Inequality
in a Hurry

The Whole Thing in One Page

Inequality is usually pictured as a ladder. Some people stand higher, some lower, and the political argument begins over how far apart the rungs should be. That picture is too static. Economic inequality is better understood as a set of flows feeding stocks, with gates between them. Income arrives as wages, profits, rents, interest and transfers. Some is consumed. Some becomes wealth. Wealth then earns returns, absorbs shocks, buys time, finances education, opens neighbourhoods and businesses, and can be passed on. Opportunity shapes who reaches each gate, and the outcome of one generation becomes part of the starting conditions of the next.

That is why income inequality, wealth inequality and inequality of opportunity are related but not interchangeable. A country can have a moderate spread of annual incomes and a large concentration of wealth. It can have high inequality but substantial movement between generations, or lower inequality with rigid barriers inside particular groups. A single Gini coefficient cannot tell you which machine you are looking at.

The causes are plural. Technology can raise the value of scarce skills and replace routine tasks. Education can widen access or ration advantage. Global trade can lift national incomes while imposing concentrated losses on particular workers and places. Firms divide revenue between labour and capital under rules shaped by competition, bargaining and institutions. Discrimination can alter hiring, pay, credit and housing. Inheritance transfers claims accumulated under yesterday's economy into tomorrow's race. Taxes, benefits, public services, labour law, planning rules and political institutions alter the distribution before and after money reaches a household.

The consequences are plural too. Some inequality accompanies innovation, risk, effort, scarcity and the ordinary fact that people make different choices. Flatten every reward and you can weaken useful incentives. But inequality can also waste talent when children cannot buy the inputs that turn ability into achievement, reduce bargaining power when losing a job is ruinous for one side and survivable for the other, and turn wealth into political and market power. The relevant question is rarely whether inequality is good or bad. It is what produced this inequality, what it does next, and whether the mechanism is worth the price.

The global story makes the argument harder. Hundreds of millions have escaped extreme poverty over recent decades, driven heavily by growth in Asia, while inequality inside many countries remains high. A person can become richer in absolute terms while falling behind the richest people around them. Global inequality can fall even as domestic inequality rises. Both statements can be true because they answer different questions.

Policy therefore begins with diagnosis, not a preferred tax rate. Education, competition, housing, health, early childhood, labour institutions, anti-discrimination enforcement, transfers and taxation act at different points in the machine. Their effects depend on incentives, incidence, administrative capacity and political response. The choice is not equality or growth. It is which inequalities a society is willing to create, tolerate, insure against, or reproduce.

The first unequal outcome becomes part of the next unequal starting point. That loop is the subject.

That is the book.

Why You Should Care

In the United States, a widely used measure of absolute mobility once produced a startling number. About 90 per cent of children born in 1940 grew up to have higher inflation-adjusted household incomes than their parents had at the same age. For children born in the 1980s, the estimate was around half. The result does not describe every country, and it does not measure movement in rank. It does show why inequality cannot be understood from a payslip alone. The distribution facing one generation helps shape the possibilities of the next.

That is the first reason to care about inequality. It changes what an outcome means. A large reward can reflect useful scarcity, exceptional effort, luck, inherited advantage, monopoly, discrimination, bargaining power, or several of these at once. The number on the payslip does not reveal the mechanism that produced it.

The second reason is that inequality compounds. A high income is useful today. Wealth changes tomorrow. A household with assets can wait for the right job, move for an opportunity, finance training, survive illness, put down a housing deposit, lend to a child or start a firm. A household with no buffer may reject a better long-term option because the short-term risk is unbearable. Two people facing the same nominal choice therefore do not face the same practical choice. Money buys consumption, but wealth also buys room to make decisions badly and recover.

The third reason is political. Economic resources do not remain politely inside economics. They pay for campaigns, lobbying, litigation, media, expertise, associations, neighbourhood access and time. That does not mean every rich person controls politics or every unequal society is captured. It means a distribution of resources can become a distribution of voice. If wealth can help shape the rules that govern wealth, inequality is capable of reproducing itself through institutions rather than through markets alone.

Then comes the uncomfortable part. Inequality is not synonymous with poverty, and reducing one does not mechanically reduce the other. If the incomes of the poorest double while those of the richest quadruple, poverty may fall while inequality rises. A policy that shrinks the top without improving the bottom may reduce inequality while leaving deprivation untouched. A useful moral argument must therefore keep at least three questions separate: how much do people have, how far apart are they, and how much do their origins determine where they end up?

The scale is concrete. The World Inequality Report 2026 estimates that in 2025 the global top tenth received 53 per cent of pretax national income and owned about three quarters of household wealth, while the bottom half received 8 per cent of income and owned about 2 per cent of wealth. These are reconstructed distributional estimates rather than a census of every household. The broad contrast is still hard to miss: accumulated ownership is more concentrated than annual income.

Yet the same era also contains an enormous fall in extreme poverty. Under the World Bank's current international poverty line, its March 2026 data vintage estimates 847 million people in extreme poverty in 2024, 10.4 per cent of the global population, with a nowcast of 10.0 per cent for 2026. The series has been revised as prices, surveys and purchasing-power comparisons change, so the exact historical count is not fixed forever. The direction is not in doubt: absolute deprivation can fall at the same time as large relative gaps persist.

So inequality is not one graph with one verdict. It is a system connecting reward, security, opportunity and power across time.

Which gap? Measured how? Before or after tax and transfers? Between households or individuals? Across a country or the whole world? Over one year or a lifetime? Does the gap reflect a productive difference, a barrier, a rent, a historical exclusion or a temporary transition? What happens to mobility if it persists? Which intervention reaches the mechanism rather than the symptom?

Those questions are harder than declaring inequality either a scandal or a necessary price of freedom. They are useful.

The Core Ideas

Inequality Is a Distribution, Not a Number

The first mistake is asking how unequal a society is before asking unequal in what.

Start with income. Economists usually distinguish market income from disposable income. Market income includes wages, self-employment earnings and returns to capital before most government redistribution. Disposable income adds cash transfers and subtracts direct taxes. If two countries have similar wage gaps but one taxes high incomes more progressively and pays larger transfers, their market-income inequality can look similar while household living standards after redistribution differ sharply.

Then change the unit. Individual earnings are not household income. A person with no wage may live with a high earner. A high earner may support several children. Household comparisons therefore adjust for size, often using an equivalence scale because two people living together do not need twice the housing, heating and appliances of one. That adjustment is defensible and consequential. Change it and measured inequality changes.

Now change the period. Annual income jumps around. Students appear poor before entering high-paying careers. Self-employed earnings swing. People retire. Temporary unemployment creates a sharp one-year drop that may not represent lifetime resources. Permanent inequality and short-run volatility are different problems, even when both widen an annual distribution.

Wealth is different again. Wealth is the value of assets minus debts: housing, businesses, pensions, shares, bonds, land and other claims, less mortgages and loans. It is much more concentrated than income because it is accumulated over years and can earn returns. A household with a modest income can have substantial housing or pension wealth. A young professional can earn well and have negative net wealth after student or business borrowing. Treating income and wealth as substitutes produces nonsense.

Opportunity is harder because it is partly counterfactual. You cannot directly observe the income someone would have earned with different parents, health, race, school, neighbourhood or childhood wealth. Researchers infer opportunity from patterns: the strength of the link between parents' and children's outcomes, differences by place or group, adoption and migration designs, school effects, experiments and natural experiments. These methods illuminate mechanisms without reducing a life to one coefficient.

How, then, do we compress a distribution?

The best-known answer is the Gini coefficient. Imagine lining everyone up from poorest to richest and plotting the cumulative share of income they receive. Perfect equality would follow a diagonal: the bottom 20 per cent receive 20 per cent, the bottom 50 receive 50, and so on. The more the actual curve bows below that line, the greater the inequality. The Gini summarises the area between the two. Zero means complete equality. One, or 100 on another common scale, means one unit receives everything.

It is elegant and incomplete. Two societies can have the same Gini with different distributions. One may have a compressed middle and a thin ultra-rich top. Another may have a broad top and a deeply poor bottom. If you care about poverty, the bottom matters more. If you care about political concentration, the top may matter more. If you care about the middle class, neither tail alone is enough.

So analysts also use percentile ratios and income shares. The 90:10 ratio compares someone near the top with someone near the bottom. The top 1 per cent share asks how much income or wealth accrues to a small elite. The Palma ratio compares the top 10 per cent's income share with the bottom 40 per cent's, deliberately focusing on the tails. Each measure answers a different question.

