The Three Eras of Global Inequality, 1820-2020 (PDF)

STONE CENTER ON SOCIO-ECONOMIC INEQUALITY WORKING PAPER SERIES No. 59 The three eras of global inequality, 1820-2020 with the focus on the past thirty years Branko Milanovic December 2022 REVISED September 2023 September 2023 Third draft The three eras of global inequality, 1820-2020 with the focus...

STONE CENTER ON SOCIO-ECONOMIC INEQUALITY WORKING PAPER SERIES No. 59 The three eras of global inequality, 1820-2020 with the focus on the past thirty years Branko Milanovic December 2022 REVISED September 2023 September 2023 Third draft The three eras of global inequality, 1820-2020 with the focus on the past thirty years Branko Milanovic 1 ABSTRACT The paper reestimates global inequality between 1820 and 1980, reappraises the results up to 2013, and presents new inequality estimates for 2018. It shows that historically, global inequality has followed three eras: the first, from 1820 until 1950, characterized by rising income differences both between and within countries; the second, from 1950 to the last decade of the 20th century, with very high global and between-country inequality; and the current one of decreasing inequality thanks to the rise of Asian incomes, and especially so Chinese. The present era has seen the emergence of the global “median” class and the greatest reshuffling in income positions between the West and China since the Industrial Revolution. Absolute income differences in the past 30 years have however increased, and the income gap between the “core” and the poor “periphery” (if China is excluded) remains large: the ratio between median income in the core and periphery in 2018 exceeds 8 to 1 in PPP terms, and 22 to 1 in nominal dollars. At the current rate of periphery’s convergence, it would take three centuries for the poor countries to have one-half of its population at, or above, the median income of the poorest core country. The evolution of global inequality in the future will much more depend on what happens to the growth rates and inequality in India and large African countries than on China. Keywords: Income distribution, globalization, convergence JEL codes: D30, D31 Total word count: Around 21,400. Graduate Center City University of New York, and the Stone Center on Socio-economic Inequality. I am grateful for comments to the participants at seminars held in 2022 where the earlier drafts of the paper were presented, originally at the European Economic History Conference in Groningen, and then University of California Riverside, Princeton University, ESADE Business School in Barcelona, Center for Economic Research and Graduate Education (CERGE) anniversary conference in New York, University of Oxford, Peking University, Center for International Relations and Sustainable Development in Belgrade, Luxembourg Income Study Summer School, Moscow students’ association “Obshchee Delo” and London School of Economics. I am grateful to the editor, Jason Hickel, and four anonymous reviewers for excellent and often very probing comments. The paper has also benefitted from detailed comments by François Bourguignon, Roy van der Weide, Nishant Yonzan and Laurence Chandy. 1 1 1. The three eras defined: a bird’s eye view of history Studying empirically global inequality—defined as inequality in real incomes between citizens of the world—helps us understand better the big changes that have occurred over the past two centuries. Figure 1 shows the level of global inequality, measured by the Gini coefficient, from 1820 to 2018. Even if there are methodological differences (for the period up to 1980, the mean per capita country income is approximated by country’s GDP per capita, and from 1988 by mean income obtained from household surveys), and quality differences (information on within-national inequalities, especially in the nineteenth century is much weaker than in the recent period), the extent of the overall change leaves very little doubt as to the dominant trends. 2 3 They sharply delineate the three eras of global development. The first era covers the period 1820-1950, and is characterized by the steadily rising global inequality. Around the time of the Industrial Revolution, global inequality was estimated at 50 Gini points.4 Compared to the inequality levels that were recorded since this was a rather modest inequality for the world as a whole. It is approximately equal to the level of inequality that today we find in very unequal countries like Brazil or Colombia. Throughout the nineteenth As discussed in Annex I, the underlying data for the years 1820-1980 come from the seminal work of François Bourguignon and Christian Morrisson (2002). Their numbers are revised and updated by using the more recent estimates of GDPs per capita and population from the Maddison project (2017 version). The new data are based on 2011 international prices rather than on 1990 prices used by Bourguignon and Morrisson. This change makes the 1820-1980 results comparable to those of the later years where I also use 2011 PPPs. The within-national inequalities for 1820-1980 are from Bourguignon and Morrisson (2002). The data for the period after 1980 are from the nationally representative household surveys, including between 111 and 136 countries and covering in all but two cases more than 90 percent of the world GDP and world population. They are based on the work by Lakner and Milanovic (2016), Milanovic (2021) and recent unpublished compilations and calculations. 2 A further methodological difference also discussed in Annex I is that the lack of nineteenth century data for many countries forced Bourguignon and Morrrisson to create country grouping (consisting in the extreme cases of more than 30 countries) and to treat global inequality as de facto inequality among the groupings (33 in total) rather than among the countries as is done from 1988 onwards. Grouping countries imparts a downward bias to global inequality and might explain its somewhat lower level of at the time when the two series are “combined” (see Figure 1). 3 There are three papers on the long-term evolution of global income distribution written after the initial Bourguignon and Morrisson paper. They are Milanovic (2011), van Zanden, Baten, Foldvari and van Leeuwen (2014), and Chancel and Piketty (2021). They each try to improve on the underlying historical income distributions, but for historical mean country incomes they all take Maddison Project data (various versions). Milanovic (2011) uses social tables for the countries for which they are available, van Zanden et al. (2014) use the GDP-wage ratios and a variety of non-income measures like the distribution of rents paid and distribution of heights, and Chancel and Piketty (2021) use, when available, fiscal data. 4 2 century, however, global inequality constantly grew reaching 62 Gini points on the eve of World War I. In the inter-war period, inequality slightly decreased, only to further go up mostly due to the effects of World War II that, in income terms, benefited the already rich countries like the United States and further impoverished China and India. Figure 1. Estimated global income inequality 1820-2018 80 75 WW2 and US dominance 70 WW1 and the Great Depression Gini index 65 60 55 The rise of Asia Industrial Revolution and the rise of the West 50 45 40 35 30 1800 1850 1900 1950 2000 2050 Year Note: 1820-1980 based on the revised Bourguignon and Morrison (2002) data series; 1988-2008 based on Lakner and Milanovic (2016); 2018-13 based on Milanovic (2022); 2018, unpublished results. For fuller explanation see Annex I. The Figure shows the break in the data sources between the series where country means are GDP per capita (up to 1980 and following the work by Bourguignon and Morrisson) and the series where country means are derived from household surveys (after 1988). The second era extends over the second half of the twentieth century. It is a period of very high global inequality maintained at a level between 67 and 70 Gini points. 3 The third era begins around the turn of the twenty-first century and extends until 2018, the last year for which we have the data. Global inequality is decreasing throughout that period, going down from 70 Gini points to 60 Gini points. The decrease, having occurred over less than 20 years, is very sharp. It is shown in Figure 1 by the strong downward slope of the line which is steeper than the (reverse) upward rising slope during the nineteenth century. The previous picture sharpens when we look in Figure 2 at the two components of global inequality, namely (i) the between-country inequality which represents inequality between (population-weighted) mean country incomes (called Concept 2 inequality), 5 and (ii) the within-country inequality, which is a population-weighted summation of all national inequalities. (We use the Theil (0) index or mean log deviation here because it is, unlike Gini, exactly decomposable. The overall picture does not change though, as can be seen from Tables A1 and A2 in Annex I.) The between-country component was rising throughout the nineteenth century, plateauing over the second half of the twentieth century, and beginning its decline at the close of the twentieth century, continuing to 2018-20. Its movement is similar, but more dramatic than the movement of global inequality. The between-inequality is, in effect, the main driving force behind the changes in global inequality, and thus in the recent period behind its decline. The three eras represent three different periods of international development. The first era is characterized by income divergence between, on the one hand, the industrializing countries of North-Western Europe, North America, and Japan, and on the other hand China, India and Africa with stagnant or even declining per capita incomes. This is the period that is in economic history known as the Great Divergence. It lasted throughout the nineteenth century. The great economic divergence had its corollaries in the great divergence in political and For the ease of exposition, it is useful to differentiate between Concept 1 international inequality which is the (unweighted) inequality in country mean incomes (often studied under the topic of country convergence or divergence), Concept 2 international inequality which is inequality in population-weighted country mean incomes, and Concept 3 or global inequality which is inequality between world citizens. In the latter two inequalities, the units of observation are individuals; the difference is that in Concept 2 inequality individuals enter the calculation with the mean income of their country, and in Concept 3 with their actual incomes. Concept 2 inequality is thus a component of Concept 3 inequality. The concepts were first defined in Milanovic (2005). 5 4 military power between the rising states and those that stagnated or declined. It coincided with the European conquest of most of Africa, colonization of India, semi-colonization of China, and Japanese colonization of Korea and Taiwan. As Braudel (1979, p. 535) writes, “History of the world between about 1400 and…1950 is one of an ancient parity [between the West and Asia] collapsing under the weight of multisectoral distortions…Compared with this predominant trend, everything else is secondary.” The between-country inequality remained at approximately the same—very high—level between 1950 and the turn of the twenty-first century. The increasing number of countries adds to the between country inequality compared to what it was before, but that “bias” –if one can indeed speak of the bias in this case—is not a dominant factor in what we observe. The second half of the twentieth century is the era of the Three Worlds, relatively well delineated in terms of their income levels and geographical spread: the First World of advanced capitalist countries, the Second World of less rich East European socialist countries and the USSR, and the Third World of poorer, and in many cases emerging from colonization, countries of Asia and Africa. To the latter are often added Latin American countries, even if they were, on average, richer and politically independent since the early nineteenth century. The third era is, as we have seen, the era of the rapidly declining between country inequality on the heels of the rising Asian mean incomes. 