Global, endogenous, unique, evolutionary: Sleuthing the Industrial

Global, endogenous, unique, evolutionary:
Sleuthing the Industrial Revolution
Gerald Silverberg
UNU-MERIT & IIASA
Explaining Economic Change
Workshop
Rome, November 12, 2014
© 2014 by Gerald Silverberg
Can an Experimental Methodology Be
Applied to a Historical Science?
• How is economic change usually “explained”
using theory-free quantitative analysis? We
regress dependent variables on independent
variables using a set of independent
observations:
Yi  f ( Xi )  i
where the Xi can be exogenous variables or earlier
observations of the Y variables.
Why does this formalism break down
in a historical discipline?
• There are few true exogenous variables, such as geography,
geology and climate
• The differentiation between independent and dependent
variables breaks down because of endogeneity
• There are few if any independent observations of the same
unchanging relationship . One usually assumes that distinct
spatio-temporal observations are independent if they are
widely separated in time or space, but in a globalized evolving
world this assumption can be violated even if interactions
appear weak
• → everything is dependent on everything else, both in space
and time. The world economy is a single, historically unique
evolving entity. Can we nevertheless make sense of it
analytically?
Example:
Acemoglu/Johnson/Robinson 2005
Regress degree of urbanization against exogenous variables (Atlantic
trading/seacoast) and some endogenous variables at earlier time such as
proxies for institutions (property rights)
A Theory of History from a SystemsTheoretic Perspective
dX / dt  fP( X )   ( X , t )   ( X , t )
where
X is a vector of endogenous variables, P exogenous parameters;
f is the nonlinear function incorporating all interactions between
the endogenous variables (in all directions;
ε is a random variable reflecting contingencies: weather, cultural
shocks (innovations, ideas, religions), outcomes of battles,
assassinations, etc., assumed exogenous but possibly influenced by
state variables;
δ is a random variable reflecting collective agency, whether arrived
at by the whim of an absolutist ruler or by the play of interests, such
as decisions about taxes, tariffs, war, alliances, etc. Individual
agency is incorporated collectively in the function f.
So What Are AJR Doing?
• mapping subsets of exogenous and
endogenous variables at time one to subsets
of endogenous variables at time two, and then
usually only linearly
• from what we know about dynamical systems,
it is very unlikely that this will pick up stable
causal relationships generated by the
nonlinear stochastic “world evolution”
operator
Forensic Approach
• Line up suspects
• Eliminate those with good alibis or contradictory evidence
Our chief suspect:
fear of technological unemployment prevented widespread adoption of
labor-saving mechanization in quasi-Malthusian closed economies, but was
overcome in late 18th c Britain in a world-historical breakthrough
circumstantial evidence:
• something changed that was unique in the situation of GB in the late 18thc.
• not coincidentally comes after GB moves to core of emerging core-periphery
world economic structure, replacing the Dutch Republic in 18th c. in entrepôt
function with productivity rising faster than wages due to backward linkages
to industry
• high price-elasticity of cotton exports (slave/Atlantic trade) more than
compensates for productivity increases by stealing world market from
international competitors (primarily India), thus overcoming fear of
technological unemployment
The Mystery of Mechanization
Left: Antikythera mechanism (1st c. BCE) for astronomical calculations displays
complex and precise gearing
Right: Bologna water-powered silk-throwing machinery (13th century CE)
Other examples: 1st c CE Hero of Alexandria, 9th c CE Arab astronomical
instruments and hydraulic mechanisms, 13th c CE Chinese hemp-spinning
machinery
Cotton-Spinning Revolution
Water Frame 1769, Mule 1779
“one thing that all these textile machines have in
common ... : they incorporated no principles,
materials or processes that would have puzzled
Archimedes” (Cardwell 1994, The Fontana History of
Technology, 185– 86)
Anti-Innovation Bias of Preindustrial Societies
(particularly wrt labor-saving innovations)
Queen Elizabeth I to William Lee, inventor of
stocking frame, 1600, in denying him a patent:
Thou aimest high, Master Lee. Consider
thou what the invention could do to my
poor subjects. It would assuredly bring to
them ruin by depriving them of
employment, thus making them beggars.
(Lee then moved to Rouen, France, where he
was enthusiastically granted a patent and
prospered. So much for the ‘extractive’ theory
of French absolutism, a side of the story
concealed in the Acemoglu/Robinson Why
Nations Fail? account.)
Machine-Breaking and Legal Restrictions on LaborSaving Innovation Typical of Preindustrial Societies
• Riots and restrictions on spinning wheel (3x
productivity enhancement) in 13th c.
