Did the West get rich because of slavery?
For forty years the textbook answer was no — slavery was a moral horror but a small slice of GDP. The last fifteen years of scholarship have made that answer harder to keep.
Start on the plantation
“In order to understand the brutality of American capitalism, you have to start on the plantation.”
— Matthew Desmond, The 1619 Project, The New York Times Magazine, August 14, 2019
Desmond’s essay won a Pulitzer the following spring. Within four years the framing had moved from a left-wing academic argument into school curricula across several states and a Hulu docuseries, and for a generation of readers it had become the default story of where American wealth came from. The argument is not that slavery was a stain on an otherwise clean ledger. It is that the ledger itself — the management techniques, the credit instruments, the accounting conventions of modern capitalism — was written on the plantation.
Two questions are welded together in this discourse, and the first job is to pry them apart. The historical question is whether Atlantic slavery’s profits and institutional infrastructure were load-bearing for industrial Britain and the early United States — a counterfactual about output and growth. The moral-political question is what is owed today. They are related, but they are not the same question, and the failure mode on both sides of this debate is to let an answer to one stand in for an answer to the other.
The historical question reduces to apparatus we already have. The machinery for “how much did this input contribute to growth?” is sources-of-growth decomposition — the capital-share accounting that lets you ask what fraction of the takeoff a given pile of capital can explain. That apparatus lives in Economics Ch.13 (Growth Theory). The historical record of the phenomenon itself — the roughly twelve million people carried across the Middle Passage, the plantation complex, the abolitions of 1807 and 1833, the labor regimes that followed emancipation — lives in Economic History Ch.9 (Atlantic Slavery and After). We are not running the decomposition yet. We are locating where it runs.
Start with the weight already attached to the framing, because this is not a fringe historians’ quarrel anymore. The Caribbean Reparations Commission published its Ten Point Plan in 2014. HR 40, the bill to study a reparations proposal, was reintroduced in the US Congress in January 2025. The University of Glasgow signed a £20 million restorative-justice agreement with the University of the West Indies in 2019 after auditing the slave wealth in its own endowment; Cambridge published its Legacies of Enslavement report in 2022. Universities, parliaments, and national commissions are treating the historical question as live because they think money turns on the answer. That is the terrain the historiography has to be argued on.
Take Desmond’s structural claim at its strongest, which is sharper than “slavery was bad and also there was an economy.” The claim is that the cotton plantation was a site of managerial innovation: depreciation schedules run against human capital, productivity measured to the picked pound under the threat of the whip, hierarchical labor control through calibrated violence. On this reading the plantation is a candidate seed for the twentieth-century industrial-management toolkit — the spreadsheet before the spreadsheet. The cheap move against it is to point out that the GDP magnitudes are contested and walk away. The cheap move for it is to say the moral stakes are obvious and therefore the economics must follow. Both cheap moves are the failure the rest of this walkthrough is built to avoid.
Where this leaves us
The popular discourse is asking a real question, and it is a question the textbook tradition gave a confident answer to fifty years ago and is no longer giving as confidently. The walkthrough’s job is to walk the dispute as it actually stands — not to flatten it into a Pulitzer-winning narrative, and not to dismiss it in a sentence about the small share of GDP. Both flattenings lose the part that is genuinely unsettled.
The framing has a genealogy older than 2019. It traces to a Trinidadian historian’s Oxford doctoral thesis, defended in 1938 — a man who would become his country’s first prime minister. His name was Eric Williams, and for forty years the economics profession thought it had answered him.
Williams, Engerman, and the long swing
“The profits obtained provided one of the main streams of that accumulation of capital in England which financed the Industrial Revolution.”
— Eric Williams, Capitalism and Slavery, 1944
Williams defended the thesis at Oxford in 1938, published it during the Second World War, and was elected the first prime minister of Trinidad and Tobago in 1962. His argument has two halves that modern discourse routinely fuses. The first: slavery and the slave trade financed the Industrial Revolution. The second: slavery was abolished when it stopped paying. The literature mostly rejected the second half decades ago. It is the first half — the financing claim — that is being re-fought right now.
Translate Williams’s financing claim into growth-theoretic terms and it becomes a precise counterfactual: British output per capita from 1750 to 1850, with the Atlantic slave complex removed from the inputs. The specific channel Williams names is the capital channel. Slave-economy profits, recycled through the merchant houses of London and Liverpool, were a large enough share of British capital formation that removing them removes the takeoff. That is a claim you can put a number on.
