Difference, Inequality and Injustice

Almost everyone agrees that inequality exists. Almost nobody agrees about what follows from that, and a surprising amount of the disagreement is not about values at all. It is about three different things being called by one word.

Difference is variation without ranking. Inequality is variation along a dimension of advantage. Injustice is inequality that should not be there.

Moving between them requires arguments, and this lesson is about which arguments — and about the specific factual questions that most normative disagreements actually turn on, which is where sociology earns its place in the argument.

One factory floor, five pairs of workers.

The first pair. Priya works nights and Ravi works days. Priya prefers nights — the shift suits her, she likes the quiet, she sleeps well. Ravi could not manage it. They differ. Neither is better off. This is difference, and there is nothing more to say about it sociologically except that people vary.

The second pair. Anil has been on the line eleven years and earns more than Sunita, who joined last spring. This is an inequality — one has more of something both value. Most people, asked, would say it is not unjust: it reflects experience, it applies to everyone the same way, and Sunita will get there. Notice what the justification is doing: it appeals to a rule that is general, transparent, and that everyone can eventually satisfy.

The third pair. Meena and Farida stand at the same machine, doing identical work at the same speed. Meena is paid 30 per cent less, because she is employed by an agency on a contract the firm adopted specifically to avoid paying the rate. Most people call this unjust, and the reason is available: the difference tracks nothing about the work, the two women cannot move between categories, and the arrangement was designed to produce the gap.

The fourth pair. The plant manager earns forty times what Farida earns. Here reasonable people genuinely disagree, and they disagree for stateable reasons: whether the job requires scarce skills, whether forty times is what it takes to attract someone, whether the manager's decisions have forty times the consequence, whether the ratio is set by a market or by a board of people in similar positions. These are partly factual disputes, and some of them can be settled with evidence.

And the fifth pair are not in the factory. Farida's daughter and the manager's daughter are both seven. One will attend a school with forty in a class; the other will not. Neither of them chose anything, neither has done anything, and by the time either is old enough to compete, the competition will have been substantially decided.

Five pairs. One is difference. Four are inequalities. They are not the same kind of thing, and an argument that treats them as one — in either direction — is not an argument about inequality but a way of avoiding one.

Three questions get fused into one, and separating them is most of the work.

The descriptive question: how much inequality is there, of what, between whom, and how is it changing? This is empirical and answerable (see 8.1.2).

The explanatory question: why does it exist and why does it persist? Also empirical, harder, and the subject of most of this Part.

The normative question: is it unjust, and what if anything should be done? This is not settled by evidence (see 7.1.2 on the warring gods).

And here is the thing that makes sociology indispensable rather than optional. Almost every normative position about inequality contains factual claims, and those claims are checkable.

"People get what they earn" — a claim about whether rewards track effort and contribution. "Anyone can make it" — a claim about mobility rates. "They chose that path" — a claim about the range of options available. "That's just how societies are" — a comparative claim, refutable by one society that is different. "Inequality is the price of growth" — an empirical claim about a trade-off. "Redistribution would not help" — an empirical claim about policy effects.

None of these settles the value question. All of them can be wrong, and when they are, the normative position built on them needs rebuilding. That is the division of labour, and it is why this Part matters even to someone who has already decided what they think.

Difference into inequality

What has to be true before a difference is an inequality.

Three things.

A dimension that is valued. Money, health, time, safety, respect, autonomy, security, education, political voice. Variation on something nobody wants more of is not inequality, which is why "people are different" is a true statement that does almost no work.

A ranking. More of it is better. Some differences resist this genuinely — musical taste, temperament, the night shift for Priya — and the political uses of the confusion run in both directions. Presenting an inequality as a mere difference ("some cultures simply value education more") is one move. Presenting a difference as an inequality — treating an unranked variation as a deficiency — is the other, and it is the mechanism 6.1.1 described.

