Economics 2015 4

Vol. 9, 2015-4 | February 04, 2015 | http://dx.doi.org/10.5018/economics-ejournal.ja.
2015-4
Income Inequality and Health: Evidence from

Developed and Developing Countries
Dierk Herzer and Peter Nunnenkamp
Abstract
We assess the effect of income inequality on life expectancy by performing separate
estimations for developed and developing countries. Our empirical analysis challenges the
widely held view that inequality matters more for health in richer countries than for health in
poorer countries. Employing panel cointegration and conventional panel regressions, we find
that income inequality slightly increases life expectancy in developed countries. By contrast,
the effect on life expectancy is significantly negative in developing countries. Even though
the quantitative effects are small, the contrast between the two country groups proves to be
robust to modifications in measurement, specification and methodological choices.
JEL I14 C23
Keywords Health; inequality; panel cointegration
Authors
Dierk Herzer, Helmut Schmidt University, Department of Economics, Hamburg
Peter Nunnenkamp, Kiel Institute for the World Economy, peter.nunnenkamp@ifwkiel.de
Citation Dierk Herzer and Peter Nunnenkamp (2015). Income Inequality and Health: Evidence from Developed
and Developing Countries. Economics: The Open-Access, Open-Assessment E-Journal, 9 (2015-4): 157. http://
dx.doi.org/10.5018/economics-ejournal.ja.2015-4
Received October 21, 2014 Published as Economics Discussion Paper November 12, 2014
Revised December 30, 2014 Accepted January 27, 2015 Published February 4, 2015
Author(s) 2015. Licensed under the Creative Commons License - Attribution 3.0
Introduction
Income disparities have widened in various developed and developing countries

during the process of economic globalization. Critics have called for redistributive
policy interventions, not only for reasons of fairness but also to avoid economic
and social costs of wide income gaps. Impaired health could add considerably to
such costs.
Wilkinson and Pickett (2006: 1768) conclude from their comprehensive review
of the literature that the large majority of more than 150 empirical studies suggest
that health is less good in societies where income differences are bigger. 1 In
particular, this applies to international cross-country studies. Moreover, the
literature suggests that the case for health-related redistribution is particularly
strong for developed countries. According to Wilkinson (1996: 4), the distribution
of income is one of the most powerful influences on the health of whole
populations in the developed world to have come to light. 2 Lynch et al. (1998:
1074) reckoned that the loss of life from income inequality in the United States is
comparable to the combined loss of life from lung cancer, diabetes, motor vehicle
crashes, human immunodeficiency virus (HIV) infection, suicide, and homicide in
1995. The so-called Whitehall studies on British civil servants showed that, even
among people who are not poor, there is a social gradient in mortality that runs
from the bottom to the top of society (Marmot 2003: S10). 3 More recently, the
meta analysis of Kondo et al. (2009) reveals a modest adverse effect of income
inequality on health. 4 Zheng (2012) finds that, while income inequality did not
have an instantaneous effect on mortality risk in the United States, long-term
_________________________
1 See also Lynch et al. (2004) who review almost 100 studies addressing the question of whether
more unequal societies are less healthy. Informative reviews of the relevant literature are also
presented by Judge et al. (1998), Wagstaff and van Doorslaer (2000), Deaton (2003), and
Subramanian and Kawachi (2004).
2 In addition to Wilkinsons own extensive work, several prominent studies supported this view,
including Rodgers (1979) and Waldmann (1992).
3 See also Anderson and Marmot (2012) who used data from the Whitehall II study to show that
promotions reduce the probability of developing heart disease.
4 As stressed by Kondo et al. (2009: 1180), even a modestly increased risk of premature mortality
has potentially important policy implications for population health as income inequality is an
exposure that applies to society as a whole.
www.economics-ejournal.org
effects proved to be detrimental once sufficiently long lags (of 512 years) are
accounted for. 5
In contrast to developed countries, income inequality may play a secondary
role in low-income countries with pervasive absolute poverty. Under such
conditions, it could be low income per se that matters most for health and
mortality, rather than income relative to other peoples incomes (Deaton 2003).
Importantly, this so-called absolute income or poverty hypothesis implies that
previous findings on inequality and health in developed countries do not
necessarily hold for developing countries. 6
The list of studies reviewed by Wilkinson and Pickett (2006) includes some
country-specific studies for lower-income countries such as Brazil, Chile, Ecuador
and Russia. Moreover, some cross-country studies include transition and
developing countries (e.g., Waldmann 1992; Karlsson et al. 2009). Nevertheless,
the empirical evidence on the health effects of income inequality continues to be
relatively scarce for developing countries, largely because of lacking data on
income inequality for a sufficiently large sample and a sufficiently long period of
time. Kondo et al. (2009: 1181) call for further investigations because of the lack
of empirical evidence from many parts of the world, including developing
countries. Furthermore, as noted by Torre and Myrskyl (2011), international
studies covering both developed and developing countries may suffer from
inappropriate pooling by not accounting for unobserved country heterogeneity. We
account for this possibility by performing separate estimations for two samples of
developed and developing countries.
To our knowledge, no study has systematically examined whether differences
in the effect of inequality on health exist between developed and developing
countries. Measuring and analyzing the effect of inequality on health for
developed and developing countries separately is the first contribution of this
study. Our analysis contributes to filling this important gap by drawing on a
_________________________
5 However, the pattern across OECD countries shown by Leigh et al. (2009: Figure 3) indicates that
the increase in life expectancy and the decline in infant mortality were more pronounced where
inequality widened. Leigh and Jencks (2007) present long-run evidence from a panel of 12 advanced
countries; they do not rule out the possibility that inequality raises life expectancy by a substantively
significant amount (page 19).
6 On the other hand, Deaton (2003: 114) argues that many of the arguments that income inequality
is a health risk are as plausible for poor as for rich countries.
relatively new data set, the Standardized World Income Inequality Database
(SWIID 2013), developed by Solt (2009). This data set combines information from
several sources, resulting in greater coverage and better comparability (for details,
see Section 3.2).
The second contribution of this study is the use of panel cointegration methods
to avoid spurious regressions in panel data. While it is well known that most
economic time series data are non-stationary, it is less known that if non-stationary
variables are not cointegrated, then using standard panel data techniques can lead
to spurious results (see, e.g., Entorf 1997; Kao 1999). Therefore, it is important to
test whether there is panel cointegration among the variables. As discussed in
more detail in Section 3, the cointegration property not only allows one to obtain
meaningful, non-spurious results, but it is also invariant to model extensions (see,
e.g., Ltkepohl 2007). The latter justifies a reduced-form model that excludes
other time-varying factors of inequality and health. Another advantage of the
cointegration approach is that it allows one to estimate long-run coefficients in a
manner that is free of endogeneity bias (see, e.g., Engle and Granger 1987).
Moreover, panel cointegration analysis, like conventional panel analysis, allows
one to control for any country-specific omitted factors that are stable over time,
thereby eliminating the time-invariant omitted-variable bias inherent in crosssectional analysis.
The third contribution is that we use a variety of estimation methods and
models as well as different measures of inequality and health to check the
robustness of our results. The overall objective of this study is to carefully
examine the effect of income inequality on health in developed and developing
countries using panel cointegration and conventional regression techniques.
Before describing in more detail our methodological approach and the data
employed (Section 3), we provide an overview of the analytical background in
Section 2, focusing on the theoretical ambiguity of the relationship between
income inequality and health outcomes. Section 4 presents the empirical results for
our samples of developed and developing countries, and Section 5 concludes. We
find that income inequality has a small, but robust and significantly positive
impact on health outcomes in developed countries. This finding contrasts with
earlier research as reviewed in Wilkinson and Pickett (2006) and recent
contributions such as Kondo et al. (2009) and Zheng (2012). Less surprisingly
perhaps, we find significantly negative effects on life expectancy in developing
countries. Even though the quantitative effects are small, the contrast between the
two country groups proves to be robust to modifications in measurement,
specification and methodological choices.
Analytical Background
Several lines of reasoning in the relevant literature suggest that a more equal
distribution of income is associated with better average health outcomes such as
longer life expectancy and lower mortality. Nevertheless, there is theoretical
ambiguity in various respects. Prestons (1975) finding of a non-linear relationship
between life expectancy and average per capita incomes across countries provided
an important building bloc of the so-called absolute income hypothesis, which has
also been coined the poverty hypothesis (see, e.g., Deaton 2003). The most
obvious explanation for this non-linearity is that it reflects diminishing returns to
increases in income (Preston 1975: 241). Increases in income would have larger
positive effects on health outcomes among poor people than on health outcomes
among rich people.
Consequently, mean-preserving income redistribution from the rich to the
poorwithin countries or between countrieswould be associated with better
average health. Health conditions among the rich might suffer to some extent from
such income transfers, but improved health conditions among the poor would
over-compensate any adverse effects on the rich. 7 The fact that diminishing
returns to personal income imply a negative association between income inequality
and health conditions at the aggregate level has been labeled a statistical artifact
by Gravelle (1998). This notion is meant to distinguish the absolute income
hypothesis from propositions according to which income inequality is directly
hazardous to health (see below). In the present context, it is more important to note
that the absolute income hypothesis requires the relationship between health and
personal income to be concave.
Even though the Preston curve is widely accepted as a stylized empirical
observation, the theoretical case for the concave relationship between health and
_________________________
7 The rich may not even suffer at all from impaired health if the so-called Preston curve becomes a
horizontal line once income exceeds a certain threshold.
income is open to debate. Grossman (1972) regards health as a durable capital

stock that produces an output of healthy time. The marginal product of health
capital increases with higher wage rates; the higher a persons wage rate, the
greater the value to him of an increase in healthy time (Grossman 1972: 241).
Grossmans model thus predicts that the demand for health and medical care
should be positively correlated with wage rates and per capita income. Similarly,
Waldmann (1992: 1291) argues that health care is plausibly a superior good.
This could prevent diminishing returns to income to the extent that additional
spending on health care translates into better health outcomes. 8 The absolute
income hypothesis would not hold under such conditions.
Another line of reasoning expects directly hazardous effects of income
inequality on health outcomes. The so-called relative income hypothesis,
according to which equal societies are healthier, draws on concepts and insights
from several disciplines, notably psychology, politics and economics. 9 Wilkinson
(1996; 1997; 2000), its most prominent proponent, argues that the epidemiological
transition from infectious diseases to chronic and degenerative diseases implies
that the major reason for differences in mortality and health shifts from (absolute)
material deprivation to (relative) social disadvantage. Social disadvantage is
supposed to give rise to psychosocial stress and relative deprivation. Unequal
societies are characterized, according to Wilkinson (2000: 4), by much more
stressful strategies of dominance, conflict and submission. At the same time,
biologists have shown that chronic stress impairs health by permanently perturbing
the physiologic balance (Sapolsky 2004).
This reasoning, though plausible, does not necessarily imply impaired health
due to income inequality. First, inequality may be closely linked with relative
deprivation, but there is little that suggests it is income inequality (Deaton 2003:
152). Second, rank matters and (upward) comparisons of ones own well-being
with higher ranked individuals in relevant reference groups may be stressful. 10
_________________________
8 Deaton (2003: 119) mentions that skepticism exists about the degree to which medical care results
in higher life expectancy.
9 See Kawachi and Kennedy (1999) for an overview of important concepts and pathways.
10 On the other hand, Leigh and Jencks (2007: 3) suspect that upward comparisons may be even
soothing if they lead people whose current economic circumstances are stressful to think that their
future circumstances could be better.
However, deprivation and adverse health effects would be contained if inequality