There is a deeper point. Measurement is not neutral description. It encodes what kind of inequality you think matters. A government that reports only the median can miss top concentration. One that reports only the Gini can miss poverty. One that reports only pretax income can miss public redistribution. One that reports only cash income can miss public healthcare and education. One that reports household averages can hide unequal control of resources inside households.

The condition established here will return at the end of the seven ideas. Once inequality is understood as a distribution produced through multiple stages, the policy problem changes. You stop asking for the lever that "reduces inequality" and start asking where in the distributional process a particular inequality is created.

Income Flows, Wealth Stores, and Wealth Pushes Back

A wage is a flow. A house is a stock. That distinction explains why wealth inequality can become much larger than income inequality and why annual earnings never tell the whole story.

Suppose two households both receive £60,000 a year. One owns a mortgage-free home and £300,000 in financial assets. The other rents and has £5,000 in savings. Their current incomes are identical. Their exposure to the future is not.

The first household can withstand a redundancy, borrow cheaply against collateral, fund a move, pay a deposit for a child, wait through a weak business year or invest in assets that rise in value. The second may need uninterrupted cash flow. A shock that is inconvenient to one can be destructive to the other. Wealth therefore functions as insurance, option value and bargaining power as well as a source of consumption.

The arithmetic of accumulation reinforces the gap. If a household saves part of its income and earns a return, its assets can compound. A household with no surplus cannot begin the process. A household carrying expensive debt can compound in the opposite direction, sending part of future income to earlier creditors. Returns are uncertain, taxes vary, asset prices fall as well as rise, and no law guarantees that wealth outruns wages. The mechanism still matters: stocks carry the effect of past flows into future periods.

Asset ownership also changes who benefits from economic growth. When house prices rise, homeowners gain on paper while renters face no matching asset appreciation and may face higher rents or deposits. When share prices rise, gains accrue in proportion to ownership. If business equity is concentrated, a broad increase in corporate profits can raise national income while distributing much of the resulting wealth to a narrow group.

This is why the labour-capital split matters, although it is not a complete theory of inequality. National output ultimately becomes income to workers, owners of capital, governments and other claimants. The ILO estimates that the global labour income share fell by about 1.6 percentage points between 2004 and 2024. That aggregate does not identify every winner and loser. Workers can own pensions and shares, owners can also work, and national patterns differ. It does show why a modest change in the division of a vast income pool can matter at scale.

Wealth also enters before the labour market. Families buy homes near sought-after schools, pay for tutoring, support unpaid internships, finance university living costs and transfer housing deposits. The advantage need not take the form of a trust fund. It can arrive as reduced risk. A graduate whose parents can cover three months of rent can search longer for a job matching her skills. Another may accept the first offer. Both appear to have made free choices. Their fallback positions differed.

Inheritance makes the mechanism explicit. Economists distinguish the direct transfer of assets from the broader transmission of advantage. Parents transmit genes, health, language, expectations, networks, neighbourhoods, schooling and money. An inheritance received at fifty may do little to explain the heir's first job, while parental wealth used at eighteen may have shaped education and housing for decades before any estate was divided.

The policy argument becomes difficult because wealth is both reward and input. Saving, investment and entrepreneurship create assets that societies usually want people to build. The ability to pass something to children is itself a powerful motive. Yet when accumulated claims purchase better starting positions, yesterday's outcomes become part of tomorrow's opportunity structure.

A flow becomes a stock. The stock changes the next flow. That is the compounding engine of inequality.

Markets Reward Scarcity, but Institutions Decide the Bargain

A popular account says pay equals productivity. A rival account says pay equals power. Both contain something important and both are too clean.

Begin with scarcity. If a skill becomes more valuable and remains hard to supply, employers bid more for it. The expansion of computing raised demand for many analytical, managerial and technical tasks while automating some routine work. Education premiums increased in several countries as demand for skilled labour moved faster than supply. This is the core of skill-biased technological change.

But technology does not arrive with a wage distribution attached. A machine can complement one worker, replace another, create entirely new tasks, reduce the value of a credential, increase the scale of a successful firm, or make supervision easier. Whether productivity gains reach workers depends on what happens inside firms and labour markets.

Consider a simple surplus. A worker can produce £50 an hour of additional value in a job. Her next best option pays £20. The employer's next best candidate would generate less value or cost more. Any wage between the worker's outside option and the value created can leave both sides better off than walking away. Economics does not supply one inevitable division of that surplus. Search costs, information, contracts, competition, norms and bargaining determine where inside the range the wage lands.

This is where institutions enter. Minimum wages set a floor. Collective bargaining can move negotiations from one worker against one employer to groups on both sides. Employment protection changes the cost of dismissal. Non-compete clauses can affect worker mobility where they are enforceable. Pay-transparency rules alter information. Occupational licensing can protect consumers and restrict entry. Immigration rules affect labour supply. Competition policy can affect employer concentration. None of these acts outside the market. They help constitute the market in which bargaining occurs.

Unions are a clear case because their distributional mechanism is visible. Collective bargaining can raise wages for covered workers and compress wage differences, particularly toward the lower and middle parts of the distribution. It can also create costs, rigidities or insider-outsider problems depending on design and context. The ILO's current work treats declining bargaining power as one contributor to the fall in labour income share, alongside technology and globalisation. That is a causal contribution, not an instruction to attribute every wage trend to union decline.

Employer power matters too. In the textbook case of many firms competing for workers, an employer that underpays loses staff. Where workers face few local alternatives, costly moves, licensing barriers, tied benefits, immigration restrictions, poor information or non-compete agreements, the threat of exit weakens. Economists call the broader condition monopsony power even when there is more than one employer. The important feature is not a literal company town. It is an upward-sloping labour supply to a particular employer, meaning the firm has some discretion over wages.

The same logic appears at the top of firms. Executive pay depends on firm scale, scarce talent, governance, bargaining and norms. Superstar markets can produce enormous rewards when technology allows the best performer, platform or manager to serve a global market. But attributing every high income to marginal productivity is as weak as attributing it all to extraction. The causal mix differs across occupations and institutions.

Education interacts with this bargain rather than sitting outside it. More schooling can raise individual productivity and signal ability. Expanding access can increase the supply of scarce skills and reduce some wage premiums. Yet education cannot guarantee equality if the economy creates many low-paid jobs, if credentials escalate, if school quality differs sharply, or if families purchase additional advantages around the public system.

The central correction is therefore to the word "market" as if it were an actor. Markets do not walk into a room and set a wage. People and firms make offers under legal and institutional constraints, with unequal alternatives and information. Scarcity creates a range of plausible rewards. Institutions help decide how the surplus is divided.

Opportunity Is Produced Long Before the Job Interview

Inequality of opportunity sounds like the gentler part of the debate. People who disagree fiercely about outcomes often agree that children should have a fair chance. Agreement weakens when the phrase is made concrete.

What counts as a circumstance for which a child should not be held responsible? Parents' income is obvious. What about neighbourhood, school quality, disability, race, sex, family stability, inherited wealth, parental education, nutrition, pollution exposure, social networks or language? By the time a person is old enough to make choices that adults regard as autonomous, many of the inputs into those choices have already been distributed.

Early childhood matters because capabilities are cumulative. Health affects attendance and cognition. Language exposure shapes later learning. Stable housing reduces disruption. Good schools can build skills and expectations. Safe neighbourhoods affect stress, peers and access to institutions. None of these determines a life, and individuals regularly outrun statistical predictions. At population scale, their effects appear in distributions.

Place provides unusually strong evidence because researchers can compare children who move. Work using US tax records has shown large differences in adult outcomes depending on where children grow up, and movers tend to acquire more of a destination area's effect the younger they arrive. That pattern is hard to explain with a story in which geography merely labels fixed family traits. Neighbourhoods are part of the production function for opportunity.

Mobility itself needs two definitions. Relative mobility asks whether a child's rank depends strongly on the parents' rank. Absolute mobility asks whether children end up with higher real incomes than their parents. A country can perform differently on the two. Rapid growth can lift many children above their parents while preserving rank. A society can reshuffle ranks without giving everyone substantial absolute gains.