• Denial of English patent to William Lee
(stocking frame), 1600, misfortunes of John
Kay (flying shuttle), 1740s
• Labor-saving rarely listed as patent grounds
before late 18th c (MacCleod)
• Frequent outbreaks of machine breaking
and mob violence against IR innovations
• Last upwelling of machine breaking 1812,
Luddite insurrection, requiring more troops
to suppress than deployed against Napoleon
Machine Breaking and Property-Rights Protection (Clark 2007)
Table 12.2 Gains from Innovation in the Textile Sector during the Industrial Revolution
Innovator
Device
Result
John Kay
Flying shuttle
James Hargreaves Spinning jenny
Richard Arkwright Water frame
Samuel Crompton Mule
Ed. Cartwright
Power loom
Eli Whitney
Cotton gin
Richard Roberts
Self-acting mule
1733 Impoverished by litigation to enforce patent.
House destroyed by machine breakers in 1753.
Died in poverty in France.
Patent denied. Forced to flee by machine
breakers in 1768. Died in workhouse in 1777.
Worth £0.5 million at death in 1792. Most of his
fortune was made after 1781, when other
manufacturers stopped honoring his patents.
(Also fled mob violence, courts overturn patent.)
No attempt to patent invention. Granted £500 by
manufacturers in the 1790s. Granted £5,000 by
Parliament in 1811.
Patent worthless. Factory burned by machine
breakers in 1790. Granted £10,000 by Parliament
in 1809.
Patent worthless. Made money later (United
States) as a government arms contractor.
Patent revenues barely covered development
costs. Died in poverty in 1864.
Two IRs During the 18th c. Must Be
Differentiated, That Only Converge in 19th c.
• The Coal-Steam Engine-Iron Complex (CIS)
– longer history from 13th c (Newcastle seacoal)
– more science based (Torrecelli, Huygens, Papin,
Hooke)
– resource enhancing
• The mechanization of cotton spinning (MOL)
– rapid takeoff (Pauli and Wyatt 1740sHargreaves/Arkwright/Crompton)
– more ‘tinkering’ based, not science based
– labor saving (by a factor of 10x-1000x)
MOL Initially Develops Independently from CIS: Steam only
overtakes water power in 1830 in GB, in 1880 in USA (Allen
2009 p. 173, Hunter 1985)
10000
1000
steam
100
water
wind
total
10
Stationary Power Sources in GB
1.0
1
1810
1860
0.8
1910
0.6
steam
fraction
1760
water
wind
0.4
0.2
0.0
1760
1780
1800
1820
1840
date
1860
1880
1900
1920
Returning to a “Systems” perspective, there has been a
remarkable convergence of four research trajectories to a coreperiphery model of the world economy/polity
• Neo-Marxist, world-systems approach (Wallerstein 1974-89,
Frank 1978, Arrighi 1994, Inikori 2002)
• (Maritime) hegemony school (Mahan 1890,
Modelski/Thompson 1988, Doran 1980, Kennedy 1989)
• New economic geography/new trade theory (Krugman 1991,
Krugman/Venables 1995, Fujita/Krugman/Venables 1999,
Baldwin/Forslid 1997)
• Network analysis of world trade network & product spaces
(Fagiolo et al. 2007, Hausmann et al. 2007)
Krugman/Venables 1995: Core-Periphery Bifurcation
as a function of agglomeration/trade cost ratio
International Trade Network (Fagiolo 2009)
Product Space Network (Hidalgo et al. 2007)
500 years of seapower hegemony
(‘trade follows the flag’)
threshold = 0.25
8
7
6
country
5
4
3
2
1
1494
1522
1550
1578
1606
1634
1662
1690
1718
1746
1774
1802
1830
1858
1886
1914
1942
1970
1998
0
Left: Dating of hegemonic epochs according to the HI ≥ threshold criterion, using actual
largest naval power in that year. The height of the bars codes for the hegemon as
follows: 1=England 2=France 3=GB 4=NL 5=Portugal 6=Spain 7=USA. Right: Herfindahl
concentration index of seapower (Silverberg 2006).
Can We Derive Some General Mechanisms for
the Capitalist Mercantilist Period?
• ‘Trade follows the flag’: seapower critical in seizing trade
advantages and locking out rivals
• ‘The tail wags the dog’: the search for extra-profits drives
economic activity out of proportion to contribution to total
product (rightly or wrongly). Partly explained by fiscal
limitations of state (kings could only levy tariffs on foreign
trade and excise taxes on select commodities, not on land)
• Strong agglomeration economies to entrepôt trade and
industrial districts
• MOL acceptable in context of import substitution and
infant industries (France, USA) or conquest of export
markets (GB), if not too damaging to domestic incumbents
So we can now return to the origins of the
Industrial Revolution by identifying some
‘smoking guns’:
• Why cotton (and not wool, silk, linen or
anything else)?
• Why Britain and not the Netherlands (or
France)?
• Why Lancashire?