within specific reference groups was low compared to economy-wide inequality.
Indeed, inequality appears to be lower in groups of people who have much in
common, such as co-workers, friends, relatives, and neighbors (Leigh et al. 2009).
Furthermore, people typically belong to various reference groups and tend to
reduce stress by deriving self-esteem from the reference group where their ranking
is highest. 11 Third, while social subordination often involves stress and an
increased risk of stress-related diseases, Sapolsky (2004: 397 and 408) concludes
from surveying the relevant literature that there are numerous and dramatic
exceptions to this profile. In unstable hierarchies, stress centers on the higher ranks
as dominant individuals constantly need to defend their position against emerging
competitors. 12
Income inequality could also impair health conditions by eroding social trust
and affecting the political process of delivering public goods. It is widely agreed
that mutual trust and social capital are associated with better health (e.g., Kawachi
et al. 1997; dHombres et al. 2010; Ronconi et al. 2012). 13 Trust and social capital
could also help contain violent crime that may have minor direct effects on
mortality and life expectancy, but could have considerable second-order effects by
creating chronic stress among potential victims (Leigh et al. 2009). It is less
obvious, however, that adverse health effects of a disrupted social fabric can be
traced back to income inequality. On the one hand, Alesina and La Ferrara (2002)
find that a higher degree of income disparity is one factor among others eroding
mutual trust among individuals in US localities. According to Leigh (2006), a
_________________________
11 Sapolsky (2004: 408) provides the example of a low-paid clerk who is the best player in the firms
sports team; the place in the former hierarchy may be dismissed as just a job, whereas the latter
may be emphasized and become a source of considerable self-esteem. Along similar lines, Miller
and Paxson (2006) conclude from previous research that subjective social status is a better predictor
of psychological stress and health than objective measures of socioeconomic status.
12 Sapolsky (2004) also notes that humans reduce stress by adjusting the psychological meaning of
rank. For instance, an anticipated winner of a tournament tends to suffer more stress and deprivation
when failing to win than a novice player surviving just the first rounds. See also Lynch et al. (2004:
18): The stress effects of dominance hierarchies seem not even to be generalizable across primate
species, let alone generally applicable to the health effects of hierarchical human social
organization.
13 Putnam (1995: 67) defines social capital as features of social organization such as networks,
norms, and social trust that facilitate coordination and cooperation for mutual benefit.
negative effect of income inequality on trust can also be observed across countries.
Uslaner (2002: 22) concludes from studies on the United States as well as crosscountry studies that the level of economic equality is the strongest determinant of
trust. 14 On the other hand, reverse causality from trust and social capital to
inequality cannot be ruled out. Indeed, Knack (2002: 71) shows that social capital
is progressive, in the sense that it helps the poorer classes more than it helps the
richer classes. Uslaner (2002: 2021) also considers the possibility of reverse
causality as high degrees of trust lead countries to spend more on redistributing
wealth from the rich to the poor. Taken together, inequality is not necessarily the
ultimate cause of impaired health. 15
Ambiguity also prevails on whether income inequality is associated with less
or more public spending on health. Saint-Paul and Verdier (1993) model
redistribution through public spending on education, showing that more inequality
is associated with higher spending on education since the median voter (being
poorer than the mean) prefers a higher rate of taxation. The reasoning of SaintPaul and Verdier would also apply to health-related spending. It cannot be ruled
out, however, that more inequality reduces the support for public spending on
health or education. This could happen if poorer population segments participate
less in the electoral process than richer population segments. The preferences of
the poor may also be underrepresented because of the political clout of the rich
elite and the associated pressure for lower taxes. 16 In any case, it should be
recalled from Deaton (2003) that is disputed that higher spending on medical care
would necessarily result in better health outcomes.
_________________________
14 Uslaner (2002) offers two reasons of why economic equality promotes trust: (i) it makes poorer
people more optimistic that they can share in societys bounty, and (ii) it creates stronger bonds
between different groups in society.
15 Note also that dHombres et al. (2010) consider income inequality as an instrumental variable for
social capital.
16 Kawachi and Kennedy (1999: 221) quote Paul Krugman to this effect. Leigh et al. (2009) also
note that public spending on health could decline with more heterogeneous preferences of voters.
This argument is based on Alesina et al. (1999) who show that the average value of public goods to
members of a community diminishes with more pronounced heterogeneity. However, heterogeneity
in Alesina et al.s (1999) analysis is mainly linked to ethnic fractionalization, while they do not find
robust negative effects of income inequality on public spending (see Deaton 2003: 1312).
In the light of theoretical ambiguity, the health effects of income inequality are
essentially an empirical issue. The subsequent panel cointegration analysis
provides a particularly useful empirical approach to help clarify the causal links
between income inequality and health outcomes. Clearly, cross-section analyses
face problems to establish the direction of causality (Kawachi et al. 1997: 1497). 17
More surprisingly perhaps, even recent panel studies do not systematically address
causality issues. 18 Reverse causality is possible, or even likely (Borghesi and
Vercelli 2004; Deaton 2003; Leigh et al. 2009). Ill health may widen income gaps
in several ways. Measures that equalize health conditions across the population,
e.g., clean water supply in relatively poor countries, are also likely to narrow
income gaps. Better health enhances peoples earning capacity by reducing
absenteeism from work and improving productivity at work. Health conditions
within poor families affect the level of education and, thus, the income potential of
their children. Income differences across countries could be reduced if health
conditions improved in poorer countries through faster diffusion of superior health
technology and drugs. Hence, it appears essential to employ empirical methods
that account for bidirectional causality.
Empirical models and data
Our objective is to examine the effect of income inequality on health in developed

and developing countries using panel cointegration techniques as well as
conventional regression techniques. In this section, we present the empirical
models and discuss some econometric issues (Subsection 3.1). Then, we describe
the data, report descriptive statistics, and present some preliminary evidence
(Subsection 3.2).
_________________________
17 Most cross-section studies conclude with similar caveats.
18 For instance, Leigh and Jencks (2007) do not pursue Granger causality tests as they find no
statistically significant relationship between inequality and health. Etienne et al. (2007: 19) conclude
that there is still at least one important dimension which needs to be investigated namely the issue of
causality.
3.1
Empirical models and econometric issues
Following common practice in (panel) cointegration studies (see, e.g., Pedroni

2007; Herzer 2008; Moscone and Tosetti 2010), we consider a parsimonious
model which includes only the two variables of empirical interest: inequality and
health. Thus, the basic model takes the form
LEit = ai + i t + bGiniit + it ,
(1)
where the subscript i refers to one of the N cross-sectional units, i = 1, 2, ..., N ,

and the subscript t refers to one of the T time points, t = 1, 2, ..., T . LEit is the most
commonly used summary measure of health statuslife expectancy at birth
(Henderson 2009), and Giniit represents the standard measure of income
inequalitythe Gini coefficient (measured on a 0 to 100 scale). Following
previous inequality-health studies (see, e.g., Leigh and Jencks 2007), we use LEit
in levels rather than in logs. The coefficient b thus captures the long-run effect of a
change in the Gini index (by one unit) on life expectancy (in years). As the
principal interest is on the long-run effect, it is not essential to be concerned about
the variable lags through which inequality will impact health. Given that there is
evidence to suggest that income inequality has lagged effects on life expectancy
(Zheng 2012), we also employ a panel Granger causality test as a robustness
check. As is well known, the basic idea of Granger causality is to test whether
lagged values of a particular variable significantly affect the contemporaneous
value of another variable. In the robustness section, we also use the infant
mortality rate as an alternative measure of population health and the top-decile
income share as an alternative measure of inequality. Finally, any country-specific
omitted factors which are relatively stable in the long run or which evolve
smoothly over time are captured by country-specific fixed effects, ai, and countryspecific time trends, it. The country fixed effects account for factors such as
climate/geography, institutions, culture, and norms, while the individual time
trends capture factors such as country-specific (medical) technological progress.
Given that all variables exhibit trends (as shown in Figures A.1 and A.2 in
Appendix 1), it is reasonable to assume that LEit and Giniit are non-stationary
integrated processes. If this assumption is correct, the linear combination of the
two variables must be stationary, or, in the terminology of Engle and Granger
(1987), LEit must be cointegrated with Giniit. Otherwise, there is no long-run
10
relationship between life expectancy at birth and income inequality. Equation (1)
would in this case represent a spurious regression in the sense of Granger and
Newbold (1974). Entorf (1997) and Kao (1999) demonstrate that the tendency for
spuriously indicating a relationship may even be stronger in panel data regressions
than in pure time-series regressions. Thus, the necessary conditions for our model
to be a correct description of the data are that LEit and Giniit are non-stationary or,
more specifically, integrated of the same order, I(1), and cointegrated.
A regression consisting of (non-stationary) cointegrated variables has the
property of superconsistency such that coefficient estimates converge to the true
parameter values at a faster rate than they do in standard regressions with
stationary variables, namely rate T rather than T (Stock 1987). The important
point in this context is that the estimated cointegration coefficients are
superconsistent even in the presence of temporal and/or contemporaneous
correlation between the stationary error term, it , and the regressor(s) (Stock
1987), implying that cointegration estimates are not biased by omitted stationary
variables (see, e.g., Bonham and Cohen 2001).
The fact that a regression consisting of cointegrated variables has a stationary
error term also implies that no relevant non-stationary variables are omitted. Any
omitted non-stationary variable that is part of the cointegrating relationship would
become part of the error term, thereby producing non-stationary residuals, and thus
leading to a failure to detect cointegration (see also Everaert 2011).
If there is cointegration between a set of variables, then this stationary
relationship also exists in extended variable space. In other words, the
cointegration property is invariant to model extensions (see also Ltkepohl, 2007),
which is in stark contrast to regression analysis where one new variable can alter
the existing estimates dramatically (Juselius 2006, p. 11). The important
implication of finding cointegration is thus that no additional variables are
required to account for the classical omitted variables problem. More specifically,
the result for the long-run relationship between life expectancy and inequality
would also hold if additional variables were included in the model (see also
Juselius 1996).
Of course, there are several other factors that may affect population health
and/or income inequality. Therefore, adding further non-stationary variables to the
model may, on the one hand, result in further cointegrating relationships. If,
however, there is more than one cointegrating relationship, identifying restrictions
11
are required to separate the cointegrating relationships. Otherwise,

multicollinearity problems may arise. On the other hand, adding further nonstationary variables to the regression model may result in spurious associations.
More specifically, if a non-stationary variable that is not cointegrated with the
other variables is added to the cointegrating regression, the error term will no
longer be stationary. As a result, the coefficient of the added variable will not
converge to zero, as one would expect of an irrelevant variable in a standard
regression (Davidson 1998).
These considerations justify a parsimonious model such as Equation (1) (if
cointegrated). All the same, we check the robustness of the results to the inclusion
of additional control variables. More specifically, we follow Leigh and Jencks
(2007) and include GDP per capita and GDP per capita squared.
The superconsistency of the cointegration estimation also implies that the
potential endogeneity of the regressors should not affect the estimated long-run
coefficients; the estimated long-run coefficients from reverse regressions should be
approximately the inverse of each other due to the superconsistency (Engle and
Granger 1987). However, although the standard least-squares dummy variable
(LSDV) estimator is superconsistent under panel cointegration, it suffers from a
second-order asymptotic bias arising from serial correlation and endogeneity in
finite samples. As a consequence, its t-ratio is not asymptotically standard normal.
To deal with this problem, one has to employ an asymptotically efficient
(cointegration) estimator. Examples of such estimators include panel versions of
the dynamic OLS (DOLS) and fully modified ordinary least squares (FMOLS)
methods. As shown by Wagner and Hlouskova (2010), the panel DOLS estimator
outperforms other asymptotically efficient panel cointegration estimators.
Therefore, this DOLS estimator is our preferred estimator, but in the robustness
section we also present results based on alternative estimation procedures.
The idea behind the DOLS estimator is to account for possible serial
correlation and endogeneity of the regressors by augmenting the cointegrating
regression (given by Equation (1)) with lead, lag, and current values of the first
differences of the I(1) regressor(s) (see, e.g., Kao and Chiang 2000; Mark and Sul
2003). Accordingly, in our case, the DOLS regression is given by:
LEit = ai + i t + b Giniit +
j = k
ij
Giniit j + eit ,
(2)
12
where is the difference operator (such that Giniit = Giniit Giniit-1) and k is the
number of leads and lags. We use one lead and lag in the DOLS estimations to
preserve degrees of freedom, as is common practice in the literature (see, e.g.,
Spilimbergo and Vamvakidis 2003; Thorbecke and Smith 2010; Herzer et al.
2012).
Another empirical issue is the likely cross-sectional dependence among the
variables. Cross-sectional dependence may be the result of a common business
cycle and other common factors such as health shocks. Examples of such shocks
that affect health in multiple countries at the same time might include major
influenza epidemics, the spread of HIV/AIDS, the introduction of new vaccines,
and the diffusion of antibiotics (Leigh et al. 2009). To control for potential crosssectional dependence, we estimate the long-run effect of income inequality on
health status using both the raw data and demeaned data; that is, in place of LEit
and Giniit, we also use
LEit ' = LEit LEt and

Giniit ' = Giniit Ginit , where
LEt = N 1 i =1 LEit and

N
Ginit = N 1 i =1 Giniit ,
N
(3)
which is equivalent to including time dummies. Moreover, we use a battery of

panel unit root and cointegration tests, including so-called second-generation panel
unit root and cointegration methods that explicitly allow for cross-sectional
dependence.
A potential disadvantage of panel cointegration methods is that they typically
require balanced panel data over a sufficiently long time period. Complete time
series data on some alternative measures of inequality and health are not available
for many countries over long periods of time. To check the robustness of our
results to alternative measures of inequality and health, we are thus forced to use
conventional panel methods. More specifically, we use data on the tuberculosis
incidence rate and the income share of the bottom quintile to estimate a standard
autoregressive distributed lag (ARDL) model of the form
13
j =1
j =0
j =0
Healthit = ai + t + j Healthit j + j Inequalityit j + j Controlsit j + eit
(4)
where Healthit and Inequalityit stand for the measures of health and inequality, ai
are country-specific fixed effects (as before), and t represents time dummies. As
control variables, we include GDP per capita and GDP per capita squared
following Leigh and Jencks (2007), as discussed above. The number of lags is set
to k = 2 when the time period is sufficiently long (about 30 years); otherwise
(when T is about 20 years) we use one lag, k = 1. The long-run effect of a change
in inequality on health is given by

