How Did Growth Begin?

The Industrial Revolution and its Antecedents

Jón Steinsson*
University of California, Berkeley

December 16, 2022

We live in an era of economic growth. Over our lifetimes, the lifetimes of our par-
ents, and grandparents, output per person in North America and much of Western
Europe has grown on average by roughly 2% per year. This steady growth has led
to a staggering transformation of material well-being. Ordinary workers in North
America and Western Europe earn about 15 times more than they did two hundred
years ago.
After several generations of steady progress, it may seem inevitable that eco-
nomic growth will continue throughout our life-times and the life-times of our chil-
dren and grandchildren. It is important, however, to realize that the era of rapid
economic growth that we live in is a very recent phenomenon. Before 1750, eco-
nomic growth was less than one tenth as rapid as it is today; and before 1500, eco-
nomic growth proceeded at a truly glacial pace (as far as we can tell using current
historical knowledge).
Our species has dominated the earth for thousands of years. Massive empires
have risen and fallen. But over the millenia before 1500, the material well-being of
ordinary workers changed very slowly if at all. Then, in a blink of an eye (from a
long-term historical perspective), economic growth increased from close to zero to
* I would like to thank Sungmin An, Zhaosheng Li, Sergio Nascimento, Shen Qui, and Daniel
Reuter for excellent research assistance. I would like to thank Charles Angelucci, Robert Allen, Gre-
gory Clark, Anton Howes, Ethan Ilzetzki, Reka Juhasz, Alain Naef, Suresh Naidu, Emi Nakamura,
Nuno Palma, Gary Richardson, Paul Schmelzing, Hans-Joachim Voth, David Weinstein, and Ludger
Woessmann for valuable comments and discussions. I would like to thank Robert Allen, Gregory
Clark, Paul Schmelzing, and Ludger Woessmann for generously sharing data and replication pro-
grams. I would like to thank Mikel Acha, Joshua Archer, Caroline Dority, Jesse Flowers, Jocob Gold-
stein, Chazel Hakim, Brian Hill, Henry Isselbacher, Benjamin Nhan, Kevin Wang, and David Zhou
for finding errors and typos. First posted in April 2020.

1
a modern rate of 2% per year. This dramatic change is what we call the Industrial
Revolution.
The Industrial Revolution first occurred in Britain in the 18th and 19th centuries,
but its roots extend back considerably in time and to other parts of the world. In this
chapter, I seek to shed light on why the Industrial Revolution occurred, why it oc-
curred in Britain, and why it occurred in the 18th and 19th centuries. Unfortunately,
there is no consensus among scholars regarding these important questions. To the
contrary, this is a topic of active, lively, and sometimes rather contentious debate.
Given this state of scholarship, I will not try to present a single narrative about
the causes of the Industrial Revolution. Instead, I will discuss four major strands
of thought: changes in institutions, the Enlightenment, advances in agriculture, and
the role of high wages and cheap coal. While the advocates of each of these strands
of thought are sometimes quite critical regarding the importance of the others, my
view is that all four of these strands (and others) emphasize important develop-
ments that likely played an important role in bringing about the Industrial Revolu-
tion.
When it comes to big questions such as why the Industrial Revolution occurred,
there are proximate causes and deeper causes. Proximate causes are typically easier
to identify than deeper causes. In the case of the Industrial Revolution, important
innovations such as the spinning jenny, the steam engine, and improvements in agri-
cultural yields are obvious proximate causes. Once the proximate causes have been
identified, they typically beg the deeper question of why they occurred: Why did
all these innovations occur in Britain at this time? Perhaps they occurred because
of favorable institutions in England or because literacy was widespread. This then
begs the question of why England had those characteristics. In this way, the quest to
understand why the Industrial Revolution occurred is like peeling an onion: there
are countless deeper layers to uncover. In this chapter, I discuss both proximate and
deeper causes. I therefore try to peel the onion at least a few layers in. Inevitably,
the deeper one gets the more speculative the discussion will become.
I will not provide a narrative history of the Industrial Revolution and other his-
torical develpments discussed in this chapter such as the Enlightenment and the
political upheaval in 17th century England. Several of the books that I cite in the
chapter provide excellent narrative accounts of these events. Let me in particular
recommend The Industrial Revolution by T.S. Ashton (1948/1997), The Lever of Riches
by Joel Mokyr (1990), The English Revolution of 1640 by Christopher Hill (1940), and
The Reformation by Peter Marshall (2009).

2
 Real Wage
110
1860
100
1450
90

80

70

60
1800
50 1650

40 1300

30

20
0 5 10 15 20
Population (millions)
Figure 1: Real Wages and Population in England from 1300 to 1860
Note: This figure replicates Figure 5 in Clark (2005). The real wage and population series are up-
dated series from Clark (2010). The real wage series is an index scaled to be equal to 100 in the
1860s.

1 Institutions
Figure 1 plots the evolution of real wages and the size of the population in England
over the period 1300 to 1860. In 1300, the population of England was about five and
a half million. Over the next 150 years, the population dropped by more than 50%
to about two and a half million as plagues ravaged England. The largest of these
plagues was the Black Death of 1348. As the population shrank, real wages more
than doubled. Over the subsequent 200 years, the population of England recovered,
and, as it did, real wages fell. In 1640, both real wages and the size of the population
in England were almost identical to their levels 340 years earlier.
As I explain in greater detail in chapter XX [Malthus chapter], the evolution of
real wages and the population in England between 1300 and 1640 strongly suggests
that there was no productivity growth in England over this period. The English
economy seems to have moved up and down a stable labor demand curve over this
period with the movements along the curve being caused first by plagues and then

3
by recovery of the population as the plagues subsided.1 The Malthus model dis-
cussed in chapter XX [Malthus chapter] captures these dynamics well. The easiest
way to see why productivity growth was close to zero over this period is that real
wages and the population returned to virtually the same point in 1640 as they had
been at in 1300. If productivity had increased over the intervening 340 years, the
labor demand curve in England would have shifted outward. In this case, the econ-
omy would not have been able to return to its earlier state since that state would no
longer have been on the labor demand curve (it would have been below and behind
the new curve). The fact that the economy returned to the same point in 1640 as
it was at in 1300, thus, implies that the labor demand curve did not shift over this
period. This, in turn, implies that there was no productivity growth over this 340
year period.
Figure 1 suggests that something important changed in the English economy in
or around the 1640s. The point for the 1650s is the first point that is clearly off the
labor demand curve that the earlier points seem to be moving along. After 1650,
the points keep moving further and further off the earlier curve. This indicates that
around 1650 productivity started to rise in England. At first, this growth was slow.
But in the late 18th century and especially after 1800, growth seems to accelerate a
great deal. Figure 1, thus, suggests that the 150 year period between 1640 and 1800
was a period of transition from stagnation to modern growth in England. By 1800,
the Industrial Revolution was in full swing and the growth rate of the economy was
staggeringly large relative to anything experienced before.
The timing of the onset of growth suggested by Figure 1 is quite intriguing. The
1640s are not just some random decade in the history of England. This is the decade
of the English Civil War between Parliamentarians and Royalists. This armed con-
flict ended with the overthrow (and execution) of Charles I and the establishment of
a Commonwealth in 1649. The monarchy was later restored in 1661, but it was then
overthrown again in the “Glorious Revolution” of 1688.
On the one hand, it is perhaps surprising that the onset of growth seems to have
occurred at a time of armed conflict. But the English Civil War and Glorious Revolu-
tion brought about sweeping institutional change in England that may have played
an important role in creating conditions conducive to economic growth. An influen-
tial argument along these lines was put forth by Douglas North and Barry Weingast
(1989).

4
1.1 The Sovereign’s Commitment Problem
Risk of expropriation has been an important barrier to growth throughout history.
Most societies have had weak states and therefore faced a constant risk of invasion.
Members of such societies have also faced risk of coercion and expropriation from
stronger members within the society. A strong state can reduce the risk of invasion.
It can also enforce order within its borders and thereby reduce general lawlessness.
In a strong state, however, a different problem arises in the form of predatory behav-
ior by the state itself—usually the sovereign and members of his elite (see Acemoglu
and Robinson (2019) for an extensive discussion).
An important downside for a sovereign of acting in a predatory manner is that
the risk of predation by the sovereign reduces the incentives of his subjects to make
investments that foster economic growth. No one will invest in improving their
land, building a waterwheel, or inventing a new technology if they expect that these
investments will be expropriated by the sovereign or members of his elite. Even the
most self-interested ruler may, therefore, desire to restrain his predatory powers and
instead establish secure property rights accompanied by moderate and predictable
rates of taxation. This strategy will encourage investment, help foster economic
growth and growth in the sovereign’s tax base. It may therefore yield higher tax
revenue for the sovereign in the long run than unrestrained despotism.
North and Weingast point out a fundamental difficulty faced by a sovereign who
would like to establish secure property rights in order to encourage growth: “the
sovereign or government must not merely establish the relevant set of rights, but
must make a credible commitment to them.” In other words, it is not enough to re-
spect property rights in the short run. The sovereign must be able to get his subjects
to believe that he is committed to respecting their property rights in the future as
well. Why might this be particularly difficult? Suppose the sovereign promises to
respect the property rights of his subjects and the subjects believe that the sovereign
will honor this promise both in the near term and further in the future. In this case,
the subjects will think that the returns that they can earn from making investments
are high. They will therefore invest and this will result in increased wealth cre-
ation. The problem is that, as the subjects amass wealth, the sovereign’s temptation
to renege on his earlier promise and confiscate their wealth grows. Eventually this
temptation may become irresistible (especially in a time of war when fiscal needs
are acute).
This problem is called a commitment problem. In a nutshell, it arises from a conflict

5
between what is in the sovereign’s interest at one point in time and what is in his
interest at a later point in time. For simplicity, let’s consider a sovereign that gov-
erns for two periods. We will refer to the first period as “ex ante” (“ex ante” is Latin
for “before”) and the second period a “ex post” (“ex post” is Latin for “after”). We
choose these names for the periods in reference to the fact that the sovereign’s sub-
jects will make an important choice (invest or not invest) between these two periods.
Ex ante, it is in the interest of the sovereign to promise that he will respect property
rights for the entire time he is in power. If this promise is believed, it provides
his subjects with incentives to invest, increase production, and thereby increase the
sovereign’s tax base. However, ex post, once the subjects have amassed wealth, it
is in the sovereign’s interest to renege on his earlier promise and expropriate his
subjects’ wealth.
The important consequence of this logic is that the subjects should not believe
the sovereign’s initial promise. If the subjects can understand the incentives of their
sovereign, they should be able to anticipate that the sovereign will have an incentive
to renege on his earlier promise once they have amassed wealth. This means that
they should not believe the initial promise. In this case we say that the sovereign’s
promise is not credible.
Commitment problems are common in a variety of settings. For example, par-
ents face a never ending string of commitment problems. A parent may say to their
child: “If you don’t finish your dinner, you won’t get dessert.” But the child may un-
derstand that this is not a credible threat. Once dinner is over, the parent’s willpower
to enforce the punishment as their child screams and cries and others at the table
get dessert may be limited. An important macroeconomic example of a commit-
ment problem is the problem of a government trying to regulate its banking system.
The government may say to the banks: “If you get into trouble, we won’t bail you
out.” But the banks may understand that this is not a credible threat. If the banks
do get into trouble, the prospect of letting them go bust may be too costly for the
government to follow through on its earlier threat.
Commitment problems can be solved in two ways. The first is through good
reputation in cases where interactions are repeated. The sovereign may build a repu-
tation over time for honoring property rights. This may eventually convince his sub-
jects that his promises are not empty. In these cases, the sovereign has an incentive
to honor his earlier promises so as to maintain his reputation because having a good
reputation benefits him in the future. Conversely, when interactions are repeated,
reneging on earlier promises has the cost that it tarnishes the sovereign’s reputation

6
and thereby leads to lower investment and tax revenue in the future. These long-run
costs may imply that expropriation is not worth it even in cases where the short-run
benefits are substantial. (Similarly, parents may endure a painful evening in the
hope that this will convince their children to listen to them in the future.) For repu-
tation to be an effective check on the sovereign’s temptation to renege, the sovereign
must value future tax revenue substantially (i.e., he can’t be too impatient and he
can’t be too worried about being overthrown). These ideas have been formalized us-
ing game theory (see, e.g., chapter 2.3 of Gibbons, 1992, on repeated games). North
and Weingast argue, however, that sovereigns have in practice seldom been able to
make credible commitments to refrain from arbitrary wealth expropriation through
reputation alone.
The second way to solve commitment problems is to create commitment devices.
Odysseus famously tied himself to the mast. Cortez began his campaign to conquer
Mexico for Spain in 1519 by ordering his men to burn the ships they arrived on.
In both cases, these acts were meant to constrain potentially destructive future be-
havior. In other words, they were devices meant to commit Odysseus and Cortez’
army to act a certain way which they might not see as in their best interest later on.
The most common form of a commitment device in modern societies is a contract.
Contracts commit the parties involved to perform actions at later dates that they
may not see as being in their interest at that point (e.g., to pay for a car that a com-
pany has already delivered). The enforcement of private contracts is an important
function of modern governments.
Constitutions are perhaps the most important commitment devices of our times.
The founders of the United States worried about possible tyranny by a strong exec-
utive. To guard against this risk, they wrote the basic structure of the government,
including the separation of powers and Congress’s control over the power of taxa-
tion, into a constitution and made this constitution quite difficult to amend. Calls
for even greater constitutional protection of liberties then led to the passing and
ratification the Bill of Rights: ten amendments to the U.S. Constitution that guaran-
tee certain basic liberties and place specific limits on government powers. The fact
that it is so difficult to change the U.S. Constitution is an important check on U.S.
Presidents that have authoritarian tendencies.
North and Weingast argue that the Glorious Revolution of 1688 created an insti-
tutional structure in England that credibly constrained the sovereign from arbitrary
and confiscatory expropriation of his or her subjects’ wealth. They describe several
ways in which the English Crown expropriated its subjects in the first few decades

7
of the 17th century. These included “forced loans” (i.e., loans granted under duress)
that were not repaid on time or in full, the sale of monopolies in settled industries
(with serious adverse consequences for incumbents in the industry), the sale of pub-
lic offices such as seats in the House of Lords (which reduced the value of existing
seats), the use of purveyance power (seizure of goods for “public purposes” with-
out full compensation), the sale of “dispensation” (i.e., threatening subjects with
arbitrary enforcement of otherwise lightly enforced regulations unless they paid a
bribe), and outright seizure of property. The English Crown was by no means alone
in employing these tactics. For example, the Kings of France employed many of the
same tactics to raise revenue.
At this time, the House of Commons was mainly made up of the wealthiest seg-
ment of the English population. It was exactly these segments of the population
that bore the brunt of the Crown’s expropriation. Parliament sought to constrain
the Crown in various ways, but with little success. Eventually, disputes between
Parliament and the Crown lead to the outbreak of civil war in England in the 1640s
in which the Parliamentarians eventually prevailed. The economic disputes I have
described were one major cause of the English Civil War. But other disputes also
played an important role, such as religious disputes between Catholics and Protes-
tants. North and Weingast argue that the Parliamentarians in England would have
been unlikely to prevail in the Civil War had the Crown had a standing army like its
counterparts on the continent.
The English Civil War lead to dramatic institutional change. Parliament abol-
ished the Star Chamber and passed an act requiring all legal disputes involving
property to be tried in common law courts. Before the Civil War, the Star Cham-
ber had various powers including final adjudication of disputes concerning royal
prerogative (i.e., royal decrees). The fact that the Crown controlled the Star Cham-
ber meant that Crown’s subjects had no independent court to turn to in the event
of disputes with the Crown. A second important institutional change was that the
Statute of Monopolies—which had been passed by Parliament in 1624—was now
enforced. This statute prohibited the creation of monopolies by sale of patents to
existing businesses. Third, institutional changes having to do with labor mobility
and land tenure overthrew the feudal organization of rural society in England. I
will discuss this last set of institutional changes in more detail in section 3.
A major risk in revolutions is that one tyrannical government will simply be sup-
planted by another. History is replete with examples of this. Arriving at a form of
government that is inclusive in terms of political and economic rights is evidently a

8
delicate balance. In the half-century or so before and after the outbreak of the Civil
War, England experienced several different forms of government. It was not at all
clear during this period that the final outcome of the revolution would be a gov-
ernment that could make a credible commitment to secure property rights for more
than a small elite. First, there was the period of autocratic personal rule by Charles
I after he dissolved Parliament in 1629. Then Parliament revolted and civil war en-
sued in the 1640s. Parliament prevailed in this war and set up a Commonwealth
which eventually devolved into a military dictatorship of Oliver Cromwell—the
Protectorate. During this period, radical elements of the Parliamentary coalition
(the Levellers) were violently suppressed. After Cromwell’s death in 1658, the Par-
liamentary coalition sought stability by restoring Charles II to the throne in 1660.
Charles II ruled largely with the consent of Parliament. But in the 1680s his suc-
cessor James II, sought to reestablish the Crown’s supremacy in part by issuing new
charters for seats in Parliament and thereby packing Parliament with members loyal
to him. His first such move was against his traditional opponents, the Whigs. But
he then overstepped decisively by moving against his traditional allies, the Tories.
This drove the Tories into opposition and lead to generalized revolt of Parliament
against the Crown.2
In the Glorious Revolution of 1688, Parliament overthrew King James II and re-
placed him with William of Orange (with the help of an invading Dutch army).
More importantly, though, the Revolutionary Settlement and the Declaration of
Rights that followed clearly established Parliamentary supremacy in England. The
Crown’s powers were severely limited after this point and would subsequently fur-
ther dwindle until the Kings and Queens of England became mere figureheads in
the 20th century. After the Glorious Revolution, Parliament had exclusive author-
ity to raise taxes and to approve the government’s budget. The Crown no longer
had royal prerogative powers, implying that legislative powers were solely in the
hands of Parliament. Judges no longer served at the pleasure of the Crown. Rather,
they could only be removed by act of Parliament. This meant that the judiciary was
independent of the Crown.
The Glorious Revolution (and the institutional changes going back to the Civil
War) dramatically increased the political power of Parliament in England. Parlia-
ment was dominated by the rich and the nobility. Why did this relatively small elite
group not establish its own form of tyranny over the rest of the population as had
happened, for example, in Hungary (Fukuyama, 2011, ch. 25)? North and Wein-
gast provide two potential explanations for this. The first is that the institutional

9
structure in England after the Glorious Revolution consisted of a delicate balance
between the Crown, Parliament, and the common law courts. Acts of expropriation
by either the Crown or Parliament were checked by the other two actors. The sec-
ond potential explanation discussed by North and Weingast is that the interests rep-
resented in Parliament were sufficiently diverse to place severe limits on schemes
meant to enrich one group at the expense of the rest of the population. In par-
ticular, Whigs represented the newly emergent merchant class as well as the more
progressive parts of the gentry. These groups opposed widespread regulation of the
economy. On the other hand, Tories represented the more conservative segments
of the landowning class, as well as the nobility. Together, the Whigs and the Tories
in Parliament, therefore, represented a relatively diverse set of interests, which may
have meant that it was difficult for Parliament to create extractive institutions.
One dramatic development that North and Weingast point to as evidence of a
change in the British government’s ability to make credible commitments after the
Glorious Revolution is the sharp increase in government borrowing and the im-
provement in the terms at which the government could borrow. These develop-
ments are often referred to as a Financial Revolution (Dickson, 1967). Figure 2 plots
the evolution of the ratio of British government debt to British GDP over the pe-
riod 1691 to 2015. Before the Glorious Revolution, the English Crown was able to
borrow only small amounts (less than 5% of English GDP). By 1700, British govern-
ment debt had risen to 25% of GDP, mostly due to military spending relating to the
Nine Years War with France. After 1700, British government debt continued to rise
dramatically, to well over 100% of GDP in the early 19th century as Britain fought
several large wars.
Figure 3 plots an interest rate series for government debt in England from 1530 to
1850 recently compiled by Schmelzing (2020). This figure shows a relatively steady
fall in the interest rates the English Crown faced throughout the 17th century. In-
terest rates spiked temporarily during the turmoil of the English Civil War and the
Nine Years War. But other than that, they are on a steady downward trend. The
Glorious Revolution is not an obvious break point. Yet, it is impressive that interest
rates continued to fall even as the quantity of British government debt outstanding
expanded massively in the 18th century. The interest rates the British government
paid on this massive debt in the middle part of the 18th century were remarkably
low by historical standard at around 3%. Even when the British government be-
came extremely indebted in the late 18th and early 19th centuries, the interest rates
it faced only rose to about 5%.

