Recent Trends in Income Inequality Joe Hasell, Salvatore Morelli, and Max Roser

part 2. PART 2

mechanisms and context underlying social inequalities in cancer 9 CHAPTER

chapter 9. Recent trends in income inequality Joe Hasell, Salvatore Morelli, and Max Roser

Introduction dimensions: overall inequality, the Inequality has evolved very different- share of income received by the top ly in different countries, with falling or This chapter provides the global 1% of the income distribution (re- constant levels of inequality in many context of economic inequalities referred to here as the “top 1%”), and countries and rising levels in others. lated to income, which is important relative poverty rates. Some context No single narrative serves to cap- to understand health and cancer in- is first provided by a brief look at re- ture this heterogeneity adequately, equalities. In recent years there has cent changes in the global distribu- but some clear regional patterns do been something of a renaissance tion of income. emerge. in the study of economic inequality, Among advanced industrial econ- Important provisos about the cov- simultaneously responding to, and omies, the availability of comparable erage and quality of available data feeding, the emergence of a public long-term data reveals a general become increasingly pertinent the and political consciousness of the increase in income inequality in the broader the range of countries and issue. Today, social scientists find final decades of the 20th century, the longer the time period being themselves equipped with a wealth after substantial declines earlier in considered. However, the interpre- of easily accessible data on inequal- the century. However, even among tation of all inequality data requires ities, much of which was unavailable this relatively homogeneous group some care. Unlike the measurement to them 20 years ago. of countries there are significant and of height or weight, trends and com- The purpose of this chapter is to noteworthy differences in terms of parisons of inequality data may ap- briefly summarize this body of evi- the timing and extent of the increase. pear quite different depending on the dence. The focus is on income ine- When global trends are considered, particular measure chosen. In this quality within countries across three the picture is much more complex. chapter these issues are highlighted,

Part 2 • Chapter 9. Recent trends in income inequality 121 including a brief discussion of some Fig. 9.1. Global inequality decomposed into inequalities between countries particular limitations that should be and within countries. The estimates were constructed by combining national considered when linking inequality household surveys, some of which referred to consumption and others to (disposable) income, at 2011 purchasing power parity exchange rates. Where data to health outcomes. surveys in the reference year were unavailable, adjacent years were also used. The inequality metric here is of the Generalized Entropy family. GE(0) Global income inequality (or Theil-L index) is a decomposable measure of overall inequality equal to the mean log deviation. The top horizontal line shows the evolution of overall Global income inequality simply re- inequality and the lower horizontal line that of within-country inequality, both flects the combination of inequality in population-weighted terms. The proportions of the between-country and between countries and within coun- within-country component of global inequality are given as percentages of tries. Between-country inequality is total inequality for each reference year. Source: compiled from Lakner and Milanovic (2016 [Table A.3]). basically due to relative rates of economic growth. Rapid growth in many developing countries, most notably in Asia, and a relative slowing of growth in high-income countries have brought about a convergence in average per capita incomes between countries in recent decades; after two centuries of divergence, this development is of historical sig- nificance (Pomeranz, 2000). At the same time, some of the processes driving this catch-up, such as globalization or technological development, have been charged with con- tributing to the rising inequality seen within many countries, both rich and poor, since the 1980s (Freeman, 1995; Bourguignon, 2015; Basu, 2016). Studying changes in the global income distribution enables these movements to be considered jointly. Lakner and Milanovic (2016) provided estimates for global ine- within-country component in recent Inequality Report painted a rather quality decomposed into separate years. However, it should be noted different picture of the evolution of within-country and between-country that this trend, as well as that for global inequality: between 1988 and components (Fig. 9.1). The estimates overall global inequality, is sensitive 2016 the global top 1% pulled away, show that although between-country to how incomes at the very top of the with average income growing 100% inequalities are diminishing, they distribution are accounted for and, compared with 60% growth in the still vastly outweigh within-country indeed, to the measure of inequali- world average (WIL, 2018). (Lakner inequalities. The increase in with- ty chosen. (Ravallion (2018), for in- and Milanovic (2016) checked their in-country inequality visible through- stance, reported a range of Atkinson out the 1980s and 1990s, although global inequality estimates for ro- significant, has been outpaced by indices that yield rising global in- bustness to making some allowance the convergence in average incomes come inequality over this period.) for these missing top incomes. In this between countries, translating into a In focusing on the share of income case, a far smaller fall was reported reduction in overall global inequality. received by top-earning percentiles over the period 1988–2008, and ine- Note that in Fig. 9.1 there is ev- of the population (as captured in quality was only observed to decline idence of a flattening-out in the administrative tax data), the World after 2000.)

