Video Lecture 1 – Production of Health
Income is unequally distributed across the world. To illustrate, the richest 20% of the people owned 74%
of the income whereas the poorest 20% of the people only owned 2% of the income in 2000. OECD
countries got richer over time and major parts of Asia moved forward. Eastern European countries, on the
other hand, went backwards in the 1990s. There are substantial regional differences with regard to the
income distribution.
In 1970, 56% of the people with an income below the poverty line lived in East Asia. The rest lived in South
Asia (30%), Africa (11%), or Latin America (3%). In 1970, many East Asian people grew out of poverty; the
share of the people with an income below the poverty line living in East Asia dropped to 32%. For Africa,
this percentage grew to 35%. Back then, it was predicted that in 2015, 68% of the people with an income
below the poverty line would live in Africa.
When considering the relation between GDP per capita and child survival rate, then we observe a clear
positive relationship. When looking at the development of this relationship between 1970 and 2000, we
see that most regions moved upward on the upward-sloping line. One exception is Sub-Saharan Africa; it
got poorer but slightly increased its child survival rate. Within regions (among countries), there is a wide
disparity with regard to the relationship between GDP per capita and child survival rate as well. Despite
the clear positive relationship, there are also outliers that do way better or worse than predicted by the
upward-sloping line. Child mortality of countries ranges from 0.3% to 30%, so we live in a world with a
100-fold difference in child mortality. At the same time, we live in world with a 100-fold difference in
income (ranging from $500 to $50,000 GDP per capita).
Income (measured by GDP per capita) and health (measured by child survival rate) are strongly associated
at the macro level. Is that also true within countries? You could ask yourself how many people in a certain
country do actually experience the country’s average levels. Within countries, we often see the same
pattern that we see at the macro level. Sometimes, the differences (or inequalities) within countries are
much bigger than the differences between related countries.
Health care systems aim at improving health distributions. When talking about health outcomes, we often
refer to the averages; for instance, average life expectancy or average mortality. However, it is important
to stress that those averages often have a lot of disparities within them. Health care is only one of the
inputs of health. Measurement of the marginal health product of health care and income is hampered by
at least three problems: it is difficult to measure population health properly, we want to measure the
marginal contribution of additional health care rather than its total contribution, and there is the issue of
reverse causality (meaning that an observed association does not necessarily imply causality) and
confounding effects (meaning that a third factor drives the correlation between two factors).
The health production function shows how health (output) is affected by things like health care, lifestyle,
schooling, environment, human biology, and genes (inputs). We need to control for all other factors to
identify the marginal contribution of medical care. There are diminishing returns to scale, such that the
marginal product of health investments decreases. Some argue that further investments do not lead to
further improvements in population health, which is referred to as flat-of-the-curve medicine. When you
plot the marginal product of health care (y-axis) against health care inputs (x-axis), you will get a
downward-sloping line; the further you move to the right (that is, the more inputs you use), the closer the
line will be to zero (that is, the lower the additional gains will be).
,With regard to health and development, a distinction can be made between the demographic transition
and the epidemiological transition.
• The demographic transition is mostly a consequence of a mortality effect and a fertility effect. If
mortality drops at all ages, then this leads to longer lives. If fertility drops, then this leads to fewer
births and a resulting age-structure shift (because there will be relatively few young people and
relatively more old people). Over the past centuries, mortality mostly declined due to
technological progress whereas fertility mostly declined due to behavioral change. In high-income
countries, the simultaneous effect of both led to a 2% increase in population growth. In
developing countries, the adoption of new technologies went much faster than the behavioral
shift, such that there was an average population growth of 4%. As a result, there are generally
much higher population growth rates in less developed countries, while there are much more
ageing consequences (due to relatively low fertility rates) in more developed countries.
o The mortality decline was not uniform by age. Especially the survival rates among
newborn improved. The life expectancy at all different ages grew, but this growth was
most spectacular for newborns.
o What were the main determinants of the mortality decline?
▪ Improved nutrition. Half of all mortality decline in the last eighteenth century was
due to increased caloric intake and growth in height.
▪ Public health. Water purification explains half of the mortality reduction in the
start of the 20th century in the US. Water and food borne diseases were very
important at that time and simple hygienic measures, like water purification, had
dramatic effects.
▪ Vaccinations. Many infectious diseases were cut back by medical developments,
such as vaccinations. Some critiques, however, argue that a great part of the
decline in infectious disease mortality can actually be attributed to the better
hygiene and cutting back water and food borne diseases, which occurred before
the introduction of effective vaccines.
▪ Medical treatment. Two-
thirds of the cardiovascular
mortality decline is due to
medical advances. In the
graph on the right, you see
that the drop in infectious
disease mortality already
occurred in the early-20th
century, whereas the drop in
cardiovascular disease
mortality occurred only in the second half of the 20th century.
o In short, the causes of decreasing mortality rates over time can roughly be summarized
as follows:
▪ Phase 1 (1750-1850): Nutritional improvements, together with economic growth
(people need food to be able to work, but they also need income to afford food).
▪ Phase 2 (1850-1930): Public health improvements, mainly through cleaner water
in cities and better sanitation.
, ▪ Phase 3 (1930-2000): Medical care improvements. It started with vaccines and
antibiotics. Now, we have other big medical technology advancements.
▪ Countries’ life expectancies converge and diverge in different periods, depending
on the phases of development they are in.
• The epidemiological transition implies that the causes of death and diseases change as well.
Specifically, expensive diseases are driving out cheap diseases. There is a relatively large increase
in old-age diseases.