The distinction matters in the United States, where the Opportunity Insights series shows a steep decline in absolute mobility across post-war birth cohorts. Its counterfactual exercises find that slower aggregate growth explains part of that decline, but the more unequal distribution of growth explains more. The result is specific to the measure and country, yet the lesson travels: whether children surpass their parents depends on both how fast the economy grows and who receives the growth.

Mobility can also be misleading if measured only through income. A child may rise from the bottom to the middle but still have no wealth buffer. Housing wealth, business ownership and pensions can be more persistent across generations than earnings. Race and discrimination can alter the conversion of income into wealth through housing markets, credit access and historical exclusion. Opportunity is therefore not one staircase with a single first rung.

Education is the most common policy answer because it is both morally attractive and economically productive. It is also frequently asked to repair inequalities produced elsewhere. Equalise school spending while housing segregates children by income and neighbourhood effects remain. Expand university places while affluent families dominate selective preparation and the rationing may move to internships, postgraduate credentials or social networks. Education can widen opportunity, but advantage is adaptive.

That does not make equal opportunity impossible. It makes it institutional. A fairer starting field has to be built through health, housing, education, transport, safety, anti-discrimination enforcement and family resources, not declared at the moment an employer reads two CVs.

The job interview is late in the story.

Discrimination Can Survive Competition

One of the neatest economic arguments against discrimination is that prejudice is expensive. If an employer refuses to hire equally productive workers because of race or sex, a less prejudiced competitor can hire them more cheaply, earn higher profits and push the discriminator out. Competition, on this account, carries its own anti-discrimination mechanism.

Sometimes it does. The trouble is that discrimination can enter through more than taste, and markets are rarely frictionless enough for arbitrage to erase every bias.

Gary Becker's classic model treated discrimination partly as a preference for avoiding members of a group, effectively a cost the discriminator is willing to pay. Later theories added statistical discrimination. When employers have incomplete information about individuals, they may use group averages or stereotypes as proxies. Even if the average belief began with some data, using it can deny individuals the chance to reveal their own productivity. If the resulting decisions change investment, experience or networks, the initial belief can help reproduce the pattern it predicted.

Then there is cumulative discrimination. A small disadvantage at one stage changes the pool at the next. Fewer interviews mean fewer job offers. Fewer early opportunities mean less experience. Less experience affects later pay. A wage gap at forty can therefore contain the history of earlier gates, which makes it difficult to infer the mechanism from a cross-sectional pay comparison alone.

This is why the raw gender or racial pay gap is not a direct measure of discrimination. It combines occupation, hours, experience, education, family choices, geography, employer sorting, discrimination and other factors. Statistical controls can narrow the unexplained portion, but "unexplained" does not mean "proved discrimination", just as "explained" does not mean fair. If discrimination helped determine occupation or experience, controlling for those variables can remove part of the mechanism you are trying to study.

Field experiments provide cleaner evidence at particular gates. Correspondence studies send matched applications differing in signals such as names associated with racial or ethnic groups and compare callback rates. Audit studies pair applicants. These designs cannot capture every stage of a career, but they can identify discrimination in the tested decision more directly than a national wage gap can.

Discrimination also appears outside employment. Mortgage access, insurance, housing search, school discipline, policing, disability access and business finance can alter economic trajectories. Historical rules can leave asset distributions that persist after the rules change. If a family was excluded from appreciating housing markets for decades, ending legal exclusion today does not hand it the missing capital gain.

The economic consequence is broader than unfairness to the excluded person. Misallocation wastes talent. If capable people are blocked from occupations where they would be productive, output falls. Hsieh and co-authors have argued that reductions in occupational barriers faced by women and Black Americans account for a meaningful share of US growth over the second half of the twentieth century. The exact decomposition depends on modelling assumptions, but the mechanism is strong: discrimination can make an economy poorer by putting the wrong people in the wrong jobs.

Competition can punish some discrimination. It cannot be assumed to erase history, information problems, networks, residential segregation, customer preferences or institutional barriers by itself.

Inequality Has No Single Effect on Growth, Politics or Wellbeing

Ask whether inequality helps growth and you can produce a plausible argument in either direction within minutes.

The pro-inequality mechanism begins with incentives. If effort, training, invention and risk can produce higher rewards, people have stronger reasons to undertake them. Large projects may require concentrated pools of saving and capital. Entrepreneurs may need the possibility of substantial upside to accept uncertain downside. A society that confiscated every return above the average would damage useful behaviour.

The anti-inequality mechanism begins with constraints. If children from poor households cannot finance education, health or moves, talent is wasted. If wealth buys monopoly protection or political influence, resources can shift from innovation toward rent-seeking. If low-income households face weak demand and high debt burdens, macroeconomic fragility can rise. If inequality generates instability, distrust or policy capture, investment may suffer.

Both are possible because inequality is an outcome category, not a cause. A wage premium for a scarce surgeon and a monopoly fortune protected by regulatory barriers are both inequality. Their effects need not share a sign.

Empirical research reflects that ambiguity. Cross-country studies have found relationships between high inequality and shorter growth spells or weaker growth in some settings, but identification is difficult. Countries differ in institutions, development, demography, conflict and policy. Redistribution itself varies: transfers that let children eat and attend school differ from badly designed subsidies or confiscatory taxes. The IMF has therefore moved away from a simple presumption that redistribution necessarily harms growth, while stopping short of claiming that every redistributive policy raises it.

Politics supplies another channel. When wealth is concentrated, affluent individuals and firms can spend more on political participation, lobbying, expertise, litigation and campaigns. That can improve policy when informed groups provide useful knowledge. It can also bend rules toward incumbents, protect rents or narrow the agenda. The empirical task is to identify influence, not to assume that money mechanically buys every outcome.

Perception matters as well. People do not judge inequality only by the Gini coefficient. They care about procedural fairness, mobility, security, whether rewards seem earned, whether the rich obey the same rules, and whether ordinary living standards are rising. A society can tolerate striking income gaps when mobility looks credible and basic institutions work, then react sharply when the same gaps come to signify inherited privilege or rule-breaking.

Health and wellbeing show similar complexity. Income is strongly associated with health, but causation runs through medical access, stress, work conditions, environment, education and behaviour, while health itself affects earnings. Claims that inequality alone shortens lives across all rich societies are more contested than popular summaries suggest. It is safer and more useful to identify the concrete pathways by which deprivation, insecurity or unequal access change outcomes.

The distribution can also affect demand for redistribution. The median voter in a simple model might demand more redistribution as the mean income moves farther above the median. Real politics is messier. Identity, beliefs about mobility, trust, race, age, institutions and views about deservingness alter preferences. High inequality can produce more redistribution, less redistribution, or political coalitions organised around issues that barely mention income.

So there is no inequality knob marked growth, cohesion or health. There are mechanisms. The discipline improves the moment the debate moves from "inequality causes X" to "this kind of inequality changes this constraint or incentive, under these institutions, with this effect."

Outcomes Become Starting Conditions

Return to the distribution from the first idea.

A wage difference is recorded as an outcome this year. Part is spent. Part may become savings, pension rights, housing equity or business ownership. Those assets affect who survives the next shock, who can move, who can borrow, who can wait, who can pay for education, who can make an unpaid investment in skills, and who can transfer resources to children. The next round of income therefore begins with a distribution partly created by the previous one.

That is the feedback loop that turns temporary differences into durable inequality.

It can work through families. Higher-income parents buy inputs and reduce risks for children. It can work through geography. Higher land and housing prices sort households into neighbourhoods, and neighbourhoods affect schools, networks and exposure. It can work through firms. Successful owners reinvest profits, purchase competitors or fund new ventures. It can work through politics. Economic resources finance efforts to shape regulation, tax rules and public spending. It can work through expectations. People who have repeatedly seen doors open or close make different bets about education, careers and institutions.

None of this means mobility stops. Fortunes disappear. New technologies destroy incumbents. Families divide wealth. Migration changes opportunity. Taxes and philanthropy move resources. Children choose differently from parents. Inequality does not inevitably harden into caste. Outcomes still alter the feasible set from which the next set of choices is made.

That distinction separates equality of opportunity from a slogan. If opportunity were fully independent of previous outcomes, society could allow enormous dispersion today and restart everyone from the same line tomorrow. No real society can do that, because children arrive inside families, property persists, knowledge transfers and places differ. The policy question becomes how strongly yesterday should determine tomorrow.

