Download Life expectancy and national income in Europe, 1900

Published by Oxford University Press on behalf of the International Epidemiological Association
ß The Author 2013; all rights reserved. Advance Access publication 6 August 2013
International Journal of Epidemiology 2013;42:1100–1110
doi:10.1093/ije/dyt122
LIFE EXPECTANCY
Life expectancy and national income in Europe,
1900-2008: an update of Preston’s analysis
Johan P Mackenbach* and Caspar WN Looman
Department of Public Health, Erasmus MC, Rotterdam, The Netherlands
*Corresponding author. P.O. Box 2040, 3000 CA Rotterdam, The Netherlands. E-mail: [email protected]
Accepted
23 May 2013
Background In the past, upward shifts of the so-called Preston curve, which
relates life expectancy to national income, have contributed importantly to worldwide increases in life expectancy. These shifts were
due to rapid diffusion of knowledge and technology for infectious
disease control from high-income to low-income countries. We assessed to what extent life expectancy growth in Europe has been
accompanied by upward shifts in the relation between national
income and life expectancy in later parts of the 20th century,
when progress in cardiovascular disease control was the main
driver of life expectancy growth.
Methods
Data on national income (gross domestic product per capita, in
1990 international dollars), life expectancy and cause-specific mortality covering the period 1900–2008 were extracted from international data banks. (Change in) life expectancy and age-standardized
mortality was regressed on (change in) national income, and the
regression parameters were used to estimate the contribution to
rising life expectancy and declining mortality in Europe as a
whole of changes in national income vs shifts in the relation between national income and health outcomes.
Results
Large upward shifts in the relation between national income and
life expectancy only occurred before 1960, and were due to rapid
declines in mortality from infectious diseases which were independent of rises in national income. These shifts account for between
two-thirds and four-fifths of the increase in life expectancy in
Europe as a whole during this period. After 1960, upward shifts
in the relation between national income and life expectancy were
much smaller, and contributed only between one-quarter and onehalf to the increase in life expectancy in Europe as a whole. During
the latter period, declines in mortality from cardiovascular disease
were mainly attributable to increases in national income.
Conclusions In contrast to earlier periods, recent life expectancy growth in
European countries appears to have been dependent on their economic growth. More rapid diffusion of knowledge and technology
for cardiovascular disease control from higher- to lower-income
countries in Europe may be needed to close the East–West life expectancy gap, but it is unlikely that this can be achieved in the
absence of more equal economic conditions.
1100
LIFE EXPECTANCY AND NATIONAL INCOME IN EUROPE
Keywords
1101
Life expectancy, national income, infectious disease, cardiovascular
disease, Europe
Introduction
Economic conditions have a central place in theories
of the determinants of population health.1–3 One of
the most well-known illustrations of the relation between economic conditions and population health is
the so-called Preston curve. This curve, named after
the American demographer who first described it in a
paper published in 1975,4 relates national income to
average life expectancy at birth for a range of countries at one point in time. It shows that people living
in richer countries on average live longer than people
in poorer countries, and that although the relation is
much steeper at lower levels than at higher levels of
income, it still holds at higher levels as well.
When Preston compared these curves for different
points in time (i.e. the 1900s, 1930s and 1960s), he
found that the curve had shifted upwards from the
1900s to the 1930s, and again from the 1930s to the
1960s, so that at any given level of national income
(adjusted for inflation), life expectancy was higher in
the 1930s than in the 1900s, and in the 1960s than in
the 1930s. This observation, together with the results
of a quantitative analysis of the relative contribution
of changes in national income vs that of shifts in the
relation between national income and life expectancy,
was interpreted to indicate that ‘factors exogenous to
a country’s current level of income probably account
for 75–90% of the growth in life expectancy for the
world as a whole between the 1930s and 1960s.
Income growth per se accounts for only 10–25%’.4
Speculatively, these ‘exogenous’ factors raising life
expectancy were identified as ‘public health programs
of insect control, environmental sanitation, health
education, and maternal and child health services
[. . .] in less developed areas, [and] specific vaccines,
antibiotics, and sulphonamides in more developed
areas. [These] technologies were not, for the most
part, indigenously developed by countries in either
group. Universal values assured that health breakthroughs in any country would spread rapidly to all
others where the means for implementation existed’.5
The importance of Preston’s paper is illustrated by
the fact that it has been reprinted as a ‘classic’ in the
Bulletin of the World Health Organization (2003) and in
the International Journal of Epidemiology (2007), on both
occasions with commentaries stressing its lasting relevance.6–12 There has, however, been little attention to
whether Preston’s results can be generalized across
the 20th century, and whether similar mechanisms
explain more recent advances in life expectancy. The
main part of his analysis deals with the 1930–60
period, characterized by rapid advances in infectious
disease control.5,13,14 More recent advances in life
expectancy, at least in high-income countries, were
largely due to declines in mortality from cardiovascular disease.15,16
Although European life expectancy data show salutary trends in recent decades,17 it is not obvious that
the factors contributing to life expectancy growth will
have diffused internationally in patterns resembling
those of earlier decades. Updates of Preston curves
for the 1990s18 and the year 200319 suggest that
some upward shifts have continued to occur, but
formal comparisons of the relationship across the
whole of the 20th century and into the 21st century
have not been carried out.
