{ This post is mostly stringing together my scattered tweets over the past couple of weeks. I’ve had numerous discussions on this subject with Vincent Geloso, Judy Stephenson, Ben Schneider, Benjamin Guilbert, Anton Howes, and Mark Koyama. But yesterday Geloso sent me the paper he’s working on for Alsatian wages and that kick-started further thoughts I shared with Geloso privately, and then with the others on Twitter. You can follow the most recent discussion below this tweet, although it’s very difficult to keep track of the many different threads. I’m generally a sceptic of Allen’s theory, but in this post it seems I ended up critiqueing the critiques as much as Allen himself. }

First, a quick preface. I love the work of Robert Allen. I love his papers on steel from the 1970s and 1980s. I have a love-hate relationship (on some days love, some days hate) with his book on the Soviet Union. I swoon over his work on English agriculture. And his little book on global economic history — is there a greater marvel of illuminating concision than that?

Allen has an old-fashioned interest in the economics of making stuff, the bread-and-butter of traditional economic history. He doesn’t shy away from learning the nuts and bolts of technology, which too many economists and historians do these days. Inasmuch as he discusses other things like institutions or culture, he doesn’t get carried away by lofty abstractions, and his point of departure is always the very concrete reasons that a firm or an industry or a country is more productive than another. I’m not rubbishing institutions or culture as explanations — I’m just saying, Allen’s virtue is to start with problems of production first.

Yet I always find myself in the peculiar position of loving his work like a fan-girl and disagreeing with so much of it.

In particular, I’m sceptical of his theory of the Industrial Revolution.

Allen has been advocating for at least 20 years now that England in the 18th century possessed a “high wage economy”. English labour costs relative to continental Europe and Asia were unusually high. This is an important part of his “induced innovation theory” for the invention and adoption of machines in the leading industries of the Industrial Revolution. In short, England’s high wages relative to its cheap energy and low capital costs biased technical innovation in favour of labour-saving equipment, and that is why it was cost-effective to industrialise in England first, before the rest of Europe (let alone Asia).

I hasten to add, Allen’s is not a monocausal theory. To the contrary, it is a complete multi-causal model, but his distinctive contribution is the high-wage economy. Here is Allen’s own flowchart taken from the book:

The theory is appealing, in part, because the technological innovations of the early Industrial Revolution were not exactly rocket science (a phrase used by Allen himself), so one wonders why they weren’t invented earlier and elsewhere. (Mokyr paraphrasing Cardwell said something like nothing invented in the early IR period would have puzzled Archimedes.)

But I’ve always had reasons to doubt it. As Mokyr has tirelessly argued, inventions were too widespread across British society to be a matter of just the right incentives and expanding markets — and this is a point now being massively amplified by Anton Howes.

There are more concrete reasons for scepticism. As Kelly, Mokyr, & Ó Gráda (2014) have pointed out, although nominal and real wages were indeed higher in Britain, Allen must assume that unit labour costs (wage divided by labour productivity) were also higher. But if the Anglo-French wage gap were matched by a commensurate labour productivity gap, then the labour cost to the employer would have been the same in the two countries. Actually Allen himself brings up the issue of unit labour cost in his book, but mostly hand-waves it away and implicitly assumes that ULC was higher in England. But that’s far from proven.

Besides, you already had capital-intensive production techniques in several sectors well before the classic industrial revolution period — especially in silk and calico-printing. Silk-throwing (analogous to spinning in cotton) was mechanised in Italy before 1700. The idea was pirated by Lombe who set up a water-powered silk-throwing factory circa 1719, and he was imitated by many others by the 1730s. Then you had heavily machine-dependent printing works for textiles (especially calicoes) in many European cities before the canonical industrial revolution period. None of these seemed to require Allen’s “high wage economy”. (Not to mention, Allen’s model has implications for the diffusion of the Industrial Revolution, and Scottish industrialisation was almost simultaneous with the English one, despite wage differences.)

Nonetheless, I had mentally reconciled Allen and Mokyr in the manner of Crafts by considering Mokyr = supply of inventions, Allen = demand.

