Tải bản đầy đủ - 0trang
The Promise of Health: Evidence of the Impact of Health on Income and Well-Being
The next section outlines a number of mechanisms by which health innovations can affect incomes.
Section 5.3 presents empirical evidence that speaks to this issue, much of it produced over the last
decade. Section 5.4 concludes.
5.2 M ECHANISMS
Better health comes in many forms and at different stages of life. The life-cycle patterns of human
capital and other investments, of labor market participation and income generation, and of fertility
decisions, mean that the effects of health improvements are likely to vary depending on when, and by
whom, they are experienced. This section outlines the various routes by which improved health can
plausibly lead to increases in measured incomes, taking account of the expected wide range of
heterogeneity in effects.
We classify health improvements loosely into three categories, corresponding to improved
nutrition, reduced morbidity, and reduced mortality. These categories clearly are not mutually
exclusive, as better nourished individuals fall sick less often, and death is often, but not always,
portended by illness. Nonetheless, with this classification in mind, we set out the routes by which
improved health can affect income—namely direct productivity effects, effects on incentives to
accumulate human and physical capital, and aggregation effects, which are important to assess in
extrapolating the effects of improvements at the individual level to improvements in population
5.2.1 Direct Effects
The obvious route from health to income is a direct productivity link. Higher levels of nutrition and
energy intake can likely increase the productivity of both manual work and tasks that require
concentration. Fogel (2004) traces increases in economic growth to surges in agricultural
productivity, and the impact this had on hunger and nutrition. Using historical records, he proposes
that there simply were not enough calories consumed to exhaust the productive capacity of workers in
nineteenth-century Europe, and that increases in food availability led directly to increases in the size
and stature, strength, and effort of the workforce. As reviewed by Strauss and Thomas (1998), the
historical evolution of height, attributed in large part to nutritional improvements, provides a useful
proxy for health status, and is closely linked to economic development.
Nutrition does not only increase physical strength, but it provides protection against both chronic
and infectious disease. Fogel develops the concept of physiological capital to denote the biological
capacity of an individual to learn, acquire skills, and work, and focuses on the impact of in utero,
infant, and early childhood nutrition and parental behaviors (such as smoking and drinking) that can
lead to impairment of the nervous system and neurological damage. Longer term health effects of low
birth-weight and malnutrition early in life include increased risk of heart disease, stroke,
hypertension, and diabetes.
Lower morbidity can similarly affect both the productivity of a worker’s time spent on the job, but
also the length of time devoted to work, and absenteeism. In countries in which children contribute to
household income, often by undertaking domestic and farm chores such as cleaning, fetching water,
and guarding livestock, reductions in morbidity and increases in nutrition can have direct impacts on
total, if not always market, income. And reductions in mortality have to be (weakly) good for a given
individual’s output. Total, if not per capita, output is also likely to rise in the presence of mortality
reductions, unless congestion costs are especially large.
Alternatively, health improvements of one individual can directly increase the incomes of others.
When parents and grandparents who tend to sick children are relieved of this task, their measured
incomes may rise as they (re-)enter the labor market. The same is true for children who would
otherwise be working if not for the fact that they have to tend to a sick parent. In both these cases, one
person’s health improvement increases the labor supply, and hence the income, of another. To the
extent that both individuals enter the labor market, measured incomes will increase further. However,
even if health improvements for one individual do not free up time of another, they could nonetheless
increase the productivity of both, for example due to the existence of either task complementarity or
increasing returns in household production.
Finally, improvements in one person’s health can lead to improvements in the health of others,
with direct or indirect effects on their incomes. This mechanism appears most likely to be operational
in the context of reducing the transmission of communicable and infectious diseases, in which either
behavioral change is important (such as the use of insecticide-treated bed nets to prevent malaria, or
condoms to prevent transmission of HIV and other STIs), or when public goods such as water and
sanitation facilities are used. There could well be increasing returns to improved health associated
with network externalities in such cases, which require that a certain critical mass of individuals
adopt preventive behaviors or gain access to public goods before the health (and incomes) of the
broader community can improve.
5.2.2 Effects on Capital Accumulation
Grossman (1972) was among the first to model health improvements as investments in human capital,
and indeed, the direct economic impacts of better health at early ages tend to show up later in life, if
only for life-cycle reasons. But improvements in health, as well as adding directly, by definition, to
the stock of human capital, can also strengthen incentives of individuals to increase the rate of human
capital investment, particularly in the form of education, either because the effective price falls, or
because the returns to education increase.