Whether increases in life expectancy can still be regarded as mainly due to upward shifts in the relation
with national income, as in Preston’s analysis, is
important for several reasons. It would suggest that
progress in population health has been uncoupled
from countries’ own economic growth, for example
because ‘universal values’ assured that all countries
were able to share in this progress. It would also confirm the widespread idea that the ‘marginal returns’
of higher national incomes in terms of higher life
expectancy have become small.20
In this paper we aim to assess to what extent life expectancy growth in Europe in later parts of the 20th
century and the first years of the 21st century has
been accompanied by upward shifts in the relation between national income and life expectancy.We focus
on Europe (instead of on the world as a whole) to
create a certain degree of homogeneity in epidemiological regime and also to exclude foreign aid as an explanation of rapid mortality declines in poor countries.
Data and methods
We defined our study area as the WHO European
region, minus Central Asia, Turkey and Israel.
National entities were the units of our analysis.
Because of changes in the political map of Europe
(which led to an increase in the number of independent states) and changes in the availability of data
(particularly on life expectancy), the number of
units included in the analysis rose from 13 in 1900
to 38 in 2008. We focused the analyses on 12 selected
points in time (i.e. 1900, 1910, 1920, 1930, 1939,
1950, 1960, 1970, 1980, 1990, 2000 and 2008), carefully avoiding the two World Wars (1914–18 and
1939–45).
Data on national income per head of population
were extracted from a dataset originally compiled by
Maddison (http://www.ggdc.net/MADDISON/oriindex.
htm). In contrast to other datasets, such as the Penn
1102
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
World Tables, this dataset is comprehensive in its
coverage across time and space, and provided us
with data on gross domestic product (GDP) per
head of population for almost all European countries
and the whole of the 20th century. These estimates
are based on extensive harmonization efforts and on
a conversion into 1990 International Geary-Khamis
dollars using multilateral purchasing power parities21
and can be found in Supplementary Table A1 (available as Supplementary data at IJE online). Average
national income for Europe as a whole (including
countries for which data were unavailable) was calculated as a population-weighted average of
Maddison’s GDP estimates for three European regions
which together cover the whole of Europe: Western
Europe, Central and Eastern Europe, and the (former)
Soviet Union.21
For each national entity we collected data on life
expectancy at birth, by sex, for each of the selected
years. The primary source of life expectancy data has
been the Human Lifetable Database (www.lifetable.
de), and where data were lacking these were supplemented by the World Health Organization Health for
All Database (http://data.euro.who.int/hfadb/), the
United Nations Demographic Yearbook Historical
Supplement, 1948-1997 (http://unstats.un.org/unsd/
demographic/products/dyb/DYBHist/HistTab09a.pdf),
the Meslé Vallin Database22 and various other
sources.23–25 The data used in the analysis can be
found in Supplementary Table A2 (available as Supplementary data at IJE online). Average life expectancy for Europe as a whole (including countries for
which data were unavailable) was estimated on the
basis of available values within three regions
(Western Europe, Central and Eastern Europe, and
the (former) Soviet Union) and their respective population shares.4,5
In order to assess the contribution of changes in
national income to changes in life expectancy, we regressed national life expectancy on GDP per capita for
each of the selected points in time. We used a simple
linear regression model, because the relation between
national income and life expectancy was approximately linear within the range of values observed
within Europe for each of these points in time.
Although addition of a squared term for GDP to the
regression model often resulted in an improvement of
the goodness-of-fit, even statistically significant deviations from linearity were small in magnitude (see
Supplementary Figure A1, available as Supplementary
data at IJE online). In order to check whether
changes in life expectancy can plausibly be attributed
to changes in national income, we also conducted a
longitudinal analysis in which we regressed the difference in life expectancy on the difference in GDP
per capita for each decade in between the selected
points in time.
We then calculated the contribution to life expectancy growth in Europe as a whole of changes in GDP
and of shifts in the relation between GDP and life
expectancy, using the approach developed by Preston.4,5 This analysis focused on four different time
periods: 1900–30, 1930–60, 1960–90, and 1990–2008.
For example, to determine the contribution of
changes in national income to changes in life expectancy between 1930 and 1960, we calculated life expectancy in Europe as a whole as expected on the
basis of national GDP in 1930 and the cross-sectional
GDP–life expectancy relation in 1960 (‘1930 GDP projected forwards’), and as expected on the basis of
GDP in 1960 and the cross-sectional GDP–life expectancy relation in 1930 (‘1960 GDP projected backwards’). The difference between life expectancy in
1930 and life expectancy expected on the basis of
1960 GDP projected backwards then indicates the
gain in life expectancy attributable to income
growth between 1930 and 1960, as does the difference
between life expectancy in 1960 and life expectancy
expected on the basis of 1930 GDP projected forwards.
Like Preston, we took the average of these two differences as our estimate of the contribution of income
growth per se to increases in life expectancy.
