CU Health & Medical Paper

Social Science & Medicine 189 (2017) 11e16Contents lists available at ScienceDirect
Social Science & Medicine
journal homepage: www.elsevier.com/locate/socscimed
Did the Great Recession affect mortality rates in the metropolitan
United States? Effects on mortality by age, gender and cause of death
Erin C. Strumpf a, b, c, *, Thomas J. Charters a, c, Sam Harper a, c, Arijit Nandi a, c
a
Department of Epidemiology, Biostatistics & Occupational Health, McGill University, 1020 Pine Avenue West, Montreal, QC H3A 1A2, Canada
Department of Economics, McGill University, 855 Sherbrooke St. West, Montreal, QC H3A 2T7, Canada
c
Institute for Health and Social Policy, McGill University, 1130 Pine Avenue West, Montreal, QC H3A 1A2, Canada
b
a r t i c l e i n f o
a b s t r a c t
Article history:
Received 6 January 2017
Received in revised form
4 July 2017
Accepted 20 July 2017
Available online 21 July 2017
Objectives: Mortality rates generally decline during economic recessions in high-income countries,
however gaps remain in our understanding of the underlying mechanisms. This study estimates the
impacts of increases in unemployment rates on both all-cause and cause-specific mortality across U.S.
metropolitan regions during the Great Recession.
Methods: We estimate the effects of economic conditions during the recent and severe recessionary
period on mortality, including differences by age and gender subgroups, using fixed effects regression
models. We identify a plausibly causal effect by isolating the impacts of within-metropolitan area
changes in unemployment rates and controlling for common temporal trends. We aggregated vital
statistics, population, and unemployment data at the area-month-year-age-gender-race level, yielding
527,040 observations across 366 metropolitan areas, 2005e2010.
Results: We estimate that a one percentage point increase in the metropolitan area unemployment rate
was associated with a decrease in all-cause mortality of 3.95 deaths per 100,000 person years (95%CI

