By Andrew Hillman

How can a statistic increase understanding if it is too complicated to follow?

The American economist Joseph Stiglitz understands the importance of good data. In his biography, written to coincide with the acceptance of his Nobel Prize in 2001, he said, “The problem of how people form their beliefs is, of course, the central question of statistics: making inferences on the basis of limited data.”

How then, can we explain this incomprehensible passage on US income inequality from Stiglitz’s 2012 book, “The Price of Inequality”:

“The top 1 percent get in one week 40 percent more than the bottom fifth receive in a year; the top 0.1 percent received in a day and a half about what the bottom 90 percent received in a year; and the richest 20 percent of income earners earn in total after tax more than the bottom 80 percent combined.”

This is statistic-barrelling – using data to induce an emotional reaction, with little concern for whether the figures are presented in an understandable way. The over-complexity can even work to the benefit of the writer. If we cannot easily interpret the statistics, we are inclined to fall back onto our intuitive judgments – in this case, that the figures sound and feel dramatically bad and so they probably are. Why ask readers to spend time understanding a complex subject, hoping they will ultimately agree with your view, when you can arrange the numbers to make the issue appear so clear-cut that it is beyond debate?

Stiglitz’s case is extreme; it would be difficult to invent a more confusing method to explain income inequality if you were trying. But it is also emblematic of an important problem: the statistics that reach the eyes and ears of the public are failing to fulfil their most crucial function. They are not providing a better understanding of what inequality actually looks like.

Is Oxfam’s statistic an insightful measure of global inequality?

It is not just a problem of over-complication. Oxfam's famous statistic – that the richest twenty-six people in the world have as much wealth as the poorest 50% – is simple enough, but it too is a poor method for providing understanding.

Think about it: what would you want to learn from inequality data? Good data can tell all kinds of complex stories, but to act as a basis for understanding inequality it should provide us with answers to four questions:

①How fairly are income and wealth distributed?

②Is inequality getting better or worse over time?

③Are the poorest people able to meet their basic needs?

④Do the richest people have too much, to the degree that they can buy political influence and distort the democratic process?

Oxfam’s statistic fails to answer these questions. The figure does, however, have one thing going for it: it sounds extraordinary. Since the statistic was first published in 2014, the charity has combined it with powerful adjectives – extreme, shocking, grotesque – to support their argument that high global inequality is a consequence of an economic system that has allowed a small number of individuals to amass vast levels of wealth.

Easy to remember and repeat, the figure has come to dominate the public consciousness, acting as the go-to statistic for demonstrating the extent of global inequality.And if there is one guarantee regarding human behaviour, it is that when a statistic refers to a surprisingly low number of people, someone is sure to frame it in terms of the form of transport – plane, coach or even minibus – that they could all fit in.

In 2017, when just the eight richest people held as much wealth as the poorest half (the figure was later revised upwards to fifty-four), Oxfam GB’s chief executive, Mark Goldring, said "It is beyond grotesque that a group of men who could easily fit in a single golf buggy own more than the poorest half of humanity."

In recent years, the method has been copied to measure within-nation inequality, with the Equality Trust claiming that the UK’s richest six individuals own as much wealth as the poorest thirteen million, and US senator Bernie Sanders claiming that the three richest families in America hold more wealth than the poorest 50%.

With Oxfam’s statistic exerting so much influence on the way we intuitively think about global inequality, you would hope that it provides a uniquely accurate and illuminating perspective. Indeed, it is frequently wielded as an indisputable view of global inequality. In 2018, former US President Barack Obama, said:

“…a few dozen individuals control the same amount of wealth as the poorest half of humanity. That’s not an exaggeration, that’s a statistic.”

But a statistic is not evidence of irrefutable truth. It is easy to manufacture aggregated figures that misconstrue a complex topic. And readers can be led astray if they are willing to accept a statistic without a proper understanding of how the data was collected and what it represents.

How is Oxfam’s statistic calculated?

So where does Oxfam’s data come from? While most inequality research looks at income data, Oxfam’s figure is based on net wealth – the value of total assets minus liabilities. The charity uses the Forbes billionaire rankings to measure the wealth of the world’s richest individuals. To determine levels of wealth in the world’s poorest regions, Oxfam relies on the annual Global Wealth Report from the Credit Suisse Research Institute even though the report itself warns of the lack of reliable data for these areas.

The 2018 report rated the quality of data as poor or very poor for 123 nations and stated that for a further 37, “there was no sensible basis for estimating wealth.” In total these countries comprise 38% of the world population. In correspondence with the journalist Dylan Matthews at Vox, Jim Davies, one of the report’s authors, said:

“We don’t think it is a good idea to focus too much attention on what is happening to the total wealth of the bottom 50% because that area of the world wealth distribution is less well known than the higher portion.”