Also, the difference between life expectancy in
Europe as a whole in 1930 and life expectancy expected on the basis of 1930 GDP projected forwards
can be taken to indicate the gain in life expectancy
over the 1930–60 period that was attributable to a
shift in the relation between national income and
life expectancy, as can the difference between life expectancy in 1960 and life expectancy expected on
the basis of 1960 GDP projected backwards. We
took the average of these two differences as our estimate of the contribution of a shift in the relation to
increases in life expectancy. A similar procedure was
followed for changes occurring in the 1900–30, the
1960–90 and 1990–2008 periods.
In order to be able to interpret these changes better,
we conducted a similar analysis for mortality by cause
of death (all causes, infectious diseases, cardiovascular diseases). For the comparison of 1900, 1930 and
1960, age-standardized mortality data were extracted
from a compendium of International Mortality Statistics, from which we used the data given for 1901–05,
1931–35 and 1961–65.26. For later periods age-standardized mortality data were extracted from the
World Health Organization Health for All Database
(http://data.euro.who.int/hfadb/). ICD-code numbers
are given in Appendix Table A3 (available as Supplementary Data at IJE online).
Results
National income was positively associated with life
expectancy during the whole study period, both for
men and for women. Figure 1 shows the observations
and the fitted regression line for the year 2008, together with the regression lines for earlier years (each
line represents a year, and has been drawn within its
LIFE EXPECTANCY AND NATIONAL INCOME IN EUROPE
1103
Figure 1 Relation between national income and life expectancy, Europe, 1900–2008: regression lines. Black dots and upper
lines represent observations and fitted regression lines, respectively, for 2008. Other lines are fitted regression lines for
earlier years (1900, 1910, 1920, 1930, 1939, 1950, 1960, 1970, 1980, 1990, and 2000). Ca, circa; corr, correlation coefficient
observed range of values). Table 1 and Figure 2 present the regression parameters as obtained in the
cross-sectional analysis.
The slope parameter was always positive, and except
for the year 1900 in which the number of countries
was small, the estimate was always statistically
significantly different from 0. The strength of the association between GDP and life expectancy first goes
up to reach a peak value in 1920, both for men and
for women, and then goes down again, to remain
relatively stable during the last decades of the 20th
century. Nevertheless, in 2008 the association was still
quite strong: the slope parameter indicates that $1000
higher national income in this year went together
with 0.5 (men) or 0.4 (women) year higher life
expectancy.
Over time, the intercept of the regression equation
also changed substantially, but the main change
occurred between 1939 and 1950, when it rose from
42.3 to 55.5 years among men, and from 47.9 to 59.4
years among women. After 1950, there was a further
steep rise until 1960, but after 1960 the intercept remained stable for men, and increased only slowly for
women (Figure 1 and Table 1a).
The results of a longitudinal analysis in which
changes in national income were related to changes
in life expectancy can be found in Table 1b. This analysis gives a less consistent picture of the relation between national income and life expectancy: most
coefficients have large standard errors, and positive
coefficients indicating that a rise in national income
coincided with a rise in life expectancy are only seen
1104
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
Table 1 Relation between national income and life expectancy, and between changes in national income and changes in
life expectancy, Europe, by sex, 1900–2008: regression parameters
a. National income and life expectancy (cross-sectional analysis)
Men
Intercept
Women
Slope
Intercept
Slope
1900
No. of
countries
13
Estimate
40.54
STD
Error
3.79
Estimate
0.00144
STD
Error
0.00133
P-value
0.3049
Estimate
41.06
STD
Error
4.20
Estimate
0.00221
STD
Error
0.00148
P-value
0.1642
1910
14
38.18
4.83
0.00348
0.00155
0.0447
38.73
5.10
0.00417
0.00164
0.0257
1920
14
35.64
4.03
0.00470
0.00126
0.0028
38.01
4.13
0.00482
0.00129
0.0028
1930
19
41.68
2.11
0.00346
0.00056
0.0000
44.91
1.97
0.00350
0.00052
0.0000
1939
18
42.28
2.69
0.00367
0.00063
0.0000
47.85
1.92
0.00328
0.00045
0.0000
1950
23
55.53
1.28
0.00180
0.00029
0.0000
59.40
1.13
0.00184
0.00025
0.0000
1960
27
64.41
1.01
0.00045
0.00016
0.0077
68.12
0.88
0.00064
0.00014
0.0001
1970
33
64.59
1.05
0.00036
0.00012
0.0049
69.96
1.02
0.00047
0.00012
0.0003
1980
28
65.59
1.05
0.00037
0.00009
0.0003
71.31
0.83
0.00044
0.00007
0.0000
1990
41
65.90
1.00
0.00037
0.00008
0.0000
73.56
0.65
0.00033
0.00005
0.0000
2000
41
66.66
0.95
0.00038
0.00006
0.0000
74.79
0.55
0.00029
0.00004
0.0000
2008
34
64.56
1.33
0.00054
0.00008
0.0000
74.33
0.74
0.00037
0.00004
b. Change in national income and change in life expectancy (longitudinal analysis)
0.0000
Men
Women
Slope
Slope
1900-10
No. of
countries
13
Estimate
0.00015
STD
Error
0.00416
P-value
0.9718
Estimate
0.00044
STD
Error
0.00402
P-value
0.9149
1910-20
13
-0.00231
0.00126
0.0945
-0.00276
0.00112
0.0316
1920-30
14
-0.00222
0.00114
0.0756
-0.00229
0.00154
0.1641
1930-39
16
0.00228
0.00122
0.0822
0.00089
0.00092
0.3517
1939-50
17
-0.00139
0.00097
0.1733
-0.00142
0.00069
0.0573
1950-60
24
-0.00203
0.00049
0.0004
-0.00159
0.00047
0.0026
1960-70
25
0.00044
0.00034
0.2092
0.00039
0.00030
0.2032
1970-80
26
0.00027
0.00023
0.2502
0.00000
0.00025
0.9863
1980-90
26
0.00019
0.00012
0.1291
0.00013
0.00012
0.2806
1990-00
40
0.00030
0.00007
0.0001
0.00020
0.00005
0.0006
2000-08
33
0.00003
0.00008
0.7609
0.00005
0.00006
0.4114
In the cross-sectional analysis (a), life expectancy in a particular year is regressed on GDP in the same year. In the longitudinal
analysis (b), change in life expectancy between 2 years is regressed on change in GDP between the same years. Supplementary
Table A4 (available as Supplementary data at IJE online), presents a similar analysis of changes in national income and life
expectancy over longer time-periods.