6.80 to
1.10), or 0.5%. Estimated reductions in cardiovascular disease mortality contributed 60% of the
overall effect and were more pronounced among women. Motor vehicle accident mortality declined with
unemployment increases, especially for men and those under age 65, as did legal intervention and
homicide mortality, particularly for men and adults ages 25e64. We find suggestive evidence that increases in metropolitan area unemployment increased accidental drug poisoning deaths for both men
and women ages 25e64.
Conclusions: Our finding that all-cause mortality decreased during the Great Recession is consistent with
previous studies. Some categories of cause-specific mortality, notably cardiovascular disease, also follow
this pattern, and are more pronounced for certain gender and age groups. Our study also suggests that
the recent recession contributed to the growth in deaths from overdoses of prescription drugs in
working-age adults in metropolitan areas. Additional research investigating the mechanisms underlying
the health consequences of macroeconomic conditions is warranted.
© 2017 Elsevier Ltd. All rights reserved.
Keywords:
United States
Economic recession
Mortality
Cause of death
Metropolitan area
Age-specific mortality
Gender-specific mortality
Great Recession
1. Introduction
Since the influential work of Ruhm (C.J. Ruhm, 2000), empirical
research conducted on a variety of high income countries has found
that mortality largely varies procyclically with the business
cycle(Ariizumi and Schirle, 2012; Buchmueller et al., 2007;
* Corresponding author. Department of Epidemiology, Biostatistics & Occupational Health, McGill University, 1020 Pine Avenue West, Montreal, QC H3A 1A2,
Canada.
E-mail address: erin.strumpf@mcgill.ca (E.C. Strumpf).
http://dx.doi.org/10.1016/j.socscimed.2017.07.016
0277-9536/© 2017 Elsevier Ltd. All rights reserved.
Gerdtham and Ruhm, 2006; Miller et al., 2009; Neumayer, 2004;
 A Tapia Granados, 2005a, b, 2012; Jose
A
Stevens et al., 2015; Jose
 A Tapia Granados and Roux,
Tapia Granados and Ionides, 2011; Jose
2009). That is, over and above long-term trends, mortality rates
decline during recessions.
Investigations of specific causes of death have been used to gain
insights into the mechanisms underlying this relationship. Deaths
from cardiovascular disease, felt to be responsive to short-term
changes in modifiable health behaviors and environmental
stressors (Grundy et al., 1999), have been widely examined and
found to vary procyclically(Buchmueller et al., 2007; Gerdtham and
12
E.C. Strumpf et al. / Social Science & Medicine 189 (2017) 11e16
Ruhm, 2006; Neumayer, 2004; C.J. Ruhm, 2000, 2007; Stevens
 A Tapia Granados, 2005a; Jose
 A Tapia Granados
et al., 2015; Jose
 A Tapia Granados and Roux, 2009). Specifand Ionides, 2011; Jose
ically, temporary increases in deaths from coronary heart disease
may explain as much as two thirds of the increase in heart disease
deaths during periods of economic growth, while other subcategories of heart disease deaths decrease during those same
periods(C.J. Ruhm, 2007). Procyclical patterns have also been
documented for external causes of death. For instance, there is
strong and consistent evidence that economic declines are associated with reductions in traffic accident mortality (Buchmueller
et al., 2007; Gerdtham and Ruhm, 2006; Neumayer, 2004; C.J.
 A Tapia Granados, 2005a, b;
Ruhm, 2000; Stevens et al., 2015; Jose
 A Tapia Granados and Ionides, 2011; Jose
 A Tapia Granados and
Jose
Roux, 2009) and some evidence of reductions in mortality from
other accidents(Buchmueller et al., 2007; Gerdtham and Ruhm,
2006; C.J. Ruhm, 2000; Stevens et al., 2015). However other causes of death do not exhibit a consistently procyclical pattern,
including cancer (generally acyclical (Buchmueller et al., 2007;
 A Tapia Granados, 2005b;
Neumayer, 2004; C.J. Ruhm, 2000; Jose
 A Tapia Granados and Ionides, 2011)), homicide (mixed
Jose
(Gerdtham and Ruhm, 2006; C.J. Ruhm, 2000; D. Stuckler et al.,
 A Tapia Granados, 2005b; Jose
 A
2009) (Neumayer, 2004; Jose
Tapia Granados and Ionides, 2011)), and suicide (mixed
(Buchmueller et al., 2007; Gerdtham and Ruhm, 2006; Nandi et al.,
2012; Neumayer, 2004; C.J. Ruhm, 2000; Stevens et al., 2015; D.
 A Tapia Granados, 2005b; Jose
 A Tapia
Stuckler et al., 2009; Jose
Granados and Ionides, 2011)).
Accumulating evidence demonstrates that the impacts of macroeconomic conditions on health are unlikely to operate primarily
via the employment experience of individuals (J. A. Tapia Granados
et al., 2014). Per capita work hours have been shown to be negatively related to mortality in some countries(Johansson, 2004).
Analyses of macroeconomic effects on mortality in elderly populations, a group with low labor force connection, have sometimes
demonstrated stronger responses than for some categories of
working age individuals(Buchmueller et al., 2007; Neumayer, 2004;
C.J. Ruhm, 2000). Recent studies have found the strongest sensitivities to macroeconomic conditions among young adults (largely
from traffic fatalities), with smaller impacts among the middle
aged(Ariizumi and Schirle, 2012; Miller et al., 2009; Stevens et al.,
2015). The procyclical mortality effects among the elderly are
more modest, though this group experiences the highest mortality
rates and ultimately the vast majority of ‘excess’ deaths attributable
to macroeconomic activity(Miller et al., 2009; Stevens et al., 2015).
The 2007e2009 Great Recession in the U.S(Business Cycle
Dating Committee, Sept 20 2010). was characterized by a larger
increase in the unemployment rate than in previous post-WWII
recessions and an atypically slow recovery(U.S. Bureau of Labor
Statistics, Dec 2010). Research on the health effects of this recession has established weak or no impacts on smoking and alcohol
consumption (Nandi et al., 2013; Tekin et al., 2013), increases in
exercise (Colman and Dave, 2013; Tekin et al., 2013) and adiposity
(Latif, 2014), and decreases in vehicle miles traveled by safer (e.g.,
older) drivers(Maheshri and Winston, 2016). County-level analysis
has shown that adverse economic conditions, as measured by
poverty rates and lower median incomes, lead to higher mortality(Gordon and Sommers, 2016). Increases in suicide attributed to
the recession have been observed in several European countries
(Barr et al., 2012; Corcoran et al., 2015; Kondilis et al., 2013; Lopez
Bernal et al., 2013; Stuckler et al., 2011), as have declines in traffic
deaths (Regidor et al., 2014; Stuckler et al., 2011) and premature
mortality (Regidor et al., 2014), although few of these studies had
designs which permitted causal inference(Parmar et al., 2016).
Recent analyses have found the relationship between
macroeconomic conditions and overall mortality to remain procyclical overall (Lindo, 2015; C. J. Ruhm, 2015), but also evidence of
a shift towards acyclicality in recent years due to countercyclical
upsurges in cancer and accidental poisoning deaths(C. J. Ruhm,
2015).
This study contributes to the existing literature on economic
conditions and health by examining the impacts of economic
conditions during the recent and relatively severe Great Recession(Business Cycle Dating Committee, Sept 20 2010). Specifically,
we estimate the effects of changes in unemployment within
metropolitan statistical areas (MSAs) on both all-cause and causespecific mortality from 2005 to 2010. MSAs are population centers and their adjacent communities with a high degree of social
and economic integration, and therefore reflect local labor markets.
Approximately 84% of the US population lives in MSAs(U.S. Census
Bureau, 2012). Given recent countercyclical findings for accidental
poisoning (C. J. Ruhm, 2015) and epidemiological data on prescription drug overdose deaths, (U.S. Centers for Disease Control
and Prevention, 2012) we investigated accidental drug poisoning
specifically. We further examine age- and gender-specific effects by
cause of death in order to better understand the mechanisms at
work. Some of the most widely-cited studies that examine the
impacts of the recent recession on mortality do not convincingly
identify causal relationships(Stuckler et al., 2011). We therefore
contribute further to this literature by using more rigorous
methods (C.J. Ruhm, 2000) to plausibly identify the effects of the
Great Recession on mortality.
2. Methods
2.1. Data and sample
We calculated mortality rates based on data from the Centers for
Disease Control and Prevention’s National Vital Statistics System(U.S. Department of Health and Human Services et al.,
1980e2010). Underlying causes of death were designated through
International Classification of Diseases (ICD) codes version 10 (ICD10). We used auxiliary information available on the state and
county of residence of the decedent, and their age, sex, race, and
time of death, to generate month-MSA-subgroup-specific mortality
rates. Within each MSA we stratified monthly mortality totals by
age (0e15, 15 to 24, 25 to 44, 45 to 64, and 65 years old), sex, and
race (white, non-white), using county of residence to map to MSAs.
We used 366 MSAs, geographic areas made up of counties with at
least one urbanized core (population 50,000) and integrated
adjacent areas (Office of Management and Budget, 2000) corresponding to the November 2008 update of area definitions(Office of
Management and Budget, 2008). Annual midyear population denominators were obtained from the Surveillance Epidemiology and
End Results (SEER) U.S. population database (Surveillance
Epidemiology and End Results, 2005e2010) between 2004 and
2011 for counties and demographic groups, aggregated to MSAs,
and used to estimate monthly counts in population strata by linear
interpolation. The final data set consisted of 527,040 observations
at the MSA-month-year-age-gender-race level, from 366 MSAs over
the period 2005e2010.
2.2. Exposure and outcome measures
Our primary exposure variable, the seasonally-adjusted MSAlevel unemployment rate, was collected from the Bureau of Labor
Statistics’ (BLS) Local Area Unemployment Statistics database(U.S.
Bureau of Labor Statistics, 2005e2010). Given the unavailability of
seasonally adjusted estimates for New England MSAs, these rates
were computed from county level data and were not seasonally
E.C. Strumpf et al. / Social Science & Medicine 189 (2017) 11e16
adjusted.
Our dependent variable was the number of deaths in each MSAmonth-year-age-gender-race subgroup, overall and by cause of
death. In addition to all-cause mortality, we included deaths due to
malignant neoplasms, major cardiovascular disease, pneumonia
and influenza, chronic liver disease, motor vehicle accidents, accidental drug poisoning, other accidents and adverse events
(including transport accidents not otherwise classified, non-drug
poisonings, deaths from errors in medical or surgical care, and
other accidents), suicide, legal intervention and homicide. The last
category, ‘other cause’ mortality, included all causes of mortality
not attributable to those in the preceding list (see eTable 1 for
relevant ICD-10 codes).
2.3. Statistical analyses


