ACCOUNTING FOR THE EFFECT OF HEALTH ON ECONOMIC GROWTH

ACCOUNTING FOR THE EFFECT OF HEALTH ON ECONOMIC GROWTH
David N. Weil
October, 2006
Abstract
I use microeconomic estimates of the effect of health on individual outcomes to construct
macroeconomic estimates of the proximate effect of health on GDP per capita. I employ a
variety of methods to construct estimates of the return to health, which I combine with cross-
country and historical data on height, adult survival rates, and age at menarche. Using my
preferred estimate, eliminating health differences among countries would reduce the variance of
log GDP per worker by 9.9 percent, and reduce the ratio of GDP per worker at the 90th percentile
to GDP per worker at the 10th percentile from 20.5 to 17.9. While this effect is economically
significant, it is also substantially smaller than estimates of the effect of health on economic
growth that are derived from cross-country regressions.
David_Weil@Brown.edu. I am grateful to Joshua Angrist, Andrew Foster, Rachel Friedberg,
David Genesove, Byungdoo Sohn, and seminar participants at Ben Gurion University, the Boston
University/Harvard/MIT seminar in health economics, Brown University, University of
California at San Diego, Clemson University, Cornell University, University of Haifa, the
Harvard Center for International Development, Hebrew University, Indiana University, the
International Monetary Fund, the NBER Economic Fluctuations and Growth group, New York
University, North Carolina State University, Ohio State University, University of Pennsylvania,
University of Wisconsin, and the World Bank for helpful discussions. Suchit Arora graciously
provided his data on height and adult survival. Doug Park and Dimitra Politi provided
superlative research assistance.

---

I. Introduction
People in poor countries are, on average, much less healthy than their counterparts in rich
countries. How much of the gap in income between rich and poor countries is accounted for by
this difference in health? The answer to this question is important both for evaluating policies
aimed at improving health in developing countries and more generally for understanding the
reasons why some countries are rich and some poor.
The United States government as well as several international organizations and private
charities, have recently embarked on ambitious efforts to improve health in developing countries.
Included in these efforts are the Bush Administration’s commitment of $15 billion over five
years to fight AIDS; the Roll Back Malaria partnership launched by the World Health
Organization (WHO), World Bank, and other international organizations in 1998; and the recent
creation of the independent Global Fund for AIDS, TB, and Malaria. The primary justification
for these programs is the potential to reduce suffering and premature death among the affected
populations. However, an important secondary justification is the potential gain in economic
development that is expected to follow from health improvements. For example, the report of
the WHO’s Commission on Macroeconomics and Health [2001] states
Improving the health and longevity of the poor is an end in itself, a fundamental
goal of economic development. But it is also a means to achieving the other
development goals relating to poverty reduction. The linkages of health to
poverty reduction and to long-term economic growth are powerful, much
stronger than is generally understood. The burden of disease in some low-
income regions, especially sub-Saharan Africa, stands as a stark barrier to
economic growth and therefore must be addressed frontally and centrally in any
comprehensive development strategy.
My goal in this paper is to quantitatively assess the role that health differences play in
explaining income differences between rich and poor countries, and thus to calculate the income
1