10
 Debt/GDP
3

2.5

1.5

0.5

0
1690 1715 1740 1765 1790 1815 1840 1865 1890 1915 1940 1965 1990 2015
Figure 2: Government Debt to GDP for Britain from 1691 to 2015
Note: This figure plots the nominal value of British government debt divided by nominal GDP.
Both series are taken from Thomas and Dimsdale (2017). The nominal GDP figures are based on
the work of Broadberry et al. (2015).

A key reason why the British government was able to raise so much debt at
such favorable rates was its ability to collect a steady and increasing stream of taxes
(O’Brien, 1988). The ability of the British government to finance large wars with debt
in the 18th and 19th centuries was a major advantage vis-a-vis its main adversaries,
notably France. This meant that it could avoid imposing ruinous taxes during war
time and instead pay for wars over a longer period of time. Arguably, the ability
to finance wars by issuing large amounts of debt was a key contributor to Britain’s
success in the wars of the 18th and 19th centuries, and, therefore, its emergence as a
dominant naval power and major empire.
Developments in private capital markets in the decades after the Glorious Rev-
olution were also dramatic. The Bank of England was founded in 1694 as a profit-
oriented private bank. Early on, its main activities were to assist the government in
finding buyers for its massively increasing debt and to issue bank notes that were
used as a means of payment for large transactions in London. The 18th century saw
an explosion of banking in England, initially in London, but later on throughout

11
 0.16

0.14

0.12

0.1

0.08

0.06

0.04

0.02

0
1530 1570 1610 1650 1690 1730 1770 1810
Figure 3: Interest Rates on Government Debt in England
Note: This figure plots decadal averages of interest rates on long-term government debt in England.
This series is constructed based on annual data from Schmelzing (2020). The interest rates are for
voluntary loans to the government. Prior to the Glorious Revolution, the crown would at times
coerce its subjects to lend it money. The terms on such forced loans are excluded.

the country. These banks provided an increasing array of services including note
issuance, issuance of travelers checks, clearing of payments, insurance, and loans
to finance real estate, infrastructure, enclosures, dowries, and consumption. This
period also saw the emergence of a wide range of financial securities which helped
finance an ever wider set of activities. Kindleberger (1993) argues that the political
revolutions in England in the 17th century contributed importantly to these financial
developments by reducing the risk of arbitrary seizure by the state of concentrated
assets such as occurred in 1640 and 1672.
North and Weingast’s argument is controversial. Clark (1996) argues that if the
Glorious Revolution substantially increased the security of property rights in Eng-
land, rates of return on capital should have fallen substantially after the Revolu-
tion and asset prices should have increased. He presents data on rates of return on
farmland and rental charges, and the real price of land from the records of English
charities. Clark points out that nothing much happened to these series around the
time of the Glorious Revolution. Figure 4 presents data compiled by Clark for the
rate of return on farmland in England. Clearly, there is no dramatic change in the

12
 10
9 Glorious
English Civil War Revolution
8
7
6
5
4
3
2
1
0
1500 1550 1600 1650 1700 1750
Figure 4: Rate of Return on Farmland in England
Note: This figure plots an estimate of the average rate of return on farmland in England from Clark
(2002, 2010). It is calculated as the ratio of rental payments on land to the price of the land.

return on farmland around the Glorious Revolution. Clark, furthermore, points out
that periods of political instability during the 16th and 17th centuries (such as the
Civil War) did not lead to much variation in these series. Clark contrasts this with
data from the town of Zele in Flanders over a similar time period. Both rates of
return on land and real land prices fluctuate a great deal in Zele in response to wars
affecting that area. Clark concludes that property rights in England were relatively
secure as early as 1600 and did not become substantially more secure as a result of
the Glorious Revolution.
Clark’s critique applies to the security of the property rights of landowners. It
may, however, have been that other types of property rights, or the rights of other
groups of people, were less secure prior to the 17th century revolutions in England.
It may, for example, have been that merchants faced less secure property rights than
landowners, and that individuals or groups that the Crown disapproved of were
more likely to be expropriated. Jha (2015) provides interesting evidence in support
of this thesis. He shows that among members of Parliament, merchants and in-
vestors in overseas joint stock companies were more likely to rebel, while proxies
for domestic wealth and ownership in domestic joint stock companies do not pre-

13
dict support for the rebellion.3 The Crown had much more scope for expropriation
of overseas activities than domestic activities because foreign trade was governed
by civil law courts, which were administered by the Crown itself, while domestic
property rights were governed by common law. Prior to the Civil War, the Crown
repeatedly engaged in expropriation of overseas companies by raising customs du-
ties on imports and revoking charters of overseas companies.
A related but quite different line of argument is that perhaps the Glorious Rev-
olution and the shift of power to Parliament ignited economic growth by allowing
property rights to be more easily reorganized and rationalized. Property rights in
medieval England (and elsewhere) involved a complex maze of interwoven rights
which complicated their reorganization in response to changing circumstances. I
discuss the nature and evolution of these rights prior to the Glorious Revolution in
greater detail in section 3. Bogart and Richardson (2011) discuss how in the 18th cen-
tury “Parliament embraced novel ideas concerning property and established proce-
dures for processing petitions from groups hoping to reorganize rights to land and
resources.” Most prominently, Parliament passed several thousand enclosure acts.
But it also passed a great number of statutory authority acts that allowed for the
construction of infrastructure such as turnpikes and canals. Britain experienced a
transport revolution during the 18th and 19th centuries which dramatically low-
ered transaportation costs and therefore allowed for expanded trade (Bogart, 2014).
This transport revolution may have in part resulted from Parliament’s willingness to
reorganize property rights and use its power of eminent domain. Rosenthal (1990)
provides an interesting counterpoint. He argues that highly profitable irrigation
projects in the French region Provence were not undertaken before the French Rev-
olution due to fragmented political authority over the right of eminent domain. In-
terestingly, this line of argument, if correct, implies that the Glorious Revolution
(and the French Revolution) ignited growth by making ancient (inefficiently orga-
nized) property rights less secure, or at least more easily renegotiable.
The Glorious Revolution and its antecedents may have been important for other
reasons as well. “Rent seeking” has arguably been a serious impediment to growth
and prosperity in most societies. Rent seeking refers to the pursuit of various forms
of privilege by favored and powerful individuals in society—such as monopoly
rights, guild protection, or other kinds of protection from competition, as well as
exception from taxation. Parliament struck down many monopolies and other types
of economic privileges in Britain during the 17th and 18th centuries. More generally,
a gradual shift in policy occurred towards free-market policies, where, as economic

14
historian Joel Mokyr has put it, “freedom” came to mean “the freedom to enter a
branch of economic activity rather than the freedom to exclude others” from that
activity (Mokyr, 2009, p. 27). These shifts likely created conditions that were more
favorable to innovation and entry by startups and, therefore, may have been an im-
portant factor in igniting economic growth.

1.2 The Role of Atlantic Trade

Was there something special about England in the 17th century that led—through a
series of revolutions and counter-revolutions—to the establishment of secure prop-
erty rights and relatively inclusive economic and political institutions? Acemoglu,
Johnson, and Robinson (2005) argue that the rise of Atlantic trade following the
Great Discoveries of Christopher Columbus and Vasco de Gama played an impor-
tant role in sparking these institutional changes. In this view, the Great Discoveries
of the 1490s were an important “root” cause of the Industrial Revolution. (The Great
Discoveries were, of course, themselves made possible by innovations in shipbuild-
ing prior to 1500.)
The starting point of Acemoglu, Johnson, and Robinson’s analysis is the notion
that “between 1500 and 1800, Western Europe experienced a historically unprece-
dented period of sustained growth, perhaps the ‘First Great Divergence”’. They
then argue that this First Great Divergence was concentrated among a small group
of countries involved in Atlantic trade. Since high quality measures of output per
person are not available for most of the countries they seek to analyze, Acemoglu,
Johnson, and Robinson use the urbanization rate as a proxy measure of output per
person. More specifically, the measure they use is the fraction of the population that
lived in cities and towns that at some point between 800 and 1800 had at least 5000
inhabitants. The logic for why the urbanization rate might be a reasonably good
proxy for output per person is that only areas with high agricultural productivity—
i.e., high agricultural output per person—could support large urban populations.
Figure 5 plots the urbanization rate for three groups of European countries from
1300 to 1800. The first group is the countries Acemoglu, Johnson, and Robinson
refer to as Atlantic traders. These are modern day Britain, France, the Netherlands,
Portugal, and Spain. The second group is other countries in Western Europe. Ace-
moglu, Johnson, and Robinson define Western Europe as the area west of the Elbe
river. The second group of countries thus includes modern day Austria, Belgium,
Denmark, Finland, Germany, Ireland, Italy, Norway, Sweden, and Switzerland. The

15
 30
Atlantic Traders
Other Western Europe
25
Eastern Europe

20

15

10

0
1250 1350 1450 1550 1650 1750 1850
Figure 5: Urbanization Rate in European Regions
Note: These data series were constructed by Acemoglu, Johnson, and Robinson (2005) from data on
urban populations compiled by Bairoch, Batou, and Chèvre (1988) and population estimates from
McEvedy and Jones (1978).

third group is Eastern Europe, which Acemoglu, Johnson, and Robinson define as
all countries east of the Elbe including Russia, but excluding modern day Turkey.
Figure 5 shows clearly that urbanization rates rose much more in the countries en-
gaged in Atlantic trade than in the rest of Europe. The urbanization rate of the
Atlantic traders rose by 150% from 8% to 20% between 1300 and 1800, while the
urbanization rate in the rest of Western Europe and in Eastern Europe only rose by
70% over this same period.
Atlantic trade—i.e., trade with the New World, the West Coast of Africa, and
Asia—yielded large profits for those skillful and lucky enough to be successful. Ta-
ble 1 summarizes Acemoglu, Johnson, and Robinson’s estimates of annual profits
from Atlantic trade in England from 1500 to 1800. The profits of £200,000 in the
first half of the 17th century were enough to make quite a few merchants and pri-
vateers very wealthy (GDP per person at the time was roughly £7). However, these
profits were not large enough to directly explain a sizable portion of the growth in
English output in the 17th and 18th centuries. In the first half of the 17th century,

16
 Table 1: Estimates of Annual Profits from Atlantic Trade in England
Before 1575 Negligible
1576-1600 40,000
1601-1650 200,000
1651-1675 500,000
1676-1700 900,000
1701-1750 1,700,000
1751-1800 5,000,000
Note: Estimates reported in Acemoglu, Johnson, and Robinson (2005). All figures
are in pounds sterling adjusted to 1600 prices using the index of craftsmen’s wages
from Phelps Brown and Hopkins (1955).

these profits amounted to only 0.7% of GDP. In the first half of the 18th century, they
amounted to 4% of GDP. In the second half of the 18th century, they amounted to
7.5% of GDP. Over this 200 year period, GDP per person in Britain grew by roughly
75%.4
While the direct effect of Atlantic trade on growth was modest, Acemoglu, John-
son, and Robinson argue that Atlantic trade had a substantial indirect effect on
growth by helping create conditions conducive to fundamental institutional change.
For such institutional change to occur, however, a second condition was also impor-
tant: Atlantic trade only resulted in fundamental institutional change in countries
that had relatively non-absolutist institutions to begin with. Acemoglu, Johnson,
and Robinson’s argument goes as follows: Atlantic trade created large profits. In
countries with absolutist initial institutions (e.g., Spain and Portugal), the sovereign
and his elite were able to dominate the trade and exclude others from it. How-
ever, in countries with less absolutist initial institutions (e.g., Britain and the Nether-
lands), Atlantic trade greatly expanded and enriched commercial interests outside
the royal circle. These groups eventually became powerful enough to demand insti-
tutional change that constrained the power of the sovereign and protected property
rights. These institutional changes were, in turn, crucial preconditions for economic
growth.
Acemoglu, Johnson, and Robinson’s theory helps us make sense of the different
growth rates experienced by different parts of Europe. The first group of countries
is Britain and the Netherlands. These countries had access to the Atlantic and had
relatively non-absolutist initial institutions. They experienced liberal revolutions in
the 16th and 17th centuries, which established more inclusive economic and political

17
 45
Britain and Netherlands
40
Spain, Portugal, and France
35

30

25

20

15

10

0
1250 1350 1450 1550 1650 1750 1850
Figure 6: Urbanization Rates Among Atlantic Traders
Note: These data series were constructed by Acemoglu, Johnson, and Robinson (2005) from data on
urban populations compiled by Bairoch, Batou, and Chèvre (1988) and population estimates from
McEvedy and Jones (1978).

institutions than other countries had (e.g., secure property rights for a larger portion
of the population and relatively free entry into profitable economic activities). These
countries grew, by far, the most between 1500 and 1800. Figure 6 illustrates this by
comparing the evolution of the urbanization rate in England and the Netherlands
with that of Spain, Portugal, and France.
The second group is Spain and Portugal. These countries had access to the At-
lantic and, in fact, were first-movers when it comes to Atlantic trade and coloniza-
tion. However, they had highly absolutist monarchies at the time. This implied that
the sovereign and his elite were able to dominate the Atlantic trade. As a result,
these countries did not experience liberal revolutions until several hundred years
later (in the 20th century) and did not experience the high growth rates of the first
group. France’s experience is intermediate between the first and the second groups.
The French monarchy was sufficiently absolutist to largely dominate the Atlantic
trade early on and prevent a liberal revolution from occurring until the late 18th
century. Growth in France took off later than in Britain and the Netherlands, but

18
earlier than in Spain and Portugal.
The third group is the city states of Northern Italy and the Baltic Sea. These
states had relatively non-absolutist initial institutions. However, they lacked access
to the Atlantic and therefore did not benefit from the Atlantic trade. They saw their
relative wealth and stature decline markedly after 1500. Finally, most other parts of
Europe both lacked access to the Atlantic and had highly absolutist initial institu-
tions. They grew very slowly between 1500 and 1800.
One thing that the ideas of both North and Weingast and Acemoglu, Johnson,
and Robinson highlight is the notion that socially beneficial institutional change
may not occur because it is resisted by groups that stand to lose privileges and
power. In principle, the aggregate gains resulting from the institutional changes
should make it possible for an agreement to be reached that compensates those
groups that lose from the change—this is what the Coase Theorem suggests. In
practice, however, such schemes appear difficult to negotiate (sometimes because
the privileges and power of the elite are not considered just by the rest of the popu-
lation). For this reason, the process of institutional change often involves substantial
amounts of conflict. And in many cases, beneficial institutional change does not oc-
cur or occurs only after long delay. Resistance to change by special interest groups
with power and privilege is surely one of the most important barriers to growth
both today and throughout history.

2 Enlightenment
The Industrial Revolution involved a surge in the rate of technical innovation. The
most famous innovations occurred in textiles (the spinning jenny, water frame,
mule, etc.) and steam power (the Newcomen engine, the separate condenser, the
rotary engine, the high pressure engine, etc.). But technical innovation was much
more widespread with important innovations appearing in many industries. In the
words of economic historian T.S. Ashton, “a wave of gadgets swept over England”
(Ashton, 1948/1997, p. 48). The roughly simultaneous quickening of technological
innovation across many parts of the economy suggests that the Industrial Revolu-
tion was not due to one or two lucky inventions in the 18th century. It suggests,
instead, that something changed that improved the general conditions for techno-
logical innovation.
Joel Mokyr and Deidre McCloskey have argued that the crucial change was one

19
of ideas and culture associated with the Enlightenment (Mokyr, 2009; McCloskey,
2006, 2010, 2016). According to this view, a culture of progress and growth emerged
in Northwestern Europe in the 18th century based on the idea that mankind can im-
prove its condition by scientific inquiry and rational thought. A central element of
the Enlightenment was an attack on religious orthodoxy and the notion that God’s
word was the final answer to all questions (but not religion per se). Instead of re-
ligious dogma, leaders of the Enlightenment advocated empiricism and reason. In
the words of Francis Bacon, “Man, being the servant and interpreter of Nature, can
do and understand so much and so much only as he has observed in fact or in
thought of the course of nature” (Bacon, 1620/1999). Immanuel Kant famously de-
clared “Sapere Aude! [e. Dare to Know] Have courage to use your own reason!”
(Kant, 1784/1986). These notions may seem commonplace today. They were, how-
ever, radical in a society dominated in intellectual matters by religion. The Catholic
Church did not like people questioning assumptions and fiercely resisted the hereti-
cal writings of Enlightenment thinkers.
The idea that the Enlightenment was a crucial precursor to the Industrial Revo-
lution because it created conditions favorable to technological innovation begs the
question: Why did the Enlightenment occur? One possible answer is that perhaps
an unusual cluster of brilliant and freethinking people happened to live in North-
western Europe in the 17th and 18th centuries. Perhaps it was simply a stroke of
luck that the likes of Francis Bacon, Isaac Newton, John Locke, Thomas Hobbes,
David Hume, Adam Smith, René Descartes, Voltaire, Jean-Jacques Rousseau, An-
toine Lavoisier, Gottfried Wilhelm Leibniz, Leonhard Euler, Immanuel Kant, and
Christiaan Huygens, just to name a few important thinkers, lived and worked in
Northwestern Europe in the 17th and 18th centuries.
This theory seems unlikely to be true. It suggests that essentially no person of
equal brilliance and freethinking ability to the people listed above lived on the entire
continent of Europe for hundreds and hundreds of years during the “Dark Ages”
(say 500 to 1200 CE), while over a dozen such people lived in a small part of the con-
tinent over the short span of the Enlightenment. If extreme talent is random, such
a huge disparity in the frequency of extreme talent would, statistically speaking, be
astronomically unlikely to occur given the large size of the population of Europe.
But if many brilliant and freethinking individuals have lived at all times and in
all places, why have so few important ideas survived from other times and places?
There are likely several reasons for this. First, the technology available to maintain
and transmit knowledge throughout much of history was primitive and expensive.

20
Paper was invented in China about 2000 years ago but took over 1000 years to reach
Europe. Before paper, documents were carved in stone or written on animal skin at
great expense. Most things were simply never written down. The primary method
for knowledge transmission was the spoken word. “The Memory of individuals
and of communities carried knowledge through time and space,” as Boorstin (1985,
p. 480) put it. This meant that knowledge spread much more slowly than it can in
modern times and was more vulnerable to being lost forever. Mokyr (1990, ch. 9)
discusses important technologies that the Chinese mastered but subsequently lost
altogether, such as technologies for time measurement and for building oceangoing
ships. Clark (2007, p. 142-144) discusses evidence for technological regress among
Aboriginals in Australia and especially Tasmania as well as among Inuits in the
Arctic. When first encountered by Europeans, these groups had a level of technical
knowledge much more primitive than their own ancestors are known to have had.
Many of the important inventions of the Industrial Revolution were quite sim-
ple and did not rely on any advanced scientific knowledge. They could therefore
have been invented by nearly anyone. In textiles, the spinning jenny and the flying
shuttle, for example, are simple inventions that yielded dramatic increases in pro-
ductivity. Millions of people spun and wove fabric for millenia before the Industrial
Revolution. How could it be that none of them thought of these simple improve-
ments? It seems likely that some of them actually did, but that knowledge of the
improvements did not manage to spread.
Another important reason why so little knowledge survives from most times and
places is that huge amounts of written material have undoubtedly been lost to the
destruction of war and fire. A famous case is the Great Library of Alexandria, one of
the largest libraries of the ancient world. Although sources differ, many say it was
seriously damaged due to specific episodes of war and fire. Some say the armies
of Julius Caesar accidentally burned down part of the library in 48 BCE. Others say
that the library was badly damaged when Emperor Aurelian suppressed a revolt by
Queen Zenobia in the third century CE.
But perhaps the most important reason why so little knowledge survives from
most times and places is that powerful actors in society actively seek to suppress
ideas and destroy knowledge that harms their interests. The Inquisition of the
Catholic Church is a famous example. Another prominent example is the censor-
ship and ideological repression of the Soviet Union and its satellite states during
the 20th century. But censorship, the burning of books, and the imprisonment and
murder of reformers and intellectuals has been endemic throughout history.

