# Understanding Income Inequality and its Implications: Why Better Statistics are Needed

## Features

• Author: Natasha Stotesbury and Danny Dorling, School of Geography and the Environment, University of Oxford
• Date: 21 Oct 2015
• Copyright: Image appears courtesy of Getty Images

A growing body of evidence points to high and rising inequality as one of our current decade’s most important global issues in light of the far-reaching implications increasingly associated with it. Recent papers detail how inequality is a source of extensive negative externalities: obstructing economic growth, hampering poverty reduction and generating large social and environmental costs (OECD, 2015; Dabla-Norris et al., 2015; Holland et al., 2009). It is thus unsurprising that President Obama described the trend towards widening inequality as the “defining challenge of our time” and, in September 2015, the United Nations enshrined inequality reduction in the list of Sustainable Development Goals.

Efforts to tackle inequality suffer from a lack of understanding certain statistical issues. This article highlights two weaknesses typically marrying studies of income inequality: ambiguous indicators and inadequate data. The aim is to underscore both the limitations of current inequality analysis and to highlight the significant relationship suggested between inequality and various environmental, educational and health-related outcomes nevertheless. We hope that this will, in turn, propel further, more rigorous, statistical research to overcome the present weaknesses, leading to a better understanding of the possible implications of income inequality.

We make the case for a clear and analytically useful indicator of income inequality – termed the ‘1st-to-10th ratio’. Calculating this ratio across twenty-five of the wealthiest economies shows that substantial geographical variations in inequality levels arise even between affluent economies. Secondly, despite enduring data limitations, coupling this measure of inequality with data on social outcomes gives rise to intriguing associations. These are in line with previous research claiming that equality improves the quality of life for almost everyone in a given population (Wilkinson, 2005; Wilkinson and Pickett, 2010; Pickett and Wilkinson, 2015).

A plethora of indicators have been developed to measure income inequality (Longford, 2014). Many suffer from significant drawbacks. For example, the Gini coefficient attaches more weight to income transfers affecting middle-income groups and hence is relatively insensitive to changes at the extremes of income distributions (Atkinson, 1970). This matters: not only is the Gini implicitly including a normative judgment, but it turns out that focusing on the tails of income distributions is crucial for understanding current inequality trends. In short, changes in the income share of the best-off and worst-off appear to be the most important potential driver of both recent changes in affluent nations’ inequality levels and of the close connection with particular outcomes. Inequalities within the population as a whole may matter less because unusually the middle 50% of earners tends to consistently capture approximately 50% of gross national income (Cobham et al., 2015).

Table 1 lays out the 1st-to-10th decile group ratio for twenty-five of the wealthiest economies. These economies were the richest in the world in 2009 with populations exceeding 1 million . Crucially, the 1st-to-10th statistic is a ratio of decile group means: the fraction of the mean equivalized household income of the top decile group to that of the bottom decile group. Therefore, simply put, this indicator is the ratio of the mean income of the richest 10% of households in a country to the mean income of the poorest 10% of households.

You may have already heard of this statistic – it is often referred to as the ’90-10 ratio’. However, many analyses tend to conflate the decile group ratio we use with the ratio of the median income of the top quintile to the bottom quintile, which is also commonly called the 90-10 ratio. The bottom quintile is the poorest fifth of households. The median of the bottom quintile is the 10th percentile. Hence the ratio of the median income of the top quintile to that of the bottom quintile is equivalent to the ratio of the 90th to the 10th percentile. That statistic completely fails to capture the extremes of income distributions. Given the aforementioned importance of measuring the incomes of the very best-off and worst-off in society, the ratio of decile group means gives more analytical leverage as an inequality statistic than the alternative 90-10 percentile ratio. In order to cut through the confusion and make a clear distinction between these crucially different inequality indicators, here we have termed the mean decile group ratio we use as the 1st-to-10th ratio.

As set out in Table 1, using the latest data to compute 1st-to-10th ratios shows that there is stark variation in inequality levels even amongst the richest economies. The 1st-to-10th ratio of the most unequal country, the United States, is almost four times that of the most equal, Denmark. The richest 10% in America capture 30% of gross national income (GNI).

Table 1.