b=
1
k
j =0
k
j =1
...(5)
j
As is well known, the dynamic fixed effects model may suffer from the socalled Nickell (1981) bias; that is, the correlation between the lagged dependent
variable and the fixed effects may bias the coefficient on the lagged dependent
variable toward zero. However, the bias becomes small when T is about 20 or
more. Judson and Owen (1999) compare the performance of different estimators in
terms of Nickel bias and recommend the LSDV estimator in unbalanced panels
with T = 30. Bun and Kiviet (2006) examine the performance of several dynamic
panel estimators in samples where both T and N are moderate or small and
conclude that none of these estimators (including GMM and LSDV) dominates the
others in terms of bias or mean squared error. We use the LSDV estimator given
the relatively long time dimension of our data.
3.2
Data and preliminary evidence
We estimate both Equation (2) and Equation (4) for developed and developing
countries separately. The data on life expectancy at birth are from the World
Development Indicators (WDI) 2013 online database. 19 Life expectancy at birth
indicates the number of years a newborn infant would live if prevailing patterns of
_________________________
19 Available at: http://data.worldbank.org/data-catalog/world-development-indicators.
14
mortality at the time of its birth were to stay the same throughout its life. Life
expectancy is the most widely used indicator of health status and has also several
advantages over other measures of health, including the following: (i) it depends
on both infant mortality and other mortality rates, thus incorporating mortality
rates at all stages in life; (ii) it is not biased by age structure; and (iii) data on life
expectancy at birth are available for a reasonably large number of countries and
time periods.
However, the use of life expectancy as an indicator of health can be criticized
on two major grounds. First, longer life expectancy does not necessarily translate
into better health. While this is theoretically correct, we find that life expectancy at
birth and healthy life expectancy at birth, i.e., life expectancy at birth adjusted for
morbidity and time spent in poor health, are highly correlated. According to data
from the Institute for Health Metrics and Evaluation, 20 the cross-country
correlation coefficient between life expectancy at birth and healthy life expectancy
at birth for women (men) was 0.994 (0.994) in 1990, and 0.990 (0.999) in 2000.
Thus, it is very unlikely that the results would change significantly if it were
possible to use healthy life expectancy in place of life expectancy. Unfortunately,
adequate data on healthy life expectancy are not available to conduct a meaningful
panel data analysis. Therefore, we use (unadjusted) life expectancy as our main
indicator of population health. As an alternative summary measure of population
health, we use the infant mortality rate (per 1,000 live births), and as a specific
measure of health status, we use the tuberculosis incidence rate (per 100,000
population). These two variables are also from the WDI 2013 online database.
The second limitation is that average life expectancy does not reveal the
variation of health conditions within countries. The health conditions of poorer
population segments tend to be worse than those of richer population segments
for economic reasons such as spending on health care and/or for reasons of social
or psychic deprivation. There is evidence to this effect from selected case studies,
typically for high-income countries, including the well-known Whitehall studies in
the United Kingdom (see, e.g., Marmot 2003; Anderson and Marmot 2012) and for
the United States (Singh and Siahpush 2006). Comparable data do not exist for a
panel analysis. However, it appears that the results achieved for average life
_________________________
20 Available at: http://ghdx.healthmetricsandevaluation.org/global-burden-disease-study-2010-gbd2010-data-downloads .
15
expectancy ought to hold for life expectancy of poor population segments, if such
data were available. This can be concluded at least tentatively when considering
the prevalence of malnutrition among children under five as a marker of
subsequent poor health and low life expectancy. The correlation between
malnutrition, which can reasonably be assumed to be prevalent among the poor
and be absent among the rich, and average life expectancy is strongly negative. 21
As discussed above, we include GDP per capita (in constant 2005 US dollars)
and GDP per capita squared as additional explanatory variables. These data are
also taken from the WDI. 22
As far as data on income inequality are concerned, several studies have used
the Gini coefficient data set constructed by Deininger and Squire (1996). At least
since the work of Atkinson and Brandolini (2001) it is well known, however, that
the Deininger-Squire data suffer from deficiencies such as sparse coverage,
problematic measurements, and the combination of diverse data types into a single
data set, thus limiting the comparability, not only across countries but also over
time. Many studies therefore rely on Gini data from the Luxembourg Income
Study (LIS) database or the World Income Inequality Database (WIID). The major
deficiency of all these sources is the lack of continuous and consistent inequality
data over time. More generally, it should be noted that the Gini coefficient, though
widely available and often used in empirical studies, is an imperfect measure on
inequality. Most importantly, the Gini coefficient is not consistent with the welfare
principle. 23
In this study, we utilize a data source that combines the strengths of the LIS
and WIID datathe Standardized World Income Inequality Database (SWIID
_________________________
21 The simple correlation coefficient is -0.45, based on 622 observations for 128 countries.
Moreover, the coefficient on malnutrition enters negative and highly significant in simple panel
regressions with average life expectancy as the dependent variable (details available upon request).
22 Note that we do not use GDP per capita in PPP terms because GDP per capita in PPP terms is
available for fewer countries and years than conventional (exchange rate converted) GDP per capita.
23 According to the welfare principle, income transfers among the poor are more consequential than
income transfers among the rich (Firebaugh 2003: 79). Moreover, the Gini coefficient is more
sensitive to inequality (or to measurement error) at the top of the income distribution (Deaton 2003:
135).
16
2013) developed by Solt (2009). 24 The SWIID combines information from the LIS
and WIID data to create an improved data set with greater coverage than the LIS
data and greater comparability than the WIID data. The logic behind the
methodology underlying the SWIID can be summarized as follows (see also
Morgan and Kelly 2013). The synchronization process for the SWIID starts by
utilizing inequality data from both the LIS and the WIID. The WIID data contain
several country-years not available from the LIS and often includes inequality
statistics based on multiple income concepts (with some including and others
excluding various cash and/or in-kind transfers) for the same country-year. The
SWIID synchronization process treats inequality as a latent variable, with data
from the LIS and the WIID acting as imperfect indicators of the underlying
concept. With knowledge from country-years in which the two data sets overlap,
the SWIID uses inequality estimates from the strongly comparable LIS data set
along with inequality estimates and information about the income concept
represented in the WIID to adjust the WIID data such that it mimics the
comparability of the LIS data. This yields data with greater comparability and
more coverage than any other available data set.
Another notable point is that the SWIID provides comparable Gini indices of
gross and net income inequality. Given that the theoretical arguments presented in
Section 2 implicitly refer to inequality in net, rather than gross, income, we use the
Gini coefficient based on net income to examine the long-run relationship between
income inequality and health. We note, however, that our results do not change
qualitatively if we use the Gini index of gross income inequality (see the working
paper version of this article, Herzer and Nunnenkamp 2014). The correlation
coefficient between the gross income Gini coefficient and the net income Gini
coefficient is only 0.50 in our sample of (19) developed countries, but
considerably higher (0.88) in our sample of (59) developing countries.
Many of the more recent income inequality studies use the SWIID Gini
coefficient (see, e.g., Desbordes and Verardi 2012; Cole 2013; Morgan and Kelly
2013). However, this index has the limitation that it is estimated, and estimates
may be biased (for several reasons). Specifically, the detailed account of the pros
and cons of different sources of data on income inequality by Jenkins (2014) raises
_________________________
24 Available at: http://thedata.harvard.edu/dvn/dv/fsolt/faces/study/StudyPage.xhtml?studyId=36908
.
17
questions about the validity of multiply-imputed SWIID observations. As argued

by Jenkins (2014), users must generally be aware of trade-offs between data
availability and data comparability. He conditionally recommends WIID from a
data issues perspective. However, a major contribution of our paper is the use of
panel cointegration techniques. These techniques require a balanced panel, as
discussed above. Therefore, we prefer SWIID over other sources.
We check the sensitivity of the cointegration estimates to the measure of
inequality by using the top-decile income share data provided by Leigh (2007). 25
Leigh adjusts top incomes series from different studies to produce a comparable
data set. However, these data are available only for a small number of high-income
countries. For developing countries, we use the income share of the bottom 20
percent of the population from the WDI. 26 As noted above, these data, as well as
the data on the tuberculosis incidence rates, are unbalanced, and cannot be
employed in panel cointegration analysis. Therefore, we estimate the long-run
effects from the ARDL model given by Equation (4).
In our main analysis, we focus on the cointegrating relationship between LEit
and Giniit. In order to apply panel cointegration techniques, we need a balanced
panel data set. The construction of such a data set involves a trade-off between the
time span and number of countries in the sample. For the sample of developed
countries, we select all high-income countries for which complete time-series data
are available over the period 19762010the longest time period with complete
data for a reasonably large number of high-income countries according to World
Bank (1995) classification. This yields a sample of 19 developed countries and 35
time-series observations per country (665 total observations). In the robustness
section, the long-run relationship between LEit and Giniit is also estimated for 21
developed countries over the period 19812005 (525 total observations); data for
more countries are not available for this period.
For the period 19762010, balanced panel data are available only for 19
developing countries among which there are only three low-income countries.
Given this small number of developing (low- and middle-income) countries
according to World Bank (1995) classification, we decided not to choose the
_________________________
25 Available at: http://people.anu.edu.au/andrew.leigh/.
26 Data availability does not allow us to conduct a panel data analysis of the effect of the share of the
bottom 20 percent on health in developed countries.
18
period 19762010 for the sample of developing countries. Instead, we choose the
period 19812005, resulting in a balanced panel of 59 developing countries
(including 18 low-income countries) and 25 time-series observations per country
(1475 total observations).
Table 1 lists the countries in our sample of developed countries along with the
average values for LEit and Giniit over the period 19762010. Japan had the
highest life expectancy, followed by Hong Kong and Switzerland. Hong Kong was
Table 1: Countries in the sample of 19 developed countries and summary statistics,
19762010
Country
Australia
Canada
Denmark
Finland
France
Germany
Hong Kong
Ireland
Israel
Italy
Japan
Netherlands
New Zealand
Norway
Singapore
Sweden
Switzerland
United Kingdom
United States
Mean
Minimum
Maximum
Average of Gini
29.71
29.44
24.15
22.77
29.22
26.84
44.45
31.36
32.64
31.80
26.59
25.28
30.72
23.52
41.36
22.46
29.64
31.90
34.53
29.91
19.22 (Finland, 1983)
49.96 (Hong Kong, 2002)
Average of LE
77.60
77.75
75.78
76.06
77.25
75.99
78.40
75.49
77.35
77.54
79.40
77.43
76.42
77.63
76.22
78.28
78.37
76.44
75.67
77.11
70.59 (Singapore, 1976)
85.16 (Israel, 2006)
19
the country with the highest Gini, followed by Singapore, the United States and
Israel, while Sweden had the lowest Gini.
Table 2 shows the countries in the sample of middle- and low-income
countries, their average Gini and their average life expectancy at birth over the
period 19812005. Life expectancy at birth was highest in Greece and lowest in
Sierra Leone. Thailand had the highest Gini and Mauritius the lowest.
In Tables 1 and 2 we also report the sample means of the variables used in the
analysis, along with the minimum and maximum values of the data. As expected,
life expectancy in developed countries is, on average, significantly higher than in
developed countries, while income inequality is lower in developed countries
compared to developing countries.
Finally, in Table 3 we test whether our data can be used to replicate some of the
previous findings reported in the literature. Following Rodgers (1979), Waldmann
(1992), and Beckfield (2004), among others, we use a pooled sampleof (21)
developed and (59) developing countries. The 80 countries included in this sample
are listed in the notes to Figures A1 and A2 in Appendix 1, where we plot LEit and
Giniit for the sample period 19812005. The dependent variable is life expectancy
at birth, and the regressors include the Gini coefficient, GDP per capita, and GDP
per capita squared. Columns 13 of Table 3 present results from specifications
without country and time dummies, while the results in columns 4-6 are based on
specifications that include country and time dummies to control for time-invariant
omitted-variable bias and common time effects. According to the results without
country and time fixed effects, inequality appears to have a strongly negative and
statistically significant effect on life expectancy, which is consistent with previous
cross-sectional studies (see, e.g., Rodgers 1979; Wilkinson 1992; Waldmann
1992). The results with country and time fixed effects are also in line with
previous studies (see, e.g., Mellor and Milyo 2001; Beckfield 2004, Leigh and
Jencks 2007): The coefficient on the inequality variable turns out to be
insignificant in columns 46.
Columns 3 and 6 show that when we add GDP per capita and GDP per capita
squared to the basic specifications in columns 1 and 4, the coefficient on GDP per
capita is positive and significant (as in column 2) and the coefficient on GDP per
capita squared is negative and significant. Accordingly, increases in GDP per
capita are associated with increases in life expectancy, but the effects diminish as
20
Table 2: Countries in the sample of 59 developing countries and summary statistics,

19812005
Average
of Gini
42.82
31.29
37.33
51.43
24.21
49.94
38.06
49.43
42.80
41.17
35.35
44.84
30.29
36.98
Average of
LE
72.26
65.71
61.58
67.59
71.52
74.37
70.31
69.19
76.22
49.97
65.37
65.80
69.74
48.80
Fiji
Georgia
Greece
Guatemala
Hungary
India
Indonesia
47.07
35.42
33.40
50.33
26.78
48.99
46.21
66.11
70.61
76.96
63.84
70.23
59.72
64.44
Jordan
Kazakhstan
Kenya
Korea
Kyrgyz
Republic
Latvia
46.92
28.47
51.07
32.50
70.23
66.74
56.66
72.48
Malawi
Malaysia
Mauritius
Mexico
Morocco
Nepal
Nigeria
Pakistan
Panama
Peru
Philippines
Poland
Portugal
Puerto Rico
Russian
Federation
Sierra Leone
South Africa
Sri Lanka
Tajikistan
Tanzania
Thailand
Trinidad and
Tobago
Tunisia
Turkey
Turkmenistan
31.02
27.63
66.65
69.47
Ukraine
Uruguay
Argentina
Azerbaijan
Bangladesh
Brazil
Bulgaria
Chile
China
Colombia
Costa Rica
Cote dÌvoire
Egypt
El Salvador
Estonia
Ethiopia
Average of
Gini
52.22
47.62
18.46
46.26
38.68
47.37
46.39
37.69
49.30
53.53
48.28
27.79
30.81
39.60
Average of
LE
46.55
71.29
69.98
71.72
65.16
57.00
46.50
61.92
73.58
66.92
65.45
72.17
74.76
75.10
33.17
52.40
54.52
43.42
29.07
37.80
57.15
66.88
38.59
58.65
70.23
63.31
50.59
69.77
38.22
37.90
48.00
30.89
68.16
69.71
66.00
62.99
28.23
42.60
68.69
73.18
21
Table 2 continued:
Lithuania
Madagascar
Average of
Gini
28.12
42.99
Average
of LE
70.91
54.01
Mean
Minimum
Uzbekistan
Venezuela
Zambia
Gini
Average of
Gini
30.32
41.74
54.91
LE
40.29
15.37
(Mauritius, 2002)
75.26
(Indonesia, 2000)
Maximum
Average of
LE
66.67
71.27
44.58
65.34
35.79
(Sierra Leone, 1994)
79.24
(Greece, 2005)
Note: Low-income countries (according to World Bank (1995) classification) are marked with an
asterisk ().
Table 3: Preliminary results, 19812005