unemployment rate and mortality without controls (model 1). In
model 2 we controlled for age, sex, and race. In model 3, we account
for time-invariant confounders which vary by MSA and shared
temporal trends in our outcomes by including MSA and quarter
fixed effects, respectively. Under this specification, the causal effects of economic conditions were identified using within-MSA
changes in unemployment as opposed to between-MSA comparisons. All analyses use robust standard errors clustered at the MSA
level.
In order to investigate the underlying mechanisms of the effect
of economic conditions on mortality, we examine heterogeneity by
demographic characteristics using model 3 separately by sex and
age, categorized as pre-working (ages 0 to 24), working (ages 25 to
64), and retired (ages 65) age groups.
2.4. Sensitivity analysis
We used Poisson regression to estimate the effects of the MSAlevel unemployment rate on mortality counts using the general
form:
log Ypjt
13
¼ bEjt þ gXjtp þ aj þ lt þ Upjt þ εpjt ;
where Ypjt is the number of deaths in socio-demographic subgroup
p, MSA j, and quarter t, and Ejt is the unemployment rate by MSA
and quarter. We controlled for time-invariant MSA characteristics
using MSA fixed effects (aj ) and common time trends using quarter
fixed effects (lt ). Additionally, we controlled for a vector of MSAquarter-subgroup indicator variables for age group, sex, and race
(Xjtp ). We used the natural log of the MSA-quarter-age-sex-racespecific population as the offset in the regression (Upjt) in order
to calculate mortality rates per 100,000 population. We estimated
the marginal effect of a one percentage point increase in the
quarterly MSA-level unemployment rate on the mortality rate and
present our findings as absolute changes in the annual mortality
rate per 100,000 population in order compare the contribution of
effects across outcomes and demographic groups.
We first estimated the crude association between the
When the assumption of equal mean and variance is not met,
the resulting overdispersion means that Poisson models may underestimate standard errors. As this was the case for some outcomes, we tested the sensitivity of our results by using negative
binomial regression, which incorporates an additional parameter to
represent unobserved heterogeneity among observations(Hilbe,
2007; Long and Freese, 2006).
3. Results
The all-cause annual mortality rate in US MSAs was 765 deaths
per 100,000, with the most prevalent forms of deaths being from
‘other-cause’ mortality (254 per 100,000), major cardiovascular
disease (252 per 100,000) and malignant neoplasms (178 per
100,000) (Table 1). Several mortality rates, including motor vehicle
accidents, accidental drug poisoning, suicide, and legal intervention
and homicide, were two to four times higher in men compared to
women. For most causes of death, the vast majority of deaths
occurred among those aged 65 and older, with the exceptions of
legal intervention and homicide, suicide, motor vehicle accidents,
Table 1
Rates and proportions of outcomes, exposure, and other covariates, 2005e2010, n ¼ 527,040 from 366 MSAs.
Outcome
Cardiovascular disease
Malignant neoplasms
Pneumonia or influenza
Other accidents & adverse events
Motor vehicle accident
Suicide
Chronic liver disease
Drug poisoning, accidental
Legal intervention & homicide
Other-cause mortality
All-cause mortality
Weighted mean deaths per 100,000 population
(Robust SEs)
Full Sample
Males
Females
Ages 0 – 24
Ages 25 – 64
Ages 65 þ
251.7
(6.1)
177.9
(3.6)
17.0
(0.6)
16.7
(0.7)
11.7
(0.5)
11.1
(0.5)
9.4
(0.3)
9.0
(0.4)
6.3
(0.3)
254.2
(7.8)
764.9
(16.0)
247.3
(5.6)
186.9
(4.1)
16.0
(0.6)
19.7
(0.7)
16.8
(0.8)
17.7
(0.7)
12.5
(0.4)
12.1
(0.5)
10.3
(0.5)
232.7
(6.6)
772.0
(16.2)
255.9
(6.8)
169.1
(3.3)
18.0
(0.6)
13.8
(0.6)
6.7
(0.3)
4.7
(0.2)
6.4
(0.2)
6.0
(0.3)
2.4
(0.1)
274.9
(8.9)
758.1
(16.0)
2.3
(0.1)
2.9
(0.0)
0.6
(0.0)
4.5
(0.2)
9.2
(0.4)
4.1
(0.2)
0.0
(0.0)
2.7
(0.2)
6.6
(0.3)
32.6
(0.7)
65.6
(1.5)
86.3
(2.1)
102.2
(1.9)
4.1
(0.1)
9.8
(0.3)
12.4
(0.6)
15.0
(0.7)
11.7
(0.4)
14.6
(0.6)
7.0
(0.3)
92.6
(2.4)
355.6
(7.9)
1691.4
(25.0)
1010.9
(10.2)
120.8
(4.7)
81.6
(3.6)
15.7
(0.5)
14.0
(0.6)
26.2
(0.7)
2.4
(0.1)
2.2
(0.1)
1597.7
(47.1)
4562.8
(49.4)
Note: SE ¼ standard error. The most common causes of “other-cause mortality” are chronic obstructive pulmonary disease, dementia, Alzheimer disease, diabetes mellitus, and
sepsis, which comprise 40% of cases for this outcome in the full sample.
14
E.C. Strumpf et al. / Social Science & Medicine 189 (2017) 11e16
and accidental drug poisoning. The overall population across MSAs
was 51% female, 79% white, 34% under age 25, 53% age 25 to 64, and
12% age 65þ. The weighted mean unemployment rate during
2005e2010 was 6.5%, which increased sharply from 4.6% in 2007 to
9.7% in 2010. In 2010, the interquartile range in the unemployment
rate was 8.3%e11.0%, with highly-impacted MSAs having unemployment rates as high as 30.4%.
The associations between increases in the MSA-level unemployment rate and all mortality outcomes are displayed in Table 2
with the estimated causal effect from model 3 being our
preferred estimate. All estimates are displayed with 95% confidence
intervals (CI). While the adjusted association from model 2 indicates increased unemployment led to a sizeable decline of 8.01
deaths per 100,000 population annually (95%CI
12.21,
3.80), this
correlation is markedly reduced by limiting identifying variation to
within-MSA changes and controlling for shared temporal trends. A
one percentage point increase in the MSA-level unemployment
rate decreased all-cause mortality by 3.95 (95%CI
6.80,
1.10)
deaths per 100,000 population annually (model 3). This is equivalent to a 0.5% decrease in mortality relative to the average all-cause
mortality rate. Increases in the unemployment rate also decreased
several cause-specific mortality rates: major cardiovascular disease
by 2.38 deaths per 100,000 (95%CI
3.39,
1.38), 0.9%; motor
vehicle accident deaths by 0.45 deaths per 100,000 (95%CI