during the 1930s, and then again from 1960 onwards.
Only the positive coefficient for the 1990–2000 period
was statistically significantly different from 0
(P < 0.001, both for men and for women).
As shown in Table 2, the largest increase in life expectancy in Europe as a whole occurred between 1930
and 1960. Among men, life expectancy increased from
51.67 to 66.83 years, or by 15.16 years. Two-thirds of
this increase can statistically be attributed to a shift in
the relation between GDP and life expectancy (estimated to have contributed 10.42 years), and one-third
to a rise in GDP (estimated to have contributed 4.74
years). The figures for women are similar: 11.47 years
out of a total rise in life expectancy of 16.48 years can
be attributed to a shift in the relation between national income and life expectancy.
Table 2 shows that the contribution of shifts in the
relation between national income and life expectancy
was even larger between 1900 and 1930, but became
much smaller after 1960 when only between onequarter (men) and one-half (women) of the increase
in life expectancy can be attributed to changes in
LIFE EXPECTANCY AND NATIONAL INCOME IN EUROPE
GDP per capita vs life expectancy
0.006
Men
cardiovascular disease mortality is statistically attributable to a rise in GDP, both among men (-0.93/-1.06
deaths per 1000 person-years) and among women
(-0.69/-0.77 deaths per 1000 person-years).
Women
0.005
Slope parameter
1105
0.004
0.003
Discussion
0.002
Summary of main findings
Large upward shifts in the relation between national
income and life expectancy only occurred before 1960,
and were due to rapid declines in mortality from infectious diseases independent of rises in national
income. These shifts account for between two-thirds
and four-fifths of the increase in life expectancy in
Europe as a whole during this period. After 1960,
upward shifts in the relation between national
income and life expectancy were much smaller, and
contributed between one-quarter and one-half to the
increase in life expectancy in Europe as a whole.
During the latter period, declines in mortality from
cardiovascular disease were mainly attributable to increases in national income.
2008
2000
1990
1970
1980
1960
1950
1939
1930
1910
1920
0
1900
0.001
Figure 2 Relation between national income and life expectancy, Europe, by sex, 1900–2008: slope parameters. Data
taken from Table 1a. For explanation see note to Table 1
intercept or slope. As we have seen in Table 1, after
1960 the intercept remained stable among men, and
increased only slightly among women, whereas the
slope parameter was stable among men and declined
among women. Between 1960 and 1990, male life expectancy in Europe increased by 3.17 years, and only
one-quarter of this increase can be attributed to a
shift in the relation with income (0.82 years).
Among women, a shift in the relation contributes
slightly more than half (2.90 years) to a total rise of
5.64 years in this period. Both in this period and in
the next, rises in income appear to have been more
important for the rise in life expectancy in Europe as
a whole.
Table 3 summarizes the results of a similar analysis
for mortality from all causes, infectious diseases and
cardiovascular diseases. The results of the underlying
regression analyses are presented in Supplementary
Tables A5–7 and Figures A1–3 (available as
Supplementary data at IJE online). The results for
all-cause mortality mirror those for life expectancy:
large declines between 1900 and 1930 and between
1930 and 1960 (statistically attributable mainly to
shifts in the relation between national income and
mortality), and smaller declines between 1960 and
1990 and between 1990 and 2008 (statistically attributable mainly to rises in national income).