This is also why redistribution after the fact is only one part of policy. Cash transfers can raise disposable income now. Progressive taxes can reduce concentration. But if unequal schools, restrictive housing supply, weak competition, discrimination, poor health access or bargaining institutions keep generating the same distribution before taxes, the state is forever treating the output of a machine whose settings remain unchanged.

The reverse is also true. "Predistribution", a term used for policies that change market outcomes before taxes and transfers, cannot eliminate every need for redistribution. Illness, disability, unemployment, family size, bad luck and life-cycle differences will still produce unequal needs and incomes. A serious policy architecture works before the market, inside the market and after the market.

The first idea warned that inequality is not a number. The seventh explains why. The distribution you measure at the end of a year is a snapshot of a moving system. It contains past institutions, current rewards and future starting positions in the same frame.

Once that is clear, the argument over inequality becomes less theatrical. You can favour markets and still care about inherited advantage. You can favour redistribution and still care about incentives. You can oppose discrimination without pretending every group gap proves it. You can celebrate falling poverty while worrying about concentrated wealth. You can want people to keep the fruits of useful work while refusing to treat every existing fortune as evidence of useful work.

The central choice is not between equality and inequality. Every functioning economy contains differences. The choice is which mechanisms are allowed to generate them, how much security exists underneath them, and how easily today's winners can write tomorrow's starting conditions.

How It Actually Works

Inequality is easiest to understand by following economic advantage through its stages. The route begins before wages are earned, passes through labour markets and households, accumulates in assets, crosses generations and borders, and returns through policy and behaviour. Each stage can create, compress or transmit differences.

Before the wage

Begin before work.

A child is born into a household with a bundle of resources: income, wealth, time, health, knowledge, housing, neighbourhood, networks and expectations. None is perfectly measured and none determines the child. Together they alter the cost of developing capabilities.

Take education. A state may provide nominally free schooling, yet the effective package includes housing near a school, transport, books, devices, tutoring, quiet space, nutrition, parental time and the ability to absorb failure. A missed exam after illness means something different if a family can pay for a resit, a tutor or another year. Public provision can narrow these differences substantially. It rarely makes the surrounding household irrelevant.

The same is true of health. Childhood illness, disability, air pollution, stress and nutrition can affect school attendance and later work. Public healthcare can insure much of the cost. It cannot guarantee equal exposure or equal capacity to navigate the system. A distribution of adult earnings therefore begins partly in institutions that do not look like labour markets.

Location then turns resources into opportunity. Housing markets sort people. Planning rules, transport, local taxation, school systems and historic segregation determine how strongly address predicts access. In places where good jobs are geographically concentrated and housing near them is scarce, rising productivity can be capitalised into land values. Workers who already own property gain an asset. Newcomers face a higher entry price. A local boom can therefore raise both wages and barriers to joining the boom.

US mobility research made this visible by comparing children who moved between areas at different ages. In the designs used by Chetty and Hendren, children's later outcomes moved toward the typical outcomes of the destination, with larger exposure effects for those who moved younger. The finding is specific to the settings and outcomes studied, not a law of neighbourhoods everywhere. It nevertheless shows that place can be part of the causal machinery of opportunity rather than a label attached to family background.

Entering the labour market

Now the child becomes a worker.

Employers observe imperfect signals: qualifications, experience, interviews, references, tests, names, accents, addresses and networks. Workers observe imperfect signals too: salary, reputation, job description and whatever can be learned before accepting. Matching is noisy.

Education changes the match in at least three ways. It can create skills. It can signal traits such as persistence or prior ability. It can grant access to occupations through formal credentials. The contribution of each varies by field and institution. This is why the wage premium to a degree is not a pure measure of what classrooms taught.

The labour market then prices scarcity. If demand for software security specialists rises faster than supply, wages rise. If a routine clerical task can be automated or moved, its wage pressure weakens. Trade, technology and organisational change therefore reshape the demand curve for tasks, not just for occupations.

The task lens is useful because jobs are bundles. Automation can replace some tasks and complement others inside the same job. A spreadsheet reduced the labour needed for arithmetic while increasing the value of workers who could interpret, model and communicate results. Industrial robots displaced particular production tasks while creating maintenance, engineering and logistics work. Generative AI is likely to be uneven for the same reason: it can complement some tasks, automate others and alter the scale at which firms operate. Its long-run distributional effect remains unresolved because adoption, task redesign, complementary investment and market structure are still changing. A current capability should not be mistaken for a settled wage distribution.

The wage bargain comes next. Suppose a worker creates more value in a firm than her outside option. The surplus has to be divided. In a competitive labour market with easy switching and good information, firms must offer close to what alternatives provide. In a thin market with few employers, costly moves or restrictive contracts, the same productivity can produce a lower wage because the worker's threat to leave is weaker.

This is where bargaining institutions alter distribution without changing the worker's physical output. A union, sectoral agreement, minimum wage, pay-transparency rule, non-compete ban, unemployment benefit or transport link can improve an outside option or change information. Employer concentration, weak benefits or immigration dependence can move it the other way. The resulting wage is economic, but it is not institution-free.

Discrimination can enter at the same gates. A callback lost because of an ethnic signal is not merely a moral injury. It removes one draw from the set of potential offers, reducing the chance of a good match and potentially weakening later bargaining. Small gatekeeping differences can accumulate into experience, occupation and wealth gaps over a career.

From pay to household income

The wage is not yet household living standards.

Households combine earners and non-earners. They may receive self-employment income, rent, dividends, interest, pensions or benefits. They pay direct taxes and may receive transfers. The distribution changes at each step.

A useful accounting sequence is market income, gross income after cash transfers, disposable income after direct taxes, and a broader concept sometimes called adjusted disposable income after adding the value of services such as publicly provided education and healthcare. Different statistical systems draw the boundaries differently. The important point is that government affects distribution through both cash and services.

This is why a country with unequal market wages can produce less unequal disposable living standards. Taxes and transfers compress the distribution, but the extent varies widely. Public services can narrow effective inequality further where they replace expensive private purchases. Their value is difficult to assign household by household, so many headline inequality measures leave them out.

Indirect taxes complicate the picture. Consumption taxes may take a larger share of current income from low-income households, though incidence depends on exemptions, spending patterns and what the revenue finances. Employer payroll taxes can be shifted partly into wages over time. Corporate taxes can fall on owners, workers or consumers depending on capital mobility and market structure. The statutory payer is not necessarily the economic bearer.

That is why serious distributional analysis studies incidence rather than reading a tax code as if every label describes the final burden.

Turning income into wealth

Now the household decides what can be saved.

High-income households can usually save a larger share because basic consumption absorbs less of their budget. Low-income households may save little or borrow, particularly after shocks. This creates a mechanical route from income inequality to wealth inequality.

But asset returns differ too. Wealthier households often have better access to diversified financial assets, tax planning, private businesses and professional advice. Poorer households may hold most wealth in a home, a bank account or pension, if they own assets at all. The return gap is not guaranteed and varies over time, but portfolio composition matters.

Housing is the bridge between middle-class wealth and inequality. In many countries the home is the largest asset for broad parts of the population. Rising house prices can therefore spread capital gains beyond the richest households. They can also increase the gulf between owners and renters and between regions. A homeowner sees appreciation as wealth. A first-time buyer sees the same price as a larger deposit requirement.

Leverage magnifies both. A household buying a £300,000 home with £60,000 equity is exposed to price changes on the full asset. A 10 per cent rise adds £30,000 before transaction costs, a 50 per cent gain on initial equity. A 10 per cent fall removes the same amount. Mortgages can therefore accelerate wealth building and wealth destruction. Access to safe credit is distributional infrastructure.

Business ownership is more concentrated. A successful company can turn one period's entrepreneurial income into a large capital stock. This is part of the reward system that finances risky ventures. It also means that scalable firms can create wealth faster than labour income alone. Digital markets can intensify the effect when one product serves millions of users at low marginal cost.

Pensions complicate wealth statistics. In funded systems, pension rights may represent substantial household assets. Some datasets count them fully, others partially, and public pay-as-you-go pension promises are usually treated differently from private financial wealth. Cross-country wealth comparisons can therefore mislead if institutional differences are ignored.

Passing advantage forward

The household ages.

Some wealth is consumed in retirement. Some pays for care. Some is given to children while parents are alive. Some passes through estates. The timing matters.