5 Figure 2. Decomposition of global income inequality 1820-2018 The age of the Three Worlds and diminished class conflict 100 The rise of Asia and of internal cleavages 90 80 Theil index 70 The age of empires and class struggle 60 50 40 30 20 10 0 1820 1850 1870 1890 1910 1929 1950 1960 1970 1980 1993 1998 2003 2011 2013 2018 Maximum between Theil Within Theil Note: 1820-1980 based on the revised Bourguignon and Morrison (2002) data series; 1993-2008 based on Lakner and Milanovic (2016); 2008-13 based on Milanovic (2022); 2018, unpublished results. For more detail, see Annex I. The vertical line indicates that the data on the left are based on country mean incomes approximated by GDP per capita, and the data on the right on the means derived from household surveys. When we look at the second component, the within national inequality, we notice its increase throughout the nineteenth century. The data are shown in Table A2, Annex I. 6 Although our data on within-national inequality in the nineteenth century coming from the estimates contained in Bourguignon and Morrisson, are not fully reliable, and in some cases are probably not better than educated guesses, the independently obtained information on withinThe residual inequality in the Gini index (see Table A2, Annex I) is decreasing, but this is due to the fact that the residual inequality in Gini includes both the proper effect of the within-inequality and the term reflecting overlap between the distributions. As the overlap term has gone down, due to the divergence in mean country incomes, the residual inequality has decreased. The decrease in the overlap component can be interpreted as the rising difference in incomes between citizens of different countries (see Yitzhaki 1994, and Milanovic and Yitzhaki 2002) which is, of course, consistent with the observed mean country income divergence. The within-country component of Theil(0) increases between 1820 and 1890 from 35 to 37 Theil points (Table A2, Annex I). 6 6 national inequalities in some key countries like the United Kingdom, France, Germany, the United States leaves little doubt as to the presumption of generally rising within-national inequalities during the period, at least in the economically advancing part of the world. 7 For other countries (many of which were colonies or had weak administrative capacity), we know much less about the evolution of inequality. Putting the two nineteenth-century developments together, namely, the divergence between incomes of nations, and often the divergence in individual or class incomes within nations, means that the increase in global inequality was driven by both forces of divergence. Things changed during most of the twentieth century. The level of global inequality was, as we have seen, extremely high but there was no clear upward or downward trend. Betweennational inequalities had remained high, thus ushering in the “tripartite” world. Withinnational inequalities shrank in large countries, such as the United States, Japan, Germany, UK and France, due to much more progressive policies regarding taxation and social transfers. Similarly, inequality in countries that experienced communist revolutions (among which, most importantly the Soviet Union and China) decreased as well. The second era was thus characterized by a combination of very high levels of between-country inequality and diminished national inequalities. It is with the rise of China that begins the third era of global inequality. The rise of China was important because it was very swift, dramatic in terms of the acceleration in its rate of growth, and involved a large number of people varying between 1/4 and 1/5 of the world population. Furthermore, the starting point of China in terms of its mean income was very low which also contributed, when China began to grow, to a fast reduction in the between-country inequality. The convergence in incomes did not involve only China but at the same time, or a bit later, extended to the rest of Asia, and especially to India which both by the size of its population and relative poverty came to play the role similar to the one that previously For the UK, see Lindert (1988), Milanovic, Lindert and Williamson (2011), Allen (2016 and 2019), Alvaredo, Atkinson and Morelli (2016); for the US, Lindert and Williamson (2020); for Germany, Bartels, Kersting, and Wolf (2021); for France, Piketty (2001). 7 7 belonged to China (see more in Annex II). However, as of approximately 2018, China’s attainment of a relatively high income level means that its growth is no longer contributing as much to the reduction in global inequality, and may soon even add to it. 8 This point is discussed below. The third era was therefore in some ways a mirror image of the first: the rise of one part of the world and the relative, although not necessarily in real terms, income decline of another. Thus in terms of its effect on global inequality, it was the reverse of the first era. In the nineteenth century, the rise of the West meant growing between-country inequalities whereas in the more recent period, the rise of Asia means a catch-up of incomes and hence a declining global inequality. The first period was one of divergence, the current one is one of convergence. In terms of within-national inequalities, the third era is characterized by the rising inequalities in many large countries including the United States, China, Russia, India and even the welfare states of continental Europe. 9 Only Latin America has bucked the trend: its very high inequality has slightly decreased or remained at around the same level.10 Summarizing the features of the third era, we could say that it is a period of the rise of Asia and thus of the global (population-weighted) income convergence and also a period of widening domestic cleavages within nations. Unlike in the nineteenth century, the betweenand within- forces work against each other, but the former (mean income convergence) is much stronger. This explains the swift decrease in global inequality. In 2018, China’s mean per capita income from household surveys was 7,000 international dollars, which was slightly below the world average of 7,600 international dollars. Using GDP per capita (again in PPP terms), China’s level of $17,450 in 2020 was higher than the world average GDP per capita of $15,500. According to the World Bank classification, China is ranked as an upper middle-income country. 8 The post-1980s increase in within-national inequalities is extensively documented. For the changes in OECD countries see OECD reports (2011, 2015); for the change in China, Xie and Zhou (2014) and Zhuang and Shi (2016); for the change in India, Subramanian and Jayaraj (2014); for “transition” economies, Milanovic (2008). 9 For the first point of view, see Gasparini et al. (2011), and Gasparini and Cruces (2013). For a more skeptical view on Latin American inequality decline see De Rosa, Flores and Morgan (2022). 10 8 The preponderant role of between-country inequality in explaining the decrease of global inequality in the past seventy years, can also be seen from a comparison of global inequality calculated using household surveys (as discussed so far) and the between-inequality component where mean incomes from household surveys are replaced by GDPs per capita (see Figure 3a). Both are expressed as before in international (PPP) dollars. The advantage of GDPs per capita is that they are available annually and we can follow the evolution of Concept 2 inequality without interruption. 11 (The disadvantages of using GDP per capita as a measure of individual income or welfare are well known, see e.g. Anand and Segal (2008) but our objective here is just to illustrate the importance of country convergence.) The movements of the two lines in Figure 3a are very similar implying that mean country incomes drive the decrease in global inequality. The greater speed with which Concept 2 measure goes down, compared to global inequality, is partly due to the fact that the convergence in terms of GDPs per capita is stronger than in terms of household mean per capita incomes (Milanovic, 2005), and partly to the fact that within-national inequalities have tended to increase after 1980, thus offsetting to some extent the decrease in Concept 2 inequality. It is noticeable though that after 2018, Concept 2 inequality no longer decreases. In 2020, due to the economic impact of covid (e.g. strong negative growth of India and of many countries in Africa: 41 had negative real per capita growth in 2020), it registered even a small increase.12 One cannot exclude the future reversal of the decline in Concept 2 inequality, and thus in global inequality as well, as recently argued by Deaton (2021), World Bank (2022), and Kanbur et al. (2022). I will return to the issue in Section 4. Figure 3b shows the same two inequality concepts with incomes expressed at the market exchange rates (MER) rather than PPPs. This is the approach that we shall use more It is worth underlining that the use of GDP per capita for the estimation of the between-component does not yield the same values as the use of mean incomes from household surveys, which is in our context a preferred measure. The two however move closely. 11 The population-weighted per capita growth for the African continent was -3%. Calculated from the World Bank World Development Indicators. 12 9 systematically below in Section 3, but it is introduced in this part of the text for the ease of comparison between the results obtained with the more traditional method of PPPs and those obtained by the use of national currencies converted into US dollars at the market exchange rate. What is easily noticeable is that the use of MERs instead of PPPs always shows greater inequality: it works almost as a shift parameter. This is as expected because price levels are generally lower in poorer countries, and when their incomes are measured at international dollars, they are, in relation to the incomes of the richer countries, higher than when measured at MERs, and thus global inequality is less. For global inequality (Concept 3), where the means are estimated from household surveys, the difference between the two measures is about 10 Gini points. For Concept 2 inequality where the means are GDPs per capita, the difference is currently about 15 Gini points. The implication is that the means from household surveys are less dispersed than the means obtained from National Accounts, a fact to which we already alluded. This was noticed in the early work on household surveys (Milanovic 2002, Table 9, p. 66; Deaton 2005). It is due both to higher saving and investment rates in richer countries and to the underestimation of income from self-employment and own consumption by National Accounts. These incomes are much better covered, or imputed, by household surveys. In effect, early observations that African countries tend to have an unusually high survey mean-to-GDP ratio, led later to the upward revision of National Accounts in several African countries.13 There is another difference that may be noticed in the movements of Concept 2 inequality estimated at PPPs and MERs. In the 1990s, the inequality measured at market exchange rates increased while it went down when using PPPs. The reason is balance of payment difficulties, slow growth and the role of Structural Adjustment Policies in Africa, Latin America and Eastern Europe that led to severe real devaluations in many countries undergoing the “therapy”. This depressed their incomes in MERs, while it did not do, or did to much less extent, in PPPs. From the turn of the century, however, the movements of Concept 2 inequality, whether measured in PPPs or MERs, is clearly downward: by some 10 Gini points according to PPPs, and by even more according to MERs. Yet, it is worth noticing that the 13 See Kouame, Kilimili and Pirlea (2019). 10 convergence of Concept 2 inequality that, as noted above, has stalled when measured at PPPs since the onset of the pandemic, had also stalled in MER terms from even earlier, that is from around 2015. We are thus currently in the position that when “global” inequality is measured by the population-weighted differences in mean country incomes, the convergence no longer takes place. This in turn means that for the convergence of global inequality to continue going down, the population-weighted within-national inequalities need to decrease. 