Shifts in the relation with national income statistically explain most of the decline of infectious disease
mortality between 1900 and 1930 and between 1930
and 1960. For example, in the 1930–60 period such
shifts explain two-thirds of the decline in infectious
disease mortality, both among men (-1.83/-2.70
deaths per 1000 person-years) and among women
(-1.85/-2.74 deaths per 1000 person-years)). On the
other hand, rising incomes appear to explain most
of the decline in cardiovascular disease mortality between 1960 and 1990 and between 1990 and 2008. In
the latter period, almost 90% of the decline in
Strengths and limitations
A major strength of our study is that we have brought
together all European life expectancy data with documented validity, as well as a relatively complete set of
harmonized national income data. Period life expectancy is a commonly used, reliable and easily interpretable summary measure of mortality conditions
pertaining to a particular point in time.27 A possible
limitation is that tempo effects may distort the measurement of life expectancy in times of rapidly declining or increasing mortality,28,29 and that at lower
rates of mortality larger declines are necessary for
one unit increase in life expectancy.27 One should
therefore be cautious in comparing results between
periods, for example in Table 2. Smaller increases in
life expectancy in more recent periods do not necessarily imply smaller mortality declines, but as Table 3
shows, absolute declines in age-standardized mortality were really much smaller in 1960–90 or 1990–2008
than in 1900–30 or 1930–60. As the analysis of the
relative contribution of shifts vs income rises produces
similar results for age-standardized mortality and life
expectancy, our overall conclusions are unlikely to be
biased by any shortcomings of the measure of life
expectancy.
Another strength is that we have added cause-specific mortality analyses to the analysis of life expectancy. The mortality data, however, are less complete,
particularly for the pre-1960 period, when data on
mortality from all causes, infectious diseases and cardiovascular diseases are lacking completely for the
Soviet Union and are incomplete for several countries
elsewhere in Europe. 26 It is likely, however, that the
decline of infectious disease mortality was as important for increases in life expectancy in the Soviet
1106
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
Table 2 Contribution to life expectancy growth in Europe as a whole of changes in national income and of shifts in the
relation between national income and life expectancy, by sex, 1900–30, 1930–60, 1990–2008
Life expectancy
Men
Women
GDP 1900
GDP 1930
Due to GDP
Average
GDP 1900
GDP 1930
Due to GDP
Average
Rel. 1900
43.65
44.68
1.03
1.76
45.84
47.43
1.58
2.05
Rel. 1930
2.49
52.50
55.02
2.52
6.66
7.59
49.18
51.67
Due to shift
5.53
6.98
Average
6.25
8.01
7.12
9.17
GDP 1930
GDP 1960
Due to GDP
Average
GDP 1930
GDP 1960
Due to GDP
Average
Rel. 1930
51.67
60.05
8.39
4.74
55.02
63.50
8.49
5.01
Rel. 1960
65.73
66.83
1.10
69.95
71.50
1.54
Due to shift
14.06
6.78
14.94
8.00
Average
10.42
15.16
11.47
16.48
GDP 1960
GDP 1990
Due to GDP
Average
GDP 1960
GDP 1990
Due to GDP
Average
Rel. 1960
66.83
69.41
2.58
2.35
71.50
75.12
3.62
2.74
Rel. 1990
67.88
70.00
2.12
75.28
77.14
1.85
Due to shift
1.05
0.58
3.79
2.02
Average
0.82
3.17
2.90
5.64
GDP 1990
GDP 2008
Due to GDP
Average
GDP 1990
GDP 2008
Due to GDP
Average
Rel. 1990
70.00
71.43
1.44
1.76
77.14
78.40
1.26
1.34
Rel. 2008
70.49
72.58
2.09
78.37
79.79
1.42
Due to shift
0.50
1.15
1.23
1.39
Average
0.82
2.59
1.31
2.65
Each block in this table can be understood as follows. For example, average male life expectancy in Europe as a whole in 1900 is
given at the intersection of ‘GDP 1900’ and ‘Relation 1900’, i.e. 43.65 years. Similarly, average male life expectancy in Europe as a
whole in 1930 is given at the intersection of ‘GDP 1930’ and ‘Relation 1930’, i.e. 51.67 years. The difference between these two
values, i.e. 8.01 years, is the change in life expectancy in Europe as a whole between these two points in time. This change has
been decomposed in two parts: a change attributable to a rise in national income (‘Due to GDP), i.e. 1.76 years, and a change
attributable to a shift in the relation between national income and life expectancy (’Due to shift’), i.e. 6.25 years. Each of these two
values represents the average of two preceding estimates. In the case of the change attributable to a rise in national income (1.76
years) it is the average of the difference between life expectancy in Europe as a whole in 1900 (43.65 years) and life expectancy
expected on the basis of 1930 GDP projected backwards (44.68 years), i.e. 1.03 years, and the difference between life expectancy in
1930 (51.67 years) and life expectancy expected on the basis of 1900 GDP projected forwards(49.18 years), i.e. 2.49 years.See
Methods for further explanation.
Union as it was for the Central and Eastern European
countries (such as Bulgaria and Romania) for which
cause-of-death data were available. The cause-ofdeath data are of uncertain validity, particularly in
the earlier parts of the study period and, over the
20th century, changes in cause-of-death classification
have occurred. Because our analysis was limited to
two broad groups of causes of death, however, the
results are unlikely to be seriously biased. We used
a careful reconstruction of the code numbers pertaining to infectious and cardiovascular diseases in previous editions of the International Classification of
Diseases (Supplementary Table A2, available as
Supplementary data at IJE online), so changes in classification between editions are unlikely to have affected our results.
Our analysis of the contribution of changes in
national income vs shifts in the relation between national income and life expectancy is based on crosssectionally observed associations, like Preston’s
original analysis. This analysis is based on deterministic functions, ignoring stochastic uncertainty, and the
cross-sectional associations may not provide a good
representation of underlying causal relationships.
