The underlying claim in Edward Baptist’s “oral economic history” of slavery, The Half Has Never Been Told: Slavery and the Making of American Capitalism, is that slave owners, through the scientific “calibration” of torture, intensified the work of slaves in order to increase labour productivity by 400% on southern cotton plantations between 1800 and 1860. I argue the intensification claim is poorly supported, exaggerated, and misleading. This is a follow-up to my previous blogpost, “Plant breeding, not working slaves harder, drove cotton efficiency gains in the antebellum US South“. Warning : I’ve tried to make this post as readable as possible, but it goes into considerable detail, so it is dense and dreadfully dull !
In my previous blog entry, I argued that plant breeding drove cotton productivity gains in the antebellum US South, exploiting the work of economists Alan Olmstead and Paul Rhode, especially their paper “Biological Innovation and Productivity Growth in the Antebellum Cotton Economy” [the gated, published version is O&R 2008]. But having just read Baptist’s The Half Has Never Been Told, I realise it really doesn’t matter whether the gains were achieved by new cotton cultivars, the invention of the cotton gin, or better soils in the western frontier of the “New South”.
The total output of American cotton increased by a certain amount in 1800-60, and the supply of slave labour devoted to cotton production increased by less than that amount. The difference implies a 400% increase in output per slave. No matter the cause of it, the four-fold increase in cotton per slave still had to get picked by someone.
Apart from land and labour, there were two limiting factors on the amount of cotton that could be grown in the first place. No matter how much cotton you could plant, it was of no use unless the “seed cotton” (the raw fiber) could be removed from the bolls of the plant ; and after the harvest, the seeds removed to produce lint. The gin alleviated the constraint on the speed of seed removal. But there was no mechanical cotton-picker until the 1930s, nothing comparable to the reaper for wheat. So the solution to the abundance of new frontier land and the potential for rapid seed removal was, to make slaves separate more fiber from the bolls faster.
The question is, which better captures what happened : (a) there was more acreage under cultivation (without necessarily better yield per acre) thanks to southwestern expansion, so there was more cotton to pick, and slaves were made to pick it faster ; OR (b) there was more yield per acre even with acreage expansion, so there was more cotton to pick, but slaves found it easier to pick it even if they also had to pick faster.
In my previous blogpost, I answered (b). If new cotton cultivars allowed slaves to draw more fiber out of more bolls on taller plants, then a lot of the work intensification was thanks to the plant.
Baptist attributes the higher pick rate to a revolution in slave-owner relations that separated “first slavery” from the second phase in the newly opened lands on the western frontiers of the old Southern states. Previously a system of passive resistance had prevailed where
In some regions of the Old South,
Even on tobacco plantations of the Old South, Baptist continues, surveillance was very often delegated to slaves “outside of direct white observation”, because the plots were small and scattered. However, with western expansion the “right hand” had figured out a profoundly new way to crush the passive resistance of the “left hand” and wring the slack out of the system :
“Full capacity utilisation”, as we might call it today, would be achieved by a fully white-surveilled and -directed system of gang labour, every bit as regimented as industrial factories. In this “pushing system”, a slave…
The primary incentives were negative, i.e., violence and torture :
Under the scientific principles of labour management, whipping was now no longer haphazard, let alone whimsical, but tightly linked to a quota-and-record system. “Enslavers used cotton-picking records to measure and record each enslaved person’s output. Such ledgers served, along with the scale and the whip, as key parts of the ‘whipping-machine’ system that raised cotton output steadily over time”. Therefore, it was not just the cotton gin, but this system of industrial management which allowed the revolution in cotton productivity.
When all the literary dressing is removed, Baptist’s argument can be summarised as: the “time and motion” system of whipping, quotas, and measurement drove each slave to economise on spare movement and pick cotton faster, faster, and faster. And the speed-up effect of this proto-Taylorism is what was responsible for the quadrupling of the daily pick rate between 1800 and 1860, from the average of 25 lbs (>11kg) per slave to 100 lbs. (>45kg).
Baptist, whilst dispensing with Olmstead & Rhode’s botanical explanation for the productivity growth, nonetheless relies crucially on their output estimates. In fact those data are the quantitative backbone of his thesis. Yet there is also another, unacknowledged debt. Whether he knows it or not, Baptist has used the work of Olmstead & Rhode to extend backward in time the efficiency argument advanced in a series of works on American slavery by Robert Fogel, the late economic historian & Nobel laureate, and his colleague Stanley Engerman.