11 Figure 3a. Global inequality and population-weighted inequality between countries (in PPP dollars, 1950-2018/20) 90 80 Gini index 70 60 50 40 30 1950 1960 1970 1980 1990 Between countries 2000 2010 2020 Global Figure 3b. Global inequality and population-weighted inequality between countries (at market exchange rates, 1975-2021) 90 80 Gini index 70 60 50 40 30 1950 1970 1990 Betweeen countries 2010 Global Note: The between-country inequality calculated from GDPs per capita and population sizes available in World Bank Development Indicators database. 12 2. The past thirty years in the world: the greatest reshuffle of individual income positions since the Industrial Revolution 2.1. The emergence of the global median class The changes that have occurred after approximately 1980 have profoundly affected the global distribution of incomes, not only when measured by composite indexes like Gini or Theil as we have done so far, but even when we look at the shape of the global income distribution. Danny Quah in the 1990s (see Quah, 1996) described the global income distribution as twinpeaked: the first, high, peak was that of very poor people, most of them from poor Asian countries; the second was a much lower peak at relatively high incomes and most people there were from the developed Western economies. The middle of the global income distribution was rather empty. This is the shape that we can observe on the global income distribution curve for 1988, shown in Figure 4a. The first peak occurs at around $PPP 600 per capita per year, the second peak at around $PPP 12,000. It is noticeable that (what one might call) the global median class is absent. The situation is markedly different in 2008. Not only has the overall curve shifted to the right implying a general increase in incomes, but the rightward shift was accompanied by the thickening in the middle of income distribution. There was simultaneously a significant change in income levels and in their distribution. The twin peaks have been replaced by a single peak, or the mode of global income distribution, at slightly above $PPP 1,000 per capita annually. The distribution however has remained skewed to the right, i.e. it has remained strongly asymmetric (even in log terms). The rightward movement has continued and even accelerated after 2008, so that by 2018 the global income distribution has acquired an almost log-normal shape that is characteristic of income distributions in individual countries.14 The evolution toward a The log-normal shape is consistent with very different levels of inequality. Both Gini and Theil are linked to the log-normal distribution through a single parameter, s, the standard deviation of log of incomes. The formula for 14 𝑠𝑠 𝑠𝑠 2 Gini is 2𝑁𝑁 − 1 and the formula for Theil is. The empirical s in 2018 is 0.51, and replacing it in the Gini 2 √2 formula gives the value of 0.64. The empirical Gini is 0.6, the difference being due to the fact that the empirical distribution does not follow perfectly the theoretical log-normal distribution. 13 symmetrical distribution can be also observed if we look at the measures of skewness: in 1988, it was 0.73; in 2008, 0.66; and in 2018, only 0.14. 15 (Skewness of 0 would imply a perfect bellshaped distribution in log incomes.) The new mode of the distribution is around $PPP 2,300 per capita annually, and within the range between the median and +/- one standard deviation (all in log terms) are concentrated about two-thirds of the world population. 16 It should be noted however that the global median, and what may be called “the global median class”, is much poorer than what is conventionally considered the middle class in advanced Western economies. The global median in 2018 is about $PPP 3,600 per capita, whereas the global median for the countries of Western Europe, North America and Oceania (WENAO) is more than five times higher ($PPP 18,500). The person having the advanced countries’ median income is placed between 90th and 91st global percentile. While a person with such income may be considered “middle class” in the Western sense of the term, it is obvious that globally speaking such a person is very highly placed, and may rather be considered a part of the global upper class. 17 The tendencies just described are present when the global income distribution is considered at market exchange rates (Figure 4b). Again, between 2008 and 2018, the middle part of the income distribution that runs from approximately $1,000 to $10,000 has become thicker, and the two peaks have been replaced by one. The 2018 peak, however is lower at MERs than at PPPs (it is around $700 at market exchange rates) due to the fact that people who are there are almost entirely from poor countries whose price levels are less than in the numeraire country (United States). For the same reason the global “median” class is poorer 𝑚𝑚 ∑(𝑦𝑦−𝑦𝑦 )3 The measure of skewness shown here is the standard one given in Stata: it is equal to 33 where 𝑚𝑚3 = 𝑠𝑠 𝑛𝑛 the third moment of the distribution and s is the standard deviation. y is log of income, 𝑦𝑦 =mean log income. 15 is 16 In absolute dollar terms, the range is rather wide: from $PPP 1,078 to $PPP 11,720. So one should be careful in interpreting it. It may be perhaps more useful to note that about one-half of the world population has incomes that are between about $PPP 1,600 and $PPP 8,600. For the discussion of absolute incomes see Section 3 below. The “thickening” in the middle of the global income distribution is often misinterpreted to mean that a global middle class (with income levels approaching Western middle classes) is emerging. While this might happen in the future, it is clear that currently the global middle, or more accurately named the global “median” class, is much poorer than the middle class in the rich countries. That confusion is addressed, among others, by Knauss (2019) who also presents a sociological analysis of the groups that have crossed the absolute poverty threshold. 17 14 when assessed at MERs than at PPPs, and the gap between the “median class” and the top of income distribution is greater. 15 2008 1988 2018 50000 10000 3000 300 1000 0.2.4 density.6.8 1 Figure 4a. Global income distribution in 1988, 2008 and 2018 (income in PPP dollars) log of annual PPP real income.6 Figure 4b. Global income distributions in 2008 and 2018 (income at MERs) density.4 2018 300000 100000 50000 10000 3000 1000 300 0.2 2008 log of annual income at market exchange rates Note: See Annex I. Both the coverage of the world population and of the world GDP is above 90 percent. 16 2.2 Global growth incidence curves 1988-2018: “winners” and “losers” The thirty-year period that we consider here was not however uniform. During the years of “high globalization” from around the fall of the Berlin Wall to the global financial crisis, the changes took a particular shape that is associated with the so-called “elephant curve” (Lakner and Milanovic, 2016). The distinctive feature of the global growth incidence curve for the period 1988-2008, shown in Figure 5a, was the relatively high real growth among the middle of income distribution (point A), where were located mostly Asian populations, and among the global top five percent (point C), and, on the other hand, a very sluggish, or almost zero real growth (point B) among the populations around the 80th and 90th percentile who were mostly from the middle or lower-middle classes from the advanced economies. The causes and the effects of this particular pattern of growth have been much discussed, including in an entire book by Anthea Roberts and Nicolas Lamp (2021) that provides a variety of political and economic interpretations of these developments.18 One can focus on either international causes of this development, underscored by the gap between points A and B, and argue that a particular type of globalization characterized by outsourcing and free movement of capital was its main cause. Alternatively one may focus on the “domestic” part of the development, namely the gap between points B and C, and see the slowdown in income growth of advanced countries’ middle classes as caused by the domestic forces of lower taxation of high incomes, skill-based technological change, or unusually high returns to capital. The results are, of course, consistent with both of these main explanations, and it is quite likely that both international and domestic factors played a role. This particular pattern of growth however has not continued in the next ten-year period that runs from the end of the Global Financial Crisis to the outbreak of covid-19. What the most recent growth incidence curve shows is a marked deceleration in real income growth for the global top one percent. The results do not show significant improvement for the people around the 80-90th global percentile either. Very much in accord with the “elephant chart”, it illustrates the continued strong growth in the middle. Moreover, growth rates appear to have been Some of the conclusions of the so-called “elephant graph” were criticized by Adam Corlett from the Resolution Foundation (2016). The authors replied in Lakner and Milanovic (2016a and 2016b). 18 17 particularly high among the poorest percentiles which gives to the global growth incidence curve a very clear downward shape characteristic of pro-poor growth (the exact, rather complicated, reasons for that are discussed in the next section). The results for the earlier fiveyear period 2008-13 are discussed in Milanovic (2022) and look very similar to what is reported here for the entire decade. The next five year period simply continued in the same vein.19 Figure 5b presents the same data using the growth incidence curve expressed in absolute PPP amounts. As noticed before (Lakner and Milanovic, 2016; Milanovic 2021) such curves tend to show sharp increases at the top of the distribution, and are often upwardsloping throughout. The reason is that incomes of the top percentiles are extremely high (compared to the rest of the population) and thus even if GIC in relative terms is downwardsloping this still translates in higher absolute income gains for the rich. The absolute GIC for the most recent period 2008--2018 illustrates this very clearly. While the relative GIC (Figure 5a) is downward sloping, the absolute GIC (Figure 5b) is upward sloping (“hockey-stick” like) across the distribution: higher percentiles have gained more than lower percentiles, even if their relative position has deteriorated. There is however one important feature that persists in for period 2008-18 in both relative and absolute terms: the global top 1 percent has gained much less not only in relative terms but even in absolute terms than during the era of “high globalization”. As will be discussed below, these lower gains are the result of the Global Financial Crisis that has affected disproportionately incomes of the nationally rich in rich countries whose citizens comprise more than four-fifth of the people in the global top one percent. For an alternative global incidence curve for the period 2008-18 that uses Indian consumption (instead of income) data, see Annex II. The data used for 2008 are from Milanovic (2022). They are a more complete and updated version of the dataset used in the Lakner and Milanovic (2016) paper. The data used for 2018 are newly compiled and unpublished. For more detail see Annex III. 19 18 Figure 5a. Global growth incidence curves for 1988-2008 and 2008-2018 (in relative terms) Averge real annual growth rate (in %) 9.0 8.0 7.0 6.0 A 5.0 C 4.0 3.0 2.0 1.0 B 1988-2008 95 Top 1% Percentile of global income distribution 96-99 90 85 80 75 70 65 60 55 50 45 40 35 30 25 20 15 5 10 0.0 2008-18 Figure 5b. Global growth incidence curves for 1988-2008 and 2008-2018 (in absolute terms) Absolute gain (in real PPP dollars) 25000 20000 15000 10000 5000 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 96-99 Top 1% 0 Percentile of global income distribution 1988-08 2008-18 Note: The fractiles are anonymous (that is, the global fractiles in 2018 are composed of country/percentiles that “fall” into that fractile in 2018 and which are normally different from the country/percentiles that were in that global fractile in 2008).The same holds for the period 1988-2008. For the distinction between anonymous and quasi non-anonymous growth in this context, see Lakner and Milanovic (2016). All incomes are expressed in PPPs. 19 However