But there has been a spate of critiques of Allen’s work recently. Humphries (2013); Gragnolati et al. (2011); and Stephenson (2016). The latter establishes through archival research that those builders’ wages for London on which so much of Allen’s reasoning is based weren’t wages at all, but fees paid to labour contractors and in fact the wages received were at least 20-30% lower. (That doesn’t really address the issue of the actual labour cost to firms, though.)

Then there’s Humphries & Schneider (2016). Most of the wages cited in the literature have been drawn from secondary literature (books, pamphlets, etc.), but Humphries & Schneider actually dug into all kinds of archival sources to show that the estimated 1 million women and children who spun yarn with wool, linen, and cotton in their rural homes were paid much lower wages than Allen’s narrative has relied on: ~4 d [pence] per day, rather than the >8d/day assumed in Allen. And one of the showcases of his theory is the series of inventions mechanising yarn spinning!

The source of the data makes H & S conclusions persuasive, but it’s also theoretically compelling. Men, especially in big cities, may have been paid higher wages, but women and children in the countryside were not. This makes early modern England much more like a “surplus labour economy” with an “unlimited supply of labour” à la Arthur Lewis. H & S describe putting-out merchants expanding their network of spinners farther and farther away from their core areas to find fresh labour, so that even as the demand for cloth rose they could avoid bidding up wages. This was probably reinforced by a cartel-like arrangement amongst the merchants. Labour market monopsonists also loom large in modern development microeconomics!

Quick thoughts on some of these critiques:

Allen v Gragnolati: At least on the question of the jenny, Allen’s Achilles’s heel may be working hours. In the spinning jenny paper where Allen argues the jenny was profitable in England but not in France, he assumes total production stays the same when spinning productivity rises with the jenny, because households reduce working hours in response. Gragnolati et al. asked him why not increase production in order to earn more income? The jenny would have been profitable in France as long as total output rose. Allen’s response was to cite his paper with Weisdorf. He argues that the existing working hours/year estimates are consistent with the idea that British households maintained a fixed consumption target, according to which they adjusted work hours in conjunction with market wage rates. Allen’s argument therefore implicitly assumes that British and French spinners had similar preferences for leisure-work tradeoff. But this is a highly uncertain result, and it goes against the narrative of the “Industrious Revolution” in which workers began supplying more labour in order to buy new consumer goods available. I can’t imagine this will hold up in future work. Besides, I violently hate the “peasant mode of production” idea… (Also see this.)

The wage gap & market size: I’ve mentioned this many times to people on both Twitter and in real life, but the role of market size in Allen’s model gets too often neglected. The question that no critic of Allen has so far posed is this: what is the wage gap between Britain and France that renders inventions in Great Britain profitable but not in France, given the two countries’ differences in market size. There’s a balance between factor savings due to inventions and the costs of invention which are reduced as a function of market size (i.e., costs divided by the number of goods possible to produce in a given market. In terms of the isocost model used by Allen which simplifies a section of Acemoglu’s “Directed Technical Change” paper, bigger the market size, bigger is the isocost’s shift to the origin for any given fixed cost of invention. BTW, the wage rate affects the slope of the French and British isocost curves in Allen.)

The real wage ratio for Paris/London used by Allen is ~50% in 1750-1775 and ~57% in 1775-1786, and this gets adjusted by Stephenson (2016) to ~62% and ~71% respectively. But the significance of the “Stephenson adjustment” can only be assessed in relation to market size differentials between France and Britain. And by “market size” we must take into consideration not only population and colonies but also internal barriers to trade.

But of course it’s entirely possible that further research might revise French wages upward or downward, making the Anglo-French wage gap smaller or bigger.

Allen & labour markets: The downward wage revision for spinners by Humphries & Schneider (2016) is also restricted to Britain and therefore does not address Allen’s international comparison.

Another potential problem is there may be a regional and sectoral heterogeneity in spinning wages, and Humphries & Schneider have very few cotton- and Lancashire-specific observations.

Allen assumes that wages in cotton were set by wool, which as late as 1770 is estimated at >90% of textile value added in the UK. This is equivalent to an assumption that the British labour market was well-integrated and labour was mobile. But this is unrealistic. There were natural, institutional, infrastructural, and possibly cultural reasons for labour to be relatively immobile at the time — at least between provinces rather than from the provinces to the cities. And this would have been even more likely under the rural putting-out system. Remember, we’re not talking about firms here, but about proto-industry — production taking place inside people’s homes.