0.61,
0.30), 3.8%; and legal intervention and homicide by 0.20
deaths per 100,000 (95%CI
0.36,
0.03), 3.2%. Our estimates
provide suggestive evidence of a small increase in mortality from
accidental drug poisoning of 0.10 deaths per 100,000 (95%CI
0.01,
0.21); 1.1%. We found little evidence that MSA-level unemployment
affects mortality due to cancer, pneumonia and influenza, chronic
liver disease, other accidents, suicide, or all other causes.
The effects of MSA-level unemployment rate on all-cause mortality vary somewhat by gender (Table 3). Interpreted as a percentchange from the gender-specific average mortality rates, a one
percentage point increase in unemployment decreased mortality
among men by 0.6% and among women by 0.4%. The decline in
motor vehicle accident mortality was greater among men than
among women: 4.0%, compared to 3.3%. Deaths from legal intervention and homicide declined by a rate of 0.35 per 100,000 (95%CI

0.64,
0.07; 3.4%) in men with essentially no change in women
(
0.04 (95%CI
0.11 to 0.04)). The absolute decline in cardiovascular disease mortality was greater among women than men,
though the percentage changes were similar: 2.57 deaths per
100,000 (95%CI
3.73,
1.41), or 1.0%, compared to 2.24 deaths per
100,000 (95%CI
3.23,
1.25), or 0.9%. Our estimates also provide
suggestive evidence of increases in mortality from accidental drug
poisoning in both groups: among men, mortality increased by 0.13
deaths per 100,000 (95%CI
0.03, 0.29), or 1.1%; among women,
mortality increased by 0.06 deaths per 100,000 (95%CI
0.02, 0.14),
or 1.0%.
Table 2
Effect of one percentage point increase in the MSA-level quarterly unemployment rate on the annual mortality rate difference per 100,000 person years, 2005e2010,
n ¼ 527,040 from 366 MSAs.
Cardiovascular disease
Malignant neoplasms
Pneumonia or influenza
Other accidents & adverse events
Motor vehicle accident
Suicide
Chronic liver disease
Drug poisoning, accidental
Legal intervention & homicide
Other-cause mortality
All-cause mortality
Model 1
Model 2
Model 3