122 Within-country income The Gini index is an indicator that Global picture since 1990 inequality attempts to summarize, in a single number, the degree of dispersion Fig. 9.2 compares the Gini index for across the entire distribution. about 2015 with that for about 1990.

Overall income inequality PART 2 The estimates are based on house- It is most easily understood in 9 CHAPTER Focusing on the within-country ine- terms of mean difference: a Gini in- hold surveys conducted at the na- quality, we first consider overall ine- dex of G% means that, if we take any tional level, drawing primarily from quality as captured by the most com- two households from the population the World Bank’s Povcal database monly used inequality indicator: the at random, the expected difference is (World Bank, 2018), with addition- Gini index. 2G% of the mean. al figures from The Chartbook of

Fig. 9.2. Gini index for about 2015 plotted against that for about 1990, including both income and consumption survey data. Only countries for which estimates of the Gini index were based on broadly comparable surveys for the two reference years were included. The closest survey to the reference year was selected, up to a maximum of 5 years difference. The size of the circles is in proportion to population size. Data for China are from Kanbur et al. (2017 [Table 1.B]). Source: compiled from Atkinson et al. (2017), Kanbur et al. (2017), and World Bank (2018).

Part 2 • Chapter 9. Recent trends in income inequality 123 Economic Inequality (Atkinson et al., dominantly based on income, thereby across the two periods. However, 2017). Given that surveys are often exaggerating somewhat their position comparing high- and low-inequal- not conducted on a regular annual relative to other low- and middle-in- ity countries, we see different pat- basis, estimates for the year closest come countries. (To address this is- terns. Among those countries with a to each reference year were select- sue, Alvaredo and Gasparini (2014), Gini index below 40% in 1990, there ed up to a maximum time difference in their analysis of the Povcal data, were substantial declines in very few of 5 years. On the basis of this rule, choose to apply a downwards adjust- during the period until 2015. Above the shortest admissible time be- ment of about 15% to Latin American this threshold, however, the inverse tween surveys was 15 years, from and Caribbean estimates, so that a holds. Fig. 9.2 therefore suggests a 1995 to 2010. In practice, Mali had Gini index of 50% is reduced to about modest convergence in the Gini in- the shortest such time span, with 43%. This option is not taken here.) dex across countries between 1990 surveys taken from 1994 and 2010; Moreover, even among surveys that and 2015. The pattern, however, it is only for 6 out of 84 countries that measure household income, there works in large part through regional the selected surveys fell less than are a range of income concepts that dynamics. There was an increase 20 years apart. may be used, relating to which kinds in inequality in most southern Asian It is important to stress that Fig. 9.2 of income are counted and how taxes countries, in most advanced industri- includes a rather heterogeneous mix and transfers are considered. al economies, and in several transi- of data points that are based on a Such heterogeneity is unavoid- tioning countries in eastern Europe. broad range of survey methodologies able if one wishes to take a global Across the Caribbean and Latin and concepts. This introduces signif- view. However, to attempt to man- America, as well as North Africa icant comparability issues (Alvaredo age this issue, only countries for and the Middle East, the Gini index and Gasparini, 2014; Lakner and which the Gini index estimates for decreased for almost all countries. Milanovic, 2016). Of these, the most the two reference years were based Countries in sub-Saharan Africa and acute divergence is between the use on broadly comparable surveys are the eastern Asia and Pacific regions of consumption versus income as the included in Fig. 9.2. This restricts the had more mixed results, with fall- measure of welfare within the survey. sample to 83 countries. ing inequality among countries with Generally speaking, lower-income The colours of the bubbles refer to higher inequality in 1990 and rising countries use consumption measures world region, with advanced industrial inequality among countries with low- and higher-income countries use in- economies (as defined by the World er inequality in 1990. come measures. Crucially, the level Bank, 2016) presented as a separate In terms of the average Gini in- of consumption is, as a rule, more group. Overall, inequality tends to be dex, these contrasting trends largely equally distributed across households higher in Caribbean, Latin American, cancelled themselves out during the than is income, with the gap increas- and sub-Saharan African countries. period; the mean index across all ing with average incomes (World The Scandinavian countries, and countries in the sample was more or Bank, 2016 [pp. 78–79]). (This gap several eastern European and central less the same in about 2015 (38.6%) reflects (i) the increasing propensity Asian countries, are positioned at the as it was in about 2015 (39.6%), a to save at higher incomes and (ii) a opposite end of the spectrum. Most fall of 1 percentage point being small more general tendency for house- advanced industrial economies have compared with the large variation in holds to smooth consumption levels a Gini index clustered at about 30– the data. However, sizeable increas- over time.) The more level distribution 35%, and the USA shows the highest es in inequality in several populous of consumption implies a downwards inequality during both periods. countries, including China, India, bias in consumption surveys relative In the figure a 45-degree line is Indonesia, and the USA, yielded a to income surveys. This is particular- plotted. Those countries lying above population-weighted average that ly true for countries with higher av- the line in Fig. 9.2 show higher ine- increased by four percentage points erage incomes that are represented quality in (or around) 2015 than 1990, (from 36.7% to 40.8%), in line with in Fig. 9.2 on the basis of consump- and those below the line lower ine- the rise seen in the within-country tion surveys, such as the Russian quality. Across all countries, we see component of global inequality in Federation. In contrast, surveys in a roughly equal split between coun- Fig. 9.1. Therefore, although in the Latin American countries are pre- tries with higher or lower inequality average country in the sample there

