Upper-tail human capital & bricolage

Thesis

The conventional human-capital story (more years of schooling raise productivity) explains essentially none of the early IR — average British literacy and schooling were unremarkable by Northwestern European standards in 1750, and Germany was more literate than Britain throughout the 18th century without industrializing first. What was unusual about Britain was the upper tail: a deep stratum of skilled artisans, mechanics, instrument-makers, millwrights, clockmakers, and engineers — the population from which inventors and (more importantly) the people who implemented inventions were drawn. The IR’s signature innovations were rarely flashes of theoretical brilliance; they were thousands of small mechanical refinements (Mokyr’s bricolage) accumulated over decades by competent workshop craftsmen who could read a drawing, file a part to tolerance, and modify a balky machine until it ran. Britain had this bench in unusual depth and density — a legacy of pre-industrial industries (clocks, instruments, naval supplies, mining) — and other societies did not.

This is a structural claim about the human-capital distribution, not the mean. It is closely associated with Mokyr’s Industrial Enlightenment but logically distinct: the upper tail mattered even for inventions that had no formal-science content. It is also the supply-side counterpart to Allen’s high-wage demand-side story, and the most direct quantitative challenge to it.

Lead proponents

Joel Mokyr — the framing is most fully developed in The Enlightened Economy (2009), with conceptual antecedents in The Lever of Riches (1990) and The Gifts of Athena (2002). Mokyr’s distinction between propositional knowledge (what is true about the world) and prescriptive knowledge (instructions for making things) underwrites the upper-tail thesis: the IR was a propositional-knowledge revolution feeding through a uniquely capable prescriptive-knowledge community of British artisans. Without the artisan community, the propositional knowledge has nowhere to land.
Hans-Joachim Voth — Mokyr’s most frequent recent collaborator on the great-inventor side, including the influential Mokyr–Voth chapter “Understanding growth in Europe, 1700–1870” in the Cambridge Economic History (2010), which sets out the upper-tail framing as a synthesis of the cultural and the empirical evidence. Ralf Meisenzahl & Mokyr’s “The Rate and Direction of Invention in the British Industrial Revolution” (NBER WP 16993, 2011) compiles biographical data on 759 “great inventors” across 1660–1830 and shows the population to be overwhelmingly artisan-trained, not university-trained.
Morgan Kelly & Cormac Ó Gráda — a series of papers with Mokyr that turn the framing into testable empirical work. The canonical Precocious Albion (2014) reconstructs British wages adjusted for skill and human capital, arguing that what looks like high wages in Allen’s data is largely high human capital per worker. The follow-up Mechanics of the Industrial Revolution (2023, JPE) tests a within-England prediction: 1831 county-level textile-employment shares regressed on pre-IR mechanical skill (proxied by the 1851 Census’s “mechanics and toolmakers” aged 60–69) and pre-IR wages. Skills elasticity +2, wage elasticity −6, joint R² ≈ 0.7 across the 41 counties. Literacy, bank density, and direct coal proximity add no significant explanatory power. Placebo industries (food, shoes, garments, woodworking) show no relationship — exactly what the thesis predicts.
Mara Squicciarini & Nico Voigtländer — Human Capital and Industrialization: Evidence from the Age of Enlightenment (QJE, 2015) is the continental complement. They use subscription density to Diderot and d’Alembert’s Encyclopédie across French départements as a proxy for the upper tail of human capital (the Encyclopédie was an expensive subscription-only work bought by improving artisans, manufacturers, and minor philosophes) and show that Encyclopédie density predicts post-1750 city growth and post-1800 manufacturing wages, while basic literacy does not. The paper is a near-clean falsification of the mass-literacy version of the human-capital story for early industrialization.
Jane Humphries — Childhood and Child Labour in the British Industrial Revolution (2010) provides the foundational reconstruction of the British apprenticeship system as the institutional channel for upper-tail skill formation. Uses ~600 working-class autobiographies as the primary evidentiary base.

The apprenticeship channel

The upper-tail thesis stands or falls on whether there was an institutional mechanism actually producing the artisans it requires. The leading candidate, since Humphries’s work, is the apprenticeship system.

The headline magnitudes are larger than older accounts conceded. Humphries reconstructs from autobiographical and indenture evidence that perhaps a third of working-class boys spent significant time in formal apprenticeship in the 18th century — not the residual share of the population it had become by 1900. Chris Minns and Patrick Wallis (Economic History Review, 2013, building on Wallis 2008) confirm from the Stationers’, Goldsmiths’, and Carpenters’ Company records that completion rates were substantially higher than older estimates suggested (perhaps 60% of indentured apprentices completed their term — or moved laterally into adult journeyman work without formally completing — versus the older 40% figure that was an artifact of London-elite-company overcounting of formal completions).