There may be confounding by third factors that are
associated with both national income and life
LIFE EXPECTANCY AND NATIONAL INCOME IN EUROPE
1107
Table 3 Contribution to mortality decline in Europe as a whole of changes in national income and of shifts in the relation
between national income and life expectancy, by sex, 1900–30, 1930–60, 1960–90, 1990–2008
Life expectancy
Men
1900-30
Increase LE
8.01
Due to GDP
1.76
Women
Due to shift
6.25
Increase LE
9.17
Due to GDP
2.05
Due to shift
7.12
1930-60
15.16
4.74
10.42
16.48
5.01
11.47
1960-90
3.17
2.35
0.82
5.64
2.74
2.90
1990-08
2.59
1.76
0.82
2.65
1.34
1.31
All cause mortality
Men
Decline ASDR
Due to GDP
Women
Due to shift
Decline ASDR
Due to GDP
Due to shift
1900-30
12.48
1.46
11.02
11.25
1.20
10.04
1930-60
11.39
3.71
7.68
10.49
2.78
7.71
1960-90
2.00
1.61
0.39
2.74
1.37
1.37
1990-08
1.76
1.28
0.49
1.20
0.73
0.47
Infectious disease mortality
Men
Decline ASDR
Due to GDP
Women
Due to shift
Decline ASDR
Due to GDP
Due to shift
1900-30
2.17
0.37
1.80
2.14
0.38
1.76
1930-60
2.70
0.88
1.83
2.74
0.89
1.85
1960-90
0.30
0.22
0.08
0.37
0.21
0.16
1990-08
0.03
0.02
0.05
0.01
0.00
0.01
CVD mortality
Men
Decline ASDR
Due to GDP
Women
Due to shift
1900-30
0.17
0.07
0.24
Decline ASDR
0.77
Due to GDP
0.04
Due to shift
0.82
1930-60
0.73
0.28
0.45
0.20
0.12
0.32
1960-90
0.10
0.36
0.26
0.75
0.52
0.23
1990-08
1.06
0.93
0.13
0.77
0.69
0.07
LE, life expectancy (in years); ASDR, age-standardized death rate (deaths per 1000 person-years); CVD, cardiovascular disease;
‘Due to GDP’ and ‘due to shift’ see explanation in note to Table 2, and in Methods; 08, 2008
expectancy, such as levels of education30 or longstanding cultural and political differences between
countries. In the post-war period, the division between countries with higher and lower national incomes more or less coincided with the division
between a capitalist Western part and a (post-)communist Eastern part of Europe.21 Unfortunately, the
number of countries is too small for separate analyses
of the relation between national income and life expectancy within these two regions, but additional
analyses controlling for region still suggest an (attenuated) relation between national income and life expectancy (see Supplementary Tables A8 and A9,
available as Supplementary data at IJE online).
There may also be bias from reverse causality if
increased life expectancy, or improved population
health generally, contributes to higher national
income.31–33 We have therefore refrained from explicitly causal interpretations.
That caution in interpretation is needed is also clear
from the fact that associations between changes in
national income and changes in life expectancy are
weaker and less consistent than cross-sectional associations.7 In our own longitudinal analyses
(Table 1b; also see Supplementary Table A2, available
as Supplementary data at IJE online) we found a
weak tendency for a rise in national income to be
associated with a rise in life expectancy briefly
during the 1930s, and then a bit more clearly from
1960 onwards. However, during the 1910s and 1920s
and during the 1940s and 1950s the association
tended to be negative, indicating a faster rise in life
1108
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
expectancy in countries with a slower rise in national
income. This implies that our cross-sectional analysis
may have overestimated the contribution of rising national incomes to rises in life expectancy in the periods 1900–30 and 1930–60, which would strengthen
our main conclusion of an over-riding contribution
of shifts in the relation between national income
and life expectancy before 1960. On the other hand,
our finding that the contribution of rising national
income to rises in life expectancy was more important
after 1960 finds some support in the results of the
longitudinal analyses.
Our analysis focused on Europe, and therefore
upward shifts in the relation between national
income and life expectancy cannot be attributed to
development aid, although some aid with implementing public health measures (e.g. by the American government or private foundations34) may have played a
role in the reduction of infectious disease mortality in
Southern European countries immediately after World
War II. It is uncertain, however, whether our findings
for the post-1960 period can be generalized to other
high- and middle-income countries. Trends in national income in this period in Central and Eastern
Europe were characterized by stagnation until 1990
and a steep decline followed by recovery after
1990.21 Our results suggest that these dramatic
changes in national income have left deep traces in
the life expectancy and mortality record of Europe,
but it remains to be seen whether changes in national
income have also been the main drivers behind life
expectancy change in high- and middle income countries in other parts of the world. The role of economic
growth in post-1960 life expectancy improvements in
low-income countries is highly controversial.35,36
Interpretation
Our European data for the period 1930–60 largely ‘reproduce’ Preston’s findings for this same period for
the whole world: the increase in life expectancy in
that time period was accompanied by a strong
upward shift of the relation between national
income and life expectancy. As mentioned in the
introduction, this has been attributed to the rapid
introduction of public health measures, whose implementation was relatively independent of a country’s
own economic growth, either because of international
development aid or because of the inexpensiveness of
the technology. We found that infectious disease mortality decline in Europe between 1930 and 1960 was
largely due to a strong downward shift of the mortality–national income relation, suggesting rapid diffusion of knowledge and technology independent of a
country’s economic conditions at the time.37
Our findings, however, suggest that in other periods
the mechanisms behind the increase in life expectancy were probably different. After 1960, growth in
life expectancy slowed down, and although the
strength of the association between national income
Figure 3 Change in national income vs change in agestandardized death rates from cardiovascular disease,
change in percentage daily smokers, and change in health
care expenditure per capita, Europe, 1990–2008. ASDR, agestandardized mortality rate (in deaths per 1000 personyears); GDP, gross domestic product (in I$ per capita); CVD,
cardiovascular disease; HCE, health care expenditure (in I$
per capita). Data on cardiovascular disease mortality,
smoking and health care expenditure come from the World
Health Organization Health for All Database (http://data.