Another important difference between Credit Suisse’s Global Wealth Report and most studies of inequality concerns exchange rates. Most researchers make conversions to US dollars using purchasing-power parity (PPP) – a method that adjusts for the variability of prices. In contrast, the Global Wealth Report uses official exchange rates. As the report points out, this choice is appropriate for fulfilling its primary purpose – estimating the financial assets at the top of the global distribution – because wealthy individuals tend to “move their assets across borders with significant frequency.”

However, the approach means the data is ill-suited for comparing wealth between poor individuals from different countries. Since the poorest half of the global population rarely possess assets that can move across borders, their wealth relative to prices within their own country, which official exchange rates typically undervalue, is more important than how far it would stretch as US dollars in America.

Using official exchange rates also makes it difficult for Oxfam’s statistic to track how inequality is changing over time. If the world’s richest fifty individuals held as much wealth as the poorest half in one year, but that figure fell to forty the following year, we would want to be able to say that inequality had increased. But even if every persons’ wealth relative to local prices stayed constant, the calculated total wealth of the poorest 50% could change in response to fluctuating exchange rates. Davies estimates that roughly half of the 8.6% fall in the wealth of the bottom 50% from 2017 to 2018 was due to currencies in poorer nations depreciating against the US dollar.

Does it make sense to use wealth as the primary measure of global inequality?

When it comes to considering what Oxfam’s statistic represents, the issue is primarily around what we can, and more importantly cannot, infer from aggregated wealth data. We intuitively think that the people with the lowest levels of wealth are also those with the lowest quality of life, opportunity and security. But this is not the case.

The reason is debt. The ability to borrow money, typically indicative of a good standard of living, is a prerequisite for high debt. As a result, the individuals at the bottom of the global wealth distribution are indebted students, mortgage-holding homeowners and entrepreneurs from rich nations. In fact, a person from America is twice as likely to belong to the bottom 10% as a person from Bangladesh. Can we consider net wealth, on its own, an insightful measure of global inequality when it regards an indebted American student as substantially poorer than a Bangladeshi textile worker living in extreme poverty?

The complications of including debt in Oxfam’s statistic point to a broader point about wealth data: it is highly contextual.

Wealth data can be a useful indicator of the resilience of poor families against unpredictable financial shocks such as job loss or emergency medical care. This is especially true in countries with weak social safety nets. But many people have negligible wealth for reasons other than poverty. For example, low wealth can be the product of lifestyle choices: the preference for living in expensive areas and travelling frequently over saving to buy a home.

In Germany, an adult is 20% less likely to own their home than in the US. Although lower home ownership increases wealth inequality, in Germany it is primarily the product of favourable renting laws rather than greater economic injustice.

America’s relatively high level of student loan debt indicates greater wealth inequality compared to Finland, where student tuition is paid by the government. But in comparison to Zambia, where just 1% of the adult population has attained tertiary education, it is more a reflection of greater educational opportunity.

Could the wealth data used by Oxfam be applied more effectively?

Because wealth data is particularly dependent on context, when we analyse it, we want to keep apart groups of people for whom negative, negligible and positive wealth imply different things. Oxfam’s statistic, which aggregates the wealth of 3.8 billion people, fails completely in this regard.

Such excessive aggregation may be necessary for creating a statistic that is simple enough to be memorable. But if that is the case, perhaps we should not be reliant on any one figure. With a small set of carefully chosen statistics we could provide a far clearer picture of global wealth inequality.

To examine how fairly wealth is shared, we could use the Global Wealth Report's data to look at how much wealth is owned by different sections of the global distribution (with the important reminder that wealth estimates have a high level of uncertainty, particularly in poorer countries).

To see how wealth inequality is changing over time, we should rely on the Global Wealth Report’s most accurate results. The 2019 edition focused on inequality, looking at how the share of global wealth belonging to the richest 10%, 5% and 1% has changed over the past twenty years.Ideally, we would also want to understand why wealth inequality is changing. This is a more complicated question that requires country-by-country investigation of the data at a more granular level.

For example, we would want to know if wealth inequality has increased or fallen within different age cohorts (18-29, 30-44, etc.), has the distribution of wealth between urban and rural areas changed, and is wealth inequality based on gender, race and socio-economic background rising or falling?

