poor), race/ethnicity (non-Hispanic black, non-Hispanic other race, Hispanic), and educational

attainment (high school graduate, college graduate). All regressions include age and calendar year fixed

effects.

Table 3

Job Displacement, Insurance Status, and Health Care Utilization

Panel A: Insurance Status

(1)

(2)

(3)

(4)

Any

Insurance

Private

Insurance

Public

Insurance

Ever Uninsured

Displaced

-0.099***

-0.113***

0.013***

0.128***

(0.005)

(0.002)

(0.005)

Panel B: Health Care Utilization

(1)

(2)

(3)

(4)

(5)

(6)

Any Office

Visit

Office

Expenditures

Any ER Visit

Expenditures

Any Rx

Expenditures

Displaced

-0.005

13.780

0.000

1.506

-0.008

-9.899

(0.006)

(19.296)

(0.003)

(5.153)

(0.005)

(6.797)

Notes: N=463,243. Standard errors (in parentheses) are clustered at the individual level. * p<.1 ** p<.05 *** p<.01.

Estimates are weighted using MEPS person-level sampling weight. All regressions include age and calendar year fixed

effects and baseline health and mental health interacted with trends. The dependent variables in columns (1) –(3) of

Panel A are equal to 1 if an individual is insured in the month of the current interview. The dependent variable in

column (4) of Panel A is equal to 1 if an individual experienced a spell with no insurance coverage between the last

interview and the current interview.

Table 4

Job Displacement, Insurance Status, Utilization, and Health: Round 1 Insurance Holders Only

Panel A: Insurance

Status

(1)

(2)

(3)

(4)

Any

Insurance

Private

Insurance

Public

Insurance

Ever

Uninsure

Displaced

-0.260***

-0.283***

0.018***

0.305***

(0.008)

(0.009)

(0.003)

(0.009)

Panel B: Health Care

Utilization

(1)

(2)

(3)

(4)

(5)

(6)

Any

Office

Visit

Office

Expenditu

res

Any ER

Visit

Expenditu

res

Any Rx

Expenditu

res

Displaced

-0.036***

-1.044

-0.010*

-17.785*

-0.033***

-23.510*

(0.009)

(42.441)

(0.005)

(7.536)

(0.008)

(10.011)

Panel C: Health

Outcomes

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Health

Fair/Poor

Arthritis

Diabetes

Cholester

Hyperten

sion

Cancer

Mental

Fair/Poor

Depr/Anx

Displaced

0.011*

-0.001

0.001

-0.001

-0.003

-0.002

0.016***

0.017***

(0.005)

(0.003)

(0.002)

(0.003)

(0.002)

(0.004)

(0.005)

Notes: N=242,814. The sample includes only individuals that were covered by health insurance through their current

main job in Round 1 of the panel. Standard errors (in parentheses) are clustered at the individual level. * p<.1 ** p<.05

*** p<.01. Estimates are weighted using MEPS person-level sampling weight. All regressions include age and calendar

year fixed effects and baseline health and mental health interacted with trends.

Table 5

Job Displacement, Insurance Status, Utilization, and Health - Lags and Leads

Panel A: Insurance Status

(1)

(2)

(3)

(4)

Any

Insurance

Private

Insurance

Public

Insurance

Ever Unins

Displaced:

Next round

-0.022***

0.000

0.023***

(0.005)

(0.003)

(0.005)

This round

-0.129***

-0.143***

0.013***

0.167***

(0.007)

(0.003)

(0.007)

Prior round

-0.095***

-0.111***

0.014***

0.117***

(0.008)

(0.007)

(0.004)

(0.008)

Panel B: Health Care Utilization

(1)

(2)

(3)

(4)

(5)

(6)

Any Office

Office

Expend.

Any ER

Visit

Expend.

Any Rx

Rx Expend.

Displaced:

Next round

0.011

7.489

0.006

0.547

0.012

1.756

(0.008)

(16.768)

(0.004)

(6.534)

(0.007)

(9.388)

This round

0.028***

17.102

0.007

1.085

0.019**

14.855

(0.008)

(18.562)

(0.004)

(6.698)

(0.007)

(10.698)

Prior round

-0.023**

-33.527

0.001

2.503

-0.020**

-30.496**

(0.008)

(34.564)

(0.004)

(8.520)

(0.007)

(9.311)

Panel C: Health Outcomes

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Health

Fair/Poor

Arthritis

Diabetes

Cholestero

Hypertensi

Cancer

Mental

Fair/Poor

Depr/Anx

Displaced:

Next round

-0.000

0.005

0.000

0.001

-0.001

-0.002

0.000

0.006

(0.004)

(0.003)

(0.001)

(0.002)

(0.001)

(0.003)