Country Income share (%) 1st decile group Income share (%) 10th decile group GNI per household (Mean GNI) Mean income of 1st decile group Mean income of 10th decile group 1st ro 10th ratio Population
United States 1.60 30.00 $146,627.63$23,460.42 $439,882.89 18.75 320050700 Singapore 1.64 28.97$263,228.49 $43,269.31$762,672.94 17.63 5411700
Israel 1.70 25.60 $124,223.20$21,117.94 $318,011.39 15.06 7733100 Greece 2.00 25.10$54,819.88 $10,963.98$137,597.90 12.55 11128000
Spain 2.10 24.40 $77,190.48$16,210.00 $188,344.77 11.62 46927000 Italy 2.20 24.70$82, 543.19 $18,159.50$203,881.68 11.23 60990300
United Kingdom 2.70 28.00 $101,135.06$27,306.47 $283,178.17 10.37 63136300 Portugal 2.60 25.90$54,823.77 $14,254.18$141,993.56 9.96 10608200
South Korea 2.20 21.90 $78,768.26$17,329.02 $172,502.49 9.95 49262700 Japan 2.91 25.71$102,986.86 $29,969.18$264,779.22 8.84 127143600
Australia 2.80 24.40 $195,787.67$54,820.55 $477,721.91 8.71 23342600 Canada 2.80 24.20$137,901.87 $38,612.52$333,722.53 8.64 35181700
New Zealand 3.10 25.70 $121,213.24$37,576.10 $311,518.03 8.29 4505800 France 3.40 25.30$102,129.00 $34,723.86$258,386.37 7.44 64291300
Ireland 3.20 23.80 $120,647.78$38,605.37 $287,127.44 7.44 4627200 Austria 3.10 21.60$115,352.70 $35,759.34$249,161.83 6.97 8495100
Switzerland 3.50 23.20 $235,426.70$82,399.35 $546,189.94 6.63 8077800 Netherlands 3.40 22.40$114,316.36 $38,867.56$256,068.65 6.59 16759200
Germany 3.60 23.50 $104,337.11$37,561.36 $245,192.21 6.53 82726600 Sweden 3.50 21.90$128,851.35 $45,097.97$282,184.46 6.26 9571100
Norway 3.30 20.60 $237,978.94$78,533.05 $490,236.62 6.24 5042700 Belgium 3.60 20.80$113,674.25 $40,922.73$236.442.44 5.78 11104500
Finland 3.90 21.50 $103,472.11$40,354.12 $222,465.04 5.51 5426300 Slovenia 3.70 20.00$55,956.64 $20,703.96$111,913.28 5.41 2072000
Denmark 4.00 20.80 $147,822.55$59,129.02 \$307,470.90 5.20 5619100

(Sources: OECD (2015), UNU-WIDER (2014), UNDP (2014), UNSD (2015), Eurostat (2014)

Wide geographical variation in inequality levels can be exploited as a natural experiment in order to begin to assess the possible implications of inequality for societies. Yet, even after inequality indicators are clarified, statistical analyses of inequality are often severely limited by the quality and availability of income data. Data on bottom and top incomes is particularly weak, for instance, because top earners considerably under-state their incomes in surveys, whilst tax evasion limits the reliability of tax receipts (Cobham et al., 2015). Indeed, the results in Table 1 stand in contrast to the findings of alternative studies, although the rankings generated are similar. For instance, data from the New York Times Income Distribution Database (2014) generates a 1st-to-10th ratio of 17.35 for the UK and 14.54 for Canada, whilst Ballas et al.’s (2009) study of Japan produces a value of 7.28. Furthermore, data on social outcomes are often missing, inaccurate or cross-nationally incomparable. Using the data summarized over the following three tables, the association between income inequality and various quality-of-life indicators is graphed, beginning with five environmental measures.

Table 2.