Fogel & Engerman (F&E) were the authors of a controversial 1974 book, Time on the Cross: The Economics of American Slavery, which made a multitude of iconoclastic, revisionist arguments which were immediately challenged by pretty much everyone. (See their own 10-point summary of the book.) Economists in particular honed in on their claim that
In the above, by “efficiency” F&E meant not simple labour productivity, but total factor productivity (TFP), or the output per unit of all inputs into the production system, e.g., land, labour, capital, everything.
Economists descended like locusts upon the efficiency claim at a conference devoted to Time on the Cross, producing a series of critical articles later collected in book form ; F&E responded to the criticism in the The American Economic Review in 1977 ; which prompted 4 articles in counterresponse in the March 1979 issue of the AER ; and to which F & E counter-counter-replied in 1980. In reaction to the numerous criticisms F&E corrected or refined the original claims and Fogel published a new book version of TOC called, Without Consent or Contract: The Rise and Fall of American Slavery, accompanied by 3 volumes of technical appendix. Finally, Fogel published a retrospective on that debate in 2003, which summarises the principal claims of WCC without much change. The history of this debate deserves a history of its own !
F&E’s claim of the productivity advantage of slave farms relative to free farms survived criticism and scrutiny. But Fogel had to shift his position on the causes of the productivity disparity. TOC had originally argued, the disparity was due to a combination of increasing returns to scale from plantation size AND both positive & negative incentives. ‘Positive’, meaning not just the lash but also rewards and ‘nice’ treatment as valuable property. Yet by the time of WCC, the “positive reinforcement” part was basically dropped. From WCC :
Here is the table of TFP comparisons worked out for 1860 (after correcting for some criticisms), reproduced from Fogel 1977 :
Also, Fogel toned down the crop-specific scale effects, and put more emphasis on scale economies in the supervision and management of a large number of coerced workers. This shows up in his reckoning of the source of the productivity advantage of large-scale slave agriculture :
But doesn’t the above support Baptist’s thesis ? Not exactly. Fogel found that in 1860 large slave plantations, on average, extracted 70-75% more work per hour out of their slaves than southern free farms, northern free farms, and small slave farms. I stress, this finding applied to 1860 only. What Baptist does is equivalent to turning this cross-sectional observation into a “time series” assertion.
We can reasonably surmise that slave cotton plantations of 1800 had rates of work extraction out of their slaves which were similar to the small slave farms of 1860. Therefore, the difference in “labour intensity” (actually labour utilisation), in 1860, between the large slave plantations and everybody else strongly suggests a ceiling on how much extra work effort per hour could have possibly been wrung out between 1800 and 1860 in the overall slave sector. Baptist may or may not realise this logical implication, but it is there.
Note : Fogel’s productivity estimates included all major southern crops, but cotton was the biggest as a share of the total and, since the efficiency advantage to gang slave farming was even bigger in sugar and tobacco according to Fogel, the 70-75% figure might be used, at a first approximation, for the “labour intensity” advantage in slave cotton. We could even just round it off to 100%. It still would not come anywhere to the 400% required by the implicit Baptist thesis.
70-100% can be proposed as the upper limit on the amount of “work intensity” that owners achieved with their slaves in the 60 years leading to 1860. But even that figure is highly uncertain. It is not known, for example, how many hours of work were devoted to cotton-picking by slaves, on average, before the 1850-60 period. As far as I know there is only one study on working hours on southern plantations, but it’s old, drawn from a very small sample, and does not specifically cover cotton. So we can’t match the evolution in hours worked per day per slave with the concomitant evolution in the pounds of cotton picked during harvest season. In other words, maybe there was intensification of work per hour, or maybe there was not and slaves simply worked more hours per day on cotton over time, especially if they worked on farms with a mix of crops.
The distinction between daily and hourly productivity can make a big difference. Everyone is impressed with the 400% increase in productivity, but suppose the slave work day picking cotton (as opposed to doing other things) rose from 8 hours/day to 12 hours/day at the same time. Then already the labour productivity growth is reduced to 270% (100/12 ÷ 25/8) !!!