To determine whether the poorest people have enough wealth to meet their basic needs, we could begin by looking at individuals with very low or negative wealth. We would then need to filter out people whose wealth does not provide a representative picture of their living standards: individuals with an income above a certain level, young adults who can rely on the support of wealthy parents if necessary, and people living in countries with nationalised medical care and strong social protection for the unemployed.

However, given that wealth data is less reliable for poorer families and a weaker indicator of quality of life, it makes sense to rely more heavily on income data for an understanding of how the world’s most impoverished people live.Income also has its methodological limitations. During stock market downturns, affluent capital investors can realise large negative incomes. Comparing incomes between countries is difficult because of the differences in non-discretionary expenses such as healthcare insurance. Overall, however, income has fewer intricacies; typically, an individual’s income is representative of their standard of living. The best source of data for this task is the World Bank. Since 1990, the organisation has tracked the number of individuals with an average daily income of less than $1.90 (the global definition of extreme poverty).

Finally, the amount of wealth held by the richest individuals is important. But saying that the world’s richest twenty-six people have, on average, one twenty-sixth of the total wealth held by the poorest 3.8 billion people is not particularly helpful. There must be a clearer way of presenting these figures.

For one, we should provide the actual wealth levels in monetary terms. And since we are talking about really big numbers, numbers that people struggle to wrap their heads around, we need to provide reference points for comparison.

Is the Gini coefficient a useful way of measuring inequality?

Oxfam’s statistic demonstrates that when we over-aggregate data in search of a single, memorable figure, we can end up trimming away the most valuable information.

The Gini coefficient – the measure of inequality favoured by many academics, the OECD and the World Bank – suffers from the same ailment of over-aggregation. It distils an income or wealth distribution to just one number, with 0 representing complete equality (everyone has equal income or wealth) and 1 being absolute inequality (one individual has all the income or wealth).

This method of translating monetary inequality onto a new synthetic scale is even more intangible than Oxfam’s statistic. The World Inequality Lab has been critical of the use of the Gini coefficient for precisely this reason, saying:

“[The Gini coefficient] is technical both in its calculation and in the mathematical knowledge required of the reader to interpret it. According to the World Bank, for example, the Gini index for consumption inequality in Vietnam in 2014 was equal to 0.38. Is this large or small? A Gini of 0.38 implies that the distance separating Vietnam from perfect inequality (which is 1 on the index) is 0.62. Is this an acceptable distance from perfect inequality? It is not easy for citizens, journalists, and policymakers to make sense of such a metric.”

Furthermore, the lab has argued that the coefficient “tends to downplay shifts happening at the top end and at the bottom of the distribution.” In contrast, the extremities of the distribution often dominate the debate on inequality: do the very richest have too much, do the very poorest have too little?

Can the World Inequality Lab provide a better way of looking at inequality?

How then should we measure inequality? Perhaps the problem has been over-complicated. To study inequality is to consider how evenly income or wealth are distributed, so if we want readers to understand the subject, maybe we should simply show them the distributions.

This is the approach favoured by the World Inequality Lab. A collaborative effort by over one hundred researchers in more than seventy countries, the Lab has been building a database of income distributions since its creation in 2015.

Historically, inequality research has relied on self-reported income figures collected through surveys. But the richest households are typically underrepresented in household surveys and are more likely than average to under-report their income or wealth. To combat this sampling bias, the World Inequality Lab supplements survey data with fiscal and administrative tax data.This approach has provided new perspectives and theories on inequality dynamics. For instance, in Brazil, recent household surveys had indicated that the pre-tax income share of the richest 10% had been falling over the past fifteen years, with President Luiz Inácio Lula da Silva’s Workers’ Party receiving high praise overseas for their labour market reforms, including the introduction of a minimum wage and the conditional transfer program Bolsa Familia.

But fiscal data collected by the World Inequality Lab suggests that this trend is illusory. Marc Morgan, one of the Lab’s economists researching inequality in Brazil, explained that these reforms were balanced against generous subsidies for affluent capital-owners. While incomes for the poor rose under the Worker’s Party’s tenure, it was the middle-class, not the rich, that lost out. Any significant reduction in inequality, Morgan argues, would have required reform of the country’s highly regressive tax system.

The first report based on this data, the World Inequality Report, has a prestigious cast of authors. There is Thomas Piketty, whose book “Capital in the Twenty-First Century” has sold over 1.5 million copies and was once described by the economist Paul Krugman as “the most important economics book of the year – and maybe of the decade”.Piketty’s book also has the somewhat less laudable accolade of the most-purchased, least-read book of all time based on the light-hearted Hawking Index. The index uses Amazon Kindle’s Popular Highlights function to estimate/guess how far through a book readers typically got before they gave up. Other chunky books that score highly are "A Brief History of Time" by Stephen Hawking and "Thinking, Fast and Slow" by Daniel Kahneman. And there are Berkeley economists Emmanuel Saez and Gabriel Zucman, who received national attention for assisting US Senator Elizabeth Warren in designing her wealth tax policy.

