How Well Did Facts Travel to Support Protracted Debate on the History of the Great Divergence between Western Europe and Imperial China?
By: Kent Deng (London School of Economics), Patrick O’Brien (London School of Economics)
Abstract: This paper tackles the issue of how reliable the currently circulated ‘facts’ really are regarding the ‘Great Divergence’ debate. Our findings indicate strongly that ‘facts’ of premodern China are often of low quality and fragmented. Consequently, the application of these ‘facts’ can be misleading and harmful.
Distributed by NEP-HIS on: 2017-03-19
Review by: Kenneth Pomeranz (University of Chicago)
Kent Deng and Patrick O’Brien have done us all a service by taking a step back from the conclusions drawn by different participants in the so-called “great divergence debate” to focus on the types and quality of our data, and on some conceptual problems with the application of modern measurements such as GDP to economies that were radically different from ours – in part because they were incompletely monetized. I find myself agreeing with most of their criticisms of both GDP and real wage comparisons, and have some of my own to add. Not surprisingly, then, I share their preference for research on comparative consumption – which was a big part of my approach in The Great Divergence – and agree that this is where we have the best prospects for making further progress on these issues. But I am not as ready as they seem to be to completely discard approaches based on GDP or real wage estimates; and perhaps more importantly, I would significantly modify their assessment of where our discussion of consumption and popular living standards currently stands.
We are, quite simply, unlikely to find any data that is good enough to lay these disputes to rest. I agree that the chances of finding significantly better consumption data are higher than are our chances of finding everything we would need to construct truly persuasive GDP estimates, and that the problems with treating wages as representative of living standards in a society like 18th century China are quite severe, even if we could resolve the more narrowly empirical problems with the wage data themselves (e.g. unstated but significant in-kind components, differences in the currencies in which Chinese wages and prices are often quoted, uncertainty about the length of the “day” in day wages, and so on). But at the moment, the consumption data also have significant problems; so while this may well be where we want to concentrate future research efforts, that does not mean that this is the only metric we should be tracking as we make our best guesses about the current state of this controversy. (We could, of course, theoretically all agree not to comment on this controversy until we know more, but that seems unrealistic, given how many other issues it touches upon.) So I would be inclined to keep the wage data, and even the GDP estimates, in play in this discussion, even though I share Deng and O’Brien’s sense that the consumption data are stronger (and even though those data are more favorable to the position I took in my 2000 book). Let me briefly review each of these areas.
Comparative GDP Estimates
I would not disagree with Deng and O’Brien’s critique of the GDP approach, or of Maddison’s figures in particular: the latter were based on far too narrow an evidentiary base for much of the world and much of history. I think, however, it is worth noting that there have been some more recent attempts to estimate GDPs, which have the advantage both of somewhat better data and of attempting to construct GDP figures for the Yangzi Delta, rather than for China. Given the points that some of us have made repeatedly about the vastly different scales of China and any single European nation, and about the advantages of comparing the most developed region of a continent-sized China with the richest regions of Europe rather than insisting on national units, these numbers seem to be worth at least some attention.