Besides, does the change in the average cotton pick rate of 25 lbs/day/slave to 100 lbs/day/slave even represent such a major intensification of work effort, compared with free industrial factory labour ? First of all, the 1800 starting point of 25 lbs is low by any standard. Even the 1860 end point is not egregiously high. For example, The Economics of Mechanical Cotton Harvesting contains the following table of labour requirement :
Note the above records the average yield of lint, the product of ginning the “seed cotton” or the raw fiber in the boll which is what gets harvested. The seed cotton yield can be inferred from the lint yield with a ratio of 3:1 [Olmstead & Rhode 2010]. Thus for the southern regions in the years 1909-36, the average rate of hand-harvest is approximately 13 lbs per hour. In relation to slave faming, the daily pick rate for 10-12 hour days would be 130-156 lbs, which is consistent with Whatley 1987. The standard deviation is about 40 lbs. In other words, free labour in the early 20th century was hand-harvesting seed cotton at a rate about 1 standard deviation above the slave average for 1860.
Of course, the intrinsic biological yield in 1909-36 must have been higher, so the pick rate ‘adjusted’ to antebellum conditions should be lower. In The Half Has Never Been Told, this kind of adjustment is never made, in reverse, for pick rates. Nonetheless I agree with Baptist that, on average, free workers (or sharecroppers) were absolutely not willing to work as hard as slaves were made to work (a view supported by Ransom & Sutch). But my point here is not about the welfare costs of slavery, but only about how intense 100 lbs/day really was.
What about outliers with free labour ? “Cotton-picking contests” provide some clues. One such event in California supplies the following information for 1935 : “The average picked for the nine hours of work among the contestants who finished the contest was 647.17 pounds”. And the “average picked by the 72 contestants in 1937 was 645 pounds per person”, including a woman who picked 544 lbs. Even adjusting for higher intrinsic crop yield these rates are consistent with the information from the slave plantation records.
From another paper by Olmstead & Rhode, “Slave Productivity in Cotton Production by Gender, Age, Season, and Scale” :
These histograms display the distributions of daily pick rates for adult male & female slaves during the 1840-62 period. Notice the asymmetry; in statistics, these distributions are said to have “positive skew”, meaning that their right tails are fatter and/or longer than the left side, unlike the ideal symmetric curve of the normal distribution. According to O&R the mode (the most frequently occurring value) is 100. But it’s clear from just eye-balling the distributions that the high-performing outliers raised the mean.
Baptist touts the role of scientific management of slaves, which implies something else: the slave owners or their managers understood the strengths and weaknesses of their slaves, and operated according to the principle of comparative advantage. The “most able workers, those with an absolute advantage in picking, are assigned to other more difficult tasks where they possess an even greater advantage” [Olmstead & Rhode 2010]. This is remarked upon as well in Metzer 1975, Fogel 1977, and Toman 2006. I quote from Metzer’s “Rational Management, modern business practises, and economies of scale in the ante-bellum southern plantations” :
So, because the slave owners allocated labour according to comparative advantage, the best slaves (whether prime-age males, or the smartest of the population) were not all harvesting cotton. Those with lower picking ability were, in part, manning the harvest crew.
But there were also many slaves who picked well below the mean. Baptist makes considerable effort to describe slaves — by name, in most instances — who were made to pick 200, 300, even 400 lbs of cotton per day. From his narrative one might forget that 100 lbs/day was the actual average f0r slave cotton plantations in 1860. But at least he tells you the average. Here is the chart presented in The Half Has Never Been Told :
The above was adapted by Baptist from Olmstead & Rhode’s scatter plot :
The variation around the mean not only is pretty big but also it gets bigger with time. So when Baptist talks about all those farms with huge daily average outputs, be sure to remember there were many grossly unproductive ones.
Lastly, a short word on Caribbean sugar. It had achieved the apex of the plantation gang labour system in the 18th century. And the picture is less clouded by revolutionary technologies like the gin or vast frontier expansion or new cultivars. It was a crop production system particularly egregious for gang labour and cruel, high-mortality working conditions.
Eltis et al. 2005 generates estimates for labour productivity growth in 1674-1801 for Caribbean sugar by inferring from movements in the prices of sugar and slaves. The estimation method is not nearly as good as the one for southern slave cotton by Olmstead & Rhode. But for what it’s worth, under several different assumptions the long-run growth of slave labour productivity in Caribbean sugar was dramatically less than US cotton 1800-60 :
The above shows fluctuations of the level of productivity, not rates of growth. When productivity was rising, growth rates were on the order of 0.2-0.4% per year. But the movements are extremely volatile, with little or no long-term trend ultimately.
Postscript : In the long tradition started by Eric Williams, Baptist makes extravagant claims about the indispensability of the American cotton-slave machine to Western industrialisation. I guess that will be addressed in part 3 !