Healthier children may well attend school more regularly, learn more efficiently while there, and
perform better on tests, all of which increase their acquisition and maintenance of human capital for a
given level of effort and cost. The incentive to invest in education therefore increases when the
effective price falls, increasing the number of years, the quality of the education sought, and the
willingness (most often of parents) to pay for that education. Although a similar mechanism could
affect the incentive of those who are past school age to invest in training, we expect that better health
among adults is unlikely to have as large an effect on human capital investment and future income.
However, the prospect of better health in adulthood may well induce greater levels of human
capital investment, among those of school age or older. Even if the effective price of education or
training does not fall, better health in adulthood can increase both the length of time over which the
returns to education are earned because of a longer working life, and the length of time over which
these returns are consumed, as life expectancy increases. Again, the effects of morbidity declines
among adults, which are likely to affect adult labor supply, and the impact of reduced adult mortality
rates, which might affect life expectancy more, could have similar impacts on human capital
investment, but for different reasons.
We note one important difference between health improvements that reduce the price of human
capital investment or the direct return to it, and those that simply affect life expectancy. In principle,
the first kind of health improvement increases the incentive to invest in human capital, while the
second, which simply affects the return to saving, induces an increase in investment in all forms of
capital. The difference may be difficult to observe in practice however, given that a majority of
individuals hold a large share of their total asset base in the form of human capital, while credit
market imperfections, which are particularly acute in developing countries, could limit the ability of
individuals to invest in any capital other than through on-the-job training. On the other hand, Bloom et
al. (2003) develop a model in which increases in life expectancy lead to higher savings rates at every
age, and present some cross-country evidence in support of this result.
5.2.3 Aggregate Effects
Most of the mechanisms by which health improvements increase income described above involve
changes in inputs to the production process—for example, through increases in the quantity or quality
of labor supply in the short term, and changes in the stock of human and possibly other capital over
the longer term. If income generation depends on other inputs that are in fixed supply, such as land,
then the per capita effects of population-wide health improvements might be much smaller than those
experienced by a single individual who enjoys a similar health improvement. These general
equilibrium effects limit the extent to which studies at the micro-level can be extrapolated, and can
also account for at least some of the relatively weak cross-country empirical evidence on the impact
of health on income.
Similarly, undisputedly positive health improvements can nonetheless reduce per capita income if
they adversely alter the dependency ratio. This seems particularly likely for a number of interventions
that directly target mortality: in rich countries mortality rates have recently fallen most among the
elderly, while in the developing world infant (under one year) and child (under 5 years) mortality
rates have been reduced through expansions of immunization and nutrition programs, and improved
sanitation. The short run impacts of such health improvements on per capita income are almost
certainly negative, as neither group contributes significantly to measured GDP.
Some authors have even suggested that the impact on total income of reductions in mortality rates
among adults of working age may not be large enough to offset the population effect, thereby reducing
per capita income. Conversely, high-mortality epidemics can have ambiguous effects on per capita
income. The Black Death in fourteenth-century Europe, and the HIV/AIDS epidemic in southern
Africa, both led in the short run to reductions in the labor force without offsetting reductions in the
capital stock, leading to higher labor productivity and greater incomes for those lucky enough to be
spared. The expansion of antiretroviral therapy (ART) across Africa in the last five years, which has
increased survival rates among HIV-infected individuals, many in the prime of working life, could
thus have lowered per capita income.
How should we interpret these arithmetic consequences of changes in population structure and
size? The natural implication is that a focus on per capita incomes is not necessarily well-placed, and
that it is important to adopt meaningful welfare measures that reflect the obvious social costs of
mortality. One straightforward adjustment is to divide total income produced by the total number of
people, living and dead, who were once alive and who would be now if not for a given mortality
risk. Let’s call this per nata income. Higher mortality thus decreases total income (the numerator) but
not the denominator.
This measure is not entirely satisfactory, of course, as it implicitly assumes fixed fertility rates. It
can thus be argued that in calculating the income effects of reductions in child mortality rates that lead
women to have fewer children, the denominator should in any average measure indeed fall. In fact,
even health improvements that reduce morbidity among children can induce parents to move along the
quality–quantity frontier (Becker 1981), reducing the desired number of births, and the denominator in
a meaningful measure of mean welfare.