Independent
variable
Gini
(1)
(2)
(3)
-0.3768*** -0.1520*** -0.1066***
(-20.66)
(-7.91)
(-5.95)
GDP per capita

($ 1000s)
GDP per capita
squared
($
1000s)
Year dummies
Country dummies
Countries
Observations
No
No
80
2000
(4)
-0.0107
(-0.97)
(5)
0.0180
(1.59)
(6)
0.0177
(1.59)
0.3887***
(30.45)
1.0157***
(25.39)
0.0193 0.2708***
(1.09)
(6.53)
No
No
80
1889
-0.0156***
(-14.72)
No
No
80
1889
-0.0038***
(-7.47)
Yes
Yes
80
1889
Yes
Yes
80
2000
Yes
Yes
80
1889
Note: The dependent variable is LE. t-statistics (based on White's heteroskedasticity consistent
standard errors) are in parentheses. *** indicate significance at the one percent level.
22
GDP per capita rises, which is consistent with the results of Preston (1975),
Deaton (2003), and Leigh and Jencks (2007).
Empirical Analysis
In this section, we estimate the long-run effect of income inequality on population

health for developed and developing countries separately. We first analyze the
effect for developed countries (Subsection 4.1), as most previous studies have
done (see, e.g., Wilkinson 1992; Wennemo 1993; Judge et al. 1998; Leigh and
Jencks 2007), and check the robustness of our results (Subsection 4.2).
Subsequently, we provide estimates of the long-run effect of inequality on health
in developing countries (Subsection 4.3).
4.1
The long-run effect of inequality on health in developed

countries
The pre-tests for unit roots and cointegration, reported in Appendix 2, suggest that
LEit and Giniit are non-stationary and cointegrated. This implies that there is a
(non-spurious) long-run relationship between life expectancy at birth and the
SWIID Gini coefficient. To estimate this relationship, we use the panel weighted
DOLS estimator suggested by Kao and Chiang (2000). As discussed above, the
DOLS estimator is superconsistent, asymptotically unbiased, and normally
distributed, even in the presence of endogenous regressors.
Table 4 presents the results of this estimation procedure both for the raw data
(column 1) as well as for the data that have been demeaned over the crosssectional dimension (column 2). The estimated coefficient based on the raw data is
positive and significant at the five percent level. When the demeaned data are used
(to account for the problem of cross sectional dependence induced by common,
unobservable factors), the coefficient becomes significant at the one percent level.
This suggests that an increase in inequality is associated with an increase in life
expectancy in high-income countries, which is in contrast to most previous studies.
What can be said about the magnitude of the estimated effect in column 2?
Multiplying the coefficient of Gini (0.0467) with the average value of the change
23
Table 4: Estimates of the long-run effect of inequality on life expectancy at birth in

developed countries, 19762010
Independent
variable
Gini
Demeaned data
Fixed effects
Individual trends
Countries
Observations
Pooled panel
DOLS
estimator
(1)
0.0195**
(1.98)
No
Yes
Yes
19
608
Pooled panel
DOLS
estimator
(2)
0.0467***
(4.34)
Yes
Yes
Yes
19
608
Group-mean
panel DOLS
estimator
(3)
0.0331***
(3.11)
Yes
Yes
Yes
19
608
Pooled panel
FMOLS
estimator
(4)
0.0368*
(1.89)
Yes
Yes
Yes
19
646
Note: The dependent variable is LE. t-statistics are in parenthesis. *** (**) [*] indicate significance
at the one (five) [ten] percent level. The DOLS results are based on a one lead/lag model.
in the Gini coefficient in the sample (0.114214) yields a value of 0.005334,

implying that inequality has contributed about 0.005334 years to the annual
increase in life expectancy for the average country in the sample. With an average
increase in life expectancy of 0.229674 years, this means that inequality has been
responsible for about 2.32 percent of the annual increase in life expectancy in the
country sample considered here. Clearly, the quantitative impact of changes in
inequality on life expectancy in developed countries is fairly small. 27
Two recent studies may help explain why we do not find negative effects of
higher income inequality on life expectancy in developed countries. First,
according to Zheng (2012), adverse effects on life expectancy are difficult to
capture unless long lags are accounted for. The cointegration approach allows us to
estimate the long-run cumulative effect of inequality. We also employ a panel
Granger causality test as a robustness check in the next section. However, the
available data do not allow us to capture the extremely long lags of up to 12 years
considered by Zheng (2012) in his study for the United States. Second, Torre and
Myrskyl (2011) find that adverse health effects of higher income inequality are
_________________________
27 As noted by Kondo et al. (2009) and Zheng (2012), however, even a small increase in the risk of
premature mortality may have considerable consequences for the countrys population as a whole.
24
restricted to younger age cohorts; for women the association vanishes already at
ages above 15. This could render our surprising finding more plausible when
taking into account that the contribution of mortality at ages below 50 on life
expectancy at birth is relatively small in developed countries (Torre and Myrskyl
2011: 4). By contrast, premature mortality tends to be much more common in
developing countries. 28
4.2
Robustness
Even though the quantitative impact is small, the finding of a significantly positive
relationship between inequality and health in high-income countries contradicts
previous findings. Hence, we perform several robustness checks. First, we
examine whether the positive relationship between inequality and health in
developed countries is robust to alternative estimation techniques. A potential
problem with the pooled results (in columns 1 and 2 of Table 4) could be that they
are based on the implicit assumption of homogeneity of the long-run effects.
While efficiency gains from the pooling of observations over the cross-sectional
units can be achieved when the individual slope coefficients are the same, pooled
estimators may yield inconsistent and potentially misleading estimates of the
sample mean of the individual coefficients when the true slope coefficients are
heterogeneous. However, this type of problem is mainly attributed to dynamic
panel data models with lagged dependent variables (Pesaran and Smith, 1995).
Moreover, a comparative study by Baltagi and Griffin (1997) concludes that the
efficiency gains from pooling appear to more than offset the biases due to
intercountry heterogeneities (p. 317). Nonetheless, we allow the long-run
coefficients to vary across countries by using the group-mean panel DOLS
estimator suggested by Pedroni (2001). This estimator involves estimating separate
DOLS regressions for each country and averaging the long-run coefficients,
b = N 1 iN=1 bi . The corresponding t-statistic is computed as the sum of the
individual t-statistics (calculated using heteroskedasticity and autocorrelation_________________________
28 Regrettably, the Human Mortality Database used by Torre and Myrskyl (2011) does not provide
any information on mortality for specific age cohorts in developing countries (except for Chile). This
prevents us from extending the analysis of Torre and Myrskyl to a broader sample, including lowincome countries.
25
consistent standard errors) divided by the root of the number of cross-sectional

units, tb = iN=1 tb / N . In addition, we use the pooled FMOLS estimator
i
suggested by Phillips and Moon (1999). Like the time series FMOLS estimator,
the panel FMOLS estimator incorporates a semi-parametric correction to the OLS
estimator, which eliminates the second order bias induced by the endogeneity of
the regressors. We report the results of these estimation methods in columns 3 and
4 of Table 4.
The results show a positive and significant effect of inequality on life expectancy. Interestingly, the panel and group-mean DOLS estimators produce similar
coefficients, suggesting that slope heterogeneity is not a serious problem in this
sample. Like the group-mean DOLS coefficient estimate (in column 3), the
FMOLS coefficient estimate in column 4 is somewhat smaller than the DOLS
estimate in column 2, and it is significant only at the ten percent level. As shown
by Wagner and Hlouskova (2010), these two estimators perform worse than the
panel DOLS estimator of Kao and Chiang (2000). Therefore, we continue our
robustness analysis using the (weighted) panel DOLS estimator (with demeaned
data).
To verify that the surprisingly positive, though small, effect of inequality on
health is not due to individual outliers, the DOLS regression is re-estimated by
excluding one country at a time from the sample. The sequentially estimated
coefficients and their t-statistics are presented in Figure 1. They fluctuate between
0.034 (due to the exclusion of New Zealand) and 0.059 (due to the exclusion of
Switzerland) and are always significant at the one percent level, suggesting that the
positive effect of inequality on health is not the result of individual outliers. At the
same time, the quantitative impact remains small independent of which country is
excluded from the sample.
It is common practice in conventional panel studies to use time-averaged data
to eliminate business cycle effects. However, as pointed out by Attanasio et al.
(2000), annual data provide information that is lost when time-averaged
observations are used. Moreover, is not obvious that averaging over fixed time
intervals will effectively eliminate business cycle effects; the length of the interval
over which averages are computed is arbitrary, and there is no guarantee that
business cycles are cut in the right way, as their length varies over time and across
countries. In addition, the use of time-averaged data decreases the number of
26
Figure 1: DOLS estimation with single country excluded from the sample
Coefficients on Gini
.060
.055
.050
.045
.040
.035
.030
2
10
12
14
16
18
16
18
t-statistics of the coefficients on Gini

5.5
5.0
4.5
4.0
3.5
3.0
2.5
2
10
12
14
observations, and hence statistical power. Despite these concerns, we re-estimate

the DOLS regression using five- year averages. The results of this estimation are
reported in column 1 of Table 5. As can be seen, the estimate using five-year
averages is close to the corresponding estimate based on annual data reported
27
Table 5: DOLS estimates for five-year averages, for a larger sample of countries, and for
specifications with control variables
5-year
averages
Independent
variable
Gini
(1)
0.0560***
(6.72)
Larger sample
over a shorter
time period
(2)
0.0332***
(2.84)
GDP per capita

($ 1000s)
GDP per capita
squared
($ 1000s)
Demeaned data
Fixed effects
Individual trends
Sample period
Countries
Observations
Yes
Yes
Yes
19762010
19
95
Yes
Yes
Yes
19812005
21
462
With GDP p.c. and

GDP p.c. squared
(3)
0.0259**
(2.48)
With GDP p.c.

and GDP p.c.
squared
(4)
0.0087***
(5.51)
-0.0610**
(-2.08)
0.0826**
(2.46)
0.0010***
(2.94)
Yes
Yes
Yes
19812005
20
440
-0.0020***
(-4.76)
Yes
Yes
No
19812005
20
440
Note: The dependent variable is LE. t-statistics are in parenthesis. *** (**) indicate significance at
the one (five) percent level. Given the very small number of time series observations, the DOLS
result in column 1 is based on a model with one lag and no leads; all other results are based on
equations with one lead and one lag. The estimates in column 2 are based on a sample that includes
the 19 countries of our main sample plus Belgium and Spain. For Ireland, complete time series data
on real GDP per capita are not available for the period 1981-2005. Therefore, Ireland is excluded
from the sample used to estimate the specifications in columns 3 and 4.
in Table 4, column 2. This is consistent with several studies showing that

cointegration estimates are remarkably stable across frequencies (see, e.g.,
Chambers 2001; Click and Plummer 2005; Herzer 2013).
In column 2 of Table 5, we estimate the coefficient on Giniit using a larger
sample of 21 high-income countries from 19812005 (which is the period used in
the next section to analyze the effect of inequality on health in developing
countries). Once again, the estimated coefficient is highly statistically significant,
though the positive impact is even smaller than in column 2 of Table 4.
28
As discussed in Section 3, the finding of cointegration implies that there are no