Three institutional features distinguished the British apprenticeship system from continental analogues. First, the 1563 Statute of Artificers had been progressively eroded by 1750, leaving apprenticeship as a contractual training relationship rather than a guild-policed one. The Statute was formally repealed in 1814; well before then, in practice, masters could take on apprentices without guild approval and apprentices could move between trades. Second, premium fees were heavily subsidized by parish poor-rate apprenticeships (the parish paid a master to take on a poor child as apprentice, both as poor-relief and as skill-formation; Humphries documents this as a major channel into trades like millwrighting, blacksmithing, and shoemaking). Third, the transferable nature of British mechanical skills — the same gear-cutting and fitting techniques worked across clockmaking, instrument-making, mining-engine work, and textile machinery — meant that an apprentice trained in one trade could move into another as the IR’s labor demands shifted.

The continental contrast is substantial. French apprenticeship remained more guild-policed through the early 18th century, and entry into mastership was structurally restricted by maîtrise fees and family-line preferences. German apprenticeship varied by territory but in much of the southern German Handwerk tradition required Wanderschaft and was tied to closed guilds. Where continental apprenticeship was strong (the Dutch Republic in the 17th century; some Swiss cantons in the 18th), it tracked relatively early industrialization. Where it was weak or restrictive (much of Mediterranean Europe; the Russian Empire), it tracked late or absent industrialization. The pattern is suggestive without being decisive — endogeneity to commercial demand for skilled labour is hard to rule out.

Continental evidence: the Encyclopédie test

The Squicciarini–Voigtländer (2015) test deserves attention because it is one of the cleanest available natural experiments on the upper-tail vs. mass-literacy question.

The setup. France in the 1770s had widely varying literacy rates across départements (the famous Maggiolo map of 1685–1690 marriage signatures, supplemented by 18th-century data, shows a literacy gradient running roughly NE–SW from Picardy down to Languedoc). It also had widely varying Encyclopédie subscription rates — known precisely from Robert Darnton’s reconstruction of subscription lists across the 1751–72 publication period. Subscription was expensive (the full Encyclopédie cost roughly the annual wage of a master craftsman), targeted at improving artisans and manufacturers, and is a clean proxy for the upper tail of practical-knowledge human capital — separable from the mass-literacy distribution.

The result. Encyclopédie subscription density predicts subsequent city growth (1750–1850) with a robust elasticity around 0.3, controlling for initial city size, region, and basic literacy. It also predicts post-1800 manufacturing wage premia. Mass literacy, by contrast, does not significantly predict either outcome. The within-France pattern is opposite to the cross-country prejudice that “Germany had higher literacy and industrialized later, so literacy doesn’t matter” — Squicciarini–Voigtländer show the same null within France while showing that the upper-tail measure does the work.

The methodological caveat is that Encyclopédie subscription is itself partly endogenous to commercial dynamism — improving manufacturers in dynamic regions both subscribed and grew. The authors instrument with distance to the Encyclopédie publishing centres (Paris, Neuchâtel) and with the existence of provincial scientific academies, which addresses the criticism partially. The result has held up in subsequent literature.

Great inventors and the artisan-engineer overlap

The Meisenzahl–Mokyr “great inventors” project (NBER WP 16993, 2011, and subsequent work) compiled biographical data on 759 individuals identified as significant British inventors across 1660–1830. The findings are sharply against the heroic-natural-philosopher narrative.

Roughly 80% of the great inventors had no university education whatsoever. Of those who did, most had read law or medicine, not natural philosophy.
The dominant educational background was apprenticeship plus self-study, often supplemented by membership in provincial scientific or improving societies (the Lunar Society of Birmingham, the Manchester Literary and Philosophical Society, the Birmingham Brassfounders’ Society).
Inventors clustered geographically in the established artisan regions — Birmingham/Black Country, Lancashire, the West Riding, London — and not in university towns.
Most inventors held trades that overlapped multiple times with mining, clockmaking, instrument-making, or mill-engineering. Trevithick was a Cornish mining engineer’s son; Stephenson was a colliery brakesman; Maudslay began as a powder-monkey at Woolwich Arsenal; Watt was an instrument maker for Glasgow University; Crompton built his own mule from clockmaker’s gears.

The implication is precise. The upper tail that mattered was not “scientifically literate elites” — it was “highly skilled artisan-engineers, often in the upper artisan stratum, embedded in mechanical trades with cumulative procedural knowledge.” This is what makes the upper-tail thesis empirically distinguishable from Mokyr’s broader Industrial Enlightenment story: most of the great inventors did not look like Industrial Enlightenment fellows and did not read in the Republic of Letters.

The wage debate and the skill composition argument

The deepest tension in the IR literature is between the upper-tail story and Allen’s high-wage induced-innovation thesis. The two cannot both be straightforwardly true: if British wages were high (Allen), British workers were expensive and machinery was substitution; if British wages were not really high once you adjust for skill (Kelly–Mokyr–Ó Gráda), the price-driven inducement story collapses.

The Kelly–Mokyr–Ó Gráda critique runs in two layers. The level layer (Precocious Albion, 2014) argues that Allen’s wage data overstate British real wages relative to continental ones because Allen does not adequately adjust for skill content. London building craftsmen (the core of Allen’s series) were not strictly comparable to Parisian or Antwerp counterparts; the British were more skilled per nominal occupational title. Adjusting for productivity per worker, the real-wage gap shrinks substantially. The within-country layer (Mechanics, 2023) shows that within England the regions that industrialized first were low-wage regions with high mechanical-skill stocks — the opposite of what the high-wage thesis predicts at the within-country level.

Allen’s response (in interviews and revisions, less in formal published rebuttal) has been (a) to dispute the skill-composition adjustment as overcorrecting and (b) to argue that the within-country and between-country levels of the argument operate differently — between-country wage differences caused machinery substitution; within-country variation in textile location reflects a different equilibrium where coal and water-power access are first-order. The within-country/between-country move is methodologically defensible but rhetorically deflationary, since it concedes that the within-country test does not run the way the original thesis predicted.

For the upper-tail thesis, the broader implication is that the British “high wages” Allen observed were partly an outcome of the upper-tail human capital that made British workers more productive — not an exogenous shock that induced labour-saving innovation. The two stories may converge if read non-mechanistically: Britain had high productivity because of its skill stock; the high productivity manifested as high wages; the high wages did then make labour-saving innovation marginally more profitable; but the causal arrow runs from skill to wages to innovation, not from exogenous demographic-induced wages to innovation.

Key arguments

The IR’s inventions were artisan inventions. Hargreaves (handloom weaver), Crompton (mule), Cort (puddling), Trevithick (mining engineer), Maudslay (instrument maker), Stephenson (colliery brakesman). These were not Royal Society fellows. They were exceptionally skilled craftsmen who built, tested, and refined machinery in workshop environments. Meisenzahl–Mokyr’s 759-inventor biographical dataset confirms this systematically.
Britain’s pre-industrial industries built the bench. Centuries of clockmaking (Lancashire and London), nautical instruments (Greenwich), brass and pewter (Birmingham), iron-mongery, locksmithing, and mining produced a population of artisans with transferable mechanical skills — gear-cutting, precision boring, drawing-from-prints. When the early IR demanded engines and machines, the workshop population was there to build them.
Apprenticeship density was the institutional channel. The British apprenticeship system was unusually large (perhaps a third of working-class boys), unusually open (loosely guild-policed, with parish-subsidized entry), and produced workers with codified, transferable skills. Humphries (2010) and Minns–Wallis (2013) document the scale and completion-rate evidence directly.
The mean doesn’t matter; the tail does. A society with average literacy 50% but a 5% share of highly skilled mechanics will industrialize faster than one with average literacy 70% and a 1% share of skilled mechanics. The IR was a tail event in human-capital space, not a mean event. Squicciarini–Voigtländer (2015) confirm this empirically across French départements: Encyclopédie subscription density predicts city growth; mass literacy does not.
It connects to the wage debate as a deflation of Allen. If British wages look high, it may be because British workers were more skilled per worker — i.e., the high-wage observation is a high-human-capital observation in disguise. The within-England test in Kelly–Mokyr–Ó Gráda 2023 runs in the opposite direction to the high-wage thesis at the within-country level: textile employment concentrated in low-wage, high-skill counties.