5.3 EMPIRICAL EVIDENCE
Perhaps the starkest differences in health and income are apparent in cross-country comparisons.
While the identification challenges that must be overcome in estimating the impact of improved health
on income and economic activity in such settings are manifold, they have nonetheless not deterred
researchers from addressing the issue at this level of aggregation. A sizeable empirical literature has
thus focused on using national-level data to investigate the relationships between health and income.
Alternatively, researchers have used natural or quasi-experiments, in which arguably exogenous
shocks to health status, associated with either sudden disease outbreaks or discrete policy-driven
improvements in the health environment, to identify the impact of health on income. At a more microlevel, randomized control trials have been employed to improve nutrition or health, and the impacts
on schooling, labor supply, and income measured. Of course, as identification has become more
reliable with these methods, the parameters of interest have focused on progressively narrower
5.3.1 Cross-country Studies
Cross-country correlations between population health (as measured by life expectancy) and income
per capita, as exemplified by the Preston Curve (Preston 1975), are strong and have remained so over
the last hundred years (Pritchett and Viarengo 2010). Figure 5.1 uses 2007 data from the United
Nations Development Programme’s Human Development Report database1 to illustrate this
relationship. While advances in technology and its diffusion have shifted the relationship between
income and health, the cross-country correlation between the variables has proven robust over time.
One strand of the literature thus employs cross-country regression techniques in an attempt to identify
a causal link from health to income at the national level. These studies have had to make strong
identifying assumptions that are sometimes difficult to sustain.
A number of authors have used cross-country regression analysis to isolate the impact of
improvements in population health on macroeconomic performance, in particular per capita GDP, that
could underlie the correlations observed in the Preston curve. Pritchett and Summers (1996) were
among the first to address this issue, using infant and child mortality rates as measures of population
health. As instruments for economic growth, they employed terms of trade shocks, the investment to
GDP ratio, the black market premium for foreign exchange, and the deviation of the official exchange
rate from its purchasing power parity level, all of which were argued to be correlated with income,
but would not directly affect health. While their estimates provide evidence that income growth does
drive health improvement, they do not rule out the reverse effect.2
FIGURE 5.1 Life expectancy per capital GDP (US dollars)
Gallup and Sachs (2001) showed that better health was correlated with larger subsequent changes
in income, suggesting that health could affect not only the level, but the growth rate of income.
Obvious endogeneity problems plague this strand of the literature, as omitted variables that determine
health could explain future income growth. They noted, however, that the geographic concentration of
infectious diseases in tropical regions of the world suggested a plausible instrument for health status,
namely distance from the equator, and found a strong impact of health on income growth using this
Geography as a proxy for health has however been challenged by a number of authors (e.g.
Acemoglu, Johnson, and Robinson 2001; Easterly and Levine 2003; Rodrik, Subramanian, and Trebbi
2004), who point out that geographic location may well influence economic growth either directly or
through historical patterns of institutional development and adoption. Once the impact of geography
on the choice of institutions is accounted for, there remains no independent link from geography (and
thus health) to income. The most acute problem these studies face is simply the availability of reliable
historical data.3 However, even with reasonable data, a remaining methodological problem is that
both health and institutional choice are potentially endogenous, and a single instrument cannot identify
the contribution of each to growth. In response, McArthur and Sachs (2001) note the relatively small
sample size employed in that study, and its limited geographic diversity, and present evidence on a
wider sample of countries and claim that both institutions and the disease environment—in particular
the prevalence of malaria—influence future economic growth. Sachs (2003) attempts to further
identify the specific impact of malaria by constructing an index of “malaria ecology” and a separate
instrument for institutional choice based on mortality rates of colonial settlers. He finds both affect
future income, suggesting again that both health and institutions matter for growth. But since malaria
ecology could easily be associated with the conditions encountered by early settlers, and hence the
choice of institutions, the validity of the estimates remains in question.
Bloom, Canning, and Sevilla (2004) present results from thirteen cross-country analyses that
complement or refine the Gallup and Sachs methodology, all of which report large effects of health on
income growth. The authors themselves try to resolve the problem of omitted variables by regressing
income growth on changes in health, and other lagged variables. This general approach does not find
significant support in the wider macroeconomic literature, however (Mankiw 1995; Weil 2007).
