The Baptist Question Redux: Emancipation & Cotton Productivity

Edward Baptist, the author of The Half Has Never Been Told, has been claiming since the publication of his book that a putative post-Emancipation drop in overall agricultural productivity in the American South is proof that it was torture, not new cotton cultivars and frontier soils, which had been largely responsible for the US cotton boom of 1800-60. But there are severe limitations to what the cliometric literature on slavery can reveal about post-Emancipation productivity specifically in cotton-picking.

The nature of the “Baptist Question”

In The Half Has Never Been Told the Cornell historian Edward Baptist attributed the very large increase in cotton-picking rates by slaves in the American south to the scientific application of torture in the gang labour system. Baptist ignored more than dismissed the statistical case made earlier by Olmstead & Rhode that the 400% increase (from 25 lbs. to 100 lbs.) in the raw cotton picked per day per slave between 1800 and 1860 was made possible by the introduction of new seeds with higher yields producing easier-to-pick plants, in the fertile frontier soils of the New South.

[For details, see my post “Plant breeding…drove cotton productivity gains in the US South”.]

In Baptism by Blood Cotton, I acknowledged that the extra cotton that was grown had to be picked by someone, even if the higher yield was due to better plants and better soils. Nobody disputes that technology and (coerced) labour were complements. Absolutely no one disputes that slaves were forced to work harder than they would have done had they been free. Nobody disputes the violent coercion part of this story (even if Baptist appears to think his critics do deny it).

But according to O & R, the new cotton varieties produced lint of longer staple length, with some varieties producing taller, and therefore easier-to-pick, plants with more bolls containing more lint. In that case, there was simply more cotton to pick for any given level of work effort induced by the lash. Recently Brad Hansen put the choices quite well :

“There are essentially two ways that this increase [in cotton picked] over time could have occurred. First, slaves could have been forced to pick closer to the maximum that they were physically capable of. Second, the maximum that they were physically capable of picking increased over time. O & R argue for the second explanation. Improved plants enabled slaves to pick more cotton in a given amount of time. In other words, slaveholders used physical coercion to force slaves to pick at maximum picking rates and through plant breeding they were able to increase this maximum amount that a person was physically capable of picking overtime.”

So it’s a question of which factor predominates in that 400% increase in daily cotton-picking rates: [a] slaves were made to pick closer to their physical maximum; or [b] the biological maximum was raised by new seeds and soils.

Baptist contends that in the traditional ‘task’ system, slaves were less supervised and therefore their labour was less than fully utilised. The gang system moved the utilisation of slave labour closer to full capacity. But it’s difficult to believe slaves had been so under-utilised in the Old South relative to maximum physical potential that you could raise utilisation rates by 400% ! (Baptist now says he does not discount the seeds explanation as partial, but in the book he barely mentions it and dismisses it in the footnotes.)

In the aftermath of his book’s publication, Baptist faced some harsh criticisms, especially about his garbling of economics. See for example Burnard (with response from Baptist and counter-response); as well as the roundtable reviews in the Journal of Economic History [ungated version], with Olmstead particularly severe in his critique; and recently Clegg which, for some of its arguments, cites my own critique. Also see the multiple blogposts by the economic historian Brad Hansen [1, 2, 3, 4].

Partly in response to such criticisms, Baptist has been loudly arguing that the decline in southern agricultural productivity after the war ‘destroys’ Olmstead & Rhode’s botanical argument. See his Twitter pronunciamentos:



And also some of his comments at the Junto blog:

“One problem here is the critique of the cotton productivity argument. Post bellum sharecropping was actually notoriously unproductive. See Fogel, who calculates a 40% drop in productivity”.

[Baptist has also now written a direct criticism of Olmstead & Rhode.]

Unfortunately, the Baptist Question can not be satisfactorily addressed by citations from the cliometric slavery debates of the 1970s and 1980s, especially regarding the effects of Emancipation. Their focus was too macro and rarely cotton-specific; and there are just too many uncertainties in their estimates of aggregate productivity to settle the decidedly micro question of cotton picking rates. As I argued previously, you need evidence from direct observation of how much cotton was picked or pickable per day or per hour for freedmen.