---

gain that would result from an improvement in the health of people living in poor countries.
Economists have identified several channels through which health affects the level of
output in a country. One channel, which I call the proximate or direct effect of health, is that
healthier people are better workers. They can work harder and longer, and also think more
clearly. Beyond this proximate effect of health, there are a number of indirect channels through
which health affects output. Improvements in health raise the incentive to acquire schooling,
since investments in schooling can be amortized over a longer working life. Healthier students
also have lower absenteeism and higher cognitive functioning, and thus receive a better
education for a given level of schooling. Improvements in mortality may also lead people to save
for retirement, thus raising the levels of investment and physical capital per worker. Physical
capital per worker may also rise because the increase in labor input from healthier workers will
increase capital’s marginal product. The effect of better health on population growth is
ambiguous. In the short run, higher child survival leads to more rapid population growth. Over
longer horizons, however, lower infant and child mortality may lead to a more-than-offsetting
decline in fertility, so that the Net Rate of Reproduction falls (Bloom and Canning [2000],
Kalemli-Ozcan, Ryder, and Weil [2000]). At a much longer horizon, Acemoglu, Johnson, and
Robinson [2001] argue that the poor health environment in some parts of the world led European
colonizers to put in place extractive institutions which in turn reduce the level of output today.
In this paper, I look only at health as a proximate determinant of a country’s income – that is, I
examine the effect of better health in enabling workers to work harder and more intelligently,
holding constant the level of physical capital, education, the quality of institutions, and so on.
Examining the effect of health on economic growth is made difficult by the endogeneity
of health itself. The mechanisms that lead to a positive dependence of health on income are
fairly obvious. People who are richer can afford better food, shelter, and medical treatment.
Countries that are richer can afford higher expenditures on public health.1 Because health is
1 Pritchett and Summers [1996], using an instrumental variables procedure, find a
significant effect of national income on health, as measured by infant and child mortality. The
2

---

endogenous, it is almost impossible to use aggregate data to determine the structural effect of
health on income. I am able, however, to use structural microeconomic estimates of the direct
effect of health on individual income, which, along with aggregate data on health differences
among countries, are all that is required to measure the direct effect of health differences on
income differences among countries.
Beyond looking at health’s effect on growth, a second goal of this research is to examine
the broader question of what determines a country’s level of income. Recent research (see
Caselli [2005] for a review) has used the technique of development accounting to parse variation
in income among countries into the pieces explained by accumulation of physical capital and
human capital in the form of education, as well as remaining residual variation due to differences
in productivity. The conclusion from this literature is that residual productivity is by far the
most significant source of income differences, explaining more than half of the variance of
income. Because existing analyses do not account for health, differences in income due to health
are included in this productivity residual, along with differences in income due to institutions,
geography, culture, and so on. By accounting for variation in health among countries, I am able
to explain a fraction of this residual productivity variation, and produce a purified version of the
productivity residual.
The rest of this paper is organized as follows. Section II discusses previous literature that
instruments that they use are terms of trade shocks, the ratio of investment to GDP, the black
market premium, and the deviation of the exchange rate from PPP. A more contentious question
is the degree to which average health varies among countries for reasons other than income.
Gallup and Sachs [2001] argue that tropical areas have fundamentally worse health environments
than do the temperate parts of the world. They claim, for example, that the fact that malaria has
been eliminated in currently rich areas (such as Spain or the Southern United States) but not in
poor ones (such as sub-Saharan Africa) does not reflect differences in income, but rather the fact
that malaria’s grip is much stronger in Africa. Under this view, these fundamental differences in
the health environment present a very strong obstacle to economic growth in the tropics. In
contrast, recent work by Acemoglu, Johnson, and Robinson [2001] take the view that differences
in the fundamental health environment between countries are not large, and that the high level of
disease in tropical countries is more a result than a cause of their poverty.
3

---

has examined the link between health and economic outcomes. Section III presents a framework
for analyzing how health affects income at the individual and national level. Section IV
discusses the aggregate health indicators used in the analysis. Section V presents a variety of
estimates of the return to health indicators. Section VI looks at the magnitude of productivity
differences among countries implied by differences in health outcomes, and Section VII looks at
the contribution of health variation to variation in GDP among countries. Section VIII
concludes.

II. Background Literature
Research examining the link between health and economic outcomes, at either the
individual or national level, has generally examined two types of health measures: inputs into
health and health outcomes. Inputs into health are the physical factors that influence an
individual’s health. These include nutrition at various points in life (in utero, in childhood, and
in adulthood), exposure to pathogens, and the availability of medical care. Health outcomes are
characteristics that are determined both by an individual’s health inputs and by his genetic
endowment. Examples include life expectancy, height, the ability to work hard, and cognitive
functioning. For the purpose of explaining income differences among countries or individuals,
the key health outcome of interest is how health affects the ability to produce output. I call this
health outcome “human capital in the form of health.” We do not observe human capital in the
form of health directly, but presumably it is some combination of ability to work hard, cognitive
function, and possibly other aspects of health. In contrast to human capital in the form of health,
there are a number of health outcomes that can be observed at either the individual level, the
national level, or both. I refer to these health outcomes as health indicators.
Comparisons of health among countries can be made by looking at either inputs to health
or health indicators. Rich countries have more of almost all health inputs than poor countries.
To give some obvious examples, the fraction of the population with access to clean drinking
4