• Cotton is an almost non-existent industry in 18th c England – Wool had
always been the dominant industry. In 1750 wool industry 10 times
larger than cotton. If factor prices and size of industry were
determinants of inventive effort, wool should have been the likely
candidate for mechanization. No a priori reason to believe wool
mechanization more difficult than cotton (contrary to Landes and
Mokyr)
• England is intermediate protectionist case in cotton textiles after 1701
(imports except for printing/dyeing processing for reexport
prohibited, domestic production restricted to ‘fustians’ (an inferior
product): linen warp and cotton weft). Instance of collective agency
with fortuitous consequences. France is so protectionist that it
eliminates preexisting cotton textile capabilities, while NL is free
trade, impeding emergence of domestic industry
• India dominates world production in quality and price of full-cotton
textiles at all quality levels
• Growing market in Atlantic economy (slave/triangular trade), but
England cannot fully supply it in printed cottons nor in the light
woolens in demand (slave ships stock up in Holland on way to Africa).
Why Cotton?
“The potential market for the successful innovator in
cotton is a more satisfying explanation for why technical
change in the eighteenth-century European textile
industry centered on cotton than the well-worn story
that cotton was easier to work and therefore, mechanize
than wool. If cotton was easier to work with, we must
also ask why it was so difficult for Europeans to spin
good-quality warp yarn by hand.”
Prasannan Parthasarathi, 2011, Why Europe Grew Rich
and Asia Did Not: Global Economic Divergence, 16001850, p. 288
 high price elasticity of cotton compared to
saturated European woolen market due to import
substitution and export-led growth (Inikori 2002)
 major obstacle to labor-saving mechanization
broken: worker opposition, fear of technological
unemployment (Ludditism before the Luddites)
(even land-owning classes feared technological
unemployment due to Poor Law tax assessments
and social unrest)
All macroinnovators (Hargreaves, Arkwright,
Crompton, Cartwright) had their workshops and
factories destroyed, and often had to relocate to
escape initial mob violence
• Reverse side of coin is India’s loss of export
markets in 18th c and loss of domestic market in
19th c.
• Conjecture: total global employment in cotton
declines sharply even while it explodes in GB.
• Principle lessons of IR:
– core countries can enjoy employment and productivity
growth simultaneously
– catch-up countries must industrialize or become
peripheralized/deindustrialize
– protectionism and mercantilism pay
Why GB and not the Netherlands?
• Pure factor price argument fails here: according to
Allen’s own data, factor prices (wages, fuel, interest
rates/capital costs) are remarkably comparable
• NL loses maritime hegemony and control of the
Atlantic economy (and even much of Asian trade) to
GB between the Anglo-Dutch wars (starting 1651) and
the Treaty of Utrecht (1713). Also increasingly
threatened by an aggressive France
• As a result, GB succeeds in imposing its mercantilist
regime (Navigation Act, Cotton Acts), while NL is free
trade in cotton as only European country
• The Glorious Revolution, Treaty of Utrecht, and
trilateral hegemonic rivalry game (NL, GB, F) make NL
a junior maritime partner to GB, GB is free to refrain
from the expense of a standing army and concentrate
on maritime trade and power (Pincus 2009, Mahan
1896), and they go to war together against France
• NL has no significant water-power resources
(although it does have sufficient peat and in principle
coal). As I have already argued, water power is still
key to textile mechanization, not steam, and still
determines mill location decisions in many countries
(US, France, Germany, Switzerland) until the mid1800s.
Why Lancashire?
• Lancashire is by all accounts an absolute backwater in early 18th
c England except for rising importance of Liverpool in slave
trade
• Wages in 1750 around 2/3 that of London (Deane 1965, Hunt
1968). Highly elastic labor supply due to rural migration and
Ireland immigration
• Does not enjoy “pithead” coal prices until major canals are built
in the 1760s
• But ample water power!
• While there is some evidence for a preexisting
watch/clockmaking industry in Lancashire, London is the world
center. 50% of London clockmakers supposedly migrate to
Lancashire during IR according to contemporary accounts
• Agglomeration economies reinforced by breakthroughs in
technology and economics of steam drive after 1790s.
Other ‘Usual Suspects’ Have Alibis
• Institutions (North/Weingeist, Acumoglu/Robinson)
– but institutions are endogenous (chicken-egg problem) and can
change rapidly (Reformation, Glorious and French Revolutions)
– Absolutist regimes often more favorable to innovation and
industrial policy than GB, though possibly less successful in
results (state-enterprise trap?)
• Science/Enlightenment Values (Mokyr)
– not clear why this eliminates decreasing returns barrier to
‘artisanal’ R&D
– refuted by data analysis of Allen
– ‘tinkering’ now accepted even by Meisenzahl/Mokyr 2012 as
dominant innovation driver in MOL
• Other suspects: Protestant Ethic, Genetic Selection...
Last Man Standing: GB Killed the Technological
Unemployment Bogeyman with Slave Trade
(Inikori), Atlantic Economy and Seapower
Hegemony (Mahan)