1.21 (
3.27, 0.84)
0.16 (
0.89, 1.22)

0.30 (
0.45,
0.15)

0.04 (
0.19, 0.10)

0.36 (
0.50,
0.23)
0.10 (0.02, 0.18)
0.25 (0.20, 0.31)
0.23 (0.11, 0.35)

0.04 (
0.15, 0.07)
0.63 (
1.16, 2.42)

0.51 (
5.40, 4.37)

4.73 (
7.03,
2.43)

1.48 (
2.25,
0.71)

0.65 (
0.83,
0.46)

0.16 (
0.31,
0.02)

0.37 (
0.50,
0.23)
0.08 (0.02, 0.13)
0.21 (0.16, 0.26)
0.18 (0.09, 0.27)

0.04 (
0.20, 0.12)

1.55 (
3.26, 0.15)

8.01 (
12.21,
3.80)

2.38 (
3.39,
1.38)

0.07 (
0.64, 0.50)

0.14 (
0.45, 0.16)

0.12 (
0.47, 0.24)

0.45 (
0.61,
0.30)
0.04 (
0.01, 0.10)

0.01 (
0.08, 0.06)
0.10 (
0.01, 0.21)

0.20 (
0.36,
0.03)
0.37 (
1.79, 1.04)

3.95 (
6.80,
1.10)
Results indicate the absolute change in the mortality rate per 100,000 person years from a one percentage point increase in the MSA-level unemployment rate with 95%
confidence interval. Estimates from Poisson regression models offset by log-transformed MSA-demographic specific population estimates.
Model 1 includes no covariates.
Model 2 includes covariates for sex, age, and race.
Model 3 includes covariates for sex, age, race, and indicator variables (fixed effects) for each MSA and quarter.
Table 3
Effect of one percentage point increase in the MSA-level quarterly unemployment rate on the annual mortality rate difference per 100,000
person years by gender, 2005e2010, n ¼ 527, 040 from 366 MSAs.
Cardiovascular disease
Malignant neoplasms
Pneumonia or influenza
Other accidents & adverse events
Motor vehicle accident
Suicide
Chronic liver disease
Drug poisoning, accidental
Legal intervention & homicide
Other-cause mortality
All-cause mortality
Males
Females

2.24 (
3.23,
1.25)

0.06 (
0.90, 0.78)

0.25 (
0.61, 0.10)

0.26 (
0.67, 0.16)

0.68 (
0.88,
0.47)
0.06 (
0.04, 0.17)

0.05 (
0.16, 0.06)
0.13 (
0.03, 0.29)

0.35 (
0.64,
0.07)

0.57 (
2.12, 0.99)

4.83 (
7.92,
1.73)

2.57 (
3.73,
1.41)

0.06 (
0.55, 0.42)

0.07 (
0.36, 0.22)
0.03 (
0.29, 0.34)

0.22 (
0.36,
0.08)
0.02 (
0.03, 0.07)
0.03 (
0.03, 0.09)
0.06 (
0.02, 0.14)

0.04 (
0.11, 0.04)

0.21 (
1.67, 1.24)

3.24 (
6.06,
0.41)
Sample size 263,520 by gender.
Results indicate the absolute change in the mortality rate per 100,000 person years from a one percentage point increase in the MSA-level
unemployment rate with 95% confidence interval. Estimates from Poisson regression models offset by log-transformed MSA-demographic
specific population estimates.
Regression models include covariates for age, race, and indicator variables (fixed effects) for each MSA and quarter.
E.C. Strumpf et al. / Social Science & Medicine 189 (2017) 11e16
As expected, the absolute decreases in all-cause mortality are
largest at older ages, reflecting higher mortality rates (Table 4).
Interpreted as a percent-change from the age-specific average
mortality rates, a one percentage point increase in the MSA-level
unemployment rate reduced all-cause mortality by 1.5% among
those under age 25, 0.6% among those ages 25e64, and 0.2% among
those ages 65 and over. Among children and young adults, increases in the MSA-level unemployment rate reduced mortality
due to motor vehicle accidents and other accidents. Adults ages
25e64 also experienced reductions in mortality due to motor
vehicle accidents, but also legal intervention and homicide, and
‘other’ causes of death. The suggestive evidence of increases in
mortality due to accidental drug poisoning (0.20 deaths per
100,000 (95%CI
0.03, 0.43); 1.4%) and suicide (0.08 deaths per
100,000 (95%CI
0.01, 0.17); 0.5%) are most evident in adults ages
25e64. While increases in the unemployment rate reduced deaths
from cardiovascular disease (
9.59 deaths per 100,000 (95%CI