And people who can be compared on it. Which requires deciding who is in the comparison, and that decision does more work than almost anything else. The same world contains falling inequality between countries and rising inequality within many of them — driven largely by very large populations in Asia growing faster than rich countries while inequality inside most countries widened. "Is inequality rising?" has no answer until you say between whom (see 7.6.4).

Vertical inequality — between individuals or households, ranked from top to bottom. What the standard measures capture.

Horizontal inequality — between groups defined by identity: region, ethnicity, religion, language, caste. Frances Stewart's argument is that these are the politically explosive ones. A society can have moderate vertical inequality and severe horizontal inequality, and the evidence associates group-based inequality — particularly when economic, political and cultural disadvantage coincide in the same group — with organised conflict far more strongly than the overall income spread does. The reason is intuitive: a group has a name, a history and an organisation; a percentile does not.

Inequality of outcome — what people end up with. Inequality of opportunity — what chances they had. Inequality of condition — the circumstances they started from.

And the three are linked in a way that makes "equal opportunity, unequal outcome" harder to hold than it sounds: one generation's outcomes are the next generation's conditions. Large outcome inequality among parents produces large condition inequality among children, which produces unequal opportunity by any definition. Equal opportunity across generations therefore places limits on how unequal outcomes can be — which is an empirical relationship, not a slogan, and 8.3 is where it gets measured.

Positional goods. Some goods are inherently relative: the top jobs, the scarce places, the desirable neighbourhood, the qualification that distinguishes you. Everyone can get richer and nobody can move up the queue, because the queue is the good. This is why rising absolute living standards can coexist with intense and unrelieved competitive pressure — and why credential inflation happens (see 9.2).

Distribution versus mobility. The same distribution can contain a great deal of movement or almost none, and these are separate facts about a society (see 8.3).

Tilly's question: why does inequality stick to categories?

The most useful single framework in this Part, and it reorients the whole subject.

Charles Tilly asked a question that had been oddly neglected. Inequality between individuals could be random — some people are luckier, quicker, harder-working. So why is inequality so reliably organised around categories? Male and female. Citizen and foreigner. Black and white. Brahmin and Dalit. Permanent staff and agency staff.

His answer: because categorical inequality is easier to sustain, and it is sustained by four mechanisms that operate together.

Exploitation. A group controls a resource, enlists the effort of others in producing value from it, and excludes them from the full value produced. The key point is that the advantaged party depends on the disadvantaged one — this is not neglect but a relationship.

Opportunity hoarding. A group secures access to a valuable resource and restricts entry: a licensed profession, a catchment area, a network of referrals, an apprenticeship that runs through families. No exploitation is needed — nobody's labour is being taken — only a closed door. Weber's social closure, made concrete (see 4.3.6).

Emulation. Categorical distinctions established in one setting are copied into others because they are ready-made and cheap to install. The distinction that organised the workshop organises the union, the club and the school. This is why the same boundaries appear across institutions that have no connection.

Adaptation. Everybody, on both sides, builds routines, relationships and identities around the existing arrangement — so that changing it disrupts things people value even when they do not endorse the inequality. This is the mechanism that makes unjust arrangements durable without anybody defending them.

Why this framework is worth carrying. It explains what individual-level accounts cannot: why inequality tracks lines that have nothing to do with individual attributes, why those lines persist across generations, why the same lines appear in unrelated institutions, and why arrangements survive the disappearance of the beliefs that created them.

And it converts a moral intuition into a research programme. For any inequality: which of the four mechanisms is operating, at what site, maintained by whom? That is answerable (see 7.5.4, 7.9.1).

From inequality to injustice

Five justifications, and the empirical question each one raises.

Desert. People deserve what they get because of effort, talent or contribution. The empirical question: does the distribution actually track those things? How much of the variation in earnings is accounted for by effort and contribution, and how much by inheritance, birthplace, timing, network and category? This is measurable, and the answers are uncomfortable for the justification.

