The Statistical Errors of the Reparations Agenda

By Phillip W. Magness. Phil Magness is a Senior Research Fellow at the American Institute for Economic Research.

"Journalist and political commentator Ta-Nehisi Coates drew attention to the political cause of slavery reparations during a heavily publicized congressional hearing this week. While commentators on both sides of the issue agree that his case was eloquently argued, one of its central claims rested on faulty economic data.

Specifically, Coates contends that the case for reparations comes from the economic measurement of the antebellum slave economy in the United States. He testified, “By 1836 more than $600 million, almost half of the economic activity in the United States, derived directly or indirectly from the cotton produced by the million-odd slaves.”

This stunning statistical claim was widely repeated in commentary on the hearing. It is, however, unambiguously false.

Coates’s numbers come from Cornell University historian Ed Baptist’s 2014 book The Half Has Never Been Told. In a key passage in the book, Baptist purports to add up the total value of economic activity that derived from cotton production, which at $77 million made up about 5 percent of the estimated gross domestic product (GDP) of the United States in 1836. Baptist then committed a fundamental accounting error. He proceeded to double and even triple count intermediate transactions involved in cotton production — things like land purchases for plantations, tools used for cotton production, transportation, insurance, and credit instruments used in each. Eventually that $77 million became $600 million in Baptist’s accounting, or almost half of the entire antebellum economy of the United States.

There’s a crucial problem with Baptist’s approach. The calculation of GDP, the main formulation of national accounts and a representation of the dollar amount of economic activity in a country in a given year, only incorporates the value of final goods and services produced. The rationale for doing so comes from accounting, as the price of the final good already incorporates intermediate transactions that go into its production and distribution. Baptist’s numbers are not only wrong — they reflect a basic unfamiliarity with the meaning and definition of GDP.

When The Half Has Never Been Told first appeared in print, economists immediately picked up on the error. Bradley Hansen of Mary Washington University kicked off the scrutiny by posting a thorough dissection of Baptist’s errors on his personal blog. Economic historians Alan Olmstead (UC-Davis) and Paul Rhode (University of Michigan) chimed in with a devastating critique of Baptist’s empirics, observing that a continuation of his “faulty methodology by summing the ‘roles’ of cotton with a few other primary products” would yield an amount that “easily exceed[ed] 100 percent of GDP” in the antebellum United States — an economic impossibility.

Stanley Engerman, perhaps the foremost living expert on the economics of slavery, weighed in next:

Baptist's economic analysis, intended to demonstrate the essential role of the slave-grown cotton economy for Northern economic growth, is weakened by some variants of double and triple counting and some confusion of assets and income flows. To go from a value of the Southern cotton crop in 1836 of "about 5 percent of that entire gross domestic product," to "almost half of the economic activity of the United States in 1836" (pp 312-22) requires his calculation to resemble the great effects claimed by an NFL club when trying to convince city taxpayers that they should provide the money to build a new stadium because of all the stadium's presumed primary and secondary effects.

The main takeaways are that (1) the actual percentage of GDP derived from slavery is measured from final goods and services that involved slave-based production, and (2) Ed Baptist clearly did not understand what he was doing when he calculated his statistic. Cotton was by far the biggest item on the list of final goods and services, and, while its output varied year by year, it is probably reasonable to place slave-based goods in the mid to high single digits, not the 50 percent claim that Coates repeated.

Unfortunately, historians who work on the “New History of Capitalism” — a school of historiography that emerged after the financial crisis of 2007–8 and that purports to study the relationship between slavery and capitalism — have proven remarkably ill-suited at grasping the fundamentals of GDP and other economic concepts.

Not to be outdone by Baptist’s erroneous 50 percent estimate, Emory University historian Carol Anderson offered an even-higher figure from the eve of the Civil War itself. According to Anderson, “80 percent of the nation’s gross national product was tied to slavery” in 1860. Following Coates’s testimony using Baptist’s erroneous numbers for 1836, several historians began circulating this estimate from Anderson’s 2016 book White Rage as evidence of the growing influence of slavery on American capitalism in the late antebellum period.

Like Baptist’s book, it too derives from a fundamentally erroneous understanding of national accounts. Anderson’s footnote points to the late historian David Brion Davis’s foreword to a 2010 book on abolitionism. Davis makes a very different claim, however, in noting that the total value of slaves on the eve of the Civil War was equal to 80 percent of a single year’s gross national product (GNP). Anderson appears to have misread Davis’s data point and transformed it into a broader claim about slavery’s share of the entire economy.

While this figure is admittedly astounding and signifies the vast amount of wealth tied up in Southern slavery, Anderson mistakes it for the recurring yearly value of slave-produced economic output. She therefore commits the basic economic error of confusing stocks — by definition, a one-time measurement — and flows, which are measured over time. As we’ve already seen from Baptist’s example though, the actual percentage of GDP (or GNP) tied up in slavery was actually a small fraction of that amount.

Basic statistical errors of this type are a pervasive feature of the “New History of Capitalism” genre of scholarship — even to the point that they are now entering into the discourse over policy discussions, as Coates’s widely touted testimony at the reparations hearing illustrates.

In each case, the historians’ demonstrably wrong GDP and GNP numbers make for a shocking claim that appears to situate slave production at the very core of all American economic activity before the Civil War. This claim appears to confirm many of the ideological expectations of the same historians, who also evince a pronounced hostility to market capitalism throughout their work. Linking historical capitalism to slavery is more of a political exercise for the present day than a scholarly inquiry into the past, and in fact the most virulent defenders of slavery in the mid-19th century actually presented their cause as an expressly anti-capitalist venture (for those interested in further explorations, I’ve written about this deficiency of the “New History of Capitalism” literature at greater length here).

There is a great moral gravity to discussions of slavery, not only as a historical problem but also an institution with persistent and adverse legacies that remain with us. It is therefore a timely and ever-present subject of scholarly inquiry and discussion.

Regardless of where one stands on the reparations debate or other causes in the modern political scene, academics owe the public an honest, accurate, and scientific assessment of slavery’s history, including its economic dimensions. That assessment is harmed when the discussion forgoes scientific rigor or even basic statistical practices to rally around a mistaken number to support a misleading and grossly inaccurate conclusion about the nature of the antebellum economy. Baptist, Coates, and the other public figures who have repeated this faulty statistic have an obligation to correct their error."