when we try to ”dissect” the change and to look at “winners” and “losers” it is important to point out that the very fact that growth rates along the global income distribution differ means that there is a significant “churning” within the distribution. In other words, while it made sense to approximate the middle of the global income distribution with the Asian populations (and especially with the Chinese) in 1988, their very growth has moved many of them toward higher percentiles. Thus the middle of the global income distribution in 2018 is not composed of the same country/percentiles that populated it in 1988. It is therefore of particular importance to look at who were the most important “winners” in the period 2008-18, that is who populated the lower global percentiles, running approximately from the first to the fourth global ventile (i.e. the bottom 20% of the global income distribution). As can be seen, their average growth was around 7 percent per capita annually (Figure 5a). The story of what we observe at the bottom of the global income distribution between 2008 and 2018 is complicated because of the churning and the fact that the sample is not balanced (i.e., not exactly the same countries are included in 2008 and 2018). It revolves around three key developments. a. Changes on the bottom The first was the rise of China whose lower parts of rural income distribution have “vacated” the global bottom quintile in large numbers. While in 2008, some 200 million of the rural Chinese were in the global bottom quintile that number has fallen to only 75 million in 2018. China has thus left “open” some 125 million “slots” in the bottom quintile (or approximately 10% of the total number of the slots in that quintile). This was due to the very high growth rates among the poor Chinese rural deciles. Incomes of all (but the lowest) Chinese rural deciles more than doubled between 2008 and 2018, with the average annual growth rates being around 10 percent per capita. 20 The income of the lowest Chinese rural decile still increased significantly (by 80 percent over ten years) even if it failed to more than double as other rural deciles did. The source for Chinese calculations are micro data from the China Household Income Project (CHIP) from 2007 and 2018. They are also included in Luxembourg Income Study. 20 20 The vacant slots were filled by other countries’ populations who in 2008 were above the bottom quintile level but because of Chinese growth were pushed back. This is the second development. Among countries whose populations were thus “relegated” to the bottom quintile, countries in the Indian subcontinent were the most important. The number of people from Bangladesh and Pakistan who “fell” into the bottom global quintile was respectively 76 and 56 million. Combined with the fact that India (rural and urban together) has 560 million people in the global bottom quintile, the Indian Subcontinent now accounts for more than onehalf of all people in the poorest global quintile. 21 The bottom quintile has thus become much more “Indian” than before, not necessarily because of low income growth among these countries’ poor but because it fell short of Chinese growth. The share of the Nigerian poor in the bottom quintile increased significantly as well, going from a bit over 20 million to almost 70 million. It is also worth pointing to two large middle-income countries with very high inequality whose poor were in increasing numbers pushed into the bottom quintile because of Chinese growth. They are Brazil whose number of people in the bottom global quintile increased from 12 million in 2008 to 21 million in 2018, and South Africa with the increase from 12 million to 16 million. The third development is straightforward: high growth among the people (or more exactly, the country/percentiles) who were in the bottom quintile in 2008 in countries that had relatively many globally poor. Table 1 singles out several such countries accounting in total for 13 percent of the globally poor in 2008. It shows, for example, that in the case of Vietnam which in 2008 had 17 million of its population in the global bottom quintile, the average annual growth rate among these people was 12.5 percent per annum. 22 21 In 2018, the size of the bottom quintile is 1.35 billion people. The growth rate cannot be calculated across the same people as we lack longitudinal data, but can be calculated across the same country/percentiles. In other words, we keep the relevant country/percentiles at their 2008 global positions (in this case, in the bottom quintile) and calculate their income growth rate. This is called the quasi nonanonymous growth rate. 22 21 Table 1. Population, average incomes and income growth of some groups that were in the bottom global quintile in 2008 Population in the bottom quintile in 2008 (in million) Ethiopia Tanzania Ghana Kenya Philippines Vietnam Total (countries listed here) Total bottom quintile Percentage of country’s population in global bottom quintile in 2008 Income per capita per year of the population in the bottom quintile (in international dollars) 1/ 64 28 14 23 14 17 160 77 65 59 56 15 20 2008 430 406 389 440 534 510 441 1249 20 408 Average annual growth rate 2008-18 (percent per capita) a/ 2018 987 701 994 855 860 1659 979 8.7 5.6 9.8 6.9 4.9 12.5 8.3 727 7.4 Note: 2008 incomes are converted into 2018 amounts using countries Consumer Price Indexes and then expressed in 2018 international dollars using the 2018 PPP exchange rates. “Group” here refers to the country/percentiles that were in the global bottom quintile in 2008. a/ Calculated across country/percentiles that were in the bottom global quintile in 2008 (quasi nonanonymous growth rate) In conclusion, the recorded high income growth of the bottom quintile between 2008 and 2018 was thus due to three very different developments: the big rise in Chinese rural incomes that “expelled” most of the Chinese rural population out of the bottom quintile, the “fall” into the bottom quintile of people who before were above it (mostly from the Indian subcontinent and Nigeria), and the fast growth among some poor groups in poor countries. To the extent that the first two developments were dominant, the improved position of the globally poor is somewhat illusory: the people who are there are better-off than those who were there before, but it is because their own growth failed to keep pace with the growth rate of the Chinese poor. b. China vs. the West 22 The best way to appreciate the difference in the growth experience between China and the rich countries which is the single most important development during 2008-18 is to look at the real growth rates of disposable (post-tax and post-transfer) income across their distributions. They are shown in Figure 6a. China’s average annual per capita rate of growth was over 8 percent for almost the entire income distribution. The growth was broadly-speaking pro-poor as indicated by the very high growth (10 percent per annum) for the lower middle classes. There is also an interesting uptick for the top 1 percent. On the other hand, American and German average growth rates over the same period were—across the distribution—about 2 percent per capita annually. In the case of the United States, an even lower growth, due to the effects of the financial crisis, is noticeable for the top 1 percent. These diverse rates of growth had a strong effect on the shape of the global income distribution. We have noticed in Figure 5a that the growth rate at the global top in the period 2008-18 was fairly low (1.3 percent p.a.), in marked contrast with the period before the Global Financial Crisis. When one takes into account that: (i) 10 to 11 percent of the US population and 4-5 percent of the German population belong to the global top one percent, (ii) between the two of them, they account for two-thirds of all people in the global top 1 percent, and (iii) their rates of growth were around or below 2 percent per annum, the low growth rate of the global top 1 percent becomes more understandable. In other words, the low growth of the global top was due to the low growth of the upper parts of national income distributions in rich countries, whose citizens overwhelmingly populate the global top. When growth is calculated in real absolute income gains (expressed in 2018 PPP terms), uniformly higher Chinese percentage growth rates translate into almost the same real gains as in the Unites States and Germany (Figure 6b). The differences between the countries, for a given percentile in distribution, are relatively small: for example, at the median, gain in Germany was a bit under $PPP 4,000, just over $PPP 3,000 in China, and $PPP 2,700 in the United States. The shape of the absolute gain curves for all three countries is upward sloping, indicating that on average higher gains accrue to the people in higher income percentiles. 23 Figure 6a. Growth rates across income distributions for China, US and Germany, 2007/08 to 2018 Real average annual per capita growth (in %) 12.0 10.0 China 8.0 6.0 4.0 Germany 2.0 USA 0.0 0 20 40 60 80 100 Income percentile Real absolute per capita gain (in 2018 PPPs) Figure 6b. Absolute income gains by percentile for China, US and Germany, 2007/08 to 2018 (in 2018 PPP dollars) 30000 25000 20000 USA 15000 China 10000 Germany 5000 0 0 10 20 30 40 50 60 70 80 90 100 Income percentile Note: Calculated from LIS data for US and Germany for 2008 and 2018, and LIS data for China 2018 and CHIP for China 2007. Nominal amounts deflated by countries’ Consumer Price Indices to obtain real growth rates shown in Figure 6a. Real absolute income gains in domestic currency normalized by 2018 PPPs to obtain the gains in Figure 6b. 24 c. The great reshuffle in the middle When we translate the rise of Asian economies and China in particular in terms of individuals’ income levels, we observe what is probably the greatest reshuffle of individual incomes since the Industrial Revolution. The China effect, which is the most important part of the global positional reshuffling, is present, even if unequally, in almost all parts of the global income distribution. We have seen that it explains to some extent the significant increase of the lowest incomes and we look next at its effects on the middle of the global income distribution. Figure 7 (upper panel) shows the positions of Chinese urban deciles and Italian deciles in 1988 and 2018. The upward movement of the Chinese urban deciles that amounted to between 24 and 29 global percentiles (meaning that people in a given Chinese urban decile leapfrogged over one-fourth or more of the world population, or got ahead over approximately 1.5 billion people) 23 is not a surprise. But obviously as the Chinese deciles have gone up in the global distribution, other countries’ deciles, if relatively close to the upward-moving Chinese deciles, had to go down. This is illustrated on the example of Italy. The bottom Italian decile has slipped by 20 global percentiles, the second and the third by respectively six and two. The other deciles were not affected as they tend to be above the part of the global distribution where the Chinese influence has been the strongest. The changes observed in the case of Italy are not unique to that country. The German bottom decile has slipped from the 81st global percentile in 1993 to 75th percentile in 2018 (Figure 7, bottom panel). 24 The second lowest decile has, like in Italy, lost in its relative position. In France (not shown here), the bottom three deciles have lost out, with the lowest one again losing the most, going down from the 73rd global percentile in 1988 to 69th percentile in 2018. In the United States, the bottom decile has lost 7 percentiles, and the positional loss has spread to the bottom 40 percent of the population. The positional losses covered in some cases the entire countries’ income distributions. Countries that in 1988 were richer than China (but not as rich as the US, Italy etc.) and were by 2018 within the “reach” of Chinese upward movement had all their deciles register positional 23 One should keep in mind that the overall world population was 5.1 billion in 1988 and 7.6 billion in 2018. 24 I use 1993 data because the 1988 data cover only the former West Germany. 