21
 Whether the ideas of brilliant and freethinking individuals survive or not de-
pends on the relative strength of the forces and technologies that destroy knowledge
and suppress new ideas and the forces and technologies that preserve knowledge
and spread new ideas. The fact that the state of human knowledge was relatively
stagnant throughout most of history suggests that the forces and technologies that
sought to suppress new ideas were powerful enough to completely counterbalance
the forces and technologies that sought to spread new ideas.

2.1 A Watershed Moment: The Invention of the Printing Press

Then something changed. But what? I contend that a watershed moment was the
invention of the movable-type printing press by Johannes Gutenberg around 1450.5
The idea is that the printing press massively altered the balance of power between
those seeking to suppress knowledge and those seeking to spread it: after its inven-
tion, the forces seeking to create and spread knowledge decisively had the upper
hand and knowledge began to expand at a rapid rate. If this hypothesis is correct,
the invention of the printing press is probably the most important invention of all
time.
The hypothesis that the invention of the movable-type printing press was the wa-
tershed moment that “changed everything” faces at least two important challenges.
First, movable-type printing had been invented in China long before Gutenberg—at
least as early as the 11th century (Boorstin, 1985, ch. 62). Why didn’t movable-
type printing revolutionize knowledge accumulation in China? An important dif-
ference between the Chinese and European experience with movable-type printing
is the fact that the Chinese language has no alphabet. Rather, it has more than thirty
thousand characters. To print books using movable type, a Chinese printer needed
type pieces for each of several thousand characters. To assemble a page, the printer
needed to retrieve each character from this vast collection and then put them back in
the right place after use. This was a laborious process that meant that movable-type
printing was not nearly as useful an invention in China as in Europe.
Second, the Industrial Revolution occurred a full 300 years after Gutenberg’s in-
vention. If movable type was the crucial element, why did the Industrial Revolution
not occur earlier? A plausible explanation for this is that European society needed
to undergo a colossal transformation before conditions were ripe for the Industrial
Revolution. The forces that opposed new ideas were very strong in Europe around
1450. It took quite some time to weaken these forces. The invention of movable-type

22
printing arguably set in motion a sequence of transformational events in Europe: the
Reformation in the 16th and 17th centuries, then the Enlightenment and Scientific
Revolution in the 17th and 18th centuries. The explosive economic growth of the
Industrial Revolution was a culmination of these changes.
Consider first the Reformation. Martin Luther is said to have nailed his Ninety-
Five Theses to the door of the All Saints’ Church in Wittenberg on October 31,
1517. This event is widely regarded as the starting point of the Reformation. In
truth, Luther’s revolution started off as an obscure academic debate mostly about
indulgences—the sale by the Catholic Church of certificates remitting some of the
punishment a person was allegedly due in purgatory because of his or her sins on
Earth. It, however, escalated, and in 1520 Luther was excommunicated by Pope
Leo X. Luther reacted by burning the papal bull of excommunication in Wittenberg,
refusing to recant his errors at the imperial diet at Worms—famously proclaiming
“here I stand, I can do no other”—translating the New Testament into powerful and
idiomatic German, and publishing a series of pamphlets calling for major ecclesi-
astic reform. These events made Luther a celebrity and national hero in Germany.
Marshall (2009) provides an excellent account of these events and the Reformation
more generally, which my account draws on heavily.
There had been attempts at radical ecclesiastic reform before Luther. Famous ex-
amples include the movements of John Wycliffe in England and Jan Hus in Bohemia
in the 14th and early 15th centuries. In these prior cases, the Catholic Church was
able to beat the movements back and maintain control of the dogma. Jan Hus was
burned at the stake in 1415. In the case of Luther’s movement, however, the Catholic
Church totally lost control. As Marshall (2009) points out, unlike the writings of
Wycliffe and Hus, Luther’s writings were printed. Luther’s pamphlets were, in fact,
enormously popular, helping to spread his ideas wider and faster than anything the
Catholic Church had ever had to deal with before. Other ecclesiastic reform leaders
of the 14th and 15th centuries were arguably deeper thinkers on doctrinal matters
than Luther—including Calvin, Zwingli, and Melanchthon. Luther, however, made
extraordinarily effective use of the printed pamphlet as a means of spreading his
ideas, writing copious numbers of pamphlets in a language that appealed to a broad
audience.
The Reformation loosened the grip of religious authorities on intellectual life in
Europe, which helped lay the groundwork for the Enlightenment and Scientific Rev-
olution (Enlightenment for short). Censorship was much less severe in some Protes-
tant countries such as England and the Netherlands than in Catholic countries such

23
as France in the 17th and 18th centuries. The existence of these Protestant pockets of
relative liberty was arguably important for the spread of the Enlightenment. Some
of the most prominent Enlightenment thinkers were French. But in many cases they
published their work in the Netherlands, England, and other Protestant regions due
to censorship in France. Voltaire is a good example. Early in his life, French au-
thorities twice sent him to the Bastille for his writings. He fled to England and
the Netherlands and published much of his most controversial early work in those
countries (much of it was banned in France). Later in life, Voltaire settled close to
the Swiss border in Southern France and used Swiss printers to publish much of his
late controversial work—such as Candide.
Another channel through which the Reformation likely contributed to the emer-
gence of the Enlightenment was by encouraging literacy. Luther advocated univer-
sal schooling to enable Protestants to read the Bible on their own (Becker and Woess-
mann, 2009). Of course, once people could read the Bible, they could also read other
things. Historical data on literacy are, unfortunately, rather limited. However, sev-
eral researchers have gathered data on the fraction of people able to sign their names
on various formal records such as court depositions. The idea is that being able to
sign one’s name is a useful proxy for basic literacy. Figure 7 presents estimates
of the evolution of the literacy rate of husbandmen—farmers of relatively modest
means—in London and Middlesex from 1560-1740 based on data of this sort from
Cressy (1980) and Houston (1982).6 According to these estimates, literacy skyrock-
eted in the 200 year period leading up to the Industrial Revolution. In 1560, only
27% of husbandmen in London and Middlesex could sign their names. By 1740, this
fraction had risen to 84%.
More fragmentary evidence seems to indicate that literacy rates were very low
in medieval Europe and in Roman times (Clark, 2007, p. 175-181). The rise of mass
literacy in the lead up to the Industrial Revolution in Northwestern Europe is to my
knowledge a unique event in history. It seems likely that this rise of mass literacy
may have played an important role in enabling increased innovation and produc-
tivity in the leadup to the Industrial Revolution.
The most important channel through which the printing press affected literacy
was, however, almost certainly the enormous fall in the price of books that it caused.
Clark and Levin (2011) have gathered data on the price of books in England from
1350 to 1835. They use these data to construct a price index that provides an estimate
of the change in the price of a book of constant quality over this period. According
to their estimate, the price of books fell by a factor of 10 between the second half of

24
 1

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2
1560 1580 1600 1620 1640 1660 1680 1710 1730
Figure 7: Literacy Rate of Husbandmen in London and Middlesex, 1560-1740
Note: Results from a regression described in notes to the text based on data from Cressy (1980)
and Houston (1982). Cressy reports the fraction of witnesses who sign ecclesiastical records, while
Houston reports the fraction of witnesses who sign court depositions.

the 14th century and the first half of the 16th century. It then fell by another factor of
two over the next 50 years. These estimates suggest that a book that cost one pound
sterling around 1400 cost only one shilling (5% of a pound sterling) around 1600.
Despite being huge, this estimate probably understates the rise in accessibility of
books over this period. As I note above, Clark and Levin are attempting to estimate
the price of books of constant quality. Early printed books were of very high quality
because they were still luxury goods being marketed mainly to wealthy individuals
(and institutions) who were used to buying and reading books that had been tran-
scribed on vellum. Gutenberg, in particular, put a great deal of effort into making
his Bible have a very similar look and feel to contemporary Bibles transcribed on
vellum. Over time, however, people got used to the new medium and it became
less important to try to replicate the old medium. This allowed for smaller print
(more words per page), cheaper paper, less calligraphy, cheaper binding, and per-
haps most importantly much cheaper forms of written material such as pamphlets
and newspapers. All these things together meant that by the 18th century writ-
ten materials (pamphlets, newspapers, magazines, and books) were affordable to a

25
much larger portion of the population than a few hundred years earlier (when the
price of a book was similar to the price of a horse). The fact that people could afford
books surely massively increased the demand for being able to read.

2.2 Protestant Prosperity

The previous section puts forth a rather complex theory of causal links from the in-
vention of the printing press to the Industrial Revolution running through the Ref-
ormation, the Enlightenment, and the rise of mass literacy. I supported this theory
with some narrative history as well as data on literacy and the price of books. My
hope is that you find this theory plausible and intriguing. However, there is surely
a great deal of variation among you—the readers of this book—as to how plausi-
ble you find this theory and how important a role you think it played in bringing
about the Industrial Revolution. After all, the amount of evidence I have presented
supporting the theory is relatively small. In truth, I don’t think the amount of ex-
isting evidence on this topic is close to conclusive. However, there does exist some
clever and intriguing evidence that supports different parts of this theory. Below, I
describe in some detail an interesting recent paper by Becker and Woessmann (2009)
that provides evidence on the role of Protestantism and literacy.
Writing at the turn of the 20th century, Max Weber observed that Protestant areas
were more prosperous than Catholic areas (Weber, 1905/2002). He then famously
argued that this difference was due to a “Protestant ethic” that resulted in a “spirit
of capitalism” among Protestants, i.e., Protestants working harder, saving more, and
pursuing economic gain for its own sake. The Reformation had introduced the no-
tion of a “calling” which suggested a religious sanctification of labor. According
to Weber, this created a strong work ethic specific to Protestants. The Protestant
“recognizes, as the only means of living a life pleasing to God ... the fulfillment
of innerworldly duties which arise from the individual’s station in life. This then
becomes one’s “calling”” (p. 29). Weber contrasted this with other religions where
religious devotion typically involved rejecting worldly affairs for ”morality through
the pursuit of monastic asceticism.”
Weber’s ideas have been hugely influential within social science over the past
100 years. The premise—that Protestant areas were more prosperous around 1900—
is true. However, it does not necessarily follow from this that Protestantism causes
prosperity. (A correlation between two variables does not prove that one causes the
other.) The complicating issue is that areas that were Protestant in 1900 may differ in

26
other ways from Catholic areas—e.g., they may have better weather, more natural
resources, or better institutions—and it may have been these other differences that
led the Protestant areas to become more prosperous.
This problem is called the omitted variables problem: There may be other unob-
served variables that cause prosperity and are correlated with Protestantism. If this
is the case, the correlation between Protestantism and prosperity may simply reflect
the influence of these other variables. For example, suppose that the true deter-
minant of differences in prosperity around 1900 was some hard-to-measure aspect
of the weather: Places with good weather were more prosperous than places with
bad weather. Suppose furthermore that Protestantism happened to have spread in
places with good weather (or that good weather helped spread Protestantism). In
this case, prosperity would be correlated with Protestantism without there being
any true causal link between the variables. The true cause of prosperity would be
weather, which just happened to be correlated with Protestantism (or also caused
Protestantism).
How can we tell whether Weber was right about Protestantism causing pros-
perity? As with all theoretical hypotheses about how the world works, we need to
look for empirical evidence that either supports or contradicts Weber’s thesis. The
scientific ideal would be if we could run a randomized controlled experiment. In
this particular context that would mean: 1) having a large sample of regions, 2) ran-
domly picking some of them and exposing these to Protestantism, 3) waiting 250
years and seeing if the regions that were randomly exposed to Protestantism (but
otherwise the same on average) ended up becoming more prosperous. Obviously,
this type of randomized controlled experiment is not feasible in practice.
Since true experiments are seldom feasible in economics, economists have devel-
oped alternative methods to provide evidence on questions of causation. The key
idea that underlies these methods is the idea of a natural experiment. For the question
we are discussing—whether Protestantism causes prosperity—a natural experiment
is some fluke of nature or fluke of history that yields variation in the propensity of
regions to be Protestant but is otherwise unrelated to prosperity. More specifically,
we are searching for a variable that has two characteristics:

1. Relevance: It is correlated with variation in Protestantism,

2. Exogeneity: It is uncorrelated with all other determinants of prosperity.

A variable that satisfies these two conditions is called a valid instrumental variable
(or instrument for short). The idea is to focus on a component of the variation in

27
Protestantism that is uncorrelated with other determinants of prosperity (uncorre-
lated with all omitted variables). The instrument is a tool to capture only this com-
ponent of the variation in Protestantism. In other words, the instrument is a tool that
allows us to vary only a single potential determinant of prosperity (Protestantism).
If the instrument is valid (i.e., uncorrelated with other determinants of prosperity),
then that means that on average all things other than Protestantism are equal as we
vary the instrument. As a consequence, if the instrument is correlated with prosper-
ity, it must be that this correlation arises because Protestantism causes prosperity.
This is a subtle idea. The next several pages will illustrate it through example.
First, let us consider what kind of variable might satisfy the relevance and exo-
genity conditions for being a valid instrument. To get some practice thinking about
this, it is useful to consider an example of a variable that is not a valid instrument
(i.e., doesn’t satisfy the conditions described above). Consider, for example, the level
of literacy at the time of the Reformation as a candidate instrumental variable. It is
plausible that the level of literacy of different regions at the time of the Reformation
predicts their propensity to adopt the new religion. Perhaps more literate regions
were more likely to adopt Lutheranism because Luther’s emphasis on reading the
Bible in vernacular appealed to them more strongly. If this was the case, literacy at
the time of the Reformation satisfies the relevance condition of a valid instrument.
It seems less likely to be true, however, that literacy at the time of the Reformation
satisfies the exogeneity condition. To satisfy this condition, it must be the case that
literacy at the time of the Reformation does not affect the subsequent prosperity of
the regions in question through any other channel than potentially the adoption of
Protestantism. It is easy to think of plausible violations of this idea. Literacy may
in fact affect subsequent prosperity through many channels. For example, literacy
allows farmers to more easily learn about new crop rotations, literacy facilitates the
use of contracts which may be crucial for many forms of investment, and literacy al-
lows traders to keep better records and learn about basic accounting practices. For
all of these reasons (and many more) it is likely that literacy affects prosperity. This
means that literacy at the time of the Reformation is not a good candidate for an
instrumental variable for Protestantism.
Finding a variable that simultaneously is relevant and exogenous can be quite
difficult. A large part of the ingenuity involved in empirical economics has to do
with thinking of clever instruments and gathering the data needed to measure them.
For the question we are interested in, Becker and Woessmann (2009) propose a very
clever instrument: distance to Wittenberg. Wittenberg was the city where Luther

28
lived and worked. Becker and Woessmann’s idea is that places closer to Witten-
berg are more likely to be Protestant because they were more strongly exposed to
Luther’s ideas. At the time, travel and the transmission of information was slow
and costly. Luther’s ideas were spread partly by thousands of students and scholars
who came to Wittenberg to hear Luther’s sermons and speeches. Many priests also
came to Wittenberg to be officially ordained. High cost of travel made these trips
much less feasible for those living further away from Wittenberg. As a consequence,
the Protestant Revolution diffused from Wittenberg. This logic makes it likely that
distance to Wittenberg satisfies the relevance condition for being a valid instrument,
at least when applied to regions within Germany.
On the other hand, it is generally accepted by historians that, before Luther, Wit-
tenberg was an unimportant place. This makes it plausible that distance to Wit-
tenberg did not affect subsequent prosperity through other channels than Protes-
tantism and therefore satisfies the exogeneity condition for being a valid instrument.
It is generally not possible to prove empirically that an instrument satisfies the ex-
ogeneity condition. Researchers can make arguments—such as the one Becker and
Woessmann make and I repeat above—that a variable is likely to be exogenous.
Researchers can also present evidence suggesting that—in our context—regions fur-
ther away from Wittenberg are not systematically different along various observable
dimensions than places closer to Wittenberg. But there are always many unobserv-
able dimensions that are left unexplored for reasons of data limitation: It is rarely
possible to make sure everything else is held constant in a natural experiment. This
means that ultimately the notion that a variable satisfies the exogeneity condition
must always to some degree be assumed.
Given a candidate instrumental variable, the basic structure of the empirical ar-
gument usually involves three important steps: a first stage regression, balance tests,
and a reduced form regression. In our case, these steps are:

1. First Stage: Is distance to Wittenberg correlated with Protestantism?

2. Balance Tests: Were areas closer to Wittenberg different in other ways?
3. Reduced Form: Is distance to Wittenberg correlated with prosperity?

Becker and Woessmann present empirical evidence that sheds light on these
questions using data on 452 counties of Prussia, the dominant German state around
1900. Let’s start by discussing their evidence for the first stage relationship. Figure
8 shows a rough depiction of the share of the population that was Protestant across

29
 Figure 8: Protestant Population Share in Prussia in 1871
Note: County-level depiction based on 1871 Population Census. See Appendix A of Becker and
Woessmann (2009) for further information about the data.

Prussian counties according to the Prussian census of 1871. The location of Witten-
berg is depicted by a dot close to the center of the figure. The figure reveals a striking
pattern. The vast majority of counties close to Wittenberg have a population share
of Protestants that is above 85%, while counties further away from Wittenberg are
much more likely to have a lower Protestant share of the population. The figure,
therefore, strongly suggests that distance to Wittenberg is correlated with Protes-
tantism within Prussia in 1871.
We can formally test whether distance to Wittenberg is correlated with Protes-
tantism by running the following (first-stage) regression

PROTi = α + βDISTi + X′i γ + ϵi , (1)

where PROTi denotes the population share of Protestants in county i, DISTi denotes
county i’s distance to Wittenberg in kilometers, Xi denotes a vector of control vari-
ables, and ϵi is an error term. The coefficient of interest in this regression is β, which

30
 Table 2: Explaining Protestantism
Dependent Variable: Share Protestants
Distance to Wittenberg in km -0.095
(0.011)
Number of Observations 452
1st-Stage F-Statistic 74.2
Notes: These estimates are taken from Table III in Becker and Woessmann (2009).
Standard errors are in parentheses. The control variables included in the regression
are: % age below 10, % Jews, % females, % born in municipality, % of Prussian
origin, average household size, ln(population size), population growth from 1867-
1871 in %, % missing education info, % blind, % deaf-mute, % insane.

measures the extent to which distance to Wittenberg is correlated with the Protes-
tant population share conditional on the controls in Xi . Table 2 presents Becker and
Woessmann’s estimate of this regression. Their estimate of β is -0.095. This means
that on average the share of Protestants drops by 9.5 percentage points for each 100
km one moves away from Wittenberg. The standard error on Becker and Woess-
mann’s estimate is 0.011. This is an estimate of the statistical uncertainty associated
with Becker and Woessmann’s estimate of β. A common convention (rule of thumb)
in economics is to consider values of β that are more than roughly two standard er-
rors away from the estimate rejected, but values within roughly two standard errors
not rejected. Adopting this rule of thumb, we can say that Becker and Woessmann’s
data reject values of β outside of the interval [−0.073, −0.117]. In particular, the no-
tion that there is no relationship between the population share of Protestants and
distance to Wittenberg—i.e., that β = 0—is soundly rejected by Becker and Woess-
mann’s data. When a variable is being used as an instrument, a stricter statistical
standard is usually applied. In these cases, a common rule of thumb is that the
F-statistic for the hypothesis that β = 0 should be larger than 10 (which roughly
corresponds to β being more than three standard deviations away from zero). The
F-statistic in Becker and Woessmann’s first stage is 74.2. Since the F-statistic is above
10, we say that distance to Wittenberg is a “strong” instrument.
To corroborate their assumption that distance to Wittenberg is exogenous, Becker
and Woessmann present a number of balance tests. The balance tests take the form
of regressions of various characteristics of Prussian counties prior to the Reforma-
tion on distance to Wittenberg. The idea is that if distance to Wittenberg is to be
considered exogenous in this context, it should not be correlated with observable