Acemoglu and Johnson’s (2007) study is probably the most innovative recent cross-country
analysis of the impact of health on income. It uses the international epidemiological transition of the
1940s, associated with the discovery and diffusion of penicillin and sulfa drugs, vaccines,
insecticides, and the creation of the World Health Organization, as an exogenous source of health
improvements. Specifically, the authors use variation in the potential health improvement associated
with full adoption of these innovations, which they argue are uncorrelated with future income growth,
as an instrument for actual health improvements. They find that population growth rates increased in
response to the innovations, as did aggregate income, but per capita income fell.
Acemoglu and Johnson’s methodology—using potential health improvements as an instrument for
actual changes in health status—has been used in a number of other studies on the impact of malaria
on income (see below). However, some commentators have questioned the mechanisms underlying
the empirical results. In particular, while capital constraints may well bind in the short run, reducing
the potential for increases in per capita income, over the forty-year period of their analysis one would
expect the capital stock, including the productivity of land, to increase accordingly, and for fertility
rates to adjust. Focusing instead on the effect of higher mortality on economic outcomes, Young
(2005) uses micro-data to calibrate a model of the AIDS epidemic in South Africa. He incorporates
fertility effects and the impact of adult mortality on the intergenerational transmission of human
capital, but still finds that the increase in capital–labor ratios associated with the disease leads to
higher per capita income, consistent with Acemoglu and Johnson’s macro-analysis.
Weil (2007) uses microeconomic estimates of the effect of health on individual incomes (some of
which are reviewed below) to calibrate a model of production which is then used to estimate the
share of cross-country variation in per capita GDP that is attributable to variations in health
indicators. That is, per capita output differs across countries because of differences in physical
capital assets, worker skills (or educational capital), the capacity of workers to work hard and long,
and to think clearly (health capital), and other factors. Using this methodology, he finds that between
about 10 and 30 percent of the variation of log GPD per capita across countries can be attributed to
differences in health human capital. His preferred estimate (23%) is “roughly the same as the share
accounted for by human capital from education, and larger than the share accounted for by physical
The strength of Weil’s approach is that it relies on well-identified microeconomic estimates of the
causal impact of health on income. One limitation, recognized by the author, is that it is difficult to use
the framework to estimate the “full” contribution of differences in health to cross-country variation in
incomes. Allocating national output as the return to various factors becomes problematic when one
factor (say health) determines investment in others (say education). While many microeconomic
studies of the effect of health on income examine this and other similar routes, Weil’s analysis
focuses on the proximate effect of health on income—that which arises due to the physical and mental
capacity of workers. That is, his analysis can be thought of as providing an estimate of the impact of
removing health differences across countries, while holding all other factors, including physical and
educational capital, fixed. Allowing those factors to endogenously respond to improvements in health
would likely lead to somewhat larger income responses.
5.3.2 Micro-level Evidence
The labyrinth of channels from health to income and back, the time lags between innovations in one
and effects on the other, and the dependence of these effects on a range of other complementary
institutional and environmental factors, all suggest the debates surrounding macro-level empirical
analyses using aggregate data will continue for some time. An alternative approach has been to
investigate the individual mechanisms by which health affects income using quasi-experimental data,
and randomized control trials.
Nutrition is perhaps most important for human growth and development, and so has been heavily
studied among children. However, nutritional intake in adulthood can directly affect economic output,
as reviewed by Thomas and Frankenberg (2002). They summarize their findings as indicating that
“[w]hile the establishment of this link is not straightforward, the weight of evidence points to
nutrition, and possible other dimensions of health, as significant determinants of economic
For example, Thomas et al. (2004) report the results of a large-scale iron supplementation
intervention in Indonesia that covered over 17,000 adults. They find that iron supplements had
significant effects on men who were otherwise anemic, and who experienced increased energy levels,
better work attendance, and up to 20 percent higher productivity.4 There was however no discernible
effect for women.
On the other hand, the physiological benefits of better nutrition do not always show up in greater
measured output. For example, Thomas and Frankenberg report the results of randomized experiments
in which iron supplementation improved energy use on the job, but did not lead to higher measured
incomes. For example, in a study of female Chinese cotton workers, Li et al. (1994) found iron
supplements were effective in increasing energy levels and that they reduced energy expenditure per
work task completed. However, the subjects did not increase output in their primary jobs, but spent
more time on non-work activities.5 This could have been due to constraints in the work-place that
made it difficult to increase output (reliance on other workers, technology constraints), or simply to
the fact that optimizing individuals chose to spend the additional energy on non-work activities.