missing trending variables and that therefore no additional variables are needed to
produce unbiased estimates. Nevertheless, we check the robustness of our results
to the inclusion of GDP per capita and GDP per capita squared. A potential
problem with this strategy is that it can introduce collinearity among the stochastic
regressors or between the additional variables and the individual time trends. To
account for this problem, we present estimates with and without individual time
trends. As can be seen from columns 3 and 4 of Table 5, the impact of inequality
on health remains positive and statistically significant when we include GDP per
capita and GDP per capita squared, regardless of whether or not individual time
trends are used in the analysis (to represent technological change). While in
column 3 the coefficients on GDP per capita and GDP per capita squared have
unexpected signs, the signs of these coefficients in column 4 are as we expect: the
coefficient on GDP per capita is positive and significant and the coefficient on
GDP per capita squared is negative and significant (in the regression without
deterministic time trends).
Next, we examine whether the results are robust to alternative measures of
inequality and mortality. Leigh and Jencks (2007) use data on the income share of
the richest 10 percent to examine the effect of income inequality on health for 12
high-income countries in an unbalanced panel between 1903 and 2003. Here we
construct a balanced panel for 8 high-income countries (Canada, France,
Netherlands, New Zealand, Sweden, Switzerland, the United Kingdom, and the
United States) for the period 19611996. Regrettably, complete time series on life
expectancy at birth and the top income share are not available for more countries
and years. We also use the infant mortality rate to check the robustness of our
results, following many studies which regard infant mortality as an indicator of the
general health status of the population. Table 6 presents the results of the DOLS
regressions using these two different measures, labeled TopDecile and IMR, both
separately and jointly. All estimates suggest that income inequality increases
health in high-income countries.
A potential problem with these estimates is that life expectancy and infant
mortality are measures of mortality rather than morbidity. While the two are highly
29
Table 6: DOLS estimates using different measures of inequality and health
Independent
variable
TopDecile
Different inequality
measure
[Dependent variable:
LE]
Different health
measure
IMR]
(1)
0.1032***
(4.90)
(2)
Gini
Demeaned data
Fixed effects
Individual
trends
Sample period
Countries
Observations
Different inequality
and different health
measure
IMR]
(3)
-0.2076***
(-3.30)
Yes
Yes
-0.0577***
(-3.20)
Yes
Yes
Yes
Yes
Yes
19611996
8
264
Yes
19762010
18
576
Yes
19611996
8
264
Note: t-statistics are in parenthesis. *** indicate significance at the one percent level. The DOLS
results are based on a one lead/lag model. For Hong Kong, complete time series data on infant
mortality are not available for the period 19762010. Therefore, Hong Kong is not included in the
sample used to estimate the coefficient in column 2.
correlated (see Section 3.b), it is morbidity rather than mortality which should be
affected most by inequality. In other words, morbidity is, by definition, a better
measure of the health response to income inequality than mortality (Soobader and
LeClere 1999). Regrettably, summary measures of morbidity which account for all
diseases (whether physical, psychosomatic or psychiatric) are not available to
conduct a meaningful panel analysis. A specific measure of morbidity is the
tuberculosis incidence rate, which is available for a sufficient number of countries
over the period 19902011. Because the data are unbalanced, we do not apply the
DOLS estimator, but estimate the long-run effect of inequality on the incidence of
tuberculosis using the ARDL model. Table 7 presents the results with and without
control variables. The estimated long-run coefficient on Gini is always negative
30
Table 7: Estimates of the long-run effect of inequality on tuberculosis incidence in

developed countries, 19902011
Exogenous variable
Gini
(1)
-0.7070 ***
(0.1310)
(2)
-0.6843***
(0.1341)
0.6203***
(0.1168)
Yes
Yes
30
547
Yes
Yes
30
537
GDP per capita ($ 1000s)

GDP per capita squared
($ 1000s)
Year dummies
Country dummies
Countries
Observations
(3)
-0.6583***
(0.1515)
3.0097***
(0.7857)
-0.0176***
(0.0057)
Yes
Yes
30
537
Note: Standard errors (calculated by the Delta method) are in parentheses. *** indicate significance
at the one percent level. The effects were estimated from an autoregressive distributed lag model
with one lag on the endogenous variable and one lag on the exogenous variable(s).
and highly significant, suggesting that inequality decreases the incidence of

tuberculosis. This corroborates the finding that inequality has a positive effect on
population health in developed countries. 29
Finally, we perform a panel Granger causality test based on a levels vector
autoregression (VAR). It is well known from the time series literature that while,
in general, the asymptotic distributions of the Wald test for Granger causality in
levels VARs with non-stationary variables are nonstandard (see, e.g., Toda and
Phillips 1993), the conventional Wald test for a bivariate cointegrated VAR model
is asymptotically distributed as chi-square and therefore valid as a test for Granger
causality (see, e.g., Ltkepohl and Reimers 1992). Since LEit and Giniit are
cointegrated, conventional Wald test statistics formed from standard VAR
estimates are thus applicable for testing causality between income inequality and
_________________________
29 Again, the coefficients of GDP per capita and GDP squared have the wrong sign. The positive
association between GDP per capita and tuberculosis may be due to endogeneity bias if higher
morbidity increases future per capita income by reducing population size.
31
Table 8: Bivariate Granger causality tests, 1976-2010

Lags
Null
hypothesis
Gini does
not cause
LE
LE does
not cause
Gini
(1)
p-value of the
Sum of the lagged
Sum of the lagged
Granger causality
coefficients
coefficients on life
chi-square statistic on the Gini coefficient
expectancy
(2)
(3)
(4)
0.0228
1
2
3
4
0.016
0.000
0.001
0.050
0.0366
0.0489
0.0603
0.0240
1
2
3
4
0.000
0.297
0.810
0.643
-0.0105
-0.0287
-0.0404
Note: This table reports the p-values of Granger-causality VAR tests. The null hypothesis is that one
(two, three, and four) lag(s) of the (demeaned) series of Gini (LE) do not help predict the series of LE
(Gini).
health. However, a potential problem is that standard panel causality tests, which
impose homogeneous coefficients of the lagged dependent variables across
countries, may be biased when in fact the coefficients of the lagged dependent
variables are heterogeneous across countries (Dumitrescu and Hurlin, 2012).
Therefore, we apply the Dumitrescu and Hurlin (2012) test of Granger noncausality for heterogeneous panels to our sample of 19 countries (over the period
1976-2010). This test involves conducting the Granger causality test for each
country individually and then averaging the individual Wald statistics.
The Schwarz criterion is used to determine the appropriate number of lags for
each country. Since this number varies from one to four across the countries, we
perform the panel Granger causality tests using one to four lags.
The p-values of the Granger causality chi-square statistics are reported in
column 2 of Table 8. The sums of the lagged coefficients of the exogenous causal
variables (obtained as averages of the individual country coefficients) are
32
presented in columns 3 and 4. As can be seen, the null hypothesis of no Granger

causality from Giniit to LEit is rejected at the five percent level or better for all lags,
and the sum of the coefficients on the lagged Gini coefficients is always positive in
the life expectancy equation. This supports our result that income inequality
increases life expectancy in developed countries.
Interestingly, the Granger causality test rejects the null hypothesis that one lag
of LEit does not help predict Giniit at the one percent level with a positive
coefficient on life expectancy. From this, it can be concluded that an increase in
life expectancy leads to an increase in inequality. However, this conclusion is not
robust to the number of lags used in the tests; for two, three, and four lags, the null
hypothesis of no Granger causality from income inequality to live expectancy
cannot be rejected.
In summary, we find robust evidence for the surprisingly positive, though
quantitatively small, effect of income inequality on population health in developed
countries. In the next subsection, we investigate whether this result also holds for
developing countries.
4.3
The long-run effect of inequality on health in developing

countries
In Table 9, we present DOLS estimates of the long-run effect of inequality on

health in developing countries using life expectancy and infant mortality as
measures of health outcomes; the measure of inequality is the Gini index. The
estimated coefficient in column 1 is negative and highly significant, suggesting
that inequality is negatively related to life expectancy. Column 2 reports a positive
and significant coefficient, suggesting that inequality is positively associated with
infant mortality. Thus, in contrast to the results for developed countries, the results
for developing countries show that inequality harms health. This is consistent with
the results of the few previous studies that have examined the effect of income
inequality on mortality in a (sub)sample of developing countries (see, e.g.,
Rodgers, 1979; Flegg, 1982; Waldmann, 1992). To quantify the effect, we
multiply the coefficient of Gini in column 1 with the average change in the Gini
index (0.0740 0.0563 = 0.004). The result implies that the average loss of life
expectancy due to inequality is just about 1.5 days per year (0.004 365 =
1.46).
33
Table 9: DOLS estimates of the long-run effect of inequality on life expectancy at birth
and in developing countries, 19812005
Independent variable
Gini
Demeaned data
Fixed effects
Individual trends
Countries
Observations
(1)
Dependent variable: LE
-0.0740***
(-13.37)
Yes
Yes
Yes
59
1298
(2)
Dependent variable: IMR
0.1701***
(9.96)
Yes
Yes
Yes
56
1232
Note: t-statistics are in parenthesis. *** indicate significance at the one percent level. The DOLS
results are based on a one lead/lag model. For Azerbaijan, Georgia, and Puerto Rico, complete time
series data on infant mortality are not available for the period 19812005. Therefore, these countries
are not included in the sample used to estimate the coefficient in column 2.
We also report results based on the income share of the bottom quintile,
labeled BottomQuintile. Balanced panel data on this measure of inequality are not
available, which prevents us from using the DOLS procedure. Instead, we employ
the ARDL model. The results in Table 10 show that the long-run coefficient on
BottomQuintile is always positive and significant, indicating that inequality has
negative consequences for health in developing countries. In addition, we used the
ARDL model to perform estimations with the incidence of tuberculosis as a
specific measure of morbidity as in Table 7 above for the sample of developed
countries. 30 The Gini index enters with a positive coefficient in the estimations for
developing countries with the incidence of tuberculosis as the dependent
variable,instead of life expectancy. This is in contrast to the corresponding result
for developed countries, even though the coefficient on the Gini index fails to
reach statistical significance at conventional levels. In other words, we again find
that the results for developed countries do not carry over to developing countries.
As a final robustness check, we re-estimate the DOLS regression (with
demeaned data) for our overall sample of developed and developing countries for
_________________________
30 For the sake of brevity, the results are not shown in detail. They are available from the authors on
request.
34
Table 10: Long-run-estimates for developing countries using the income share of the
bottom quintile of the population as the inequality measure, 1983-2011
Exogenous variable
BottomQuintile
(1)
0.4091*
(0.0538)
(2)
0.4082**
(0.1974)
-0.3345
(0.3012)
Yes
Yes
32
230
Yes
Yes
32
226
GDP per capita ($ 1000s)

GDP per capita squared
($ 1000s)
Year dummies
Country dummies
Countries
Observations
(3)
0.5086**
(0.2481)
0.0372
(1.0622)
-0.0515
(0.0865)
Yes
Yes
32
226
Note: The dependent variable is LE. Standard errors (calculated by the Delta method) are in
parentheses. ** (*) indicate significance at the five (ten) percent level. The long-run effects were
estimated from an autoregressive distributed lag model with two lags on the endogenous variable and
two lags on the exogenous variable(s), given the relatively long period covered in the data.
the period 1981-2005 using an interaction term between the Gini coefficient and
GDP per capita. Since GDP per capita is part of the interaction term, GDP per
capita is included in the DOLS regression. In addition, to be consistent with the
specification in column 4 of Table 5, we include GDP per capita squared, while
the country-specific time trends are excluded from the regression (to avoid
multicollinearity problems). This yields the following equation (t-statistics in
parenthesis, *** (**) indicate significance at the one (five) percent level): 31
LEit = 0.0249 * * * Gini + 0.0010 * *

GDP per capita ($ 1000s) Gini
( 2.45)
(4.94)
+ 0.0380 * * GDP per capita ($ 1000s)
_________________________
31 For Azerbaijan, Estonia, Ireland, Kazakhstan, Kyrgyz Republic, Lithuania, Poland, Russian
Federation, Tajikistan, Tanzania, Turkmenistan, Ukraine, and Uzbekistan, complete time series data
on GDP per capita are not available for the period 1981-2005. Therefore, these (13) countries are not
included in the sample used to estimate the interaction term regression.
35
(1.98)
0.0023 * * * GDP per capita squared ($ 1000s)
(7.37)
(6)
While the coefficient on the Gini variable is negative and statistically

significant, the coefficient on the interaction term is positive and statistically
significant. The implication of this result is that the effect of inequality on life
expectancy is negative for countries with lower levels of GDP per capita and
positive for countries with higher levels of GDP per capita.
To specify the threshold at which the effect becomes positive, we differentiate
the above equation with respect to Gini, dLE/dGini = 0.0249 + 0.0010 GDP per
capita ($ 1000s). Setting the derivative of LE with respect to Gini equal to zero and
solving for GDP per capita yields a threshold of about 24,900 US$. Among the 80
countries in the sample, 16 countries have average incomes above the threshold at
which inequality increases life expectancy. 32 All these countries are high-income
or developed countries according to World Bank (1995) classification.
Conclusions
The widely held belief that more unequal societies are less healthy is politically
highly relevant. Calls for redistributive policy interventions in order to improve
health and ensure longer life expectancy would be justified, particularly if wide
income gaps represent a major aspect of inequality within countries. This provided
the motivation to re-assess Wilkinsons (1996) verdict that the distribution of
income is one of the most powerful determinants of the health of whole
populations. While a few recent studies have doubted this verdict for developed
countries, the scant evidence available so far for developing countries posed the
important question of whether the experience of developed countries would also
hold for lower-income countries. Our empirical analysis addressed these issues. In
addition, we attempted to overcome several limitations of previous research by
_________________________
32 The 16 countries above the threshold are Australia, Belgium, Canada, Denmark, Finland, France,
Germany, Ireland, Italy, Japan, Netherlands, Norway, Sweden, Switzerland, the United Kingdom,
and the United States.
36
employing panel cointegration techniques, which allowed us to account for

