Racial Capitalism and Structural Accounts

8.4.2 established that unequal treatment occurs and that its magnitude differs several-fold between comparable countries. It did not explain why the arrangement exists, why it is so durable, or why it survives the disappearance of the beliefs that produced it — the finding that measured attitudes improved for decades while measured hiring discrimination did not.

This lesson covers the accounts that try to explain that, and it grades each on the same criterion the rest of this course uses: what does it predict that could turn out false?

A strike, and a slogan that should not be possible.

In 1922, on the goldfields of the Rand in South Africa, white mineworkers went on strike. The mine owners had proposed to reduce costs by allowing Black workers into semi-skilled jobs that had been reserved for whites — jobs paying several times what Black miners were paid for the work they were permitted to do.

The strike escalated into an armed rising. It was put down by the state with aircraft and artillery, with more than two hundred people killed.

And the banner carried by the strikers read: "Workers of the world unite and fight for a white South Africa."

Sit with that sentence, because it disposes of one theory immediately.

The simplest account of racial division in the workplace is that employers create it to weaken labour — divide and rule. Here the employers were trying to remove the colour bar, for the ordinary reason that it forced them to pay more; and the workers fought, with weapons, to keep it.

The pattern is not unique to South Africa. Exclusionary agitation against Chinese labour in the nineteenth-century American west was driven substantially by white workers and their organisations, not by employers, who were importing that labour precisely because it was cheaper. Craft unions in several countries operated racial exclusion in their membership rules for decades. In each case, the group defending the racial boundary was higher-priced labour, and the group attacking it was capital.

Which does not mean employers never benefit from division. They frequently have, demonstrably, and the strike-breaking use of workers from an excluded group is documented across many industries — followed, reliably, by the union excluding those workers afterwards, which guarantees their availability as strike-breakers next time.

So the honest starting point is that racial boundaries in labour markets have been made and defended by different parties at different times, for different reasons — and any theory that assigns the beneficiary in advance will be wrong about half the cases.

The accounts

One — the split labour market, which is the most testable.

Bonacich's theory begins by rejecting a two-class picture and proposing three: employers, higher-priced labour, and cheaper labour.

Ethnic antagonism arises, on this account, when labour of different price is available for the same work. The price difference need not be about ethnicity at all initially — it can arise from differences in resources, information, political rights, or the alternatives available at home.

Employers prefer the cheaper labour. Higher-priced labour, threatened with displacement, has two available responses:

Exclusion — keep the cheaper group out entirely, through immigration restriction or a bar on entry to the trade. A caste arrangement — admit them but confine them by rule to a defined set of inferior positions, so they cannot compete for the protected ones. This is a job colour bar, and it is precisely what the Rand strikers were defending.

What the theory predicts, and this is why it is worth having. Antagonism should be most intense where a price differential exists for the same work and higher-priced labour has the political capacity to act on it — and it should be absent where either condition fails. That is a testable claim, and the historical cases it was built on fit it well.

Its weakness is also specific. It explains what happens once a price differential is racialised. It does not explain why the differential falls along racial lines in the first place — which usually requires a prior history of conquest, slavery, colonial recruitment or immigration policy. The theory presupposes what most needs explaining, and its defenders accept this.

Two — the divide-and-rule account, and the evidence for and against.

The claim that a racially divided workforce weakens labour as a whole, so that all workers are worse off than under solidarity.

The evidence for it that is worth knowing is a finding that across American metropolitan areas, greater racial inequality was associated with lower incomes for white workers too, and with weaker unions — consistent with division costing everyone rather than transferring from one group to another.

And the evidence against it is the story. The Rand and the Californian exclusion movements are cases where the higher-priced group's action was directly self-interested and successful in the short run. Both can be true: division can lower the wage floor for everyone in aggregate while delivering real protected advantages to a particular group of workers in a particular labour market. The mistake is to insist on one.

Three — the public and psychological wage.

Du Bois's argument from Black Reconstruction (see 4.5.4): that white workers in the American South accepted low material wages compensated in part by a public and psychological wage — deference, access to public spaces and functions, treatment by police and courts, and the schools and public institutions attached to their status.

The argument's power is that it identifies a currency other than money, which is why it explains behaviour that looks economically irrational.

Its difficulty is measurement. Deference and standing are real and are hard to quantify, and the causal claim — that this compensation was what secured white workers' acquiescence — is not straightforwardly testable. It remains one of the most influential arguments in the field and one of the least settled empirically, and both halves of that should be said.

Four — racial capitalism.

The term has two distinct uses, and conflating them causes a great deal of confusion.

The South African usage came first and is the more specific. It described a documented mechanism: apartheid's system of reserves, pass laws and migrant labour kept the cost of Black labour below its full reproduction cost, because the reserves absorbed the costs of raising children, of illness and of old age, which the employer would otherwise have had to cover in the wage. Racial domination was, on this account, doing identifiable economic work, and the argument was made with reference to specific legislation and specific accounts.

The later and now more widespread usage makes a broader historical claim: that capitalism did not become racial at some later date but emerged from a European order already structured by racialised hierarchies, so that racialism is constitutive of it rather than an addition to it.