This round

0.012**

0.005

0.001

-0.002

-0.003*

0.014***

0.016***

(0.004)

(0.003)

(0.001)

(0.002)

(0.004)

Prior round

0.016***

0.006

0.002

-0.002

-0.003

-0.005*

0.012***

0.022***

(0.004)

(0.002)

(0.003)

(0.002)

(0.004)

(0.005)

Notes: N = 463,243. DispN1 is an indicator for one period prior to displacement, Disp0 is an indicator for the period of displacement, and Disp1M is an indicator for 1

or more periods after displacement. Standard errors (in parentheses) are clustered at the individual level. * p<.1 ** p<.05 *** p<.01. Estimates are weighted using

MEPS person-level sampling weight. All regressions include age and calendar year fixed effects and baseline health and mental health interacted with trends.

APPENDIX

Table A1

Job Displacement, Insurance Status, Utilization, and Health by Type of Displacement

Panel A: Insurance Status

(1)

(2)

(3)

(4)

Any

Insurance

Private

Insurance

Public

Insurance

Ever

Uninsured

Laid off or

-0.113***

-0.128***

0.013***

0.145***

Job Ended

(0.007)

(0.006)

(0.003)

(0.006)

Business

-0.055***

-0.068***

0.013**

0.078***

Closed

(0.009)

(0.004)

(0.009)

Panel B: Health Care Utilization

(1)

(2)

(3)

(4)

(5)

(6)

Any Office

Visit

Office

Expend.

Any ER

Visit

ER Expend.

Any Rx

Rx Expend.

Laid off or

-0.007

0.234

0.001

0.626

-0.01

1.831

Job Ended

-0.006

-23.691

-0.003

-5.206

-0.006

-8.038

Business

0.015

-7.927

0.002

11.11

0.004

3.693

Closed

-0.012

-22.813

-0.005

-13.314

-0.01

-9.223

Panel C: Health Outcomes

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Health

Fair/Poor

Arthritis

Diabetes

Cholesterol

Hyperten

sion

Cancer

Mental

Fair/Poor

Depr/Anx

Laid off or

0.007

0.004

0.002

0.001

-0.003

0.010***

0.021***

Job Ended

(0.004)

(0.002)

(0.001)

(0.002)

(0.001)

(0.003)

Business

0.008

0.009

0.003

-0.003

-0.002

-0.000

0.007

0.023***

Closed

(0.006)

(0.005)

(0.002)

(0.003)

(0.004)

(0.006)

Notes: N=463,243. Standard errors (in parentheses) are clustered at the individual level. * p<.1 ** p<.05 *** p<.01.

Estimates are weighted using MEPS person-level sampling weight. All regressions include age and calendar year fixed

effects and baseline health and mental health interacted with trends.

Table A2

Summary Statistics By Displacement Status: Round 1 Holders

Never Displaced

Displaced

Demographic Characteristics

Male

0.56

0.60

Age

41.5

39.5

Black, Non-Hispanic

0.11

0.13

Other Race, Non-Hispanic

0.06

Hispanic

0.09

0.11

High School Dropout

0.07

0.09

High School Graduate

0.29

0.32

Any College

0.64

0.58

Round

Physical Health

Health Fair/Poor

7.4

6.4

9.1

8.9

Arthritis

8.3

15.2

8.0

15.2

Diabetes

3.9

5.0

3.3

4.7

High Cholesterol

8.0

11.8

6.7

10.4

Hypertension

12.4

16.5

11.6

15.6

Cancer

1.7

4.3

1.3

3.4

Mental Health

Mental Health Fair/Poor

2.7

3.3

4.0

5.5

Depression/Anxiety

7.3

15.1

9.4

20.4

Insurance Coverage

Any Insurance

100.0

95.1

100.0

68.7

Private Insurance

100.0

94.3

100.0

65.1

Public Insurance

2.0

2.4

1.9

4.8

Ever Uninsured

1.7

5.3

3.5

34.1

Health Care Utilization

Any Office-Based Visit

48.7

39.4

48.1

37.0

Expenditures on Office-Based Visits

214.3

217.4

206.2

222.7

Any ER Visit

3.9

2.8

5.5

3.4

Expenditures on ER Visits

34.0

29.1

44.2

46.6

Any Prescription Drug

49.3

43.7

47.3

39.1

Expenditures on Prescription Drugs

141.0

183.7

140.2

161.9

Observations

49150

42576

3589

3314

Notes: The data are from the 1996 through 2010 waves of the Household Component of the

Medical Expenditure Panel Survey. The sample includes individuals aged 21 to 65 that are

employed and are covered by health insurance through their employer in the first round of the

survey. Observations are weighted using MEPS person-level sampling weights.