13.96,
5.22); 0.6%) among adults age 65þ, we find no evidence
that deaths due to external causes were significantly affected
among this group.
Results of our sensitivity analyses are presented in eTable 2. The
choice of regression specification between negative binomial and
Poisson generally had no meaningful impact on estimated mortality rate differences.
4. Discussion
We estimated the effect of the change in unemployment rates
2005e2010 on all-cause and cause-specific mortality among the
total U.S. metropolitan population. To better understand the underlying mechanisms, we also examined these relationships separately by sex and age. Controlling for demographic confounders,
time trends, and fixed differences across MSAs reduces the range of
alternative explanations and strengthens a causal interpretation of
our estimates.
Our results are consistent with findings from previous studies
documenting that all-cause mortality varies procyclically with
macroeconomic conditions. We find that certain categories of
cause-specific mortality also follow this pattern, and are more
pronounced for certain gender and age groups. The reductions in
cardiovascular disease mortality driven by increases in MSA-level
unemployment account for 60% of the all-cause effect and were
slightly more important among women than men. Legal intervention and homicide mortality also declined, particularly for men and
adults ages 25e64. Consistent with previous literature, we document important reductions in motor vehicle accident mortality
15
linked to increases in area-level unemployment, effects that are
most notable among men and those under age 65. We find suggestive evidence that mortality due to accidental drug poisoning
increases during periods of higher unemployment, though the
estimated effects are small and have borderline statistical significance. Lastly, we find that certain prevalent causes of death,
including cancer, pneumonia and influenza, and suicide, are
generally not sensitive to area-level economic conditions over this
period.
Several recent U.S. studies have reported an attenuation of the
procyclical relationship between unemployment and mortality in
recent years, with some evidence pointing to the rise in accidental
drug poisoning deaths during periods of economic decline(Miller
et al., 2009; C. J. Ruhm, 2015; Stevens et al., 2015). Our estimates
provide suggestive evidence that mortality due to accidental drug
poisoning increased more in metropolitan areas with larger growth
in unemployment during the Great Recession, particularly among
adults ages 25e64. While the increase in drug overdose deaths
continued to increase before, during, and after our study period
(Rudd et al., 2016), our findings suggest that within-MSA increases
in unemployment may have contributed to this trend.
There are several limitations to our analyses. First, the study
period comprises only six years and short periods have been linked
to volatility and imprecision in similar estimates(Ionides et al.,
2013; C. J. Ruhm, 2015). However, in most cases our results are
consistent with previous research conducted over longer time periods. Second, recent studies have demonstrated differences in the
relationship between economic conditions and mortality by
decomposing their samples narrowly by age. Our broader age
groups may average over such fine-grained variations. Third, our
focus on metropolitan regions enabled us to avoid the weaknesses
associated with state-level analyses (Lindo, 2015) and to link this
study with previous work that found that changes in the MSA-level
unemployment in the same period were not consistently associated
with changes in health behaviors(Nandi et al., 2013). However, the
relationship between area-level economic conditions and mortality
may differ between MSAs and rural areas. This may be particularly
important for deaths due to motor vehicle accidents (higher speeds
and vehicle-miles traveled) (Zwerling et al., 2005) and accidental
drug poisoning, as death and injury from opioid misuse are
concentrated in states with large rural populations(Keyes et al.,
2014). Fourth, the relationship between mortality rates and the
MSA-level unemployment rate is different from that between
individual-level mortality and employment status. We did not
address labor force participation as several other studies have done
 A Tapia Granados,
(C. J. Ruhm, 2015; Stevens et al., 2015; Jose
Table 4
Effect of one percentage point increase in the MSA-level quarterly unemployment rate on the annual mortality rate difference per 100,000 person years by age, 2005e2010,
n ¼ 316,224 from 366 MSAs.
Cardiovascular disease
Malignant neoplasms
Pneumonia or influenza
Other accidents & adverse events
Motor vehicle accident
Suicide
Chronic liver disease
Drug poisoning, accidental
Legal intervention & homicide
Other-cause mortality
All-cause mortality
Ages 0 – 24
Ages 25 – 64
Ages 65 þ
0.02 (
0.04, 0.08)
0.01 (
0.03, 0.06)

0.04 (
0.06,
0.01)

0.11 (
0.22,
0.00)

0.57 (
0.75,
0.39)

0.04 (
0.14, 0.06)
0.00 (
0.00, 0.01)
0.03 (
0.04, 0.10)

0.20 (
0.44, 0.04)

0.09 (
0.32, 0.14)

0.99 (
1.59,
0.39)

0.72 (
1.22,
0.23)

0.13 (
0.53, 0.28)

0.03 (
0.10, 0.05)

0.15 (
0.36, 0.06)

0.45 (
0.63,
0.27)
0.08 (
0.01, 0.17)

0.10 (
0.20, 0.01)
0.20 (
0.03, 0.43)

0.24 (
0.42,
0.07)

0.97 (
1.87,
0.08)

2.22 (
3.94,
0.50)

9.59 (
13.96,
5.22)
0.24 (
2.20, 2.68)

0.55 (
1.95, 0.85)
0.30 (
0.94, 1.54)

0.14 (
0.39, 0.12)
0.09 (
0.07, 0.26)
0.30 (0.04, 0.56)

0.01 (
0.09, 0.08)

0.04 (
0.16, 0.08)
1.56 (
3.79, 6.91)