Incentives. Unequal rewards are necessary to get demanding roles filled by capable people. The functionalist argument from 5.1.1, and Tumin's reply still stands: it assumes the talent pool is open, which inherited advantage prevents; it takes the existing reward structure as evidence of functional importance, which is circular; and it does not explain the size of the differentials, which vary enormously between societies with similar productivity. The empirical questions: does the pool select on talent? Do smaller differentials produce worse allocation?

Need. Distribution should follow need. The empirical questions are about measurement — whose needs, established how (see 7.2.2) — and this is Sen's territory (see 8.1.3).

Choice. People chose differently, so different outcomes are legitimate. The empirical question is what the choice set was, and whether the "choices" were made under constraint, with what information, at what age, by people who could see the alternatives.

Entitlement. If each step in a history of transfers was legitimate, the resulting distribution is legitimate however unequal — Nozick's position, and its force comes from being about process rather than pattern. The empirical question it raises is the one it invites: was each step legitimate? Applied to real societies with histories of conquest, slavery, expropriation, colonial extraction and legally enforced exclusion, the principle turns out to be extraordinarily demanding — it requires rectification of every unjust transfer, which no society has attempted.

And the pole opposite Nozick is Rawls, whose difference principle permits inequalities only where they work to the benefit of the least advantaged, derived from what people would choose without knowing which position they would occupy.

Both positions are philosophical and neither is settled by data. But notice: Rawls needs to know whether a given inequality actually benefits the worst-off, and Nozick needs to know whether the historical transfers were just. Both cash out into empirical questions, and both have been used by people who did not check.

What people believe about inequality is systematically wrong, in a patterned direction.

People underestimate how unequal their society is. In a well-known American study, respondents estimated that the wealthiest fifth held around 59 per cent of the country's wealth; the actual figure at the time was around 84. Asked what they would prefer, they chose a distribution far more equal than either — with the top fifth holding around 32 per cent.

And the preference was remarkably consistent across income, politics and sex. People who disagreed sharply about policy agreed substantially about what a desirable distribution looked like, and all of them were badly wrong about the actual one.

People also misplace themselves. Across many countries, respondents cluster their self-placement towards the middle of the distribution regardless of where they actually sit. The rich systematically think they are less rich than they are, and the poor systematically think they are less poor. The mechanism is straightforward and sociological: people estimate their position from the people they know, and social networks are strongly sorted by class (see 2.2.7).

Three consequences worth carrying.

Public opinion about inequality is partly an opinion about a false picture. Preferences expressed under a mistaken belief about the facts are not a straightforward guide to what people want.

Correcting the picture changes attitudes less than you would expect, and the experimental literature finds informing people about actual inequality shifts stated concern more than it shifts policy support — which is itself a finding about how policy attitudes are held.

And this is why 8.1.2 comes next. You cannot argue about how much inequality is acceptable without knowing how much there is, and almost nobody does.

Because the strongest arguments in this area are the ones that separate the three questions, and almost nobody does.

Four habits.

One — when someone says "inequality", ask: of what, between whom? Income, wealth, health, safety, time, respect, voice — these move differently and sometimes in opposite directions. Between individuals, households, groups, regions, countries or generations. A claim without both is not yet a claim.

Two — separate the descriptive claim from the normative one, in both directions. "Inequality has risen" is a factual statement that does not by itself establish anything is wrong. "That's unfair" is a value judgement that does not by itself establish anything is happening. People who agree about the facts can disagree about the values, and that disagreement is legitimate (see 7.1.2).

Three — find the factual claims inside the normative argument, and check them. Every justification listed above rests on something checkable. This is the single most useful thing sociology contributes to a political argument, and it does not require taking a side.

Four — ask Tilly's question. Which categorical boundary is this inequality organised around, and which of the four mechanisms maintains it — exploitation, opportunity hoarding, emulation, adaptation? This turns indignation into a research question, which is 1.1.1's whole programme applied to the discipline's central subject.