(Ransom & Sutch dispute there was a post-emancipation drop in productivity. See my note in the comments section.)

Which productivity are you talking about ?

There is a lot of loose talk of ‘productivity’ by historians who really need to understand the following (very elementary) distinctions in its meaning:

  • total output
  • output per worker (per year)
  • output per worker per day
  • output per worker per hour
  • total factor productivity (TFP), or output per input of all factors of production (land, labour, capital)

Today, when most people talk about productivity, they usually mean labour productivity in the sense of output per worker-hour. But the slavery debate initiated by Time on the Cross was couched in terms of TFP — the cliometricians of the era were arguing about the North-South differences in total factor productivity for the year 1860. By contrast, the productivity estimated by Olmstead & Rhode is output (cotton picked) per worker per day on southern farms between 1800 and 1860.

Care needs to be taken in comparing all these things. You can have land productivity rising while labour productivity is falling, with no net gain in TFP. You can also have the reverse.

You can have hourly labour productivity rising or staying constant, even when TFP is stagnant or falling. You can have daily labour productivity growing without any growth in hourly productivity, which implies that workers simply worked more hours per day. Conversely, if daily labour productivity shoots up while the work day gets shorter or stays the same, that could mean more work was being squeezed out of workers for every hour they worked.

The distinction between daily and hourly productivity can make a big difference. Everyone is so 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) !!!

Then there is the distinction between the productivity of the agricultural sector as a whole and the crop-specific productivity for cotton. Most of the cliometric slavery debate focused on all-agriculture productivity. Of course cotton was an important part of the southern economy, but it was still less than 30% of its total agricultural output (TOC2, pp 131-138, WCC-TP1 pg 262n). So you cannot simply deduce changes in crop-specific productivity from changes in sectoral productivity — especially on farms where many crops were grown at any given time — at least not without having some detailed data on the allocation of labour time to specific crops within individual farms. So even in the case of large plantations where 60% of the monetary value of output was derived from cotton (WCC-TP1 pg 254; small farms cotton share: 29%), you still can not infer how much labour-time was devoted to producing that crop and therefore you can not infer the cotton-specific labour productivity from the all-crop information.

( However, there is an alternate method, the price dual, for estimating productivity changes for a single crop like cotton. I leave this as a note in the comments section.).

Further micro-evidence on how long it took to pick cotton

In my earlier post I presented some evidence from the 1930s about how much labour-time was required to hand-pick a pound of raw cotton just before picking became mechanised. That evidence is consistent with Whatley 1987 which found that the “average picking rate of 125 pounds of seed cotton per 10-hour man-day is the quote found most frequently” in US Department of Agriculture bulletins of the 1930s. Neither seems terribly far removed from the average of 100 lbs. in 1860 found by Olmstead & Rhode.

Both are also consistent with other evidence. From Craig and Weiss, “Hours at Work and Total Factor Productivity Growth in the Nineteenth-Century U.S. Agricuture” (chapter 1 of New Frontiers in Agricultural History; my uploaded PDF copy):


According to the above, the amount of labour-time required to produce 100 lbs of cotton lint steadily declined. Of course, it’s not the same as picking rates of raw seed cotton.

However, recently the economic historian Trevon Logan (Logan 2015) published manual cotton-picking rates from 1952-65 on the Mississippi farm belonging to his own sharecropping grandfather:

“Although the agricultural productivity of African Americans in the antebellum era is well-investigated, after Reconstruction we know much less about how productive African Americans were in the fields.

“The cotton picking books retained by my deceased paternal grandparents are quite detailed and allow me to estimate the individual productivity of nine children who were involved in intensive manual cotton picking from 1952 to 1965. This rich quantitative data forms the basis for this study. In addition to estimating the overall productivity (pounds of cotton picked per person per day), I also use the records to investigate gender differences in agricultural productivity.