31
economic characteristics of Prussian counties prior to the Reformation that may
themselves have affected the subsequent prosperity of these counties. Formally,
the balance test regressions take the form
Yi = α + βDISTi + X′i γ + ϵi , (2)
where Yi is an economic characteristic of county i.
Becker and Woessmann present nine such balance tests. Table 3 presents the re-
sults for four of these. The first column presents results for a regression where the
dependent variable Yi is an indicator for whether a county was a free imperial city
(Reichsstädte) in 1517—i.e., the variable is equal to one if the county was a free impe-
rial city at this time and zero otherwise. The results presented in the second column
are for a regression where the dependent variable is a measure of urbanization in
1500: urban population per km2 . The dependent variable for the third column is
an indicator for whether there was a university in the county and the dependent
variable for the fourth column is an indicator for whether there was a school in the
county. All four of these measures are chosen as proxies for economic or educational
development.
In all four cases, the estimated coefficient on distance to Wittenberg is small in
magnitude and also small relative to the coefficient estimate’s standard error. In
none of these four cases (or the other five reported in Becker and Woessmann) is the
coefficient estimate larger in absolute value than two times the standard error. As
we discussed above, economists typically consider values further than roughly two
times the standard error away from an estimated coefficient to be rejected by the
data at hand. Values closer than this cannot be rejected due to statistical uncertainty.
The fact that the estimated coefficients in the four balance tests reported in Table
3 are smaller in absolute value than two times their standard errors means that in
none of these four cases can the value zero be rejected. In cases like these, we say
that the coefficients are not statistically significantly different from zero. This means
that none of these balance tests provide statistically significant evidence against the
idea that distance to Wittenberg is exogenous. Distance to Wittenberg, therefore,
passed these balance tests, which corroborates Becker and Woessmann’s claim that
it is a valid instrument.
We now turn to the third step of the argument: is distance to Wittenberg corre-
lated with prosperity in the 19th century? We can formally test this by running the
following (reduced-form) regression
PROSPi = α + βDISTi + X′i γ + ϵi , (3)

32
 Table 3: Was Wittenberg a Random Place?
Dependent Variable: Imperial City Urbanization University School
in 1517 in 1500 in 1517 in 1517
Distance to Wittenberg 0.0034 0.00006 -0.0019 -0.0073
(in 100 km) (0.0071) (0.00013) (0.0047) (0.0099)
Number of Observations 452 452 452 333
R2 0.0005 0.0004 0.0004 0.002
Notes: These estimates are taken from Table IV in Becker and Woessmann (2009). Standard errors
are in parentheses. The dependent variables are: an indicator for whether a county was a free im-
perial city, urban population per km2 , indicator for whether county had a university, and indicator
for whether the county had a school.

where PROSPi is a proxy for prosperity of county i in the late 19th century. Becker
and Woessmann (2009) present results for three different proxies for prosperity.
Their main proxy is income tax revenue per capita in 1877. The other two mea-
sures they use are the logarithm of average annual incomes for male elementary
school teachers in 1886 and the share of the labor force working in manufacturing
and services in 1882. Table 4 presents estimates for all three of these measures. The
estimates for all three proxies indicate that distance to Wittenberg is indeed corre-
lated with prosperity. Specifically, counties that are further away from Wittenberg
on average have lower income tax revenue per capita, lower teacher salaries, and a
smaller share of the workforce working outside of agriculture.
Let’s consider the estimate for income tax revenue in more detail. The point
estimate in this case is -6.0. This means that income tax revenue per capita falls on
average by 6 pfennig (0.06 Marks) for each 100km one moves away from Wittenberg.
The standard error on this coefficient estimate is 2.3. The estimate is, therefore, con-
sidered statistically significantly different from zero (since it is further away from
zero than two times the standard error). But is the relationship quantitatively mean-
ingful? The average distance to Wittenberg across counties in Becker and Woess-
mann’s data is 326km and the standard deviation of the distribution of distance
to Wittenberg is 149km. Moving one standard deviation further away from Wit-
tenberg, therefore, on average lowers per capita tax revenue by roughly 9 pfennig.
Average per capita tax revenue across counties was 198 pfennig and the standard
deviation of per capita tax revenue 70 pfennig. Moving one standard deviation fur-
ther away from Wittenberg, therefore, on average lowers per capita tax revenue by
roughly 0.13 standard deviations. While this is not huge, it is certainly a meaningful

33
 Table 4: Does Distance to Wittenberg Predict Prosperity
Dependent Variable: Income Tax Log Teacher Share Manuf.
per capita Income and Services
Distance to Wittenberg -6.0 -1.00 -0.78
(in 100 km) (2.3) (0.48) (0.36)
Number of Observations 426 452 452
Notes: These estimates are produced using replication code and data provided by Ludger
Woessmann. They are the reduced form estimates that correspond to the instrumental variables
(IV) estimates presented in Table V of Becker and Woessmann (2009) except that I have included
“% missing eduction info” as an additional control for consistency with the first stage reported
in Table 2. Income tax per capita is measured in pfennig. Standard errors are in parentheses.
The control variables included in the regression are: % age below 10, % Jews, % females, % born
in municipality, % of Prussian origin, average household size, ln(population size), population
growth from 1867-1871 in %, % missing eduction info, % blind, % deaf-mute, % insane.

amount. Another way to gauge quantitative magnitude is to ask how much income
tax revenue per capita falls on average when one moves from the 10th percentile of
the distribution of distance to Wittenberg to the 90th percentile. This turns out to be
roughly 22 pfennig, or about 11% of average per capita tax revenue.
The results in Tables 2 and 4 show that distance to Wittenberg is correlated with
Protestantism and is also correlated with prosperity. If distance to Wittenberg is a
valid instrument, then it is not correlated with any other determinant of prosperity
(i.e., it is uncorrelated with all omitted variables). This means that when we move
from cities that are close to Wittenberg to cities that are far away from Wittenberg,
the only potential determinant of prosperity that is changing on average is Protes-
tantism. The fact that prosperity is changing as we move away from Wittenberg
must therefore be due to the change in Protestantism (everything else is being held
constant on average). We can therefore conclude that Becker and Woessmann’s data
support Weber’s theoretical conjecture that Protestantism causes prosperity. Obvi-
ously, this conclusion relies heavily on Becker and Woessmann’s claim that distance
to Wittenberg is a valid instrument.
By combining information from the first stage regression and the reduced form
regression, we can produce a quantitative estimate of the effect of Protestantism on
prosperity. In other words, we can answer the question: How much does prosperity
increase if Protestantism increases by (say) ten percentage points? The easiest way to
see this is to consider the causal chain from distance to Wittenberg to prosperity. The
first link in the chain is from distance to Wittenberg to Protestantism. The second

34
 Table 5: IV Estimates of the Effect of Protestantism on Prosperity
Dependent Variable: Income Tax Log Teacher Share Manuf.
per capita Income and Services
Share Protestant 0.62 0.11 0.08
(0.24) (0.05) (0.04)
Number of Observations 426 452 452
Notes: These estimates are taken from Table V in Becker and Woessmann (2009). Income tax
per capita is measured in pfennig. Coefficients in the second column are multiplied by 100.
Standard errors are in parentheses. The control variables included in the regression are: % age
below 10, % Jews, % females, % born in municipality, % of Prussian origin, average household
size, ln(population size), population growth from 1867-1871 in %, % blind, % deaf-mute, %
insane.

link in the chain is from Protestantism to prosperity. The assumption that distance
to Wittenberg is exogenous—i.e., uncorrelated with all determinants of prosperity
other than Protestantism—then implies that

Effect of DIST on PROSP

= (Effect of DIST on PROT) × (Effect of PROT on PROSP).

Rearranging yields

Effect of DIST on PROSP

Effect of PROT on PROSP = . (4)
Effect of DIST on PROT
The numerator on the right-hand-side of this equation is the regression coefficient in
the reduced form regression, while the denominator is the regression coefficient in
the first stage. This tells us that dividing the reduced form regression coefficient by
the first stage regression coefficient yields an estimate of the causal effect of Protes-
tantism on Prosperity.
There is a more direct way to estimate the effect of Protestantism on prosper-
ity. This is a procedure called an instrumental variables (IV) regression. Using this
one-step method makes it easier to construct appropriate standard errors for the es-
timate. Becker and Woessmann report such IV estimates, which I reproduce in Table
5. We see that raising the share of Protestants by 10 percentage points increases in-
come tax revenue per capita by 6.2 pfennig. This estimate is statistically significantly
different from zero.
My discussion of Becker and Woessmann’s work to this point has focused on es-
tablishing a causal link between Protestantism and prosperity. This is not, however,

35
 Table 6: Effect of Protestantism on Literacy
Dependent Variable: Share
Literate
Share Protestant 0.19
(0.03)
Number of Observations 452
Notes: This IV estimate is taken from Table III in Becker and Woessmann (2009).
Standard errors are in parentheses. The control variables included in the regression
are: % age below 10, % Jews, % females, % born in municipality, % of Prussian
origin, average household size, ln(population size), population growth from 1867-
1871 in %, % blind, % deaf-mute, % insane, % missing education info.

the main point of Becker and Woessmann’s paper. Their central argument is not to
confirm Weber’s theory but rather to critique it and propose an alternative. Weber
did not merely argue that Protestantism caused prosperity. He argued that the rea-
son why Protestantism caused prosperity was that it fostered a “Protestant ethic”
that resulted in a “spirit of capitalism” among Protestants. It is this part of Weber’s
theory that Becker and Woessmann take issue with. They instead propose that the
channel through which Protestantism causes prosperity is that Protestantism en-
couraged literacy.
To support this idea, Becker and Woessmann report an IV estimate of the causal
effect of Protestantism on literacy in Prussia in 1886. I reproduce this estimate in
Table 6. The finding is that Protestantism did in fact have a large causal effect on
literacy in late 19th century Prussia. The IV regression coefficient is 0.19 and is
highly statistically significant. This estimate implies that conversion of the entire
population of a region from Catholicism to Protestantism resulted in a 19 percentage
point increase in literacy in the late 19th century.
Becker and Woessmann argue that this effect on literacy is large enough to ex-
plain the bulk of the difference in prosperity between Catholic and Protestant areas,
leaving little room for other explanations such as Weber’s notion of the Protestant
ethic. They point out that modern causal estimates of the effect of education on in-
come (based on IV regressions) turn out to yield estimates that are very similar to
regular (OLS) regression estimates of this relationship (Card, 1999). In other words,
the omitted variables bias in modern OLS estimates of the effect of education on in-
come turns out to be close to zero. Becker and Woessmann argue that if the same is
true in their setting, they can adjust their income measures for the effect that Protes-

36
tantism has on income through literacy by estimating the effect of literacy on earn-
ings using a simple (OLS) regression and subtracting the estimated effect of literacy
from their income measures. Once they have done this, they can then estimate the
causal effect of Protestantism a second time using this adjusted measure of income.
When Becker and Woessmann do this, they find that Protestantism has very little
(additional) effect on income. In other words, once they have adjusted for the effect
of Protestantism that flows through literacy, they find that there is very little effect
left over for other potential explanations. While interesting, this argument relies on
the rather strong assumption that the causal effect of literacy can be estimated using
an OLS regression in their setting.
To summarize, Becker and Woessmann (2009) provide causal evidence for two
of the causal links in the theory that I presented in section 2.1. They show that the
Protestant revolution caused a large increase in literacy and a more modest increase
in prosperity. To be precise, they estimate the relative effect of Protestantism in re-
gions that were more strongly treated by the Reformation. How large an aggregate
effect the Reformation had is actually not clear from their evidence alone. The ef-
fects of the Reformation almost surely also affected areas that remained Catholic,
implying that the aggregate effect may be much larger than the relative effect.
As I have emphasized above, Becker and Woessmann’s conclusions rely heavily
on the assumption that distance to Wittenberg is a valid instrument for the effect
of Protestantism on prosperity and literacy. Earlier in this section, I discussed how
the printing press likely played an important role in the success and spread of the
Reformation. Rubin (2014) presents systematic causal evidence supporting this no-
tion. Furthermore, Dittmar (2011) presents causal evidence that cities with printing
presses grew substantially faster in the 16th century. Together these results suggest
that the printing press may be an important omitted variable in Becker and Woess-
mann’s analysis. In other words, perhaps it is greater exposure to the printing press
that caused Protestant areas to be more prosperous, but Protestantism itself had no
effect on growth.
Dittmar and Rubin use distance to Mainz—the city where Gutenberg established
the first printing press—as an instrument for the printing press in their analyses.
One simple way to assess whether the printing press is an important omitted vari-
able in Becker and Woessmann’s analysis is to include distance to Mainz as a control
in the IV regressions that Becker and Woessman run. I have rerun the regressions
presented in Tables 2, 4, and 5 with this extra control. The first stage regression
is virtually unchanged. However, the reduced form and IV regressions are con-

37
siderably weakened. Focusing on the results for income tax per capita, including
distance to Mainz as a control leads the coefficient on distance to Wittenberg in the
reduced form to fall from -6.0 to -1.2 and become insignificantly different from zero.
Likewise, the IV estimate of the causal effect of Protestantism on income tax per
capita falls from 0.62 to 0.11 and becomes insignificantly different from zero. In the
reduced form, the coefficient on distance to Mainz is actually much more statisti-
cally significant than the coefficient on distance to Wittenberg when they are both
included in the regression.
These results illustrate that it is difficult to disentangle the impact of the printing
press and the impact of Protestantism on growth. It may be that the printing press
affected growth through its influence on the spread of Protestantism. But it may also
be that the printing press influenced growth through other channels and the results
of Becker and Woessmann are driven by the fact that the printing press caused both
growth and Protestantism. In this case, Protestantism itself may not have had any
direct influence on growth.

3 The Agricultural Revolution

The population of England more than tripled between 1700 and 1850 from about 5.5
million to about 17.5 million. Throughout much of this period, imports of food in ex-
cess of exports made up a very small fraction of total food consumption in England.
Given this, the explosion of the population could not have occurred without an ex-
plosion in agricultural output. This explosion in agricultural output was driven by
large increases in both crop yields and agricultural labor productivity. Average crop
yields in wheat, barley, and oats roughly doubled over this period (Broadberry et al.,
2015, Table 3.06). The fraction of the English population engaged in agriculture fell
substantially from about 55% to well below one third (Allen, 2000), implying that
labor productivity in English agriculture must have risen substantially. The large
increase in crop yields and agricultural labor productivity that occurred over the
150 year period after 1700 is referred to as the Agricultural Revolution.
Why did the Agricultural Revolution occur? Let’s begin by considering the prox-
imate causes and then work our way towards deeper causes. The proximate causes
of the Agricultural Revolution are not particularly controversial. Several develop-
ments contributed. The most important of these are likely to have been improve-
ment of land, more intensive crop rotations, and the breeding of higher yield ani-

38
mals and crops. During the 18th century, 8 million acres of ‘waste’ were converted
to ‘pasture’ more than doubling the area of improved pasture in England (Allen,
2009, p. 62-63). This conversion involved substantial improvement of the land as
fields were enclosed by walls and cleared of surface stone. In addition, large areas
were drained so as to make them suitable for crops or to improve yields.
The traditional crop rotation in England was a three field rotation. Each field ro-
tated between a winter grain such as wheat or rye, a summer grain such as barley or
oat, and lying fallow. Thus, a third of the arable land lay fallow at any given point
in time. Letting land lay fallow on a regular basis is one way to allow the land to
maintain its fertility over long periods of time. During the Agricultural Revolution,
the three field rotation was replaced by a more intense system of “convertible hus-
bandry” (sometimes also called “alternative husbandry” or the “Norfolk system”).
This system involved alternating fields between arable and pasture, the sowing of
grass seeds, as well as the introduction of new crops, most notably roots (such as
turnips) and legumes (such as clover). A prototypical example of this new rotation
was a four-course rotation of wheat, turnips, barley, and clover (Timmer, 1969).
This system had two primary benefits. First, it allowed farmers to keep much
larger stocks of animals resulting in more meat, dairy, and wool (turnips and clover
were used mainly as fodder for the animals). Second, it allowed farmers to dras-
tically reduce the amount of land that was left fallow (or completely eliminate the
fallow) without this resulting in diminished soil fertility. Legumes have nitrogen-
fixing qualities and therefore improve soil fertility. In addition, the fact that large
herds of animals were grazed on the land as part of the rotation (and kept through
the winter) yielded large amounts of natural fertilizer in the form of animal drop-
pings. The new system actually did more than just maintain the fertility of the soil.
It slowly increased the nitrogen content of the the soil, which improved grain yields
over time (Allen, 2008).
But why did these large improvements only occur in the 18th and 19th centuries,
and not earlier? Narrative evidence as well as data on crop yields and labor produc-
tivity indicate that improvements in English agriculture were much slower prior
to the 18th century. Economic historian Jan de Vries has pointed out that the key
principles employed to intensify the crop rotation were known long before the 18th
century and used in centers of advanced agriculture such as Flanders and Lom-
bardy. In fact, these principles were described in Roman agricultural handbooks (de
Vries, 1976, p. 39-40). For some reason, however, these methods were not widely
employed in early modern England.

39
 One possibility is that the knowledge of these principles was not widespread. As
we discuss in section 2, the technology for maintaining and spreading knowledge
was quite primitive prior to the invention of movable-type printing around 1450. De
Vries discusses how many books were published about improved agricultural pro-
ductivity in the 17th and 18th centuries, the most famous perhaps being Jethro Tull’s
1731 book Horse-Hoeing Husbandry. Spreading innovative ideas about efficient crop
rotations in the 18th century took time and a great deal of effort by agricultural re-
formers. Furthermore, adapting these ideas to local conditions required significant
experimentation in each location (Timmer, 1969).
An influential traditional view among economic historians holds that the slow
growth in agriculture before the 18th century was due to institutional inflexibility
resulting from the open field system of farming that was widespread in England
(and much of Northern Europe). According to this view, the Enclosure Movement
was a crucial institutional development that unleashed the productive potential of
English agriculture and, according to some, caused the Industrial Revolution by
freeing up labor for use in other activities than food production.

3.1 The Enclosure Movement

Before 1500 much of the land of England and Northern Europe was farmed under
the ancient system of open fields. In this system, the unit of settlement was a village
as opposed to a farm. Around each village there were typically three large fields.
These fields were divided into a large number of small strips of land. Each farmer
in the village farmed several of these strips. The three fields served as crop rotation
units in the traditional three field rotation system. In addition to these three fields,
the village had some meadow, pasture, woodlands, and ’wastes.’ This land was
managed communally by the villagers. The villagers grazed their livestock on these
“commons” as well as on the field that lay fallow and the other fields after harvest
time. The system of small strips within larger fields and the practice of opening the
fields to grazing after harvest meant that all farmers in the village must in practice
follow the same crop rotation and the same schedule regarding sowing and harvest-
ing the field (McCloskey, 1972; Allen, 2001a).
By 1850, this ancient system had been largely swept away in England by some
5,000 acts of Parliament and a similar number of voluntary agreements. It was re-
placed by a system of privately owned enclosed plots of land “free of village di-
rection” (McCloskey, 1972). This process is commonly referred to as the Enclosure

40
Movement. The shift from ambiguous communal property rights in the open field
system to a system of well-defined private property rights spread from England to
other countries in the late 18th century and the 19th century.
In England, enclosures had been taking place at least since the thirteenth cen-
tury. However, judging by the experience of the South Midlands, as documented
by Allen (1992), most enclosures occurred in two waves: There was an initial wave
between 1550 and 1650 and a second, much bigger wave between 1750 and 1850. En-
closures before 1750 were mostly by voluntary agreement. Landowners would sell
each other strips so that they ended up with consolidated holdings. They would
then divide up the village commons and renounce common grazing rights on each
others land. An important complication was that all those that had grazing rights
in the village needed to be party to the agreement (as did the tithe owner). Often
enclosure coincided with increases in the concentration of land ownership. Large
landowners or men who had made their fortune in trade would buy out all others
in a village, bar common grazing, and disband collective decision making.
The large wave of enclosures that occurred between 1750 and 1850 were over-
whelmingly by acts of Parliament. In this case, unanimity was not required. Agree-
ment by the owners of roughly 75% of the land in question was sufficient. In these
cases, an appointed commissioner would survey the land, redraw boundaries, and
allocate the new parcels to all those that owned land or grazing rights in the village
in proportion to the value of their prior holdings. Common grazing and collective
control was abolished.
The causes and consequences of the Enclosure Movement have been hotly de-
bated ever since the events transpired. Some see the enclosure movement as a cru-
cial institutional change where property rights were ‘rationalized,’ which improved
incentives for innovation and thereby caused the Agricultural Revolution. Others
see the enclosure movement as a massive act of expropriation by large landown-
ers of a complex system of ancient peasant rights: the poorest classes of people—
cottagers and the landless—typically received no land or very small parcels, but lost
formal or informal rights to graze a cow, keep some fowl, and gather firewood on
the common. In many cases, this meant that life was no longer viable in the village
for these people.