Country 1st to 10th Ratio Water consumption Meat consumption Carbon dioxide emissions Consumption of motor gasoline Waste
United States     18.75                219.48 117.60 17.02 10.08 717
Singapore     17.63                239.80 71.10 4.32 1.35 ..
Israel 15.06                107.82 102.00 8.95 2.93 633
Greece 12.55                116.02 80.60 7.56 2.06 502
Spain 11.62                115.85 93.10 5.79 0.84 446
Italy 11.23                151.90 86.60 6.70 1.25 485
United Kingdom 10.37 101.45 82.80 7.09 1.77 489
Portugal 9.96 88.20 90.30 4.71 0.85 433
South Korea 9.95 147.22 62.20 11.84 1.42 365
Japan 8.84 121.00 48.80 9.29 2.77 356
Australia 8.71 182.46 121.10 16.52 5.23 637
Canada 8.64 178.20 92.20 14.14 8.55 382
New Zealand 8.29 185.44 126.90 7.12 4.31 621
France 7.44 86.82 88.70 5.19 0.99 542
Ireland 7.44 138.85 80.50 7.88 2.19 589
Austria 6.97 72.88 106.40 7.77 1.65 577
Switzerland 6.63 90.15 74.70 4.63 2.93 707
Netherlands 6.59 73.25 72.70 10.06 2.04 528
Germany 6.53 61.19 87.90 8.92 1.88 602
Sweden 6.26 96.78 81.90 5.52 2.43 460
Norway 6.24 178.41 65.90 9.19 1.60 499
Belgium 5.78 65.09 76.80 8.85 0.94 442
Finland 5.51 74.22 74.40 10.16 2.42 494
Slovenia 5.41 81.73 82.00 7.50 1.83 412
Denmark 5.20 64.31 75.20 7.25 2.04 746

Sources: Fig. 2.1, FAO (2015); Fig. 2.2 UNEP (2015), Fig. 2.3 World Bank (2015a), Fig. 2.4 International Energy Statistics (2015), Fig. 2.5 OECD (2014). Indicators: Water consumption - municipal water withdrawal refers to the annual quantity of water withdrawn primarily for the direct use by the population. Meat consumption - the total amount of meat available for human consumption per year in kilograms, excluding fish and seafood consumption. Note: a value of 82 relates to approximately eating one beef steak a day every day all year. Carbon dioxide emissions - emissions stemming from the burning of fossil fuels and the manufacture of cement. Waste - municipal waste collected per capita served expressed in kilograms. Municipal waste consists to a large extent of waste generated by households (and as such is distinct from measures of agricultural and industrial waste).

Across the five charts the correlations are by no means conclusive. Indeed, any positive correlation between inequality and environmental degradation is considerably driven by the United States, which is a consistent outlier. With bubble size corresponding to population, the US is easy to identify as the largest bubble. Dismissing the United States as an idiosyncratic case however would be dangerous given the global trend towards increasing inequalities. It is possible that the extent of environmental degradation in the United States foreshadows what will occur in other countries as, and if, inequality rises. Furthermore, if the United States were not shown as a single bubble in the chart, and rather data for each state within the United States were shown, then the patterns would appear to be more consistent (Wilkinson and Pickett 2010).

Data on indicators such as water consumption are notoriously unreliable given variations in national measurement methodologies. Cultural differences play a role too. For example, the Japanese’s relatively low consumption of meat is potentially at least in part a result of the relatively high consumption of fish in Japan. In response, Table 3 below summarizes data from the Programme for International Student Assessment (PISA) and the PIAAC Survey of Adult Skills, which are far more suited to cross-national analysis. The former is the world’s global metric for quality, equity and efficiency in school education. Approximately 510,000 students completed the assessment in 2012 representing about 28 million 15 year olds in the schools of the 65 participating countries. Similarly, the Survey of Adult Skills is an internationally standardised survey conducted in 33 countries.

The six figures generated from the more comparable and unadjusted PISA and PIAAC data produce intriguing results. The correlation between inequality levels and mathematical, literacy and problem solving ability appear to be significantly stronger when older respondents are considered. This is interesting because it hints at the possibility that more unequal countries’ education systems fail to foster long-term understanding to the same extent that education in more equal countries appears to have a longer lasting effect on young peoples’ ability. Even if future studies aiming to unearth a causal relationship are inconclusive, it is still startling that the average 16-24 year old in the United Kingdom would, given four illustrations of the plan of a cardboard box, fail to identify the correct shape. Similarly, they would be unable to use a bibliographic search on a library website to find the author of a titled book.

Table 3.