Three things must be said about the second. It is a genuine and important corrective to accounts that treat racial domination as incidental to economic development, and it foregrounds the role of slavery and colonial extraction in the formation of capitalism, which orthodox accounts frequently minimised (see 6.7.1).

The specific historical claim about pre-modern European hierarchies is contested among historians, including sympathetic ones.

And the term has been applied so widely that it now often functions as a name for "capitalism and racism together" rather than as a mechanism. As a claim about origins it is substantive and arguable. As an explanation of present variation it forbids very little — it does not predict why hiring discrimination should be several times larger in one rich democracy than in another, which is the variation most in need of explaining.

Five — racial formation.

Omi and Winant's framework: race is not a fixed thing but an unstable, politically contested set of meanings, remade over time through racial projects that connect a representation of race to an organisation of resources. The state is the central site, because it is where classifications are made and rights attached (see 8.4.1).

Its strength is that it explains change and variation — why categories differ between countries and shift within them, which fixed accounts cannot.

Its limitation is that it is a framework rather than a mechanism. It tells you where to look and what to call things; it does not on its own predict which projects will succeed.

Six — systemic racism and the account of colour-blind ideology.

Bonilla-Silva's argument: societies are racialised social systems in which economic, political, social and ideological levels are structured by racial categories, so that racial outcomes are produced by the ordinary operation of institutions rather than by individual prejudice.

And its most useful component is the analysis of how this is justified once explicit prejudice becomes unsayable. Four recurring frames are identified in ordinary talk: abstract liberalism (invoking equal opportunity and individual choice to oppose measures that would change outcomes); naturalisation (treating segregation and sorting as natural preference — "people just prefer their own"); cultural racism (explaining outcomes by group culture rather than group biology); and minimisation (holding that discrimination is no longer a significant obstacle).

Why this matters empirically: it predicts precisely the pattern 8.4.2 documented — explicit prejudice declining continuously while discriminatory behaviour does not — and it supplies a mechanism for how that combination is sustained.

And the criticism deserves stating. A framework in which any outcome can be described as evidence of a racialised system risks being unfalsifiable (see 7.2.3). The defence is that the specific mechanisms are specifiable and testable — and the audit literature, the redlining evidence below and the wealth decompositions are exactly that. The framework is worth what its mechanisms are worth, which is the right standard to hold it to.

The case where structural accounts are strongest

The wealth gap, where the mechanism is documentable and requires no current discriminator.

In the United States, the racial gap in household wealth is far larger than the gap in income — the income ratio between white and Black median households is in the region of 1.6 to 1.7 to one, while the median wealth ratio has been of the order of six to eight to one. And it has closed far less over time than the income gap has.

Why wealth behaves differently is 8.3.3's answer: it accumulates, earns returns, is inherited, and is dominated by housing for most households. Which makes it the variable most sensitive to historical policy, because a rule applied decades ago is still compounding.

And the historical policy is documented rather than inferred.

Federal mortgage insurance in the United States from the 1930s operated with underwriting guidance that treated the entry of other racial groups into a neighbourhood as a threat to property values, and encouraged restrictive covenants. Government-sponsored maps graded neighbourhoods for lending risk, with the lowest grade — marked in red — assigned in large part on the racial composition of the residents.

The effect was to channel decades of subsidised, long-term, low-deposit mortgage lending — the instrument that built the mid-century middle class — overwhelmingly to white households, at precisely the moment when suburban property was appreciating fastest.

And the long-run effects have been estimated with a credible design. Comparing neighbourhoods on either side of the graded map boundaries — a boundary discontinuity design (see 7.4.3) — researchers found persistent differences in segregation, homeownership rates, house values and credit access decades later, on either side of lines drawn in the 1930s.

This is what a strong structural claim looks like. A documented rule, applied at a datable moment, with a specified mechanism of transmission (housing wealth compounding and being inherited), and an identification strategy that separates the rule's effect from everything correlated with it. No currently living person needs to discriminate for the effect to persist, which is the definition of structural, and it is checkable.

Three errors, one on each side and one shared.

Functionalist reasoning. "This arrangement benefits group X, therefore it exists and persists." Benefit is not a mechanism (see 5.1.1). The redlining case works because the mechanism is specified; a claim that racial inequality persists because capital benefits, with no account of through whose decisions, is an assertion.

Demanding an individual discriminator. The mirror error: refusing to accept a structural claim unless someone is currently doing something wrong. The whole point of the wealth evidence is that the rule was abolished and its effects were not, because they had been converted into assets. A structure does not need a live perpetrator.

And the shared error: using "structural" as a synonym for "large" or "important". It has a specific meaning — an effect produced by the arrangement of rules, resources and institutions rather than by the decisions of identifiable actors — and a claim that uses the word without specifying the arrangement has said only that the speaker takes the problem seriously.

Because these accounts differ in what they would have you do, and in what would show them wrong.

Four questions for any structural account of racial inequality.

What is the mechanism, stated as a sequence? Who decides what, on what basis, with what consequence that persists.

What does it forbid? If every possible outcome is consistent with it, it is a description (see 7.2.3).