25 losses. This was for example the case of Croatia and Serbia. The lowest deciles in the two countries dropped by between 20 and 30 global percentiles. The loss, even if less dramatic, extended throughout their entire distributions and affected even the top income deciles. Poland that had a remarkably good economic performance over the thirty-year period considered here was affected negatively too. Its bottom 40 percent of population lost out, even if the top’s position improved: the Polish top decile moved from the 88th global percentile to the 95th. Very unequal middle income countries such as Brazil showed a different type of change. There, the losses were the largest for the middle income deciles. The rich remained outside of China’s “reach”. 1 2 3 4 5 6 7 National decile Global income percentile 75 80 85 90 95 100 1988 8 9 Global income percentile 60 70 80 90 100 China urban 10 2 3 4 5 6 7 National decile 1993 1 2 3 8 9 4 5 6 7 National decile 1988 Germany 1 Italy 2018 Global income percentile 60 70 80 90 100 Global income percentile 20 40 60 80 100 Figure 7. Positions of country/deciles in global income distribution in 1988 and 2018 10 2018 8 9 10 2018 United States 1 2 3 4 5 6 7 National decile 1988 8 9 10 2018 Note: The graph shows the global income position of each national income decile (running from the poorest, 1, to the richest, 10) in 1988 and 2018. The data for Germany are for 1993 and 2018. 26 d. Relative constancy on the top Unlike the middle of the global income distribution, the composition of the top has remained more stable. To assess this we look at the composition of country/percentiles that were in the global top 5 percent (top ventile) in 2008 and 2018. The global top 5 percent contains between 320 and 330 million people in both years and is more representative of the globally affluent than the more rarified global top 1 percent. The latter, by definition, includes only the very rich and their incomes are more likely to be affected by the underestimation of the returns to capital (see Yonzan et al, 2022). Thus, the top 5 percent are a more relevant group and their incomes are more accurately measured. When we take twelve countries with the largest absolute participations in the global top ventile in 2008 (they are given in the first twelve positions, going from left to right, in Figure 8), we note that there are only two newcomers in 2018 and, consequently, two dropouts. Ten out of twelve countries are the same with approximately the same number of people in both years. The United States is by far the most important: in both years, about 40% of the globally affluent are US citizens. The United Kingdom, Japan and Germany come next, with their positions slightly shifting between the two years, each participating by between 5 and 8 percent among the globally affluent. The newcomers among the top twelve countries are the urban Chinese whose participation has increased from 1.6 percent to 5 percent, and the Spanish citizens, going up from 1.6 to 1.9 percent. The two countries that have dropped out are Russia and Taiwan. We note a strong persistence in terms of both countries and the number of their citizens who are part of the globally affluent. WENAO (Western Europe, North America and Oceania) plus Japan have about 250 million of their citizens amongst the globally affluent.25 They account for 87 percent of the group in 2008, and 78 percent in 2018. Thus the global West’s role has remained preponderant. This includes all WENAO citizens, i.e., not only from among the WENAO countries that are among the top twelve countries by the number of people being part of the global top ventile. 25 27 Among Asian countries (exclusive of Japan), the Chinese urban population is the most important. The shares of the Indian and Indonesian urban populations in the top global ventile have also risen between 2008 and 2018: in the case of India from 1.3 to 1.5 percent; in the case of Indonesia, from 0.3 to 0.5 percent. Overall, the importance of the non-Japanese parts of Asia among the globally affluent has increased but—with the exception of the urban Chinese— their numbers are still modest. The same is true for the other parts of the world (Africa, Latin America, and Eastern Europe) which, with the exceptions of Brazil and Russia, never had a significant participation among the globally affluent. Figure 8. The composition of the global top 5 percent in 2008 and 2018 (in percent) 45.0 40.0 Share of global top ventile (in %) 35.0 30.0 25.0 20.0 15.0 10.0 5.0 0.0 USA GBR DEU JPN FRA CAN 2008 ITA AUS KOR BRA RUS TWN CHN-U ESP 2018 Note: Each bar shows the share of that country in the global top ventile. For example, about 40% of the population in the global top ventile comes from the United States. 28 3. Is convergence real? 3.1. Introducing absolute measures of inequality and market exchange rates So far we have looked mostly at convergence in real income (measured in PPP dollars) between world citizens, and indirectly between countries with their incomes again expressed in PPP dollars. The use of PPPs is normally combined with relative measures of inequality (Gini or Theil). The use of international dollars that are made comparable both across countries at a given point in time, and across time, is probably the best way to assess differences in living standards. PPP dollars in principle value the same good or service equally regardless of where they are produced and consumed. There are indeed significant problems of index numbers that remain (see Deaton 2005, 2010; Deaton and Aten 2017; also International Comparison Program (ICP) (worldbank.org) ) and PPPs have been used in several variants, most commonly GK (GearyKhamis) or EKS (Eltetö, Köves and Szulc) that differ by the weights they assign to different components. Despite these problems they are the best metric for comparisons of welfare over time and place. But they are not the only metric. There are at last two other different ways that one can look at issues of global inequality. First, instead of using international dollars, incomes can be measured by using US dollars, where domestic currency is converted in US dollars at the market exchange rate (as we have done in a few instances above). 26 Second, instead of relative measures of inequality, one could use absolute measures, that is instead of looking at relative distance in incomes, one can look at their absolute distances. This gives, as shown in Table 2, four different combinations. Obviously, this could be done in every other convertible currency (the euro or yen) but the use of US dollar is the most common. 26 29 Table 2. Different ways of assessing global inequality Conversion of local currency International dollars income into Measure of inequality Relative A Compares welfare Most common approach Absolute C Compares welfare Tends to increase with the rise in mean income US dollars at market exchange rate B Compares ability to purchase internationally-priced goods D Compares ability to purchase internationally-priced goods The most common approach (A) has also the tendency, compared to the alternatives, to show convergence. This happens for two reasons. Price levels rise as countries become richer. Most of it is due to the increase in prices of the non-tradable goods (and thus in wages). For example, price levels of rich countries are close to the price level of the United States (whose price level is used as a numeraire=1) while price levels of poorer countries are much lower. In 2018, the price level of Switzerland was 1.32 and the price level of India 0.29. When we use PPPs, we boost incomes of people living in poor countries—simply by acknowledging that their price levels are low. With MERs, however, income gaps between rich and poor countries are invariably greater than with PPPs. The second reason has to do with the use of relative measures vs. absolute. Assuming that we have “solved” the problem of what exchange rate to use, growth as such tends to produce rising absolute income differences. Consider the simplest example of two countries or individuals with incomes of 10 and 5 in period 1. The relative income gap is 2 to 1, the absolute gap is 5 units. Let the rich double their income, while the poor more than double theirs so that the new incomes are 20 and 12. The relative gap has decreased, the absolute increased. This is something that we normally find, especially over the long-term, when the average income grows. The absolute Gini coefficient is equal to the relative Gini (G), that we have used so far, multiplied by the mean income μ. According to the latest version of the Maddison Project, the mean income of the world in 1820 was just around $PPP 1,000 while in 2018 it was $PPP 7,600. The mean value has therefore increased by 7.6 times, and it is immediately obvious that the 30 changes in the absolute Gini, the product Gμ, will be dominated by the movements in μ (since the movement of Gini, as we have seen, varies comparatively little, between 0.5 and 0.7). The use of absolute measure may make sense over the short time-periods when the mean does not change much, or over the long-time periods if the economy is stagnant, but its informational value rapidly deteriorates otherwise 27 To give an intuitive example. US Gini in 1860 was 0.51, in 2018, it was 0.41. 28 It does make intuitive sense that US inequality today is less than it was at the time when 13% of its population was enslaved, and thus living at the subsistence. But, the average absolute real income distance (measured by the absolute Gini) has increased from $PPP 2,245 to $PPP 22,687, driven by the twelve-fold increase in real mean American income. In other words, when we take two randomly chosen Americans in 1860, their absolute income distance was just about one-tenth of the income distance between two randomly chosen Americans today, but the main reason for this is simply that incomes of people today are much greater (in real terms) than in the past. 29 However, absolute measures do contain an important informational content over short periods when we are interested in how the fruits of growth are distributed: do they accrue mostly to the rich, the middle, or the poor. This may be especially relevant for the period of globalization when we may want to study absolute gains from globalization. Indeed, as we have seen in Figure 5b, the absolute Atkinson and Brandolini (2004) calculate absolute, relative, and “intermediate” (linear combination of absolute and relative) measures of global and inter-national inequality. Yet a different way to look at global inequality is to contrast, at individual level, its importance to the importance of national inequalities. Milanovic and Roemer (2016) do this kind of analysis by varying the implicit weights that individuals give to global inequality (at one extreme they worry only about it) and to national inequalities (at the other extreme, individuals may care only about inequality in their own country). 27 The US Gini for 1860 is from Lindert and Williamson (2016, pp. 115-116); for 2018, it is calculated from Luxembourg Income Survey micro data based on US Current Population Survey. 28 There is an additional technical problem with the use of absolute measures for the calculation of global inequality. For every given benchmark year, we convert national currency incomes into international dollars using that year’s PPP exchange rate which may be derived from various ICP exercises (say, from the most recent 2017, or from the earlier one conducted in 2011). But for temporal comparison between two years we use US Consumer Price Index as a deflator since in both years US price level is used as a numeraire. This is not an entirely accurate deflation because the price basket on which US CPI is based is not the same as the implicit price basket on which global price level is based. 29 31 global growth incidence curve during the past thirty years was strongly upward sloping, indicating that the absolute gains were pro-rich even if the relative gains often were not. Next we shall look at measures B and C thus complementing what was done so far using mostly measure A (Table 3). 30 The calculation is conducted for the period 1988-2018 only because of lack of historical data on nominal dollar incomes (note that the Maddison data series has always been expressed only in PPPs; therefore the underlying income fractiles have also been expressed in PPP terms only).31 The results in Table 3 all come from the nationally representable household surveys, that is, the same surveys that were used to calculate the A measures of inequalities. The PPP incomes are expressed in 2011 PPPs, as in the rest of the paper, and in the case of MER measures, nominal dollar incomes are deflated to 2011 (real) US dollars using the US Consumer Price Index. 