Country 1st to 10th Ratio Population Maths ability at age 15 Maths ability at ages 16-24 Literacy ability at ages 16-24 Problem Solving ability at age 15 Problem Solving ability at ages 16-24
United States 18.75 32005000 481 240.0 498 260.9 50.5
Singapore 17.63 5411700 573 542
Israel 15.06 7733100 466 486
Greece 12.55 11128000 453 477
Spain 11.62 46927000 484 254.3 488 263
Italy 11.23 60990300 485 250.8 490 260.2
United Kingdom 10.37 63136300 494 253.1 499 262.1 50.5
Portugal 9.96 10608200 487 488
South Korea 9.95 49262700 554 280.9 536 292.9 69.5
Japan 8.84 127143600 536 280.5 538 296.5 72.2
Australia 8.71 23342600 504 269.0 512 282.9 61.1
Canada 8.64 35181700 518 267.1 523 274.4 59.0
New Zealand 8.29 4505800 500 512
Ireland 7.44 4627200 501 257.6 523 270.2 52.3
France 7.44 64291300 495 262.9 505 274.6
Austria 6.97 8495100 506 277.4 490 275.9 58.9
Switzerland 6.63 8077800 531 509
Netherlands 6.59 16759200 523 283.0 511 292.1 64.1
Germany 6.53 82726600 514 273.9 508 277.7 58.1
Sweden 6.26 9571100 478 278.2 483 282.8 66.5
Norway 6.24 5042700 489 269.2 504 273.3 61.1
Belgium 5.78 11104500 515 509 64.3
Finland 5.51 5426300 549 284.8 524 296.7 66.7
Slovenia 5.41 2072000 501 481
Denmark 5.20 5619100 500 272.5 496 275.7 58.2

Sources: Fig. 3.1, Fig. 3.3, Fig. 3.5: OECD (2012); Fig. 3.2, Fig. 3.4, Fig. 3.6 OECD (2013) Tables A2.7, A2.3, A2.10b respectively. Indicators: Maths ability at age 15 - mean PISA score for Mathematics in 2012 survey. Maths ability at ages 16-24 - adjusted mean proficiency in numeracy among 16-24 year olds (the adjusted mean includes adults who were not able to provide background information because of language difficulties, or learning or mental disabilities). Literacy ability at age 15 - mean PISA score for Reading in 2012 survey. Literacy ability at ages 16-24 - adjusted mean proficiency in literacy among 16-24 year olds. Problem solving ability – percentage of population at or above Level 2 (baseline level) in problem solving ability.

The final table in this short article looks to data on indicators of health and social-cohesion. These categories are particularly illustrative of the data limitations faced by inequality researchers. For example, international statistics on mental health disorders are far from complete, whilst the level of anxiety disorders recorded is highly dependent on the extent to which countries attempt to diagnose them in the first place. Nevertheless the potential for a link to be demonstrated between greater income inequality and negative social outcomes remains, especially in light of the considerably higher rates of crime, murder, infant mortality and health disorders in the United States.

Table 4.

Country 1st to 10th Ratio Population Infant mortality rate Mental health disorders Anxiety disorders Crime rates Murder rates
United States 18.75 320050700 7 47.4 1920.1 3531.3 4.7
Singapore 17.63 5411700 3 20.0 314.2 0.2
Israel 15.06 7733100 4 17.6 13.8 2767.7 1.8
Greece 12.55 11128000 4 57.4 2198.1 1.7
Spain 11.62 46927000 4 19.4 134.4 798.7 0.8
Italy 11.23 60990300 5 18.1 229.9 1603.7 0.9
United Kingdom 10.37 63136300 5 356.1 3642.0 1.0
Portugal 9.96 10608200 4 53.3 1872.4 1.2
South Korea 9.95 49262700 4 254.7 5230.4 0.9
Japan 8.84 127143600 3 18.0 335.3 206.4 0.3
Australia 8.71 23342600 4 119.8 1675.6 1.1
Canada 8.64 35181700 5 126.3 1682.3 1.6
New Zealand 8.29 4505800 6 39.3 25.4 4352.8 0.9
Ireland 7.44 4627200 4 19.1 1213.4 1.2
France 7.44 64291300 4 37.9 439.4 1720.3 1.0
Austria 6.97 8495100 4 32.9 3089.3 0.9
Switzerland 6.63 8077800 4 33.2 1689.9 0.6
Netherlands 6.59 16759200 4 31.7 87.1 1895.9 0.9
Germany 6.53 82726600 4 25.2 428.8 2531.0 0.8
Sweden 6.26 9571100 3 45.1 1121.4 0.7
Norway 6.24 5042700 3 36.7 738.7 2.2
Belgium 5.78 11104500 4 29.1 41.2 2364.0 1.6
Finland 5.51 5426300 3 19.2 5093.0 1.6
Slovenia 5.41 2072000 3 9.7 886.7 0.7
Denmark 5.20 5619100 4 21.7 1044.2 0.8