Does it explain the variation? The most demanding fact in this Topic is that discrimination differs several-fold between comparable rich democracies (see 8.4.2). An account that explains why racial inequality exists in general, but not why it is much larger in one place than another, has explained the easier half.

And does it explain persistence without explicit prejudice? This is the fact requiring explanation, and the accounts that do best on it are the ones specifying institutional mechanisms rather than attitudes.

Where this leaves the field, stated proportionately.

The split labour market theory is the most testable and explains a specific class of cases well, while presupposing the racialisation it starts from.

Divide-and-rule and the protected-advantage account are both partly right, in different labour markets, and the insistence on one is the error.

Racial capitalism in its specific South African form identified a documented mechanism; in its broad form it is a substantive claim about origins that forbids little about the present.

Racial formation is the best framework for change and variation and is not itself a mechanism.

And the systemic account is worth exactly what its specified mechanisms are worth — which, in the case of housing, credit and hiring, has turned out to be a great deal.

One closing point that connects this Topic to the next. Everything here has treated race and class as interacting. Neither reduces to the other: the split labour market shows class actors defending racial boundaries, and the wealth evidence shows racial policy producing class positions. Which is why 8.7 exists, and why the question Fraser and Honneth posed in 8.1.3 — whether these are one axis or two — is still open when you get there.

In 1922 white miners on the Rand struck under the banner "Workers of the world unite and fight for a white South Africa", against employers trying to remove a colour bar that raised their costs. Racial boundaries in labour markets have been made and defended by different parties at different times, so any theory that assigns the beneficiary in advance is wrong about half the cases.

Two questions must be kept apart: who benefits, and why it persists. Benefit is not a mechanism — the standard error is functionalist reasoning.

Six accounts. The split labour market — employers, higher-priced labour and cheaper labour, with antagonism arising where a price differential exists for the same work and higher-priced labour responds with exclusion or a caste arrangement. Testable, fits its cases, and presupposes the racialisation it begins from. Divide-and-rule, supported by evidence that greater racial inequality accompanied lower white incomes and weaker unions, and contradicted by the cases where higher-priced labour drove exclusion — both are partly right. Du Bois's public and psychological wage, which identifies a currency other than money and remains among the most influential and least settled arguments in the field.

Racial capitalism has two uses: the specific South African account, in which reserves and migrant labour kept the cost of labour below its reproduction cost — a documented mechanism — and the broader claim that racialism is constitutive of capitalism's origins, which is a substantive historical argument that forbids little about present variation.

Racial formation explains change and variation through contested racial projects with the state as the central site, and is a framework rather than a mechanism. And the systemic account — with its four frames of abstract liberalism, naturalisation, cultural racism and minimisation — predicts exactly the pattern 8.4.2 found: explicit prejudice falling while behaviour does not. It is worth what its specified mechanisms are worth.

The strongest structural case is wealth. The American racial wealth ratio has been of the order of six to eight to one against an income ratio nearer 1.7, because wealth compounds and is inherited. Mid-century mortgage insurance guidance and neighbourhood risk maps channelled subsidised lending away from Black neighbourhoods, and a boundary discontinuity design finds persistent differences in segregation, homeownership, values and credit decades later. A documented rule, a specified transmission mechanism, and an identification strategy — and no living discriminator required.

Three errors: functionalist reasoning; demanding a live perpetrator before accepting a structural claim; and using "structural" to mean "large".

Split labour market — three classes, with antagonism arising from price differentials for the same work.

Exclusion / caste arrangement — the two responses of higher-priced labour: keeping the cheaper group out, or confining it by rule.

Divide and rule — the account in which employers benefit from a divided workforce.

Public and psychological wage — Du Bois: non-monetary compensation in deference, access and treatment.

Racial capitalism — in its specific form, racial domination lowering the reproduction cost of labour; in its broad form, the claim that racialism is constitutive of capitalism's emergence.

Racial formation / racial project — race as contested meanings, remade by projects linking representation to the distribution of resources.

Racialised social system — a society whose institutions systematically produce racial outcomes without requiring individual prejudice.

Colour-blind frames — abstract liberalism, naturalisation, cultural racism, minimisation.

Redlining — the grading of neighbourhoods for lending risk substantially by racial composition.

Boundary discontinuity design — comparing outcomes either side of an administrative line to identify its effect.

Structural effect — an outcome produced by the arrangement of rules and resources rather than by identifiable current decisions.

One — apply the split labour market. Find a historical or current case of hostility between worker groups and ask whether a price differential for the same work exists, and which group has the political capacity to act.

Two — demand the mechanism. Take any structural claim about racial inequality and write it as a sequence: who decides what, on what basis, producing what that persists. If you cannot, note where it breaks.

Three — trace an asset. Take one historical housing or lending rule in your own country and follow its effects forward using 8.3.3's channels. The question is what it would look like in a wealth distribution today.

Four — test falsifiability. For each of the six accounts, write one observation that would count against it. Some will be much easier than others, and that ranking is informative.

Five — explain the variation. Take the fact that measured hiring discrimination differs several-fold between comparable rich democracies, and ask which of the six accounts could explain it. This is the hardest question in the Topic and the one most worth sitting with.