And IF labour markets were highly local and fragmented, then that ironically supports Allen’s view against the criticism found in Humphries & Schneider (2016). What matters is not the ‘national’ wage set by spinning wool, but the specifically cotton wages paid specifically in Lancashire. (There might have even been variation within Lancashire.) You can have local labour ‘shortages’, when the labour market is not national.

In the surplus labour scenario envisioned by Humphries & Schneider, rising demand for labour need not bid up wages. But if labour markets were fragmented and local, then the Lewis-like model need not apply.

It is not convincing to argue that ‘labour-saving’ is seldom if ever mentioned in patent applications or in inventors’ records. Demand for cotton goods was growing faster than wool, and on first approximation this implies demand for cotton-specific labour was growing and cotton-specific wages could be bid up in local labour markets. The motive for producing more output faster is equivalent (especially in a constant returns-to-scale industry like textile putting-out) to demanding more labour.

It could be argued that even if labour was not very mobile, the putting-out merchants were. They could have widened their spinning network and induced more households to spin cotton rather than wool, as the demand for cotton grew. But (again) that implies higher wages in cotton than in wool.

Besides, under the putting-out system, expanding output (=increasing labour inputs proportionately) could raise transaction costs rapidly and prohibitively. Merchant-middlemen had to deliver raw material to each household, retrieve the yarn, then deliver the yarn to weavers, and then retrieve the cloth. Thus there was a natural constraint on output.

Edit (12 July 2018): Allen has replied to Stephenson. I have a short comment on Allen’s response to Stephenson.

Edit (7 Dec. 2016): Besides the question, “why was it not invented in France?”, I would also ask, “under what factor price ratios did the technology diffuse to France?” and “how did the British technology that was finally adopted in France differ from the original inventions in Britain”?

Allen’s model has very strong implications not just for invention, but for diffusion of technology. The initial invention is factor-biased, but subsequent modifications/learning-by-doing/microinventions economise on all factors of production. (Allen’s example: more energy efficiency even though coal was cheap.) When unit costs fall enough, then even in economic environments with quite different factor price ratios, the adoption of the modified technology should be possible.

That’s his theory anyway. But that adds another method of testing. Were the factor price ratios in France after 1800 ‘correct’ for the technology that was adopted? The jennies and the water-frames that were adopted in France after 1790 almost certainly had more spindles and (in the case of the frame) were more energy-efficient.

Other blogs on Allen:

How (much) were British workers paid? Evidence beyond wage rates

Spinning little stories: Why cotton in the Industrial Revolution was not what you think

The High Wage Economy: the Stephenson critic

England circa 1700: low-wage or high-wage, which blogs about the new working paper by Humphries & Weisdorf (2016). Vincent Geloso’s summary: “preindustrial labor markets had search costs; workers were willing to sacrifice on the daily wage rate (lower w) in order to obtain steady employment (greater L) and thus the proper variable of interest is the wage paid on annual contracts”.

Vollrath on Allen vs Mokyr

Howes on Allen vs Mokyr

Vincent Geloso’s summary of the Twitter ménage à entre-quatre-et-huit.

By the way, evidence from Spain does lend some support to Allen. Martínez-Galarraga & Prat (2015) find relatively high wages were a factor in Catalan industrialisation. Also see their post explaining their findings at Nada es Gratis (in Spanish). Some other historical induced innovation evidence include: Hornbeck & Naidu (2014) and Hanlon (2015).

Edit 7 Dec 2016: Vincent Geloso has another post with a new summary.

Edit 6 Dec 2016: Several people have mentioned Temin‘s debate with Habakkuk as an argument also against Allen. The USA was also a high wage economy relative to UK in the 19th century, and this was also true for other land-abundant settler economies such as Canada, Australia, and possibly Argentina in the 19th century. However, Temin’s actual argument was that wages in the USA were higher than in the UK, but then so were interest rates and the cost of capital. Allen, by contrast, argues that wages in England > France, but both energy and capital were cheaper in England than in France. The USA, as it happens, was abundant in land and natural resources and manufacturing was quite resource-intensive. Cf. Wright (1990). That’s consistent with Allen’s induced innovation idea. Also see Allen (2012).