Sources: Fig. 4.1 World Bank (2015b), Fig. 4.2 WHO (2011), Fig. 4.3 WHO (2014), Fig. 4.4 UNODC (2014), Fig. 4.5 UNODC (2015). Indicators: Infant mortality rate - the number of infants dying before reaching one year of age, per 1,000 live births in a given year. Mental health disorders - disorder prevalence as a percentage indicates the estimated lifetime prevalence of any DSM-IV/CIDI disorder (as classified by the WHO to allow cross national comparisons). Anxiety disorders - DALYs (Disability-Adjusted Life Years) in '000s (one DALY can be thought of as one lost year of healthy life). Crime rates - number of persons brought into formal contact with the police and/or criminal justice system for all crimes per 100,000 of population. Murder rates - intentional homicide rate per 100,000 population.

In summary, this article maintains that more statistical analysis is crucial for understanding economic inequality and its implications. Firstly, income inequality statistics must be explicit as to whether or not they capture the incomes of the very best-off and worst-off in societies, given that the relative income shares of these groups are the driving force behind current inequality trends. Secondly, severe data limitations currently hamper the ability of statisticians to research the potential implications of inequality. Nevertheless, even the handful of figures in this article support the conclusion that inequality tends to be related to a wide range of negative social outcomes. It is of course possible that many intervening factors are also key, such as levels of trust, good governance and the historical and political circumstances that have resulted in the wide variations shown here - as well as many other factors we have not considered. Whilst it is stressed that this analysis is only showing suggestions of correlations, these are nonetheless necessary prerequisites for possible causal “dose-response” relationships to be discovered. Hence they support calls for better data collection to be carried out, as a precursor to more sophisticated statistical analysis.

References

Atkinson, A. (1970). “On the Measurement of Inequality.” Journal of Economic Theory (2): 244-263.

Ballas, D., Dorling D., Nakaya, T., Tunstabll, H. and Hanaoka, K. (2013) ‘Income inequalities in Japan and the UK’, Journal of Social Policy and Society.

Cobham A., Schlogl L. and Sumner A. (2015) “Inequality and the Tails: The Palma Proposition and Ratio revisited”. UN/DESA Working Paper No. 143.

Dabla-Norris E., Kochhar K., Ricka F., Suphaphiphat, N. and Tsounta, E. (2015). Causes and Consequences of Income Inequality: A Global Perspective. IMF Staff Discussion Note [Online]. Available from: www.imf.org/external/pubs/ft/sdn/2015/sdn1513.pdf.

Eurostat (2014) Number of private households [Online]. Accessed August 2014. Available from: www.appsso.eurostat.ec.europa.eu/nui/setupDownloads.do.

FAO (2015) AQUASTAT database [Online]. Accessed June 2015. Available from: www.fao.org/nr/water/aquastat/main/index.stm. Data are from latest year available spanning 2002-2013.

Holland, T., Peterson, G., & Gonzalez, A. (2009). “A Cross-National Analysis of How Economic Inequality Predicts Biodiversity Loss”. Conservation Biology 23(5): 1304-13013.

IMF (2015) World Economic Outlook Database [Online] Accessed June 2015. Available from: www.imf.org/external/pubs/ft/weo/2015/01/weodata/index.aspx.

International Energy Statistics (2015) US Energy Information Administration [Online]. Accessed June 2015. Available from: www.eia.gov/cfapps/ipdbproject/iedindex3.cfm?tid=5&pid=62&aid=2&cid=regions&syid=2009&eyid=2013&unit=TBPD. Data are from 2013 for all countries except Singapore (2012) (latest year available).