41
3.2 Did Enclosures Cause the Agricultural Revolution?
Contemporary accounts of agrarian reformers and early economic historians
viewed the open field system as being woefully inefficient. The system of scat-
tered strips meant that each farmer’s diligence or lack thereof had spillover effects
onto his neighbors’ strips (externalities). Also, the communal system of grazing led
to over-use of the commons (the commons problem). Furthermore, the open field
system was thought to be resistant to progress. According to Ernle (1912/1961),
“open-field farmers were impervious to new methods.” The main reason given for
this was the cumbersome nature of decision-making in an open-field village where
unanimity was required to make any change to the crop rotation or land use.
Proponents of this traditional view see the Enclosure Movement as a crucial
cause of the Agricultural Revolution. If this view is correct, the success of the En-
closure Movement in dramatically increasing productivity in English agriculture is
an important piece of evidence supporting the importance of capitalism (i.e., clearly
defined property rights) as a precondition for innovation and economic growth.
This traditional view has faced several challenges over the past 50 years. Eco-
nomic historian Robert Allen gathered evidence on the adoption of new crops and
on crop yields in the South Midlands over the period 1450-1850 (Allen, 1992, 2009).
Allen’s evidence indicates that open field farmers were far from ‘impervious to new
methods.’ They made changes to land use—e.g., introduced new crop rotations—
where this was appropriate. Yet, enclosed farms seem to have done so more com-
pletely, especially when it came to conversion of land from arable to pasture. Fur-
thermore, Allen’s data indicates that while crop yields were higher on enclosed
farms than open farms, this difference was a modest fraction of the increase in yields
since the middle ages. In other words, open farms managed to attain a large fraction
(something like 80%) of the increase in crop yields that enclosed farms attained.
This evidence certainly undermines the traditional view that institutional rigid-
ity of the open field farms made improvements impossible to implement. However,
it does not necessarily undermine the argument that enclosure was a crucial driver
of increased productivity in agriculture. Imitation is easier than innovation. It may
have been that enclosed farms led the way in terms of innovative practices in agri-
culture. The institutional rigidities of open farms may have been severe enough to
prevent innovation but not severe enough to prevent imitation of successful changes
implemented by nearby (enclosed) neighbors.
A different empirical argument focuses on the timing of increases in labor pro-

42
ductivity in agriculture. To measure labor productivity, one must be able to measure
both labor input and production. An important challenge in estimating labor pro-
ductivity in English agriculture over this period is lack of data on agricultural pro-
duction at the national level. Since direct data on agricultural production is scant,
economic historians have sought to estimate agricultural production indirectly by
first estimating food consumption and then making an adjustment for net imports
of food. Consider the following simple model for agricultural production:

Yat = ϕt Cat Nt , (5)

where Yat denotes agricultural output at time t, ϕt (the Greek letter phi) denotes the
ratio of agricultural production to agricultural consumption at time t, Cat denotes
food consumption per capita at time t, and Nt denotes the size of the population at
time t. This equation states that agricultural output in England (the left hand side)
must equal agricultural consumption (Cat Nt ) adjusted for net imports (ϕt ).
To get an expression for labor productivity in agriculture, we divide both sides
of equation (5) by the number of people working in agriculture:
Yat Nt
= ϕt Cat , (6)
Nat Nat
where Nat denotes the number of people working in agriculture at time t. The left
hand side of this equation is labor productivity in agriculture.
Wrigley (1985) uses this model to estimate the change in labor productivity in En-
glish agriculture from 1520 to 1800. He makes two simplifying assumptions. First,
he assumes that net imports of food into England are zero (i.e., ϕt = 1). Second,
he assumes that food consumption per person in England was constant over this
period. Given these two assumptions, the evolution of labor productivity is sim-
ply given by the evolution of the inverse of the fraction of the English population
engaged in agriculture.
Figure 9 plots Wrigley’s estimates of labor productivity in agriculture. Given
Wrigley’s assumptions, the sharp drop in the fraction of the agricultural population
in England over the 18th century implies that labor productivity in English agricul-
ture rose substantially. The increase from 1700 to 1750 is 20% and from 1750 to 1800
the increase is a further 26%. The large increase over the second half of the 18th cen-
tury provides indirect support for the importance of the Enclosure Movement since
this is when the large wave of Parliamentary enclosures occurred.
An important weakness of Wrigley’s estimates is that they rely on assumptions
that are unlikely to be true. The eighteenth century was a period of rapid change

43
 250

Wrigley (1985) adjusted

200 Wrigley (1985)
Allen (2000)

150

100

50

0
1200 1300 1400 1500 1600 1700 1800
Figure 9: Labor Productivity in Agriculture in England
Note: The solid gray line is from Wrigley (1985). The dashed gray line adjusts the solid gray line to
account for net imports and exports of food using estimates from Allen (2000). The solid black line
is from Allen (2000).

and this change is likely to have affected per capita food consumption. In particular,
per capita food consumption is likely to be significantly affected by changes in real
wages and changes in the relative price of food. In addition to this, towards the end
of the 18th century, England started to import a significant amount of food on net.
Allen (2000) relaxes the two simplifying assumptions that Wrigley makes. He
proposes a simple model of food demand and uses data on real wages and the rela-
tive price of food to estimate how per capita food consumption changed over time.
He also incorporates estimates of net imports of food. The black line in Figure 9
plots Allen’s estimates of labor productivity in agriculture. Allen’s estimates turn
out to be substantially different than Wrigley’s, especially during the 18th century.
According to Allen’s estimates, labor productivity rose by 34% in the first half of
the 18th century (before the largest wave of enclosures), but then fell by 8% in the
second half of the 18th century (during the enclosure wave). This time pattern is
clearly less consistent with the Enclosure Movement having played a crucial role in
the Agricultural Revolution.

44
 The main reason behind Allen’s lower estimated growth of labor productivity
in the second half of the 18th century is the fact that the relative price of food rose
during this period. This indicates that the supply of food was not able to keep up
with increases in demand resulting from the sharp increase in the English popula-
tion over this period. Increased net exports also play a (relatively modest) role. The
dashed gray line in Figure 9 adjusts Wrigley’s estimates for net imports of food but
maintain the assumption of constant per capita food consumption.
While Allen’s estimates are preferable in that they allow the demand for food to
vary with prices and income, a potential drawback of these estimates is that they are
sensitive to errors in the estimation of these variables. There is considerable debate
among economic historians about the evolution of real wages in England in the 18th
century. Furthermore, recent evidence suggests that an “industrious revolution”
occurred in the lead-up to the Industrial Revolution, where hours worked per day
and days worked per year increased (Humphries and Weisdorf, 2019). If this is the
case, measures of real wages understate the change in real income over this period
and therefore understate the increase in agricultural labor productivity.

3.3 Rural Growth and Urban Growth: Did One Cause the Other?
A second aspect of the traditional view about the Agricultural Revolution is the
notion that the Agricultural Revolution caused the Industrial Revolution. Ragnar
Nurkse stated this view in a particularly colorful manner in the following passage:

Consider what happened in the original home of industrial develop-

ment, in England in the eighteenth century. Everyone knows that the
spectacular industrial revolution would not have been possible without
the agricultural revolution that preceded it. And what was this agricul-
tural revolution? It was based mainly on the introduction of the turnip.
The lowly turnip made possible a change in crop rotation which did not
require much capital, but which brought about a tremendous rise in agri-
cultural productivity. As a result, more food could be grown with much
less manpower. Manpower was released for capital construction. The
growth of industry would not have been possible without the turnip and
other improvements in agriculture. (Nurkse, 1953, p. 52-53)

It is clearly true that industrial development cannot occur on a large scale with-
out improvements in agricultural productivity. If each farmer grows only enough

45
food to feed his family, there is no surplus of food to feed people in cities. In such
a circumstance, there can be no cities and no industry. This notion does not, how-
ever, imply that productivity growth in agriculture is necessarily the key cause of
industrial development. It is quite possible that causation runs the other way.
Let’s spell this out in a little more detail. One possibility is that the chain of events
is as follows: Productivity in agriculture increases. This pushes some people out
of agriculture because the agricultural sector can now feed more people than only
those that work in that sector. Equivalently, less farmers are needed to feed a given
population. This means that a larger fraction of the population is free to pursue
other activities. To the extent that some of these extra people devote themselves to
trade, industry, and innovation, this may initiate industrial growth.
Another possibility, however, is that the chain of events is quite different: Re-
turns to urban activities (trade and industry) rise for some reason. This generates
an incentive for people in rural areas to move to cities. In other words, the cities
pull people in from the country-side. The increased wealth in the cities results in in-
creased demand and higher prices for agricultural products. The increased demand
and higher prices of agricultural products strengthens the incentives of farmers to
improve the land and innovate, which in turn leads to improvements in agricultural
productivity.
In the first of these two stories, the Agricultural Revolution causes the Industrial
Revolution. In the second, however, the Industrial Revolution (or the rapid increase
in overseas trade we discuss in section 1) causes the Agricultural Revolution. A
third story is that both revolutions have a common external cause. For example, the
dramatic fall in interest rates in England after the Glorious Revolution (see Figure
3) encouraged investment both in urban and rural improvements and may thereby
have jump-started both the Agricultural and Industrial Revolutions.
How can we tell which of these stories is likely to have been most important?
Allen (2009) points out that the first two of these stories differ sharply in their impli-
cations for urban wages. In the first story, agriculture pushes people into the cities. If
this was the dominant story in 18th century England, wages in English cities should
have been pushed down due to a glut of people arriving from the countryside. In
the second story, however, it is the boom in the cities that pulls people in from the
country-side. If this was the case, we should have seen high wages in the cities. (The
third story also implies high urban wages.)
To assess whether real wages were high or low in English cities during the Agri-
cultural Revolution, it is useful to compare them with real wages in other areas that

46
 2.0

1.8

1.6

1.4

1.2

1.0

0.8

0.6

0.4
London Amsterdam Antwerp
0.2
Paris Valencia Milan
0.0
1200 1300 1400 1500 1600 1700 1800

Figure 10: Real Wages in European Cities

Note: These data are estimates of respectability ratios constructed by Robert Allen. The respectabil-
ity ratios divide nominal income of a laborer that works 250 days per year by the cost of 3.15
respectability baskets. See the text and Allen (2001b, 2009) for more details. Each point is a decadal
average.

did not experience an agricultural revolution. To this end, Figures 10 and 11 present
estimates of real wages and labor productivity in agriculture, respectively, in several
European countries over the period 1300 to 1850. These estimates were developed
by Robert Allen. The estimates of labor productivity in Figure 11 are from Allen
(2000) and are based on the same methods as those Allen used to produce the agri-
cultural labor productivity series for England that we plot in Figure 9.
The methods that underlie the estimates of real wages in Figure 10 were first
developed in Allen (2001b) and subsequently refined in Allen (2009). Comparing
real wages across space is tricky because the lowest cost way to attain a certain level
of well-being is different in different places. Recall that measures of real wages are
calculated by dividing a measure of nominal wages (e.g., wages in pounds sterling
or in grams of silver) by a measure of the price level. The price level is meant to
measure the cost of attaining a certain level of well-being. The thing that is tricky
about this is that attaining a certain level of well-being in two different places may

47
entail purchasing quite different baskets of goods. Suppose, for example, that we
are interested in calculating the cost of subsistence in two places. This involves
calculating the cheapest way to get enough calories and other vital nutrients to live
in these two places. This basket may differ a great deal depending on the relative
price of different types of food. In areas that are well suited for growing potatoes,
the cheapest way to subsist may involve buying a large amount of potatoes, while in
areas that are better suited for growing oats (say) the subsistence basket will involve
more oats and less potatoes.
Allen has constructed measures of the cost of subsistence and the cost of a “re-
spectable” standard of living for quite a number of European (and non-European)
cities. Figure 10 plots his estimates of “respectability ratios” for six European cities.
The respectability ratio is constructed by first calculating the nominal income of a
laborer that works for 250 days in a year. This level of income is then compared
with the cost of purchasing a basket of goods that yields a “respectable” standard of
living. The details of this basket differ across locations due to differences in climate
and cuisine. Table 7 presents the respectability basket for northwestern Europe. For
areas further south in Europe certain items (such as beer and butter) are replaced by
other items (wine and olive oil).
The respectability basket in Table 7 is for a single adult male. However, a typical
laborer’s income supported not only the laborer himself but also his wife and chil-
dren. To take account of this Allen assumes that three respectability baskets were
needed to support a family. Finally, to account for the cost of housing, Allen adds a
further 5% to the cost of the respectability basket. The respectability ratio is there-
fore equal to 250 times the nominal daily wage in a city divided by the cost of 3.15
respectability baskets in the city.7
Figure 10 reveals that the 15th century was a high point for living standards
across Europe. At this time, laborers could comfortably afford Allen’s respectability
basket (the respectability ratio was above one). Over the course of the 16th cen-
tury, living standard fell throughout Europe. But starting around 1600, a marked
divergence occured. Living standards remained low or continued to fall in Paris,
Valencia, and Milan, while they rose substantially in London and Amsterdam. The
pattern of low living standards in Paris, Valencia, and Milan is representative of
many other European cities. It is England and the Low Countries that are the out-
liers in terms of high living standards in the 17th and 18th centuries.
Let’s now compare these estimates of living standards with the estimates of la-
bor productivity in agriculture reported in Figure 11. This figure reveals that Eng-

48
 Table 7: Allen’s Respectability Basket for Northwestern Europe
Quantity Spending Nutrients Per Day
per person Share Calories Grams of
per year Protein
Bread 234 kg 36.0% 1,571 64
Beans/Peas 52 l 5.5% 370 28
Meat 26 kg 12.8% 178 14
Butter 5.2 kg 4.0% 104 0
Cheese 5.2 kg 3.3% 54 3
Eggs 52 1.1% 11 1
Beer 182 l 20.0% 212 2
Soap 2.6 kg 1.7% – –
Linen 5m 4.8% – –
Candles 2.6 kg 2.9% – –
Lamp Oil 2.6 l 4.3% – –
Fuel 5.0 M BTU 4.6% – –
Total 2,500 112
Note: The information in this table is taken from Table 2.1 in Allen (2009). The
spending shares are for the prices in Strasbourg from 1745-54. Spending shares in
other places and times may be somewhat different. Wine was substituted for beer
and olive oil for butter depending on local custom. The fuel ration was set at 2 M
BTU in Spain and Italy due to warmer climate. Wheat bread was used in London,
Paris, Valencia, and Milan, while rye bread was used in Antwerp and Amsterdam.

land and the Netherlands were the only two countries in Western Europe to experi-
ence an Agricultural Revolution in the 18th century. Other countries experienced no
growth or even a decline in agricultural labor productivity during this period. Mod-
ern day Belgium is an interesting outlier. It managed much higher productivity in
agriculture prior to the 18th century than other countries, but was then overtaken
by England and the Netherlands in the 18th century. Labor productivity in other
countries remained close to medieval levels.
Comparing Figures 10 and 11, we see that the Agricultural Revolution occurred
in exactly the places where real wages were the highest. Allen (2009) argues that
these facts are more consistent with the view that urban growth caused the agri-
cultural revolution than with the alternative view that the agricultural revolution
caused urban growth. The high wages in London and Amsterdam were pulling
people into these cities from the countryside and increasing demand for agricul-

49
 0.90

0.80

0.70

0.60

0.50

0.40

0.30

0.20
England Netherlands Belgium
0.10
France Spain Italy
0.00
1200 1300 1400 1500 1600 1700 1800
Figure 11: Labor Productivity in Agriculture in Europe Regions
Note: These data are from Allen (2000).

tural products. Allen’s argument suggests that it may have been the high demand
for agricultural products that induced increased innovation in agriculture and that
this may have helped bring about the Agricultural Revolution. We will discuss the
idea of such “induced technical change” in more detail in section 4.

3.4 The Decline of Feudalism and the Rise of Capitalism

Medieval England was a feudal society. Formally, all land was owned by the Crown.
In practice, however, the Crown granted most of the land to noble lords in exchange
for periodic fees and an obligation to provide soldiers in times of war. The land
was organized into manors. In most cases, a manor consisted of a village and the
surrounding fields and commons. In terms of ownership, the manor consisted of
three types of land: the demesne, freeholds, and villein land. The demesne was
under the direct control of the lord. Freeholds were farmed by free peasant farmers,
while the villein land was farmed by serfs (also referred to as villeins). The lord
owned the villein land and to a large extent owned the labor power of the serfs

50
that farmed that land, although the lords were bound by certain customs in their
treatment of the serfs.
While medieval freeholds and villein farms were usually small, the demesne was
often much larger. On manors with many serfs, the demesne was largely farmed
by the serfs (for the benefit of the lord). However, on manors with large demesne
and few serfs, the lord hired wage labor. A royal inquisition from the 13th century
indicates that in the South Midlands 32% of land was demesne, 40% was villein,
while 28% consisted of freeholds (Allen, 1992, ch. 4). Clearly a large fraction of the
population of medieval England was shackled in serfdom.
The 600 year period between the Norman Conquest of 1066 and the English Civil
War was a period of gradual but ultimately radical change in the structure of prop-
erty rights and the extent of rule of law in England, as well as the degree of personal
freedom of the English population. Early in this period, property rights were limited
at all levels. At the top of the hierarchy, the lord could not sell or bequeath his manor
without consent of his superior lord (often the Crown), and in some cases such con-
sent required the payment of a fee. The same was true of freeholds. These could not
be sold, leased, or bequeathed without the consent of the lord of the manor.
Over time, however, the property rights of lords vis-a-vis the Crown and free-
hold farmers vis-a-vis the lord strengthened. For freehold farmers, the gradual in-
crease in the importance of royal courts was a crucial development. Prior to the rise
of royal courts, the only source of justice available was seigneural or manor courts
(i.e., courts administered by the lord). This obviously meant that individuals had
no recourse when it came to disputes with the lord, which gave the lord a great deal
of power. Fukuyama (2011, ch. 17) discusses how after the Norman Conquest, the
Kings of England gradually established courts of appeal: the royal courts. This was
advantageous to the King for several reasons: it helped establish his prestige and
legitimacy; and it generated a non-trivial amount of revenue for the Crown. Over
time, the rulings of the royal courts established the English Common Law; common
because it applied universally across all of England in contrast to precedents from
seignueral courts.
By the 14th century, freehold farmers had become the full owners of their prop-
erty: they could use it as they pleased, lease it, bequeath it, and sell it (with rel-
atively minor restrictions). The situation of serf farmers, however, was much less
favorable. Their only protection was custom. They could not seek justice from royal
courts if disputes arose with the lord. The lord could expel them from their land as
he pleased. The serf was subject to arbitrary fines (tallage), restrictions on marrying

51
outside the manor (merchet), arbitrary requests for labor service on the demesne,
corporal punishment, and imprisonment by the lord. Finally, serfs could be sold as
chattel. See Allen (1992, ch. 4) for more detail.
Allen (1992) argues that this sorry state of affairs for serfs in England “collapsed
under the impact of the Black Death of 1348-9 and the subsequent century of pop-
ulation decline.” At first after the Black Death, lords sought to fill vacant farms
by promoting landless cottagers. However, as depopulation continued, the grip of
lords on their vassals loosened. A peculiar feature of serfdom in England was that
a serf’s subordinate position only applied vis-a-vis his own lord. In other fiefs, the
serf was no different from a free person. Furthermore, there was no ‘Fugitive Serf
Law’ in England to facilitate the return of serfs that ran away. Finally, other lords
were quite keen to attract and retain any tenant that was able to pay rent. As a con-
sequence, serfs that managed to flee their lords left their servitude behind and by
1500 feudal serfdom was almost completely eliminated in the midlands.
The labor scarcity of the fifteenth century also led lords to switch from labor in-
tensive arable farming to less labor intensive pastoral farming. This was often done
through enclosure. In some cases, areas were enclosed that had been voluntarily
abandoned. But in others, enclosures involved eviction and subsequent depopula-
tion of the manor. These 15th and early 16th century enclosures involved significant
expropriation of peasant property and customary rights, destruction of villages, and
expulsion of their inhabitants. They therefore line up well with the Marxist view of
enclosures as exploitation of the poor by the rich.
Allen (1992) argues that the exploitative and depopulating nature of these early
enclosures “so alarmed official opinion that the Crown began protecting peasant
farmers.” This was done through several means. The most effective was the gradual
extension of secure property rights to tenant farmers. Copyholds and beneficial
leases—the main forms “rental agreements” of this era—became secure forms of
tenure. This process limited greatly the ability of lords to expropriate their vassals
through enclosure. With enforceable titles, enclosures could only proceed with the
agreement of peasant tenants.
Traditionally, copyholders were free tenants that held villein land ‘at the will
of the lord according to the custom of the manor.’ Generally, copyholders paid a
modest annual rent and a more substantial fee to the lord when the land was sold
or passed to an heir. The determination of these fees was somewhat ambiguous. In
many cases, they were governed by custom. But lords had some ability to alter them
arbitrarily and could block sales by stipulating a high enough fee. More importantly,