euro.who.int/hfadb/). Each dot represents a country
LIFE EXPECTANCY AND NATIONAL INCOME IN EUROPE
and life expectancy became weaker, the enormous rise
in incomes statistically explains most of the increase
in life expectancy. Whereas some authors have argued
that national income has become less important, perhaps to be replaced by income inequalities as main
determinants of life expectancy,20,38 our results suggest that national income still counts. Changes in the
relation between national income and life expectancy
contributed relatively little to the rise in life expectancy in this period, particularly among men.
Among women, the contribution of shifts in intercept
was a bit larger, and if this indicates at least some diffusion of cardiovascular disease control independent
of national income growth, any beneficial effects on
men must have been counteracted by other factors,
e.g. their greater vulnerability to economic crises,39
perhaps in the form of psychosocial stress40 and/or
excessive alcohol consumption.41
On the whole, however, our results suggest that diffusion of knowledge and technologies for cardiovascular disease control has been strongly dependent
on a country’s economic growth. Whereas this may
indicate a lack of solidarity between European countries, it also points to the possibility that effective cardiovascular disease control required substantial
investments. Figure 3 presents the results of a longitudinal analysis for changes in mortality from cardiovascular disease during the 1990–2008 period, for
which data on a few potential determinants are available as well. Figure 3a shows that the cross-sectional
relation between national income and cardiovascular
disease mortality may indeed have a causal component: stronger rises in national income in this period
are accompanied by stronger declines in cardiovascular disease mortality, and the few countries with
declining national incomes saw their cardiovascular
disease mortality rate go up. Although rises in national income are associated weakly with declines
in smoking (Figure 3b), there is a much stronger
association with rises in health care expenditure
(Figure 3c). This could be interpreted as confirming
the importance of expensive technology for controlling cardiovascular disease. Treatment costs for these
conditions have gone up strongly in high-income
1109
countries, and although this has been cost effective42
it is perhaps unsurprising that countries with fewer
resources have been unable to raise their expenditure
to the same levels.
Conclusions
In contrast to earlier periods, recent life expectancy
growth in European countries appears to have been
dependent on their economic growth. More rapid diffusion of knowledge and technology for cardiovascular disease control from higher- to lower-income
countries in Europe may be needed to close the
East–West life expectancy gap, but it is unlikely that
this can be achieved in the absence of more equitable
economic conditions.
Supplementary Data
Supplementary data are available at IJE online.
Funding
This study was partly supported by Netspar, the
Network of Pensions, Aging and Retirement
(Tilburg, the Netherlands).
Acknowledgements
The authors thank France Meslé and Jacques Vallin
(Institut National des Etudes Démographiques, Paris,
France) and Domantas Jasilionis (Max Planck
Institute for Demographic Research, Rostock,
Germany) for sharing some of their data. The authors
also thank Samuel Preston (Sociology Faculty,
University of Pennsylvania, Philadelphia, USA) and
David Leon (London School of Hygiene and Tropical
Medicine, London, UK) for their insightful comments
on a previous version of this paper.
J.P.M. will act as guarantor for this paper.
Conflict of interest: None declared.
KEY MESSAGES
In the past, upward shifts of the so-called Preston curve, which relates life expectancy to national
income, have contributed importantly to worldwide increases in life expectancy.
In Europe, such large upward shifts only occurred before 1960, and were due to rapid declines in
mortality from infectious diseases which were independent from rises in national income.
After 1960, upward shifts in the relation between national income and life expectancy were much
smaller, and contributed only between one-quarter and one-half to the increase in life expectancy in
Europe as a whole.
More rapid diffusion of knowledge and technology for cardiovascular disease control from higher- to
lower-income countries in Europe may require more equitable economic conditions.
1110
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
References
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
Krieger N. Epidemiology and the People’s Health. Theory and
Context. Oxford: Oxford University Press, 2011.
Kunitz SJ. The Health of Populations. General Theories and
Particular Realities. Oxford: Oxford University Press, 2007.
McKeown TF. The Origins of Human Disease. Oxford:
Blackwell, 1988.
Preston SH. The changing relation between mortality and
level of economic development. Popul Stud (Camb) 1975;
29:231–48.
