Summary : New cultivars of cotton led to an unprecedented rise in the productivity of US southern cotton in the 60 years before the American Civil War. The Economist magazine may have said some stupid things about the subject in its review of Edward Baptist’s book on slavery, but it was fundamentally right to question the claim that antebellum efficiency gains were due to slaves being worked ever harder by their masters. (Note : I have a follow-up post, “Baptism by Blood Cotton“, which critiques in detail the central claims of Baptist’s book. There is also a brief note on slaves-as-wealth in Piketty & Slave-Wealth.)
I’ve said for a long time now that The Economist “newspaper” mostly produces glib, shallow hack pieces, despite somehow maintaining an august reputation to the contrary. Recently it published a review of The Half Has Never Been Told : Slavery and the Making of American Capitalism by historian Edward E. Baptist. In one sentence the reviewer used some vaguely injudicious language, and this caused outrage in the comments section and in social media. The Economist quickly “retracted” the review but quoted it in its entirety all the same. I reproduce the section which contains the economic-historical argument :
I’ve italicised the part which incited the outrage, but the final sentence is apparently the real howler in the eyes of the offended.
I’m more interested in the economics of the above claims. I haven’t read this just-published book, so I can only go by the review, and as far as I can tell neither Baptist nor the offended have disputed the reviewer’s characterisation of the book’s thesis.
The Economist‘s elf was lazily speculating, a priori, about what could have been the determinants of efficiency in southern cotton production. It’s possible the reviewer is familiar with some of the arguments and debates surrounding Time on the Cross : The Economics of American Slavery, one of whose many controversial arguments was that slaves were becoming ever more valuable property in the antebellum American South and were therefore better treated than commonly supposed. [Edit : Fogel restrictively argued, “The material (not psychological) conditions of slaves compared favourably with those free industrial workers”.] But I doubt that literature is known to the reviewer, because then he would have been familiar with recent research on the sources of increased efficiency of cotton agriculture in 1800-60.
According to the paper by Alan Olmstead and Paul Rhode, “Biological Innovation and Productivity Growth in the Antebellum Cotton Economy” [published version], the labour productivity of southern cotton grew 2.3% per year in the period 1800-60 :
The conclusions about the growth of labour productivity are based on the following data :
In short, cotton output grew significantly faster than labour inputs, i.e., slave labour productivity was rising quite steadily. The bottom of the table displays the market prices of cotton and slaves, which serve as a check on the productivity observation. Cotton prices were falling but a major input in the production process, slaves, was getting more expensive. This fact is strongly consistent with rising labour productivity.
Personally I can’t believe 2.3% per annum growth in productivity over 60 years could possibly be achieved by something other than major technological change. 2.3% per year is huge. To put that in perspective, consider that output per worker in English arable agriculture rose ~1% per year in the period 1800-1850. [See Table 15, Broadberry et al.] Or consider Olmstead & Rhode’s claim (page 35) that the mechanical reaper increased the productivity of wheat harvest by 50-100%. That is puny compared with the quantum leap in cotton picking productivity : 2.3% over 60 years amounts to almost 400%. How the hell does one achieve a quadrupling of output per worker just by increased effort alone, no matter how coercively obtained ???
Olmstead & Rhode attribute the productivity gain to the introduction of particular cultivars of the Mesoamerican cotton species G. hirsutum, more colloquially known as “upland cotton”. Its characteristics are well summarised by Stephen Yafa in Cotton: The Biography of a Revolutionary Fiber :
Yafa further notes that “green-seeded upland would become the one that blanketed the American South” ; and today “it accounts for about 95 percent of all the cotton grown and used around the world”.
Olmstead and Rhode devote several pages to the trial-and-error process by which various cultivars of G. hirsutum were carefully selected, hybridised (especially with a Mexican cultivar), and adapted to the soil and climactic conditions of the American south. (A more detailed literary account, full of observations by contemporaries on the importance of new cotton varieties, is found here.) They also document the failure of another promising cotton species G. barbadense, also known as “Sea Island cotton”, to take hold on account of its low yield.