A transfer of £40,000 toward a housing deposit at age thirty may affect a child's life more than a £100,000 inheritance at age sixty. Early transfers can alter where someone lives, whether they buy or rent, how much they borrow, whether they start a business and which risks they take. Economists interested in intergenerational inequality therefore look beyond inheritance tax records to inter vivos gifts, parental spending and family guarantees.

Networks are transmitted too. Parents know employers, professions, universities and neighbourhoods. They can explain which internships matter, how to negotiate an offer or what a mortgage broker will ask for. These are not always purchased advantages, and many arise from ordinary parental care. At scale they still make family background economically productive.

The intergenerational link is visible in rank correlations. If parental income rank strongly predicts child income rank, relative mobility is low. But the coefficient does not tell you why. Genes, family culture, wealth, schools, neighbourhoods, discrimination and networks are bundled together. Good research tries to separate channels rather than treating persistence as proof of a single mechanism.

Absolute mobility asks a different question: did the child earn more in real terms than the parent? Strong economic growth can raise absolute mobility even where rank persistence remains. Weak growth can reduce it even if ranks shuffle. The two measures answer different political hopes.

Across borders

Now widen the map.

A worker in a middle-income country may be poor relative to someone in Britain and rich relative to the poorest people in her own country. Global inequality combines inequality between countries with inequality within them.

For much of modern history, where you were born explained an enormous share of your position in the world distribution. The industrial revolution opened vast gaps between early industrialisers and much of Asia, Africa and Latin America. During the late twentieth and early twenty-first centuries, rapid growth in populous Asian countries narrowed many between-country gaps.

This is the source of an apparent contradiction. Global interpersonal inequality can fall while inequality rises within some nations. If hundreds of millions in China or India move closer to rich-country incomes, the global distribution compresses even if top incomes inside the United States, China or India rise faster than their national medians.

Branko Milanovic popularised one version of this story with the "elephant curve", describing uneven global income growth across percentiles during the high-globalisation period. The chart was useful and frequently overread. Its shape depends on years, data and population composition. It does not prove that gains to Asian middle classes caused stagnation among western workers. It shows that distributional change must be examined at more than one geographic scale.

Migration makes the geographic lottery tangible. Moving from a low-productivity country to a high-productivity one can multiply earnings for the same person because institutions, capital, technology and firms differ. That is a distributional fact of enormous magnitude. It also produces domestic political conflicts over labour supply, public services, housing and identity that belong to the immigration debate in greater depth.

Trade works similarly. Comparative advantage can raise total income while creating concentrated losses in exposed industries and regions. If gains are broad and adjustment costs are local, national welfare can rise while particular workers experience lasting damage. The distribution of compensation, retraining, mobility and new investment then determines whether the aggregate gain becomes a broadly shared gain.

When inequality changes behaviour

A distribution does not sit still after being measured. People respond to it.

At the bottom, lack of liquidity can shorten horizons. A household with no buffer may pay more for credit, reject training with delayed returns or stay in a poor job because a missed pay packet is unaffordable. These choices can look impatient when they are rational responses to risk.

In the middle, housing and education can become positional. If access to a scarce school or neighbourhood depends on relative purchasing power, households may spend more to avoid falling behind even when absolute quality has improved. The contest is over rank, so one household's purchase changes the pressure on another.

At the top, wealth expands the menu of investment and influence. Large investors can diversify, hold illiquid assets, finance litigation and withstand long horizons. Founders can self-finance. Political donors can spend on advocacy. Again, none of this implies a single motive or guaranteed outcome. It means resources alter feasible strategies.

Firms respond too. If consumers have unequal willingness to pay, companies segment markets. If workers have unequal outside options, wages can differ. If capital is concentrated, firms may find it easier to raise large sums from a smaller investor base. Distribution enters business models.

Governments respond to the distribution and to voters' beliefs about it. They set taxes, transfers, education, health, pensions, minimum wages, competition rules, housing policy and inheritance rules. Those decisions alter the next distribution. They also alter behaviour, which changes the tax base, labour supply, saving, investment and prices. Distributional policy is therefore a general-equilibrium problem even when political debate presents it as moving money from one column to another.

The policy stack

Because inequality is produced at several stages, policy acts at several stages.

Before market income is earned, governments influence childhood resources through healthcare, childcare, schools, housing, transport and family support. These policies aim at capabilities and opportunity.

Inside markets, governments set rules for competition, labour bargaining, discrimination, corporate governance, credit and property. These shape wages, prices, profits and access.

After market outcomes, taxes and transfers change disposable resources. Social insurance protects against unemployment, disability, illness and old age. Public services add resources that cash measures often miss.

Across generations, inheritance rules, property taxation, education finance and housing supply affect how strongly wealth reproduces itself.

No layer can carry the whole load. Heavy redistribution can compensate for unequal market outcomes while leaving barriers untouched. Pure opportunity policy can take decades and cannot insure adults against bad luck. Labour rules can improve bargaining but cannot replace disability benefits. Competition can reduce rents but cannot eliminate inherited wealth. Education can raise skills but cannot guarantee enough high-productivity jobs.

The practical argument is therefore about a portfolio and its trade-offs.

Transfers can reduce poverty quickly but may affect work incentives depending on withdrawal rates and design. Progressive taxation raises revenue and compresses post-tax incomes but unusually high marginal rates can change labour supply, avoidance, migration or the timing and form of income. Wealth and inheritance taxes target stocks and transfers but face valuation, liquidity and enforcement problems. Minimum wages can raise low pay with employment effects that depend on level, market power and local conditions. Unions can compress wages and improve voice while creating rigidities if poorly designed. Housing liberalisation can widen access to productive places but creates local losers among incumbent owners and renters during transition.

Policy choices also differ by development level. In a low-income country, reliable electricity, basic health, roads, schools, property administration and broad growth may dominate arguments about top marginal rates. In a rich country, the binding constraints may be housing scarcity, childcare, ageing, market power or inherited wealth. "Reduce inequality" is not a policy specification.

How we know

Inequality data are reconstructed from imperfect sources. Household surveys measure broad populations well but often miss the richest and can understate capital income. Tax records capture high incomes more precisely where filing systems are comprehensive but omit people and forms of wealth outside the tax base. National accounts provide totals without identifying who receives them. Wealth surveys struggle with valuation, offshore assets, private businesses and pension rights.

Modern distributional datasets combine these sources, which improves coverage and introduces modelling choices. The World Inequality Database, World Bank Poverty and Inequality Platform, Luxembourg Income Study, OECD and national statistical agencies therefore produce measures that can differ without one being fraudulent.

Causal evidence is harder. Researchers use reforms, lotteries, migration, experiments, matched applications, administrative records and long panels to separate mechanisms from correlation. Results travel imperfectly across countries and periods. Treat exact levels cautiously, but do not mistake measurement difficulty for ignorance. The main patterns, including the greater concentration of wealth than income and the persistence of family background, survive across many methods.

What People Get Wrong

“The Gini coefficient tells you how fair a country is”

It tells you how dispersed a chosen distribution is under a chosen definition. It does not tell you whether the gap came from innovation, discrimination, inheritance, rent extraction or age. It does not tell you whether the poorest are destitute or comfortable. It does not reveal top concentration especially well, and it changes with taxes, transfers, household adjustment and the unit measured.

Fairness requires a theory of causes, procedures and minimum standards. The Gini is a thermometer, not a verdict.

It also loses information by design. Move £1 of income from the middle to the bottom and the coefficient may change differently from moving £1 from the top to the middle, yet neither movement tells you whether anyone crossed a poverty threshold or gained meaningful security. For public debate, the sensible dashboard pairs a summary measure with median income, poverty, top shares and mobility. One number can discipline attention. It cannot carry a moral philosophy.

“If inequality rises, the poor must be getting poorer”

Not necessarily. Suppose the poorest household's income rises from £10,000 to £15,000 while the richest rises from £100,000 to £180,000. The bottom is better off in absolute terms and the gap is wider.

This distinction matters globally. Extreme poverty fell dramatically from 1990 even while income and wealth remained highly unequal. Poverty and inequality overlap because low incomes sit at one end of the distribution, but one asks about deprivation and the other about relative shares or gaps. Good policy should know which problem it is trying to solve.

The reverse error matters too. Lower inequality is not automatically a social improvement. A deep recession can compress incomes if profits and top earnings collapse faster than low wages. War can destroy fortunes. An authoritarian state can equalise measured cash incomes while restricting freedom and leaving privileged access to housing, goods or political power outside the statistics. Direction alone is not enough. Ask what happened to living standards and why the distribution changed.