32 30 There is little point in looking at measure D which is driven by changes in nominal magnitudes. The absolute measures of inequality over the entire 1820-29018 period are neither particularly informative nor empirically easily done, or even feasible. This is due first, to the structure of the Bourguignon-Morrisson dataset where the world is divided into thirty-three blocks and some blocks contain the (identical) estimated income shares for 20 or 30 countries (see Annex 1). With Maddison’s GDP per capita data in international dollars, these estimated income shares can be translated into per capita incomes of various fractiles but Maddison data do not contain historical GDPs at market exchange rates. In addition, such “market” exchange rates cannot be found, or even be considered meaningful for many Asian and African countries in the 19th century. Furthermore, more recently, during most of the 20th century, a great number of ”developing” and many “developed” countries had a system of multiple exchange rates, so the very concept of what is a relevant market exchange rate is unclear. Take, for example, Egypt in 1955 or Turkey in the 1960s, or socialist countries until 1990: even if one had all exchange rates used in each country and each year, taking the mean of them would be hardly relevant, if the shares of transactions conducted at each were vastly different. Taking the black market rate as the market exchange rate (even if available) would grossly underestimate real incomes 31 The latter is an approximation because the global price index that theoretically underlies incomes expressed in nominal dollar terms does not necessarily follow the US consumer price index. The problem is similar to the one regarding the strictly speaking non-transitivity of PPPs of different years because of the divergence between the global and US-only price indexes. 32 32 Table 3. Absolute and relative, PPP and MER estimates of global inequality Inequality measures 1. Relative Gini with PPP exchange rates a/ 2. Relative Gini with market exchange rates 3. Absolute Gini with PPP exchange rates b/ 4. Absolute Gini with MERs c/ Global mean income 5. In 2011 PPPs 6. In 2011 US dollars 1988 1993 1998 2003 2008 2013 2018 69 69 68 69 66 62 60 78 80 79 80* 77 73 72 2796 2796 2933 3114 3497 3600 4114 2911 2972 2864 3134* 3483 3343 3477 4029 4047 4288 4532 5219 5836 6843 3760 3714 3620 3934* 4498 4599 4866 a/ From Table A2, Annex I. b/ Equal to line 1 times line 5 (divided by 100). c/ Equal to line 2 times line 6 (divided by 100. * = year 2005. Figure 9. Absolute Gini (for incomes at PPP dollars or at market exchange rate) 5000 4500 Dollars 4000 3500 3000 2500 2000 1988 1993 1998 2003 PPP 2008 2013 2018 MER Note: The figure shows one-half of the average income gap between two random individuals in the world when incomes are measured either in real international (PPP) dollars or dollars at the nominal market exchange rate deflated to 2011 US dollars. There are two important conclusions that can be drawn from these results. First, relative Gini with market exchange rates is about 10 points higher than the same Gini with PPPbased incomes (as shown also in Figures 3a and 3b). This result has been noted before 33 (Milanovic 2005; Anand and Segal 2008; and more recently Milanovic 2022; Annex 2). The MERbased relative Gini decreases more or less in step with PPP-based relative Gini. The issue is therefore not convergence, but the level of inequality at which convergence takes place. Second, the absolute Gini with PPP-based incomes increases throughout as the global mean income goes up. In the three decades considered here, the world real mean PPP income has increased by 70 percent while the mean absolute distance (or the absolute Gini) has increased by 47 percent. Thus, indeed, the absolute income distance between two random individuals in the world is much higher in 2018 than thirty years before. Figure 9 shows that absolute Ginis whether calculated for incomes in PPP terms or in real dollars move in about the same way but that since 2018, the PPP-based absolute Gini is substantially higher than the dollar-based absolute Gini. This is the reflection of the changed nature of the global economy characterized by higher growth of populous countries with price levels lower than the US numeraire. The share of such countries (mostly China and India) increased in the world economy, and thus the total global income, whether calculated from the surveys or National Accounts, is much higher in PPP than in dollar terms. For example, in 2018, the world GDP in PPP terms was 125 trillion, while in nominal dollars it was only 85 trillions. It is then not surprising that the average absolute income distance between two random individuals is about $PPP 8,200 vs. $7,000. 33 Figure 10 shows the global income distribution in 2008 and 2018 on an absolute scale. In both cases, the histogram contains fifty bins/intervals. The fact that the global distribution is very unequal means that when bins are of equal width (say, between x and x+a and then between x+a and x+2a and so forth) many higher-income bins will contain very few people while the bottom bins will be “crowded”. This is easily noted in Figure 10 where in 2008 close to two-thirds of the world population was “squeezed” in the lowest bin (up to $PPP 5,000 per capita annually).34 The share of people in the lowest bin decreased to under 60 percent in 2018, 33 Note that the absolute income difference between two random individual is equal to twice the absolute Gini. $PPP 5,000 per capita per year is about $PPP 14 per day which is at the low end of the national poverty lines in rich countries (see Ravallion, 2010, Figure 2, p. 35). 34 34 and correspondingly the second lowest bin gained in importance, from containing about 12 percent of the world population to almost 20 percent. So, most of the “action” between 2008 and 2018 was confined to the bottom two bins; the share of the world population that is in the bottom two bins was about three-quarters in both years. The same data are shown numerically in Table 4. Here the bin sizes vary in order to highlight better the composition of the distribution. It is thus easily noticeable that in both years, the bin with the largest share of the world population (almost one-third) was that with per capita incomes between $PPP 1,000 and $PPP 3,000. What has dramatically changed (as discussed in Section 2.2a above) was the decrease in the number of people in the two bottom groups (below $1,000). This has obvious implications or the calculation of global poverty: if we use a relatively low poverty line (of say, about $PPP 700 as currently used by the World Bank, or even up to $PPP 1,000), the decrease of global poverty is substantial. But if we use a poverty line that is higher (say, around $PPP 7,000 or $PPP 20 per day which is approximately a poverty line used in rich countries), then the decline in global poverty becomes much more modest. 35 Figure 10. Global income distribution in 2008 and 2018 (in absolute PPP amounts) Global income distribution 2018 0 0 20 20 Percent 40 Percent 40 60 60 80 80 2008 0 50000 100000 150000 200000 250000 Per capita income in PPP dollars 0 50000 100000 150000 200000 250000 Per capita income in PPP dollars Note: The figure shows the histogram of global income distribution in 2008 and 2018, created with fifty bins of equal width. The vertical axis gives the percentage of the world population in each bin. The horizontal axis gives per capita incomes in PPP dollars. The graph shows that the lowest bin contains almost two-thirds of the world population in 2008 and less than 60% in 2018. 36 Table 4. Global income distribution in 2008 and 2018 (displayed by income bins, in absolute PPP terms) Lower income bound 400 1000 3000 5000 7000 9000 12000 15000 18000 21000 24000 27000 30000 35000 40000 45000 50000 70000 90000 Percentage of world population 2008 2018 Upper income bound 400 1000 3000 5000 7000 9000 12000 15000 18000 21000 24000 27000 30000 35000 40000 45000 50000 70000 90000 open 9.5 21.3 31.4 11.1 5.8 3.3 4.0 2.9 2.2 2.0 1.4 1.0 0.9 1.0 0.6 0.4 0.3 0.6 0.2 0.2 2.4 11.0 31.3 15.2 9.0 6.0 6.6 4.3 3.3 2.2 1.9 1.2 1.4 1.1 0.8 0.5 0.6 0.8 0.2 0.3 3.2. Core vs. the periphery A different way to look at convergence may be to ask the question whether an increasing proportion of people belonging to the non-core economies enjoys the standard of living associated with the core economies. 35 There are two steps that need to be made here: to define the core economies, and to define what would be the “threshold” income at the core that may be considered as an aspirational income level among the periphery. Obviously, if an increasing share of people from periphery enjoy income that is at least equal to the threshold core income, one can argue that there is income convergence. Survey data with their The core-periphery distinction has, of course, a very long tradition going back to the World Systems Theory, and André Gunder Frank (1966) and Samir Amin (1974). For a more recent discussion along the similar lines, see Knauss (2019) and Hickel (2017). 35 37 household per capita (that is, individual) income levels enable us to answer that question with much greater precision, and more meaningfully, than we would be able to do using GDPs per capita to assess whether a given peripheral country has come closer or overtaken a core country. For core countries, I use the conventional definition that includes Western Europe, North America and Oceania, and Japan (i.e., the group which coincides with the “old” OECD countries). For the threshold income, I use the median income in the poorest core country. In 1988, the poorest core country was Portugal, and the median income in Portugal was $PPP 4,660; in 2018, the poorest core country was Greece and the median income in Greece was $PPP 8,898. The next question we ask is, how many people from peripheral countries had income equal or greater than the core threshold in 1988 and 2018. Table 5 shows the results. As expected, China’s success stands out: from practically having no people with income above the core threshold in 1988, almost one-quarter of China’s population (inclusive of both urban and rural) is above the threshold in 2018. China has overtaken Latin America which thirty years ago was far ahead of it. The convergence is also notable in the rest of Asia (outside of Japan and China): the proportion of the population above the core threshold income more than tripled. Elsewhere, the success was more muted. Latin America improved, but rather marginally from 16 percent to a bit over 20 percent; so did Africa, but with just 3 percent of its population above the core threshold, it remains very much behind other parts of the world, and exceedingly poor. (I shall return to the topic of Africa’s growth in the last part of the paper). Finally, eastern Europe and the countries of the former Soviet Union have declined compared to the core. It is also worth noting that among the core countries themselves, the percentage of those above the threshold is about 83 percent in both years. When we look at the situation in 2018, there is thus a very clear pecking order that is revealed by the data in Table 5. The core is still far ahead of any other part of the globe; former communist countries (the erstwhile “Second World”) remain in the second place even if their relative position has deteriorated; China, having overtaken Latin America, is now No. 3. Far behind everybody else is Africa. Overall, if we look only at the core vs. the periphery (without 38 China), the relationship has only modestly improved in favor of periphery. The percentage of the “peripherals” with income above the core threshold has increased by 3.7 points (from 6.7 percent to 10.4 percent) which gives an average annual improvement of 0.12 percentage points. At that rate, for the periphery (outside of China) to come to a position where, say 50 percent of its population, enjoys at least the core threshold income would require more than 300 years. 