Key evidence

Apprenticeship records. Humphries (2010) reconstructs scale and demographics from ~600 working-class autobiographies; Minns–Wallis (2013) reconstruct completion rates from London company records. The system was large, open by European standards, and produced workers with transferable mechanical skills.
Patent inventor biographies. Khan, Sokoloff, and others have parsed early-19th-century British and American patents by inventor occupation, showing the dominance of artisan-mechanic inventors over university-trained ones in the early IR. Meisenzahl–Mokyr (NBER WP 16993, 2011) extend this with full biographical reconstruction of 759 great inventors 1660–1830, finding ~80% with no university education and most apprenticed into mechanical trades.
Wage-structure comparisons. Kelly, Mokyr, & Ó Gráda’s Precocious Albion (2014) argues that British skilled-to-unskilled wage ratios and skilled-worker quality were unusually high by 1700. British workers were also physically larger, healthier (a consequence of better nutrition and the protective effect of the Poor Law), and more literate than continental counterparts — the measurable human-capital substrate that underwrote the bench-of-mechanics story.
Continental adoption patterns. Early-19th-century French and German industrialists routinely imported British mechanics to build and maintain their first machinery (the so-called “British workmen abroad” pattern documented by W. O. Henderson, Britain and Industrial Europe 1750–1870). This is hard to explain on a pure factor-price account but exactly what the human-capital story predicts.
County-level English evidence (Kelly, Mokyr, Ó Gráda 2023). The 1831 share of male employment in textiles across the 41 counties of England is regressed on pre-IR mechanical-skill supply (proxied by the 1851 Census’s “mechanics and toolmakers” aged 60–69, apprenticed in the 1790s) and on pre-IR wages. Elasticities: skill supply +2, wages −6; joint R² ≈ 0.7. Literacy, local bank density, and direct coal proximity add no significant explanatory power. Placebo industries (food, shoes, garments, woodworking) show no relationship — exactly what the thesis predicts. Real wages rose sharply in the industrializing north and declined in the previously prosperous south; the widely-cited “stagnant IR-era real wages” finding is an aggregation artifact of opposite-moving regions. Primary-source texture: Watt wrote to his partner Roebuck that “the principal hindrance to erecting engines… is always smith-work”; Richard Cobden’s remark that British manufacturing greatness “grows out of the skilled labour of the men working in metals” is the contemporary framing the paper revives.
Continental human-capital evidence. Squicciarini–Voigtländer (2015) on French Encyclopédie subscriptions: upper-tail human capital predicts post-1750 city growth and post-1800 manufacturing wages; mass literacy does not. Within-France variation, with instruments for endogeneity.

Major critiques

Selection on the outcome. Britain industrialized; therefore, in retrospect, Britain’s craftsmen look exceptional. Quantitative comparisons of pre-industrial artisan density across European countries are hard and contested. The Squicciarini–Voigtländer continental result helps because it does not rely on Britain at all, but it remains the case that the leading evidence (the Meisenzahl–Mokyr inventor dataset, the Humphries apprenticeship reconstruction, the Kelly–Mokyr–Ó Gráda county tests) is overwhelmingly British.
Which way does the causation run? Did dense skilled-craftsman populations cause industrialization, or did proto-industrial demand (textiles, mining, naval supply) for skilled labour build the population? The structural answer is “both, mutually,” but that softens the causal claim into a co-evolution story whose start-date and originating push are hard to pin down. The Kelly–Mokyr–Ó Gráda IV strategy (instrumenting mechanical skill stocks with mid-18th-century Watchmakers’ Company apprenticeship fees, 1700 population density, and distance to coal) addresses the within-England version; the cross-country version has no comparable instrument.
What about France? France had Vaucanson, the École des ponts et chaussées, an excellent technical-engineering tradition. Why didn’t this bench produce French industrialization? Defenders point to the French bench being more state-employed and less commercial, and to apprenticeship being more guild-restricted — but these are post-hoc. The Squicciarini–Voigtländer within-France result mitigates this by showing that where in France the upper-tail human capital was concentrated, those départements did industrialize earlier — suggesting the cross-country level may be aggregating heterogeneous regional dynamics.
Allen’s response. Within-country tests for the high-wage hypothesis are not the only level at which the thesis runs; cross-country differences in factor prices may still drive cross-country differences in machinery substitution. Allen also disputes the skill-composition adjustment in Precocious Albion as overcorrecting, arguing London building craftsmen were comparable to continental counterparts in the relevant senses. The dispute is technical (regression specifications and skill-content adjustments) and unresolved.
From the Industrial Enlightenment school. The bricolage story understates how often workshop tinkerers were embedded in (or interacting with) scientifically literate networks — the Lunar Society’s binding of Boulton and Watt to Priestley and Erasmus Darwin is the textbook case. The two stories may be complementary halves of one, not alternatives. Mokyr’s own work treats them as such; the disaggregated form of the thesis (the bricolage component is empirically distinguishable from the propositional-knowledge component) is more contested.
The biological-environmental confound. British workers were larger and healthier than continental counterparts; some of what Precocious Albion attributes to skill may be physical productivity from better nutrition (the Komlos heights literature). The two are mutually reinforcing — better-nourished workers complete apprenticeships more reliably — but the decomposition is not clean.

Status

Mainstream as a partial story; gaining ground over the past decade. The upper-tail framing has become harder to ignore in any synthesis since the Kelly–Mokyr–Ó Gráda critiques of Allen’s wage data and the Meisenzahl–Mokyr reconstruction of the great-inventor population. It sits naturally as the supply-side counterpart to the high-wage demand-side story and as the artisanal complement to Mokyr’s epistemic story. The most productive recent work — Squicciarini–Voigtländer’s continental tests, the Minns–Wallis apprenticeship reconstruction, and the JPE 2023 within-England test — is moving the upper-tail thesis from a synthetic argument toward a quantitatively testable framework. Where it still struggles is at the cross-country comparative level: explaining why Britain rather than the Dutch Republic or France produced the deepest artisan bench remains underspecified.