“Wealth inequality is just income inequality accumulated”

Income matters, but the conversion is not mechanical. Saving rates differ. Asset portfolios differ. House prices, share prices and business values move. Debt magnifies gains and losses. Inheritance transfers stocks. Tax systems treat labour and capital differently. A household can have high income and little wealth, or moderate income and substantial housing and pension assets.

Wealth also changes behaviour before it produces income. It insures risk, secures credit and finances opportunities. That feedback is why wealth deserves separate treatment rather than being filed as delayed salary.

The distinction also changes policy. A temporary wage subsidy acts on a flow. Pension enrolment, housing access, matched saving, business ownership and inheritance rules act on stocks or on the routes into them. If the concern is that half the population has little capacity to absorb a shock, a modest change in annual income inequality may leave the central problem untouched. Balance sheets can matter more than payslips.

“People are paid what they produce”

Productivity constrains sustainable pay, but it does not uniquely determine it. Many jobs produce value jointly, making individual marginal products difficult to observe. Workers and firms bargain over surplus. Outside options, employer competition, information, institutions and norms affect the split.

The opposite claim, that wages are only power, fails too. A firm cannot indefinitely pay every worker more than the value the business can create. Scarcity and productivity matter. Distribution emerges from productive value interacting with bargaining and institutional design.

This is why the same occupation can command different wages across otherwise comparable places and periods. Productivity may differ, but so can employer concentration, union coverage, wage floors, credential rules and workers' ability to move. The observed wage is the result of the whole bargaining environment. Treating it as a pure productivity reading erases institutions; treating it as pure power erases the constraint that firms must create enough value to pay it.

“Education can solve inequality”

Education can raise productivity, widen opportunity and increase mobility. It cannot perform every distributional job.

If housing prices segregate access to good schools, labour markets create weak outside options, capital ownership is concentrated, discrimination blocks entry, and scarce top positions remain scarce, more credentials may shift rather than remove competition. Universal secondary education once distinguished workers; later it became a baseline. Advantage can migrate to university prestige, postgraduate study, internships or networks.

Education is one of the strongest equalising tools available. Treating it as a substitute for housing, health, labour, competition and family policy asks a school to repair the whole economy.

There is also a composition problem. Education can improve everyone's skills while leaving relative inequality little changed if access improves at every level and employers continue to ration the most attractive jobs. That is still a gain. A better educated population can be richer and healthier even if the income ranking remains dispersed. The mistake is judging education only by whether it compresses one distribution.

“High inequality proves low mobility”

The two are correlated in many datasets but not identical. Inequality describes the spread of outcomes. Mobility describes movement through the distribution over time or across generations.

A society could have large rewards and frequent movement between positions. Another could have a narrower distribution with rigid ranks. In practice, high inequality can make mobility harder when wealth buys opportunity, but the relationship depends on institutions.

The famous "Great Gatsby Curve" plots greater income inequality against lower intergenerational mobility across countries. It is a useful empirical pattern, not a law of nature and not proof that one coefficient causes the other.

Mobility measures can also hide distance. Moving from the 10th percentile to the 30th is substantial rank mobility but may still leave a person with low resources. Falling from the 99th to the 90th is downward mobility while remaining rich. A society interested in opportunity should inspect both rank persistence and the real living standards attached to those ranks.

“There is one optimal amount of inequality”

Economics offers no universal number.

Some dispersion rewards scarce skills, effort, innovation and risk. Some reflects age and life-cycle differences. Some comes from luck. Some comes from barriers, market power or inherited advantage. The welfare cost of a given Gini therefore depends on what sits underneath it and what living standards sit beneath the bottom.

The serious objective is not to choose a national Gini as though setting a thermostat. It is to preserve useful incentives, reduce deprivation and waste, widen opportunity, insure bad luck, restrain rents and decide how strongly one generation's success should determine the next one's start.

That answer will differ across societies because institutions and starting points differ. A country with mass poverty, weak tax administration and scarce capital faces different trade-offs from a rich country with mature welfare systems and expensive housing. Even within one country, the desirable response to a surgeon shortage is different from the response to a protected monopoly. "Inequality" groups both gaps together; policy has to separate them again.

There is no neutral baseline hiding underneath these choices. Property rights, bankruptcy rules, company law, education systems, labour contracts and planning rules already distribute risks and claims before any minister announces a redistributive programme. The absence of a new policy is therefore not the absence of distributional policy. It is a decision to keep the existing rule set. The useful comparison is between alternative institutions and their effects, not between intervention and an imaginary untouched market.

This matters most when people agree on the headline and disagree on the mechanism. Two voters can both want less inequality while one favours stronger wage bargaining and the other favours wider ownership of capital. Two can accept large income gaps while disagreeing sharply about inherited wealth. The absence of one optimal number does not end the argument. It forces the argument onto causes, consequences and institutional choices, where it belongs.

Use It

Ask which distribution

Whenever someone says inequality rose or fell, ask what was measured.

Income or wealth? Individual or household? Market income or disposable income? Before or after housing costs? One year or lifetime? National or global? Mean, median, Gini, top share or percentile ratio?

This is not pedantry. Different measures can move in different directions. A claim that survives this interrogation is more likely to describe a real mechanism rather than a graph selected for a political purpose.

Then ask for the denominator and the date. Wealth shares can jump when asset prices move. Household income can shift with inflation, employment and family composition. International comparisons can change when currencies or purchasing-power estimates are revised. A statistic becomes much more informative once you know the population, resource concept and period behind it.

Separate floor, gap and mobility

Three questions should become automatic.

How well off are the people at the bottom in absolute terms? How large is the gap between parts of the distribution? How strongly does origin predict destination?

A society can improve the floor while widening the gap. It can compress incomes while leaving poor mobility. It can have high mobility in rank but weak absolute growth. Policy arguments become clearer when speakers are required to say which objective they value.

This three-part test also prevents rhetorical substitution. A politician may answer a mobility problem with a poverty statistic, or answer a poverty problem with a falling Gini. Both can be true and irrelevant. Write the three objectives on separate lines before judging a policy. Then ask which line it moves and what it costs on the others.

Follow the outside option

When a wage, price or contract looks unfair, inspect the alternatives available to each side.

Can the worker switch employers without moving? Can the tenant move without losing access to school or work? Can the small supplier survive losing one buyer? Can the borrower refinance? Can the founder wait six months for capital?

Bargaining power often hides inside the cost of saying no. Improving competition, information, mobility or insurance can change distribution without administratively setting every outcome.

The lens works upward too. A chief executive negotiating with a weak board, a dominant platform negotiating with small sellers, or a landlord in a severely supply-constrained market may have unusually strong outside options. Before calling the resulting price or pay efficient, inspect whether the other side had credible alternatives.

Trace the stock behind the flow

A salary figure is incomplete until you ask what assets and debts sit behind it.

Two households with the same income may have different housing equity, pensions, student debt, family support and emergency savings. Those stocks alter risk and future opportunity.

This lens is especially useful when evaluating housing, education or entrepreneurship. Ask who can absorb the downside. The ability to survive a failed attempt is itself an economic advantage.

It also changes how you interpret risk-taking. One founder may be willing to leave a salary because family wealth covers rent. Another may reject an identical venture with identical expected returns because failure would threaten housing. Calling the first more entrepreneurial describes behaviour while missing part of the balance sheet that made the behaviour affordable.

Test whether a gap is a signal or a mechanism

Group differences are facts to explain, not explanations.

A pay gap may signal discrimination, occupational sorting, hours, experience, geography, family constraints or several channels. A school gap may reflect teaching, selection, neighbourhoods, health or household resources.

Do not stop at the gap. Look for a mechanism that predicts what should happen if one input changes, then seek evidence from reforms, experiments, movers or matched comparisons. Description becomes useful when it generates a test.

Use the same discipline with top incomes. If the hypothesis is scarce talent, ask whether pay rises where scale and measurable performance rise. If the hypothesis is market power, ask what happens when entry or competition increases. If the hypothesis is inherited advantage, look for effects of family wealth holding current income or ability as constant as the design allows. A mechanism earns confidence by surviving a prediction it could have failed.

Ask what reproduces the result

The most important inequality question is often not why a gap exists today but what makes it persist.

Does wealth buy access to appreciating housing? Do networks reproduce hiring advantages? Does poor transport isolate workers from jobs? Does market power protect incumbent profits? Do schools depend on local property values? Does political influence protect a rent?