36 Table 5. Peripheral population with income equal or greater than the core threshold income Population with income above the threshold (in million) 3.2 36.8 36.8 1988 Total population Percentage above the threshold Population with income above the threshold (in million) 25.5 537.5 205.4 2018 Total population (in million) Percentage above the threshold 1. Africa 224.7 1.4 863.4 3.0 2. Asia (excl. Japan) 2567.2 1.4 3934.3 13.7 Asia (excl. Japan 1465.6 2.5 2562.8 8.0 and China) China 0 1101.6 0 332.1 1371.5 24.2 3. Latin America 61.2 375.7 16.3 114.3 555.4 20.6 4. Eastern Europe 47.4 151.8 31.2 109.4 359.4 30.4 (incl. Soviet Union) Total periphery (1 to 148.6 3319.4 4.5 786.7 5712.5 13.8 4) Periphery without 148.6 2217.8 6.7 454.6 4341.0 10.4 China Core 688.7 825.3 83.4 855.4 1031.8 82.9 Note: Total population included is 4.15 billion in 1988 and 6.75 billion in 2018. The coverage of global population by surveys is thus 94 percent in 1988 and 91 percent in 2018. When we break the world income distribution into the two distributions, one for core and another for the periphery (without China), the gap between the two parts is revealed to be very high. The median per capita income in PPP dollars is just above $18,000 for the core, and only about $2,300 for the periphery (see the dashed lines in Figure 11a; also Table 6). The ratio of the medians is thus 8.2 to 1. In MER terms the gap between the medians is obviously even greater: 22.6 to 1 (Figure 11b). But perhaps even more revealing is to look at the percentage of 36 Calculated as (50-10.4)/0.12=330. 39 the core population living at incomes below the periphery’s median, and its equivalent, the percentage of the periphery’s population having an income above the core’s median: in both cases, such overlap is minimal. At MERs, both overlaps are around 1 percent (or less). In PPP dollars, the share of periphery’s population with incomes greater than the core’s median is 2.9 percent, and the reverse (the poor from core countries with income less than the periphery’s median) is 1.4 percent. 37 The number of people from core countries with incomes lower than the periphery’s median is 16.6 million. This include the extremely poor people who might have fallen between “the cracks” of the welfare system and those with a very low annual income but not necessarily poor if measured by long-term income or wealth. 37 40 Figure 11a. Distribution of income in the core and periphery countries in 2018 (incomes in PPP dollars per capita) density 2 3 4 Ratio 8.2 1 Core 100000 50000 10000 3000 1000 300 0 Periphery log of annual per capita PPP income Figure 11b. Distribution of income in the core and periphery countries in 2018 (incomes in dollars at market exchange rate) density 2 3 4 Ratio 22.6 1 Core 100000 50000 10000 3000 1000 300 0 Periphery log of annual per capita income at market exchange rate Note: Core is defined as Western Europe, North America, Oceania and Japan. Periphery is defined as all other countries except China. The vertical dashed lines are drawn at the points of median incomes. The areas under the curves reflect population sizes. 41 Table 6. Income overlap between the core and the periphery In international dollars Median income (1) Periphery (2) Core Ratio (2) to (1) Percentage of the core’s population with income below the periphery’s median Percentage of the periphery’s population with income above the core’s median In dollars at the market exchange rate 2242 18055 8.2 1.4 759 17167 22.6 0.6 2.9 1.0 4. The present and the future: the period of big external shocks The time when this paper is being written is unlike any recent period. It is characterized by three large external shocks that are still enduring, and whose consequences are impossible to forecast with any certainty. The first was the shock of covid that began in the late 2019, and has been going on for the three past years. It has had important effects on countries’ growth rates (for example, making India’s per capita growth rate in 2020 almost minus 10 percent) and thus on the Concept 2 inequality. However, it is too early to say how these effects will play out over the medium term: many of the GDP declines were almost immediately (in the next years) reversed by similarly-sized increases. The second important exogenous shock was the deterioration of the US-Chinese relations which, given that these are the two largest economies in the world accounting together for almost 45% of the global GDP, will also have an impact on global inequality. Again, here too it is impossible to foretell whether the medium- to long-term impact of the trade conflict would be to slow down Chinese, and even American, economy or not; or to change the evolution of their domestic inequalities. The third shock was that of the Russia-Ukraine war that so far has not only seriously weakened both economies (e.g. with Ukraine’s economy estimated to have contacted by almost a third in 2022) but has affected the rest of the world through economic sanctions, creation of politically-motivated economic blocs, and higher prices of energy and food. It is 42 obvious that if the war continues the effects will cascade and will affect global inequality not only through the differentials in countries’ growth rates, but also through changes in real within-national inequalities. For example, the higher relative prices of food and energy will disproportionately affect poorer households in all countries because the share of expenditures they make on such goods is much higher than that of the richer households. Leaving these exogenous shocks aside (because their effects are, as mentioned, impossible to forecast), we can at least pinpoint to two longer-term developments that do not depend directly on the effects of the shocks. They are the changing roles of China, India, and Africa in the global income distribution. As we have seen, China’s role in shaping the global income distribution was without a doubt determinant since the early 1980s. But the very fact that China has through its advance shaped the distribution means also that China’s relative position has shifted markedly upwards. Hence its growth cannot any longer be globally inequality-reducing as much it was in the past. While in 1988, Chinese urban deciles covered the range between the 17th and 68th global percentile, thirty years later, they range from the 38th to the 93rd percentile. The medianincome urban Chinese has advanced from around the 50th global percentile in 1988 to about the 70th global percentile in 2018. 38 This has also meant that the inequality-reducing role of China has been reduced as China has grown richer, and that at the present, Chinese growth may be broadly neutral as far as global inequality is concerned, even if, as explained in the footnote below, the situation is a bit more complicated. 39 Chinese rural deciles covered in 1988 the range from the first (globally poorest [sic]) percentile to the 56th; in 2018, they ranged from the 7th to the 81st global percentile. The median-income rural person advanced from the 15th to the 43rd global percentile. 38 In global inequality studies, we obviously deal with individual incomes, not with countries’. Thus, the increase of income of a poor Chinese (or even of a poor American) may be inequality-reducing, while higher income of a rich Chinese might increase global inequality. The exact formula (Milanovic, 1994) for the individual infinitesimal income increase that raises Gini in general is 𝑑𝑑𝑑𝑑 2(𝑛𝑛−𝑖𝑖+1) 〉 𝑑𝑑𝑑𝑑 = 𝑖𝑖 〈1 − 𝐺𝐺 − 39 𝑌𝑌 𝑛𝑛+1 where G=Gini coefficient, dmi = change in income the i-th individual (when individuals are ranked from the poorest, 1, to the richest, n=100) and Y=total income of the community (in this case, the world). For Gini to go up, the expression must obviously be positive; thus we require dG>0. Note that the expression depends on what is the initial Gini: the higher the initial Gini, the more “difficult” it is to contribute to its further increase. We can treat i’s 43 Figure 12 shows the annual (marginal) China effect on Concept 2 international inequality, measured by GDPs per capita, from 1952 to 2020 (see Note to Figure 12 for the way the marginal effect is calculated). The years when Chinese growth has contributed to the reduction of Concept 2 inequality are those when the graph line is below the horizontal axis (line of 0), and the opposite for the years when China’s growth has added to global inequality. The latter has happened in the years when China’s growth was strongly negative, and China was poor, as during the Great Leap Forward in 1961, and, to a lesser extent, during the Cultural Revolution that lasted the decade from 1966 to 1976. 40 After the reforms in 1978, as China’s growth picked up, the very opposite occurred: China was the most important engine of global inequality reduction. Figure 12 shows it by the line lying below 0. Closer to the present however that effect weakens. It weakens because China is now sufficiently rich that its high growth no longer has the inequality-reducing effect that it had in the past. It is important to underline that this has nothing to do with the slowdown of Chinese growth as such but with the changed position of the Chinese population in the global distribution of income. In other words, the engine of global inequality reduction that China was from 1978 until approximately 2020, is no longer as powerful as before. as percentiles running from 1 to n. With the current global Gini of 60, dG>0 only for i>81, that is, if incomes of people above the 81st percentile go up (everything else being the same). This could of course also be interpreted that for the Gini to go up the rate of growth among people above the 81st global percentile must be greater than the rate of growth among those poorer than them. The key question then becomes: how may Chinese are above this point? The answer is that in 2018, one-quarter of the urban Chinese were above this point and only 4 percent of the rural Chinese. Consequently, China can still contribute to the reduction in global inequality, but only if that growth comes from the lower part of the urban distribution or from (all but the top 4 percent) of the rural distribution. 40 China’s bad performance added almost 1 Gini point to between-country inequality in 1961. 44 Figure 12. The effect of Chinese GDP per capita growth on Concept 2 inequality, 1952-2020 Marginal China effect -.01 -.005 Gini 0.005.01 Annual change in China effect 1940 1960 1980 year 2000 2020 Note: The graph shows the marginal (yearly) change in Concept 2 inequality due to the addition of China (to all other countries). We can distinguish three periods. In the 1952-1978 period, adding China increases international inequality because China is poor. The China effect does not change much from one year to another as China grows more or less at the same rate as the world. The exceptions are the Great Leap Forward in 1961 and several years of the Cultural Revolution that impoverished China and thus added to global inequality. In 1978-2000, adding China also increases international inequality because China is still poor. But that effect diminishes as China gets richer. So incrementally (from year to year) China’s superior performance helps reduce global inequality. After 2000, the addition of China reduces global inequality but that incremental (marginal) effect becomes substantially weaker from one year to the next. Around 2017-18, China’s incremental/marginal effect becomes close to zero. China’s growth from that point has an almost neutral effect on Concept 2 inequality. Calculated from GDPs per capita obtained from the World Bank’s World Development Indicators. The end of the benign (pro-equality) China effect highlights in turn the key roles that will be played by India and by the populous African countries in the future. In order for the population-weighted convergence to continue India and large African countries need to grow faster than the world, and especially faster than the rich OECD countries. 41 The question has been asked before, can Africa’s growth in the rest of the twenty-first century replicate recent The importance of populous countries growing fast is obvious because only they can make a serious dent in global inequality. 