The cliometric turn put the number on it, and the number was small. Stanley Engerman estimated in 1972 that slave-trade and West India profits ran around half a percent of British national income. Robert Fogel and Engerman’s Time on the Cross (1974) found American slavery to be profitable and efficient as a business but not load-bearing for the macroeconomy. The consensus benchmark settled at four to five percent of British GDP at the end of the eighteenth century — non-trivial, real, but not the engine. The intellectual lineage of this cliometric program — Fogel’s quantitative method arriving alongside the broader counter-revolution in economics — lives in History of Economic Thought Ch.10 (Counter-revolution).
Write British output as $Y = A \cdot F(K, L)$, with $A$ the level of technology and institutions, $K$ capital, $L$ labor. The Engerman tradition asks: what fraction of $K$ at the start of industrialization came from Atlantic slave profits, and how much of the takeoff does the standard model attribute to that fraction?
$$\Delta Y / Y \approx \alpha \cdot (\Delta K / K), \qquad \alpha \approx \tfrac{1}{3}$$With a capital share $\alpha$ around one-third and slave profits around four to five percent of GDP, the contribution the model attributes to the slave-capital channel comes out in single digits. The structural question Stage 3 reopens is whether the model is even measuring the right channel — whether $A$ itself, the legal and financial substrate, was built on slavery in ways $F(K, L)$ cannot see.
Industrialization is a fire that takes off in the 1780s. Williams says the fuel was Atlantic slave profits. Engerman measured the pile and found it about four or five percent of the energy in the system — the fire burns without it. The neo-Williamsite revival does not deny Engerman’s measurement of that pile. It argues he measured the wrong pile.
Give the Engerman position its full strength, because Stage 3 needs something solid to push against. The claim is not that slavery did not matter. It is that the specific macroeconomic channel through which slavery would have caused industrialization — profits, accumulated and reinvested as capital — is measurable, and the measurement does not support the strong Williams claim. The takeoff had many simultaneous inputs: coal geography, the stability of British property rights, a scientific and tinkering culture, deep market integration. Slavery was one input among several, and on the capital-flow accounting it was not the decisive one. This is a serious finding, arrived at with serious method.
It held the field for forty years, and the second half of the Williams thesis fared even worse. Seymour Drescher’s Econocide (1977) demolished the “abolished-when-it-stopped-paying” claim by showing that the British slave economy was thriving and profitable at the moment Britain chose to destroy it — abolitionists were dismantling a going concern, not euthanizing a dying one. By the early 2000s the textbook treatment was settled: slavery was a profitable, efficient, morally monstrous business that sat alongside the Industrial Revolution rather than under it.
Where this leaves us
Williams’s strong claim — slavery as the financing of industrial takeoff in a capital-flow sense — does not survive the cliometric work. But the Engerman tradition’s confident step from “slavery was a small share of GDP” to “slavery was not load-bearing for industrialization” rests on an assumption: that GDP-share is the right thing to measure. That assumption is exactly what cracked.
In 2014 two books arrived in the same season — one by a Harvard historian, one written into the broader New History of Capitalism — arguing that the Engerman tradition had spent forty years measuring the wrong thing. They did not dispute the size of the pile. They disputed which pile.
War capitalism and the eleven-percent number
“The early industrial capitalism of the cotton empire rested on violence and bodily coercion — the domination of masters over slaves and of frontier capitalists over indigenous inhabitants.”
— Sven Beckert, Empire of Cotton: A Global History, 2014 (Bancroft Prize; Pulitzer finalist)
Beckert reframes cotton as a single transatlantic system: African slave-raiding kingdoms, Caribbean and US Southern plantations, Liverpool credit houses, Manchester mills — one circulation of capital, violence, and coerced labor. The Manchester mill is not standalone British engineering; it is one node in a circuit. Beckert calls the whole system war capitalism, and the name is the argument: the thing that came before industrial capitalism was not peaceful commerce but armed expropriation.
The revival runs on two strands. The first is structural and institutional. Beckert in Empire of Cotton, Walter Johnson in River of Dark Dreams, and the broader New History of Capitalism argue that the financial and institutional plumbing of industrial takeoff — bills-of-exchange systems, Lancashire mill finance, New York merchant credit chains, the early American banking system — was wired into the Atlantic slave economy in ways a standard sources-of-growth decomposition is structurally blind to. The load-bearing point: if slavery’s contribution operated through the institutional channel, then GDP-share returns the wrong answer, because it is pointing at the wrong outcome variable. The apparatus that is blind to this channel is the one in Economics Ch.13 (Growth Theory).
The second institutional door is the one Daron Acemoglu, Simon Johnson, and James Robinson opened with their settler-mortality work — the basis of their 2024 Nobel. Their argument is that whether a former colony industrialized depends on which institutions the colonizers installed, and those depended on the labor regime they imposed. A meaningful slice of the institutional variance that work attributes to colonial labor regimes runs through slavery as the regime in question. The formal apparatus — the instrumental-variable identification, the extractive-versus-inclusive frame — lives in Economics Ch.18 (Institutional Economics).