10.66 (
20.79,
0.52)
Sample size 210,816 in 0e24 and 25e44 and 105,408 in 65þ.
Results indicate the absolute change in the mortality rate per 100,000 person years from a one percentage point increase in the MSA-level unemployment rate with 95%
confidence interval. Estimates from Poisson regression models offset by log-transformed MSA-demographic specific population estimates.
Regression models include covariates for sex, race, and age, and indicator variables (fixed effects) for each MSA and quarter.
16
E.C. Strumpf et al. / Social Science & Medicine 189 (2017) 11e16
2005b), though with little difference in their conclusions. Lastly, we
do not directly control for cross-MSA migration. Differential
migration rates that are correlated with changes in unemployment
rate would pose a challenge a causal interpretation of our results.
Such concerns are lessened, however, since internal US mobility is
decreasing over time and is pro-cyclical (Molloy et al., 2011). Evidence of an increase in mobility during the Great Recession is only
seen within-MSAs. Cross-region migration is falling due to a decline
in the geographic specificity of returns to occupations and an increase in workers’ ability to learn about other locations before
moving (Kaplan and Schulhofer-Wohl, 2017).
In this analysis, we show important differences in the impacts of
economic conditions on mortality across age, sex, and causes of
death. Additional research investigating the mechanisms underlying the health consequences of macroeconomic conditions is warranted. Further attention to variation in this relationship across
causes of death, population subgroups, and urban vs. rural contexts
could inform evidence-based policies. Evaluations of the ability of
social and health care policies to mitigate adverse effects are also
needed.
Acknowledgements
ES and SH were supported by Chercheur boursier awards from
bec e Sante
. AN was supported by
the Fonds de Recherche du Que
The Canada Research Chairs Program. These funders had no role in
study design; the collection, analysis and interpretation of data; the
writing of the articles; or the decision to submit it for publication.
Appendix A. Supplementary data
Supplementary data related to this article can be found at http://
dx.doi.org/10.1016/j.socscimed.2017.07.016.
References
Ariizumi, H., Schirle, T., 2012. Are recessions really good for your health? evidence
from Canada. Soc. Sci. Med. 74, 1224e1231.
Barr, B., Taylor-Robinson, D., Scott-Samuel, A., McKee, M., Stuckler, D., 2012. Suicides
associated with the 2008-10 economic recession in England: time trend analysis. BMJ Br. Med. J. 345.
Buchmueller, T.C., Grignon, M., Jusot, F., Perronnin, M., 2007. Unemployment and
Mortality in France, 1982-2002: Centre for Health Economics and Policy Analysis. McMaster University.
Business Cycle Dating Committee, Sept 20 2010. Business Cycle Dating Committee.
National Bureau of Economic Research, Cambridge, MA.
Colman, G., Dave, D., 2013. Exercise, physical activity, and exertion over the business
cycle. Soc. Sci. Med. 93, 11e20.
Corcoran, P., Griffin, E., Arensman, E., Fitzgerald, A.P., Perry, I.J., 2015. Impact of the
economic recession and subsequent austerity on suicide and self-harm in
Ireland: an interrupted time series analysis. Int. J. Epidemiol. 44, 969e977.
Gerdtham, U.G., Ruhm, C.J., 2006. Deaths rise in good economic times: evidence
from the OECD. Econ. Hum. Biol. 4, 298e316.
Gordon, S.H., Sommers, B.D., 2016. Recessions, poverty, and mortality in the United
States: 1993e2012. Am. J. Health Econ. 2, 489e510.
Grundy, S.M., Pasternak, R., Greenland, P., Smith, S., Fuster, V., 1999. Assessment of
cardiovascular risk by use of multiple-risk-factor assessment equations: a
statement for healthcare professionals from the American Heart Association
and the American College of Cardiology. J. Am. Coll. Cardiol. 34, 1348e1359.
Hilbe, J.M., 2007. Negative Binomial Regression. Cambridge University Press, New
York.
Ionides, E.L., Wang, Z., Tapia Granados, J.A., 2013. Macroeconomic Effects on Mortality Revealed by Panel Analysis with Nonlinear Trends, pp. 1362e1385.
Johansson, E., 2004. A note on the impact of hours worked on mortality in OECD
countries. Eur. J. Health Econ. 5, 335e340.
Kaplan, G., Schulhofer-Wohl, S., 2017. Understanding the long-run decline in
interstate migration. Int. Econ. Rev. 58, 57e94.
, M., Brady, J.E., Havens, J.R., Galea, S., 2014. Understanding the
Keyes, K.M., Cerda
ruraleurban differences in nonmedical prescription opioid use and abuse in the
United States. Am. J. Public Health 104, e52ee59.
Kondilis, E., Giannakopoulos, S., Gavana, M., Ierodiakonou, I., Waitzkin, H., Benos, A.,
2013. Economic crisis, restrictive policies, and the population’s health and
health care: the Greek case. Am. J. Public Health 103, 973e979.
Latif, E., 2014. The impact of macroeconomic conditions on obesity in Canada.
Health Econ. 23, 751e759.
Lindo, J.M., 2015. Aggregation and the estimated effects of economic conditions on
health. J. Health Econ. 40, 83e96.
Long, J.S., Freese, J., 2006. Regression Models for Categorical Dependent Variables
Using Stata. Stata Press books, College Station.
Lopez Bernal, J.A., Gasparrini, A., Artundo, C.M., McKee, M., 2013. The effect of the
late 2000s financial crisis on suicides in Spain: an interrupted time-series
analysis. Eur. J. Public Health 23, 732e736.