124 was no significant change in the Gini Fig. 9.2 does not enable us to trace vey data for about 2000. In terms of index between 1990 and 2015, the the different paths taken by countries additional information, we see that average person lived in a country over the 25 years between the ob- the fall among Caribbean and Latin that had meaningful rises in inequal- servations; repeating the exercise American countries was concen- PART 2

ity. (The sample covers less than for the period after 2000 yields some trated after 2000 (as confirmed in 9 CHAPTER half of the countries in the world but additional information (Fig. 9.3). In Fig. 9.4), whereas the rises in east- represents about 85% of the global this example, surveys were selected ern Europe seen in Fig. 9.2 occurred population; although better global with a maximum period of 3 years in the post-Soviet period during the coverage might be expected to affect between each reference year; this 1990s. Among advanced industrial the unweighted mean reported here, increased the number of countries economies, the rise seen over the the population-weighted mean would plotted to 93, but India was exclud- full period from 1990 was still contin- be unlikely to change much.) ed because of the absence of sur- uing into the new millennium.

Fig. 9.3. Gini index for about 2015 plotted against that for about 2000. Both income and consumption survey data were included. Only countries for which estimates of the Gini index were based on broadly comparable surveys for the two reference years were included. The closest survey to the reference year was selected, up to a maximum of 3 years difference. The size of the circles is in proportion to population size. Data for China are from Kanbur et al. (2017 [Table 1.B]). Source: compiled from Atkinson et al. (2017), Kanbur et al. (2017), and World Bank (2018).

Part 2 • Chapter 9. Recent trends in income inequality 125 Fig. 9.4. Gini index in selected Latin American countries for the period 1981–2012. Figures refer to equivalized household income, defined as market income plus transfers, less taxes on wage income. Source: SEDLAC (CEDLAS and the World Bank) (2018).

Long-term picture in advanced extent and timing of any increase typology masks important differenc- industrial economies differed significantly between coun- es. For example, Sweden (Fig. 9.5a) tries. To highlight these differences, stands out in terms of the extent of the For several advanced industrial we group countries into those follow- rise in inequality seen there, follow- economies, we can benefit from ing loosely similar trends over this ing a different trend from its Nordic longer-term Gini index series con- period. Several countries underwent neighbours and joining the ranks of forming to a more homogeneous a more or less continuous increase countries such as New Zealand and set of definitions. In this section we in inequality between the 1980s and the United Kingdom. More modest, refer to inequality of equivalized the 2010s (Fig. 9.5a). Another cluster but still significant, increases in in- disposable household income, that is, income after taxes and transfers is formed by several Nordic coun- equality in Denmark, Finland, and have been paid, measured at the tries, which began their climb some- Norway contrast with the steady lev- household level, but adjusted to ac- what later in the 1990s, and from a els in France and the Netherlands count for the size and composition of lower starting point (Fig. 9.5d). (Fig. 9.5c), which contribute to a rela- the household. We primarily draw on Indeed, a geographical distinc- tive convergence between continen- data presented in The Chartbook of tion is often made between high-in- tal European and Nordic countries Economic Inequality (Atkinson et al., equality English-speaking countries, from the 1980s. The step increase 2017). Fig. 9.5 demonstrates a gen- the more moderate continental in inequality seen at a greater mag- eral rise in overall income inequality European countries, and low-ine- nitude in Canada, New Zealand, and among advanced industrial econo- quality Nordic countries. Although it the United Kingdom, and at a lower mies since the 1980s. However, the is informative, even this very loose magnitude in Finland and Germany

126 Fig. 9.5. Gini index in high-income countries for the period 1960–2015. In most cases figures refer to disposable (after taxes and transfers) household income, equivalized for household composition. For Canada, the unit of analysis is the family; for Italy, figures are per capita. Data for Denmark and the USA are from LIS (2018). Source: LIS (2018). a b PART 2 9 9 CHAPTER 9