In poor countries, there is very strong association between income and health. It is possible that both the
economic development and the health development go hand in hand for other (unobserved) reasons.
Does all income growth result in better health? In other words, is wealthier always healthier? Economists
have tried to find out under what conditions a higher income results in better health.
• Some researchers observed the strong positive
cross-sectional association between income and
health and they wanted to quantify that to make it
useful for policymaking. According to this research,
a 10% increase in GNP was, on average, associated
with one additional year of life expectancy, an 8.3%
lower infant mortality rate, a 14.2% lower child
mortality rate and a 1.5% lower crude death rate.
However, these relationships between income and
health outcomes are concave (if you use a linear
scale instead of a logarithmic scale) and differ
across years (countries do not necessarily move along the curve, but the curve itself shifted
upward over time). This is illustrated in the graph. Why is the curve shifting upwards? Because
technological improvements make health production cheaper; at each level of income, for later
years, higher levels of life expectancy can be obtained. Curves that relate income to health
outcomes (such as life expectancy) are often referred to as Preston curves.
• Economists tried to compute the relative marginal productivity of income, assuming that a higher
income leads to higher health investments.
o The Kakwani model measured health outcomes (h) as a function of the national income
(y): ℎ𝑖 = 𝛼 + 𝛽 log(𝑦𝑖 ) + 𝜀𝑖 . If both health outcomes and income are in logarithmic
terms, β measures the elasticity; the percentage change in health outcome as a result of
a percentage change in income. Kakwani found strong positive income elasticities of
population health, but also that these fall with rising income (which relates to the
concavity of the curve).
o Anand and Ravallion (1933) extended the Kakwani’s analysis by also including other
variables (Z): ℎ𝑖 = 𝛼 + 𝛽 log(𝑦𝑖 ) + 𝛾𝑍𝑖 + 𝜀𝑖 . When controlling for these other factors, in
particular the poverty rate and public health expenditure per capita, the income level
becomes insignificant (such that we cannot reject the null hypothesis that β equals zero).
This suggest that more income does not by definition leads to better health outcomes.
o Pritchett and Summers (1933) used a panel dataset of 60 developed countries over the
period 1960-1985. They examined the relationship between the logarithm of infant and
child mortality and the logarithm of income while controlling for certain factors. The
, difference with the previous models is that this model also allows for country-specific
fixed effects (αi) and time-specific effects (δt): log(ℎ𝑖𝑡 ) = 𝛼𝑖 + 𝛽 log(𝑦𝑖𝑡 ) + 𝛾𝑋𝑖𝑡 + 𝛿𝑡 +
𝜀𝑖𝑡 . The country-specific fixed effects pick up unobserved reasons (like climate or culture)
why some countries have higher or lower health outcomes that are unrelated to the
factors included in the model. The time-specific effects measures, for each year, a trend
that is common for all countries. In order to establish a causal relationship, the
researchers used instrument variables (which, in this case, are variables that influence
income but do not directly affect health) for income growth; for instance, they used the
terms of trade (ratio of exports over imports). They estimated a long-term causal income
elasticity of -0.2 for infant mortality and -0.4 for child mortality. This implies that a 10%
increase in income reduces the infant mortality by 2% and the child mortality by 4%.
Based on this causal interpretation, the researchers were able to conclude that half
million deaths in 1990 were due to poor economic performance in the 1980s.
According to Cutler, Deaton and Lleras-Muney (2006), the association between income and health is
strong, both between and within countries. However, income growth is neither a necessary nor a
sufficient condition for health improvement; for instance, countries might grow in income level but not in
health, and vice versa. In other words, wealthier does not always imply heathier. Instead, knowledge,
science and technology are key for improvement. The authors conclude that accelerated production of
new knowledge may even increase the health gap by income, because of its lagged adoption in low-
income countries.
As said earlier, flat-of-the-curve medicine implies that the relationship between income and health
diminishes at higher levels of income (as represented by a flatter curve). Are there still gains in mortality
at higher levels of income? Maybe we should focus more on other (quality) gains associated with higher
levels of income, such as morbidity rates, rather than focusing only on mortality. Wilkinson put forward
the income inequality hypothesis, which implies that at high levels of income, it is the (re)distribution of
income rather than the level of income within a country that becomes more important. Finally, one might
argue that we should not focus on all types of mortality, but only on the subset of mortality that is
amenable to health care. With regard to this last point, when focusing on the subset of causes of death
that are potentially avoidable, Heijink et al. (2011) still find a clear negative relationship with per capita
health spending in many OECD countries. Avoidable mortality comprises deaths from certain conditions
that should not occur in the presence of timely and effective healthcare. A recent study by Mackenbach
et al. (2017) also focused on mortality from conditions amenable to medical care and found a negative
association between the avoidable mortality rates and both GDP and health care expenditure as
percentage of GDP. No such negative relationship was found for non-amenable mortality. So, there is no
flat-of-the-curve for avoidable mortality. The authors also find that the mortality rate is higher for men
than for women and that it is higher for lower-educated people than for higher-education people. The
inequalities between sex groups have declined over time, but this is not the case for the inequalities
between educational levels. The mortality rates did not only decrease over time, but also as health care
expenditure as share of GDP increases (which can be explained by the fact that this share increased over
time). Always remember that the predicted (linear) lines are comprised of many observations (such as
countries) that might differ greatly from each other.
As mentioned earlier, Wilkinson put forward the income inequality hypothesis. This hypothesis concerns
the question whether, at high levels of income, it is the inequality of incomes within countries that matter