“On average, children in the Logan family picked approximately 120 pounds of cotton per person per day in their late teen years. These estimates are quite similar to recent estimates of slave productivity produced by Olmstead and Rhode (2010), who calculate productivity using detailed plantation records from the late antebellum era from the same region. On average, the children in the Logan family were more than 95% as productive as their enslaved predecessors in the field. Gender differences in productivity were quite small within the Logan family, disappearing almost entirely by late pubescence. Given that the method of picking cotton was largely unchanged over time — both the enslaved and Logan children picked cotton by hand in a process that underwent very little technological innovation — the estimates imply that the extraordinary productivity of slaves persisted long after Emancipation.”

Conveniently, Logan 2015 reproduces some of the methods of Olmstead & Rhode, and the picking rates are expressed as percentage of slave averages recorded by O&R:




It must be kept in mind, the above picking rates should not be considered completely comparable with those of antebellum slave plantations. First, the cotton plants in the 1950s may have had an intrinsically higher biological yield than those of 1860. Second, the incentives might have been very different — global cotton prices were lower than a century earlier. And as Logan himself points out, the cotton seeds were a separate source of revenue for the Logan family. Finally — of course — this is just one family.

Nonetheless you can not blithely assume that cotton-picking rates collapsed with the abolition of slavery.

Books cited

For more on productivity issues from the cliometric literature, see the notes in the comments section (very tedious).

About pseudoerasmus
This entry was posted in cotton, Edward Baptist, historians of capitalism, Slavery, The Half has never been told and tagged , , , , , , , , , . Bookmark the permalink.

8 Responses to The Baptist Question Redux: Emancipation & Cotton Productivity

  1. Some notes on the estimates of productivity in the cliometric debates

    What was measured for 1860 ?

    According to Robert Fogel, southern agriculture as a whole had a TFP advantage of about 35% over northern agriculture [WCC, pg 76]. Back in 1971 and 1974, he and Stanley Engerman had computed agricultural TFP as a geometric mean (a special kind of weighted average) of the productivities of labour, capital, and land. Each of these productivities was estimated separately and directly from the Parker-Gallman dataset of over 5000 farms in over 400 counties of the South.

    The crucial point here is that the labour productivity component of TFP was output per worker-year — the labour input was expressed in terms of the number of “effective hands”, because data on man-hours or man-days were not available. Don’t take my word for it. Fogel himself said his TFP “indexes measured the labor input not in man-hours but in man-years”, so the “more intensive utilization of labor shows up not as greater labor input but as higher level of productivity” (Ch. 12, pg. 244, WCC-TP1).

    In the face of criticisms (e.g., David & Temin 1974) that the output-per-worker advantage of the South could be spurious because slaves might have worked longer in any given year and on any given day than free farmers, Fogel (1977) turned to the estimates of the length of the work year produced by John F. Olson, which would be published many years later as chapter 11 of WCC-TP1.

    Olson estimated that, overall, slaves worked approximately 10% fewer hours per year in 1860 than northern farmers in the same year. The slave average of 2800 hours/year was derived from the following:

    • the average days worked per year gleaned from only seven antebellum southern plantations is 281 — which is only about 75% of the potential (see Table 11.3, pg 226, WCC-TP1);
    • average of 9.9 hours per day, gleaned from data on cotton farm operators in Eastern and Delta cotton regions in the 1930s (see Table 11.2, pg. 224)

    Based on this information, Fogel concluded that the southern advantage was not in slaves being made to work longer per day and per year. And by ruling out other factors (such as locational rents) Fogel felt confident to conclude “greater intensity of labor per hour, rather than more hours of labor per day or more days of labor per year, is the reason the index of total factor productivity is 39 percent higher for gang-system plantations than for free farms” (WCC pg 78).

    I went into this detail because this 2800 figure is an important assumption in the calculations of post-Emancipation southern productivity.

    What was measured for 1880 ?