endogeneity concerns.
We found that income inequality has a significantly positive impact on
population health in developed countries. Even though wider income gaps increase
life expectancy only by a quantitatively small margin, this result proved to be
robust to modifications in measurement, specification and methodological choices.
It was surprising to find wider income gaps to cause slightly better health
outcomes in developed countries, although the findings of Mellor and Milyo
(2001) as well as Leigh and Jencks (2007) pointed into the same direction.
Our parsimonious empirical model leaves it to future research to explore in
more detail the channels through which inequality affects health. Interdisciplinary
research is called for to evaluate the role played by factors such as trust,
psychosocial stress, health care and public goods provision, political participation
and representation, and different reference groups as possible intermediating
variables. Specifically, concepts and insights from psychology, biology, political
science and economics may help explain the surprisingly positive effects of
inequality on health in developed countries. E.g., certain stress-related health risks
may center on higher-income ranks if hierarchies are unstable and dominant
individuals constantly need to defend their position. More inequality may be
associated with higher government spending on health care, as the MeltzerRichard theorem would predict. Health conditions, including for poorer population
segments, may also improve in line with Miller and Paxson (2006) if the superiorgoods character of medical and environmental services induces stronger demand
for such services by richer people beyond a certain income threshold. Specifically,
high income inequality in developed countries may be associated with higher
quality and better organization of health systems which could benefit also poor
segments of the population through spillover effects.
In contrast to developed countries, we found that people living in developing
countries with wider income inequality have a significantly lower life expectancy
than people living in developing countries with a more equal distribution of
income. The health-impairing effect of income inequality in developing countries
is quantitatively small, but fairly robust. Again, the exact transmission channels
would have to be identified in future interdisciplinary research. However, it
appears that the case for redistributive policy interventions to improve health and
increase life expectancy is stronger for developing countries than for developed
37
countries. This is in some conflict with predictions from the absolute income or
poverty hypothesis, according to which health primarily depends on the incidence
of poverty in low-income countries. Unfortunately, our findings suggest that
progressive income taxation might be advisable for health reasons exactly where
wide income gaps tend to be most difficult to redress via taxationdue to
insufficient administrative capacity and political resistance of local elites.
Consequently, the preferred policy response may still consist of targeted pro-poor
interventions with regard to the provision of health services. Improving the
education of poor population segments could provide another indirect handle to
tackle the health-impairing effects of income inequality.
Generally speaking, health policies in both developing and developed
countries should take into account that income disparity is just one manifestation
of inequality. Paraphrasing Deaton (2003: 152), inequality may be important for
health, even though the quantitative impact of income inequality on health is rather
small and working in opposite directions. In addition to helping identify
transmission channels, interdisciplinary research could also provide further
insights into the links between different aspects of inequality and health conditions
in developed and developing countries. Furthermore, deeper insights may be
gained once persistent data constraints are relaxed. Continued efforts to collect
information on mortality by age cohorts as well as life expectancy adjusted for
morbidity and time spent in poor health are of particular importance in this regard.
Finally, the measurement of health conditions should be refined in order to reveal
differences within countries, notably between particularly poor and richer
population segments in developing countries.
Acknowledgements: We would like to thank an anonymous referee as well as Michael

Grimm and Hermann Sautter for most useful comments and helpful suggestions.
38
Appendix 1: Main Variables by Country over the Period 19812005

Figure A.1: Life expectancy at birth by country over the period 19812005
Note: The gure includes all (80) countries for which complete time series data on life expectancy
and the Gini coefficient are available over the period 19812005. The countries from left to right are:
Argentina, Australia, Azerbaijan, Bangladesh, Belgium, Brazil, Bulgaria, Canada, Chile, China,
Colombia, Costa Rica, Cote dÌvoire, Denmark, Egypt, El Salvador, Estonia, Ethiopia, Fiji, Finland,
France, Georgia, Germany, Greece, Guatemala, Hong Kong, Hungary, India, Indonesia, Ireland,
Israel, Italy, Japan, Jordan, Kazakhstan, Kenya, Korea, Kyrgyz Republic, Latvia, Lithuania,
Madagascar, Malawi, Malaysia, Mauritius, Mexico, Morocco, Nepal, Netherlands, New Zealand,
Nigeria, Norway, Pakistan, Panama, Peru, Philippines, Poland, Portugal, Puerto Rico, Russian
Federation, Sierra Leone, Singapore, South Africa, Spain, Sri Lanka, Sweden, Switzerland,
Tajikistan, Tanzania, Thailand, Trinidad and Tobago, Tunisia, Turkey, Turkmenistan, Ukraine, the
United Kingdom, the United States, Uruguay, Uzbekistan, Venezuela, and Zambia.
39
Figure A.2: Gini coefficient by country over the period 19812005
Note: The gure includes all (80) countries for which complete time series data on life expectancy
and the Gini coefficient are available over the period 19812005. The countries from left to right are:
Argentina, Australia, Azerbaijan, Bangladesh, Belgium, Brazil, Bulgaria, Canada, Chile, China,
Colombia, Costa Rica, Cote dÌvoire, Denmark, Egypt, El Salvador, Estonia, Ethiopia, Fiji, Finland,
France, Georgia, Germany, Greece, Guatemala, Hong Kong, Hungary, India, Indonesia, Ireland,
Israel, Italy, Japan, Jordan, Kazakhstan, Kenya, Korea, Kyrgyz Republic, Latvia, Lithuania,
Madagascar, Malawi, Malaysia, Mauritius, Mexico, Morocco, Nepal, Netherlands, New Zealand,
Nigeria, Norway, Pakistan, Panama, Peru, Philippines, Poland, Portugal, Puerto Rico, Russian
Federation, Sierra Leone, Singapore, South Africa, Spain, Sri Lanka, Sweden, Switzerland,
Tajikistan, Tanzania, Thailand, Trinidad and Tobago, Tunisia, Turkey, Turkmenistan, Ukraine, the
United Kingdom, the United States, Uruguay, Uzbekistan, Venezuela, and Zambia.
40
Appendix 2: Panel unit root and cointegration tests

Appendix 2.1: Panel unit root tests
One of the most commonly employed tests for unit roots in panels is that of Im,
Pesaran and Shin (2003), the IPS test. It tests the null hypothesis that all of the
individuals of the panel have a unit root against the alternative that some fractions
are (trend) stationary using the augmented Dickey-Fuller (ADF) regression for the
ith cross-section unit
ki
xit = z it i + i xit 1 + j ij xit j + it
(A.1)
j =1
here ki is the lag order, zit represents deterministic terms, such as fixed effects or
fixed effects combined with individual time trends, and is the first-difference
operator. To test the unit root null hypothesis, H 0 : i = 0 , i =1, 2, , N,
against the alternative of (trend) stationarity, H 1 : i < 0 , i = 1, 2, , N1 ; i = 0 ,
i = N1 + 1 , N1 + 2 , , N, a standardized t-bar statistic is constructed as
t =
N [t NT ]
v
(A.2)
where t NT is the average of the N (=19) cross-sectional ADF t-statistics, and

and are, respectively, the mean and variance of the average of the individual tstatistics, tabulated by Im et al. (2003).
However, the IPS test procedure assumes cross-sectional independence and
can thus lead to spurious inferences if the errors, it, are not independent across i
(e.g., due to common shocks or spillovers between countries). Therefore, we also
employ the cross-sectionally augmented IPS test proposed by Pesaran (2007). This
test is designed to filter out the cross-section dependency by augmenting the ADF
regression with the cross-section averages of lagged levels and first differences of
the individual series. Accordingly, the cross-sectionally augmented ADF (CADF)
regression is given by
ki
ki
j =1
j =0
xit = z it ' i+ i xit 1 + j ij xit j + i xt 1 + ij xt j + vit ,
(A.3)
41
x . The cross-sectionally
where xt is the cross-section mean of xit, xt = N
i =1 it
augmented IPS statistic is the simple average of the individual CADF statistics and
is defined as
1
CIPS = t-bar = N
Ni
t
i =1
(A.4)
where ti is the OLS t ratio of i in Equation (A.3). The corresponding critical

values are given by Pesaran (2007).
The results of the two tests for the variables in levels and in first differences
are reported in Table A.1. Both tests fail to reject the unit root null hypothesis in
levels, whereas the unit root hypothesis is rejected for the first differences. From
this we conclude that LEit and Giniit are integrated of order 1, I(1) the necessary
condition for cointegration in a bivariate context.
Table A.1: Panel unit root tests
Variables
Deterministic terms
IPS statistics
CIPS statistics
Constant, trend
0.58
-2.16
Constant, trend
0.81
-2.10
Constant
-6.00***
-2.41***
Constant
-3.83***
-2.62***
Levels
LE
Gini
First differences
LE
Gini
Note: Three lags were selected to adjust for autocorrelation. The IPS statistic is distributed as N(0, 1).
The relevant five- (one-) percent critical value for the CIPS statistics is -2.71 (-2.85) with an intercept
and a linear trend, and -2.20 (-2.36) with an intercept. *** denote significance at the one percent
level.
42
Appendix 2.2: Panel cointegration tests

We use several panel cointegration tests to examine whether there is a long-run
relationship between live expectancy at birth and income inequality. The first is
the two-step residual-based procedure suggested by Pedroni (1999, 2004), which
can be intuitively described as follows. In the first step, the hypothesized
cointegrating relationship
LEit = ai + i t + bi Giniit + it
(A.5)
is estimated separately for each country. In the second step, the residuals from
these regressions are tested for stationarity based on
ki
it = iit 1 + j ij it j + wit
(A.6)
j =1
To test the null hypothesis of a unit root or no cointegration, H 0 : i = 1 ,

Pedroni proposes two classes of test statistics. The first category pools the
autoregressive coefficients across different countries during the unit-root test and
thus constrains the autoregressive parameters to be homogeneous across countries,
i = . Pedroni refers to these within-dimension-based statistics as panel
cointegration statistics. The second class of statistics averages the individually
estimated autoregressive coefficients for each country, thus allowing the
autoregressive parameter to be heterogeneous across countries. Pedroni refers to
these between-dimension-based statistics as group-mean panel cointegration
statistics. The panel cointegration statistics include a non-parametric variance ratio
statistic (panel v), a non-parametric Phillips and Perron type -statistic (panel ), a
non-parametric Phillips and Perron type t-statistic (panel PP) and a Dickey-Fuller
type t-statistic (panel ADF). Similarly, the group-mean panel cointegration
statistics include a Phillips and Perron type -statistic (group ), a Phillips and
Perron type t-statistic (group PP) and an ADF type t-statistic (group ADF). The
standardized distributions for the panel and group statistics are given by
j N
v
N (0, 1)
(A.7)
43
where j is the respective panel or group statistic, and and are the expected
mean and variance of the corresponding statistic, tabulated by Pedroni (1999).
However, standard panel cointegration tests such as those of Pedroni (1999,
2004) assume cross-sectional independence and can have size distortions when
this assumption is violated. To test for cointegration in the presence of possible
cross-sectional dependence, we use a two-step residual-based procedure in the
style of Holly et al. (2010). In the first step, we apply the common correlated
effects (CCE) estimator of Pesaran (2006) to the static cointegrating regression.
Like the cross-sectionally augmented IPS test, the CCE estimator allows for crosssectional dependencies that potentially arise from multiple unobserved common
factors. The cross-sectionally augmented cointegrating regression for the ith crosssection is given by
LEit = ai + i t + bi Giniit + g1i LEt + g 2i Ginit + eit
(A.8)
where the cross-sectional

averages LEt = N 1 Ni LEit and Ginit =
1 N
N i Giniit serve as proxies for the unobserved factors. In the second step, we
compute the cross-sectionally augmented IPS statistic for the residuals from the
individual CCE long-run relations, it = LEit i t bi Giniit , including an
intercept. In doing so, we account for unobserved common factors that could be
correlated with the observed regressors in both steps.
A drawback of residual-based (panel) cointegration tests is that they are
generally not invariant to the normalization of the cointegrating regression.
Therefore, we also use the Larsson et al. (2001) procedure, which is based on
Johansens (1988) maximum likelihood estimation procedure. Like the Johansen
time-series cointegration test, the Larsson et al. panel test treats all variables as
potentially endogenous, thus avoiding the normalization problems inherent in
residual-based cointegration tests. In addition, the Larsson et al. procedure allows
the long-run elasticities to differ from the short-run elasticities and hence does not
impose a possibly invalid common factor restriction. It involves estimating the
Johansen vector-error-correction model for each individual country:
i
yit = i yit 1 + ik yit k + zit i+ it ,
(A.9)
44
where yit is a p 1 vector of endogenous variables ( yit = [ LEit ; Giniit ]'; p is the
number of variables), and i is the long-run matrix of order p p. If i is of
reduced rank, ri < p, it is possible to let i = i i , where i is a p ri matrix,
the ri columns of which represent the cointegrating vectors, and i is a p ri
matrix having p rows which represent the error-correction coefficients. The null
hypothesis is that all of the N (=19) countries in the panel have a common cointegrating rank, i.e., at most r (possibly heterogeneous) cointegrating
relationships among the p variables: H 0 : rank ( i ) = ri r for all i = 1, ... , N .
The alternative hypothesis is that all the cross-sections have a higher rank:
H 1 : rank ( i ) = p for all i = 1, ... , N . To test H 0 against H 1 , a panel
cointegration rank trace-test statistic is computed by calculating the average of the
individual trace statistics, LRiT {H (r ) H ( p )} :
LR NT {H (r ) H ( p)} =
1 N
LRiT {H (r ) H ( p)} ,
N i =1
(A.10)
and then standardizing it as follows:

LR {H (r ) H ( p )} =
N LR NT {H (r ) H ( p )} E ( Z k )
Var ( Z k )
) N (0, 1)
.
..(A.11)
The mean E ( Z k ) and variance Var ( Z k ) of the asymptotic trace statistic are
tabulated by Breitung (2005) for the model (with an intercept and a trend) we use.
However, it is well known that the Johansen trace statistics are biased toward
rejecting the null hypothesis in small samples. To avoid the Larsson et al. test, as a
consequence of this bias, also overestimating the cointegrating rank, we
additionally compute the standardized panel trace statistics based on small-sample
corrected country-specific trace statistics. Specifically, we use the small-sample
correction factor suggested by Reinsel and Ahn (1992) to adjust the individual
trace statistics as follows:
T ki p
LRiT {H (r ) H ( p )}
(A.12)
45
where ki is the lag length of the models used in the test.