Longford, N.T. (2014). Statistical Studies of Income, Poverty and Inequality in Europe: Computing and Graphics in R using EU-SILC (London: Chapman and Hall).

New York Times Income Distribution Database (2014) assembled by Janet Gornick (LIS), Thierry Kruten (LIS), Branko Milanovic (LIS), David Leonhardt (NYT), and Kevin Quealy (NYT) [Online]. Accessed July 2015. Available from: www.lisdatacenter.org/resources/other-databases.

OECD (2012) Pisa 2012 Results [Online]. Accessed September 2015. Available from: http://www.oecd.org/pisa/keyfindings/pisa-2012-results-overview.pdf.

OECD (2013) Skills Outlook [Online]. Accessed September 2015. Available from: www.oecd.org/site/piaac/Skills%20volume%201%20(eng)--full%20v12--
eBook%20(04%2011%202013).pdf. Original Source: Survey of Adult Skills (PIAAC) (2012).

OECD (2014) StatsExtracts Datebase [Online]. Accessed June 2015. Available from: www.stats.oecd.org. Data are from latest year available spanning 2009-2013. Data missing for Singapore.

OECD (2015) In It Together: Why Less Inequality Benefits All (OECD Publishing: Paris).

Pickett, K. E and Wilkinson, R.G. (2015). “Income Inequality and Health: A Causal Review.” Social Science and Medicine 128: 316-326.

UNDP (2014) Human Development Reports Data [Online]. Accessed June 2015. Available from: www.hdr.undp.org/en/content/population-total-both-sexes-thousands. [Original source: UNDESA (2013a). 2012 Revision. World Population Prospects. Population Division Database. Detailed Indicators. Accessed June, 14.2013. http//:esa.un.org/unpd/wpp/unpp/ ]

UNEP (2015) The UNEP Environmental Data Explorer [Online]. Accessed June 2015. Available from: www.geodata.grid.unep.ch/. Original source: Food and Agriculture Organization of the United Nations (FAO) - FAOStat. http://ede.grid.unep.ch. Data are from 2011 (latest year available for all countries).

UNODC, United Nations Office on Drugs and Crime (2014) Global Study on Homicide 2013 [Online]. Accessed July 2015. Available from: http://www.unodc.org/documents/data-and-analysis/statistics/GSH2013/2014_GLOBAL_HOMICIDE_BOOK_web.pdf. Data are from latest year available spanning 2011 to 2012.

UNODC, United Nations Office on Drugs and Crime (2015) Statistics on Crime [Online]. Accessed July 2015. Available from: http://www.unodc.org/unodc/en/data-and-analysis/statistics/crime.html. Data are from the latest year available spanning 2004 to 2013.

UNSD, United Nations Statistics Division (2015) UNSD Demographic Statistics [Online]. Accessed October 2015. Available from: www.data.un.org/Data.aspx?d=POP&f=tableCode:326 .

UNU-WIDER (2014) World Income Inequality Database (WIID3.0b) [Online]. Accessed October 2015. Available from: www.wider.unu.edu/research/Database/en_GB/wiid/.

WHO, World Health Organisation (2011) WHO World Health Mental Survey: Global Perspectives on the Epidemiology of Mental Disorders. Kessler, R. and Üstün, T. (eds.). Chapter 24.

WHO, World Health Organisation (2014) Global Health Estimates 2014 Summary Tables [Online] Accessed July 2015. Available from: http://www.who.int/healthinfo/global_burden_disease/estimates/en/index2.html.

Wilkinson, R. (2005). The Impact of Inequality (London: Routledge).

Wilkinson, R. and Pickett, K. (2010). The Spirit Level (London: Allen Lane).

World Bank (2015a) World Development Indicators [Online]. Accessed July 2015. Available from: www.databank.worldbank.org/data/home.aspx. Original source: Carbon Dioxide Information Analysis Center, Environmental Sciences Division, Oak Ridge National Laboratory, Tennessee, United States. Data are from 2011 (latest year available for all countries).

World Bank (2015b) World Development Indicators [Online]. Accessed July 2015. Available from: http://wdi.worldbank.org/table/2.21#. Original source: Estimates developed by the UN Inter-agency Group for Child Mortality Estimation (UNICEF, WHO, World Bank, UN DESA Population Division) at www.childmortality.org.

View all

View all