For instance, a recent paper by Stephen Broadberry, Hanhui Guan and David Daokui Li suggests that Britain must have overtaken the Yangzi Delta in per capita GDP by the first quarter of the 18th century. This is, of course, materially different from my claim in The Great Divergence that the Yangzi Delta had not fallen significantly behind until well into the second half of the 18th century, and maybe not until 1800 (though my conclusions were based on estimates of consumption, longevity, and basic human well-being, rather than GDP). Nor has this paper found a way around all the problems with historical GDP and comparisons between very different societies to which Deng and O’Brien have pointed.(These include the radically different market baskets of these two societies, which had relatively little trade between them, the very different degree to which the government tracked various kinds of production, and others; to which I would add, large differences in the degree to which various goods and services passed through the market.) Still, I would not want to simply discard such work, given the difficulties that dog other approaches. Moreover, I think it is noteworthy that a debate between an early and a late 18th century divergence represents a considerably different intellectual landscape than the one we would have if we relied on Maddison’s GDP numbers, or on the non-quantitative work of David Landes, Deepak Lal, and various others – or for that matter, on an earlier attempt by Guan and Li to estimate comparative GDPs, which had previously claimed that a huge gap already existed in the 15th century. An earlier paper by Debin Ma suggested that the per capita GDP of the Lower Yangzi — which he defines slightly more broadly than I do, so that it includes some poorer areas – probably exceeded that of China in general by about 50% in 1750: such an adjustment, patched onto Maddison’s data, would leave the Delta behind Britain, but by considerably less (in percentage terms) than, say, Germany lags Norway today, or the U.S. lags Luxembourg (when nobody argues that this makes either Germany or the U.S. a “backward” economy fundamentally different from the “advanced” ones, as we used to think China was relative to early modern Europe.). Similarly, a study by Li and Van Zanden, based on data for the 1820s, finds Holland well ahead of two counties in the Delta for which there are particularly good data in those years; but as they note, the 1820s were a period of both agricultural crisis due to natural disasters and depression in the cloth trade, the second biggest sector of that region. They suggest that if we had data for 1800, it would show a smaller gap, but still a significant one, with Holland perhaps 50 or even 70% higher in per capita GDP. Admittedly, that is far from the rough parity I had originally suggested at 1800, and would now be inclined to put at somewhere around 1750 instead; there are some plausible adjustments that I think would narrow the gap further, but that is not really the point for now. Instead I would emphasize that despite continuing disagreements and continuing data problems – the latter of which will probably never be fully solved – we have made some progress in narrowing the range of plausible answers about when and how much divergence occurred in these terms. Even if GDP is a seriously flawed measure for purposes of this debate, I am not sure we want to throw it out entirely.
Real Wage Comparisons
The wage data is similarly vexed, and Deng and O’Brien have, I think, explained very convincingly some reasons for skepticism. But let me add two more, extend the discussion of another, and suggest a possible implication of how we might read this imperfect data in conjunction with the even more imperfect GDP data.
The first additional cause for skepticism in Robert Allen’s own reconstruction of real wages based on prices and wages recorded by the supercargo of an East India Company ship docked in Guangzhou during the trading season of 1704. Using the same basic approach as he and his co-authors use on the larger data set they compiled a bit later, Allen arrives at the conclusion that the real wages here were roughly equal to those in London at the same time – and Guangzhou, it is generally agreed, was not quite as wealthy as the Yangzi Delta. (Allen 2004). Let me make clear that I am not saying that these data give the correct picture, and the other data a false one: there are certainly reasons why the Guangzhou data could be unrepresentative. But these are at least prices and wages that we know were actually paid by individuals in private markets, and that should be relatively free of unreported in-kind benefits. (Chinese workers would certainly not have lived on a British ship, and are unlikely to have eaten British food.) By contrast, the Chinese data in the later and better-known 5 author paper on early modern wages around the world are from administrative reports of prevailing wages, with no evidence that the reporters actually made any sort of survey, and no data at all on in-kind compensation. I would thus not be quick to throw out the Guangzhou material just because it is a smaller database – and that it gives a figure which is so much higher than those derived from the other data seems lke a good reason to doubt the data overall. There is also something highly suspicious about the lack of difference between the wage rates for different parts of China in the administrative data, when every qualitative source we know of agrees that there were very large regional differences in material living standards. Again, this is not sufficient reason to discard these data when we have so few; on the contrary, they represent the best large-scale data set that we have, and the authors have used them to make interesting claims. But I think we need to be very careful about how we use them for this particular debate.
It is crucial that most Chinese did not rely on wages for their income – as O’Brien and Deng note. But we also know quite a bit about the likely relationship between these wages and peasant earnings. All our evidence suggests that tenant farmers earned a great deal more than agricultural laborers, even if the latter were able to find year-round employment. This conclusion is confirmed whether we look at estimates of money earnings, convert wages into cash and compare to farm yields net of rent and other expenses, or consider key social indicators – including the especially lowly status of wage laborers, and the fact that most tenants were able to marry and raise families, and most agricultural laborers could not. And in a society without strong guilds or unions, there is no reason to think that unskilled wages in town would be much higher than in the countryside. (This would be particularly true in the Yangzi Delta, where towns were very densely distributed across the map, so that almost any rural resident could bid for an unskilled job in a town without going very far. ) Indeed, in the admittedly limited data we have, the difference between the earnings of a farmer with an average-sized tenancy and a wage laborer are on the order of 2.5 – 3.0:1. Under the circumstances, wages are a very poor guide to popular living standards: and comparing them to wages in England or Holland, where proletarians made up a large portion of the labor force by the 18th and especially the 19th century, represents a comparison between the bottom of the income distribution in one place and something approaching to middle in the other. 