52
villein tenure was traditionally at will. Lords could therefore evict copyhold tenants
at any time even in violation of manoral custom without the tenant having any
redress.
Over the course of the quarter millennium following the Black Death, the royal
courts gradually extended the property rights of tenant farmers in England. They
came to enforce the right of tenants to sell their copyhold in exchange for a cus-
tomary fee. They also eliminated the right of lords to evict tenants in violation of
custom. This meant that copyholds either became completely inheritable (copyhold
of inheritance) or a lease ’for lives’ (copyhold for lives). Copyholds for lives were
usually granted for three lives—the farmer’s, his wife’s, and his son’s. When the
son took over the farm, the copyhold would be renewed and new lives added (his
wife and son) in exchange for a fee. By 1600, copyholds had therefore acquired a
substantial proprietary interest in the soil they tilled.
Allen (1992) refers to this process as the ‘rise of the yeoman.’ He argues that the
rise of the yeoman with a substantial proprietary interest in the soil was a key cause
of the Agricultural Revolution. Before this development, most land was farmed
by serfs. Since the serfs didn’t have secure property rights, they had very limited
incentives to innovate. If they did innovate, the lord was in a position to rob them
of the entire fruits of their innovation. The lord might want to promise to respect
property rights. But in this regard, he faced the commitment problem we discussed
in section 1.
The Civil War and subsequent period of turmoil marks the formal end of feu-
dalism in England. The first act of Parliament after the Restoration in 1660 abol-
ished the feudal duties of lords vis-a-vis the Crown (Hill, 1961, p. 127). This same
act, however, affirmed and secured the remaining rights of the lords vis-a-vis their
copyholders and leasees. It is therefore a significant step in the ‘modernization’ of
property rights in England.
The Civil War marked the high point of the ‘yeoman farmer’ in England. Over
the subsequent two hundred years, the yeoman farmer gradually disappeared as
the ownership of land in England was radically transformed. This process involved
both a massive increase in the concentration of land ownership—the rise of the
‘Great Estates’—and also a substantial increase in average farm size. The classic
yeoman farmer of the 17th century farmed a moderately sized farm (on the order of
60 acres) and used mostly family labor. By 1800, most farmland was organized into
farms that were much larger (on the order of 200 acres) and worked mostly by wage
labor. This transformation resulted in a rural society consisting of three classes:

53
wealthy landowners, large-scale tenant farmers, and landless laborers (a rural pro-
letariat). The shift from a feudal to a capitalist agrarian society therefore involved
both a transformation of property rights and a transformation of labor relations.
The causes of the massive increase in the concentration of landownership that
occurred in England in the 18th and 19th centuries have generated much debate
among economic historians. Allen (1992, ch. 5) proposes an interesting explanation:
the development of the modern mortgage. He argues that the traditional system
of modest rents and periodic fees used by lords in England can be thought of as a
way for the lords to borrow from their tenants: the lords received a large up-front
payment instead of a higher annual rent. This system was obviously inefficient
in that poor tenants were not natural sources of financing. But the lords had no
alternative absent a modern mortgage market.
The 17th century saw the gradual development of the modern mortgage. Before
that time, mortgages were very short term (six months) and extremely onerous for
debtors: if a debtor defaulted even by one day he forfeited his land to the creditor
but still owed the principal of the loan! Gradually, the royal courts allowed debtors
who had defaulted to recover their land upon repayment of the loan. The most sig-
nificant shift occurred after the Restoration. Many Royalists had at that point lost
their land to creditors (it had been confiscated by Parliament in the Civil War and
the Royalists had purchased it back with the help of mortgages that they subse-
quently defaulted on). The royal courts allowed these Royalists to repossess their
land upon repayment. By the end of the 17th century, it had become an established
part of the common law that debtors could automatically and indefinitely extend
their mortgages as long as they paid interest.
This financial innovation allowed lords to finance the purchase of small free-
holds. More importantly, it allowed them to cease renewing copyhold agreements
and instead let them ‘run out.’ The lord could then reorganize the land into rent-
maximizing (large) farms and lend them to tenants at market rates. This process
involved the lord forgoing the copyhold renewal fees that were otherwise a signifi-
cant source of his income. His ability to mortgage parts of his land allowed him to
maintain his prior level of consumption until the higher rents restored his income
(and paid off the mortgages).
The mortgage market no doubt significantly facilitated the concentration of land
in England and its reorganization into large capitalist farms. However, if this type
of transformation had always been profitable, it seems likely that it would have
occurred gradually through self-financed purchases and copyhold conversions by

54
lords even before the development of mortgages. The absence of a gradual process
of this type prior to the Civil War suggests that other institutional barriers were
important. The feudal rights of the Crown vis-a-vis lords and the threat of arbitrary
confiscatory actions by the Crown against successful lords along the lines discussed
in section 1 were likely important in this respect.
The idea that the Black Death caused the collapse of feudalism in England was
challenged by historian Robert Brenner in a wonderfully insightful and important
paper (Brenner, 1976). Brenner points out that while feudal serfdom declined in
Western Europe after the Black Death, it intensified in Eastern Europe. If labor
scarcity led to the collapse of serfdom in Western Europe, why did the same not
happen in Eastern Europe?
Acemoglu and Wolitzky (2011) present an interesting response to this argument.
They develop a theoretical model of labor coercion. In their model, lords use force
to coerce serfs into accepting unfavorable terms of ‘employment.’ The coercion is
modeled as the lords expending resources to reduce the outside option of the serfs.
Acemoglu and Wolitzky show that in their model an increase in the scarcity of labor
(e.g., due to a large plague) can either increase coercion or reduce it depending on
the relative importance of two effects. On the one hand, labor scarcity increases
the value of output which encourages increased coercion by lords. On the other
hand, labor scarcity may increase the outside option of the serfs which discourages
coercion by lords.
Acemoglu and Wolitzky’s model suggests that the divergent paths of serfdom in
Western and Eastern Europe after the Black Death are due to different relative effects
of this shock on the value of output and the outside option of serfs. As we discuss
above, certain peculiar aspects of feudalism in England led to a large improvement
in the outside option of serfs in that country after the Black Death: lords found them-
selves competing for tenants in an environment without a ‘Fugitive Serf Law.’ This
however begs the question why the lords in England did not collectively manage to
enact and enforce a Fugitive Serf Law and other repressive collective measures. In
Eastern Europe, collective action among the lords succeeded along these lines lim-
iting the increase in the outside option of serfs (see discussion in Brenner, 1976, and
Aston and Philpin, 1987). The concurrent expansion of Baltic and North Sea trade
may also have played an important role in these divergent paths by increasing the
demand for grain in Eastern Europe. Finally, recent palaeoecological research that
uses radiocarbon-dated pollen data from roughly 250 coring sites (lakes and wet-
lands) across Europe to assess changes in land use suggests that mortality dues to

55
the Black Death was lower in Eastern Europe than in Western Europe (Izdebski et al.,
2022).
Our discussion in this section has focused on the shift in England from feudal to
capitalist economic institutions. To understand why this shift occurred in Europe as
opposed to somewhere else in the world, it is useful to push the discussion further
back in time. Early human societies were tribal. Fukuyama (2011, ch. 16) discusses
the decline of tribal kinship based organization of society and the rise of individu-
alism in Europe. In tribal societies, property rights and personal freedom (e.g., to
marry) are typically severely constrained by a person’s status and various obliga-
tions towards family members. The object of these customs is often to guarantee
that the land and other assets of a kinship group remain within the group.
Fukuyama argues that the early decline of these kinship obligations and the
rise of individualism is unique to Europe. He furthermore argues—citing Goody
(1983)—that this shift was driven by the material interests of the Catholic Church.
In the sixth century, the church took a strong stand against various practices that
helped maintain property within families in Europe: marriage of close kin, mar-
riage of widows to relatives of their dead husbands, the adoption of children, and
divorce. The resulting decline in these practices lead to a large increase in widows
and spinsters without male heirs. At the same time the church strongly promoted
donations of land and property to itself. Within a short period, the church had be-
come the owner of a substantial fraction of productive land in Europe.
According to Fukuyama, a byproduct of these policies of the church was to
weaken kinship ties in Europe and strengthen the rights of individuals. Feudal-
ism arose to fill the void left by the decline in kinship based systems of protection
against violence. Much later, the rise of individual property rights and the decline
of feudalism was aided by the relative absence of a complex set of kinship based en-
tails. The feudal relationship of vassalage between lord and serf was a relationship
between individuals, not between a lord and a kin group.
Acemoglu and Robinson (2019) argue that Europe benefited from a fortuitous
fusion of two sets of institutions: state institutions inherited from the Roman Em-
pire and participatory norms and institutions inherited from the Germanic tribes
that overthrew the Western Roman Empire. They argue that the key to political de-
velopment is a Shackled Leviathan: a strong state that is held in check by a strong
society. This relies on a balance of power between the state and the people, which is
hard to obtain. In most places, the state is either too strong and becomes despotic or
social norms and customs limiting the emergence of centralized power—what they

56
call “the cage of norms”—are too strong in which case no state emerges.

4 High Wages and Cheap Coal

Perhaps the most curious aspect of the Industrial Revolution is how geographically
concentrated it was. With relatively minor exceptions, all the important industrial
innovations of the 18th and early 19th century were British inventions. These in-
clude the original atmospheric steam engine invented by Thomas Newcomen, John
Smeaton’s efficiency improvements of that engine, the separate condenser (James
Watt), the high pressure steam engine (Richard Trevithick), the flying shuttle (John
Kay), the spinning jenny (James Hargreaves), the water frame (Richard Arkwright),
the mule (Samuel Crompton), and the coke smelting furnace (Abraham Darby) to
name only a few of the most important innovations. Some of these inventions (most
notably the steam engine) relied on scientific discoveries that had mainly been made
on the continent. But industrial engineering innovation overwhelmingly occurred
in Britain. In short, the Industrial Revolution occurred in Britain.
Why did an explosion of industrial innovation all of a sudden occur in Britain in
the 18th century? In 1500, Britain was a relatively unimportant island off the coast
of Northern Europe with a very modest population in comparison to, say, France,
Northern Italy, the Lower Yangzi Delta in China, the Kanto plain in Japan, or Gu-
jarat in India. By 1800, however, Britain was in the midst of transforming itself into
an industrial powerhouse like no other the world had every seen, leaving all other
regions of the world far behind. Explaining the concentrated burst of industrial in-
novation in Britain in the 18th and early 19th centuries is a key element of explaining
the Industrial Revolution.
We have already discussed several potential contributors to the Industrial Rev-
olution: changes to institutions, the Enlightenment, and the Agricultural Revolu-
tion. Britain’s scorecard on being a “special place” when it comes to these three
potential contributors is uneven. Consider agriculture. Figure 11 shows that within
Europe, England was a typical place when it comes to agricultural productivity un-
til the 17th or 18th century. Belgium held a large lead in agricultural productivity
within Europe before 1700. But agricultural productivity was much higher still in
the Yangzi Delta (as evidenced by its huge population density). The Enlightenment
set Europe apart from other regions of the world intellectually. But it was experi-
enced by a large part of Europe as opposed to just Britain. Of the three potential

57
contributors, the one that most favors Britain is institutional change. Institutional
change in Britain over the centuries proceeding the Industrial Revolution—but par-
ticularly in the 17th century—resulted in much greater liberty and a more inclusive
and capitalist-oriented government in the 18th century than in most other times
and places. But is this enough to explain why the Industrial Revolution occurred in
Britain?
In a provocative and highly influential book, Robert Allen argues that while in-
creased liberty and a more inclusive and capitalist-oriented government may have
been an important contributor to the emergence of the Industrial Revolution, this
was not the whole story (Allen, 2009). According to Allen, a crucial additional fac-
tor that uniquely favored Britain was the fact that wages in Britain in the 18th cen-
tury were higher than in most other places in Europe and the price of energy was
lower. This combination of wages and prices gave entrepreneurs in Britain a strong
incentive to produce labor-saving innovations, and, in particular, to invent ways to
replace labor with machinery driven by cheap coal. Allen writes: “The Industrial
Revolution, in short, was invented in Britain in the eighteenth century because it
paid to invent it there, while it would not have been profitable to invent in other
times and places.”
Another influential idea, emphasized by historian Kenneth Pomeranz, is that the
huge relaxation of ecological constraints on growth that resulted from the discovery
of the New World and steam power where a crucial contributor to industrializa-
tion in Britain (Pomeranz, 2000). According to Pomeranz, “the British story ... is
unimaginable without two crucial discontinuities—one created by coal and one by
colonies.” A weakness of this idea is that it does not help explain the explosion of
innovation—growth in productivity—which is arguably a crucial feature of the In-
dustrial Revolution. In a world constrained by ecology (i.e., a Malthusian world),
a relaxation of ecological constraints results in population growth, but not growth
in productivity and real wages. We will discuss ways in which the ecological boon
associated with the discovery of the New World and the exploitation of fossil fuels
where extremely important. But since they don’t help explain the sudden accelera-
tion of technical progress, it is hard to see them as providing an explanation for the
Industrial Revolution.

58
4.1 Directed Technical Change
Allen’s (2009) argument discussed above may seem curious in that high wages are
said to have been a key advantage for British industry. How can this be? Shouldn’t
high wages, if anything, make British industry less competitive and thereby discour-
age investment in Britain? Simple economic models indeed have this prediction.
Those models, however, ignore the effects of high wages on innovation. Allen’s
argument is that high wages spurred labor-saving innovation. But if high wages
spur labor-saving innovation, doesn’t that mean that low wages spur innovation
that economizes on other factors of production (e.g., land and capital)? Is there any
reason to believe that the overall level of innovation should be higher when wages
are high than when wages are low?
These are difficult questions that are not fully addressed in Allen’s book. An
interesting paper by the economist Daron Acomoglu presents a model of “directed
technical change” that sheds some light on these questions (Acemoglu, 2002). Ace-
moglu’s model indeed has the implication that more innovation will be directed to
factors of production that are more expensive. High wages therefore induce labor-
saving innovation, while high land prices induce land-saving innovation. Whether
the overall level of technical progress is higher when wages are high or when land
prices are high, however, depends on a number of other aspects of reality. In par-
ticular, this depends on the “innovation possibilities frontier.” This is a fancy way
of saying that the cost and technical difficulty of making different types of innova-
tions matters for the direction and overall level of technical progress. For example,
if innovation that economizes on land or capital is for some reason very costly or
technically difficult, then conditions that favor economizing on land and capital but
discourage labor-saving innovation (i.e., low wages) will result in a low overall level
of technical progress.
One way to make sense of Allen’s argument is that industrialization had to fol-
low a certain particular path that began with labor-saving innovations. The pre-
industrial world featured very few machines. Innovation that economized on ma-
chines was therefore not really possible (there were no machines to economize on).
The initial phase of industrialization must therefore be labor saving: the invention
of machines. Once machines had been invented, capital-saving technical progress
became a possibility and indeed a great deal of such innovation occurred (e.g., the
massive efficiency improvements of steam engines and textiles machinery that we
discuss in more detail later in this section). But to get the whole thing started, ma-

59
chines needed to be invented and this could only happen in a high-wage economy
because it was only in a high-wage economy that it was profitable to invent the inital
(highly innefficient) machines.
What about land? Land certainly featured prominently in the pre-industrial
economy. Furthermore, our discussion of the Agricultural Revolution in section
3 indicates that innovation that economized on land was possible (e.g., intensive
crop rotations). An interesting feature of the cross-country data that we reviewed
in that section is that a subset of European countries underwent a substantial Agri-
cultural Revolution, while others did not. It so happens that the countries in which
agriculture was revolutionized were exactly the countries that were most exposed
to the growth in trade and commerce associated with the massive expansion of At-
lantic trade in the early modern period. One way to make sense of this cross-country
variation in the extent to which different countries experienced an Agricultural Rev-
olution is that it reflects the importance of directed technical change: The countries
that saw a rapid expansion of wealth and population associated with Atlantic trade
experienced a sharp increase in food demand. This increase in food demand and the
associated increase in food prices increased the profitability of technical change in
agriculture and therefore spurred on an Agricultural Revolution in these countries.
But if high food demand in England and the Netherlands in the 18th century
spurred the Agricultural Revolution, why didn’t high demand for food associated
with Malthusian population pressure in earlier times spur land-saving innovation?
It seems that such innovation did occur to a greater extent in Asia than in Europe.
Agricultural yields in certain regions of Asia (e.g., the lower Yangzi Delta) were
high enough to support dramatically higher population densities than anywhere in
Europe. Furthermore, substantial innovation was possible in Europe as evidenced
by the Agricultural Revolution. Why didn’t it occur earlier? Something must have
held back land-saving innovation in Europe. It is not well understood what this
was.

4.2 The Steam Engine, Coal, and the Energy Transformation

Before the Industrial Revolution, the possibility of sustained increases in the living
standards of ordinary workers was seriously hampered by ecological constraints:
Land was (and is) in virtually fixed supply; production was very land intensive;
and this meant that the economy faced sharply diminishing returns to labor. The
Malthus model discussed in chapter XX [Malthus chapter] explains this logic in de-

60
 Table 8: Energy Consumption in England and Wales
1560s 1700s 1750s 1800s 1850s
Farm Animals 21.1 32.8 33.6 34.3 50.1
Population 14.9 27.3 29.7 41.8 67.8
Firewood 21.5 22.5 22.6 18.5 2.2
Wind 0.2 1.4 2.8 12.7 24.4
Water 0.6 1.0 1.3 1.1 1.7
Coal 6.9 84.0 140.8 408.7 1,689.1
Total 65.1 168.9 230.9 517.1 1,835.3
Total less coal 58.2 84.9 90.1 108.4 146.2
Notes: Energy is measured in petajoules. The source of these estimates is Ta-
ble 2.1 of Wrigley (2010). The numbers for farm animals and the population
are estimates of calories consumed. Wind energy includes sail ships. The
firewood number for 1850s is actually for the 1840s.

tail. In such an environment, increases in technology lead to a larger population in

the long run, but not higher wages.
The invention of the steam engine radically altered this state of affairs by allow-
ing production to become much less land intensive than before. Before the Industrial
Revolution, it was not only food production that was land intensive. The produc-
tion of both heat and mechanical energy was also highly land intensive. Table 8
contains estimates of energy consumption in England and Wales between the 1560s
and the 1850s. In the 16th century, a very substantial portion of energy consump-
tion was firewood, mostly for heating. The source of a large majority of mechanical
energy was human and animal power, which required food as an input. Expansion
of industry with the associated increase in demand for energy would therefore have
required a great deal of extra land. The advent of steam engines powered by coal
dramatically changed this. Table 8 shows how the Industrial Revolution caused a
30-fold increase in energy consumption in England and Wales and an increase in the
share of coal in energy consumption from about 10% in the 1560s to more than 90%
in the 1850s. Clearly, this massive increase in energy consumption would not have
been possible without the use of coal.
Wrigley (2010) argues that the shift from an organic economy constrained by lim-
ited land to an energy-rich economy powered by coal (and later other fossil fuels)
was the most fundamental change that occurred over the course of the Industrial
Revolution. He furthermore argues that the key puzzle regarding the Industrial

61
Revolution is not how it got started but rather why it did not end quickly. The an-
swer to this is that the shift to fossil fuels as a source of energy dramatically relaxed
the ecological constraint that had previously held back growth in living standards.
From this point of view, the steam engine is by far the most transformative tech-
nological innovation of the Industrial Revolution, perhaps the innovation that rid
the economy of the crucial bottleneck holding back growth, and that thereby made
all the difference. For this reason, the story of how the steam engine was developed
is of particular interest. Was there something special about 18th century England
that resulted in the steam engine being invented then and there? Or was this just a
stroke of good luck that could have occurred elsewhere or at another time?
Thomas Newcomen’s first successful steam engine was put into service in Dud-
ley in 1712, where it drained a mine. It was an atmospheric engine in which steam
was condensed to form a vacuum under a piston. The piston was then pushed
down by the pressure of the atmosphere above it. This design was based on the
scientific discovery by Torricelli in the 17th century that the atmosphere had weight
and the subsequent discovery of Papin that condensing steam can produce a vac-
uum. Enlightenment advances in science are therefore one of the crucial precursors
to the invention of the steam engine. One important reason the Romans (for exam-
ple) didn’t invent the steam engine was that they didn’t have access to knowledge
of 17th century science.
But as Allen (2009) describes, Newcomen’s engine was not a simple application
of earlier scientific ideas. It was a highly complex machine (by the standards of
the time) and it’s invention required a major project of research and development.
It took Newcomen over a decade of experimentation and considerable expense to
arrive at a successful design. This large investment of time and money was worth-
while in Britain only because Britain had a large coal industry which created a sub-
stantial demand for drainage and because the huge amount of fuel needed to oper-
ate the highly inefficient early steam engines was virtually free in the coal fields.
In the 17th and 18th centuries, the British coal industry grew enormously—by
a factor of roughly 60—to become by far the largest in the world (Allen, 2009, p.
81-82). Allen argues that the cause of this huge growth was the rapid expansion
of London during this period that resulted from Britain’s success in international
trade. The growth of London led to a large increase in the demand for fuel for
heating purposes. Firewood was the traditional fuel used to heat houses. The large
increase in the demand for firewood caused a ‘timber crisis’ around London in the
17th century (Nef, 1932; Hatcher, 1993).