In contrast to iron supplementation, the impact of increases in caloric intake is less clear, with
some randomized experiments showing no effect, and others a small impact. Observational studies
(i.e. those with non-random variation in caloric intake) have shown positive correlations between
calories and productivity, but suffer from problems of unobserved heterogeneity.6 For example,
Foster and Rosenzweig (1994) report that caloric intake is correlated with hourly productivity, but
the causality is likely in the opposite direction, as the variation they exploit rests in the strength of
effort incentives—higher powered incentives lead to more effort, higher income, and higher caloric
consumption. Croppenstedt and Muller (2000), however, estimate a positive impact of health and
nutritional status on the productivity of peasant farmers in Ethiopia.
Childhood nutrition can affect future incomes either through a direct effect on future productivity or
by inducing improved attendance and performance at school. Some studies can estimate only the net
impact of these two channels, while others have been able to isolate the effect of the link through
education from the more direct effect.
Hoddinott et al. (2008) report results from a long-term study of the impact of child nutrition in
Guatemala, in which children in two villages were offered a highly nutritious dietary supplement,
while those in two comparison villages were offered a supplement of minimal nutritional value. Sixty
percent of village residents, by then aged between 25 and 42, were tracked more than twenty years
later. Wages rates of those from the treatment villages were found to be 46 percent higher than those
of control villages, but they nonetheless did not have significantly higher earned incomes.
Case and Paxson (2008) find that the labor market pays a wage premium to individuals of greater
height, itself a robust indicator of childhood nutrition. They go further and argue that the premium is
not a reward for height itself, either due to discrimination or the productivity that height might
portend, but that height serves as a marker for cognitive ability, which is rewarded by the market.
Similarly, Behrman and Rosenzweig (2004) use data on twins and find that fetal growth is associated
with future height and years of completed schooling in adulthood. These studies suggest that in utero
and early childhood health can have important effects on the efficiency with which children learn
while at school, and hence their future economic performance.
In a similar vein, Case, Fertig, and Paxson (2004) control for parental income, education, and
social class and find that children who experience poor health attain lower education, worse health,
and lower social status in adulthood. Poor health in childhood thus acts as a mechanism of intergenerational poverty transmission.
In the developing world, childhood malnutrition and morbidity not only can reduce returns on the
intensive margin by compromising learning efficiency at school, but they can have large effects on the
extensive margin by reducing school enrollment and attendance. In a field experiment in western
Kenya, Miguel and Kremer (2004) assigned simple and cheap de-worming treatment to randomly
selected students across randomly selected schools. They found that absenteeism in treated schools
was reduced by one-quarter, and that by reducing transmission of the infection the intervention also
improved the health and school participation of untreated students in both treated and nearby
untreated schools. However, despite the impact on school participation, the intervention did not have
a significant effect on test scores, so the contribution to human capital and future economic activity
In contrast, Field et al. (2009) use temporal and geographic variation in the implementation of
intensive iodine supplementation in Tanzania to study the impact on schooling of reductions in iodine
deficiency syndrome. They find that children treated in utero attain between one-third and one-half a
year more education relative to their (untreated) siblings and peers, and that this effect is particularly
great for girls. They also infer from the data that this increase in years of schooling attained is driven
in part by better performance while in school, and the associated higher likelihood of passing tests
which allow grade progression, and that the program did not affect the age at which children begin
formal education. Thus iodine supplementation can be interpreted as reducing the price of completed
school grades, and hence the demand for education.
Like iodine deficiency syndrome, anemia can impact childhood educational attendance and
attainment. Bobonis et al. (2006) report the results of a field experiment in which iron supplements
were randomized across 2–6-year-old children in Delhi slums, where the pre-intervention rate of
anemia was 69 percent. This study focused on very young children, for whom test score data were of
little relevance. They find that child weight increased, and that pre-school participation rates rose, at
least in the short run. On the other hand, the longer run impacts of the intervention were harder to
interpret, due to sample attrition and non-random sorting of new cohorts.