USA

Italy New Zealand Spain UK Sweden Australia Canada Germany

2 Japan 2 2 2

2 2 29 9 9 99 2 2 2 29 9 9 99 2 2 2 9 9 9 99 2 2 2 9 9 9 99 2 2 2 c 9 d 99 9 99

Portugal

Switzerland France

Netherlands Norway 2 2 Finland 2 2 Denmark

2 2 29 9 9 99 2 2 2 29 9 9 99 2 2 2 9 9 9 99 2 2 2 9 9 9 99 2 2 2

(Fig. 9.5b), also merits attention; the Inequality before and after taxes effect of redistribution rather impre- recent levelling out in these countries and transfers cisely. Between the countries there increasingly serves to accentuate are important differences in what the exceptionality of the USA among Using Organisation for Economic Co- is counted as a transfer, particular- high-income countries. operation and Development (OECD) ly in relation to pension systems. Taken together, advanced indus- data from 2014 (OECD, 2018), Moreover, any market responses trial economies are today consider- Fig. 9.6 shows the Gini index for to tax policy are already included in ably more unequal places than they market income in red and that for market income inequality.) were in the 1980s. In very recent disposable income in blue. The size The large variation in the redistrib- years, however, the trend is less of the gap between the two mea- utive effect of taxes and transfers sys- clear, with the Gini index rising, fall- sures captures the effect of the systems in different countries means that ing, and levelling out in roughly equal tem of taxes and transfers (both pub- the resulting level of overall inequal- proportions among advanced indus- lic and private) on reducing overall ity of disposable income has a large trial economies in the post-2008 pe- inequality. (As noted by Morelli et al. degree of independence from mar- riod (World Bank, 2016 [Table 4.1]). (2015), this difference captures the ket incomes. For instance, inequality

Part 2 • Chapter 9. Recent trends in income inequality 127 Fig. 9.6. Gini index of market and disposable income, where figures refer to equivalized household income. Most observations are from 2014, but if data from 2014 were not available earlier observations are shown (the earliest is 2011, for China, India, and the Russian Federation). Estimates for the Netherlands are provisional, according to the OECD. *, market income Gini index for China, Hungary, Mexico, and Turkey refers to income after taxes and before transfers. Source: OECD (2018).

in Chile, India, and Mexico is compa- and transfers that is considerably low- of inequality in incomes before taxes rable to that in Finland if we consider er than that of the United Kingdom. and transfers (e.g. lower than inequal- incomes before taxes and transfers. This is, however, not to downplay ity after taxes and transfers in both However, these countries contrast the role of market incomes in shap- the United Kingdom and the USA). sharply in terms of their levels of dis- ing disposable income inequality. The More generally, increasing concerns posable income inequality. Despite Republic of Korea is a case in point about the political or economic limits starting out with the least equal dis- here: despite minimal redistribution, to redistributive taxation have brought tribution of market income among it lies towards the middle of the rank- attention to the level of inequality of OECD countries, Ireland achieves a ings in terms of disposable income in- market incomes as an important level of income inequality after taxes equality because of its very low level issue in its own right, as well as to

128 the potential role of predistribution standard Gini index estimates for the such that most European countries policies to encourage a more equal country based on survey data alone today remain much more equal plac- spread of incomes, assets, and op- underestimate the increase in ine- es than in the early 20th century. After portunities before the operation of the quality between 1970 and 2006 by strong rises in inequality since the PART 2