41 45 Asian (and Chinese) growth experience? This matters not only because Africa is relatively poor, but because Africa is the only continent whose population is expected to grow in this century and perhaps even in the next. We obviously cannot answer with any certainty the question regarding the likelihood of Africa’s future fast growth, but if we can look at the past and Africa’s post-1950 record, and use that as a possible guide regarding the future, we cannot be too optimistic. Table 7 shows that only six African countries have succeeded in registering five or more years of consecutive per capita growth of at least 5 percent. This rate of growth can be seen as a reasonable objective which, if maintained over at least a decade, will allow to achieve perceptible convergence. But it was an objective that, as the data show, was unattainable for almost all African countries. In addition, the exceptional episodes listed here involved mostly very small countries (in terms of population) and countries whose growth largely depended on one export commodity (oil in the case of Gabon and Equatorial Guinea, and cocoa in the case of Cote d’Ivoire). It is only Ethiopia, herself rebounding for the disastrous effects of the civil war and secession of Eritrea, that was a populous country (with more than 100 million people) exhibiting high growth for a long period (13 consecutive years). 46 Table 7. African countries with high growth (defined as 5% per capita per year for at least five consecutive years); period 1950-2020 Country Period Number of years Botswana 1969-83 15 Cote d’Ivoire 2013-17 5 Cape Verde 1994-98 5 Ethiopia 2005-17 13 Gabon 1962-66 5 Gabon 1970-76 7 1993-2005 13 Equatorial Guinea Note: Calculated from the World Bank World Development Indicators. This simple exercise suggests that for large African countries (ranked by the size of their populations in 2022: Nigeria, Ethiopia, Egypt, Congo, Tanzania, South Africa) to take over the role of China in this century and the next, one needs to envisage an entirely different growth record. At first sight, this seems very difficult to imagine particularly when we realize that 5 percent per capita annually implies an overall growth rate of 7 to 8 percent (given that the population will be rising at about 2 percent per annum). Moreover, it needs to be achieved under the conditions where an exceedingly young population means that many would be outside of the working age. The rate of income growth per person of working age would need to be even greater than 5 percent, and of course to perdure for a sizeable period. There is an additional issue. According to many recent publications (see Baarsch et al, 2020; Diffenbaugh and Burke 2019; Taconet et al. 2020) climate change is likely to particularly negatively affect Africa, increasing income inequality both between and within nations (e.g. a study of Mozambique by Silva et al., 2015). The negative effect on most of Africa is not only due to greater climate fluctuations in the tropical and sub-tropical areas but also because the continent depends, more than the rest of the world, on agriculture. It is obviously agriculture that is most influenced by the rise in temperature, droughts, floods and climate volatility. Since, as we have seen, the future of global inequality and income convergence crucially depends on the performance of African countries, it is not unreasonable, in light of all these difficulties, to 47 be pessimistic about the future developments, and, in the worst case scenario, think that the period of the past thirty years may be seen just as an interlude, a brief reprieve from a longerterm implacably high global inequality. On a positive note however, one needs to recall that the Asian success was not seen before it took place and that many prominent economists (most famously Gunnar Myrdal in his The Asian Drama: An Inquiry into the Poverty of Nations published in 1968) were overwhelmed by the bleak prospects for Asian growth, given the apparent overpopulation of the countries and their slow technological development. Not only that these forebodings did not materialize, but Asia (and this does not apply to China only) became a continent with the exceptionally high rates of growth. These errors of prediction should give us pause when we look at the difficulties of Africa’s growth in the next fifty or more years, and perhaps too easily dismiss possibilities of a sharp break with the past. 5. Conclusions Studying global inequality in incomes over the past two centuries is in effect studying global economic, and to some extent, political and military, history of the world. The numbers are dry, but behind these numbers –and determining these numbers—are big historical changes: the rise and decline of different parts of the world. Before the Industrial Revolution global income inequality, as far as we know, was relatively low. There were poor and rich people in various countries and empires, but there were no systematic differences in income levels between different parts of the world. This is, of course, what we believe today, based on the empirical data that begin with 1820. But this is known only by extension, not through direct estimation of inequality of the world in say, 1700. Thus while this conclusion seems reasonable much more remains to be done in studying inequality between and within countries in the decades and centuries before the Industrial Revolution. Social tables that list salient classes with their population shares and incomes will be, I believe, the main source for further advances in our empirical knowledge of pre-industrial inequalities. 42 For a very good survey of social tables and their use for the estimation of historical inequality, see the recent paper by von Fintel, Links and Green (2023). 42 48 The Industrial Revolution represented a fundamental break: it launched some countries, principally Western Europe and North America, to a much higher growth path, and left others more or less at the same level at which they were before. Military disparity between the countries, and thus colonization, reinforced the cleavage, and created a world composed of a core and a periphery. Thus the gap between nations widened, and together with it also the gap between the “haves” and the “have-nots” within nations. Throughout the nineteenth century and up to the World War I global inequality increased, driven by both of these forces. After World War II, global inequality stabilized, albeit at an unprecedently high level. Within-national inequalities decreased, but this was insufficient to make much of a difference to the dominant force of unequal world development. The tripartite world of the second half of the twentieth century began to crumble after 1980. The rise of China, and around the same time or a bit later of India, Indonesia, Bangladesh etc., was indeed a mirror image of the Industrial Revolution. But, like in a mirror-image, while the Industrial Revolution increased between country gaps, the rise of Asia reduced them. In absolute terms, however, the gaps (measured by the average income distance between two random individuals in the world), increased. They also remained very large, and in some cases even widened, between the “core” of the world economy (the rich countries of Europe, North America and Oceania) and parts of Africa and Asia. The period from around 1980 to 2020 thus witnessed the largest reshuffle of individual income positions since the Industrial Revolution. Global upper-middle and high income positions that were “populated” almost exclusively by the citizens of the Western countries and Japan, began to be taken over by the populations from the “rising Asia”. The reshuffle has not yet affected, to a significant extent, the very top of the global income distribution (the top 5 percent or the top 1 percent), but if the differential in real income growth between Asia and the West persists, the reshuffle will be felt there too. The positional reshuffle is, by historical standards, dramatic because of the numbers of people involved in these upward movements. If the numbers of people going up were smaller, the extent of the reshuffle would be obviously less. But when millions of people with similar incomes overtake a person, he or she falls down, positionally, very fast. The positional decline does not imply, of course, a real income decline. 49 Very often it goes together with an absolute improvement in real income. But not with an absolute improvement that is equal to that of other people with similar incomes from other countries. Does positional decline matter? In many ways, not. We often do not know our national income positions, much less so global. But in other ways, it matters. Lots of consumption is “globalized”: there are international goods and services that are affordable only to the select segments of the globally affluent or of the global middle class. One does not need to know exactly where he or she falls in the global income ladder; yet inability to easily purchase certain “global” goods and services (foreign travel, latest smart phone, subscription to the popular show, attendance of a sporting event) will soon, even if indirectly, convey that message. The current period is therefore one of dramatic developments where the progression of incomes in China without doubt plays the key role. In fact, never in history have so many people’s incomes increased so much so fast. Whatever happens in the future will not erase the magnitude of this historical success. China’s role in the reshuffling of global incomes is not over, but its role in reducing global inequality is at, or is soon coming to, an end. The parts of the world whose income convergence now becomes of key importance are India and half a dozen populous African countries, which are also the only parts of the world likely to register massive population growth—which indeed makes them even more important for the matters of global inequality. Will they be additionally hampered by the climate change? Will Africa in the twentyfirst century replicate Asia of the latter part of the twentieth? This is the critical question for the continued income convergence—which is not merely a numerical objective, but a means towards the creation of a more equal and peaceful world eloquently envisaged two and a half centuries ago by Adam Smith: At the particular time when these discoveries [of the Americas] were made, the superiority of force happened to be so great on the side of the Europeans that they were enabled to commit with impunity every sort of injustice in those remote countries. Hereafter, perhaps, the natives of those countries may grow stronger, or those of Europe may grow weaker, and the inhabitants of all the different quarters of the world may arrive at that equality of courage and force which, by inspiring mutual fear, can alone 50 overawe the injustice of independent nations into some sort of respect for the rights of one another (Wealth of Nations, Chapter 7). 51 Annex I. Revisions of the Bourguignon-Morrisson dataset The original Bourguignon-Morrisson database consists of 33 country/blocs each having its own distribution and covering the period 1820-1992, at approximately ten-year intervals (there are in total 11 benchmark years; all of them except 1992 which is not used here are listed in Tables A1 and A2). The distributions are bloc-specific and in principle blocs should consist of countries that have similar distributions. Since the compositions of the blocs do not change, the implicit assumption is that the income distribution in each country within the bloc evolves in the same manner throughout two centuries. This is obviously a huge assumption, but, given the lack of data, it is difficult to improve upon it. The country/blocs may consist of individual countries, like France, Mexico, Nigeria, Poland, the United States, France etc. In that case GDPs per capita of the bloc are straightforward: they come from the Maddison original database where GDPs were expressed in 1990s PPPs. The distributions are also straightforward because they relate to individual countries. Most of the blocs however include more than one country. There, the GDPs per capita (on which is imposed a given distribution) are, in principle, population-weighted averages coming from the same Maddison database. But in same cases, when the number of countries in a bloc is large and GDPs for all countries are not available, the average GDP per capita of the bloc might reflect only GDPs of a few countries, i.e. of the countries for which the Maddison database provides GDPs per capita in that