Following a similar strategy to that employed by Field et al., Bleakley (2007a and 2007b) uses
two specific historical health interventions—hookworm eradication in the southern United States and
malaria eradication campaigns in selected countries in the Americas (US, Brazil, Colombia, and
Mexico)—to estimate the effect of health on income. In his study of the impact of hookworm
eradication efforts under the Rockefeller Sanitary Commission in the American South in the early
twentieth century, Bleakley measures pre-existing infection rates in 1913 by location (on average, 40
percent of school-aged children were infected prior to the intervention). In a similar fashion to the
approach adopted by Acemoglu and Johnson (2007), the author uses variation in infection rates by
location, which reflect the potential benefits from eradication, to identify the impact of changes in the
health environment on economic outcomes. He finds that areas with higher pre-existing infection rates
saw greater increases in school enrollment, attendance, and literacy after the intervention. For
example, he finds that before 1910 the impact of infection rates in 1913 on school attendance is
negative, but that by 1920 there is no impact of 1913 infection rates on attendance. That is, those
areas that had the most to gain from the eradication saw enrollment rates increase more. Similar
results are found for literacy.
There may have been other changes in the economic environment that could have led to similar
trends over this period, but it is argued that these should have affected adults in different areas in a
similar way. However, no similar pattern is found among adults across the affected areas, who, by the
nature of the disease, had virtually no pre-existing infection.
A similar exercise is performed using the malaria eradication campaigns in the United States
c.1920 and in Brazil, Colombia, and Mexico c.1955 (Bleakley 2006). Pre-existing prevalence rates
provide the exogenous variation permitting him to identify the impact of childhood exposure to
malaria on future adult literacy and incomes. He finds that for individuals born well before the
relevant eradication campaign, those born in more malarious regions had lower wages and literacy
later in life, but for those born well after the campaigns, pre-eradication malaria prevalence had little
effect on future wages and literacy. He concludes that “persistent childhood malaria infection reduces
adult income by 40 to 60 percent.”
Interestingly, Bleakley is able to differentiate between the impact of morbidity and mortality on
future income. He finds that eradication of vivax malaria (which causes high morbidity, but relatively
few deaths) leads to significant increases in human capital formation and future income, but that
eradication of falciparum malaria (which is relatively fatal) produces no such gains. His preferred
rationalization of this result is that reductions in mortality rates increase the marginal benefit of human
capital acquisition (as there are more years in which to earn a return on human capital investments),
but this might have little impact on the level of investment if marginal costs are increasing steeply.7
On the other hand, a reduction in morbidity makes it easier to attend school and to learn while there,
thereby flattening the marginal cost curve, and leading to significant increases in human capital
Bleakley uses his estimates to extrapolate across countries, and estimates that malaria can account
for about 10–16 percent of the income gap between the US and Latin America. For the Americas at
least then, this evidence suggests that eradicating malaria would modestly narrow the cross-country
income gap by inducing higher growth in Latin America. He concludes that “…while reducing
malaria could bring substantial income gains to some countries, the estimated effect is approximately
an order of magnitude too small to be useful in explaining the global income distribution” (2006: 26).
According to this research, improving health could be important for growth, but is unlikely to be a
Using a similar approach, Cutler et al. (2007) examine the impact of a malaria eradication
program across Indian states during the 1950s and find that the program increased literacy and
primary school completion by ten percentage points, accounting for about half the observed gains in
these measures over the period spanning the intervention in malarious regions. Hong (2007) and
Lucas (2009) both find significant affects of either exposure to malaria or its eradication on a variety
of economic outcomes such as schooling, literacy, labor force participation, and/or wealth.
Almond (2006) uses a negative historical health innovation as opposed to a policy-driven
improvement to again estimate the link from in utero health to economic performance later in life. In
particular, he studies the effects of the 1918 influenza epidemic, which was harsh and short, and
whose incidence varied geographically. In comparing cohorts exposed in utero to the flu with those
who were born just before or conceived just after, he found the former had “reduced educational
attainment, increased rates of physical disability, lower income, lower socioeconomic status, as well
as accelerated adult mortality compared with other cohorts.” Similarly, those born in states where the
pandemic was less severe fared better than those in other states.
The role of health on the incentive to invest in education has attracted both theoretical and
empirical attention of researchers. For example, Kalemli-Ozcan et al. (2000) and Soares (2005)
incorporate both education and fertility choices in models of demographic transition and the impact of
health on growth. The essential feature of these models is that longer life expectancy increases the
returns to investment in general, including education. Those cross-country analyses above that assess
the impact of life expectancy on income are unable to identify a large effect partly because much of
the improvement in life expectancy over the twentieth century was associated with reductions in
infant and child mortality, which occurs before most educational decisions are made.