taxes and transfers system (Atkinson, more than one half.) As well as affect- 1980s, the BRICS group of countries 9 CHAPTER 2015). ing levels, this may potentially also (Fig. 9.8) today displays levels of top understate inequality trends over 1% shares that are generally higher Top income shares time, as demonstrated by Atkinson et than that of high-income countries, in al. (2011), for instance, in the case of parallel with the Gini index. In recent years, increasing attention the USA. Top income shares there- The correlation between top in- has been paid to the share of income fore serve as an important counter- come shares and Gini index is, how- received by the highest-earning pro- part to the foregoing observations ever, far from perfect; indeed, the portions of the population (referred using Gini index estimates based on correlation has weakened somewhat to here as “top income shares”) survey data alone. since 2000 (Morelli et al., 2015). The (Atkinson and Piketty, 2007, 2010). Another benefit of tax data is persistently high top 1% share in This approach has several advan- that they are often available over far Brazil, for instance, contrasts with tages. First, it addresses a lack of longer timeframes than household the falling Gini index seen in Fig. 9.4. sensitivity in the Gini index to shifts survey data are. Consequently, for In the Russian Federation, large de- at the extremes of the distribution many countries a much longer-term clines in the Gini index based on con- (Osberg, 2017; WIL, 2018). Second, view of inequality trends is avail- sumption surveys during the 1990s in practice, the use of the top income able with top income shares than (coinciding with the post-Soviet eco- shares measure has gone hand in with the Gini index calculated from nomic collapse) directly contrast with hand with the use of administrative household surveys. The downside is the rocketing top 1% share over the tax data and national accounts ag- that, at present, long-term data are same period. gregates, as opposed to household available for a limited range of coun- Overall, the picture of recent ine- survey data alone. (Alvaredo et al. tries, somewhat skewed towards quality trends painted by top income (2016) set out a methodology for in- advanced industrial economies, and shares is somewhat less benign than corporating both survey and tax data this makes summary statements of that given by standard Gini index es- to impute an income distribution con- global scope difficult. Fig. 9.7 shows timates. Among the 22 countries for sistent with total household income the share of (pre-tax) income re- which top 1% share estimates were as reported in national accounts. ceived by the top 1% across three available in the World Inequality Some of the top income shares se- groups of high-income countries: Database (WID, 2018) about both ries in the World Inequality Database English-speaking countries, central 2000 and 2015, more countries saw (WID, 2018) are derived in accord- Europe (plus Japan), and Nordic meaningful increases than falls; this ance with this distributional national countries. Fig. 9.8 shows the same is in direct contrast to the change in account (DINA) approach.) Such fis- series for the so-called BRICS the Gini index over the same period cal data avoids one major shortcom- grouping of countries (Brazil, the (Fig. 9.3). This disparity is at least ing of survey data: that of underreport- Russian Federation, India, China, partly due to selection, however, ing of incomes or non-response by and South Africa). In both cases with those countries that have rising those at the very top of the income the data are drawn from the World Gini index values in Fig. 9.3 overrep- distribution, and the underestimation Inequality Database (WID, 2018). resented within this small sample. of inequality that this may imply. Many of the general observations Nevertheless, given the acknowl- The Gini indexes reported in the noted for the Gini index (Figs. 9.2– edged weaknesses of Gini index es- previous section were based ex- 9.4) still hold. English-speaking coun- timates based on household survey clusively on household surveys. tries have, in general, seen a promi- data to capture movements at the (Analysis by Atkinson et al. (2011) nent rise in the top 1% share in recent extremes of the distribution, data on shows that a combination of survey decades, with more muted increases top income shares, where available, data and administrative tax data for in continental European and Nordic provide an indispensable additional top incomes in the USA implies that countries. This rise came after a fall, perspective.

Part 2 • Chapter 9. Recent trends in income inequality 129 Fig. 9.7. Top 1% share of pre-tax income (all income received by individual owners of capital and labour, before tax/ transfers but after pensions) in high-income countries for the period 1915–2014. The Italian series on top income share was extended to 2014 (provisional estimates) using adjusted council-level data on incomes reported in income tax returns, kindly provided by Demetrio Guzzanti. Source: WID (2018). a

United States

United Kingdom Canada

Ireland Australia

Germany

France Japan Italy Spain

Netherlands

Sweden Norway Finland Denmark

130 Fig. 9.8. Top 1% share of pre-tax income (defined as for Fig. 9.7) in BRICS countries for the period 1915–2015. BRICS, Brazil, the Russian Federation, India, China, and South Africa. Source: WID (2018). PART 2 CHAPTER 9 CHAPTER

Relative poverty rates comparative feature that makes the and southern European countries measure an indicator of inequality. have generally higher levels of Whereas top income shares track In practice, relative poverty inequality than their continental the incomes of a fixed (upper) pro- lines are typically defined as some European counterparts, which in portion of the population, poverty fraction of the contemporaneous turn are more unequal than Nordic rates do the reverse: they fix a level median income. We now consider countries. of income (a poverty line) and track the percentage of individuals with In terms of trends over time the proportion of the population that disposable incomes less than 60% (Fig. 9.10), few generalizations are falls beneath that level. In the case of the national median, adopted by possible, with the exception that, of an absolute poverty rate, that the European Union (among oth- when the most recent observations poverty line is set so as to maintain ers) as its headline poverty indi- are compared with those about 1980, a constant purchasing power over cator. When those countries with no country with available data for this time, at a level considered neces- estimates available between 2012 period has seen a meaningful fall in sary to achieve a certain minimum and 2014 are considered (Fig. 9.9), poverty rates. However, some coun- standard of living. Relative poverty poverty rates range from 11% in tries stand out for the increase that rates instead refer to a poverty line Czechia to almost 30% in Peru and has occurred over this period; Israel that is tied in some way to the aver- South Africa. As with other dimen- is particularly notable in this respect, age standard of living of the time. In sions of income inequality, lower-in- and Germany, Spain, and Taiwan, real terms, that threshold may rise come countries generally feature China also had increases of several and fall with the overall fortunes of more heavily at the top end of the percentage points. There was also a the population in question. It is this rankings. Again, English-speaking significant jump in Finland from the