---

water, the number of physicians per capita, and the nutrient composition of the diet all differ
markedly between rich and poor countries. Similarly, rich countries today have better health
inputs than they did in the past. Comparisons of health indicators tell much the same story as
comparisons of inputs (these data are discussed further below). These facts suggest that
unobservable health outcomes, including human capital in the form of health, are also better in
rich than poor countries.
A large microeconomic literature examines the effects of varying health inputs on health
outcomes themselves, human capital attributes that are contingent on health outcomes, and
wages. In many studies, more than one of these groups of dependent variables is examined.2
Several studies (see Behrman et al. [2003], Alderman, Hoddinott, and Kinsey [2006], Maccini
and Yang [2005], Chavez, Martinez, and Soberanes, [1995]) have examined the long-run effects
of childhood nutrition, using a variety of natural and man-made experiments that provide
exogenous variation in nutrition. They find that better nutrition leads to improvements in school
completion, IQ , height, and wages. Studies of the effect of adult nutrition ( Strauss [1986],
Strauss [1997], Basta et al. [1979], Thomas et al. [2004]) similarly find positive effects on labor
input and wages. Bleakley [2007] and Miguel and Kremer [2005] find that treatment with
deworming drugs increases school attendance. Several other studies that have examined the
effects of childhood nutrition and health inputs are discussed in Section V.A.
Using these microeconomic estimates of the effects of variation in health inputs on
wages, it is possible to calculate the contribution of variance in a single health input to variation
in income among countries. For example, Behrman and Rosenzweig [2004], using variation in
birth weight among monozygotic twins as an instrument, find significant effects of intrauterine
nutrition on adult wages. Using their estimate, one can show that eliminating variation in birth
2 See Thomas and Frankenberg [2002] for an extensive review.
5

---

weight among countries would reduce the variance of log GDP per worker by 3.1 percent.3
Extending this methodology to encompass other health inputs is difficult. Comparing
rich to poor countries, there are large differences along most of the dimensions considered in the
microeconomic studies listed above, and many more as well. A complete analysis of the effects
of equalizing health inputs (and thus health) among countries would require data on how each of
these inputs differed among countries as well as a microeconomic estimate of the effect of each
input on labor productivity. Neither the data nor the relevant estimates to undertake this exercise
currently exist. A further theoretical problem with conducting such an exercise is that to add
3 Behrman and Rosenzwieg actually report a smaller number, but the two figures can be
reconciled as follows. Define yi as the log of GDP per worker in country i, bi as average birth
weight, and as predicted log GDP per worker based on the equation
,
where the value of $ is derived from the twins data. To assess how much of the world variance
in log income is accounted for by variance in birth weight, Behrman and Rosenzwieg look at the
ratio var( ) / var(y) = $2 var(b) / var (y). Using their estimated values of $=.00413 along with
data for a cross-section of countries where var(y) = 1.16 and var(b) = 32.6, this equation yields a
value of .0005, which they report as “less than one percent.”
The problem with this measure is that because is not constructed by least squares, it is
not orthogonal to the error term. That is, if , represents the factors other than birth weight that
affect log wages, then in the equation yi = + ,i, the two terms on the right hand side are not
orthogonal. A natural measure of the influence of birth weight on the variance of income is the
proportional reduction in log variance that would result from equalizing birth weight among
countries, that is, (var ( ) + 2cov ( ,,)) / var(y). This can be expanded in turn as
The term cov(y,b)/var(b) is the coefficient from a regression of log GDP on average birth weight,
which Behrman and Rosenzweig report as 0.136. Plugging this value along with the above data
into this equation yields a value of 3.1 percent.
6