Jayachandran and Lleras-Muney (2009) present evidence that reductions in adult mortality, which
increase the life expectancy of school-aged children, can have important effects on educational
attainment. They exploit a sharp fall in maternal mortality, from 1.8 to 0.5 maternal deaths per
hundred live births over a short seven-year period in Sri Lanka that translated into a 4.1 percent
increase in life expectancy of 15-year-olds, and find that female literacy increased 2.5 percent and
years of schooling rose by 4.0 percent. Their empirical strategy is strengthened by a number of unique
features of the environment: first, as well as occurring after education decisions have been made,
maternal mortality affects relatively young adults, and reducing it saves potentially many future years
(and hence induces a stronger human capital acquisition response); second, the authors are able to
compare outcomes for women across districts (in which the reductions in maternal mortality rates
varied); and third, the outcomes for both women and men (whose mortality rates and life expectancy
changed less over the same period) can be usefully compared.
At the cross-country level, Fortson (2007) uses data from Demographic and Health Surveys in
southern Africa to assess the effect of adult mortality on the incentive to invest in human capital. She
finds that living in an area with higher HIV prevalence, and hence a higher perceived risk of
premature death, is associated with lower educational attainment and slower grade progression. This
effect is not observed just for orphans, but for non-orphans as well. Reduced life expectancy appears
to dampen the incentive to invest in human capital.
Most often it is adults who make education decisions on behalf of their children, so reductions in
adult mortality can increase the incentive to invest in human capital simply because those who would
finance it are alive to do so. For example, Case, Paxson, and Ableidinger (2004) find that
orphanhood reduces the school enrollment of children compared with other children with whom they
live, suggesting that biological ties seem to matter in the allocation of investment.
This chapter has outlined a number of mechanisms by which improvements in health could lead to
increases in income, and has documented a growing body of empirical evidence that assesses the
strength of these links. From a policy perspective, it is not necessarily useful to dwell on the question
of whether health determines income or vice versa as both directions of causality are likely to be
operative. The more useful question is, “Are the income returns to some health improvements likely
to be enough to tip the balance in favor of interventions that would otherwise not pass a cost–benefit
test?” It remains very difficult to address this question in general, the answer to which depends on the
nature of the health improvement, the intervention itself, and the economic and institutional
Basic public economics suggest that policymakers first investigate the prevalence of market
failures that inhibit the adoption of the kinds of interventions that might otherwise have large health,
and income, effects. Obvious candidates are public goods (e.g. vector control), and perhaps goods
that exhibit increasing returns to scale in production, such as in the provision of some private goods
including e.g. iodized salt, which might be best provided, financed, or simply mandated by
Alternatively, the correlation between health and income suggests that health policy should be an
integral part of social policy in general and of redistributional policies in particular. First,
independent of any causation between the two, either (poor) health status can be used as a marker of
(low) income, so income transfers in the form of tax credits and subsidies might be directed to the
sick and infirm; or publicly provided health-improving goods and services, of little value to those
who are not sick, could serve as an efficient self-targeted means of redistributing to the poor.
However, the scope for errors of both inclusion and exclusion is wide, and some health conditions
concentrated particularly among the poor, and some services valued particularly by them, might have
better targeting properties than others.
Second, absent issues of targeting efficiency, the causal link from health to income, and the fact
that income does not fully explain health outcomes, suggests that the provision of health services to
the poor could be an efficient means of improving their well-being. There is a taste of a merit–good
argument to this reasoning: the fact that education, itself correlated with income, is an important
independent determinant of health, suggests that demand for certain health services among the poor
might not be as high as it could be. Whether this means the poor should be provided with health
services or education is not clear—perhaps both?
In the poorest regions of the world, the potential gains to health improvement can only be realized
if complementary institutions such as schools with effective teachers and well-functioning labor and
credit markets exist. The promise of health, however, is matched by the challenges of delivering
health services, an issue that we have not addressed here but which must be incorporated into any
policy choice involving public provision or financing of health interventions. In these environments
especially, the mobilization of resources to finance improved health, and the income gains it
promises, must go hand in hand with the design of incentives for effective service delivery, embedded
in responsive and well-governed institutions.
ACEMOGLU, DARON and JOHNSON, SIMON (2007), “Disease and Development: The Effect of Life
Expectancy on Economic Growth,” Journal of Political Economy, 115(6): 925–85.
—— and ROBINSON, JAMES A. (2001), “The Colonial Origins of Comparative Development: An
Empirical Investigation,” American Economic Review, 91(5): 1369–401.