    Here are some statements taken from Robert Fogel, Without Consent or Contract: The Rise and Fall of American Slavery which have been loosely cited by Baptist:

    “…the labor productivity of southern farms declined sharply between 1860 and 1880. Indexes based on three different bodies of data yield the same basic result… the range of the estimated decline is reasonably narrow-from 31 to 43 percent.44 …. Present estimates indicate that about two-thirds of the reduction in labor productivity was due to a decrease in the efficiency of labor, which is measured by the index of total factor productivity. The balance was due to a decline in the amount of improved land, work animals, and other capital that was available to each agricultural laborer”. [pg 99]

    “The preliminary comparisons between productivity in 1860 and 1880 indicate that small white farms were about 5 percent less productive in 1880 than they had been in 1 860 and that almost the entire decline is concentrated in the category of black farmers, who worked mainly on slave plantations before the war and on small farms after it. The labor productivity of black farmers was 60 percent lower in 1880 than it had been in 1860.48”. [pg 100]

    None of the statements is about productivity in cotton, but about southern agriculture in general. Fogel’s two sources are:

    • Jon R. Moen, “Changes in the Productivity of Southern Agriculture between 1860 and 1880”, ch. 15, WCC-TP1.
    • Donghyu Yang, “Explanations for the Decline in Southern Per Capita Income, 1860-1880”, ch. 38, WCC-EM.

    Moen’s calculations of southern agricultural labour productivity (output per worker-year) for 1860 and 1880 are presented in Table 15.6 from page 330:

    moen t15_6

    The above assume that the labour force participation rate did not change after Emancipation — the number of working days and hours per year stayed the same, and the percentage of men, women, and children working stayed the same. According to Moen, if you accounted for the claim by Ransom & Sutch that black labour force participation fell between 28% and 37% after Emancipation, then the worst-case drop in output per worker (Q/L5) falls from to 43% to 33-39%. It’s not clear and I am not sure, but that computation appears to assume 2800 hours per worker-year.

    The various subscripts of L refer to the number of “effective hands”, i.e., workers adjusted for differences in age, sex, etc., and this is where a lot of uncertainty creeps in. L2 refers to the raw number of actual workers dwelling on farms listed in the 1880 Census; L3: adds a correction for a supposed undercount of family labour on white-operated farms, estimated statistically; L4: correction for the number of non-resident hired labourers, inferred from the Census; L5: correction for the fact that L4 implies too low a percentage of blacks in the southern labour force in 1880. For me L3 and L5 seem the most debateable.

    It’s not clear what should really be compared in order to assess the fall in agricultural output per worker due to the end of slavery. The largest slave farms in 1860 versus the black-operated farms in 1880 ? That might tell us something about the disappearance of scale economies from gang labour, except that Moen’s chapter also tells us blacks held both less land and less valuable land in terms of both improved and unimproved acres than whites (which would partially explain the lower black labour productivity). Compare all farms in 1860 with all farms in 1880 ? That may actually be more appropriate, since that would speak to the South’s ability in general to produce without slaves.

    I am not saying the estimates cited are not useful or accurate enough for the very bird’s-eye-view purposes for which they were constructed. But given the uncertainties and debateable assumptions they are not appropriate for inferring anything about cotton-picking rates after Emancipation.

    (Other comments worth making??? Factor shares in geometric index are assumed to be constant between 1860 and 1880. Prices estimated from production function for the south.)