But even if these wages do not tell us much about comparative living standards, they might nonetheless tell us something about trends in comparative labor productivity. Significantly, estimates of rural incomes and rural labor productivity suggest that the Yangzi Delta was still on a par with England and Holland on these measures even as late as the 1820s. This would place it far above the rest of Europe – including, of course, a number of countries that began mechanized industrialization and sustained per capita growth well before China did. Delta agriculture was also still well ahead in total factor productivity in agriculture as late as 1820. When taken together with the evidence already discussed which suggests a relatively late divergence, and therefore a fairly sudden widening of the gap once it manifested itself – since nobody doubts that it was quite large by the mid-19th century – this would at least suggest that we should not probably be looking at agriculture to explain the divergence. It also suggests, as Robert Allen has argued in his discussion of British industrialization, that it was higher wages in the growing urban sector that pulled up rural wages, necessitating labor-saving innovations in agriculture, rather than agriculture creating an enlarged urban work force by shedding workers on its own. It does not appear that urban demand exerted a comparably strong effect on rural Chinese wages, even though the barriers to rural-urban migration – whether in the form of exclusionary institutions or urban dis-amenities – were comparatively weak in Qing times; instead, the most likely reason not to leave was that, as noted above, most rural people earned far more than unskilled wage laborers in either city or countryside, and nothing was pushing up urban wages fast enough to overcome this disincentive. A significant gap in urban real wages – if confirmed by further studies that can more fully overcome the problems described by Deng and O’Brien – might then be significant not as a sign of a difference in living standards that already existed, but as a sign of urban changes in Europe that were beginning to create such differences.
(to be continued…)
 Broadberry, Guan and Li 2014..
 Maddison 2001: 42, suggesting that Western Europe overtook China ca. 1300.
 Guan and Li had previously argued (2010) that China was far behind by the 15th century, if not earlier, and had fallen even further behind over the succeeding centuries. See also Landes 1998, Lal 1998.
 Ma 2004. Compare Maddison 2003: 262.for an interpellated UK figure.
 Li and Van Zanden 2012:973,
 It is true, for instance, that not every Chinese worker in Guangzhou was able to offer his services to the foreigners who docked there, perhaps reducing competition and driving up wages. Bt quite a few could – numerous memoirs from foreigners who visited Guangzhou in this period speak about foreigners being besieged by crowds of potential porters and other service providers. Nor is it clear why any restrictions would have driven up the price of labor more than it did that of the many kinds of provisions for which Lockyer records prices paid, and which Allen uses to create the denominator of his real wage.
 I review some of this data in Pomeranz 2006, and Pomeranz 2011. On the strength of tenant rights, which helped make this income differential durable, see for instance Rawski 1972 Li Wenzhi and Jiang Taixin 2005;Yang Guozhen 1988; Zhang Peiguo 2002.
 On the distribution of towns across space see Xue Yong 206: 319, and 302-346, 432-475 for a discussion of various estimates of urbanization in the Yangzi Delta more generally.
 Pomeranz 2011.
 Allen 2009b; Li and Van Zanden 2012.
 For intra-European comparisons, see Allen 2000..
 Li and Van Zanden 2012:975; Allen 2009b.
 Allen 2009a.
 On Chinese guilds in this period, see Moll-Murata 2009, Pomeranz 2013. Though our data is thus far inconclusive, China does not appear to have had a pronounced “urban graveyard effect” – that is clear evidence of worse health and higher mortality in cities, providing a disincentive to migration that had to be overcome by significant wage differentials, as was the case in early modern and industrializing Europe.