If the reproduction mechanism is strong, one-off redistribution may fade. If it is weak, temporary support may be enough. Policy should match the persistence mechanism.

This is the difference between treating a leak and repainting the wall. A transfer can compensate a household for an unequal outcome this year, which may be exactly what is needed. If the same household remains cut off from good jobs by transport, housing or discrimination, the transfer does not remove the generator. Compensation and structural repair are different tasks, and a serious system may need both.

Then ask who bears the transition. Opening housing supply can reduce scarcity over time while construction disrupts existing residents. Stronger competition can lower rents while destroying incumbent profits and some jobs. Education reform can improve opportunity for children while doing nothing for adults whose chances were already rationed. A distributional policy can be directionally right and still require insurance for the people who lose during the change.

Finally, distinguish the policy's target from its political sales pitch. A housing reform may be sold as growth policy and operate mainly through lower land scarcity. A childcare subsidy may be sold as family support and alter labour-force participation. A competition reform may be sold as consumer policy and raise worker outside options. Distributional effects often arrive through mechanisms whose official labels say little about inequality.

That is useful because it keeps diagnosis ahead of ideology. Instead of asking whether a proposal belongs to the left or right, ask which constraint it changes, who gains bargaining room, who loses a rent, who bears transition costs and whether the effect persists after people adapt.

The limits

Inequality statistics compress lives. They cannot tell you whether one person's higher income reflects effort, luck, family support, risk, discrimination or social value. They can show patterns that demand explanation, not allocate moral desert person by person.

Causal evidence travels imperfectly. A minimum wage that works well in one labour market need not have the same effect at another level or in another country. A housing reform can improve access and still impose transition costs. A transfer can reduce poverty and create withdrawal-rate problems. Distributional policy has behavioural and political responses.

There is also no complete separation between equality and other values. People care about freedom, security, privacy, family autonomy, innovation, community and democratic voice. An institution can improve one dimension while worsening another. The correct comparison is between feasible systems with trade-offs, not between the existing world and a frictionless ideal.

The one thing to keep

Keep the loop.

Today's distribution is not the scoreboard at the end of the game. It is part of the equipment handed out before the next round begins.

Income can become wealth. Wealth can become security, education, location, networks, credit and political influence. Those advantages affect the next income distribution, which creates the next stock of wealth. The loop can be weakened by growth, competition, public services, insurance, taxation, mobility and chance. It can be strengthened by exclusion, scarcity, rents and inheritance.

That changes the central question. Stop asking whether inequality is good or bad in the abstract. Ask which differences reward useful behaviour, which reflect barriers or rents, what floor sits beneath them, and how strongly they write the starting conditions of people who did not choose them.

That is the choice hidden inside the distribution.

Terms

Absolute mobility. The extent to which people end up with higher real incomes or living standards than their parents. It differs from moving up in rank and is highly sensitive to broad economic growth.

Adjusted disposable income. Household disposable income plus the imputed value of certain public services, commonly health and education. It captures resources that cash-only measures miss.

Bargaining power. The ability of one side in a negotiation to obtain better terms because its alternatives, information, patience or institutional position are stronger. Outside options are often the decisive variable.

Capital income. Income generated by ownership of assets, including interest, dividends, rents and some business profits. Its distribution depends heavily on asset ownership and on how returns are taxed and measured.

Disposable income. Income available after direct taxes and cash transfers, under the definition used by a statistical system. It is closer to household spending power than pretax earnings.

Economic rent. A return above what is necessary to keep a resource in its current use. Rents can arise from scarcity, monopoly, regulation, land or other barriers and are central to debates about unearned gains because cutting a rent need not weaken productive effort.

Equivalised income. Household income adjusted for household size and composition. The adjustment recognises that shared households enjoy some economies of scale.

Equality of opportunity. The idea that outcomes should depend less on circumstances people did not choose. Measurement usually examines how strongly background predicts later outcomes, though no statistic can cleanly divide circumstance from effort.

Gini coefficient. A summary measure of dispersion derived from the Lorenz curve. Zero represents perfect equality and one represents maximal concentration under the standard form. It does not identify where in the distribution the gap lies.

Great Gatsby Curve. The observed cross-country association between higher income inequality and lower intergenerational mobility. It is a pattern, not a universal causal law.

Gross income. Income before direct taxes, often after including certain transfers depending on the statistical convention. Definitions differ across datasets.

Human capital. Skills, knowledge, health and capabilities that raise a person's productive potential. Education is one source, not the whole concept, and family and public investment both contribute.

Income. A flow of economic resources received over a period, including wages and potentially self-employment, transfers and capital income depending on the measure.

Intergenerational elasticity. A measure of how strongly parents' economic outcomes predict children's outcomes. Higher persistence generally implies lower relative mobility.

Labour income share. The share of national income accruing to labour compensation rather than capital and other claims. Measuring self-employment income requires adjustment.

Lorenz curve. A graph showing the cumulative share of income or wealth held by cumulative population shares ordered from poorest to richest.

Market income. Income generated through labour and capital before most government redistribution. Exact treatment of pensions and transfers varies by dataset.

Median income. The income of the person or household in the middle of the distribution. It is less affected by extreme top incomes than the mean.

Monopsony. A labour-market condition in which employers have wage-setting power because workers face limited or costly alternatives. It does not require a literal single employer; frictions can give many firms some wage-setting power.

Palma ratio. The income share of the top 10 per cent divided by that of the bottom 40 per cent. It emphasises distribution at the two ends.

Percentile. A position in an ordered distribution. The 90th percentile is the point above which roughly 10 per cent of observations lie.

Predistribution. Policies and institutions that alter market outcomes before taxes and transfers, such as education, competition rules, wage-setting institutions and labour standards. The term focuses attention on how primary incomes are generated.

Progressive taxation. A tax structure in which the average tax rate rises with the tax base or income under the relevant definition. Progressivity does not by itself reveal economic incidence.

Relative mobility. Movement in economic rank compared with parents or peers. A person can rise in absolute income without rising in relative rank.

Rent-seeking. Effort devoted to capturing or protecting economic rents through political, legal or strategic means rather than creating additional social value. It matters because high returns can reflect successful protection of scarcity as well as productive innovation.

Social insurance. Collective protection against risks such as unemployment, disability, illness and old age, usually financed through taxes or contributions.

Top income share. The fraction of total measured income received by a top group such as the richest 10, 1 or 0.1 per cent.

Wealth. The stock of assets minus liabilities at a point in time. It includes financial and real assets under the chosen statistical definition and can provide security and collateral before producing any cash income.

Wealth-income ratio. Aggregate wealth relative to annual national or household income. It describes the size of accumulated asset stocks compared with current flows and helps separate changes in asset values from changes in annual earnings.

Inter vivos transfer. A transfer of wealth made while the giver is alive, such as a housing deposit or business gift. It can affect opportunity long before a formal inheritance and is easy to miss in estate-based measures of inherited advantage.

Go Deeper

Branko Milanovic, Global Inequality: A New Approach for the Age of Globalization (Harvard University Press, 2016). Start here for the global picture. Milanovic explains why inequality between countries and within countries can move in opposite directions, and why location remains such a powerful determinant of income. The book is readable, empirical and especially useful for escaping a purely national frame. Its treatment of global percentiles also shows why a person can gain strongly in absolute terms while losing position relative to a national elite.

Thomas Piketty, Capital in the Twenty-First Century (Harvard University Press, 2014). Read this for the modern historical argument about wealth, capital and inheritance. Its long-run datasets transformed the public debate, while several interpretations and projections remain contested. It is long, which is partly the reason this book exists, but the historical sweep is unmatched. Read the data chapters even if you disagree with the policy conclusions, because the separation of income flows from accumulated capital is foundational.

Raj Chetty, Nathaniel Hendren, Patrick Kline and Emmanuel Saez, “Where Is the Land of Opportunity? The Geography of Intergenerational Mobility in the United States,” Quarterly Journal of Economics 129, no. 4 (2014). Read the original evidence behind the modern geography-of-opportunity literature. The paper is technical, but its maps and research design show how administrative data can turn mobility from a slogan into a measurable local phenomenon. Pay particular attention to how the authors distinguish relative mobility from expected outcomes for children from low-income families.