The results of these tests are presented in Table A.2. All test statistics indicate
that LEit and Giniit are cointegrated (or exhibit a single cointegrating vector).
Table A.2: Panel cointegration tests
Panel -statistic
Panel -statistic
Panel PP-statistic
Panel ADF-statistic
Group -statistic
Group PP-statistic
Group ADF-statistic
CIPS statistic for the residuals of
the CCE long-run relations
Panel trace statistics (unadjusted)

Panel trace statistics (adjusted)
33.39***
-1.99**
-4.83***
-7.54***
-1.78**
-3.78***
-4.15***
-3.37***
Cointegration rank
r=0
5.63***
4.50***
r=1
0.56
0.05
Note: *** (**) indicate a rejection of the null hypothesis of no cointegration/no cointegrating vector
at the one (five) percent level. The relevant one-percent critical value for the CIPS statistic is -2.36.
All other test statistics are asymptotically normally distributed. The right tail of the normal
distribution is used to reject the null hypothesis in the panel -statistic and the panel trace statistic,
while the left tail is used for the other statistics. One lag was used to form the panel trace and the
CIPS statistics. For all other statistics, the number of lags was determined by the Schwarz criterion
with a maximum of seven lags.
46
References
Alesina, A., R. Baqir, and W. Easterly (1999). Public goods and ethnic divisions.
Quarterly Journal of Economics 114(4), 12431284.
http://qje.oxfordjournals.org/content/114/4/1243.full.pdf+html
Alesina, A., and E. La Ferrara (2002). Who trusts others? Journal of Public Economics,
85(2), 207234.
http://www.sciencedirect.com/science/article/pii/S0047272701000846
Anderson, M., and M. Marmot (2012). The effects of promotions on heart disease:
Evidence from Whitehall. Economic Journal 122(561), 555589.
http://onlinelibrary.wiley.com/doi/10.1111/j.1468-0297.2011.02472.x/pdf
Atkinson, A.B., and A. Brandolini (2001). Promise and pitfalls in the use of secondary
data-sets: income inequality in OECD countries as a case study. Journal of Economic
Literature 39(3), 771799.
http://pubs.aeaweb.org/doi/pdfplus/10.1257/jel.39.3.771
Attanasio, O.P., L. Picci, and A.E. Scorcu (2000). Saving, growth, and investment: a
macroeconomic analysis using a panel of countries. Review of Economics and
Statistics 82(2), 182211.
http://www.mitpressjournals.org/doi/pdf/10.1162/003465300558731
Baltagi, B.H., and J.M. Griffin (1997). Pooled estimators vs. their heterogeneous
counterparts in the context of dynamic demand for gasoline. Journal of Econometrics
77(2), 303327.
http://www.sciencedirect.com/science/article/pii/S0304407696018027#
Beckfield, J. (2004). Does income inequality harm health? New cross-national evidence.
Journal of Health and Social Behavior 45(3), 231248.
http://hsb.sagepub.com/content/45/3/231.full.pdf+html
Bonham, C.S., and R.H. Cohen (2001). To aggregate, pool, or neither. Testing the rationalexpectations hypothesis using survey data. Journal of Business and Economic
Statistics 19(3), 278291.
http://www.tandfonline.com/doi/abs/10.1198/073500101681019936#.VJAs93v2tOY
Borghesi, S., and A. Vercelli (2004). Globalisation, inequality and health. International
Journal of Global Environmental Issues 4(1/2/3), 89108.
http://inderscience.metapress.com/content/1c81jv0em9vkehph/?genre=article&issn=1
466-6650&volume=4&issue=1-2-3&spage=89
Breitung, J. (2005). A parametric approach to the estimation of cointegrating vectors in
panel data. Econometric Reviews 24(2), 151173.
http://www.tandfonline.com/doi/abs/10.1081/ETC-200067895#.VJAttXv2tOY
47
Bun, M.J.G., and J.F. Kiviet (2006). The effects of dynamic feedbacks on LS and MM
estimator accuracy in panel data models. Journal of Econometrics 132(2), 409444.
Chambers, M.J. (2001). Cointegration and sampling frequency. Department of Economics
Discussion Paper No. 531, University of Essex.
http://essex.ac.uk/economics/discussion-papers/Papers-text/dp531.pdf
Click, R.W., and M.G. Plummer (2005). Stock market integration in ASEAN after the
Asian financial crisis. Journal of Asian Economics 16(1), 528.
Cole, W.M. (2013). Strong walk and cheap talk: The effect of the international covenant of
economic, social, and cultural rights on policies and practices. Social Forces 92(1),
165194.
http://sf.oxfordjournals.org/content/early/2013/06/12/sf.sot074.full.pdf+html
Davidson, J. (1998). Structural relations, cointegration and identification: some simple
results and their application. Journal of Econometrics 87(1), 87113.
Deaton, A. (2003). Health, inequality, and economic development. Journal of Economic
Literature 41(1), 113158.
http://pubs.aeaweb.org/doi/pdfplus/10.1257/002205103321544710
Deininger, K., and L. Squire (1996). A new data set measuring income inequality. World
Bank Economic Review 10(3), 565591.
http://wber.oxfordjournals.org/content/10/3/565.full.pdf+html
Desbordes, R., and V. Verardi (2012). Refitting the Kuznets curve. Economics Letters
116(2), 258261.
Dumitrescu, E.-I., and C. Hurlin (2012). Testing for Granger non-causality in
heterogeneous panels. Economic Modelling 29(4), 14501460.
dHombres B., L. Rocco, M. Suhrcke, and M. McKee (2010). Does social capital
determine health? Evidence from eight transition countries. Health Economics 19(1),
5674.
http://onlinelibrary.wiley.com/doi/10.1002/hec.1445/pdf
Engle, R.E., and C.W.J Granger (1987). Co-integration and error-correction:
Representation, estimation, and testing. Econometrica 55(2), 251276.
https://ideas.repec.org/a/ecm/emetrp/v55y1987i2p251-76.html
48
Entorf, H. (1997). Random walks with drifts: Nonsense regression and spurious fixedeffects estimation. Journal of Econometrics 80(2), 287296.
Etienne, J.-M., A. Skalli, and I. Theodossiou (2012) Do economic inequalities harm
health? Evidence from Europe. Centre for European Labour Market Research,
Discussion Paper 200713. Aberdeen, 2007.
http://aura.abdn.ac.uk/bitstream/2164/131/1/ISSNpercent200143-07-13.pdf
Everaert, G. (2011). Estimation and inference in time series with omitted I(1) variables.
Journal of Time Series Econometrics 2(2), 126.
http://www.degruyter.com/view/j/jtse.2011.2.2/jtse.2011.2.2.1054/jtse.2011.2.2.1054.
xml?format=INT
Firebaugh, G. (2003). The new geography of global income inequality. Harvard University
Press, Cambridge, MA.
Flegg, A.T. (1982). Inequality of income, illiteracy, and medical care as determinants of
infant mortality in developing countries. Population Studies 36(3), 441458.
http://www.jstor.org/discover/10.2307/2174055?sid=21105469654993&uid=3737864
&uid=4&uid=2
Granger, C.W.J., and P. Newbold (1974). Spurious regressions in econometrics. Journal of
Econometrics 2(2), 111120.
http://www.sciencedirect.com/science/article/pii/0304407674900347
Gravelle, H. (1998). How much of the relation between population mortality and unequal
distribution of income is a statistical artefact? British Medical Journal 316(7128),
382385.
http://www.bmj.com/content/316/7128/382
Grossman, M. (1972). On the concept of health capital and the demand for health. Journal
of Political Economy 80(2), 223255.
http://www.jstor.org/discover/10.2307/1830580?sid=21105469654993&uid=4&uid=2
&uid=3737864
Henderson, J.W. (2009). Health economics and policy. Mason, Ohio: SouthWestern
Cengage Learning, 2009.
Herzer, D. (2008). The long-run relationship between outward FDI and domestic output:
evidence from panel data. Economics Letters 100(1), 146149.
Herzer, D. (2013). Cross-country heterogeneity and the trade-income relationship. World
Development 44(4), 194211.
http://www.sciencedirect.com/science/article/pii/S0305750X12002288
49
Herzer, D., H. Strulik, and S. Vollmer (2012). The long-run determinants of fertility: one
century of demographic change 19001999. Journal of Economic Growth 17(4), 357
385.
http://link.springer.com/article/10.1007%2Fs10887-012-9085-6
Herzer, D., and P. Nunnenkamp (2014). Income inequality and health: evidence from
developed and developing countries. Discussion Paper No. 2014-45, November 12,
Economics - The Open-Access, Open-Assessment E-Journal.
http://www.economics-ejournal.org/economics/discussionpapers/2014-45
Holly, S., M.H. Pesaran, and T. Yamagata (2010). A spatio-temporal model of house
prices in the USA. Journal of Econometrics 158(1), 160173.
Im, K.S., M.H. Pesaran, and Y. Shin (2003). Testing for unit roots in heterogeneous panels.
Journal of Econometrics 115(1), 5374.
Jenkins, S.P. (2014). World income inequality databases: An assessment of WIID and
SWIID. Working Paper 2014-31. Institute for Social and Economic Research.
https://www.iser.essex.ac.uk/research/publications/working-papers/iser/2014-31.pdf
Johansen, S. (1988). Statistical analysis of cointegrating vectors. Journal of Economic
Dynamics and Control 12(23), 231254.
http://www.sciencedirect.com/science/article/pii/0165188988900413
Judge, K., J.-A. Mulligan, and M. Benzeval (1998). Income inequality and population
health. Social Science & Medicine 46(4), 567579.
Judson, R.A., and A.L. Owen (1999). Estimating dynamic panel data models: A guide for
macroeconomists. Economics Letters 65(1), 915.
Juselius, K. (1996). An empirical analysis of the changing role of the German Bundesbank
after 1983. Oxford Bulletin of Economics and Statistics 58 (4), 791819.
http://onlinelibrary.wiley.com/doi/10.1111/j.1468-0084.1996.mp58004010.x/abstract
Juselius K. (2006). The cointegrated VAR model: Methodology and applications. Oxford:
Oxford University Press.
Kao, C. (1999). Spurious regression and residual-based tests for cointegration in panel
data. Journal of Econometrics 90(1), 144.
Kao, C., and M. Chiang (2000). On the estimation and inference of a cointegrated
regression in panel data. Advances in Econometrics 15, 179222.
http://www.emeraldinsight.com/doi/abs/10.1016/S0731-9053%2800%2915007-8
50
Karlsson, M., T. Nilsson, C.H. Lyttkens, and G. Leeson (2010). Income inequality and
health: Importance of a cross-country perspective. Social Science and Medicine 70(6),
875885.
Kawachi, I., and B.P. Kennedy (1999). Income inequality and health: Pathways and
mechanisms. Health Services Research 34(1 Pt 2), 215227.
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1088996/
Kawachi, I., B.P. Kennedy, K. Lochner, and D. Prothrow-Stith (1997). Social capital,
income inequality, and mortality. American Journal of Public Health 87(9), 1491
1498.
Knack, S. (2002). Social capital, growth, and poverty: A survey of crosscountry evidence,
in: Grootaert, C., and van Bastelaer, T. (eds.) The Role of Social Capital in
Development: An Empirical Assessment. Cambridge: Cambridge University Press,
4284.
Kondo, N., G. Sembajwe, I. Kawachi, R.M. van Dam, S.V. Subramanian, and Z. Yamagata
(2009). Income inequality, mortality, and self rated health: Meta analysis of multilevel
studies. British Medical Journal, 339(b4471), 11781181.
http://www.bmj.com/content/339/bmj.b4471
Larsson, R., J. Lyhagen, and M. Lthegren (2001). Likelihood-based cointegration tests in
heterogeneous panels. Econometrics Journal 4(1), 109142.
Leigh, A., C. Jencks, and T.M. Smeeding (2009). Health and economic inequality, in: W.
Salverda, Nolan, B., and Smeeding, T.M. (eds.) The Oxford Handbook of Economic
Inequality. Oxford: Oxford University Press, 384405.
http://onlinelibrary.wiley.com/doi/10.1111/1368-423X.00059/abstract
Leigh, A. (2006). Does equality lead to fraternity? Economics Letters 93(1), 121125.
Leigh, A. (2007). How closely do top income shares track other measures of inequality?
Economic Journal 117(524), F619F633.
http://onlinelibrary.wiley.com/doi/10.1111/j.1468-0297.2007.02099.x/abstract
Leigh, A., and C. Jencks (2007). Inequality and mortality: Long-run evidence from a panel
of countries. Journal of Health Economics 26(1), 124.
Leigh, A., C. Jencks, and T.M. Smeeding (2009). Health and economic inequality, in:
Salverda, W., Nolan, B., and Smeeding, T.M. (eds.) The Oxford Handbook of
Economic Inequality. Oxford: Oxford University Press, 384405.
51
Ltkepohl, H., and H.E., Reimers (1992). Granger-causality in cointegrated VAR

