11. A composite view on inequality and well-being

Auke Rijpma
Utrecht University

By the end of the 2010s, the global economy was producing far more than at any time in the past. However, it is not equally evident that all the world’s citizens have improved their living conditions at the same pace. Has everyone reaped the benefits of increased GDP?

While GDP and productivity are important measures of material progress, for nations as well as individuals, there is a widely felt need to look “beyond GDP” to understand people’s well-being. To this end, statistical agencies and academic researchers are expanding data collection and analyses to include different types of well-being metrics. Since well-being is inherently multidimensional, this implies tracking a wide range of indicators to obtain a more complete picture of development (Stiglitz, Sen and Fitoussi, 2009[1]; Boarini and Mira D’Ercole, 2013[2]).

Economic and social historians are no exception to this trend. They had already gathered much data about development and well-being in the long-run. Thanks to the efforts to create historical national accounts, country-level series on GDP and productivity are widely available (Maddison, 2001[3]). These series are continuously being improved and are now going back further in time (Bolt and van Zanden, 2014[4]). Equally important, however, is that historical statistics have always included other well-being measures such as real wages, height and mortality (Steckel and Floud, 1997[5]; Feinstein, 1998[6]; Bengtsson and van Poppel, 2011[7]). Efforts to collect, harmonise and analyse such data are resulting in an ever-clearer picture of long-run developments. The previous How Was Life? volume presented the state of the art in this field (van Zanden et al., 2014[8]). Besides GDP per capita, it provided insights into the long-term development of wages, educational attainment, health (life expectancy and height), safety, political freedoms, the environment and gender inequality.

While long-run developments in well-being are clearer than ever, one key challenge is to move beyond country averages to focus on the distribution of well-being outcomes within countries. For one, this is necessary because these averages will always miss part, or sometimes even most, of the distribution of well-being. Moreover, inequality is often seen as an obstacle to achieving societal progress in general (OECD, 2015[9]). However, most of this research focuses on income inequality, while inequality in other dimensions, such as wealth, health and education, could be equally salient.

This volume therefore has a twofold contribution compared to the previous How Was Life? report (van Zanden et al., 2014[8]). The first is to expand the set of average (country-level) well-being indicators for which we have long-term series. These indicators are working hours, extreme poverty and improved per capita GDP series. The second contribution is to go beyond the income and gender inequalities covered in the previous volume to look at inequalities in education, length of life and wealth.

The goal of this chapter is to provide an integrated view based on this new historical data. The chapter focuses on the new indicators covered in this volume that have sufficient country coverage, such as working hours and extreme poverty, but also builds on some of evidence from the previous How Was Life? report. Above all, however, this chapter analyses the development of inequality of well-being at the world level. After introducing the data and concepts, the chapter describes trends in within-country inequality, and then develops a set of composite indicators to summarise progress in the full dataset of historical well-being indicators underpinning this book.

The underlying data and concepts used are described in detail in the relevant chapters of this volume and in (van Zanden et al., 2014[8]). This section briefly summarises the most important points about the newly added data, and describes how these data are used in this chapter.

First, the measures of inequality in length of life and education are to an extent affected by a “right truncation” in the distribution as, at some point, obtaining more education or greater longevity becomes difficult. This means that, as the average level of education or life expectancy increases and comes closer to the truncation point, inequality mechanically decreases, a pattern that does not necessarily hold for income and wealth. It should therefore be remembered that inequality trends in these indicators might be driven by countries’ positions relative to the maximum achievable levels.

The poverty indicator used here is extreme poverty, measured using a cost-of-basic-needs approach. It measures what share of the population cannot afford a basic consumption bundle. In this approach, the poverty metric deviates from the World Bank’s “one-dollar-a-day” method, which uses an average of national poverty lines as the threshold (Chen and Ravallion, 2010[10]; Ravallion, Datt and van de Walle, 1991[11]). From a long-term perspective, however, the two indicators provide broadly consistent pictures (Ravallion, 2020[12]). Unless stated otherwise, the regional and global series of inequality indicators in this chapter are population-weighted averages of country-level Gini coefficients. Consequently, the resulting figures cannot be interpreted as a proper Gini for the world or a world region, i.e. a Gini coefficient calculated over all individuals in the world, irrespective of where they live. While the population-weighted averages can give an impression of regional developments in inequality, this limitation should be kept in mind.

More generally, all regional and global trends presented in this chapter are population-weighted averages based on all countries for which data are available in the Clio-Infra database, i.e. stretching beyond the 25 countries covered in this book. To limit the risk that the trends in the global and regional series presented in this volume simply reflect the greater availability of data over time, data for countries with missing data are imputed by (log-linear) interpolation and extrapolation.1 Regional and global averages are reported only if at least 40% of the population in a region or the world is covered by non-imputed data.

To summarise the developments in the many well-being indicators gathered in this book, this chapter makes use of composite indicators like the one that was used in the previous How Was Life? volume (van Zanden et al., 2014[8]). In this chapter, this composite indicator is updated with the new indicators of the present volume. The other novelty is that two separate composite indicators are presented: one for the indicators on average well-being in countries, and one for the measures of inequality of well-being within countries.

Composite indicators can be controversial. The heart of the issue is that the indicators being aggregated are conceptually different and have different units and scales. Combining them into one number requires forcing the variables to a common scale, selecting an aggregation function and choosing weights. The trade-offs in composite indicators implied by this procedure amount to statements on the relative importance of each indicator for overall well-being (Ravallion, 2012[13]; Ravallion, 2012[14]). Since people can hold different opinions on the relative importance of each dimension of well-being, it is very difficult to devise a completely satisfactory solution to this issue.

That said, composite indicators have their own benefits. Above all, they are useful tools to summarise the large number of indicators gathered for a volume such as this. It is likely that readers will summarise the masses of data they see, either on their own account, or because the introduction, conclusion or executive summary highlight certain patterns at the expense of others. The composite indicators presented in this chapter do this in a systematic, disciplined and transparent way.

This chapter relies on the same approach to composite indicators as used in the previous How Was Life? volume (Rijpma, 2014[15]). A latent variable model is used to extract one or more common factors from the variables entering the model. It does this by finding the shared information between the indicators in a way that distinguishes between countries as best as possible. To this end, the procedure assigns higher weights to indicators that are highly correlated, and vice versa.

The main disadvantage to this approach is that such a statistical procedure is not guaranteed to provide correct, or even satisfying trade-offs. When each indicator captures a unique part of well-being, a latent variable model can give problematic results, because such variables might have a low correlation. There are of course also advantages to the approach. For one, the functional form of aggregation implicit in the latent variable approach comes down to a linear aggregation with minimal transformations. The indicators are standardised only to have a mean of zero and a standard deviation of one to facilitate computation. This keeps the trade-offs simple and transparent (Ravallion, 2012[13]; Chakravarty, 2003[16])). The specific model used here can moreover deal with missing data, which is an important issue for composite indicators, because they need full data for a given year and country to be calculated. This is of course difficult to achieve with historical data without an imputation procedure. Finally, the statistical approach used in this chapter can also provide estimates of uncertainty, including that caused by the imputation of missing data, and can do so at the regional level for which much of the data is reported. Details can be found in (Rijpma, 2016[17]) and (Rijpma, 2014[15]), which are in turn based on (Jackman, 2009[18]) and (Høyland, Moene and Willumsen, 2012[19]).

A detailed discussion of the sources and data quality of individual measures is provided in the other chapters of the current and previous volumes (van Zanden et al., 2014[8]). In this section, only a number of general features are noted.

Overall, data quality improves as we move closer to the present. For the period after World War II, the data behind most indicators are gathered either by statistical agencies or by researchers using similar methods to statistical agencies. Prior to World War II, estimates are frequently based on research using imperfect data. In the earliest decades, in particular the first half of the 19th century, the quality of the data is lower as data becomes scarcer, and guesstimates become inevitable.

The quality of data behind the new indicators that are analysed in this chapter should be summarised briefly. For working hours, data falls in the highest-quality category from the 1930s onwards. When data are available for earlier periods in Western Europe and the Western Offshoots, they are also typically of fairly high quality (research using the same methods as statistical agencies). Coverage outside these regions before the 1930s is however limited. Measures of extreme poverty are of high or fairly high quality since the 1950s. Data for the 19th century are typically worse, with guesstimates becoming common, often based on estimates of per capita GDP and income inequality. It is also important to note that price data in socialist countries can be unreliable, because they could be set by the government and goods were not always available to be bought at those prices. Because of this, guesstimates remain necessary in these countries until well into the 20th century.

Regarding the new inequality measures, the available data on inequality in length of life are of good quality. However, coverage is limited outside the advanced economies of Western Europe and its Offshoots before the 1950s. The data underlying the educational inequality estimates are of fairly high quality from the 1950s onwards. Data are scarcer for the first half of the 20th century, and guesstimates are frequent in the 19th century. Data on wealth inequality, finally, are very scarce outside Western Europe and the Western Offshoots. Even as late as the 1990s, estimates are available for less than 10% of the countries in the Clio-Infra datasets. For this reason, this indicator is unfortunately not part of the inequality analysis of this chapter.

Figure 11.1 shows the world population-weighted average over time for six new indicators added in this volume, the new historical GDP per capita series described in Chapter 2, as well as the income inequality series included in van Zanden et al. (2014[8]). The developments in these indicators confirm one of the overall conclusions of the previous How Was Life? volume: that of improvements in well-being in the world over the past 200 years. The new GDP per capita series utilising better PPPs still shows huge progress in productive capabilities over the past 200 years. While this progress is not distributed evenly over the globe, the average economic well-being of the world’s citizens has improved considerably.

Globally, extreme poverty decreased over almost the entire period. Progress was particularly fast in the 1950s, 1970s, 1990s and 2000s, when on average, extreme poverty declined by five percentage points or more over each 10-year period. Recently, the decline has been concentrated in Asia and sub-Saharan Africa; prior to that, the declines were concentrated in the other regions.

Weekly working hours in manufacturing declined throughout the period as well. While country coverage outside Western Europe and the Western Offshoots before 1950 is limited, from that moment onwards working hours declined across the globe. In a forty-year period, they declined from 50 hours to less than 44 hours per week in the 1990s. For countries where data are available from an earlier date, working hours were higher still, commonly as high as 60 in the 19th century. The decline was set in motion in the second half of the 19th century, with the largest progress made in the first half of the 20th century. Very recently, weekly working hours have been increasing again.

The inequality indicators display a more diverse pattern. On average, income inequality decreased from the early 20th century up until the 1980s, after which it started increasing again. The coverage for wealth inequality allows us to make statements only for the 1990s and later, when it was clearly increasing. However, a slow decline in wealth inequality before this period can be observed for the countries for which data are available.

Length-of-life inequality decreased throughout the period for which we have sufficient data, for both men and women. Only in the 1990s can a slowdown be observed, mostly for men. This decrease was mostly concentrated in Eastern Europe and the former Soviet Union, where inequality increased for men. Educational inequality too has been decreasing for most of the period covered by our data, with a slowdown becoming visible only in the 2010s. These changes in inequality suggested by Figure 11.1 are investigated in more depth in the next section.

To understand these trends as well as those for the composite indicator discussed below, it is useful to look at the correlation of the indicators with per capita GDP (Figure 11.2). Compared to the core set of indicators in the previous volume, correlations with per capita GDP are somewhat weaker. Whereas correlation coefficients of 0.5 or higher were consistently reported for indicators such as (average) real wages, average years of education, or life expectancy, such large coefficients are less frequent now.

Extreme poverty does have a fairly strong negative correlation (-0.5/-0.7) with GDP per capita, although well below the strong correlation suggested by (Dollar and Kraay, 2002[20]). Figure 11.2 shows a consistent, negative correlation of working hours with GDP per capita, which is however not very strong and measured with some uncertainty. The correlation is clearly negative only from the 1960s onwards and has become weaker in recent decades, due to increases in working hours in a number of high-income regions.

Looking at the inequality indicators, the correlation of income inequality with per capita GDP has changed considerably over the course of the 20th century. In the early part of the century, the correlation was positive, meaning that high-income countries also experienced high-income inequality. This relation had reversed after World War II, so that advanced economies tended to be more equal. Today the correlation is close to zero.

Length-of-life inequality and educational inequality show a negative correlation with GDP per capita, implying that richer countries had lower inequality in health and education. In the 19th century, however, the correlation with length-of-life inequality is measured with substantial uncertainty, implying that it is probably safest to say that, in that period, length-of-life inequality did not display a strong relation with per capita GDP. Conversely, the correlation of educational inequality with per capita GDP is consistently negative. For both inequality indicators it is worth remembering that the levels of average years of education and life expectancy are strongly correlated with per capita GDP in the same period (van Leeuwen and van Leeuwen-Li, 2014[21]; Zijdeman and Ribeiro de Silva, 2014[22]).

Figure 11.3 shows the overall relation of the three inequality indicators with GDP per capita. Overall, all inequality measures display a negative correlation with GDP per capita. Strong negative relations are observed for educational inequality and length-of-life inequality. Income inequality and wealth inequality, however, show a more complex pattern, with a positive relation at low levels of GDP per capita, and a negative one at higher levels.

The discussion of the main global trends in well-being inequality in the previous section suggests that the 20th century may have been characterised by a U-shape in inequality in a number of dimensions. From a high point at the start of the 20th century, income inequality declined to a low in the 1970s-80s, after which it started rising again, This U-shaped pattern matches the findings of recent research on income and wealth inequality (Piketty, 2014[23]; Scheidel, 2017[24]). While educational inequality and length-of-life inequality declined more consistently, there too a slowdown can be observed more recently.

Figure 11.4 looks at the average change in the Gini coefficient in income inequality by region to pinpoint the moment when income inequality started declining. Looking at Western Europe and its Offshoots, the 1920-29 period was, on average, still characterised by increasing inequality. From the 1930s to the 1970s in Western Europe, and until the 1960s in the Western Offshoots, income inequality declined. Therefore, according to the data presented here, the decline in income inequality in the advanced economies did not start immediately after World War I, but only began in earnest during the 1930s. Likewise, the rise of income inequality in the Western Offshoots in the 1960s seems to pre-date the breakdown of the Keynesian consensus and the adoption of neoliberal policies. More precise data than the decennial estimates presented here should however be used to determine this turning point more precisely.

Moving to other regions, the decline in income inequality in the 1930s and 1940s affected most regions with the exception of South and Southeast Asia, although in the Middle East and North Africa (MENA) the decline was very small. The recent increase of income inequality also extended to Asia and particularly Eastern Europe and the former Soviet Union, although not to sub-Saharan Africa, the MENA and the Latin America and the Caribbean.

Educational inequality declined throughout the period (Figure 11.5). The decline was strongest in the 1950s-80s period but has recently slowed down to a point of near-stagnation. The decline in educational inequality was less pronounced in Western Europe and the Western Offshoots, regions where educational inequality was already fairly low in the 1870s. However, the decline in educational inequality affected all regions throughout the period. The U-shaped pattern for income inequality does not apply to educational inequality.

Finally, we look at regional changes in length-of-life inequality (Figure 11.6). Data coverage is an issue here, with few countries outside Western Europe and the Western Offshoots having enough data to allow for calculating regional averages over a long period. What stands out from Figure 11.6 is that, in most of the world’s regions, the trend is towards lower inequality. This trend is common to all regions and concentrated in the middle of the 20th century. The slowdown of this decline in the late 20th century is also fairly widespread, but particularly pronounced in Eastern Europe and the former Soviet Union, where length-of-life inequality even increased.

Figure 11.7 compares inequality between 1910 and 2000/2010, two points in time for which data coverage is relatively good for all inequality indicators. What stands out is that, in the case of income, inequality in 1910 is very weakly correlated with inequality in 2000. This also means that that there are countries featuring high inequality in 1910 that are ranked relatively low in 2000, and vice versa. Countries such as Sweden, France, Japan, Italy and the Netherlands moved from being highly unequal countries in terms of income to being relatively equal countries in 1990. Countries like South Africa, Brazil and India show the opposite development.

The other dimensions of inequality display a different pattern. Often, inequality in 1910 is predictive of inequality in 2000/2010. Moreover, there is a degree of convergence, as very unequal countries in 1910 achieved stronger reductions in inequality than the most equal countries in the same year. What this means for multidimensional inequality is that countries that became more equal in terms of income inequality have also become more equal in other dimensions. Countries that became more unequal in terms of income over the 20th century compensated for this to some extent in other dimensions.

In Figure 11.8, these developments are analysed further by considering the entire inequality trajectory for the 25 “Clio-Infra countries” – countries with good historical data that together cover a large share of the population from various world regions.2 By splitting the sample by the level of inequality in 1910, this figure shows whether high-inequality countries developed differently from low-inequality ones. High-inequality countries are defined as those with above median inequality in 1910. Figure 11.8 shows that in the case of educational inequality, high and low inequality countries in 1910 converged. Despite fewer observations, developments in length-of-life inequality are similar. For income inequality, the pattern is however different. Countries with low income inequality became somewhat more unequal on average, and countries with high income inequality became somewhat more equal. In the 1960s and 1970s, both high- and low-income inequality countries converged to a lower level of income inequality, after which increases were common to many countries.

As a final way of analysing trends in the well-being indicators covered by the two How Was Life? volumes, this section looks at composite indicators. Two distinct indicators are used. The first covers the country averages indicators (levels) from the two volumes. The indicators taken from van Zanden et al. (2014[8]) are average real wages, height, life expectancy, average years of education, biodiversity, democracy and homicide rates. The old GDP per capita series are replaced by the new ones that incorporate the more recent PPP estimates presented in Chapter 2. Data on working hours and extreme poverty are added from the present volume.

The second composite indicator is based on the inequality indicators: income inequality from the previous volume, and length-of-life inequality, educational inequality and gender equality from the present volume. Due to low country coverage before the 1990s, wealth inequality data are not included in this analysis. Because male and female length-of-life inequalities are very highly correlated and because gender equality is already included, only male length-of-life inequality is included (due to its higher coverage).

The weights of the composite indicators implied by the latent variable models used to construct these two composites are shown in Table 11.1. Most variables contribute as expected, the one exception being biodiversity; for a discussion, see Rijpma (2014[15]). The contribution of extreme poverty and working hours is similar in magnitude to that of the core variables from the previous volume, with the exception of average years of education and life expectancy, which make a larger contribution. The inequality measures all enter in the expected direction (for gender equality, a higher score implies less equality, so it is expected to have the opposite sign). This means that high values in the composite indicator correspond to higher inequality. The model assigns a somewhat lower weight to income inequality than to the other inequality indicators.

Before discussing developments in these composite indicators, it is useful to compare the composite indicator of country averages to the one included in Rijpma (2014[15]). The overall pattern of the two indicators is very similar. For one, the factor loadings in Panel A are largely unchanged from Rijpma (2014[15]). Moreover, Annex Figure 11.A.1 shows that the regional developments implied by each composite indicator are typically very close, differences being visible only in the early period, but still well within the wide confidence intervals for this period. Larger differences exist at the country level but are typically minor.

A number of factors explain this similarity. First, the country coverage for the new indicators is usually lower than for the old set, and the model does not favour variables with a high degree of “missingness”. These variables would mostly add to the uncertainty of the estimates. Furthermore, each new variable added to a composite indicator has, by definition, a lower impact than the variables previously included. In other words, since we started with nine variables, adding two additional ones was never likely to result in large changes to the composite indicator, unless they have a very large weight.

Above all, though, Panel A suggests that the indicators used in Zanden et al. (2014[8]) were also the most important ones, and the ones for which data are available in sufficient quantity and quality. Because of this, the latent variable model presented in the previous volume was already capable of capturing the shared information from the old set of variables, while the new variables added in the present volume mostly confirm this pattern. Arguably, this means that some of the variables included in the composite indicator might be redundant, a point further discussed below.

Given the above, the story told in van Zanden et al. (2014[8]) still holds when looking at the composite indicator for country averages shown in Figure 11.9. While there is sometimes considerable uncertainty in the regional estimates, overall, the world has seen great progress, with all regions showing considerable increases in the composite well-being indicator over the 200-year period. Progress is also greater and more equally distributed than that shown by GDP per capita. In terms of the composite indicator, there are no regions or countries that are worse off today than the best-performing countries were in 1820 (Table 11.2). No country in the year 2000 does worse than the United States in 1820. If we were to make this comparison in terms of GDP per capita, some countries would fare worse (see Chapter 2).

However, strong and widespread gains do not mean that these gains proceeded at the same pace across regions. Western Europe and the Western Offshoots performed better than the other regions throughout the period. By the middle of the 19th century, they already had the highest scores on the composite well-being indicator. Both regions kept their lead over other regions throughout the period. That said, clear cases of convergence also occurred, as East Asia, Eastern Europe and the former Soviet Union, the Middle East and North Africa, and Latin America and the Caribbean began to catch up with the two leading regions. This catching-up process began roughly in the middle of the 20th century. Finally, while there has been substantial progress, South and Southeast Asia and, above all, sub-Saharan Africa have not converged to the levels achieved in the leading regions.

Figure 11.10 shows the regional development of the composite indicator of well-being inequalities. The composite inequality indicator is measured with a reasonable degree of certainty, though for some regions, East Asia and the Western Offshoots in particular, it is hard to make comparisons over time and with other regions with substantial certainty.

The overall trend is, again, one of improvement, with inequality declining in all world regions. No substantial trends towards higher inequality can be observed when looking at multiple inequality indicators at once. Western Europe and the Western Offshoots perform best throughout the period. It is noticeable that our composite measure of inequality for Western Europe falls below the levels of the Western Offshoots around 1950. The composite indicator of multiple inequality indicators shows that this reversal had already occurred in the early 20th century.

The composite indicator of well-being inequality also displays regional convergence. Differences between regions in 2000 were substantially smaller than in 1900. Eastern Europe and the former Soviet Union in particular have become far less unequal. While sub-Saharan Africa and the Middle East and North Africa do converge with the best-performing regions, they remain the two regions with higher inequality throughout the period. For the Middle East and North Africa, this is a striking development, as the region has made much more progress, both in terms of per capita GDP and of the composite indicator of average well-being levels. Its poor performance in terms of length-of-life inequality, gender equality, and to a lesser extent income inequality, are behind this pattern.

Table 11.3 shows a trend towards declining inequality in most countries. In 1870, most of the countries shown were very unequal, with those featuring positive values of the composite indicator being in the top 50% of countries in terms of inequality. According to the composite indicator, highly developed countries like Sweden, Germany and the United Kingdom are today’s low inequality countries.

Generally, the How Was Life? volumes have told a story of long-term progress. Looking at the many long-term well-being indicators produced by historical research generally shows that life has been improving throughout the world, on a range of indicators. The new and updated indicators presented in this volume have by and large confirmed this picture. Extreme poverty and working hours have declined, and the improved estimates of GDP per capita have not overturned the general conclusion drawn in the previous volume.

While large parts of the past two centuries, and the second half of the 20th century in particular, conform to this general improvement, this is not the whole story. Progress on some indicators has stalled in recent decades, the rise in working hours being one of the examples presented here. Moreover, these gains are not distributed evenly over the world, with some regions, above all Western Europe and the Western Offshoots, attaining higher well-being levels throughout the 1820-2020 period, and other regions lagging far behind.

The addition of within-country inequality indicators to the picture of well-being sketched in this volume strengthens the above observations. While within-country income inequality decreased since the 1930s in many world regions, it has increased more recently. The other inequality indicators analysed in this volume display a global trend towards more equality.

Suggestions for future research focus mostly on the inequality dimension. For one, more and better data on inequality in dimensions other than income are clearly needed, especially for countries outside Western Europe and the Western Offshoots. Currently, long-run series on inequality other than income have poorer country coverage, especially in the case of length-of-life inequality.

Another issue deserving attention is how to measure inequality when there are “soft limits” to the distribution, as in the case for length of life and years of education. These limits mean that progress in the overall level of an indicator almost automatically reduces inequality, as more and more people approach the limits of educational attainment or longevity. Further research on how to measure inequality in these circumstances may change the conclusions reached in this chapter.

To summarise trends in inequality in well-being in a composite indicator, this chapter has taken a highly practical approach. All inequality indicators were linearly combined, which is not an entirely satisfactory approach. Working with a more limited set of variables, and considering alternative aggregation functions, e.g. (Jones and Klenow, 2016[25]; UNDP, 2010[26]; Atkinson, 1983[27]), could open up new avenues here. This effort should be combined with an investigation of whether any variables in the composite indicators are redundant. This would ease the data requirements of alternative aggregation procedures.

Finally, one important reason for looking at within-country inequality is that country-level averages tell an incomplete story of well-being. Well-being is experienced by individuals. To address this, historical micro-level data, preferably in multiple dimensions for each individual, are needed. This will require substantial data collection efforts and new methods of analysis.

References

[27] Atkinson, A. (1983), The Economics of Inequality, Clarendon Press, Oxford.

[7] Bengtsson, T. and F. van Poppel (2011), “Socioeconomic inequalities in death from past to present: An introduction”, Explorations in Economic History, Socioeconomic Inequalities in Death, Vol. 48/3, pp. 343-56, https://doi.org/10.1016/j.eeh.2011.05.004.

[2] Boarini, R. and M. Mira D’Ercole (2013), “Going beyond GDP: An OECD perspective”, Fiscal Studies, Vol. 34/3, pp. 289-314, https://doi.org/10.1111/j.1475-5890.2013.12007.x.

[4] Bolt, J. and J. van Zanden (2014), “The Maddison project: Collaborative research on historical mational accounts”, The Economic History Review, Vol. 67/3, pp. 627-51, https://doi.org/10.1111/1468-0289.12032.

[16] Chakravarty, S. (2003), “A generalized human development index”, Review of Development Economics, Vol. 7/1, pp. 99-114, https://doi.org/10.1111/1467-9361.00178.

[10] Chen, S. and M. Ravallion (2010), “The developing world is poorer than we thought, but no less successful in the fight against poverty”, The Quarterly Journal of Economics, Vol. 125/4, pp. 1577-1625, https://doi.org/10.1162/qjec.2010.125.4.1577.

[20] Dollar, D. and A. Kraay (2002), “Growth is good for the poor”, Journal of Economic Growth, Vol. 7/3, pp. 195-225, https://doi.org/10.1023/A:1020139631000.

[6] Feinstein, C. (1998), “Pessimism perpetuated: Real wages and the standard of living in Britain during and after the industrial revolution”, Journal of Economic History, Vol. 58/3, pp. 625-58, https://doi.org/10.1017/S0022050700021100.

[19] Høyland, B., K. Moene and F. Willumsen (2012), “The tyranny of international index rankings”, Journal of Development Economics, Vol. 97/1, pp. 1-14, https://doi.org/10.1016/j.jdeveco.2011.01.007.

[18] Jackman, S. (2009), Bayesian Analysis for the Social Sciences. Wiley Series in Probability and Statistics, Wiley, Chichester.

[25] Jones, C. and P. Klenow (2016), “Beyond GDP? Welfare across countries and time”, American Economic Review, Vol. 106/9, pp. 2426-57, https://doi.org/10.1257/aer.20110236.

[3] Maddison, A. (2001), The World Economy: A Millennial Perspective, Development Centre Studies, OECD Publishing, Paris, https://dx.doi.org/10.1787/9789264189980-en.

[9] OECD (2015), In It Together: Why Less Inequality Benefits All, OECD Publishing, Paris, https://dx.doi.org/10.1787/9789264235120-en.

[23] Piketty, T. (2014), Capital in the Twenty-First Century, The Belknap Press of Harvard University Press, Cambridge Massachusetts.

[12] Ravallion, M. (2020), “On measuring global poverty”, Annual Review of Economics, Vol. 12, pp. 167-188, https://doi.org/10.1146/annurev-economics-081919-022924.

[14] Ravallion, M. (2012), “Mashup indices of development”, The World Bank Research Observer, Vol. 27/1, pp. 1-32, https://doi.org/10.1093/wbro/lkr009.

[13] Ravallion, M. (2012), “Troubling tradeoffs in the human development index”, Journal of Development Economics, Vol. 99/2, pp. 201-9, https://doi.org/10.1016/j.jdeveco.2012.01.003.

[11] Ravallion, M., G. Datt and D. van de Walle (1991), “Quantifying absolute poverty in the developing world”, Review of Income and Wealth, Vol. 37/4, pp. 345-61, https://doi.org/10.1111/j.1475-4991.1991.tb00378.

[17] Rijpma, A. (2016), “What Can’t Money Buy? Wellbeing and GDP Since 1820”, Working Paper, Centre for Global Economic History, No. 0078, Utrecht University, https://ideas.repec.org/p/ucg/wpaper/0078.html.

[15] Rijpma, A. (2014), “A composite view of well-being since 1820”, in How Was Life?: Global Well-being since 1820, OECD Publishing, Paris, https://dx.doi.org/10.1787/9789264214262-17-en.

[24] Scheidel, W. (2017), The Great Leveler: Violence and the History of Inequality from the Stone Age to the Twenty-First Century. The Princeton Economic History of the Western World, Princeton University Press, New Jersey.

[5] Steckel, R. and R. Floud (1997), Health and Welfare During Industrialization. A National Bureau of Economic Research Project Report, The University of Chicago Press, Chicago, IL.

[1] Stiglitz, J., A. Sen and J. Fitoussi (2009), Report by the Commission on the Measurement of Economic Performance and Social Progress, http://www.stiglitz-sen-fitoussi.fr/documents/rapport_anglais.pdf.

[26] UNDP (2010), Human Development Report 2010: The Real Wealth of Nations: Pathways to Human Development, United Nations Development Programme, New York, http://www.hdr.undp.org/sites/default/files/reports/270/hdr_2010_en_complete_reprint.pdf (accessed on 2 February 2021).

[21] van Leeuwen, B. and J. van Leeuwen-Li (2014), “Education since 1820”, in How Was Life?: Global Well-being since 1820, OECD Publishing, Paris, https://dx.doi.org/10.1787/9789264214262-9-en.

[8] van Zanden, J. et al. (2014), How Was Life? Global Well-Being Since 1820, Paris: OECD Publishing, http://doi.org/10.1787/9789264214262-en.

[22] Zijdeman, R. and F. Ribeiro de Silva (2014), “Life expectancy since 1820”, in How Was Life?: Global Well-being since 1820, OECD Publishing, Paris, https://dx.doi.org/10.1787/9789264214262-10-en.

Notes

← 1. Of the indicators presented here, per capita GDP and wealth are log-linearly interpolated; all other indicators are imputed with linear interpolation. The composite indicator has its own imputation procedure, which is explained in Rijpma (2016[17]).

← 2. 75% of the population is covered on average, with lower percentages further back in time. The sub-Saharan Africa and Middle East and North Africa regions have lower coverage.

Metadata, Legal and Rights

This document, as well as any data and map included herein, are without prejudice to the status of or sovereignty over any territory, to the delimitation of international frontiers and boundaries and to the name of any territory, city or area. Extracts from publications may be subject to additional disclaimers, which are set out in the complete version of the publication, available at the link provided.

© OECD 2021

The use of this work, whether digital or print, is governed by the Terms and Conditions to be found at http://www.oecd.org/termsandconditions.