Preston SH. Mortality Patterns in National Populations, With
Special Reference to Recorded Causes of Death. New York:
Academic Press, 1976.
Caldwell JC. Mortality in relation to economic development. Bull World Health Organ 2003;81:831–32.
Bloom DE, Canning D. Commentary: The Preston Curve
30 years on: still sparking fires. Int J Epidemiol 2007;36:
498–99.
Kunitz SJ. Commentary: Samuel Preston’s ‘The changing
relation between mortality and level of economic development’. Int J Epidemiol 2007;36:491–92.
Leon DA. Commentary: Preston and mortality trends
since the mid-1970s. Int J Epidemiol 2007;36:500–01.
Mackenbach JP. Commentary: Did Preston underestimate
the effect of economic development on mortality? Int J
Epidemiol 2007;36:496–97.
Riley JC. Commentary: Missed opportunities. Int J
Epidemiol 2007;36:494–95.
Wilkinson RG. Commentary: The changing relation between mortality and income. Int J Epidemiol 2007;36:
492–94.
Mackenbach JP, Looman CW. Secular trends of infectious
disease mortality in The Netherlands, 1911–1978: quantitative estimates of changes coinciding with the introduction of antibiotics. Int J Epidemiol 1988;17:618–24.
Gwatkin DR. Indications of Change in Developing
Country Mortality Trends. Popul Dev Rev 1980;6:615–44.
Olshansky SJ, Ault AB. The fourth stage of the epidemiologic transition: the age of delayed degenerative diseases.
Milbank Q 1986;64:355–91.
Vallin J, Meslé F. Convergences and divergences in mortality. A new approach to health transition. Demogr Res
2004;2:11–44.
Leon DA. Trends in European life expectancy: a salutary
view. Int J Epidemiol 2011;40:271–77.
World Bank. World Development Report 1993. Washington:
World Bank, 1993.
Deaton A. Health in an Age of Globalization. Cambridge:
National Bureau for Economic Research, 2004.
Rodgers GB. Income and inequality as determinants of
mortality: An international cross-section analysis. Popul
Stud 1979;33:343–51.
Maddison A. The World Economy: Historical Statistics. Paris:
Organisation of Economic Cooperation and Development,
2003.
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
Vallin J, Meslé F. The Segmented Trend Line of Highest
Life Expectancies. Popul Dev Rev 2009;35:159–87.
Vallin J, Mesle F, Adamets S, Pyrozhkov S. A new estimate of Ukrainian population losses during the crises of
the 1930s and 1940s. Popul Stud 2002;56:249–63.
Duthé G, Badurashvili I, Kuyumjyan K, Meslé F, Vallin J.
Mortality in the Caucasus: An attempt to re-estimate
recent mortality trends in Armenia and Georgia. Demogr
Res 2010;22:691–732.
Meslé F, Shkolnikov V, Vallin J. Mortality by cause in the
USSR in 1970-1987: the reconstruction of time series. Eu
J Popul 1992;8:281–308.
Alderson M. International Mortality Statistics. New York:
Facts on File, 1981.
Preston SH, Heuveline P, Guillot M. Demography. Measuring and Modelling Population Process. Oxford: Blackwell
Publishing, 2001.
Bongaarts J, Feeney G. How long do we live. Popul Dev
Rev 2002;28:13–29.
Luy M. Mortality tempo-adjustment: An empirical application. Demogr Res 2006;15:561–90.
Caldwell JC. Routes to Low Mortality in Poor Countries.
Popul Dev Rev 1986;12:171–220.
Fogel RW. The Escape From Hunger and Premature Death,
1700-2100. Europe, America, and the Third World.
Cambridge: Cambridge University Press, 2004.
Acemoglu D, Johnson S. Disease and Development: The Effect
of Life Expectancy on Economic Growth. Cambridge (MA):
National Bureau for Economic Research, 2006.
Swift R. The relationship between health and GDP in
OECD countries in the very long run. Health Econ 2011;
20:306–22.
Shaplen R. Toward the Well-Being of Mankind: Fifty Years of
the Rockefeller Foundation. New York: Doubleday, 1964.
Cutler DM, Deaton A, Lleras-Muney A. The determinants
of mortality. J Econ Perspect 2006;20:97–120.
Pritchett L, Summers LH. Wealthier is healthier. J Hum
Resources 1996;31:841–68.
Omran AR. The epidemiologic transition theory. A preliminary update. J Trop Pediatr 1983;29:305–16.
Wilkinson RG. Income distribution and life expectancy.
BMJ 1992;304:165–68.
Leon DA, Saburova L, Tomkins S et al. Hazardous alcohol drinking and premature mortality in Russia: a
population based case-control study. Lancet 2007;369:
2001–09.
Marmot M, Bobak M. Psychosocial and biological mechanisms behind the recent mortality crisis in Central and
Eastern Europe. In Cornia GA, Paniccia R (eds). The
Mortality Crisis in Transitional Economies. Oxford: Oxford
University Press, 2000. pp. 127–48.
Leon DA, Shkolnikov VM, McKee M. Alcohol and
Russian mortality: a continuing crisis. Addiction 2009;
104:1630–6.
Cutler DM, McClellan M. Is technological change in
medicine worth it? Health Affairs 2001;20:11–29.