The second strand is quantitative, and it is the one that should make a cliometrician sit up. Maxine Berg and Pat Hudson — Hudson the author of the standard British Industrial Revolution textbook in 1992, not an outsider — argue in Slavery, Capitalism and the Industrial Revolution (2023) that the slave-plantation complex was so embedded across textiles, iron, copper, shipping, banking, insurance, and consumer culture that any account treating slavery as peripheral is empirically untenable. Their estimate of the slave economy’s share of British GDP at the end of the eighteenth century is closer to eleven percent than to four or five. Eleven versus five is not a rounding difference. It is a doubling of the basis-parameter of the forty-year consensus.
Argue the neo-Williamsite case at its strength, and the first thing strength requires is separating the strong popular version from the strong defensible one. The popular rendering reaches for “cotton was half of US GDP.” The strong version does not need that number and is better without it. What it needs is the institutional channel made concrete: marine insurance markets that priced human cargo; mortgage credit secured against enslaved people as collateral; bills-of-exchange networks settling London against Kingston and Charleston; double-entry plantation bookkeeping with its depreciation conventions for amortizing a human being. The claim is not that cotton produced fifty percent of British GDP. It is that the institutions through which capital was deployed were shaped by the plantation economy — and that shaping is invisible to a Solow-style decomposition that only counts the capital, not the architecture the capital moved through.
Beckert’s “war capitalism” framing makes the sequence concrete. Capitalism arrived in two stages. First came violent expropriation by private actors backed by state navies — African captives, indigenous lands, plantation extraction. Then came industrial capitalism, built on top. The first stage is not a moral footnote to the second; it is the institutional substrate the second grew out of. This is the system that Economic History Ch.9 (Atlantic Slavery and After) narrates as the productivity machine the Atlantic built. And the intellectual lineage — the New History of Capitalism as a deliberate departure from the cliometric program, conducted in the empirical-turn register the discipline adopted after 2008 — lives in History of Economic Thought Ch.17 (Modern Pluralism).
Where this leaves us
Put the institutional argument and the quantitative argument together and the field is in a different place than it was in 1974. The Engerman benchmark is no longer the modal estimate; Berg and Hudson’s eleven percent is a serious challenger from inside the discipline, not a polemic from outside it. Beckert’s structural argument identifies a real measurement gap, and the gap runs in the direction of underweighting slavery’s role. The honest revision: magnitude contested but defensibly larger than the late-twentieth-century consensus held; institutional channel real and underweighted by the standard tools; the load-bearing claim more defensible than the cliometric tradition allowed.
That is the news from inside the discipline, and it favors the revival. The news from outside is less flattering for the side this stage just argued at strength: some of the most viral popular claims do not survive arithmetic scrutiny — and the people pointing this out have not, until recently, been the people the revival wanted to hand the microphone to. Stage 4 hands it to them anyway, because the credibility of the larger case depends on it.
The compounding moral debt
“Until we reckon with our compounding moral debts, America will never be whole.”
— Ta-Nehisi Coates, “The Case for Reparations,” The Atlantic, June 2014
Coates’s essay is the bridge from historiography to politics. Its empirical center of gravity is not 1748 but 1948: the practical exclusion of Black veterans from the GI Bill, redlining maps that closed off mortgage credit, the contract-buying schemes that stripped Black families in postwar Chicago of equity. The essay is built more on twentieth-century housing policy than on eighteenth-century GDP shares — and that turns out to matter for what the magnitude debate can and cannot decide.
Separate the two questions one more time, because Stage 4 is where conflating them does the most damage. The magnitude question: how arithmetically wrong is “cotton was half of American GDP”? The policy question: granting whatever the magnitude turns out to be, what follows about present-day obligations? These have different answers and different evidence, and the move that wrecks the discourse is to let a verdict on one settle the other.
On magnitude, the critics are right and the walkthrough has to say so. Edward Baptist’s The Half Has Never Been Told arrived at its enormous figure by counting the value of enslaved people as part of the cotton sector’s output — but enslaved people were the producing capital, not the final product. Counting them alongside land, intermediate inputs, and the final crop double-counts the same value and inflates the figure several-fold. The actual share of antebellum US GDP attributable to cotton is around five percent, not fifty. Phillip Magness — ideologically positioned, technically correct on the arithmetic — made the case sharply; Paul Olmstead and Alan Rhode reached the same corrected number independently in a 2018 Journal of Economic History critique. The crucial point for engagement: the neo-Williamsite case at strength — Beckert’s institutional argument, Berg and Hudson’s eleven percent — does not depend on Baptist’s figure. Arguing the strong case at strength does not require defending the arithmetic errors of the weak one.