Maheshri, V., Winston, C., 2016. Did the Great Recession keep bad drivers off the
road? J. Risk Uncertain. 52, 255e280.
Miller, D.L., Page, M.E., Stevens, A.H., Filipski, M., 2009. Why are recessions good for
your health? Am. Econ. Rev. 99, 122e127.
Molloy, R., Smith, C.L., Wozniak, A., 2011. Internal migration in the United States.
J. Econ. Perspect. 25, 173e196.
Nandi, A., Charters, T.J., Strumpf, E.C., Heymann, J., Harper, S., 2013. Economic
conditions and health behaviours during the ‘Great Recession’. J. Epidemiol.
Community Health 67, 1038e1046.
Nandi, A., Prescott, M.R., Cerd
a, M., Vlahov, D., Tardiff, K.J., Galea, S., 2012. Economic
conditions and suicide rates in New York City. Am. J. Epidemiol. 175, 527e535.
Neumayer, E., 2004. Recessions lower (some) mortality rates:: evidence from Germany. Soc. Sci. Med. 58, 1037e1047.
Office of Management and Budget, 2000. Standards for defining metropolitan and
micropolitan statistical areas. In: Office of Management and Budget,
pp. 82227e82238 (Federal Register).
Office of Management and Budget, 2008. OMB Bulletin No. 09e01. Executive Office
of the President (Washington, D.C).
Parmar, D., Stavropoulou, C., Ioannidis, J.P., 2016. Health outcomes during the 2008
financial crisis in Europe: systematic literature review. BMJ 354, i4588.
Regidor, E., Barrio, G., Bravo, M.J., de la Fuente, L., 2014. Has health in Spain been
declining since the economic crisis? J. Epidemiol. Community Health 68,
280e282.
Rudd, R.A., Aleshire, N., Zibbell, J.E., Gladden, R.M., 2016. Increases in drug and
opioid overdose DeathseUnited States, 2000-2014. MMWR Morb. Mortal. Wkly.
Rep. 64, 1378e1382.
Ruhm, C.J., 2000. Are recessions good for your health? Q. J. Econ. 115, 617e650.
Ruhm, C.J., 2007. A healthy economy can break your heart. Demography 44,
829e848.
Ruhm, C.J., 2015. Recessions, healthy no more? J. Health Econ. 42, 17e28.
Stevens, A.H., Miller, D.L., Page, M.E., Filipski, M., 2015. The best of times, the worst
of times: understanding pro-cyclical mortality. Am. Econ. J. Econ. Policy 7,
279e311.
Stuckler, D., Basu, S., Suhrcke, M., Coutts, A., McKee, M., 2009. The public health
effect of economic crises and alternative policy responses in Europe: an
empirical analysis. Lancet 374, 315e323.
Stuckler, D., Basu, S., Suhrcke, M., Coutts, A., McKee, M., 2011. Effects of the 2008
recession on health: a first look at European data. Lancet 378, 124e125.
Surveillance Epidemiology and End Results, 2005-2010. Download US Population
Data-1969-2011. National Cancer Institute.
Tapia Granados, J.A., 2005a. Increasing mortality during the expansions of the US
economy, 1900e1996. Int. J. Epidemiol. 34, 1194e1202.
Tapia Granados, J.A., 2005b. Recessions and mortality in Spain, 1980e1997. Eur. J.
mogr. 21, 393e422.
Population/Revue Eur. de De
Tapia Granados, J.A., 2012. Economic growth and health progress in England and
Wales: 160 years of a changing relation. Soc. Sci. Med. 74, 688e695.
Tapia Granados, J.A., House, J.S., Ionides, E.L., Burgard, S., Schoeni, R.S., 2014. Individual joblessness, contextual unemployment, and mortality risk. Am. J. Epidemiol. 180, 280e287.
Tapia Granados, J.A., Ionides, E.L., 2011. Mortality and macroeconomic fluctuations
mogr. 27, 157e184.
in contemporary Sweden. Eur. J. Population/Revue Eur. de De
Tapia Granados, J.A., Roux, A.V.D., 2009. Life and death during the Great depression.
Proc. Natl. Acad. Sci. 106, 17290e17295.
Tekin, E., McClellan, C., Minyard, K.J., 2013. Health and Health Behaviors during the
Worst of Times: Evidence from the Great Recession. National Bureau of Economic Research, Cambridge, MA. Working Paper 19234.
U.S. Bureau of Labor Statistics, 2005-2010. Bureau of Labor Statistics Local Area
Unemployment Statistics (LAUS). U.S. Department of Labor, Washington, D.C.
U.S. Bureau of Labor Statistics, Dec 2010. Sizing up the 2007-9 recession: comparing
two key labor market indicators with earlier downturns. In: U.S. Bureau of
Labor Statistics. Office of Publications and Special Studies, Washington, DC.
U.S. Census Bureau, 2012. Patterns of Metropolitan and Micropolitan Population
Change: 2000 to 2010. 2010 Census Special Reports, Washington, DC.
U.S. Centers for Disease Control and Prevention, 2012. CDC grand rounds: prescription drug overdoses – a U.S. epidemic. MMWR Morb. Mortal. Wkly. Rep. 61
(1), 10e13.
U.S. Department of Health and Human Services, 1980-2010. Centers for disease
control and prevention, & national center for health statistics. In: National Vital
Statistics System (Ed.), Vital Statistics of the United States. Hyattsville, MD.
Zwerling, C., Peek-Asa, C., Whitten, P.S., Choi, S.W., Sprince, N.L., Jones, M.P., 2005.
Fatal motor vehicle crashes in rural and urban areas: decomposing rates into
contributing factors. Inj. Prev. 11, 24e28.
Journal article review
This paper should be 1-2 pages typewritten, double spaces 1” margins
After reading the journal article answer the following questions:
1. What is the purpose of this article?
2. Why is it important to investigate or examine the subject of the article?
3. How are the authors carrying out the task? Are their methods and comments
appropriate and adequate to the task?
4. What do they claim to have found out? Are the findings clearly stated?
5. How does this advance knowledge in the field?
6. What is your opinion about the article? Is it of value, useful or groundbreaking
in any way?