1. Acemoglu. Disease and Development: The Effect of Life
Expectancy on Economic Growth
Authors: Daron Acemoglu and Simon Johnson
Abstract: We exploit the major international health improvements from the 1940s to estimate the
effect of life expectancy on economic performance. We construct predicted mortality using
preintervention mortality rates from various diseases and dates of global interventions. Predicted
mortality has a large impact on changes in life expectancy starting in 1940 but no effect before
1940. Using predicted mortality as an instrument, we find that a 1 percent increase in life
expectancy leads to a 1.7–2 percent increase in population. Life expectancy has a much smaller
effect on total GDP, however. Consequently, there is no evidence that the large increase in life
expectancy raised income per capita.
The innovation in our approach is to exploit the international epidemiological transition, which led
to potentially exogenous differential changes in mortality from a number of major diseases across
the world. Poor economic development and health status is correlated when regressed cross-
country. However, this is also due to lots of surrounding factors, and causality is hard to
distinguish. Poor health status leads to reduced productivity, and countries with lower levels of
welfare lead to worse health systems and bad health states.
There are diminishing returns to effective units of labor, as land and capital are supplied
inelastically. Aggregate productivity benefits from improved health might be exaggerated in micro-
estimates.
From 1940s, there is an international epidemiological
transition which improved LE worldwide. We can see that LE
is converging in countries with varying economic status
(see figure 1), but we don’t see this in GDP per capita.
The paper does not argue that welfare is not increased,
this is indeed the case. However, output per capita is
not increased.
Background and data
Improvements in public health:
- Development countries were/are plagued by
major diseases with few effective drugs.
Penicillin (against infections), a vaccine against yellow fever and DDT (against malaria).
- Establishing World Health Organization as a driving force behind public health
- Change in international values
Data on cross-country disease mortality rates before 1940s was collected. Also, the global
intervention dates from each disease are important data.
,OLS regression is performed to
get correlation and intuition for
the effects. This is however not
the causal effect, but because it
is not via IV it overestimates the
effects as there is no exogenous
shock (1.17 in table 3). ‘While
panel A shows a positive
relationship between life
expectancy and total income,
panels B and C show that this
increase in total GDP is
insufficient to compensate for
the increase in total population
and working age population. As
a result, there is a negative
(sometimes significant)
relationship between GDP per
capita and GDP per working age population and life expectancy. There is no evidence of a positive
effect of life expectancy on GDP per capita in table 3.’
Predicted mortality and First stages
Due to two factors of endogeneity,
reverse causality and omitted
variable problems, OLS estimates
are unlikely to uncover causal
effects. We want exogenous
variation in LE to build an
instrumental variable.
Pre-international epidemiological
transition, there was variation in
disease prevalence across the world.
Based on this baseline, we expect
different effects of the interventions
on LE in different countries. Map baseline
cross-country disease prevalence against global
intervention dates for specific diseases.
Predicted mortality uses a country’s initial
mortality rate until there is a global
intervention/ post-intervention, the mortality rate declines to frontier mortality rate.
,Exclusion criterium is that predicted mortality rates do not impact future changes in population or
income, this is checked via robustness of results.
Alternative instruments are constructed to show robustness and gave similar results. The zeroth
stage estimates investigate the relationship of interventions and disease mortality, they seem to
be strongly correlated.
First-stage estimates the relationship
between change in predicted mortality and
the change in (log) LE. In figure 4, the
richest countries are excluded and still
there is a strong, negative relationship
between these variables.
Main results
The main results are a 2SLS estimate (second stage) of log life expectancy on six outcomes, most
importantly the log GDP per capita and per working age population.
Results population: ‘we conclude that there is a large, relatively precise, and robust effect of life
expectancy on population (increase in life expectancy leads to increased population). The elasticity
of population in response to life expectancy at birth is estimated consistently to lie be- tween 1.65
and 2.15, which is similar to the OLS estimates.’
Results births and age composition: large effects of life expectancy on total births (2.15-2.9). Effect
of life expectancy on working age population very similar to effect on total population.
Results on GDP: Reduced form (change pred. mortality on log GDP) shows slight decline, which
indicates that countries with larger declines in predicted mortality experienced somewhat higher
GDP growth between 1940 and 1980.
Short run GDP per capita decreases, total GDP not influenced
No significant effects seen of change in life expectancy on total GDP. This already predicts the
results of GDP per capita, as the population grew. The relation between change in life expectancy
and GDP per capita is negative.
A concern was that the total population grew (child mortality prevented) but working age
population did not grow. By studying the GDP per working age capita, we can conclude that also
here the relation is negative.
Ashraf et al.(2007) shows that even when health has positive effects on long-run income per
capita,population dynamics will lead to considerable delays before any increase in income
percapita is observed
, ‘Overall, the 2SLS
estimates show no
evidence that the large
increase in life
expectancy in many
parts of the world
starting in the 1940s led
to a significant increase
in GDP per capita.
Instead, the increase in
life expectancy was
associated with a
significant increase in
population and a
considerably smaller
increase in total GDP.’
‘Our results indicate that the increase in life expectancy led to a significant increase in population;
birth rates did not decline sufficiently to compensate for the increase in life expectancy. We find a
small positive effect of life expectancy on total GDP over the first 40 years, and this effect grows
somewhat over the next 20 years, but not enough to compensate for the increase in population.
Overall, the increases in life expectancy (and the associated increases in population) appear to
have reduced income per capita. There is no evidence that the increase in life expectancy led to
faster growth of income per capita or output per worker. This evidence casts doubt on the view
that health has a first- order impact on economic growth.’