The report is unafraid of sharing detailed analysis with its readers. It primarily looks at sub-sections of the pre-tax income distribution, showing how income levels change between countries and over time for the richest 1%, richest 10%, 81st to 90th percentile etc.

The Lab's detailed approach makes for a hefty final product but it is compelling and easy to follow – certainly easier going than Piketty’s best-selling book – and it allows the authors to provide contextual information that is specific to each nation.

For example, when looking at changes in income shares in China since 1980, the report splits the data to examine the dichotomous impact of globalisation on urban and rural communities. For Russia, the report sub-categorises private wealth data to show the rapid rise in offshore wealth since the millennium; work by Zucman estimates that over half of wealth belonging to Russia’s richest 0.01% is now held in offshore accounts.

What are the limitations of the World Inequality Lab’s work?

All methods have weaknesses, and as the American economist James K. Galbraith has argued, the World Inequality Report has its own collection of limitations and caveats. He points out that tax data is only as reliable as a country’s laws and the institutions responsible for tax collection. Indeed, for Sub-Saharan Africa, where most of the world’s poorest people live, tax data is limited and unreliable, and so the World Inequality Report is entirely dependent on survey data.

Galbraith also demonstrates the difficulty in comparing tax data over time. He shows that, according to the Lab’s data, America experienced a sudden increase in inequality in the late 1980s – a finding that is likely driven not by a genuine change in incomes but by the Tax Reform Act of 1986, which made avoidance more challenging by reducing the complexity of the tax system. And apparent differences in inequality between countries could be merely the product of different tax systems. As Galbraith point out:

“whatever the relative demerits of the US tax code, the Internal Revenue Service in the US has a reputation for effective enforcement that is not to be found in Italy or France.”

The use of pre-tax income, instead of alternative measures such as disposable income or consumption levels, creates further difficulties in comparing inequality between countries with different tax policies. Although it includes transfers such as pension, unemployment and disability payments, pre-tax income does not account for the full impact of redistribution, and so inequality may be overestimated in countries with more progressive tax policies.

So perhaps, in administrative tax data, the World Inequality Lab has not discovered a new panacea to the challenge of analysing inequality data.Hopefully it is becoming clear that when it comes to global inequality, there is no single perfect method, nor an all-encompassing and definitive answer. But the feature that makes the lab’s work so uniquely valuable is not its methodology but its presentation.

In addition to the report, the Lab has published inequality data to an unprecedented level of detail and provides a sleek interface that allows users to filter information, plot it visually and download it in a raw format. This approach offers readers what is starkly lacking from most analysis of inequality: the ability to engage with and explore the data that interests them most. For example, the Luxembourg Income Study Database (considered the most extensive and harmonized archive of inequality data by the social development division at the United Nations) only provides readers with Gini coefficients and key figures.Many of these key figures are just as nebulous as the Gini coefficient. There are two forms of the Atkinson coefficient – a measure that determines which sections of the distribution contribute the most to inequality; and the 90/10, 90/50 and 80/20 ratios – which divide the income of the 90th (or 80th) percentile by the income of the 10th (or 20th or 50th) percentile.

Can Dollar Street help us understand how people around the world live?

Still a key problem remains: what good is the World Inequality Lab’s work to improve public understanding of the global income distribution without equal understanding of how income levels affect quality of life and opportunity?

For Gapminder, a Swedish think tank – or, as they prefer to be called, a ‘fact tank’ – educating people about the lives of those on the other side of the world has been the central focus for the past decade. “Factfulness: Ten Reasons We're Wrong About the World – and Why Things Are Better Than You Think”, written by Gapminder’s founder Hans Rosling and published posthumously in 2018, received endorsements from Barack Obama and Bill Gates for its ability to explain common misconceptions and provide the facts on global development in an engaging way.Gates was so enamoured with the book that he offered to pay for copies for all recent US graduates who requested the book.

And yet, Gapminder’s finest contribution to the battle against ignorance might be Dollar Street, a data visualisation tool with the emphasis on the visual. Speaking in 2017, the creator, Anna Rosling Rönnlund explained, “I sent out photographers to 264 homes in fifty countries, and in each home the photographers take the same set of photos. They take the bed, the stove, the toys and about 135 other things.” The families are surveyed to calculate their consumption levels in US dollars and then allocated a hypothetical house on Dollar Street, with the poorest living at one end and the richest at the other.