62
 16
Wood
14
Coal
12

10

0
1400 1450 1500 1550 1600 1650 1700 1750 1800
Figure 12: Real Prices of Wood and Coal in London
Note: This figure is based on Figure 4.3 in Allen (2009). Prices are decadal averages. The prices of
wood and coal are deflated by the consumer price index described in Allen (2001b). Allen reports
these prices in terms of the price level in Strasbourg in 1754-54. The prices in the figure should
therefore be interpreted as the price of wood and coal in grams of silver per million BTU if the
overall price level was that of Strasbourg in the mid 18th century.

Figure 12 presents the evolution of the real price of wood and coal in London
from 1400 to 1830—i.e., the price of wood and coal divided by a general consumer
price index. The price of wood roughly doubled in real terms over the course of the
17th century and stayed high thereafter. This large increase resulted from the explo-
sive growth in London’s population, which rose from 200,000 in 1600, to 575,000 in
1700, and 960,000 in 1800 (Wrigley, 2010).
High wood prices spurred demand for an alternative to firewood for heating
houses. The alternative that emerged was coal. But to switch from heating houses
with wood to coal was not a trivial change. The coal-burning house had to be in-
vented (Allen, 2009). Burning coal emits sulphurous fumes that render a house un-
inhabitable unless the fumes are vented out effectively. Furthermore, burning coal
requires an enclosed space with a generous draft to maintain both high heat and
substantial oxygen flow to the fire. The fire in a medieval house was typically on an

63
 10
9
Grams of Silver per million BTUs
8
7
6
5
4
3
2
1
0
Amsterdam London Paris Strasbourg Newcastle Beijing
Figure 13: Price of Energy in the Early 1700s
Note: This figure replicates Figure 4.1 in Allen (2009).

open hearth in the middle of a large room. For coal to be used for residential heat-
ing, narrow (and tapered) chimneys with an enclosed fireplace and grates under the
fuel were essential. Arriving at such a design involved a great deal of experimenta-
tion. The explosive growth of London over this period with the associated building
boom aided the needed experimentation.
The combination of high prices for firewood and low prices for coal were an
essential driver of the shift to coal-burning houses in London and therefore of the
rise of the coal industry in England. The rise of the coal industry provided large de-
mand for drainage, which helped induce the invention of the steam engine. Another
byproduct of the rise of the coal industry was that the price of energy in Northern
and Western England was by far the lowest in the world. Figure 13 presents the
price of energy in several European cities in the early 1700s. The price of energy in
London is based on coal and was roughly 5 grams of silver per million BTUs. This
price mostly reflects the cost of transport of coal from Newcastle to London. The
price of energy in Newcastle was below one gram of silver per million BTUs, much
lower than elsewhere in Europe. This low price of energy in Northern and Western
Britain helped spur not only the invention of the steam engine, but also other energy
intensive innovations including coal based metallurgy.
These developments—the rise of the coal industry and the associated low price
of energy—are important factors that set Britain apart even from other Western Eu-

64
ropean regions that shared many other advantages with Britain. Consider, in partic-
ular, the Dutch Republic: it was similarly exposed to the Enlightenment and to the
Reformation, it benefited from the growth in international trade, and it saw liberal
political reform. Together, these developments led to explosive growth of Amster-
dam and high wages as in Britain (see Figure 10). So, why was there no Industrial
Revolution in the Netherlands? This is a tough question and no fully satisfactory
answer has yet been developed. But one important difference is that Amsterdam
had a different backstop energy supply: peat. When Amsterdam grew, its popu-
lation increasingly switched to using peat for heating. This kept energy prices in
Amsterdam relatively low. But the increased use of peat did not have any of the
auxiliary longer-term benefits for technological innovation that coal mining had in
Britain.
Allen’s argument is that high wages and cheap coal created strong incentives for
technical change aimed at substituting energy and capital for labor in Britain dur-
ing the Industrial Revolution. Critics of this view have, however, pointed out that
a great many inventions in Britain over this period actually aimed to economize on
energy and capital. Allen counters this point by drawing a distinction between what
economic historian Joel Mokyr has called macro-inventions and micro-inventions
(Mokyr, 1990). Macro-inventions are inventions of something entirely new—such
as Newcomen’s steam engine—that then set in motion a long train of subsequent
improvements. Micro-inventions, on the other hand, to refer to the many incremen-
tal improvements on existing machines or techniques that aim to reduce cost and
improve efficiency. Allen (2009) argues that the critical macro-inventions that got
the the Industrial Revolution going—e.g., the steam engine, the spinning jenny, and
the coke smelting furnace—did indeed radically change the relative importance of
different factors of production by substituting energy and capital for labor. High
wages and cheap coal gave the British much stronger incentives to invent these
game-changing technologies than people in other regions.
Once the new technologies had been invented, however, a process of improving
them immediately began. The nature of the technical change involved in this pro-
cess of improvement was quite different from that of the original macro-inventions:
it involved improvements that saved all inputs to production, not the least fuel. The
history of the steam engine illustrates this well. The initial Newcomen steam engine
was extremely inefficient. It used an enormous amount of fuel for each horsepower-
hour of mechanical energy it produced. Figure 14 plots the evolution of the coal
consumption of pumping engines in the 18th and 19th century. The original New-

65
 50

45
Newcomen
40 Watt
35 Cornish Average
Cornish Best
30

25

20

15

10

0
1725 1750 1775 1800 1825 1850
Figure 14: Coal Consumption of Pumping Engines
Note: This figure replicates Figure 7.1 in Allen (2009). Coal consumption is measured in pounds per
horsepower-hour.

comen engine needed 45 pounds of coal to produce a single horsepower-hour of

mechanical energy. This staggering level of inefficiency meant that the only task
for which it could profitably be used was to drain coal mines (where coal was es-
sentially free). However, as the steam engine was made more energy efficient, the
scope of its use dramatically increased: first to draining other types of mines and
then more widely to powering machinery. The invention of the original steam en-
gine therefore shifted the incentives of inventors. Now they faced strong incentives
to invent technologies that improved energy efficiency.
The scale of the efficiency improvements of steam engines in the 18th and 19th
centuries is staggering. Coal consumption fell from 45 pounds per horsepower-hour
in the early 18th century to less than one pound per horsepower-hour in the late 19th
century (Allen, 2009, p. 164-165). Little is known about the early improvements in
the efficiency. The drop from 30 to 17.6 pounds per horsepower-hour in 1769 is due
to various improvements made by the engineer John Smeaton. The drop from 17.6 to
below 10 pounds per horsepower-hour was due to James Watt’s separate condenser.

66
 Table 9: Stationary Power Sources in Great Britain
1760 1800 1830 1870 1907
Steam 5,000 35,000 160,000 2,060,000 9,659,000
Water 70,000 120,000 160,000 230,000 178,000
Wind 10,000 15,000 20,000 10,000 5,000
Total 85,000 170,000 340,000 2,300,000 9,842,000
Notes: This table replicates Table 7.1 in Allen (2009).

The drop from just below 10 pounds in the 1790s to below 2 pounds per horsepower-
hour in the 1830s was due to several inventions. The most important of these was
the high-pressure engine of Richard Trevithick. Other important inventions include
the use of steam expansively and Arthur Woolf’s invention of the drop valve and
compounding.
The original Newcomen engine was not well suited for powering machinery be-
cause its rocking motion was too irregular. The first solution to this problem was to
combine the Newcomen engine with water power, i.e., use the Newcomen engine
to return water to an upstream reservoir so that it could pass over a water wheel
repeatedly. This technology was widely used in the second half of the 18th century.
In the 1780s, however, James Watt took up the challenge of creating a rotary steam
engine that produced smooth motion. One of his innovations was ’double-action’:
injecting steam alternatively into each end of the cylinder. This along with a reor-
ganized system of rods allowed for ’parallel motion’: the piston to both push and
pull the beam. Watt’s third important innovation was ‘sun and planet’ gears that
were used to rotate the drive shaft and double the rotation speed. Finally, there was
the ‘centrifugal governor’, which was used to stabilize the speed of the engine. The
result was a rotary steam engine that could power machinery.
However, Watt’s rotary steam engine was not very fuel efficient. It used 12-15
pounds of coal per horsepower-hour (Allen, 2009, p. 172). This is one reason for the
slow adoption of steam power in industry before 1830. Table 9 presents estimates of
the evolution of installed power in Britain by power source. Even in 1830, less than
half of installed power is steam power. Another reason for the slow adoptions of
steam is the impressive improvements in the efficiency of water power in the 18th
and early 19th centuries (Mokyr, 1990, p. 90-92). The major breakthroughs include
the breast wheel (John Smeaton, 1750s), the sliding hatch (John Rennie, 1780s), and
the water turbine (Benoit Fourneyron, 1837). In the 1840s, however, William Mc-

67
Naught added high pressure and compounding to Watt’s rotary engine design. This
cut the fuel consumption of rotary engines to 5 pounds per horsepower-hour. After
this, the use of steam power expanded rapidly and “the general mechanization of
British industry began” as Allen (2009) puts it.

4.3 Cotton
The cotton industry was the marvel of the early Industrial Revolution. Its growth
both in terms of productivity and output was breathtaking. Between 1770 and 1815,
the British cotton industry grew by 7% per year. This yielded total growth of roughly
2200% over this 45 year period (Harley, 1982; Crafts and Harley, 1992). Between
1780 and 1860, total factor productivity (TFP) in the British cotton industry grew by
2.6% per year and accounted for a quarter of all TFP growth in the British economy
(Crafts, 1985). This rapid growth in productivity led to a large fall in the relative
price of cotton cloth. Figure 15 plots the price of cotton (or linen) cloth relative to
the price of bread in England from 1690 to 1900. From 1750 to 1820, the relative price
of cotton cloth fell by a factor of ten.
The story of the growth of the cotton industry is one of mechanization of an ex-
tremely labor-intensive activity. Table 10 reports estimates constructed by Robert
Allen of the breakdown of the cost of producing a pound of relatively coarse cot-
ton yarn using different technologies. The first column provides this breakdown
for traditional hand methods used prior to the onset of mechanization. Using these
traditional methods, Allen estimates that a cotton spinner could spin roughly one
pound per day and was payed 7 pence for this. Preparing the cotton for spinning by
cleaning it and carding it into rovings took a comparable amount of time. In addi-
tion to this, the “putting out” system of production—where most spinning occurred
in people’s homes as a part-time activity alongside farming and other activities—
involved a substantial amount of administration. Allen estimates the total labor
costs of producing a pound of cotton yarn using these traditional methods to be
17.19 pence, a bit more than the cost of the raw cotton. In sharp contrast, these tra-
ditional methods involved a minimal amount of capital expenses: 0.93 pence per
pound according to Allen’s estimate. These estimates therefore imply that the labor
share of value added in the production of cotton yarn before mechanization was
95% (17.19/(17.19+0.93) = 0.949).
The most famous innovations in the cotton industry were to spinning: James
Hargreaves’ spinning jenny (1764), Robert Arkwright’s water frame (late 1760s),

68
 12

10

0
1690 1710 1730 1750 1770 1790 1810 1830 1850 1870 1890
Figure 15: Prices of Cotton (or Linen) Cloth Relative to Bread in England
Note: This figure is based on Figure 4 in Allen (2009). Prices are decadal averages.

and Samuel Crompton’s mule (1779). But many other aspects of textile produc-
tion were also thoroughly transformed by mechanization. The carding machine
patented by Robert Arkwright—an improvement on an earlier machine invented
by Lewis Paul—improved labor productivity in the production of cotton yarn by a
similar amount as Arkwright’s more famous water frame. Other important inno-
vations include the cotton gin (Eli Whitney) for separating cotton fibers from seeds,
bleaching (Claude Berhollet), metal printing (Thomas Bell), and the power loom in
weaving (Edmund Cartwright). Mokyr (1990, p. 96-103) and Allen (2009, ch. 8)
provide a more detailed discussion of innovation in textiles during this period.
Most if not all of these innovations represented “biased” technical change in the
sense that they disproportionately saved labor. This can be clearly seen for the in-
novations involved in spinning in Table 10. Comparing the first and second column
shows the impact of the spinning jenny. According to Allen’s estimates, it lowered
the cost of spinning by 2/3, from 7 pence per pound to 2.33 pence per pound. The
spinning jenny was a much more expensive machine than the traditional spinning
wheel. The cost of capital therefore rose, but only from 0.93 to 1.88 pence per pound.

69
 Table 10: Real Cost of Cotton Yarn (16 count)
Hand Method 24-Spindle Arkwright Glasgow
Jenny Mill Mill
1760 1775 1784 1836
Labor:
Cleaning and Carding 7.00 7.00 2.69 0.16
Spinning 7.00 2.33 2.57 0.34
Reeling, bundling, etc. 0.47 0.47 2.19 0
Administrative 2.72 2.72 0.41 0.02
Total Labor 17.19 12.52 7.86 0.52
Materials:
Raw Cotton 16.88 16.88 16.88 16.70
Other 0 0 1.20 0.53
Capital 0.93 1.88 2.00 0.47
Total Cost 35.00 31.28 27.94 18.22
Labor Share of Value Added 95% 87% 80% 53%
Notes: Costs are reported pence per pound at 1784 prices. The data are from Table 8.1 of Allen (2009).
The “count” of cotton yarn measures the number of hanks (770m) of yarn that make one pound of
yarn. The higher is the count, the finer the yarn. 16 count yarn is relatively coarse, similar to the yarn
used for modern jeans.

Clearly, the spinning jenny substituted capital for labor in the production of cotton
yarn. The overall cost saving was 3.72 pence per pound and the labor share of value
added fell from 95% to 87%.
Arkwright’s water frame reduced the cost of spinning by roughly the same
amount as the spinning jenny (column 3 of Table 10). Arkwright’s other innova-
tions, however, reduced the cost of cotton yarn a substantial amount further. In
addition to inventing the water frame and improving the carding machine, Ark-
wright also invented the cotton mill, i.e., the factory. Factory production allowed
for more efficient use of power sources such as water wheels and later steam en-
gines. It eliminated the administrative labor associated with the putting-out system
of domestic production. It also allowed for more discipline in the workplace (Clark,
1994). Workers that worked in a domestic setting or in workshops without disci-
pline tended to work irregular hours and do a substantial amount of socializing
at work. Arkwright imposed strict discipline in his cotton mills: workers were re-
quired to show up on time, work certain hours, work at a constant rate, and refrain

70
from socializing. This raised their productivity and allowed for more efficient use of
machinery. However, workers disliked discipline and demanded a wage premium
to work under such conditions. Factories with discipline and workshops without
discipline coexisted throughout the 19th century. Eventually, however, discipline
came to dominate the workplace (although often in a less extreme form than in the
early cotton mills).
The third column in Table 10 presents a breakdown of the cost of producing a
pound of cotton yarn in one of Richard Arkwright’s cotton mills in the mid 1780s.
The cost of spinning (using water frames) is roughly 1/3 of its value with traditional
methods. The cost of cleaning and carding (using Arkwright’s carding machine) is
also roughly 1/3 of its value prior to mechanization. There are also large savings of
administrative labor from moving production to a factory setting. Total labor costs
have fallen to only 7.86 pence per pound. Capital costs are 2 pence per pound, only
slightly higher than in column 2, but there are some extra materials costs. Overall,
Arkwright’s mill reduced the cost of producing yarn by roughly 7 pence per pound
relative to traditional methods. This equals 20% of the traditional cost of producing
yarn, but roughly 40% of the traditional cost of yarn less the cost of raw cotton.
Arkwright’s innovations were highly biased towards saving labor. The labor share
in value added in his mill was 80%, much lower than the 95% labor share using
traditional methods.
The early inventions of the spinning jenny, water frame, and carding machine
set in motion a torrent of further innovation that drove costs lower and lower over
time. The most important of these inventions was Samuel Crompton’s mule, which
combined the best features of the spinning jenny and the water frame. In addition
to improving efficiency, the mule allowed much finer yarn to be spun using a mech-
anized process and thus reduced the cost of high-count yarn by enormous amounts.
The final column of Table 10 presents a breakdown of the cost of producing yarn in
a cotton mill in Glasgow in 1836. This mill used the self-acting mule—which had
been invented by Richard Roberts in the 1820s—and was vastly more efficient than
Arkwright’s mill from 1784 along all dimensions. The total cost of a pound of cot-
ton yarn was down to 18.22 pence per pound, with 16.70 of that being the cost of
raw cotton. The improvements between the mid 1780s and the mid 1830s had been
strongly labor saving just like the early inventions. This is evident from the fact that
the labor share of value added had fallen from 80% to only 53%.
The vast majority of the innovation in spinning cotton occurred in Britain. Fur-
thermore, even though the technology became widely known in other countries

71
quite quickly and adoption was in some cases even subsidized by foreign govern-
ments, widespread adoption in other countries lagged behind Britain by roughly 40
years. Why did all this innovation happen in Britain and not (say) in France? Allen
argues that the crucial reason is that wages were much higher in Britain. Higher
wages in Britain made the operation of these more capital intensive (labor saving)
technologies more profitable in Britain than in France. In short, he argues that in-
venting and using the spinning jenny, water frame, and mule was profitable in
Britain but was not profitable in France, since wages were much lower in France.
Allen (2009, p. 203) argues that the annual profit rate of an Arkwright-type mill in
Britain in the 1780s was 40%, while in France it was only 9% (lower than the 15%
rate of return fixed capital in business at the time).
Britain had an additional important advantage over France: a large watch-
making industry. The water frame and mule required precision parts (gears) to
operate. Arkwright employed a large number of watch-makers to build his ma-
chines. This may be one reason why the cotton industry arose in Lancashire: most
of the world’s watch movements were made in Lancashire at the time. France did
not have a comparable watch-making industry. It is therefore likely that the cost
of building the needed machines was substantially higher in France than in Lan-
cashire. Landes (2000) argues that the substantial size of the watch-making industry
in Britain was due to high wages in Britain that led to high demand for clocks and
watches.
Allen’s high wage hypothesis is controversial. Kelly, Mokyr, and O Grada (2014)
argue that British wages were higher in the 18th century because British workers
were more productive. If true, this implies that unit labor costs were no higher
in Britain than in France, undermining Allen’s thesis of high labor costs in Britain.
Kelly, Mokyr, and O Grada present evidence on relative productivity in reaping and
threshing—two physically demanding, low skill, agricultural activities. These esti-
mates indicate that British workers were about 70% more productive than French
workers, enough to cancel out a large part of the difference in wages. Whether this
difference is due to differences in the inherent productivity of the workers (as Kelly,
Mokyr, and O Grada argue) or the incentives the workers faced is not clear. Our
discussion of agriculture in section 3 suggests that incentives might have mattered.
The main channel Kelly, Mokyr, and O Grada suggest for higher British productiv-
ity of unskilled workers is better nutrition. But evidence from modern developing
countries does not support the notion that higher wages increase the productivity
of poor workers enough to prevent unit labor costs from rising (Banerjee and Du-

72
flo, 2011). Furthermore, it is not clear that these productivity differences existed in
domestic cotton spinning—a sedentary occupation performed largely by women in
their homes.
Gragnolati, Moschella, and Pugliese (2011) argue that Allen understates the prof-
itability of the spinning jenny in France by assuming that output per worker remains
constant and hours fall when the spinning jenny is adopted as opposed to assuming
that hours stay constant and output rises (or something in between). Their calcula-
tions suggest that under the assumption that hours stay constant, the spinning jenny
was also profitable in France. This critique does not, however, dispute the notion
that the spinning jenny was more profitable in Britain. The overall level of profitabil-
ity is hard to pin down. There are no doubt many factors left out of the relatively
simple calculations that Gragnolati, Moschella, and Pugliese do (and also those that
Allen does). The fact that the spinning jenny was (substantially) more profitable in
Britain therefore continues to provide a possible reason why it was invented and
adopted there and not in France.
Stephenson (2018) argues that the English wage data that Allen (and other eco-
nomic historians) have used for the the 18th century overstate actual wages. The
issue is that the underlying data are payments to contractors as opposed to wages
earned by worker. Stephenson (2018) argues that the actual wages are lower by 20
to 30 percent due to markups by the contractors. Allen (2019) disputes this. But
even if it is true, it does not necessarily imply a smaller wage premium in England
relative to (say) France. This depends on whether French data also suffers from the
same problem, which is unfortunately not clear. Suppose, however, that Stephen-
son is correct and the English wage premium is say 20% smaller than Allen’s data
suggest. Even in this case, Stephenson’s adjustment only closes a relatively modest
portion of the wage gap between London and Paris. Allen’s estimates of respectabil-
ity ratios in different European cities (Figure 10) indicate that the difference in real
wages between London and Paris in 1750 was 120%.8

5 Concluding Thoughts
The Industrial Revolution and its antecedents are a complex topic about which
much is still poorly understood. Scholars disagree strongly about many important
issues and scholarship is rapidly evolving. It is therefore with some trepidation that
I venture to provide a summary of my own views about this topic based on my

73
study of the literature over the past few years.9 To the extent that there is a water-
shed event that made the Industrial Revolution possible, I think it is the invention
of movable type printing by Johannes Gutenberg around 1450. This event radically
altered the balance that had existed between forces seeking to create and spread
knowledge on the one hand and forces that destroyed knowledge and suppressed
new ideas on the other hand. Movable type printing gave the forces seeking to cre-
ate and spread knowledge a decisive upper hand and set in motion an intellectual
transformation: first the Reformation, then the Enlightenment and Scientific Revo-
lution, and ultimately the Industrial Revolution.
Changes to institutions in the leadup to the Industrial Revolution were surely
very important. But my views on this differ from North and Weingast (1989) in
two significant ways. First, I think North and Weingast overemphasize the Glorious
Revolution. I am sympathetic to Allen’s (1992) narrative regarding the importance
of gradual expansion of liberty and property rights in England over the 500 year pe-
riod leading up to the Glorious Revolution. Second, it is not clear to me that the pri-
mary growth enhancing characteristic of the post Glorious Revolution regime was
more secure property rights. I would rather point to Parliament’s policy of facili-
tating the reorganization and rationalization of property rights through enclosures,
statutory authorities, and estate acts (Bogart and Richardson, 2011). The feudal sys-
tem of property rights that existed before was likely woefully inefficient. The post
Glorious Revolution Parliament can be viewed as a new technology to facilitate re-
organization (lower transaction costs in the language of the Coase Theorem).
Finally, I am sympathetic to the view that directed technical change was impor-
tant. First, Robert Allen’s (2009) ideas about high wages and cheap coal in Britain
creating unusually strong incentives for the invention of labor saving devices are in
my view the most plausible existing theory for why the Industrial Revolution was so
extremely geographically concentrated in Britain. Second, the fact that Agricultural
Revolutions occurred exactly in the places that had the highest wages (Britain and
the Netherlands) exactly when the demand for food in these places rose, suggests
to me that the Agricultural Revolution was another example of directed technical
change.

74
Notes
1
The labor demand curve is a relationship between wages and labor supplied, not wages and the
population. Changes in labor supplied will differ from changes in the population if days worked
per year vary over time or if the fraction of the population that works varies. Humphries and Weis-
dorf (2019) estimate how days worked per year evolved over time in England from 1260 to 1850 by
combining data on earning of workers on annual contracts and data on wages of day laborers. They
document substantial changes in days worked per year over the period 1300 to 1640: days worked
per year dropped substantially in the 14th century (especially after than Black Death) and then rose
gradually after that. Days worked were quite similar in 1300 and 1640. The conclusion I reach in the
text based on wage and population data is therefore robust to allowing for variation in days worked
per year over this period.
2
See Hill (1940) and Hill (1961) for a more detailed account of the revolutions of 17th century
England and a Marxist perspective on their economic causes.
3
An empirical challenge Jha must overcome is that ownership in an overseas joint stock company
may be correlated with other characteristics that cause support for the rebellion. Jha (2015) shows
that owners of shares are very similar to non-owners along a range of characteristics. He also exploits
variation in propensity to own shares resulting from coincidence of the timing of an initial public
offering of overseas shares and the time when an individual turned 21 and was able to sign a legally
binding contract to purchase such shares.
4
The calculations in this paragraph make use of estimates of nominal English GDP from Broad-
berry et al. (2015) and estimates of the popluation of England from Clark (2010). For compatibility
with the profit numbers reported by Acemoglu, Johnson, and Robinson (2005), the GDP estimates
are adjusted to 1600 prices using the index of building craftsmen’s nominal wages from Clark (2005).
English GDP in 1630s was £35 million. English population in the 1630s was 5.2 million. This means
that GDP per person was roughly £7.
5
Gutenberg did not invent the printing press in a single year. He finished the first copies of his
famous Bible—his first major work—in approximately 1454. But sources vary even as to the exact
completion date for the Gutenberg Bible. By the time Gutenberg finished printing the Bible, he
had been working on his technology for many years. He had several rounds of “venture capital
funding” for at least 15 years prior to publishing the Bible. In numerous cases, his investors sued
him for failing to deliver a product to market or related infractions. Much of what is known about
Gutenberg’s life is from these lawsuits. Boorstin (1985, ch. 63) provides an excellent discussion of the
technical challenges Gutenberg overcame and the long and arduous process involved.
6
Cressy (1980) reports the fraction of witnesses who sign ecclesiastical records. He reports this
data broken down by several different occupations and locations. Houston (1982) reports the fraction
of witnesses who sign court depositions by gender but not by occupation or location. I pool the data
from Cressy and Houston and run a regression with time fixed effects, location fixed effects, and
occupation fixed effects. I use a separate location fixed effect for the data from Houston and treat the
male and female data from Houston as two separate “occupations.” Cressy also reports literacy rates
for women, which I also treat as a separate “occupation.” Figure 7 reports the time fixed effects from
this regression when the occupation “husbandmen” and the location “London/Middlesex” are the
omitted categories.
7
Allen has since made some adjustments to the way he calculates these baskets. Fore a discussion,
see Allen (2015).
8
Other critiques include Humphries (2013) and Humphries and Schneider (2019). Allen’s re-
sponses can be found in Allen (2015) and Allen (2020). Both responses are in my view convincing.
9
This summary is written in the spring of 2020.

75
References
A CEMOGLU , D. (2002): “Directed Technical Change,” Review of Economic Studies, 69,
781–809.
A CEMOGLU , D., S. J OHNSON , AND J. R OBINSON (2005): “The Rise of Europe: At-
lantic Trade, Institutional Change, and Economic Growth,” American Economic Re-
view, 95, 546–579.
A CEMOGLU , D. AND J. A. R OBINSON (2019): The Narrow Corridor, New York, NY:
Penguin Press.
A CEMOGLU , D. AND A. W OLITZKY (2011): “The Economics of Labor Coercion,”
Econometrica, 79, 555–600.
A LLEN , R. C. (1992): Enclosure and the Yeoman, Oxford, UK: Clarendon Press.
——— (2000): “Economic Structure and Agricultural Productivity in Europe, 1300-
1800,” European Review of Economic History, 3, 1–25.
——— (2001a): “Community and Market in England: Open Fields and Enclosure
Revisited,” in Communities and Markets in Economic Development, ed. by M. Aoki
and Y. Hayami, Oxford, UK: Oxford University Press, 42–69.
——— (2001b): “The Great Divergence in European Wages and Prices from the Mid-
dle Ages to the First World War,” Explorations in Economic History, 38, 411–447.
——— (2008): “The Nitrogen Hypothesis and the English Agricultural Revolution:
A Biological Analysis,” Economic Journal, 68, 182–210.
——— (2009): The British Industrial Revolution in Global Perspective, Cambridge, Eng-
land: Cambridge University Press.
——— (2015): “The High Wage Economy and the Industrial Revolution: A Restate-
ment,” Economic History Review, 68, 1–22.
——— (2019): “Real Wages Once More: A Response to Judy Stephenson,” Economic
History Review, 72, 738–754.
——— (2020): “Spinning Their Wheels: A Reply to Jane Humphries and Benjamin
Scheider,” Economic History Review, forthcoming.
A SHTON , T. S. (1948/1997): The Industrial Revolution, Oxford, UK: Oxford Univer-
sity Press.
A STON , T. AND C. P HILPIN (1987): The Brenner Debate: Agrarian Class Structure and
Economic Development in Pre-Industrial Europe, Cambridge, UK: Cambridge Uni-
versity Press.
B ACON , F. (1620/1999): “The New Organon: Or, True Directions Concerning the
Interpretation of Nature,” in Selected Philosophical Works, ed. by R.-M. Sargent,
Indianapolis, IN: Hackett.

76
B AIROCH , P., J. B ATOU , AND P. C H ÈVRE (1988): La population des villes europeenees
de 800 a 1850: Banque de donnees et analyse sommaire des resultats. (The Population
of European Cities, 800-1850: Data Bank and Short Summary of Results.), Centre of
International Economic History Series, No. 2, Geneva: Li: Librarie Droz.
B ANERJEE , A. V. AND E. D UFLO (2011): Poor Economics, New York, NY: Public Af-
fairs.
B ARRO , R. J. (1987): “Government Spending, Interest Rates, Prices, and Budget
Deficits in the United Kingdom, 1701-1918,” Journal of Monetary Economics, 20,
221–247.
B ECKER , S. O. AND L. W OESSMANN (2009): “Was Weber Wrong? A Human Capital
Theory of Protestant Economic History,” Quarterly Journal of Economics, 124, 531–
596.
B OGART, D. (2014): “The Transport Revolution in Industrializing Britain: A Sur-
vey,” in Cambridge Economic History of Britain, 1700-1870, ed. by R. Floud and
J. Humphries, Cambridge, UK: Cambridge University Press.
B OGART, D. AND G. R ICHARDSON (2011): “Property Rights and Parliament in In-
dustrializing Britain,” Journal of Law and Economics, 54, 241–274.
B OORSTIN , D. (1985): The Discoverers, New York, NY: Random House, Vintage
Books edition.
B RENNER , R. (1976): “Agrarian Class Structure and Economic Development in Pre-
Industrial Europe,” Past & Present, 70, 30–75.
B ROADBERRY, S., B. M. C AMPBELL , A. K LEIN , M. O VERTON , AND B. VAN
L EEUWEN (2015): British Economic Growth 1270-1870, Cambridge, UK: Cambridge
University Press.
C ARD , D. (1999): “The Causal Effect of Education on Earnings,” in Handbook of Labor
Economics, Voluma 3A, ed. by O. Ashenfelter and D. Card, Amsterdam, Holland:
Elsevier, 1801–1861.
C LARK , G. (1994): “Factory Discipline,” Journal of Economic History, 54, 128–163.
——— (1996): “The Political Foundations of Modern Economic Growth: England,
1540-1800,” Journal of Interdisciplinary History, 26, 563–588.
——— (2002): “Land Rental Values and the Agrarian Economy: England and Wales,
1500-1914,” European Review of Economic History, 6, 281–308.
——— (2005): “The Condition of the Working Class in England, 1209-2004,” Journal
of Political Economy, 113, 1307–1340.
——— (2007): A Farewell to Alms, Princeton, NJ: Princeton University Press.
——— (2010): “The Macroeconomic Aggregates for England 1209-1869,” Research in
Economic History, 27, 51–140.

77
C LARK , G. AND P. A. L EVIN (2011): “How Different Was the Industrial Revolution?
The Revolution in Printing, 1350-1835,” Unpublished Manuscript, University of
California, Davis.
C RAFTS , N. F. (1985): British Economic Growth During the Industrial Revolution, Ox-
ford, UK: Clarendon Press.
C RAFTS , N. F. AND C. K. H ARLEY (1992): “Output Growth and the British Indus-
trial Revolution: A Restatement of the Crafts-Harley View,” Economic History Re-
view, 45, 703–730.
C RESSY, D. (1980): Literacy and the Social Order: Reading and Writing in Tudor and
Stuart England, Cambridge, UK: Cambridge University Press.
DE V RIES , J. (1976): The Economy of Europe in the Age of Crisis, 1600-1750, Cambridge,
UK: Cambridge University Press.
D ICKSON , P. (1967): The Financial Revolution in England: A Study in the Development
of Public Credit, 1688-1756, New York, NY: St. Martin’s Press.
D ITTMAR , J. (2011): “Information Technology and Economic Change: The Impact
of the Printing Press,” Quarterly Journal of Economics, 126, 1133–1172.
E RNLE , L. (1912/1961): English Farming: Past and Present, London, UK: Heinemann
Educational Books and Frank Cass and Company.
F UKUYAMA , F. (2011): The Origins of Political Order, New York, NY: Farrar, Straus,
and Giroux.
G IBBONS , R. (1992): Game Theory for Applied Economists, Princeton, NJ: Princeton
University Press.
G OODY, J. (1983): The Development of the Family and Marriage in Europe, New York,
NY: Cambridge University Press.
G RAGNOLATI , U., D. M OSCHELLA , AND E. P UGLIESE (2011): “The Spinning Jenny
and the Industrial Revolution: A Reappraisal,” Journal of Economic History, 71,
455–460.
H ARLEY, C. K. (1982): “British Industrialization Before 1841: Evidence of Slower
Growth During the Industrial Revolution,” Journal of Economic History, 42, 267–
289.
H ATCHER , J. (1993): The History of the British Coal Industry, Volume I, Before 1700:
Towards the Age of Coal, Oxford, UK: Oxford University Press.
H ILL , C. (1940): The English Revolution 1640, London, UK: Lawrence and Wishart
Ltd.
——— (1961): The Century of Revolution, New York, NY: Van Nostrand Reinhold
(International) Co. Ltd.
H OUSTON , R. (1982): “The Development of Literacy: Northern England, 1640-

78
 1750,” Economic History Review, 35, 199–216.
H UMPHRIES , J. (2013): “The Lure of Aggregates and the Pitfalls of the Patriarchal
Perspective: Critique of the High Wage Economy Interpretation of the British In-
dustrial Revolution,” Economic History Review, 66, 693–714.
H UMPHRIES , J. AND B. S CHNEIDER (2019): “Spinning the Industrial Revolution,”
Economic History Review, 72, 126–155.
H UMPHRIES , J. AND J. W EISDORF (2019): “Unreal Wages? Real Income and Eco-
nomic Growth in England, 1260-1850,” Economic Journal, forthcoming.
I ZDEBSKI , A., P. G UZOWSKI , R. P ONIAT, AND ET AL . (2022): “Palaeoecological
Data Indicates Land-Use Changes Across Europe Linked to Spatial Heterogene-
ity in Mortality During the Black Death Pandemic,” Nature Ecology & Evolution,
https://doi.org/10.1038/s41559-021-01652-4.
J HA , S. (2015): “Financial Asset Holdings and Political Attitudes: Evidence from
Revolutionary England,” Quarterly Journal of Economics, 130, 1485–1545.
K ANT, I. (1784/1986): “An Answer to the Question: What is Enlightenment?” in
Philosphical Writings, New York, NY: Continuum, 263–269, Lewis White Beck,
translation.
K ELLY, M., J. M OKYR , AND C. O G RADA (2014): “Precocious Albion: A New In-
terpretation of the British Industrial Revolution,” Annual Review of Economics, 6,
363–389.
K INDLEBERGER , C. P. (1993): A Financial History of Western Europe, Second Edition,
Oxford, UK: Oxford University Press.
L ANDES , D. S. (2000): Revolution in Time: Clocks and the Making of the Modern World,
London, UK: Viking.
M ARSHALL , P. (2009): The Reformation: A Very Short Introduction, Oxford, UK: Ox-
ford University Press.
M C C LOSKEY, D. (2006): The Bourgeois Virtues: Ethics for an Age of Commerce, Chicago,
IL: University of Chicago Press.
——— (2010): Bourgeois Dignity: Why Economics Can’t Explain the Modern World,
Chicago, IL: University of Chicago Press.
——— (2016): Bourgeois Equality: How Ideas, Not Capital or Institutions, Enriched the
World, Chicago, IL: University of Chicago Press.
M C C LOSKEY, D. M. (1972): “The Enclosure of Open Fields: Preface to a Study of
the Impacts of the Efficiency of English Agriculture in the Eighteenth Century,”
Journal of Economic History, 32, 15–35.
M C E VEDY, C. AND R. J ONES (1978): Atlas of World Population History, New York,
NY: Facts on File, Inc.

79
M OKYR , J. (1990): The Lever of Riches, New York: Oxford University Press.
——— (2009): The Enlightened Economy: An Economic History of Britain 1700-1850,
New Haven: Yale University Press.
N EF, J. U. (1932): The Rise of the British Coal Industry, London, UK: George Routledge
& Sons.
N ORTH , D. C. AND B. R. W EINGAST (1989): “Constitutions and Commitment: The
Evolution of Institutions Governing Public Choice in Seventeenth Century Eng-
land,” Journal of Economic History, 49, 803–832.
N URKSE , R. (1953): Problems of Capital Formation in Underdeveloped Countries, Oxford,
UK: Blackwell.
O’B RIEN , P. (1988): “The Political Economy of British Taxation, 1660-1815,” Eco-
nomic History Review, 41, 1–32.
P HELPS B ROWN , H. AND S. V. H OPKINS (1955): “Seven Centuries of Building
Wages,” Economica, 22, 195–206.
P OMERANZ , K. (2000): The Great Divergence: China, Europe, and the Making of the
Modern World, Princeton, NJ: Princeton University Press.
R OSENTHAL , J.-L. (1990): “The Development of Irrigation in Provence, 1700-1860:
The French Revolution and Economic Growth,” Journal of Economic History, 50,
615–638.
R UBIN , J. (2014): “Printing and Protestant: An Empirical Test of the Role of Printing
in the Reformation,” Review of Economics and Statistics, 96, 270–286.
S CHMELZING , P. (2020): “Eight Centuries of Global Real Rates, R-G, and the
’suprasecular’ decline, 1311-2018,” Bank of England Working Paper 845.
S TEPHENSON , J. Z. (2018): “’Real’ Wages? Contractors, Workers, and Pay in London
Building Trades, 1650-1800,” Economic History Review, 71, 106–132.
T HOMAS , R. AND N. D IMSDALE (2017): “A Mille-
nium of UK Data,” Bank of England OBRA dataset,
http://www.bankofengland.co.uk/research/Pages/onebank/threecenturies.aspx.
T IMMER , C. P. (1969): “The Turnip, The New Husbandry, and the English Agricul-
tural Revolution,” Quarterly Journal of Economics, 83, 375–395.
W EBER , M. (1905/2002): Protestant Ethic and the Spirit of Capitalism, New York, NY:
Penguin Books, Peter Baehr and Gordon C. Wells, translation.
W RIGLEY, E. A. (1985): “Urban Growth and Agricultural Change: England and
the Continent in the Early Modern Period,” Journal of Interdisciplinary History, 15,
683–728.
——— (2010): Energy and the English Industrial Revolution, Cambridge, UK: Cam-
bridge University Press.

80