Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
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.