The tagline ‘see how people really live’ could not be more apt. By exploring the photos, it is easy to compare the lives of low- and middle-income families – do they have a proper toilet, where do they get water, do they own a mobile phone? By travelling along Dollar Street, reading different families’ stories and observing the frequency with which refrigerators, computers and cars appear, it becomes easier to see how a family’s income affects the way they live.

What separates insightful ways of simplifying complex subjects from misleading ones?

A defining feature for a measure of inequality appears to be its attitude towards aggregation. Dollar Street leaves the data untouched, allowing the reader to see how specific families live. In contrast, the Gini coefficient compresses the entire population into a single figure.

The aggregation of data to produce statistics shares strong similarities with the art of abstraction – the approach used by artists and designers to condense a subject down to its essential and defining features.

The Danish artist Olafur Eliasson offers an example of abstraction. He has argued that for most people, the threat posed by climate change is so intangible, remote and incomparable to their everyday experience that they struggle to properly comprehend it. With ‘Ice Watch’, his solution has been to bring the implications of climate change closer. The project takes giant blocks of ice, separated from ice sheets in Greenland, and places them in the centre of cities across Europe. ‘Ice Watch’ focuses on the audience’s relationship with climate change – Eliasson is seeking to make the melting polar ice caps feel more real and personal.

For statisticians, researchers and journalists aiming to illustrate global inequality – a subject that, like climate change, can often feel shapeless and distant – abstraction is a delicate balancing act. They need to reduce the data down to something memorable, that makes it feel tangible and important; but they must also avoid over-simplifying or misrepresenting the subject.

The Gini coefficient fails on both counts. If anything, having global inequality simplified to just one number makes it feel less real; more like an academic curiosity than the study of how people across the world live. And it does not illuminate the subject: the casual reader has no hope of understanding what a Gini coefficient of 0.45 means for people living at the top, in the middle and at the bottom of the distribution.

Oxfam’s statistic succeeds in stirring emotion within the reader. But when it comes to increasing understanding, the statistic falls short. In fact, it may even be detrimental. For one thing, it encourages a simplistic narrative that billionaires and a rigged system are the principal sources of the world’s economic injustice. This perspective can make those of us living in the middle (and when many of us identify as ‘in the middle’ what we often really mean is ‘nearly but not quite the top’) feel powerless and even absolve us of responsibility.

Secondly, it elicits extreme reactions. Many people who view global inequality as a critical issue are seduced by the statistic’s emotional draw, and therefore see it as definitive proof that the distribution of wealth must be urgently addressed. In contrast, individuals who do not view global inequality with the same urgency can easily dismiss the figure as over-aggregated or focused on the wrong indicator: wealth rather than income. And so the debate becomes further polarized when, if people were provided with more nuanced analysis, it would stand a good chance of drawing them both towards a more informed and moderate viewpoint.

Gapminder succeeds where other approaches have failed by focusing on making the subject feel as personal as possible. Dollar Street concentrates on individual families and takes photos (a very intimate form of data in itself) of the most tangible elements of their lives: their personal possessions.

With thousands of photographs from hundreds of different homes, Gapminder is not expecting the casual reader to work through every image on Dollar Street. But by presenting the data unaggregated, it maintains the connection between the reader and real families across the world, and encourages them to explore the unrefined data for themselves.

For the World Inequality Lab, exploration is also an essential concept for drawing the reader closer to the subject. Their report mimics the experience of travelling around the world, allowing readers to move from country to country, diving into the data and discovering the income dynamics in each. We find it easier to learn and remember when we can engage and interact with the data ourselves. Like Dollar Street, the World Inequality Lab capitalises on this fact to create a more memorable experience.

And yet, their work is only valuable if it can usurp alternative approaches and capture the attention of the general public. This will be challenging. Oxfam’s statistic, for instance, is firmly entrenched in the public zeitgeist. However, the way we scrutinise data is changing fast.

The coronavirus pandemic has brought with it a deluge of models, graphs, figures and projections, and for the most part, news organisations have quickly learned how to present this information clearly and accurately. Whereas we once accepted a statistic created by an advocacy group with a clear self-interest in advancing a particular perspective on global inequality, people are now developing the habit of dissecting figures produced by even the most reputable of sources such as the World Health Organisation and national governments.

If this mindset continues post-coronavirus – that statistics are only valuable when we also understand where they came from, how they were put together and what they might be missing – Dollar Street, the World Inequality Lab and the global inequality debate would be big beneficiaries.