  2. A note on the “price dual” labour productivity for cotton

    The one observation made by Robert Fogel about post-bellum productivity specifically in cotton production (not picking) is:

    “Since the breakup of the gang-system farms was responsible for the decline in productivity, it might be thought that the price of cotton should have been higher than it actually was in 1880. Had the price of inputs remained constant, a 20 or 30 percent decline in total factor productivity would have led to a 20 or 30 percent rise in the price of cotton. However, the prices of the main inputs into cotton production — labor, land, and most other items of capital — fell by amounts that nearly offset all of the upward pressure on cotton prices caused by the change in productivity. Indeed, the ratio of input prices to the price of cotton is still another way of measuring total factor productivity, and this index indicates a 35 percent decline in the efficiency of cotton production.” [WCC, pg 101]

    “According to the calculations of Moen, TP, #15, total factor productivity [in overall southern agriculture] fell by 13 percent between 1860 and 1880 if measured by quantities, and by 35 percent if measured by the price dual. Cf. EM [Yang], #38; Jaynes (1986)” [pg. 439, footnote 49]

    The above refer to the quick-and-dirty dual method. The “primal method” which was used to compute total factor productivity by Fogel & Engerman is expressed in terms of physical quantities. (At least theoretically: you still have to use prices to make physical quantities addable in terms of common units.) But the price dual allows you to infer changes in productivity as long as you have output and input prices over time. This is more or less the same method used in Eltis et al. which found zero labour productivity growth in Caribbean sugar over the long run, despite a very harsh gang labour system.

    The simplest computation for the price dual labour productivity change in agriculture between 1860 and 1880 is found in Yang (WCC-EM pg 275) that was referenced by Fogel :


    The monthly wage rate for 1860 used above and in Moen (WCC-TP1, pg. 340) is for free male labour in the South. I don’t understand why that should be the basis for representing the cost of slave labour input ! (Eltis et al. used slave prices for the entire period.) It’s possible to argue it’s merely an accounting difference: the slave owners expropriated the difference between the market wage and what slaves consumed, and the difference just showed up in capital’s share of income. But one of the arguments made by Fogel was there was no wage at which free workers (whether antebellum whites or postbellum freedmen) would ever work in gangs as the slaves had been forced to do on plantations. So what could the 1860 market wage even mean for the purposes of the price dual ?

    You can substitute the cost of slave consumption as an imputed wage. Ransom & Sutch (pg. 212) estimates the $ value of slave consumption for 1859 at $29-32 per year. So say $3 per month. Assume cotton prices were unchanged between 1860 and 1880. Then simply replicating Yang with monthly slave consumption in 1860 and the market wage rate in 1880 would imply a large increase in labour productivity in cotton ( 9.26 ÷ 3 ). This figure is not to be taken seriously. It’s just intended to show, if the estimate were redone with slave ‘wages’ for 1860, the price dual almost certainly would imply no drop in labour productivity for the cotton sector by 1880.

    Note on the postbellum deterioration of the cotton-growing environment

    In Creating Abundance: Biological Innovation and American Agricultural Development, Alan Olmstead and Paul Rhode report that the cotton-growing environment of the South in the years after the Civil War deteriorated from worm infestation:



  3. Pingback: Baptism by Blood Cotton | Pseudoerasmus

  4. Pingback: Piketty & Slave Wealth | Pseudoerasmus

  5. Pingback: Was slavery necessary to western industrialisation ? | Pseudoerasmus

  6. Pingback: Plant breeding, not working slaves harder, drove cotton productivity gains in the US South | Pseudoerasmus

  7. Pingback: McCloskey: Cotton wasn’t crucial to the British industrial revolution | Pseudoerasmus

  8. Ransom & Sutch on Post-Emancipation Productivity

    Edit 6-Nov-15:Roger Ransom and Robert Sutch in One Kind of Freedom: the Economic Consequences of Emancipation, which is often taken to be the “Time on the Cross” of Emancipation, dispute there was a decline in productivity at all:


    [One of Fogel’s students had a clever counterargument to the above, for which see the comments section.]

    Yang (ch. 38, WCC-EM) excluded Ransom & Sutch’s estimates for the decline in labour force participation. If you adjusted Richard Easterlin‘s estimate of a 31% decline in output per worker in southern agriculture with Ransom & Sutch’s estimates of 28-37% decline in LFP, then the decline in agricultural output/worker would be too small to explain the decline in Southern income per capita because agriculture was such a large share (72%) of the total, and manufacturing and services would have had to grow at implausibly high rates.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.