The new Southern cultivars yielded more lint per boll and there were more bolls per plant. Not only that, some of the varieties also produced taller plants which were easier to pick. These photos show the dramatic difference that height could make (click to enlarge, source) :
For me the vastness of the efficiency gain alone speaks eloquently to some other factor besides ekeing more labour out of slaves until the pips squeaked. O & R nonetheless go through the motions and address the obvious alternate hypotheses for the increase in the cotton pick rate : (1) efficiencies of scale from the slave gang system ; (2) better management (including the possibility that slave owners are getting better at squeezing work out of slaves) ; and (3) the expansion of cotton production into western territories of the United States.
For the purposes of this blogpost the most important is #2 and O & R exploit a natural experiment in the daily picking rates of the high-yield Upland cotton cultivars versus the lower-yielding Sea Island cultivars :
Note that the Y-axis has got a log-scale ; the differences in the growth of pounds picked per worker-day of the two cultivars are colossal. In view of the above, it’s difficult to believe the euphemistic “managerial efficiency” was driving the secular rise in the pick rate of Upland cotton.
O & R also checked the improvements in yield on the same plantations over time before and after they switched from Sea Island to Upland. Unfortunately they produce no time series graphics for these observations, but they report large increases over time in the samples of plantations covered.
Finally, the botanical-genetic-technological explanation of the efficiency gains in US cotton production better accounts for the dramatic conquest of the world market for raw cotton by growers of the American south toward the mid-19th century. American cotton had 80% of the world share. Why didn’t this new American technology get diffused earlier and faster ? The Upland cultivar was adapted to the geoclimactic conditions of the American south and could not be rapidly adapted by the other primary cotton-growing regions of the world, such as India, Egypt or the Caribbean. Britain, the premier producer and exporter of cotton textiles, directly controlled many of these regions, but its manufacturers chose to import primarily American cotton.
So, YES, The Economist was quite fatuous and grossly ignorant in tendering the flaccid hypothesis that slaves might have been better treated, as a first-order counterargument to the claim that slaves were driven ever harder. But it was perfectly reasonable in doubting the latter claim in the first place.
At his blog the political scientist Chris Blattman rightly chides The Economist for not bothering to know the literature on the subject. (His post has been reblogged at the The Washington Post.) But then instead of getting to the heart of the matter — the efficiency of antebellum cotton agriculture — Blattman digresses about the literature on the efficacy and effects of coerced labour, with one link and graphic after another proving…. definitely not that an intensified effort at coercion of labour can quadruple worker output over six decades in a very labour-intensive production system.
On Twitter Blattman uttered to me the truism that “technology and (coerced) labour are complements”, as though no one has ever before decomposed the sources of economic growth into labour inputs, capital inputs, technological change, etc. The complementarity is beside the point. The point is that driving slaves harder, alone, could not possibly have accounted for the massive productivity gain in cotton.
PS — There were also some technological improvements in cotton agriculture unrelated to plant breeds that contributed to the efficiency of cotton production. Two most important are “scrapers” and machine seeders. But I do not stress those here. Also, the better soils of the more westerly lands in the “New South” were important.
EDIT : I’m waiting for Edward Baptist’s book to come out on Kindle, but Google Books has some excerpts. Unless there’s more to it than pp 127-8, it would appear the author uses the Olmstead productivity data, but merely hints at the botanical argument and dismisses that with a wave of the hand. He makes it sound (at least in those pages) as though it were only a matter of a single variety, Petit Gulf, of Upland cotton. But the plant breeding issue is definitely much deeper and broader. From another paper by Olmstead & Rhode (O&R 2007):
O&R 2007 continues :
O&R 2007 goes into some detail on the plant-breeding aspect of the cotton cultivars. There’s definitely much more to it than just Petit Gulf.