Part 2 • Chapter 9. Recent trends in income inequality 131 Fig. 9.9. Relative poverty rates (percentage of the total population living in households with equivalized disposable income < 60% of the median) for about 2014. Source: LIS (2018).

mid-1990s onwards, which moved number of seemingly technical is- quality, including its relation to health the poverty rate towards the higher sues in the definition and meas- outcomes including cancer. Particu- end among Nordic countries, match- urement of inequality. Following lar caution has been sounded in the ing the movement seen in the Gini Atkinson and Bourguignon (2015 [p. inequality literature against the un- index. After earlier falls in the pov- xxxiv]), we can consider a checklist critical use of secondary databases erty rate, Canada and France have of questions in assessing the compa- of distributional statistics in econo- largely rebounded in recent years. rability of two inequality data points: metric studies involving inequality In contrast, in the United Kingdom, (i) inequality of what (pre- or post-tax measures as an independent varia- a marked increase throughout the income, wealth, consumption, other ble (Atkinson and Brandolini, 2001; 1970s and 1980s has been reversed dimensions of well-being); (ii) be- Jenkins, 2015). Given the facility of since 2000, leaving the country with tween whom (individuals, families, sources such as the World Income poverty rates more in line with those or households, with various ways of Inequality Database assembled by of the continental European coun- accounting for household composi- the United Nations University World tries. The available data for Ireland tion); (iii) according to which sources Institute for Development Econom- extend back long enough to show, at (surveys, tax data); and (iv) accord- ics Research (UNU-WIDER, 2018), least, that this fall was mirrored there. ing to which measure (Gini index, top and its standardized counterpart, shares, etc.). it is easy to lose sight of the quality Using economic inequality Such issues, as we have already of the ultimate data upon which the data in health inequalities seen, impinge upon us even in at- sources are constructed, the signif- research tempting to describe recent inequal- icant comparability issues between To maintain a meaningfully con- ity trends; they are even more im- many of the underlying sources that sistent metric across countries and portant in attempting to investigate are available, and the interpolation years, attention must be paid to a the social impacts of economic ine- used to fill in for those that are not.

132 Fig. 9.10. Relative poverty rates (< 60% of the median) for the period 1978–2014. *, including Taiwan, China. Source: LIS (2018). a E E b C C E E

2 2 PART 2 USA CHAPTER 9 CHAPTER Spain Australia 2 Canada 2 Greece Italy

Ireland United France Kingdom Germany Switzerland Austria Netherlands

R R

9 99 2 2 2 9 99 2 2 2

c d EC

Israel 2 2 Mexico

2 2

Poland Hungary Taiwan, China Finland Norway Denmark Sweden Czechia R R

9 99 2 2 2 9 99 2 2 2

Notwithstanding improvements in re- social phenomena, such as health, inequality series may well be – is cent versions of these sources, Atkin- attention to causality is surely war- unlikely to yield significant results, son and Bourguignon (2015 [p. xxxiii]) ranted. For instance, causal interpre- regardless of the true relationship. still advise “careful inspection” before tations of the negative cross-coun- Moreover, depending on the caus- their use. try association between population al path being hypothesized, some Clarity about proposed causal health and income inequality, as inequality measures may be more apt mechanisms and how these relate observed in several studies, should than others. If, for instance, extreme to available distributional data is also be informed by the general absence economic disparities are thought to needed. In discussing health ine- of an effect when moving to panel pose a threat to health equality via quality, Deaton (2013) made a strong or time series data, as surveyed in specifically political channels, as case for the position that “facts and O’Donnell et al. (2015). However, Deaton (2013) suggests, then ine- correlations, without an understand- any empirical approach must take quality measures that pay particular ing of causation, are neither sufficient into account the strengths and weak- attention to top income shares may to guide policy nor to make ethical nesses of the various inequality data be more relevant. judgments”. Even if we view the facts used. Atkinson and Bourguignon Conclusions of inequality as of ethical import in (2015 [p. xxxvii]) further remind us and of themselves (Atkinson, 2015), that a very noisy regressor – as the In recent decades, levels of in- in seeking to connect them to other foregoing discussion suggests many come inequality have risen in most