---

together the different effects examined singly in microeconomic studies would require a very
strong assumption of linearity, that is, that there are no interactions among the different health
inputs.4
A second branch of the literature has attempted to answer the question “how much do
differences in health contribute to differences in income” by looking at data on health outcomes
rather than health inputs, and examining data at the national rather than individual level. These
papers present regressions of GDP per capita, GDP growth, or TFP on some measure of health
outcomes, as well as a standard set of controls. Bloom, Canning, and Sevilla [2004] report the
results of 13 such studies, which mostly reach similar quantitative results (see also Jamison, Lau,
and Wang [2005]) . Their own estimate, which comes from regressing residual productivity
(after accounting for physical capital and education) on health measures in a panel of countries is
that a one-year increase in life expectancy raises output by 4 percent.
Papers in this group suffer from severe problems of endogeneity and omitted variable
bias. For example, Bloom, Canning, and Sevilla attempt to deal with the endogeneity of health
and other inputs into production by using lagged values of these variables as instruments. The
identifying assumption required for this strategy to work – that the error term in the equation
generating health is serially correlated while the error term in the equation generating income is
not – is not explicitly stated or defended.5 More generally, the problem with the aggregate
regression approach is that, at the level of countries, it is difficult to find an empirically usable
source of variation in health, either in cross section or time series, that is not correlated with the
4 There are many cases where this required assumption of linearity is known not to hold.
For example, three different inputs – nutrition, sanitary conditions, and access to medical care –
are to some extent substitutes in combating infectious diseases. Adding together the effects of
improving one input at a time would overstate the net effect of improving all three.
5 See Mankiw [1995] for a discussion of this issue.
7

---

error term in the equation determining income.6
In this paper I pursue the same question that is addressed by the aggregate regressions, but
using a different methodology. Specifically, I construct a framework in which estimates of the
effect of variation in health inputs on individual wages can be used to generate estimates of how
differences in health, as measured by observable outcomes, contribute to differences in national
income. In other words, I use the available microeconomic estimates to create an estimate of the
importance of health at the macroeconomic level.
III. Empirical Framework
III.A Production and Wages
Start with a Cobb-Douglas aggregate production function that takes as its arguments
capital and a composite labor input,
(1)
where Y is output, K is physical capital, A is a country-specific productivity term, and i indexes
countries. The labor composite, H , is determined by
i
6 A few studies have attempted to solve the endogeneity problem by finding instruments
for health. Sachs [2003] uses a geographically based measure of “malaria ecology” to
instrument for the current prevalence of the disease, and finds that malaria has a large effect on
the level of GDP per capita. He is not able to look at the effect of overall health. There is the
further problem that a high value of the malaria ecology index may be proxying for other omitted
aspects of a tropical climate that negatively affect income. Acemoglu and Johnson [2006]
construct a measure of countries’ predicted improvements in life expectancy during the decades
around World War II based on the experiences of other countries with similar disease
environments. Using these predicted values as instruments, they find that life expectancy
improvements had a negative effect on GDP per capita and GDP per working age person.
8

---

(2) H = h v L ,
i
i i
i
where h is per-worker human capital in the form of education, v is per-worker human capital in
i
i
the form of health, and L is the number of workers.7 As discussed above, v is not the totality of
i
i
an individuals’s health; rather, it is only the aspects of individual health that are relevant for the
production of output.

The wage paid to a unit of the labor composite, w , is its marginal product,
i
(3)
.
The wage earned by worker j is a function of his own health and education, as well as the
national wage of the labor composite. In logs,
(4)
,
where 0 is an individual-specific error term. Thus, individual wages are proportional to the
i,j
individual’s level of human capital in the form of health.
7 Notice that implicit in this formulation is the notion that a worker with more education
or health supplies more units of the same basic labor input as workers who are less educated or
healthy. In the case of education, this assumption is hard to justify, since one worker with a
Ph.D. is hardly a perfect substitute for four workers who have no education. In the case of
health, the assumption may be marginally more satisfactory: one healthy worker who can work
faster or longer may indeed be a substitute for several unhealthy workers.
9

---