Argue the reparations case at strength on its own terms, which are not Solow-decomposition terms at all. Coates is making a historical-political argument about a continuous extractive system running from 1619 through Jim Crow, redlining, and the present-day racial wealth gap. The strongest version concedes the point Stage 2 established — that British and American slavery’s share of GDP in 1800 was real but bounded — and then notes that the case does not rest on those numbers. It rests on a larger twentieth-century body of evidence: the discriminatory federal housing policies of the New Deal era, the practical exclusion of Black veterans from the GI Bill, the contract-selling racket in postwar Chicago. The empirical link to the present is dated and traceable: the racial wealth gap in 2024 has identifiable origins, many of them well within the last hundred years and attributable to specific named policies.
The 1833 British abolition compensation loan is the iconic image of the whole question. Parliament voted £20 million — roughly five percent of British GDP, around forty percent of the year’s budget — to compensate slaveholders for the loss of their property, and nothing to the people who had been the property. HM Treasury finished paying off the debt in 2015. The University College London Legacies of British Slavery database now maps every recipient by name. Whatever the GDP magnitudes were in 1800, that ledger entry — settled inside living memory, with no comparable transfer ever made to descendants — sits in a different register from the cliometric debate, and it is the register Coates is writing in.
The honest version of the policy question includes the skeptics at their strongest, and there are three distinct objections worth stating in full rather than waving at. Thomas Sowell presses the political-mechanism problem: a reparations transfer moves resources from a present-day population that is not the historical perpetrator to one that is not the historical victim, which runs straight into the adjudication problems — standing, inheritance of liability, statute of limitations — that civil law was built precisely to avoid. Magness presses the magnitude problem from the policy side: if the public case for reparations is built on the inflated cotton claims, then the policy edifice is overbuilt on a foundation the arithmetic will not bear, and tying the moral argument to a falsifiable number invites the whole thing to be dismissed when the number falls. Glenn Loury presses the alternative-channels problem: framing African American economic outcomes as primarily downstream of slavery underweighs family structure, cultural capital, and the comparison-group evidence from Black immigrant populations whose trajectories diverge from the descendants-of-slavery population. None of these three disposes of Coates’s case. All of them are part of the honest argument the policy question actually contains.
“Until we reckon with our compounding moral debts, America will never be whole.”
— Ta-Nehisi Coates, “The Case for Reparations,” The Atlantic, 2014
The reparations case doesn’t live or die on cotton’s GDP share
Treating the magnitude question and the reparations question as a single argument is one of the worse moves in this discourse. They are separable. Coates’s case is built primarily on twentieth-century policy, not on 1800 GDP shares — so the cotton-was-fifty-percent claim is wrong and the reparations argument survives the correction.
The verdict, on three axes
Historiography. The cliometric consensus that held from 1972 into the early 2000s — slavery as a small share of GDP, therefore not load-bearing for industrialization — has not survived the last fifteen years intact. The institutional channel is real, and the standard sources-of-growth apparatus measures it badly because GDP-share is the wrong outcome variable for an institutional-channel claim. Berg and Hudson’s eleven percent is a serious recalibration of the basis parameter, not a settled new consensus. The strong Williams thesis — slavery as the engine of British industrialization — remains overstated. But the soft consensus that slavery was a moral horror essentially separable from the macroeconomic story is no longer defensible. The textbook treatment is due for revision in the direction the neo-Williamsite revival has been pushing.
Magnitude. On the viral popular claims — cotton as half of American GDP, slavery as the constitutive arithmetic engine of American capitalism — the numbers do not work, and the walkthrough says so even while crediting the broader argument at strength. Baptist’s double-count inflates the figure several-fold; Olmstead-Rhode and Magness reach the corrected roughly-five-percent figure independently. The credibility of the larger case depends on not defending the smaller errors.
Reparations. The historical case is stronger than the standard textbook grants — it rests on the twentieth-century policy record and the 1833 ledger, not on 1800 GDP shares. But the step from “slavery and its successor regimes were materially extractive at scales the standard tools underweight” to “reparations are the appropriate remedy” depends on additional moral premises that the historiography does not adjudicate. The two parts have to be argued separately. Conflating them, in either direction, is the failure mode on both sides.
The pattern in this dispute — a contingent, peripheral system playing a larger role in core outcomes than the standard apparatus measures cleanly — reappears one question over, at the Great Divergence. Kenneth Pomeranz, Robert Allen, and Joel Mokyr argue over whether contingent access to a coerced-labor periphery and to New World resources was the difference between Britain and the Yangzi Delta. That terrain sits in Economic History Ch.6 (The Great Divergence). It is the same dispute as this one, asked from the other direction: not “did slavery make the West rich?” but “could the West have gotten rich without the periphery at all?”