Part 2 • Chapter 9. Recent trends in income inequality 133 advanced industrial economies. tradicting trends shown in the Gini context, whereas in this chapter we Long-term data show that this in- index based on household survey have exclusively discussed relative crease was preceded by a sustained data alone. However, the restricted inequality (Atkinson and Brandolini, decline from the early 20th century coverage prevents us from making a 2010). In analysing survey data, onwards, tracing a broad U-shaped full like-for-like comparison at pres- Kharas and Seidel (2018) found that trend (Atkinson and Piketty, 2007, ent. More generally, it is important to the incomes of those at the 5th per- 2010) over the century in English- remember that the way we choose to centile of the global distribution in speaking countries, with more muted operationalize our common notions 1993 grew considerably faster until increases in continental European of inequality, and how we measure 2013 than the incomes of those at countries. These increases contin- this, may considerably affect the re- the 99th percentile (see Lakner and ued in many countries into the 21st sulting picture of inequality. A focus Milanovic, 2016). This sounds con- century, but Gini indexes have been on wealth inequality, for instance, broadly stable since the global finan- would paint a far starker picture of siderably less progressive, however, cial crisis of 2007–2008. the state of economic inequality, with when presented in absolute terms; When the view is broadened to the poorest 40% of households typi- such increases translate to only tens the global level, the picture is much cally owning less than 5% of house- of dollars per year at the bottom of more heterogeneous. Declines in hold net wealth in OECD countries the distribution but to thousands of many countries balance out rises in (Balestra and Tonkin, 2018) and dollars per year at the top. others, at least in terms of the un- top 1% wealth shares far outstrip- In this chapter we have given a weighted mean. However, many of ping those of income (WIL, 2018). summary of recent inequality trends, the world’s most populous countries Available data on wealth inequality emphasizing the differences that are had significant increases in inequali- are too scarce to enable any confi- seen across countries and regions. ty, resulting in an increasing popula- dent statements to be made about We have also tried to indicate some tion-weighted average since 1990. In global trends. Tentative first esti- of the limitations of the existing data terms of inequality among all global mates of the global top 1% share and indicate where care is needed citizens, this increase in the with- of wealth from the World Inequality in their interpretation. Such consid- in-country component was outpaced Report (WIL, 2018) largely mirror erations should form a background by convergence in mean incomes of those presented for income, howev- countries, resulting in a decline in er, with rises throughout the 1980s to any understanding of inequality global inequality that has gathered and 1990s before flattening out in the trends, and are of particular impor- pace since 2000. new millennium. tance to those seeking to study the As we have argued, data on top Many would point out the im- interaction between economic ine- income shares cast doubts on some portance of absolute differences quality and other social phenomena, of these conclusions, at times con- in income, particularly in a global including health outcomes.

134 Key points • In most advanced industrialized economies, within-country income inequality has risen since the 1980s after falling earlier in the 20th century. However, there were significant differences between countries in

terms of the timing and extent of the rise. PART 2 CHAPTER 9 CHAPTER • Globally, the picture is much more complex, with recent falls in inequality in many high-inequality countries resulting in an average Gini index today that is quite similar to that of about 1990. • Significant rises in inequality since 1990 in several populous countries, including China, India, and the USA, mean that the average person lived in a country that had meaningful rises in inequality. • Given several concerns about data quality and interpretation, it is important to consider multiple perspectives on inequality. In particular, figures on top income shares that incorporate tax data and national accounts are a key complement to standard Gini index estimates based on survey data alone and, in some cases, present notably less benign trends in recent years.