And one last thing about the five pairs. The argument about the manager's forty-fold salary is the one that gets the attention — it is dramatic, and it has two clean sides. The argument about the two seven-year-olds is the one that decides most of what happens, and it has no clean sides at all, because nobody involved has done anything. Which of those a society spends its attention on is itself a sociological fact.

Difference is variation without ranking; inequality is variation along a dimension of advantage; injustice is inequality that should not be there — and moving between them requires arguments, which are constantly skipped.

Three questions get fused: how much (descriptive), why (explanatory), and is it wrong (normative). The third is not settled by evidence — and it almost always contains factual claims that are. "People get what they earn", "anyone can make it", "they chose it", "that's how societies are", "it's the price of growth" are all checkable, and when they fail the position built on them needs rebuilding.

A difference becomes an inequality when there is a valued dimension, a ranking, and a defined comparison group — and the third does enormous work: inequality between countries has fallen while inequality within many has risen, so "is inequality rising?" has no answer until you say between whom.

Key distinctions: vertical against horizontal inequality — Stewart's argument that group-based inequality, especially where economic, political and cultural disadvantage coincide, predicts organised conflict far better than the overall spread; outcome, opportunity and condition — linked because one generation's outcomes are the next's conditions, which places limits on how unequal outcomes can be if opportunity is to be equal; positional goods, where everyone can gain absolutely and nobody can move up the queue; and distribution against mobility.

Tilly's durable inequality is the framework to carry. Inequality attaches to categorical pairs, maintained by exploitation (a relationship, not neglect), opportunity hoarding (a closed door, no exploitation required), emulation (ready-made boundaries copied between institutions) and adaptation (routines built around an arrangement, which makes it survive the beliefs that created it).

Five justifications, each with an empirical question: desert (does the distribution track effort and contribution?), incentives (is the pool open, and do smaller differentials allocate worse?), need, choice (what was the choice set?), and entitlement — Nozick's principle being far more demanding than it appears once real histories of expropriation are examined. Rawls needs to know whether an inequality benefits the worst-off; Nozick needs to know whether the transfers were just. Both cash out empirically.

And people are systematically wrong about the facts: underestimating concentration substantially, preferring something far more equal than either their estimate or reality, and placing themselves towards the middle wherever they actually sit — because they estimate from networks sorted by class.

Difference / inequality / injustice — unranked variation; variation on a valued dimension; inequality that should not exist.

Vertical / horizontal inequality — between individuals; between identity-defined groups.

Outcome / opportunity / condition — what people end with, the chances they had, the circumstances they started from.

Positional good — one whose value depends on relative position, so absolute gains cannot be shared.

Durable inequality — Tilly: inequality organised around categorical pairs.

Exploitation / opportunity hoarding / emulation / adaptation — the four mechanisms that sustain categorical inequality.

Social closure — restricting access to opportunities to an in-group.

Desert / incentives / need / choice / entitlement — the five standard justifications, each resting on a checkable claim.

Difference principle — Rawls: inequalities permitted only where they benefit the least advantaged.

Entitlement theory — Nozick: a distribution is just if its history of transfers was just.

Perception gap — the systematic underestimation of inequality and clustering of self-placement towards the middle.

One — sort your own examples. Write down five inequalities you have noticed this month. For each, mark whether it is a difference, an inequality you think justified, or one you think unjust — and write the reason in each case. The reasons will be one of the five justifications.

Two — find the factual claim. Take an argument about inequality you disagree with. Identify the empirical claim inside it, and write down what evidence would test it. Then do the same for an argument you agree with.

Three — apply Tilly. Take one inequality organised around a category — a contract type, a nationality, a caste, a sex. Ask which of the four mechanisms is doing the work, and where.

Four — test the perception gap. Before looking anything up, estimate what share of your country's wealth is held by the richest fifth, and what share you would prefer. Then look it up.

Five — do the two seven-year-olds. Write down everything that will differ between the two children in the story by the time they are eighteen, and mark which of those either child had any influence over. Then decide which of the five justifications could apply to that list.