Kelly, Mokyr & Ó Gráda (2023) — The Mechanics of the Industrial Revolution

Citation. Kelly, Morgan, Joel Mokyr, and Cormac Ó Gráda. “The Mechanics of the Industrial Revolution.” Journal of Political Economy 131(1), 2023: 59–94. Working-paper version: UCD Centre for Economic Research WP2020/16 (June 2020).

Summary

The major follow-up to Precocious Albion (2014), translating the upper-tail human-capital thesis from a synthetic argument into a county-level empirical test. Kelly, Mokyr, and Ó Gráda construct a panel of the 41 counties of England and model the share of male employment in textile production in 1831 as a function of pre-IR county characteristics. Pre-IR mechanical-skill supply is measured by the share of men aged 60–69 in the 1851 Census described as “mechanics and toolmakers” by county of birth — men who would have been apprenticed in the 1790s to masters trained a generation earlier.

Headline findings (all from the paper’s main specifications):

• Mechanical skill supply and low wages jointly explain ~70% of the variation in 1831 textile employment across counties. Literacy, access to capital (bank density), and direct proximity to coal have little additional explanatory power.
• Elasticity of textile employment with respect to skill supply: +2; with respect to wages: -6. In the authors’ formulation: “high wages acted as a powerful disincentive to successful mechanization.”
• Real wages rose sharply in the industrializing north and collapsed in the previously prosperous south. The widespread claim that living standards were static during the IR (Feinstein 1998) is “a statistical artefact of aggregating together two regions that were moving in sharply opposite directions.”
• Placebo tests pass. In industries where the technology was static (food processing, shoes, garments, woodworking), neither low wages nor accumulated skill predicts employment concentration — as the thesis predicts.
• Instrumental-variables robustness. The supply-of-skill measure is instrumented with (a) mid-18th-century apprenticeship fees for watchmakers (from Watchmakers’ Company records, Moore 2003), (b) 1700 population density relative to farmland, and (c) distance to coal. Skills IV coefficients are largely unchanged; the IV strategy is the paper’s answer to the endogeneity concern that new industries might have attracted rather than been caused by skills.

On coal — the paper’s position is more nuanced than “coal didn’t matter”:

“Coal was vital to early British success not so much for its ability to supply motive power but as a cheap source of heat that allowed metalworking to emerge to an extent unrivaled elsewhere.”

Coal’s direct effect on textile employment is small because early textile machinery ran on water power and by 1851 coal was cheaply available everywhere in Britain. But coal’s indirect effect — enabling centuries of metalworking, which trained the artisan population that built the IR’s machinery — is central. Coal appears in the IV specification as a predictor of skill supply, not of industrialization directly. This distinction is easy to misread.

On what the paper does NOT claim. The authors explicitly reject the reading “low wages and high mechanical skills caused the Industrial Revolution”:

“We see British industrialization as a lengthy and historically contingent process of slowly accelerating technological change, in which practical procedural knowledge grew alongside formal propositional knowledge in a process that stretched back centuries.”

The paper is a test of a specific within-England empirical prediction consistent with their broader upper-tail-human-capital thesis. It does not attempt to explain why Britain rather than France or China industrialized — at the between-country scale, the paper separately notes, English unskilled wages were not lower than French (after productivity adjustment) but English skilled wages were — producing a sustained flow of English artisans to France from the 1750s onward.

Key claims

• The dependent variable that matters empirically is within-England textile-industry location, 1831. Pre-IR low wages and high mechanical-skill stocks together explain ~70% of its variance across the 41 counties; other standard explanatory variables (literacy, banks, direct coal proximity) do not.
• Measured effects: skills elasticity +2, wages elasticity -6.
• Coal matters indirectly (enabling centuries of metalworking → artisans) not directly (as motive power for textile machinery).
• The “static real wages during the IR” finding (Feinstein 1998) is aggregate-artifact: industrializing-north wages rose, stagnant-south wages fell.
• The finding supports the authors’ broader upper-tail human-capital view and is the most direct within-England empirical challenge to Allen’s high-wage thesis.
• The authors resist reducing their finding to a slogan; they frame the paper as one piece of a longer-run procedural-knowledge-accumulation story, not as a new monocausal theory.