Part 2 • Chapter 9. Recent trends in income inequality 135 References

Alvaredo F, Atkinson AB, Chancel L, Piketty T, Basu K (2016). Globalization of labor markets OECD (2018). OECD income distribution Saez E, Zucman G (2016). Distributional national and the growth prospects of nations (English). database (IDD): Gini, poverty, income, methods accounts guidelines: concepts and methods used Policy Research working paper; no. WPS and concepts. Organisation for Economic Co- in WID.world. WID.world Working Paper 2016/1. 7590. Washington (DC), USA: World Bank. operation and Development. Available from: Available from: http://piketty.pse.ens.fr/files/ Available from: https://openknowledge. http://www.oecd.org/social/income-distribution- Alvaredoetal2016WIDworld.pdf. worldbank.org/bitstream/handle/10986/23929/ database.htm. Globalization00prospects0of0nations. Alvaredo F, Gasparini L (2014). Recent trends in pdf;sequence=1. O’Donnell O, Van Doorslaer E, Van Ourti T inequality and poverty in developing countries. (2015). Health and inequality. In: Atkinson T, In: Atkinson T, Bourguignon F, editors. Handbook Bourguignon F (2015). The globalization of Bourguignon F, editors. Handbook of income of income distribution, volume 2. Handbooks in inequality. Princeton (NJ), USA: Princeton distribution, Volume 2. Handbooks in economics. Economics. Amsterdam, Netherlands: North- University Press. Amsterdam, Netherlands: North-Holland. Holland. Deaton A (2013). What does the empirical Osberg L (2017). On the limitations of some Atkinson AB (2015). Inequality: what can be evidence tell us about the injustice of health current usages of the Gini index. Rev Income done? Cambridge (MA), USA: Harvard University inequalities? In: Eyal N, Hurst S, Norheim OF, Wealth. 63(3):574–84. https://doi.org/10.1111/ Press. https://doi.org/10.4159/9780674287013 Wikler D, editors. Inequalities in health: concepts, roiw.12256 measures and ethics. Oxford, UK: Oxford Atkinson AB, Bourguignon F (2015). Handbook University Press. Pomeranz K (2000). The great divergence: of income distribution. Vol. 2A. Elsevier. China, Europe, and the making of the modern Freeman RB (1995). Are your wages set in world economy. Princeton (NJ), USA: Princeton Atkinson AB, Brandolini A (2001). Promise Beijing? J Econ Perspect. 9(3):15–32. https://doi. University Press. and pitfalls in the use of “secondary” data-sets: org/10.1257/jep.9.3.15 income inequality in OECD countries as a case Ravallion M (2018). Inequality and globalization: study. J Econ Lit. 34(3):771–99. https://doi. Jenkins S (2015). World income inequality a review essay. J Econ Lit. 56(2):620–42. https:// org/10.1257/jel.39.3.771 databases: an assessment of WIID and SWIID. doi.org/10.1257/jel.20171419 J Econ Inequal. 13(4):629–71. https://doi. Atkinson AB, Brandolini A (2010). On analyzing org/10.1007/s10888-015-9305-3 SEDLAC (CEDLAS and the World Bank) (2018). the world distribution of income. World Bank SEDLAC (Socio-Economic Database for Latin Econ Rev. 24(1):1–37. https://doi.org/10.1093/ Kanbur R, Wang Y, Zhang X (2017). The America and the Caribbean) Statistics. Center wber/lhp020 great Chinese inequality turnaround. Bank of for Distributive, Labor and Social Studies and the Finland, Institute for Economies in Transition. World Bank. Available from: http://www.cedlas. Atkinson AB, Hasell J, Morelli S, Roser Available from: https://helda.helsinki.fi/bof/ econo.unlp.edu.ar/wp/en/estadisticas/sedlac/ M (2017). The chartbook of economic bitstream/123456789/14667/1/dp0617.pdf. estadisticas/#1496165297107-cedda6d3-6c7d. inequality. Available from: https://www. chartbookofeconomicinequality.com. Kharas H, Seidel B (2018). What’s happening UNU-WIDER (2018). World income inequality to the world income distribution? The elephant database. United Nations University World Atkinson AB, Piketty T, editors. (2007). Top chart revisited. Global Economy & Development Institute for Development Economics Research. incomes over the twentieth century: a contrast Working Paper 114. Brookings Institute. Available from: https://www.wider.unu.edu/ between continental European and English- Available from: https://www.brookings.edu/ project/wiid-world-income-inequality-database. speaking countries. Oxford University Press. wp-content/uploads/2018/04/workingpaper114- elephantchartrevisited.pdf. WID (2018). World Inequality Database. Available Atkinson AB, Piketty T, editors. (2010). Top from: https://wid.world. incomes: a global perspective. Oxford (NY), Lakner C, Milanovic B (2016). Global income USA: Oxford University Press. distribution: from the fall of the Berlin Wall to WIL (2018). World income report 2018. World the Great Recession. World Bank Econ Rev. Inequality Lab. Available from: https://wir2018. Atkinson AB, Piketty T, Saez E (2011). Top wid.world. incomes in the long run of history. J Econ Lit. 30(2):203–32. https://doi.org/10.1093/wber/ 49(1):3–71. https://doi.org/10.1257/jel.49.1.3 lhv039 World Bank (2016). Poverty and shared LIS (2018). Inequality and poverty key figures. prosperity 2016: piecing together the poverty Balestra C, Tonkin R (2018). Inequalities puzzle. The World Bank. Available from: http:// in household wealth across OECD Cross-national Data Center in Luxembourg. Available from: https://www.lisdatacenter.org/ www.worldbank.org/en/publication/poverty-and- countries: evidence from the OECD wealth shared-prosperity. distribution database. OECD Statistics data-access/key-figures/. Working Paper No. 88. Available from: Morelli S, Smeeding T, Thompson J (2015). World Bank (2018). PovcalNet (online analysis http://www.oecd.org/officialdocuments/ Post-1970 trends in within-country inequality tool). Washington (DC), USA: World Bank. publicdisplaydocumentpdf/?cote=SDD/ and poverty: rich and middle income countries. Available from: http://iresearch.worldbank.org/ DOC(2018)1&docLanguage=En. In: Atkinson T, Bourguignon F, editors. Handbook PovcalNet/. of income distribution, volume 2. Handbooks in Economics. Amsterdam, Netherlands: North- Holland.

136