Anthony B. Atkinson, Inequality: What Can Be Done? (Harvard University Press, 2015). Read this for policy. Atkinson treats inequality as an institutional outcome and proposes interventions across technology, labour markets, capital ownership, taxation and social security. You need not accept every proposal to benefit from the discipline of matching instruments to mechanisms. It is especially useful as a counterweight to policy debates that begin and end with income-tax rates. The book also forces a valuable distinction between changing market outcomes and compensating for them afterwards, which is the policy architecture used here. Its proposals are intentionally ambitious, so read it as a worked example of mechanism-based reform rather than a menu whose every item must travel unchanged across countries. The lasting lesson is methodological: distribution can be altered through rules governing technology, wages and ownership before the tax-and-transfer system begins its work. That makes it the strongest of the four recommendations for readers who want to move directly from diagnosis to institutional design.

Notes and Sources

Measurement and the global distribution

The definitions of income, wealth, Gini coefficients, Lorenz curves, percentile shares and household equivalisation follow standard treatments used by the OECD, World Bank and national statistical agencies. Cross-dataset comparisons require care because surveys, tax data, national accounts, pension treatment and household definitions differ.

The current global top-share figures are drawn from Lucas Chancel, Ricardo Gómez-Carrera, Rowaida Moshrif and Thomas Piketty, eds., World Inequality Report 2026. The report estimates that in 2025 the global top 10 per cent received 53 per cent of income and owned about 75 per cent of personal wealth, while the bottom 50 per cent received about 8 per cent of income and owned about 2 per cent of wealth. These are distributional estimates assembled from multiple sources rather than direct observation of every household.

The World Bank Poverty and Inequality Platform provides the current international poverty series and country distribution data. In the March 2026 vintage, the World Bank estimated that 847 million people, 10.4 per cent of the global population, lived below its current extreme-poverty line in 2024, with a nowcast of 10.0 per cent for 2026. The book uses this series only to separate absolute deprivation from relative inequality, and notes that data vintages and poverty lines are periodically revised.

Wealth, labour income and accumulation

The labour-income-share discussion uses the International Labour Organization's 2025 technical work and current ILO material. The ILO reports that the global labour income share fell by 1.6 percentage points from 2004 to 2024 and attributes the trend to several structural forces, including technology, globalisation and weakened worker bargaining power. The manuscript treats this as an aggregate accounting shift, not as a complete explanation of household wage inequality.

Thomas Piketty's Capital in the Twenty-First Century supplies the long historical treatment of wealth accumulation and inheritance. The manuscript uses the stock-flow distinction without adopting a mechanical claim that the return on capital must always exceed economic growth or that one inequality path is inevitable.

Wages, technology and institutions

The skill-biased-technological-change discussion draws on Daron Acemoglu's review of technical change and the labour market and on David Autor's later task-based work. The task framing is used to avoid the false binary that technology either destroys jobs or leaves distribution untouched.

The labour-market-power discussion follows modern monopsony economics and the broad principle that wages can depend on outside options and employer competition as well as productivity. The ILO's work on collective bargaining supports the claim that bargaining institutions can compress wage inequality. No universal employment effect is asserted for unions or minimum wages.

Opportunity, place and mobility

Raj Chetty, Nathaniel Hendren, Patrick Kline and Emmanuel Saez, “Where Is the Land of Opportunity?” supplies the US geographic mobility evidence. Chetty and co-authors' later work on absolute mobility estimates that about 90 per cent of children born in 1940 earned more than their parents, compared with about half of those born in the 1980s under their household-income measure. Their counterfactual analysis attributes most of the decline to the more unequal distribution of growth rather than slower aggregate growth alone.

The OECD's 2018 A Broken Social Elevator? provides comparative evidence on intergenerational and life-course mobility. Its 2025 To Have and Have Not: How to Bridge the Gap in Opportunities adds newer cross-country estimates showing that measured family circumstances account for a substantial but incomplete share of income inequality, with wide national variation. The book treats the Great Gatsby Curve as an empirical association rather than a universal causal law.

Discrimination and misallocation

The distinction between taste-based and statistical discrimination follows the labour-economics literature descending from Gary Becker and Edmund Phelps. Kevin Lang and Ariella Kahn-Lang Spitzer's review, “Race Discrimination: An Economic Perspective,” summarises the identification problem and the role of audit and correspondence studies.

Chang-Tai Hsieh, Erik Hurst, Charles Jones and Peter Klenow's work on occupational misallocation supports the claim that reduced barriers faced by women and Black Americans contributed materially to US aggregate growth. The exact share is model-dependent, so the narrative states the mechanism without presenting one estimate as settled.

Consequences and policy

Jonathan Ostry, Andrew Berg and Charalambos Tsangarides, Redistribution, Inequality, and Growth, supports the cautious conclusion that redistribution is not generally associated with shorter growth spells once redistribution and market inequality are distinguished, except that extreme cases and policy design matter. The IMF's current inequality material continues to treat excessive inequality as a potential threat to sustainable growth and macroeconomic stability.

Anthony Atkinson's Inequality: What Can Be Done? informs the policy-stack framing. The manuscript deliberately keeps detailed tax design within the boundary of Tax in a Hurry and uses taxation only as one stage in a wider distributional system.

Current verification

Current World Inequality Report, World Bank, OECD, ILO and IMF material was rechecked on 11 August 2026 during the fresh audit. Current figures are used sparingly because definitions, survey coverage and data vintages change. The conceptual claims do not depend on a single current estimate.

Bibliography

Books and major syntheses

Atkinson, Anthony B. Inequality: What Can Be Done? Cambridge, MA: Harvard University Press, 2015.

Milanovic, Branko. Global Inequality: A New Approach for the Age of Globalization. Cambridge, MA: Harvard University Press, 2016.

Piketty, Thomas. Capital in the Twenty-First Century. Translated by Arthur Goldhammer. Cambridge, MA: Harvard University Press, 2014.

Chancel, Lucas, Ricardo Gómez-Carrera, Rowaida Moshrif, and Thomas Piketty, eds. World Inequality Report 2026. World Inequality Lab, 2026.

Organisation for Economic Co-operation and Development. A Broken Social Elevator? How to Promote Social Mobility. Paris: OECD Publishing, 2018.

Organisation for Economic Co-operation and Development. To Have and Have Not - How to Bridge the Gap in Opportunities. Paris: OECD Publishing, 2025.

Research papers

Acemoglu, Daron. “Technical Change, Inequality, and the Labor Market.” NBER Working Paper 7800, 2000.

Chetty, Raj, Nathaniel Hendren, Patrick Kline, and Emmanuel Saez. “Where Is the Land of Opportunity? The Geography of Intergenerational Mobility in the United States.” Quarterly Journal of Economics 129, no. 4 (2014): 1553-1623.

Chetty, Raj, and Nathaniel Hendren. “The Impacts of Neighborhoods on Intergenerational Mobility I: Childhood Exposure Effects.” Quarterly Journal of Economics 133, no. 3 (2018): 1107-1162.

Chetty, Raj, David Grusky, Maximilian Hell, Nathaniel Hendren, Robert Manduca, and Jimmy Narang. “The Fading American Dream: Trends in Absolute Income Mobility Since 1940.” Science 356, no. 6336 (2017): 398-406.

Hsieh, Chang-Tai, Erik Hurst, Charles I. Jones, and Peter J. Klenow. “The Allocation of Talent and U.S. Economic Growth.” Econometrica 87, no. 5 (2019): 1439-1474.

Lang, Kevin, and Ariella Kahn-Lang Spitzer. “Race Discrimination: An Economic Perspective.” Journal of Economic Perspectives 34, no. 2 (2020): 68-89.

Ostry, Jonathan D., Andrew Berg, and Charalambos G. Tsangarides. Redistribution, Inequality, and Growth. IMF Staff Discussion Note SDN/14/02. Washington, DC: International Monetary Fund, 2014.

Institutional and data sources

International Labour Organization. Policy Measures to Address Inequalities and Increase the Labour Income Share. G20 Technical Paper No. 2. Geneva: ILO, 2025.

International Labour Organization. Employment and Social Trends 2026. Geneva: ILO, 2026.

International Monetary Fund. “Introduction to Inequality.” IMF topic resource. Accessed 11 August 2026.

World Bank. March 2026 Update to the Poverty and Inequality Platform. Washington, DC: World Bank, 2026.

World Bank. Poverty and Inequality Platform. Dataset version current 24 March 2026. Accessed 11 August 2026.