processes. The case of the term structure. Economics Letters, 40(3), 263268.
http://www.sciencedirect.com/science/article/pii/016517659290002G
Ltkepohl, H. (2007). General-to-specific or specific-to-general modelling? An opinion on
current econometric terminology. Journal of Econometrics 136(1), 319324.
Lynch, J., G.A. Kaplan, E.R. Pamuk, R.D. Cohen, K.F. Heck, J.L. Balfour, and I.H. Yen
(1998). Income inequality and mortality in metropolitan areas of the United States.
American Journal of Public Health 88(7), 10741080.
Lynch, J., G.D. Smith, S. Harper, M. Hillemeier, N. Ross, G.A. Kaplan, and M. Wolfson
(2004). Is income inequality a determinant of population health? Part 1. A systematic
review. The Milbank Quarterly 82(1), 5-99.
http://www.ncbi.nlm.nih.gov/pubmed/15016244
Mark, N.C., and D. Sul (2003). Cointegration vector estimation by panel DOLS and longrun money demand. Oxford Bulletin of Economics and Statistics 65(5), 655680.
http://onlinelibrary.wiley.com/doi/10.1111/j.1468-0084.2003.00066.x/abstract
Marmot, M. (2003). Understanding social inequalities in health. Perspectives in Biology
and Medicine 46(3), S9S23.
http://muse.jhu.edu/journals/pbm/summary/v046/46.3xmarmot.html
Mellor, J.M., and J. Milyo (2001). Reexamining the evidence of an ecological association
between income inequality and health. Journal of Health Politics, Policy and Law
26(3), 487522.
http://jhppl.dukejournals.org/content/26/3/487.abstract
Meltzer, A.H., and S.F. Richard (1981). A rational theory of the size of government.
Journal of Political Economy 89(5), 914927.
https://ideas.repec.org/a/ucp/jpolec/v89y1981i5p914-27.html
Miller, D.L., and C. Paxson (2006). Relative income, race, and mortality. Journal of
Health Economics 25(5), 9791003.
Morgan, J., and N.J. Kelly (2013). Market inequality and redistribution in Latin America
and the Caribbean. Journal of Politics 75(3), 672685.
http://journals.cambridge.org/action/displayAbstract?fromPage=online&aid=8948630
Moscone, F., and E. Tosetti (2010). Health expenditure and income in the United States.
Health Economics 19(12), 13851403.
http://onlinelibrary.wiley.com/doi/10.1002/hec.1552/abstract
52
Nickell, S. (1981). Biases in dynamic models with fixed effects. Econometrica 49(6),
14171426.
Pedroni, P. (1999). Critical values for cointegration tests in heterogeneous panels with
multiple regressors. Oxford Bulletin of Economics and Statistics 61(Special Issue
Nov.), 653670.
http://onlinelibrary.wiley.com/doi/10.1111/1468-0084.0610s1653/abstract
Pedroni, P. (2001). Purchasing power parity tests in cointegrated panels. Review of
Economics and Statistics 83(4), 727731.
http://www.mitpressjournals.org/doi/pdf/10.1162/003465301753237803
Pedroni, P. (2004). Panel cointegration: Asymptotic and finite sample properties of pooled
time series tests with an application to the PPP hypothesis. Econometric Theory 20(3),
597625.
http://journals.cambridge.org/action/displayAbstract?fromPage=online&aid=226437
&fulltextType=RA&fileId=S0266466604203073
Pedroni, P. (2007). Social capital, barriers to production and capital shares: Implications
for the importance of parameter heterogeneity from a non-stationary panel approach.
Journal of Applied Econometrics 22(2), 429451.
http://onlinelibrary.wiley.com/doi/10.1002/jae.948/abstract
Pesaran, M.H., and R. Smith (1995). Estimating long-run relationships from dynamic
heterogeneous panels. Journal of Econometrics 68(1), 79113.
http://www.sciencedirect.com/science/article/pii/030440769401644F
Pesaran, M.H. (2006). Estimation and inference in large heterogeneous panels with a
multifactor error structure. Econometrica 74(4), 9671012.
Pesaran, M.H. (2007). A simple panel unit root test in the presence of cross-section
dependence. Journal of Applied Econometrics 22(2), 265312.
http://onlinelibrary.wiley.com/doi/10.1002/jae.951/pdf
Phillips, P., and H. Moon (1999). Linear regression limit theory for nonstationary panel
data. Econometrica 67(5), 10571112.
Preston, S.H. (1975). The changing relation between mortality and level of economic
development. Population Studies 29(2), 231248.
http://ije.oxfordjournals.org/content/36/3/484.full
53
Putnam, R.D. (1995). Bowling alone: Americas declining social capital. Journal of
Democracy 6(1), 6578.
http://muse.jhu.edu/login?auth=0&type=summary&url=/journals/journal_of_democra
cy/v006/6.1putnam.html
Reinsel, G.C., and S.K. Ahn (1992). Vector autoregressive models with unit roots and
reduced rank structure: Estimation, likelihood ratio test and forecasting. Journal of
Time Series Analysis 13(4), 353375.
http://onlinelibrary.wiley.com/doi/10.1111/j.1467-9892.1992.tb00113.x/pdf
Rodgers, G.B. (1979). Income and inequality as determinants of mortality: An
international cross-section analysis. Population Studies 33(2), 343351.
http://www.tandfonline.com/doi/abs/10.1080/00324728.1979.10410449#.VJFG8Xv2t
OY
Ronconi, L., T.T. Brown, and R.M. Scheffler (2012). Social capital and self-rated health in
Argentina. Health Economics 21(2), 201208.
http://onlinelibrary.wiley.com/doi/10.1002/hec.1696/pdf
SaintPaul, G., and T. Verdier (1993). Education, democracy and growth. Journal of
Development Economics 42(2), 399407.
http://www.sciencedirect.com/science/article/pii/030438789390027K
Sapolsky, R.M. (2004). Social status and health in humans and other animals. Annual
Review of Anthropology 33, 393418.
http://www.annualreviews.org/doi/abs/10.1146/annurev.anthro.33.070203.144000
Singh, G.K., and M. Siahpush (2006). Widening socioeconomic inequalities in US life
expectancy, 1980-2000. International Journal of Epidemiology 35(4), 969979.
http://ije.oxfordjournals.org/content/35/4/969.full
Solt, F., (2009). Standardizing the world income inequality database. Social Science
Quarterly 90(2), 231242.
Soobader, M.-J., and F.B. LeClere (1999). Aggregation and the measurement of income
inequality: Effects on morbidity. Social Science & Medicine 48(6), 733744.
Spilimbergo, A., and A. Vamvakidis (2003). Real effective exchange rate and the constant
elasticity of substitution assumption. Journal of International Economics 60(2), 337
354.
Stock, J.H. (1987). Asymptotic properties of least squares estimators of cointegrating
vectors. Econometrica 55(5), 10351056.
54
Subramanian, S.V., and I. Kawachi (2004). Income inequality and health: What have we
learned so far? Epidemiologic Reviews 26(1), 7891.
http://epirev.oxfordjournals.org/content/26/1/78.full.pdf+html
Thorbecke, W., and G. Smith (2010). How would an appreciation of the Renminbi and
other East Asian currencies affect China's exports? Review of International
Economics 18(1), 95108.
Toda, H.Y., and P.C.B. Phillips (1993). Vector autoregressions and causality.
Econometrica 61(6), 13671393.
Torre, R., and M. Myrskyl (2011). Income inequality and population health: A panel data
analysis on 21 developed countries. Working Paper 2011-006. Max Planck Institute
for Demographic Research. Rostock.
Uslaner,
E.M.
(2002).
The
moral
foundations
http://www.gvpt.umd.edu/uslaner/uslanermoralfoundations.pdf
of
trust.
Wagner, M., and K. Hlouskova (2010). The performance of panel cointegration methods:
results from a large scale simulation study. Econometric Reviews 29(2), 182223.
http://www.tandfonline.com/doi/abs/10.1080/07474930903382182#.VJFJanv2tOY
Wagstaff, A., and E. van Doorslaer (2000). Income inequality and health: What does the
literature tell us? Annual Review of Public Health 21, 543567.
http://www.annualreviews.org/doi/pdf/10.1146/annurev.publhealth.21.1.543
Waldmann, R.J. (1992). Income distribution and infant mortality. Quarterly Journal of
Economics 107(4), 12831302.
http://qje.oxfordjournals.org/content/107/4/1283.full.pdf
Wennemo, I. (1993). Infant mortality, public policy and inequality a comparison of 18
industrialised countries. 1950-85. Sociology of Health and Illness 15(4), 429446.
http://onlinelibrary.wiley.com/doi/10.1111/j.1467-9566.1993.tb00354.x/pdf
Wilkinson, R.G. (1992). Income distribution and life expectancy. British Medical Journal
304(6820), 165168.
Wilkinson, R.G. (1996). Unhealthy societies: The afflictions of inequality. London and
New York: Routledge.
Wilkinson, RG. (1997). Comment: income, inequality, and social cohesion. American
Journal of Public Health 87(9), 15041506.
55
Wilkinson, R.G. (2000). Mind the Gap: Hierarchies, health, and human evolution.
London: Weidenfeld and Nicolson.
Wilkinson, R.G., and K.E. Pickett (2006). Income inequality and population health: A
review and explanation of the evidence. Social Science and Medicine 62(7): 1768
1784.
World Bank (1995). World Development Report 1995. Workers in an integrated world.
Oxford: Oxford University Press.
http://wwwwds.worldbank.org/external/default/WDSContentServer/IW3P/IB/2013/02/27/000425
962_20130227125710/Rendered/PDF/148660REPLACEMENT0WDR01995.pdf
World Bank (2000). World Development Report 200/2001. Attacking poverty. Oxford:
Oxford University Press.
http://wwwwds.worldbank.org/external/default/WDSContentServer/IW3P/IB/2007/07/12/000310
607_20070712141534/Rendered/PDF/226840CHINESE015005150001601PUBLIC1.
pdf
Zheng, H. (2012). Do people die from income inequality of a decade ago? Social Science
& Medicine 75(1), 3645.
56
Please note:
You are most sincerely encouraged to participate in the open assessment of this article. You
can do so by either recommending the article or by posting your comments.
Please go to:
http://dx.doi.org/10.5018/economics-ejournal.ja.2015-4
The Editor
Author(s) 2015. Licensed under the Creative Commons License Attribution 3.0.

Economics 2015 4

Uploaded by

Document Informationclick to expand document informationjurnal

Document Informationclick to expand document information

Copyright:

Available Formats

Economics 2015 4

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Economics 2015 4

Uploaded by

Copyright:

Available Formats

Vol. 9, 2015-4 | February 04, 2015 | http://dx.doi.org/10.5018/economics-ejournal.ja.

Income Inequality and Health: Evidence from

Income disparities have widened in various developed and developing countries

income is open to debate. Grossman (1972) regards health as a durable capital

However, deprivation and adverse health effects would be contained if inequality

Empirical models and data

Our objective is to examine the effect of income inequality on health in developed

Empirical models and econometric issues

Following common practice in (panel) cointegration studies (see, e.g., Pedroni

where the subscript i refers to one of the N cross-sectional units, i = 1, 2, ..., N ,

are required to separate the cointegrating relationships. Otherwise,

LEit ' = LEit LEt and

LEt = N 1 i =1 LEit and

which is equivalent to including time dummies. Moreover, we use a battery of

Healthit = ai + t + j Healthit j + j Inequalityit j + j Controlsit j + eit

Data and preliminary evidence

questions about the validity of multiply-imputed SWIID observations. As argued

Table 2: Countries in the sample of 59 developing countries and summary statistics,

Table 3: Preliminary results, 19812005

GDP per capita

In this section, we estimate the long-run effect of income inequality on population

The long-run effect of inequality on health in developed

Table 4: Estimates of the long-run effect of inequality on life expectancy at birth in

in the Gini coefficient in the sample (0.114214) yields a value of 0.005334,

consistent standard errors) divided by the root of the number of cross-sectional

t-statistics of the coefficients on Gini

observations, and hence statistical power. Despite these concerns, we re-estimate

GDP per capita

With GDP p.c. and

With GDP p.c.

in Table 4, column 2. This is consistent with several studies showing that

As discussed in Section 3, the finding of cointegration implies that there are no

Table 6: DOLS estimates using different measures of inequality and health

Table 7: Estimates of the long-run effect of inequality on tuberculosis incidence in

GDP per capita ($ 1000s)

and highly significant, suggesting that inequality decreases the incidence of

Table 8: Bivariate Granger causality tests, 1976-2010

presented in columns 3 and 4. As can be seen, the null hypothesis of no Granger

The long-run effect of inequality on health in developing

In Table 9, we present DOLS estimates of the long-run effect of inequality on

GDP per capita ($ 1000s)

LEit = 0.0249 * * * Gini + 0.0010 * *

While the coefficient on the Gini variable is negative and statistically

employing panel cointegration techniques, which allowed us to account for

Acknowledgements: We would like to thank an anonymous referee as well as Michael

Appendix 1: Main Variables by Country over the Period 19812005

Figure A.2: Gini coefficient by country over the period 19812005

Appendix 2: Panel unit root and cointegration tests

xit = z it i + i xit 1 + j ij xit j + it

where t NT is the average of the N (=19) cross-sectional ADF t-statistics, and

xit = z it ' i+ i xit 1 + j ij xit j + i xt 1 + ij xt j + vit ,

where ti is the OLS t ratio of i in Equation (A.3). The corresponding critical

Appendix 2.2: Panel cointegration tests

To test the null hypothesis of a unit root or no cointegration, H 0 : i = 1 ,

LEit = ai + i t + bi Giniit + g1i LEt + g 2i Ginit + eit

where the cross-sectional

yit = i yit 1 + ik yit k + zit i+ it ,

and then standardizing it as follows: