Bones, Bacteria and Break Points: The Heterogeneous Spatial
Effects of the Black Death and Long-Run Growth∗
Remi Jebwab†
George Washington University
Noel D. Johnson‡
George Mason University
Mark Koyama§
George Mason University
February 11, 2016
Preliminary and Incomplete
Abstract: The Black Death killed approximately 40% of Europe’s population between
1347-1353. Recent studies suggest that this mortality shock played a major role in shifting
Europe onto a path to sustained economic growth. Using a novel dataset that provides
information on spatial variation in plague mortality, we explore the short-run and long-run
impact of the Black Death on city growth. We find evidence for aggregate convergence. On
average, Europe’s cities recovered their pre-Black Death population within two centuries.
However, there was considerable heterogeneity in the response to the shock. The Black
Death led to the creation of new cities in areas that were relatively less urbanized before
it hit. Furthermore, the Black Death led to an urban reset: cities with better geographical
and non-geographical endowments relatively benefited, while other cities collapsed. Our
analysis thus suggests that the Black Death may have permanently affected the aggregate
level and spatial distribution of economic activity, potentially contributing to long-run
growth in Europe.
Key words: Black Death; Multiple Equilibria; Urbanization; Little Divergence; Epidemics; Path
Dependence; Urban Reset; Long-Run Growth
JEL classification: D5, J1, N00, O1, R1
∗
We are grateful to Ran Abramitzky, Dominick Bartelme, Hoyt Bleakley, Paul David, Mark Dincecco, Jeremiah Dittmar,
James Fenske, Walker Hanlon, Avner Greif, Saumitra Jha, Tony Yates, Anthony Yezer and seminar audiences at the George
Mason-George Washington Economic History Workshop (Spring 2015), George Mason University’s Public Choice Seminar,
George Washington University, the University of Michigan, Stanford University’s Economic History Workshop, and WEHC
(Kyoto 2015) for helpful comments. We gratefully acknowledge the support of the Institute for International Economic
Policy and the Elliott School of International Affairs at GWU and the Mercatus Center at GMU. We thank Trey Dudley,
Zhilong Ge, and Jessi Troyan for research assistance.
†
Corresponding Author: [email protected] George Washington University, Department of Economics, 2115 G Street,
NW, Washington, DC 20052, USA.
‡
[email protected]. Department of Economics, George Mason University, VA 22030.
§
[email protected]. Department of Economics, George Mason University, VA 22030.
1. INTRODUCTION
The Black Death was the greatest demographic shock in European history: approximately 30–40% of
the population died in just 5 years. In recent accounts of the Great Divergence, the Black Death plays
a crucial role in explaining the economic rise of western Europe (e.g. Acemoglu & Robinson, 2012;
Voigtländer & Voth, 2013b,a). However, to our knowledge, no existing research makes use of actual
city-level data on the demographic impact of the Black Death to study subsequent local economic
development across all of western Europe. This is the objective of this paper.
We construct a novel dataset combining estimates of Black Death mortality, city populations, and
numerous geographic, economic and institutional variables to study both the short-run and long-run
impact of the Black Death on urban growth and hence on economic development. These data allow
us to test several theories of the role the Black Death played in subsequent economic growth as well
as theories about the spatial evolution of economic activity across time. We establish that there was
aggregate convergence from the Black Death but local divergence. The effects of the Black Death were
highly heterogeneous: while some cities did not recover their pre-Black Death populations until the
19th century, the Black Death stimulated city growth in areas that were previously predominantly
rural.
Between 1300 and 1400, a ten percent higher Black Death mortality rate was associated with a 8.4 %
fall in urban population. In the long-run, between 1300 and 1700, the overall impact of the Black
Death was close to zero in the aggregate. This result is robust to the inclusion of controls for locational
fundamentals, increasing returns and institutions and state and country fixed effects. Instrumental
variable approaches based on the timing of infection suggest that our baseline estimate indeed reflects
the causal impact of the plague on urban development.
Although the aggregate impact of the Black Death disappeared by 1600. Recovery from the Black
Death was highly heterogeneous. Cities located on the coasts recovered faster than did inland cities.
Larger cities and cities located in the Hanseatic League or one of the major European monarchies
also grew faster in response to the shock of the Black Death. We conduct an analysis on cities that
only emerged after 1300 at both the city and the cell-level and find evidence that the Black Death
simulated urban development in areas which we were previously undeveloped. In summary, our
analysis provides novel quantitative evidence that the Black Death was not just a one-off mortality
1
shock: it changed the aggregate level and spatial distribution of economic activity in Europe.
Our findings contribute to three literatures. First, we build on work on the origins of long-run economic
growth in premodern Europe. In growth theory demographic shocks play a dual role. Population
growth promotes economic growth if high population densities encourage human capital accumulation
or technological progress (Kremer, 1993; Lagerlöf, 2003; Becker et al., 1999). However, population
growth has a negative effect on per capita income if land or capital is inelastically supplied. In such
a Malthusian environment any positive income shock is temporary: fertility increases and mortality
decreases, so that any increases in the stock of capital (and income) per capita are eventually negated
and income is stable and low in the long-run (Ashraf & Galor, 2011). Countries only achieve sustained
economic growth if the pace of technological development is sufficient to induce investment in human
capital and if, in the long-run, a demographic transition limits population growth (Galor & Weil, 2000;
Hansen & Prescott, 2002; Galor, 2011). In models of unified growth, a demographic shock like the
Black Death can increase incomes sufficiently to accelerate the transition from a Malthusian to a
post-Malthusian economy.
Numerous scholars have argued that the Black Death marks a watershed in European history, after
which it is possible to detect trends that would eventually lead to the onset of sustained economic
growth.1 Acemoglu & Robinson (2012, 101) argue that ‘the Black Death is a vivid example of a critical
junction, a major event or confluence of factors disrupting the existing economic or political balance
in society’; they claim that this shock helped lead the emergence of more inclusive institutions in
England while cementing the existence of extractive institutions in Eastern Europe.2 Our use of finely
grained city-level data which covers all of Europe over many centuries allows us to shed light on the
Black Death’s contribution to the so-called Little Divergence of northwestern Europe from southern
Europe after 1400 (Allen, 2001; Acemoglu et al., 2005; van Zanden, 2009; Broadberry, 2013). In
a related paper to ours Boerner & Severgnini (2014) study the spread of the Black Death as a way
of measuring levels of trade between European cities but they do not investigate the impact of the
plague itself on economic development.
1
See Gottfried (1983); Herlihy (1997); Epstein (2000); Pamuk (2007) for historical accounts that emphasize the
centrality of the Black Death to the eventual economic rise of western Europe. Epstein, for example, argues that the Black
Death was the ‘exogenous event which contributed to the feudal economy’s transition from a low level ‘equilibrium trap’
to a higher growth path’ (Epstein, 2000, 54).
2
This argument was originally developed by Brenner (1976) and is the subject of a considerable historiography (Ashton
& Philpin, 1985).
2
Two recent papers lay out the theoretical mechanisms that explain how the Black Death could have
set north-western Europe on a different growth path. Voigtländer & Voth (2013b) construct a model
in which a shock that raised per capita income like the Black Death could have moved Europe into a
different, permanently higher income, equilibrium. Greater demand for manufactured goods produced
in cities and an intensification in warfare further contributed to higher levels of urbanization. As cities
had higher death rates, they argue that this produced a ‘horsemen effect’—an s-shaped income-death
schedule—which enabled Europe to attain higher per capita income in the pre-industrial period. We
make a novel contribution to this literature by testing this theory at the micro-level, using a novel
dataset of plague mortality. By using city-level data, we are able to better identify the effects of the
Black Death on economic development, thereby providing confirmation at the micro-level for some of
the main macro-level predictions of Voigtländer & Voth (2013b).
In another paper, Voigtländer & Voth (2013a) argue that the demographic shock of the Black Death had
a differential effect in northern Europe—where the land could be turned over to pastoral agriculture—
than in southern Europe where the land was more suited to arable farming. As pastoral farming
increased the demand for female labor, this contributed to a labor and marriage market equilibrium in
which individuals married late and restricted fertility (the European marriage pattern); a phenomenon
that in turn raised per capita incomes.3 They argue that this helped give rise to the Little Divergence
in incomes that took place between northern Europe and southern and eastern Europe in the period
between 1400 and 1800.
The second literature that we contribute to is that on shocks and long-run persistence in economic
development. A number of important recent additions to this literature have demonstrated that
demographic shocks can have long-run impacts (Young, 2005; Acemoglu & Johnson, 2007; Nunn,
2008; Bloom et al., 2014). The Black Death differs from the mortality shocks previously considered in
the literature in its magnitude—an overall mortality rate of 40% dwarfs other morality shocks that we
are aware of such as the 1919 influenza pandemic which killed between 3-5% of the world’s population
or World War Two which killed approximately 3.7%.4 We explore both the short-run and the long-run
3
The European marriage pattern or EMP is studied by Hajnal (1965); Macfarlane (1986); Moor & Zanden (2010).
Voigtländer & Voth (2013a) build on the argument of Kussmaul (1985) that pastoral agriculture provided the economic
incentives necessary to maintain the EMP.
4
Estimates for the total numbers killed in World War 2 (1937-1945) vary between 61 and 80 million. The larger
number would represent approximately 3.7% of world population at the time which is small compared to the Black Death.
Obviously, the mortality rate in selected countries was much higher: approximately 8% in the borders of prewar Germany,
3
economic impact of the Black Death. Our main finding is that, while there was aggregate convergence
within three centuries, at a local level the effects of the Black Death were highly heterogeneous.
The Black Death was also different than other types of shocks associated with warfare or bombings
because it was a purely demographic one that left man-made capital untouched. In this respect our
setting has some resemblance to papers that study the impact of the Holocaust (Acemoglu et al., 2011)
or large scale expulsions of populations. Acemoglu et al. (2011) find that the Holocaust had negative
long-term economic effects as Jews comprised a large proportion of the middle class throughout the
Soviet Union. Chaney & Hornbeck (2015) find that the expulsion of the Morisco’s had a positive impact
on per capita income because it made more land available. Similarly, Jebwab et al. (2015) do not find
a negative effect associated with the expulsions of European and Asian settlers in Kenya.5 Unlike the
expulsion of Jewish or Muslim populations, however, the Black Death did not target a specific group,
but affected the entire economy. This aspect of our analysis has important implications for considering
the potential impact of modern pandemic diseases such Aids and malaria which have had a noticeable
demographic impact in sub-Saharan Africa (Piot et al., 2001; Sachs & Malaney, 2002; Young, 2005;
Weil, 2014). It thus connects with a literature on the consequences of both recent diseases (Almond,
2006; Bleakley & Lange, 2009; Alsan, 2015) and the historical impact of pandemic diseases (Findlay
& Lundahl, 2006; Bosker, Brakman, Garretson, Jogn & Schramm, 2008; Alfani, 2013).
Third, our analysis sheds light on the determinants of long-run urban development. This literature
finds episodes where large shocks do not change the spatial distribution of economic activity and
cases where shocks do lead to urban reset. Prominent theories in economic geography suggest that
this depends on whether locational fundamentals (i.e. physical geography), increasing returns (i.e.
economic geography) or institutions determine the location of cities. If locational fundamentals
and as high as 13.7 % in the 1941 borders of the Soviet Union and 17% in Poland. Nevertheless, even these numbers are
significantly lower than the estimates that exist for the Black Death. Other major mortality shocks include the Ukrainian
and Chinese famines of the mid-twentieth century Death tolls in both famines are controversial. Recent estimates for
mortality in the Ukrainian famine of 1931-1933 are around 4 million or approximately 10% of the population (see Snyder,
2010). Estimates for the Chinese Great Famine (1959-1961) range from 16.5 to 45 million or between 2.5 and 6.8% of
the population (Meng et al., 2015). During the Bengal famine of 1943 about 6.6% of Bengal’s population died. The worst
genocides in recent history are responsible for comparable loss of life. For example, about 11% of Rwanda’s population
was killed in 1994 and 25% of Cambodia’s population during the rule of Pol Pot. These death tolls are significantly smaller
as a fraction of the population than the Black Death.
5
Studies of the impact of expulsions of Jews from Nazi Germany include Akbulut-Yuksel & Yuksel (2015) on human
capital; Waldinger (2010, 2012) on scientific output. These studies are less directly comparable to ours as the numbers
involved were much smaller.
4
dominate then there is a unique equilibrium location for a given set of cities and economic activity
will be centered around these locations irrespective of demographic shocks, no matter how large they
are. But, in the presence of increasing returns, multiple equilibria may exist and a large enough shock
can cause the spatial system to reset (e.g. Krugman, 1991a,b).
Davis and Weinstein’s (2002) seminal study of Japanese urban development in the aftermath of World
War 2 provides evidence for the importance of locational fundamentals. In particular they show that
after experiencing complete destruction, both Hiroshima and Nagasaki returned to the same relative
positions in Japan’s distribution of cities within 20 years. Similarly, Glocker & Sturm (2014) and
Miguel & Roland (2011) study the wartime destruction of Germany and Vietnam respectively, to test
how these shocks affected the relative ranking of cities. Miguel & Roland (2011) find rapid rates
of recovery and convergence suggesting a high degree of persistence in urban location. Maloney &
Caicedo (2015) also find a high degree of persistence in the location of economic activity in the New
World between the precolonial period and today.
Other research, however, demonstrates the importance of increasing returns in enabling local shocks
to have permanent spatial effects (Bosker et al., 2007; Bosker, Brakman, Garretsen & Schramm, 2008;
Redding et al., 2011; Bleakley & Lin, 2012). For example, Rauch & Michaels (2013) contrast England,
where the urban network was reset after the Fall of the Roman Empire, with France, where Roman
city locations remained settled throughout the early middle ages. They argue that the locations of
French cities may have been shaped by path dependence and hence impeded subsequent economic
development. In England, however, the urban network was reset allowing English cities to relocate
along coasts and rivers that were more beneficial for economic activity.
There is also evidence that institutions play a role in determining how shocks affect economic outcomes
and particularly urban development. Acemoglu et al. (2005) show that the opening up of the Atlantic
trade with the New World stimulated economic development in those countries where the power of
the sovereign was initially weak (England, the Netherlands) but not where he was initially powerful
(Spain and Portugal). Campante & Glaeser (2009) argue that the contrasting developmental paths
of Buenos Aires and Chicago in the twentieth centuries can be partly explained in terms of differing
political institutions. Nunn & Puga (2012) demonstrate that in the presence of an extractive institution
like slavery, ‘bad’ geography—in this case ruggedness—could be beneficial in sub-Saharan Africa.
Finally, Dincecco & Onorato (2015) study the effects of war in stimulating urban development in
5
preindustrial Europe.
Relative to this literature, a major contribution of our paper is that we exploit a shock that was
exogenous at a local level (something that is not necessarily the case with the discovery of the New
World). Furthermore, by studying a shock that killed people but did not destroy physical capital, we
have an ideal framework to identify the impact of a demographic shock in a Malthusian environment.
Finally, our analysis allows us to indirectly speak to the ‘optimality’ of urban reset. Rauch & Michaels
(2013) argue that the persistence of the Roman network in France was inefficient. One problem with
this claim is that there are costs of switching to a new urban network and these may be high. The
old network may have been inferior relative to a possible new one but, in the presence of positive
coordination costs, it may not have been efficient to change the location of cities. The Black Death
was a sufficiently large shock that it could have led to a coordinated relocation of urban activity. And
while there is no metric to characterize the optimality of an urban network, our evidence suggests
that the new cities that emerged did so in better locations.
The structure of the remainder of the paper is as follows. Section 2 describes our data and provides
necessary background information concerning the Black Death. In Section 3 we report our baseline
results. Section 5 demonstrates that city-level recovery was highly heterogenous, explores what factors
drove city growth in the aftermath of the Black Death, and provides evidence that there was a partial
‘urban reset’. Section 6 concludes.
2. DATA AND BACKGROUND
2.1
Data
Data on Black Death mortality come from Christakos et al. (2005) who compile information from a
wide array of historical sources.6 These data yield estimates of mortality for 185 locations. Of these,
we can match 139 mortality rate locations to our city database. Therefore, we have 139 cities for
which we have an estimate of plague mortality and an estimate of their population in 1300. We have
a percentage estimate of the mortality rate for 89 of these 139 cities. For example, Florence had an
estimated mortality rate of 60%. In other cases the sources report more qualitative estimates i.e. that
6
We verify this data by consulting Ziegler (1969), Russell (1972), Pounds (1973), Gottfried (1983), and Benedictow
(2005).
6
‘about half’ or ‘at least half’ of the population died in which case we code our estimate as 50% or that
the city was ‘desolated’ or ‘abandoned’ in which case we attribute a mortality rate of 80%. Details on
how these data were assembled are provided in our Appendix.
Our main source of urban population data is the Bairoch (1988) dataset of city populations. The
Bairoch dataset reports estimates for 1,797 cities between 800 and 1850. It provides estimates for
every century up to 1700 and then for each fifty year interval up to 1850. We use 1,792 of these cities
as 5 cities, in northern Norway, Finland, and in the remote West Atlantic, cannot be matched to the
GIS data that we employ to create our geographical controls. The criterion for inclusion in the Bairoch
dataset is a city population greater than 1,000 inhabitants at any point between 800 and 1850. This
dataset has been widely used by a range of scholars studying premodern urbanization and economic
development.
We follow Bosker et al. (2013) and Voigtländer & Voth (2013b) in updating the Bairoch dataset where a
consensus of historians have provided revised estimates of the population of a particular city, including
Bruges, Paris, and London. Specifically, we supplement the Bairoch (1988) dataset using several
sources for pre-plague population including Chandler (1974, 1987), Nicholas (1997), and Campbell
(2008).7 In our regressions we will use both the corrected dataset and the original Bairoch dataset.
For control variables we group our data into three categories: (1) measures of locational fundamentals,
(2) factors that might generate increasing returns, and (3) institutional variables. To control for
differences in physical geography that could affect the success of a city we collect the following
geographic variables: a dummy variable for whether a city is within 10 km of the coast or a major
river; every city’s elevation in meters; and the cereal and grazing suitability of the surrounding 25 km.
We use a distance of 25km because we are interested in the potential agricultural productivity of each
city’s hinterland as this was a major determinant of city size in a world of high trade costs. We also
control for a city’s longitude, latitude and its average temperature between 1500-1600 (the earliest
data for which estimates of the level of temperature exist).
Increasing returns can stem from a variety of sources. First, there are preexisting trade routes. To
7
The historical populations of Paris and London in 1300 are now considered to have been higher than Bairoch’s
estimates. We use the figure of 228,000 for Paris and 60,000 for London. On the other hand, the population of Bruges is
now thought to be smaller than the number given by Bairoch. We assign them a population of 12,000 in 1000, 15,000 in
1100, 25,000 in 1200 and 35,000 in 1700. Further details are confined to the Web Appendix
7
capture the importance of trade routes we include a dummy variable if a city belonged to the Hanseatic
league (Dollinger, 1970) or was within 10 km of a medieval (land) trade route or the intersection
of a trade route (Shepherd, 1923). Second, prior investments in physical capital provide a source
of agglomeration effects. Previous research has shown that the Roman road network remained the
major road network throughout medieval Europe (Bosker et al., 2013). We include a dummy if a
city was within 10 km of a major Roman road or an intersection of a Roman road (McCormick et al.,
2013). Thirdly, human capital may be another source of agglomeration affects as demonstrated by
Dittmar’s (2011) study of the effect of the printing press on city growth rates. Therefore we include a
dummy variable for whether a city had a medieval university (Shepherd, 1923). We control for log
city population in 1300 as this may also influence subsequent city growth. If there are agglomeration
effects and these dominate congestion effects, then we would expect larger city cities to growth faster
than smaller cities. If, on the contrary, urban growth is limited by a fixed factor of production as
Dittmar (2015) suggests was the case then we would expect the coefficient on log population in 1300
to be negative.
Following De Long & Shleifer (1993), Acemoglu et al. (2005) and others we also expect institutional
factors to affect city growth in preindustrial Europe. To measure institutional differences between
European cities we assign each city to its political jurisdiction in 1300 based on Nussli (2011). We
also use modern political boundary fixed effects in our analysis as a robustness check. We distinguish
between cities that were located in monarchies such as England and France and autonomous cities,
that is either city republics such as Florence and Venice or cities which had de facto self-governance
such as Lübeck. These were the major distinctions in types of polities in the medieval period. We rely
on Stasavage (2014) and Bosker et al. (2013) to code cities as autonomous. The historical literature
suggests one might expect autonomous cities to follow policies that were more conducive to trade and
commerce (Stasavage, 2014). However, there is also evidence that city states like Venice and Florence
imposed high taxes and extractive policies on the surrounding countryside (Epstein, 2000).
Since administrative centers may have followed different development paths, we code whether a
city was a capital city in 1300 using Bosker et al. (2013). Finally, the fourteenth century saw an
intensification in warfare which may have affected economic growth as suggested by Voigtländer &
Voth (2013b). As it is impossible to obtain precise numbers on excess mortality due to warfare for the
medieval period, we follow a recent scholarship in collecting data on the location of conflicts (e.g.
8
Dincecco & Onorato, 2015; Iyigun et al., 2015). Our main source is Wikipedia’s list of all battles that
took place between 1300 and 1600. For each battle we assign a geo-coordinate based on either the
location of the battle or the location of the nearest town or city mentioned in the entry. We exclude
naval battles and conflicts which cannot be located (such battles were typically extremely minor).
We create a dummy variable that measures the distance to major battles between 1300-1350 and
1350-1600. This provides a proxy for both the disruption caused by warfare to nearby areas and for
the ‘safe-harbor’ effect identified by Dincecco & Onorato (2015) that might have led to urban growth
as rural citizens moved to cities for greater security.
2.2
Epidemiological Background
Until recently the etiology of the Black Death was a matter of scholarly dispute (e.g. Twigg (1984);
Cohn (2003)). Epidemiologists have now identified the DNA of skeletons from mass graves associated
with the Black Death as Yersinia pestis (Haensch et al., 2010; Bos et al., 2011; Schuenemann et al.,
2011).8 This decisive evidence means that we can be confident in inferring the characteristics of the
Black Death on the basis of what we know about outbreaks of bubonic plague in the modern period.
Modern bubonic plague is spread by the fleas that live on black rats Rattus Rattus. The bacteria Yersinia
pestis normally lives in the digestive tracts of fleas. If the bacteria multiple sufficiently in the stomach
of the flea then this causes a blockage. Fleas that are thus blocked regurgitate this bacteria into the
skin of whatever animals they bite as they feed (Gottfried, 1983, 6-7). Rats are the primary host for
the bacteria and rat fleas only spread to humans (and other secondary hosts) when the rat population
has become exhausted. The associated Bubonic plague has a mortality rate of around 70-75%.
Identification of the causal impact of the Black Death requires Black Death mortality to be uncorrected
with factors that could influence subsequent city growth. The fact that Bubonic plague is spread by
the fleas of rats introduces a random component to the spread of the disease. A single infected rat on
a ship was sufficient to spread the disease from one city to another. Importantly, population density is
not a determinant of the spread of bubonic plague. Benedictow notes that ‘it is a unique feature of’
the Bubonic plague ‘that the densities of rats and rat fleas overrule the effects of the density of the
susceptible human population that is the decisive factor for the dynamics of epidemic spread in the
case of all diseases that spread directly between human beings by cross infection’ (Benedictow, 2005,
8
We discuss alternative explanations of the Black Death proposed by historians and the modern scientific evidence in
our Web Appendix where we also provide more details and citations concerning our data sources.
9
284). Rat density is not necessarily correlated with population density as rats are territorial animals.
In rural areas a single rat colony may cohabit with a single household. However, in urban areas people
live closer together and the ratio between rats and humans tends to be lower. As Benedictow (2005)
argues: the ‘severity of impact on human population does not increase with mounting density of
human settlement’ (Benedictow, 2005, 33).9 These features of the plague distinguish it from almost
all other epidemic diseases such as influenza and smallpox and help to account for why mortality was
unrelated to factors such as population density.
The high mortality rates and contemporary descriptions also suggest that pneumonic plague was
present in some cases. Bubonic plague is predominantly spread from infected fleas to humans.10
Pneumonic plague can be transmitted person to person via airborne transmission. It has a mortality
rate of 95-100% following a two to three day incubation period. Many individuals die in the first
24 hours, before they develop the infectious cough associated with the disease (Benedictow, 2005,
28). This extremely high level of mortality means that the spread of pneumonic plague is typically
much more restricted than that of bubonic plague. It is worthy of note, therefore, that while bubonic
plague is uncorrelated with population density, pneumonic plague may be correlated with the density
of human settlement. We are not able to distinguish between the two forms of plague due to a lack of
evidence. This is unlikely to be a major of source of bias in our analysis as historians and scientists
agree that the rapid spread of the plague is consistent with the outbreak being largely one of bubonic
plague (see Benedictow (2005, 62-76) and Scott & Duncan (2001, 30)).11 In more formal analysis
below we show that plague mortality was uncorrelated with population density and other geographic
and economic variables.
2.3
Historical Background
The Black Death arrived in Europe in 1347. Over the next five years it spread across the continent
killing between 30 and 50% of the population.12 Death rates were comparable in the cities and in rural
9
Similarly Gottfried notes that ‘R. rattus was well-suited to both the thatched roofs of peasant dwellings and the high
roof beams and dark corners of urban houses’ (Gottfried, 1983, 7).
10
Bubonic plague can also be spread indirectly from fleas to infected animals which can transmit the disease to humans
by biting them. However, fleas were the predominant vector of transmission.
11
Benedictow observes that ‘it is clear primary pneumonic plague cannot play a major independent epidemic role and
that the Black Death cannot have been wholly or mainly an epidemic of this kind’ (Benedictow, 2005, 30).
12
Conventionally the death rate was estimated at 1/3 of Europe’s population. More recent studies suggest that the
overall death rate was considerably higher than this. Benedictow (2005) argues for a mortality rate as high as 60%. This
higher estimate, though controversial, has not been rejected by other scholars. One reviewer notes that ‘Benedictow’s
10
areas, and in general, historical accounts are unable to explain variation in mortality rates (Ziegler,
1969; Gottfried, 1983; Theilmann & Cate, 2007; Cohn & Alfani, 2007).13 Black Death mortality was
not correlated with city size or population density. To illustrate, Venice had extremely high mortality
(60%) while Milan escaped comparatively unscathed (15% mortality).14 Highly urbanized Sicily
suffered heavily from the plague. However, equally urbanized Flanders (modern-day Belgium) had
relatively low death rates, while the more rural northern Netherlands was devastated. Southern
Europe and the Mediterranean was hit especially hard, but so were the British Isles and Scandinavia.
Figure 4a suggests that there is no relationship between Black Death mortality and city population in
1300.
The spread of the plague was rapid and its precise trajectory was largely determined by chance. As
we confirm in our formal analysis, the only systematic predictor of plague spread and virulence was
latitude as the plague spread from southern Europe to northern Europe. In other respects, the precise
movement of plague was determined largely by chance. For example, it was largely coincidence that
the plague would spread first from Kaffa in the Black Sea to Messina in Sicily rather than elsewhere as
the ships carrying the plague could have stopped at other ports in the Mediterranean. Similarly, it was
partly coincidental that the plague spread first from Messina to Marseilles. The arrival of the plague in
Marseilles ensured its speedy transmission through much of western Europe in the year 1348.15 When
it arrived in a country it moved quickly. For instance, it arrived in southern England in June 1348 in
Dorset in the southwest of the country. It reached London by November. It hit the north of England by
early 1349, peaking in that part of the country in the summer of 1349 (Theilmann & Cate, 2007).
mortality estimates may eventually come to be regarded as the standard, in spite of readers’ doubts that the remarkably
similar die-off across regions is due in part to rejecting data indicating lower figures through source criticism. The estimates
are internally consistent with his assessments of plague case-fatality (circa 80 percent, p. 350) and prevalence. If plague
lethality is over 50 percent in modern populations, then 80 percent is not implausible for medieval times, considering the
nutritional stresses of the fourteenth century’ (Noymer, 2007, 624).
13
Prior to the Black Death there had been no major outbreak of epidemic disease for several centuries and as a result
neither the medical profession or political authorities were able to respond effectively.
14
There is no indication that variation in sanitation or hygiene explanations this pattern. Gottfried notes ‘it would be a
mistake to attribute too much to sanitation. The failure of Venice’s excellent sanitation to stem the deadly effect of the
plague has been discussed’ (Gottfried, 1983, 69).
15
For example, Benedictow notes that ‘The surprisingly early invasion of Marseilles by the Black Death may have
affected profoundly its pattern of strategic advance in Southern and Western Europe, and also, as we shall see, the invasion
of the British Isles. It was another momentous event in the sinister history of the Black Death . . . From this commercial
centre, communication lines radiated both by sea and by land. This city was an excellent bridgehead for the Black Death’s
offensive against south-western and Western Europe’ (Benedictow, 2005, 73).
11
The Black Death affected all segments of the population.16 Medical knowledge was rudimentary and
ineffective: Boccaccio, for instance, wrote that ‘[a]ll the advice of physicians and all the power of
medicine were profitless and unavailing’ (Boccaccio, 2005, 1371). Individuals were unable to protect
themselves from the disease. Institutional measures of prevention were nonexistent: the practice of
quarantine was not employed until later in the fourteenth century.17
Figure 1a presents estimates of total population and urbanization for the 17 (modern) countries which
contain the cities we use in our main analysis. Europe only regained its pre-plague population by
around 1600. Urbanization, in contrast, rose in the aftermath of the Black Death from around 7% to
10% by 1400, and 12% in 1600. The source of the increase in urbanization is evident in Figure 1b
which depicts the evolution of Europe’s urban population between 1100 and 1600. Approximately
50% of the acceleration in urban growth after 1353 is largely attributable to the growth of cities
that either did not exist or were below the threshold of 1,000 inhabitants in 1300. These aggregate
statistics suggest it was the emergence of new urban centers that was important in explaining the
growth in urban population relative to total population between 1353 and 1600. These data also
imply that the effect of the Black Death on Europe’s economic structure was highly heterogeneous:
many cities had not recovered their 1300 populations by 1600 while other cities rapidly expanded in
these centuries.
To see this more clearly, we examine the 139 cities for which we have Black Death mortality data. For
these 139 cities we can tentatively construct an estimate of their population in 1353—immediately
after the Black Death—by multiplying population estimates for 1300 by Black Death mortality rates.
Figure 2 shows how the distribution of these city sizes changed over time relative to the base of 1300.
The mortality shock of the Black Death is evident in the rightward shift of the line for 1353. What
is most interesting is that while the peak of the distribution remained below 100 for both 1400 and
1500, many cities not only recovered their 1300 population but grew far larger, while some cities did
not recover at all. Among the 139 cities in 1600, 52 had a population below their 1300 population;
30 have a population below 2/3’s of their 1300 level and 18 cities had a population below 1/2 of
16
It is commonly asserted that the Black Death killed indiscriminately. Recent research examining the skeletons from
plague pits in London suggests that on average plague victims were likely to be older, frailer, or less well-nourished
(DeWitte & Wood, 2008).
17
The term quarantine was first used in the city of Ragusa, part of the Venetian empire in 1377. It was adopted as a
standard policy by Venice in 1423 (Gensini et al., 2004, 257).
12
their 1300 level. In the next section we conduct a more formal regression analysis to investigate these
findings further.
3. AGGREGATE CONVERGENCE AND LOCAL DIVERGENCE
To study the effects of the Black Death on city growth we estimate:
%∆Popi,t = α + β Di,1347−1353 + X0i δ + γc + εi,t
(1)
where %∆Popi,t is the percentage population growth in city i over period t −1 to t: Pop t −Pop t /Pop t−1 ,
Di,1347−1353 is a measure of the city-level mortality rate of the Bleak Death between 1347 and 1353, Xi
is a vector of city specific controls and γc represent either historical or modern country fixed effect
which we employ in some of our regressions.
3.1
Short-Run Effects: 1300-1400
First, we focus on the impact of the Black Death on city population between 1300 and 1400. Column
(1) of Table 1 measures the short-run impact of the plague. The coefficient from our baseline OLS
regression is -0.84. This should be examined in relation to the immediate effect for the period 13471353 which is 1 by construction. A value greater than -1 in magnitude might occur if in addition to
the death caused by the plague, people left cities struck harder by the plague in the period between
1353-1400. The coefficient that we obtain suggests that a 10% higher mortality rate during the Black
Death was associated with an 8.4% smaller population growth rate. This suggests that there was little
recovery in population in the decades following the onset of the plague. This finding is consistent
with the observations of historians like Nicholas (1999, 99) and Hohenberg (2004, 14) who write of
an ‘urban crisis’ in the wake of the Black Death.
3.2
Long-Run Effects: 1300-1700
What was the effect of the Black Death on long-run urban development? Columns (4)–(8) of Table 1
examine the long-run impact of the Black Death. There was recovery in those cities hit hardest by
the Black Death in 1400-1500 and 1500-1600. The coefficient we obtain in Column (8): of 0.09 is
not significantly different from zero and demonstrates that by 1700 there was convergence in the
aggregate. In most of the following analysis we focus on the period between 1300 and 1600 to
13
minimize contamination from potentially confounding events such as the discovery of the Americas
as studied by Acemoglu et al. (2005), the adoption of the printing press Dittmar (2011) or the
introduction of the potato (Nunn & Qian, 2011).
In their analysis of Japanese city growth Davis & Weinstein (2002) find full convergence from the
shock of World War 2 by 1960. Their coefficient of interest is precisely estimated and the standard
errors associated with their estimates are small suggesting that convergence occurred at the city level.
This is best illustrated by the fates of Hiroshima and Nagasaki. Ranked 7 and 11 respectively in the
distribution of Japanese cities in 1940, they fell to 26 and 25 in 1945 in the wake of the atomic
bombs. However, by 1960 their relative ranking was again 6 and 11. In contrast, we find long-run
convergence in the aggregate but we do not obtain city-by-city convergence. In fact, at a local level
there was considerable variation in population recovery. The coefficient we obtain in Column (7) is
not precisely estimated. (We report the relevant confidence intervals in square brackets below each
estimate.) It is not just that the coefficient we obtain is small, but that the confidence intervals for
Column (7) and associated standard errors are very wide. The experience of European cities after the
Black Death was highly heterogenous.
4. INVESTIGATING CAUSALITY
A natural question raised by this analysis is whether these results reflect the causal impact of the Black
Death. We now explore a range of strategies to investigate causality.
First, we show that the parallel trends assumption is satisfied. Columns (2)-(3) of Table 1 indicate that
prior to 1300, there is no difference in growth between cities most affected and those comparatively
unaffected by the Black Death. The cities in our sample that were hit hard by the Black Death and
those that escaped comparatively unscathed were following similar development paths prior to the
demographic shock that they experienced in the 14th century.18
Next we show that Black Death mortality is uncorrelated with observable city-level characteristics
(Table 2). Variation in mortality rates cannot be explained in terms of locational fundamentals (see
Note, the parallel trends assumption is also verified if we: (i) test the effect for the period 1100-1300 (N = 59), rather
than just 1100-1200 as in column (2) or 1200-1300 as in column (3), and (ii) test the effect for the period 1000-1100 (N
= 56).
18
14
Column (1)), nor are they accounted for by our measures of increasing returns (Column (2)), or by
institutional factors (Column 3). The only significant variable explaining mortality rates is latitude
which has a significant negative relationship with mortality rates. This reflects the fact that the Black
Death hit southern Europe first, and was, in general, more virulent there than in northern Europe.19
This analysis provides some support for our claim that the Black Death was a exogenous shock. Table 3
further investigates the causal impact of Black Death mortality on subsequent urban development. The
OLS estimates we report could be subject to either a downwards or an upwards bias. A downwards bias
would occur if cities which experienced high death rates during the Black Death were subsequently
likely to grow more slowly. It would lead us to underestimate the causal impact of the Black Death on
later city growth. Alternatively our baseline estimates would be subject to an upwards bias if higher
plague mortality was linked to faster urban growth. From the perspective of understanding whether
or not the plague had a causal impact on economic development, it is most important to show that
our results are not subject to an upwards bias.
We show in rows 2–5 that our results are robust to the inclusion of controls for geographic features
including ruggedness and temperature which might positively affect both plague mortality and future
city development. Our controls for measures of increasing returns include the presence of a university
or a state capital, factors that could plausibly be associated with both higher plague mortality and
faster city growth. We also control for the location of all subsequent battles in row 6. We do this
for several reasons: historians note that the immediate impact of the Black Death was to reduce the
intensity and scale of conflict (in the Hundred Years War, for example, as documented by ?) but,
at the same time, in the longer-run as Voigtländer & Voth (2013a) and Dincecco & Onorato (2015)
argue, warfare both intensified after 1400 and is seen as a factor that helped to increase levels of
urbanization in late medieval and early modern Europe.20 One factor influencing both the spread of
19
Note, that if city level characteristics were entirely uncorrelated with the Black Death, then the coefficient of variation
would be zero. The R-squared we obtain in Column (1) is, at 0.15, considerably higher than zero. However, it falls to
0.07 when we exclude latitude and temperature (whose correlation with latitude is 0.89. In Column (4), when we drop
latitude and temperature, the coefficients of the other controls also remain insignificant and the R-squared decreases to
0.15. The R-squared doe not entirely decrease to 0 because some variables are still somewhat correlated with latitude
(correlations around 0.4-0.6). This reflects the fact that the Black Death originated in and was more severe in southern
Europe and that cities in southern Europe were different from those in northern Europe. Our controls capture some of this
variation. We will demonstrate below that our results are robust to the inclusion of country fixed effects.
20
Our results are unchanged if we also include the number of soldiers involved in each battle, but for a smaller sample
(available upon request).
15
disease and subsequent recovery is access to sea-based trade. Since the Black Death spread out from
the Mediterranean, we distinguish between three coastlines in the regression results we report in row
7: the Atlantic, the Baltic and the Mediterranean. Across all of these regressions, our findings remain
consistent with what we obtain using our baseline specification.
In row 8, we include a dummy variable for whether or not a city had a Jewish community. If, as
some historians have suggested, Jews had better hygiene practices than Christians, it is possible that
mortality rates might be lower in cities with large Jewish communities. The coefficient we obtain does
shrink slightly to -0.64, but we think this is more likely due to other factors influencing the location of
Jewish communities rather than due to differential mortality rates between Jews and Christians as we
know that most Jewish communities in this period were very small. Regression 8 also controls for
subsequent Jewish persecution as the presence of a Jewish community could have a positive impact
on growth if they provided important commercial or financial services.
Another approach to control for the presence of unobservables is to use fixed effects. Rows 9-12
implement a series of country level fixed effects. In row 10 we employ modern country fixed effects to
control for spatially or regionally correlated unobservables. Modern country borders differ substantially,
however, from the political units of the fourteenth century so in regression 11 we assign a separate
dummy variable to each of the 57 independent polities in our dataset using the information on borders
provided by Nussli (2011). The sheer number of state raises a potential problem for our analysis,
however, as many of these polities were city states or small principalities with only a single major city.
As a result, there is insufficient variation and our estimates become less precise. The effect we obtain
is almost significant at the 10% level (p value = 0.107) and similar in magnitude. Our preferred fixed
effect estimator is reported in row 12 where we restrict the sample to states with strictly more than 3
cities in our main dataset.21
Next we implement two IV strategies. The first exploits the timing of the Black Death. Timing provides
exogenous variation in Black Death mortality as there is evidence that plague virulence was correlated
with time. Cities that were affected earlier, all else equal, tended to experience higher death rates.
Figure 5 provides support for our instrumental variable strategy. It plots Black Death Mortality against
the data that the city was first infected. In general, cities infected later had lower mortality estimates.
These are the Kingdom of Aragon (N = 7), England (23), France (26), The Papal States (5), Portugal (5), Sicily-Naples
(4) and Sicily-Trinacria (4).
21
16
Using the date of first infection as an IV we obtain a coefficient of -1.63 (row 13). This is negative
and precisely estimated. However, this unconditional IV estimate is significantly larger than our OLS
coefficient and the F-statistic (18.4) is on the low side suggesting that the exclusion restriction may
be violated. This is unsurprising. The Black Death spread from the south and as the urban network
of medieval Europe was still centered around the Mediterranean Sea it is likely that month of first
infection is partially correlated with market access, trade, and city size. We therefore employ month of
first infection conditional on our control variables as our preferred IV specification. By including city
size and a quartic in latitude and a quartic in longitude to control for the geographical origins of the
plague, we are able to isolate the random component of the spread of the infection. Our conditional
IV yields a coefficient that is close to -1 and a considerably higher F-statistic (row 14).22 This result is
consistent with our baseline OLS estimates and with our hypothesis that the spread of the Black Death
was largely exogenous to city-level characteristics.
Our second IV makes use of the variation in mortality generated by the month of first infection within
a single year. The Black Death was at its most virulent during the summer months (Benedictow, 2005,
233-235). Rat fleas can survive for between 6 months and a year without a host. But they become most
active when it is fairly warm and humid—their activity peaks at temperatures between 15◦ C − 20◦ C
and in conditions of high humidity while the cold limits their activity (Gottfried, 1983, 9). The average
duration of the Black Death was 7 months (Figure 6). Therefore, cities that became infected close to
the winter escaped relatively unscathed compared to cities that were infected during the spring or
early summer. Support for this is provided by detailed month-by-month estimates of mortality from a
number of cities. Christakos et al. (2005) analyze similar data from the city of Piacenza. They find
that mortality rates peaked in September, three months after the first infection was recorded in July.
By November and December mortality from the plague fell significantly.
23
In row 15 we report the results of our analysis using our second IV: month of first infection. We
22
We obtain a similar coefficient if we use the year of first infection as an IV rather than the number of months since
first infection. Our results also hold if we include either linear, quadratic, or cubic terms for latitude and longitude instead
of quartics. We can also control for the start year or start year dummies to exploit the variation in the number of months
within a same year. As a placebo test we also see if our conditional IV predicts population growth in period century
1200-1300. The effect is insignificant and small which is inline with our placebo analysis in Table 1.
23
Figure 11 plots data from the two proxies for Black Death mortality: the number of testaments during 1348 from fives
Italian cities alongside data on the number of vacant benefices in Barcelona. Both sources of data reveal that mortality
peaked in the summer months. The analysis we present in Figure 7 likewise suggests that this IV has some explanatory
power.
17
obtain qualitatively similar results, although their is some concern about the validity of of the IV as
the F-statistic is low and the coefficient is around twice the size of the OLS coefficient. We therefore
condition our IV on our all of our baseline controls and include a quartic in latitude and longitude.
The coefficient we obtain is close to -1 and highly significant (although the F-stat remains low). In
rows 17-18 we implement both instruments in a single regression. Taken together these results are
consistent with our OLS estimates and in conjunction with the results of Table 3 suggest that the effect
of the Black Death on city populations was causal.
4.1
Robustness Checks
In this section we demonstrate that our baseline results are robust to a host of potential concerns
about our data (Table 4), the empirical specification (Table 5), and sampling issues (Table 6).
Table 4 reports robustness checks concerning the nature of our city population data and our mortality
data. Historical data of the kind we employ is noisy and characterized by measurement error. The
Bairoch (1988) dataset is widely used by economists and economic historians but it also has well
known issues and subsequent scholars have corrected it for certain cities (see the discussion in the
Appendix). The first row of Table 4 reports our baseline estimate. Rows 2 and 3 report estimates using
only uncorrected Bairoch data.
A second concern is with our estimates of plague mortality. The data provided by Christakos et al.
(2005) comprises four types of evidence: (1) numerical estimates of mortality (2) descriptive estimates
of mortality (3) desertion rates; and (4) estimates of mortality based on the records of clergymen.24 To
ensure that our results are not driven by an single type of estimate, in row 5, we exclude description
based estimates of mortality. This reduces the number of observations to 115, but the coefficient we
obtain is almost identical to our baseline estimates. It is also possible to exclude estimates based on
desertion rates (row (6)) and on estimates based on mortality among the clergy (row (7)) without
affecting our results. The regression in row 8 presents our results based solely on numerical estimates
of death rates during the plague. As a further robustness exercise we show that our results remain
stable when we exclude ‘heaped’ estimates that record a mortality rate of 25% or 50% (row (9)) (as
discussed by Voigtländer & Voth, 2013a).
To ensure that are results are consistent across both the cities for which we have historical estimates
24
We provide more details on how these estimates are constructed in the Web Appendix.
18
for the mortality rate and other cities in the Bairoch database, in rows 10-12 we spatially extrapolate
mortality rates from our core sample of cities in order to generate estimates for plague mortality
for the entire sample of cities in Bairoch. We obtain an estimate of between -0.71 and -0.78 across
specifications which is consistent with our baseline analysis.
Table 5 investigates the robustness of our results to various empirical specifications. We show that
our results remain unchanged when we cluster our standard errors at the modern country level (12
clusters), the 1300 state level (57 clusters) or employ Conley (2008) standard errors with a radius
of 100 km in order to account for spatial auto-correlation in the error term (Regressions 2-4). In
rows 5-6 we control for population change in the previous centuries. This reduces our sample size
and results is a slightly higher coefficient. The regression we report in row 7 drops our population
weights and instead excludes cities with less than 5000 individuals in 1300. In rows 8 and, 9 we use a
Log-Log specification. Row 10 estimates the relationship between population in either 1400 or 1600
on estimated plague mortality and initial population. This confirms the result that plague mortality
had a large effect on 1400 population, but no effect on 1600 population on average. Finally, in row 11
we implement a panel specification with city and year fixed effects for all cities in existence between
1200 and 1600.
Finally, we address issues concerning our sample in Table 6 to ensure that our analysis is not driven
by outliers or by specific regions or countries. Rows 1-4 drop cities which experienced dramatic
population changes between 1300 and 1400. The coefficients we obtain become smaller in magnitude
as expected, but they retain their sign and statistical significance. Excluding cities that either were
unscathed by the plague or experienced extremely high mortality rates ( ≥ 80% ) does not affect our
estimates (rows 6-7). Finally, we show that the results are robust when we omit the modern countries
with the largest numbers of observations in our dataset: France, Germany, Italy, and the UK (rows
7-10). The results do not change across these regressions implying that our findings are generalizable
to Europe as whole.
5. HETEROGENEOUS RECOVERY
Historical evidence suggests the heterogeneity of the response to the Black Death shock was important.
For example, consider the fate of two cities: Hamburg and Montpellier. Hamburg had a population of
19
about 8,000 individuals in 1300. It was struck hard by Black Death, experiencing a mortality rate
of approximately 58%.25 However, it recovered and indeed boomed in the subsequent half-century,
growing so rapidly that it had a population of 22,000 by 1400. Like the rapid recovery of Hiroshima
and Nagasaki after World War 2, such impressive growth in the wake of a major demographic shock
is in line with theories that emphasize the role locational fundamentals play in determining urban
location and development (Davis & Weinstein, 2002). By the seventeenth century, Hamburg was a
major center of international trade (Lindberg, 2008).
In contrast, for other cities the demographic shock of the Black Death appears to have led to a period
of prolonged decline. Montpellier, for instance, had a population of 35,000 in 1300. Like Hamburg
it was struck hard by the plague, experiencing a 50% mortality rate. However, unlike Hamburg, it
did not recover: its population in 1400 was 17,000, a 45% decline. It fell from being the 4th largest
French city to being the 20th. Moreover, the decline of Montpellier continued for centuries; the city
did not exceed its 1300 population until 1850. Relative to Hamburg, the fate of Montpellier is more
in keeping with models of multiple equilibria in urban location in which a large enough shock can
cause the urban system to reset (e.g. Krugman, 1991b). Furthermore, the experience of Hamburg
calls into question theories that emphasize locational fundamentals as the most important factors
in driving city growth as the same fundamentals present before 1300 evidently could not ensure
economic success in the post-Black Death period. We ask which experience—that of Hamburg’s or
that of Montpellier’s—was more representative of Europe as a whole?
Table 7 studies the heterogeneous effects of the Black Death on city growth. Our interest is in how
geographical, economic and institutional factors mediated urban recovery. We report the interaction
effects for each factor with Black Death mortality for our sample of 139 cities for which we have
mortality data and using our extrapolated mortality data. Column (1) studies the recovery during the
period 1300-1600 which is the main focus of our analysis as city growth after 1600 was affected by
the rise of Atlantic trade and the Reformation. Columns 2 and 3 extend the analysis out to 1700 and
1750 respectively to see whether the factors affecting population recovery had a differential impact in
the period leading up to the Industrial Revolution.
Several factors stand out in Table 7. The first is that coastal cities recovered faster. A second notable
factor was initial city size. Threshold effects mattered: cities with above the median population in
25
Ziegler (1969, 86) estimates the death toll in Hamburg as between a half and two-thirds.
20
1300 tended to recover more quickly from the Black Death. Cities like Paris, London and Venice were
large in 1300 and bounced back from the Black Death shock relatively quickly but for many small
cities the Black Death represented a permanent shock to their populations.
Third, cities within the Hanseatic League in northern Germany recovered rapidly from the plague.
The coefficient reported in Table 7 suggests that a one standard deviation increase in Black Death
morality was associated with approximately 64 % (≈ 4 × 0.16) faster growth for a Hansa city than for a
non-Hanseatic city. This is unsurprising as the Hanseatic league was at its peak as a commercial trading
network during the fourteenth and fifteenth centuries.26 The fourth factor identified as important in
Table 7 is whether a city belonged to a major territorial state like England, France, Castile, or Aragon.
If it was, then it also recovered more quickly This finding is inline with the literature in economic
history which emphasizes the rise of territorial states in the late middle ages and early modern period
(Tilly, 1990; Gennaioli & Voth, 2015)
5.1
Malthusian Dynamics and Horsemen Effects
A long scholarly tradition sees the Black Death as a major break point in European history (Ziegler,
1969; McNeil, 1974; Cantor, 2001; Cohn, 2003; Acemoglu & Robinson, 2012). In their recent analysis
Voigtländer & Voth (2013b) highlight the plague as one of the three horsemen of riches that helped
move Europe onto a trajectory that would eventually lead to modern economic growth.
Building on Voigtländer & Voth (2013b), we now consider the impact of the plague on subsequent
economic development. Table 8 analyzes the impact of the Black Death on both the set of cities that
existed in 1300 and on the rise of new cities after 1300. We find evidence that areas where we estimate
the Black Death was more severe were more likely to be locations of urban growth in the post-plague
period.
In Rows 1 and 2 of Table 8 we present our baseline analysis for 1300-1400 (Column 1), 1300-1600
(Column 2), and 1300-1700 (Column 3). In rows 3-5 we include all cities in the dataset between
1300-1700 and not only those cities that only existed in 1300. This gives us an unbalanced panel of
13,22 cities. By construction, estimates of Black Death death mortality rates do not exist for these
26
See Dollinger (1970). The political decline of the Hanseatic league occurred in the fifteenth century (see Rotz, 1977).
However, they remained economically successful until the end of the sixteenth century when ‘Amsterdam emerged as
northern Europe’s economic center, and Dutch control of the Baltic spice trade at this time marked the end of the dominant
role of the Hanseatic cities within northern European trade’ (Lindberg, 2008, 647).
21
cities so we use our extrapolated mortality rates to provide estimates of plague mortality. As expected,
the negative coefficient for the short-run impact of the plague becomes much smaller (as many cities
in the sample did not exist in 1300).
Rows 6-8 focus only on those cities that enter the dataset post 1300. This provides us with an
unbalanced panel of between 105 and 894 cities. Using this sample enables us to study the impact
of the Black Death on the growth of cities that had not reached the threshold of 1000 individuals in
1300. We find that the long-run impact of extrapolated mortality during the plague of the fourteenth
century was positive on subsequent urban growth in cities that only enter the Bairoch database in
subsequent centuries. They experienced between 7% and 13% faster growth between 1300 and 1600
depending on the specification we employ.
We obtain the same results by conducting a cell-level analysis. We report the results of doing this in
Table 9 and interpret these as a test of whether the ‘horsemen effects’ theorized by Voigtländer & Voth
(2013b) were different in different parts of Europe. We divide our map of Europe into 1 degree x 1
degree cells. There are 565 cells for which we know the Black Death mortality rate. Among these 565
cells, 344 had a city in our dataset at at least one point between 1300-1600, and 243 had a city in
1300. In Column (1) we estimate the effects of Black Death mortality on subsequent city growth at
the cell level focusing on the growth of those 243 cities that existed in 1300. We confirm our city level
results, obtaining a coefficient of -0.90 for the period 1300-1400, and a coefficient of 0.56 for the
period between 1300-1600. The comparatively high standard errors associated with this coefficient
(0.78) suggest that the Black Death had positive long-effect effect for some cities, but not all cities.
We next differentiate between cells which had a high level of urban activity in 1300 and those that
did not. In a standard Malthusian-style model in which individuals have Stone-Geary preferences, we
expect richer cells to have higher demand for manufactured goods and hence greater urban activity.
We examine the effect of Black Death death on the growth of cities that are not in our dataset for
1300 but which passed the threshold of 1,000 inhabitants (and hence enter the dataset) at some point
between 1400-1600. We call these cities ‘new cities’. This analysis distinguishes between the effect of
the Black Death on areas which were more or less economically developed in the period before the
Black Death. We find that Black Death mortality has a larger effect on subsequent urban growth in
cells which did not have sizable urban activity in 1300.
In order to compare cells which had cities in 1300 with those that did not we focus on the absolute
22
growth in urban populations between 1300–1400 and 1300–1600. Specifically, Column (2) focuses
on the absolute growth in population for cities in the 243 cells which already had at least one city in
1300. We obtain a negative coefficient of -0.06. This suggests that the Black Death did not lead to the
rise of new urban centers in areas which were already highly developed in 1300.
In Column (3), we study the 101 cells that did not contain a city in 1300 but in which a city passed
the threshold of 1,000 inhabitants (and hence entered the dataset) at some point between 1300-1600.
In contrast to Column (2), we find no effects on mortality rates on growth in the short-run, but strong
positive effects in the long-run. The coefficient we obtain of 0.24 implies that a morality rate of 50%
during the Black Death in a particular cell was associated with 12,000 additional inhabitants by 1600.
Finally, in Column (4) we look at the combined effect of Black Death mortality on cells with and
without cities in 1300. We find a negative effect in the short-run as expected. The coefficient we
obtain for the long-run effect suggests that this effect is positive, although not precisely estimated
because there is a lot of heterogeneity. This finding is consistent with the results of Table 3.
To summarize, we find that the Black Death prompted urban growth in those cells which were
previously rural but had no effect on average in those cells where there were already large cities in
1300. Hence in addition to a tale of two cities there was also a tale of two cells. The effects of the
Black Death on city growth growth were greatest in areas which were comparatively undeveloped
before the plague thereby providing some evidence for an urban reset.
6. CONCLUSION
Numerous scholars have argued that the Black Death marks a watershed in European history, after
which it is possible to detect trends that would eventually lead to the onset of sustained economic
growth (Gottfried, 1983; Herlihy, 1997; Epstein, 2000; Pamuk, 2007; Acemoglu & Robinson, 2012).
Existing accounts have focused on the fertility response to the Black Death (Moor & Zanden, 2010;
Voigtländer & Voth, 2013a), its effect on real wages and GDP (Allen, 2001; Broadberry, 2013), or its
impact on political institutions (Epstein, 2000; Acemoglu & Robinson, 2012). Recent work has focused
on the effect of the Black Death on urbanization in Europe in comparison to China (Voigtländer &
Voth, 2013b).
23
Our unique dataset of urban mortality allows us to explore the spatial dimension of the impact of the
Black Death on urban development in preindustrial Europe. Using a range of identification strategies
involving two IV strategies we find that the Black Death had a large impact on urban population in the
short-run. On average cities recovered on from the Black Death within two centuries. However, this
aggregate convergence masks divergence at a local level. Some cities recovered rapidly while others
declined. We find that cities recovered fastest from the Black Death if they had reached a threshold
size by 1300. This is consistent with an account where increasing returns play a role in determining
the location of urban activity. Cities that belonged to the Hanseatic league or to more centralized
monarchies also recovered more quickly. We conduct a cell-level analysis which suggests that the
Black Death shock led to the reallocation of urban activities to areas which previous lacked large
urban settlements.
24
References
Acemoglu, Daron & James A. Robinson (2012), Why Nations Fail, Crown Business, New York.
Acemoglu, Daron & Simon Johnson (2007), ‘Disease and Development: The Effect of Life Expectancy on
Economic Growth’, Journal of Political Economy 115(6), 925–985.
Acemoglu, Daron, Simon Johnson & James Robinson (2005), ‘The rise of Europe: Atlantic trade, institutional
change, and economic growth’, American Economic Review 95(3), 546–579.
Acemoglu, Daron, Tarek A. Hassan & James A. Robinson (2011), ‘Social structure and development: A legacy of
the Holocaust in Russia’, The Quarterly Journal of Economics 126(2), 895–946.
Akbulut-Yuksel, Mevlude & Mutlu Yuksel (2015), ‘The Long-Term Direct and External Effects of Jewish Expulsions
in Nazi Germany’, American Economic Journal: Applied Economics Forthcoming.
Alfani, Guido (2013), ‘Plague in seventeenth-century Europe and the decline of Italy: an epidemiological
hypothesis’, European Review of Economic History .
Allen, Robert C. (2001), ‘The great divergence in European wages and prices from the middle ages to the first
world war’, Explorations in Economic History 38, 411–447.
Almond, Douglas (2006), ‘Is the 1918 influenza pandemic over? longâĂŘterm effects of in utero influenza
exposure in the postâĂŘ1940 u.s. population’, Journal of Political Economy 114(4), 672–712.
Alsan, Marcella (2015), ‘The effect of the TseTse fly on African development’, American Economic Review
105(1), 382–410.
Ashraf, Quamrul & Oded Galor (2011), ‘Dynamics and stagnation in the Malthusian epoch’, American Economic
Review 101(5), 2003–41.
Ashton, T.H. & C.H.E. Philpin, eds (1985), The Brenner Debate, Agarian Class Structure and Economic Development
in Pre-Industrial Europe, Cambridge University Press, Cambridge.
Bairoch, Paul (1988), Cites and Economic Development, from the dawn of history to the present, University of
Chicago Press, Chicago. translated by Christopher Braider.
Baten, Jörg, Dorothee Crayen & Hans-Joachim Voth (2013), ‘Numeracy and the impact of high food prices in
industrializing Britain, 1780–1850’, Review of Economics and Statistics 96(3), 418–430.
Becker, Gary S., Edward L. Glaeser & Kevin M. Murphy (1999), ‘Population and economic growth’, American
Economic Review 89(2), 145–149.
Benedictow, Ole J. (2005), The Black Death 1346–1353: The Complete History, The Boydell Press, Woodbridge.
Bleakley, Hoyt & Fabian Lange (2009), ‘Chronic Disease Burden and the Interaction of Education, Fertility, and
Growth’, The Review of Economics and Statistics 91(1), 52–65.
Bleakley, Hoyt & Jeffrey Lin (2012), ‘Portage and Path Dependence’, The Quarterly Journal of Economics
127(2), 587–644.
Bloom, David E., David Canning & Günther Fink (2014), ‘Disease and development revisited’, Journal of Political
Economy 122(6), 1355–1366.
Boccaccio, Giovanni (2005, 1371), The Decameron, Dodo Press, London. Translated by J.M. Rigg.
Boerner, Lsrs & Battista Severgnini (2014), Epidemic trade. LSE Economic History Working Paper No. 212/2014.
Bos, Kirsten I, Verena J Schuenemann, G Brian Golding, Hernán A Burbano, Nicholas Waglechner, Brian K
25
Coombes, Joseph B McPhee, Sharon N DeWitte, Matthias Meyer, Sarah Schmedes et al. (2011), ‘A draft
genome of yersinia pestis from victims of the black death’, Nature 478(7370), 506–510.
Bosker, Maarten, Eltjo Buringh & Jan Luiten van Zanden (2013), ‘From Baghdad to London: unravelling urban
development in Europe and the Arab world 800-1800’, Review of Economics and Statistics 95(4), 1418–1437.
Bosker, Maarten, Steven Brakman, Harry Garretsen & Marc Schramm (2007), ‘Looking for multiple equilibria
when geography matters: German city growth and the WWII shock’, Journal of Urban Economics 61(1), 152–
169.
Bosker, Maarten, Steven Brakman, Harry Garretsen & Marc Schramm (2008), ‘A century of shocks: The evolution
of the German city size distribution 1925-1999’, Regional Science and Urban Economics 38(4), 330–347.
Bosker, Maarten, Steven Brakman, Henry Garretson, Herman De Jogn & Marc Schramm (2008), ‘Ports, plagues
and politics: explaining Italian city growth 1300-1861’, European Review of Economic History 12, 97–131.
Brenner, Robert (1976), ‘Agarian class structure and economic development in pre-industrial Europe’, Past and
Present 70(1), 30–75.
Broadberry, Stephen (2013), Accounting for the great divergence.
Campante, Filipe & Edward L. Glaeser (2009), Yet another tale of two cities: Buenos Aires and Chicago, Working
Paper 15104, National Bureau of Economic Research.
Campbell, Bruce M. S. (2008), ‘Benchmarking medieval economic development: England, Wales, Scotland, and
Ireland, c.1290’, Economic History Review 61(4), 896–945.
Cantor, Norman F. (2001), In the Wake of the Plague: The Black Death and the World it Made, Free Press, New
York.
Chandler, Tertius (1974, 1987), Four Thousand Years of Urban Growth, St David University Press, Lewiston, NY.
Chaney, Eric & Richard Hornbeck (2015), ‘Economic growth in the Malthusian era: Evidence from the 1609
Spanish expulsion of the Moriscos’, Economic Journal Forthcoming.
Christakos, George, Richardo A. Olea, Marc L. Serre, Hwa-Lung Yu & Lin-Lin Wang (2005), Interdisciplinary
Public Health Reasoning and Epidemic Modelling: The Case of Black Death, Springer, Berlin.
Cohn, Samuel K. (2003), The Black Death Transformed: Disease and Culture in Early Renaissance Europe, Hodder
Arnold, London.
Cohn, Samuel K. & Guido Alfani (2007), ‘Households and plague in early modern Italy’, The Journal of
Interdisciplinary History 38(2), pp. 177–205.
Conley, Timothy G. (2008), spatial econometrics, in S. N.Durlauf & L. E.Blume, eds, ‘The New Palgrave Dictionary
of Economics’, Palgrave Macmillan.
Davis, Donald R. & David E. Weinstein (2002), ‘Bones, Bombs, and Break Points: The Geography of Economic
Activity’, American Economic Review 92(5), 1269–1289.
De Long, J. Bradford & Andrei Shleifer (1993), ‘Princes and merchants: European city growth before the
industrial revolution’, Journal of Law & Economics 36(2), 671–702.
DeWitte, Sharon N & James W Wood (2008), ‘Selectivity of Black Death mortality with respect to preexisting
health’, Proceedings of the National Academy of Sciences 105(5), 1436–1441.
Dincecco, Mark & Massimiliano Gaetano Onorato (2015), Military conflict and the rise of urban Europe. working
26
paper.
Dittmar, Jeremiah E. (2011), ‘Information technology and economic change: The impact of the printing press’,
The Quarterly Journal of Economics 126(3), 1133–1172.
Dittmar, Jermiah (2015), Cities, markets, and growth: The emergence of Zipf’s law.
Dollinger, Philippe (1970), The German Hansa, MacMillan, London. translated and edited by D.S. Ault and S.H.
Steinberg.
Edwards, Jeremy & Sheilagh Ogilvie (2012), ‘What lessons for economic development can we draw from the
Champagne fairs?’, Explorations in Economic History 49(2), 131–148.
Epstein, S. R. (2000), Freedom and Growth, the rise of states and markets in Europe, 1300–1700, Routledge,
London.
Erb, Karl-Heinz, Veronika Gaube, Fridolin Krausmann, Chritoph Plutzar, Alberte Bondeau & Helmut Haberl
(2007), ‘A comprehensive global 5 min resolution land-use data set for the year 2000 consistent with national
census data’, Journal of Land Use Science 2(3), 191–224.
Findlay, Ronald & Mats Lundahl (2006), Demographic Shocks and the Fator Proportions Model: From the
Plague of Justinianto the Black Death, in R.Findlay, R. G. H.Henriksson, H.Lindgren & M.Lundahl, eds, ‘Eli
Heckscher, International Trade, and Economic History’, MIT Press, Cambridge MA.
Fischer, Gunther, Harrij van Nelthuizen, Mahendra Shah & Freddy Nachtergaele (2002), Global Agro-Ecological
Assessment for Agriculture in the 21st Century: Methodology and Results, Food and Agriculture Organization of
the United Nations, Rome.
Galor, Oded (2011), Unified Growth Theory, Princeton University Press, Princeton, New Jersey.
Galor, Oded & David N. Weil (2000), ‘Population, technology, and growth: From Malthusian stagnation to the
demographic transition and beyond’, American Economic Review 90, 806–828.
Gennaioli, Nicola & Hans-Joachim Voth (2015), ‘State capacity and military conflict’, Review of Economic Studies
. Forthcoming.
Gensini, Gian Franco, Magdi H. Yacoub & Andrea A. Conti (2004), ‘The concept of quarantine in history: from
plague to SARS’, Journal of Infection 49(4), 257 – 261.
Glocker, Daniela & Daniel Sturm (2014), War-time destruction and the persistence of economic activity. memo.
Gottfried, Robert S. (1983), The Black Death: Natural and Human Disaster in Medieval Europe, The Free Press,
New York.
Haensch, Stephanie, Raffaella Bianucci, Michel Signoli, Minoarisoa Rajerison, Michael Schultz, Sacha Kacki,
Marco Vermunt, Darlene A. Weston, Derek Hurst, Mark Achtman, Elisabeth Carniel & Barbara Bramanti
(2010), ‘Distinct clones of yersinia pestis caused the Black Death’, PLoS Pathogens 6(0).
Hajnal, John (1965), European marriage patterns in perspective, in D. V.Glass & D. E. C.Eversley, eds, ‘Population
in History: Essays in Historical Demography’, London, U.K.
Hansen, Gary D. & Edward C. Prescott (2002), ‘Malthus to solow’, American Economic Review 92(4), 1205–1217.
Herlihy, David (1997), The Black Death and the transformation of the West, Havard University Press, Cambridge,
MA.
Hohenberg, Paul M (2004), The historical geography of European cities: an interpretive essay, in V.Henderson &
27
J.Thisse, eds, ‘Handbook of Regional and Urban Economics’, Vol. 4, Elsevier, pp. 3021–3052.
Iyigun, Murat, Nathan Nunn & Nancy Qian (2015), Climate change and conflict: Evidence from 1400-1900.
Memo.
Jarvis, A., H.I. Reuter, A. Nelson & E. Guevara (2008), Hole-filled seamless srtm data v4. International Centre
for Tropical Agriculture (CIAT).
Jebwab, Remi, Edward Kerby & Alexander Moradi (2015), ‘History, path dependence and development: Evidence
from colonial railroads, settlers and cities in Kenya’, Economic Journal Forthcoming.
Kremer, Michael (1993), ‘Population growth and technological change: One million b.c. to 1990’, The Quarterly
Journal of Economics 108(3), 681–716.
Krugman, Paul (1991a), ‘History versus Expectations’, The Quarterly Journal of Economics 106(2), 651–67.
Krugman, Paul (1991b), ‘Increasing Returns and Economic Geography’, Journal of Political Economy 99(3), 483–
99.
Kussmaul, Ann (1985), ‘Time and space, hoofs and grain: The seasonality of marriage in england’, Journal of
Interdisciplinary History 15(4), 755–779.
Lagerlöf, Nils-Petter (2003), ‘Mortality and Early Growth in England, France and Sweden’, Scandinavian Journal
of Economics 105(3), 419–440.
Lindberg, Erik (2008), ‘The rise of Hamburg as a global marketplace in the seventeenth century: A comparative
political economy perspective’, Comparative Studies in Society and History 50, 641–662.
Luterbacher, J., D. Dietrich, E. Xoplaki, M. Grosjean & H. Wanner (2004), ‘European seasonal and annual
temperature variability, trends, and extremes since 1500’, Science 303, 1499–1503.
Macfarlane, Alan (1986), Marriage and Love in England: modes of reproduction 1300-1400, Basil Blackwell,
Oxford.
Malanima, Paolo (2010), Energy and population in europe: The medieval growth (10th-14th centuries). Institute
of Studies on Mediterranean Societies (ISSM) and National Research Council (CNR).
Malanima, Paulo (2009), Pre-Modern European Economy: One Thousand Years (10th-9th Centuries), Brill, Leiden.
Maloney, William F. & Felipe Valencia Caicedo (2015), ‘The persistence of (subnational) fortune’, Economic
Journal Forthcoming.
McCormick, Michael, Guoping Huang, Giovanni Zambotti & Jessica Lavash (2013), Roman road network.
DARMC Scholarly Data Series, Data Contribution Series 2013-5. DARMC, Center for Geographic Analysis,
Harvard University, Cambridge MA 02138.
McEvedy, Colin & Richard Jones (1978), Atlas of World Population History, Facts of File, New York.
McNeil, William H. (1974), Plagues and Peoples, Basil Blackwell, Oxford.
Meng, Xin, Nancy Xian & Pierre Yared (2015), ‘The institutional causes of china’s great famine, 1959-1961’,
Review of Economic Studies Forthcoming.
Miguel, Edward & Gérard Roland (2011), ‘The long-run impact of bombing Vietnam’, Journal of Development
Economics 96(1), 1 – 15.
Moor, Tine De & Jan Luiten Van Zanden (2010), ‘Girl power: the European marriage pattern and labour markets
in the North Sea region in the late medieval and early modern period’, Economic History Review 63(1), 1–33.
Nicholas, David (1997), The later medieval city, 1300-1500, Longman Limited, Harlow.
28
Nicholas, David (1999), The Transformation of Europe 1300-1600, Arnold Publishers, London.
North, Douglass C., John Joseph Wallis & Barry R. Weingast (2009), Violence and Social Orders: a conceptual
framework for interpreting recorded human history, Cambridge University Press, Cambridge.
Noymer, Andrew (2007), ‘Contesting the cause and severity of the Black Death: A review essay’, Population and
Development Review 33(3), 616–627.
Nunn, Nathan (2008), ‘The long-term effects of africa’s slave trades’, Quarterly Journal of Economics, Vol
123(1), pg. 139–176.
Nunn, Nathan & Diego Puga (2012), ‘Ruggedness: The blessing of bad geography in Africa’, The Review of
Economics and Statistics 94(1), 20–36.
Nunn, Nathan & Nancy Qian (2011), ‘The potato’s contribution to population and urbanization: Evidence from
a historical experiment’, The Quarterly Journal of Economics 126(2), 593–650.
Nussli, Christos (2011). URL: http://www.euratlas.com/about.html.
Pamuk, Şevket (2007), ‘The Black Death and the origins of the ‘Great Divergence’ across Europe, 1300-1600’,
European Review of Economic History 11(3), 289–317.
URL: http://www.jstor.org/stable/41378468
Piot, Peter, Michael Bartos, Peter D. Ghys, Neff Walker & Bernhard Schwartlander (2001), ‘The global impact of
HIV/AIDS’, Nature 410(6831), 968–973.
Pounds, Norman J. G. (1973), An Historical Geography of Europe: 450 B.C.–A.D. 1330, Cambridge University
Press, Cambridge.
Ramankutty, N.J., J. Foley, J. Norman & K. McSweeney (2002), ‘The global distribution of cultivable lands:
Current patters and sensitivity to possible climate change’, Global Ecology and Biogeography 11, 377–392.
Rauch, Ferdinand & Guy Michaels (2013), Resetting the Urban Network: 117-2012, Economics Series Working
Papers 684, University of Oxford, Department of Economics.
Redding, Stephen J., Daniel M. Sturm & Nikolaus Wolf (2011), ‘History and Industry Location: Evidence from
German Airports’, The Review of Economics and Statistics 93(3), 814–831.
Rotz, Rhiman A. (1977), ‘The Lubeck Uprising of 1408 and the Decline of the Hanseatic League’, Proceedings of
the American Philosophical Society 121(1), 1–45.
Russell, Josiah Cox (1972), Medieval Regions and their Cities, David & Charles, Newton Abbot, Devon.
Sachs, Jeffrey & Pia Malaney (2002), ‘The economic and social burden of malaria’, Nature 415(6872), 680–685.
Schuenemann, Verena J, Kirsten Bos, Sharon DeWitte, Sarah Schmedes, Joslyn Jamieson, Alissa Mittnik, Stephen
Forrest, Brian K Coombes, James W Wood, David JD Earn et al. (2011), ‘Targeted enrichment of ancient
pathogens yielding the ppcp1 plasmid of yersinia pestis from victims of the Black Death’, Proceedings of the
National Academy of Sciences of the United States of America 108(38), E746–E752.
Scott, Susan & Christopher J. Duncan (2001), Biology of plagues evidence from historical populations, Cambridge
University Press, Cambridge.
Shepherd, William R. (1923), Historical Atlas, Henry Holt and Company, New York.
Snyder, Timothy (2010), Bloodlands: Europe Between Hitler and Stalin, The Bodley Head, London:.
Spufford, Peter (2002), Power and Profit: the merchant in medieval Europe, Thames & Hudson, London.
Stasavage, David (2014), ‘Was Weber right? the role of urban autonomy in Europe’s rise’, American Political
29
Science Review 108, 337–354.
Theilmann, John & Frances Cate (2007), ‘A plague of plagues: The problem of plague diagnosis in medieval
england’, The Journal of Interdisciplinary History 37(3), 371–393.
Tilly, Charles (1990), Coercion, Capital, and European States, AD 990-1990, Blackwell, Oxford.
Twigg, Graham (1984), The Black Death: a biological reapprasisal, Batsford Academic and Educational, London.
van Zanden, Jan Luiten (2009), The Long Road to the Industrial Revolution. The European Economy in a Global
Perspective, 1000-1800, Brill Publishers, Leiden.
Voigtländer, Nico & Hans-Joachim Voth (2013a), ‘How the west invented fertility restriction’, American Economic
Review 103(6), 2227–2264.
Voigtländer, Nico & Hans-Joachim Voth (2013b), ‘The three horsemen of riches: Plague, war, and urbanization
in early modern Europe’, Review of Economic Studies 80, 774–811.
Waldinger, Fabian (2010), ‘Quality matters: The expulsion of professors and the consequences for phd student
outcomes in Nazi Germany’, Journal of Political Economy 118(4), 787–831.
Waldinger, Fabian (2012), ‘Peer effects in science: Evidence from the dismissal of scientists in Nazi Germany’,
Review of Economic Studies 79(2), 838–861.
Weil, David (2014), The impact of malaria on African development over the longue duree, in E.Akyeampong,
R.Bates, N.Nunn & J. A.Robinson, eds, ‘Africa’s Development in Historical Perspective’, Cambridge University
Press, pp. 89–130.
Young, Alwyn (2005), ‘The Gift of the Dying: The Tragedy of AIDS and the Welfare of Future African Generations’,
Quarterly Journal of Economics 120(2), 423–466.
Zanden, Jan Luiten Van, Eltjo Buringh & Maarten Bosker (2012), ‘The rise and decline of European parliaments,
1188–1789’, Economic History Review 65(3), 835–861.
Ziegler, Philip (1969), The Black Death, Collins, London.
30
FIGURE 1: Evolution of Europe’s Total Population and Urbanization Rate, 1100-1600
Panel A shows the respective evolutions of the total population (millions) and urbanization rate (%) of the
17 European countries of our main analysis. Panel B presents the evolution of the total urban population
(millions). The modern countries in our sample are Austria, Belgium, Czech Republic, Denmark , France,
Germany, Italy, Luxembourg, Norway, Poland, Portugal, Spain, Sweden, Switzerland, the Netherlands, and
the United Kingdom. The main sources for total population are Malanima (2009) and Malanima (2010).
The urbanization rate is defined as the share of all the localities above 1,000 in the total population. Total
(urban) population in 1353 is proxied by the total (urban) population in 1300 times the average (urban)
mortality rate in 1347-1353, which we estimate as 40% (37.5%). City population data is from Chandler
(1974, 1987) and Bairoch (1988).
FIGURE 2: Distribution of City Sizes for the Existing Cities in 1300, 1300-1600
This figure shows the Kernel distribution of city sizes (base 100 in 1300) for the 139 existing cities in 1300
31
for which we also know their Black Death mortality rate, for the years 1353, 1400, 1500 and 1600. These
cities belong to the 17 European countries of our main analysis (see the notes below Figure 1 for a list).
Their population in 1353 is proxied by their population in 1300 times their Black Death mortality rate in
1347-1353. The sources for the city population data are Chandler (1974, 1987) and Bairoch (1988). For ease
of readability we exclude 21 outlying observations whose population is more than 3 times their population
in 1300.
FIGURE 3: Black Death Mortality Rates in 1347-1353
!
!
!
!
!
!
!!
!
!
!
!!
!
!
!
! !
!!
!!!!!
!!! !
!!!!
!
Cities With Mortality Data
! !
(mortality rates)
!
!
!
!
!!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
25 - 50
!
50 - 75
!
75 - 100
!
!
!
!
!
!
!
!!!
!
!!! !
!
!
!
!
!!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
! !
! !
! ! ! !
!
!
! ! ! !
!
!!
! !! ! ! !
!
!!
!
!
! !
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
! !
!
!
!
0 - 25
!
!
! !
!
!
!
!
!
!!
! !!
!!!
!! !
!
!
!
!!!
! !!
!!!
!
!
! !!
This map plots the location of all 139 cities for which we have mortality estimates for the Black
Death in 1347-1353.
FIGURE 4: Morality Rates, City Populations and Recovery
b. Heterogenous Response of Urban Growth
a.
Rates and 1300 Population.
Mortality rates for 139 existing cities in 1300 for
which we have mortality data. This figure
shows that mortality rates are uncorrelated
with population in 1300 Sources: see web
appendix
to the Black Death, It plots % change in city
population for the 139 cities for which we
have mortality data between 1300-1600. To
improve visibility we excluded the top and
bottom 5% of observations. Source: see
web appendix
FIGURE 5: Timing of the Onset of the Black Death and Black Death Mortality: IV(1)
This figure depicts the relationship between Black Death mortality and the timing of the onset of
the Black Death. It demonstrates the validity of our first IV. Cities which were affected by the Black
Death earlier had higher morality rates. Source Christakos et al. (2005).
FIGURE 6: Duration of the Black Death
This figure shows that the average duration of the Black Death was 7 months. We use this
relationship to construct our second IV which relies on the month of Black Death onset as a source
of exogenous variation. Source Christakos et al. (2005).
FIGURE 7: Month of Black Death Onset and Black Death Mortality: IV (2)
This figure shows the relationship between Black Death mortality and the month of onset (within
year). The Black Death was more virulent during the summer months. Given the duration of the
Black Death, month of onset, therefore, provides exogenous variation in the mortality rate. Source
Christakos et al. (2005)
FIGURE 8: Cell Level Analysis: Testing for ‘Horsemen Effects’
This figure illustrates our cell level analysis. We use 1 by 1 degree grids. We distinguish between
cells that had a city in 1300 from those cells which acquired a city between 1300 and 1600 or did
not have a city throughout the period of analysis.
TABLE 1: BLACK DEATH MORTALITY RATES AND CITY GROWTH, 1100-1700
Dependent Variable: Percentage Change in City Population (%) in Period t
t:
1300-1400
(1)
1100-1200
(2)
1200-1300
(3)
1400-1500
(4)
1500-1600
(5)
1300-1500
(6)
1300-1600
(7)
1300-1700
(8)
β
-0.84***
[0.31]
[-1.5;-0.2]
-0.15
[0.37]
[-0.9;0.6]
0.05
[0.62]
[-1.2;1.3]
0.46**
[0.21]
[0.1;0.9]
0.39
[0.29]
[-0.2;1.0]
-0.41
[0.32]
[-1.0;0.2]
-0.13
[0.51]
[-1.1;0.9]
0.09
[0.85]
[-1.6;1.8]
59
0.00
87
0.00
139
0.02
139
0.01
139
0.01
139
0.00
139
0.00
Obs.s
R2
140
0.11
Notes: This table shows the effect β t of the Black Death mortality rate (%) in 1347-1352 on the percentage change in city population (%) for various
periods t. The main sample consists of 140 cities (i.e. localities ≥ 1,000 inh.) that already existed in 1300 and for which the mortality rate is
available. We use the population of each city in the initial year as regression weights. Robust SE’s: * p<0.10, ** p<0.05, *** p<0.01. The 95%
confidence level intervals are shown into brackets. See Web Data Appendix for data sources.
TABLE 2: CITY CHARACTERISTICS AND BLACK DEATH MORTALITY RATES
Black Death Mortality Rate (%, 1347-1352)
Dependent Variable:
(1)
(2)
(3)
(4)
Locational Fundamentals:
Average Temperature 1500-1600 (d)
Elevation (m)
Cereal Suitability 25 Km
Grazing Suitability 25 Km
Coast 10 Km Dummy
Rivers 10 Km Dummy
Longitude (d)
Latitude (d)
-1.09
-0.02
2.80
6.66
3.76
-1.10
-0.14
-1.51**
[1.26]
[0.01]
[1.79]
[7.48]
[3.27]
[2.97]
[0.23]
[0.74]
-1.22
-0.02
2.86
5.58
3.02
0.28
0.11
-1.75**
[1.50]
[0.01]
[1.86]
[8.51]
[4.08]
[3.46]
[0.38]
[0.83]
-1.99
-1.46
-0.94
-1.44
6.05
2.52
-1.52
-4.03
5.71
1.99
[2.21]
[6.23]
[7.04]
[5.14]
[5.72]
[3.25]
[5.78]
[4.22]
[6.12]
[4.76]
3.32
1.49
0.46
2.01
-2.10
[5.65]
[4.81]
[4.29]
[4.60]
[3.37]
Increasing Returns:
Log City Population in 1300
Maj. Roman Rd (MRR) 10 Km Dummy
MRR Intersection 10 Km Dummy
Any Roman Rd (ARR) 10 Km Dummy
ARR Intersection 10 Km Dummy
Medieval Route (MR) 10 Km Dummy
MR Intersection 10 Km Dummy
Market & Fair Dummy
Hanseatic League Dummy
University Dummy
-0.10
3.42
0.27
2.60
-0.70
1.59
-3.92
-6.06
0.92
3.18
[1.64]
[7.32]
[7.12]
[4.83]
[5.47]
[3.19]
[5.17]
[3.95]
[4.88]
[4.41]
Institutions:
Monarchy in 1300 Dummy
State Capital in 1300 Dummy
Autonomous City in 1300 Dummy
Parliamentary Activity in 1300-1400
Battle w/i 100 Km in 1300-50 Dummy
Obs.; R2
3.78
0.80
-3.46
1.80
-4.70
36
140; 0.15
140; 0.06
[4.67]
[4.19]
[3.98]
[3.89]
[3.08]
140; 0.07
140; 0.21
Notes: This table shows the effects of various city characteristics proxying for locational fundamentals, increasing returns and institutions
on the Black Death mortality rates (%) in 1347-1352. We use the main sample of 140 cities. Columns (1)-(4) represent four different
regressions. Robust SE’s: * p<0.10, ** p<0.05, *** p<0.01. See Web Data Appendix for data sources.
TABLE 3: BLACK DEATH MORTALITY RATES AND CITY GROWTH, INVESTIGATION OF
CAUSALITY
Dependent Variable: Percentage Change in City Population (%) in Period t
(1) t = 1300-1400
Regression:
1. Baseline
2. Controls: Locational Fundamentals
3. Controls: Increasing Returns
4. Controls: Institutions
5. Controls: All
6. Row (5) + Battle in 1350-1600
7. Row (5) + Dummies for Each Coastline
8. Row (5) + Dummies if Jewish Presence and Persecution
9. Fixed Effect for Holy Roman Empire
10. Fixed Effects for 13 Countries in 2015
11. Fixed Effects for 57 States in 1300
12. Fixed Effects for 7 States in 1300 (for States with > 3 Cities)
13. Unconditional IV 1: #Months of Infection since Oct 1347 (F: 18.4)
14. Conditional IV 1: #Months of Infection since Oct 1347 (F: 41.3)
15. Unconditional IV 2: Month of First Infection Dummies (F: 9.2)
16. Conditional IV 2: Month of First Infection Dummies (F: 4.4)
17. Unconditional IV: Both Sets of Instruments (F: 8.5)
18. Conditional IV: Both Sets of Instruments (F: 7.0)
-0.84***
-0.76***
-0.72***
-0.89***
-0.67***
-0.81***
-0.70***
-0.64***
-0.80***
-0.59**
-0.71†
-1.09**
-1.63**
-1.01***
-1.96***
-1.14***
-1.71***
-1.10***
[0.31]
[0.28]
[0.27]
[0.28]
[0.22]
[0.21]
[0.21]
[0.23]
[0.30]
[0.28]
[0.43]
[0.43]
[0.72]
[0.37]
[0.56]
[0.33]
[0.52]
[0.29]
(2) t = 1300-1600
-0.13
0.24
-0.29
-0.18
-0.30
-0.44
-0.34
-0.15
-0.04
-0.35
-0.59
-0.37
-1.09
0.25
-1.26
-1.13
-0.89
-0.39
[0.51]
[0.51]
[0.46]
[0.51]
[0.57]
[0.57]
[0.56]
[0.60]
[0.50]
[0.53]
[1.20]
[0.70]
[1.23]
[1.18]
[0.93]
[0.77]
[0.83]
[0.81]
Notes: This table shows the effect β t of the Black Death mortality rate (%) in 1347-1352 on the percentage change in city population (%) for
various periods t. We use the main sample of 140 cities. Row 1: Main results from Table 1. Rows 2-5: Controlling for locational fundamentals,
increasing returns, institutions, or all of them together. Row 6: All controls + one dummy for whether a battle occurred within 100 km from the
city in 1350-1600. Row 7: All controls + three dummies if within 10 km from the Mediterranean Sea, the Atlantic Ocean or the North-Baltic Sea.
Row 8: All controls + two dummies if Jews were likely to be present in the city in 1347 and were persecuted in 1347-1352. Row 9: Adding a
dummy for the Holy Roman Empire. Row 10: Adding 13 country FE. Rows 11-12: Adding 57 state in 1300 FE (row 12: excl. the states with ≤ 3
obs.). Rows 13-14: Instrumenting by the number of months between the city-specific date (month-year) of first infection and October 1347, the
month Messina was first infected (row 14: also adding all the controls and quartics of latitude and longitude). Rows 15-16: Instrumenting by
twelve dummies for the month at the peak of the infection (approx. 3.5 months after the month of first infection), while simultaneously controlling
for the year of infection (row 16: also adding all the controls and quartics of latitude and longitude). Robust SE’s: † p<0.15, * p<0.10, ** p<0.05,
*** p<0.01. See Web Data Appendix for data sources.
37
TABLE 4: BLACK DEATH MORTALITY RATES AND CITY GROWTH: MEASUREMENT
ISSUES
Dependent Variable: Percentage Change in City Population (%) in Period t
Regression:
(1) t = 1300-1400
1. Baseline
2. City Population Data: Bairoch Only
3. Dummies for Type of Mortality Data
4. Excluding Description-Based Mortality Data (Obs. = 115)
5. Excluding Desertion-Based Mortality Data (Obs. = 119)
6. Excluding Clergy-Based Mortality Data (Obs. = 135)
7. Raw Mortality Data (Obs. = 89)
8. Excluding Mortality Equal to 25% or 50% (Obs. = 98)
9. Extrapolated Rates Based on 140 Cities in 1300 (Obs. = 467)
10. Extrapolated Rates Based on 186 Localities (Obs. = 467)
11. Extrapolated Rates Based on 89 Cities w/ Raw Data (Obs. = 467)
-0.84***
-0.71*
-0.91***
-0.81**
-0.96***
-0.84***
-0.96**
-0.78**
-0.73***
-0.71***
-0.71***
[0.31]
[0.43]
[0.33]
[0.36]
[0.34]
[0.31]
[0.41]
[0.33]
[0.22]
[0.22]
[0.27]
(2) t = 1300-1600
-0.13
0.04
-0.24
-0.31
-0.21
-0.11
-0.38
-0.13
0.30
0.33
0.78
[0.51]
[0.53]
[0.55]
[0.58]
[0.57]
[0.51]
[0.67]
[0.52]
[0.50]
[0.48]
[0.61]
Notes: This table shows the effect β t of the Black Death mortality rate (%) in 1347-1352 on the percentage change in city population (%) for various
periods t. We use the main sample of 140 cities. Row 1: Main results from Table 1. In Rows 2: we employ the raw Bairoch population data. Row 3:
Including dummies for the type of mortality data: raw number (N = 89), literary description (25), desertion rates (21) and clergy mortality rates
(5). Rows 4-7: Excluding the description-based, desertion-based or clergy-based mortality rates, or all of them altogether. Row 8: Dropping the
observations with mortality rates equal to 25% or 50%. Rows 9-11: Using spatially extrapolated mortality rates based on 140 cities in 1300, 186
localities in 1300, and 89 cities in 1300 with existing raw mortality data respectively. The regressions are estimated for the 467 cities that existed
in 1300. Robust SE’s: * p<0.10, ** p<0.05, *** p<0.01. See Web Data Appendix for data sources.
38
TABLE 5: BLACK DEATH MORTALITY RATES AND CITY GROWTH: SPECIFICATION
CHECKS
Dependent Variable: Percentage Change in City Population (%) in Period t
Regression:
(1) t = 1300-1400
1. Baseline
2. SEs Clustered at Country in 2015 Level (N = 13)
3. SEs Clustered at State in 1300 Level (N = 57)
4. Conley-Type SEs (100 km)
5. Adding % Change in Pop. in 1200-1300 (Obs. = 87)
6. Adding % Changes in Pop. in 11-1200 & 12-1300 (Obs. = 59)
7. Dropping Pop. Weights & Excl. Cities < 5,000 in 1300 (Obs. = 111)
8. Regressing Log Change in Pop. on Log Mortality (Obs. = 134)
9. Regressing Log Change in Pop. on Log Mortality & Log Initial Pop.
10. Regressing Pop. on the Estimated Number of Dead and Initial Pop.
11. Panel Model with City FE and Year FE in 1200-1600 (Obs. = 564)
-0.84***
-0.84**
-0.84***
-0.84***
-1.10***
-1.28**
-0.78**
-0.34***
-0.33***
-0.87*
-1.07*
[0.31]
[0.37]
[0.29]
[0.20]
[0.38]
[0.52]
[0.32]
[0.11]
[0.10]
[0.46]
[0.56]
(2) t = 1300-1600
-0.13
-0.13
-0.13
-0.13
-0.14
-0.38
-0.23
-0.12
-0.15
0.00
-0.05
[0.51]
[0.39]
[0.57]
[0.60]
[0.61]
[0.78]
[0.42]
[0.16]
[0.16]
[0.64]
[0.52]
Notes: This table shows the effect β t of the Black Death mortality rate (%) in 1347-1352 on the percentage change in city population (%) for various
periods t. We use the main sample of 140 cities. Row 1: Main results from Table 1. Rows 2-3: Clustering observations at the country in 2015 level
(N = 13) and the state in 1300 level (N = 57). Row 4: Using Conley-type standard errors specifying that cities are spatially autocorrelated within
100 km from each other. Rows 5-6: Controlling the percentage change(s) in city pop. in 1200-1300 only or in both 1100-1200 and 1200-1300
respectively. Row 7: Dropping the population in 1300 weights and excluding the cities below 5,000 inh. in 1300. Rows 8-9: We regress the log
difference in city pop. on log mortality (losing 6 cities for which it is 0), and then also controls for log initial city pop. Row 10: Regressing city
pop. on the number of dead (estimated as the pop. in 1300 times the mortality rate in 1347-1352) and the initial city pop. Row 11: Running a
panel model with city FE and year FE in 1200-1600. We show the relative effects of the mortality rate in 1300 and 1600. Robust SE’s: * p<0.10, **
p<0.05, *** p<0.01. See Web Data Appendix for data sources.
TABLE 6: BLACK DEATH MORTALITY RATES AND CITY GROWTH: SAMPLING ISSUES
Dependent Variable: Percentage Change in City Population (%) in Period t
Regression:
(1) t = 1300-1400
1. Baseline
2. Dropping Top 1% Change in Pop. (Obs. = 139)
3. Dropping Top 5% Change in Pop. (Obs. = 133)
4. Dropping Top 10% Change in Pop. (Obs. = 126)
5. Dropping if Mortality = 0% (Obs. = 134)
6. Dropping if Mortality ≥ 80% (Obs. = 137)
7. Excluding France (N = 33; Obs. = 107)
8. Excluding Germany (N = 29; Obs. = 111)
9. Excluding Italy (N = 22; Obs. = 118)
10. Excluding the United Kingdom (N = 19; Obs. = 121)
11. Excluding Spain (N = 14; Obs. = 126)
-0.84***
-0.89***
-0.59**
-0.55**
-0.86**
-0.89***
-0.75**
-1.04***
-1.09***
-0.80**
-0.92***
[0.31]
[0.31]
[0.23]
[0.25]
[0.34]
[0.33]
[0.38]
[0.36]
[0.41]
[0.311]
[0.32]
(2) t = 1300-1600
-0.13
0.01
-0.08
-0.29
-0.34
-0.15
-0.21
-0.28
0.14
-0.15
-0.07
[0.51]
[0.50]
[0.49]
[0.46]
[0.57]
[0.55]
[0.63]
[0.57]
[0.77]
[0.510]
[0.50]
Notes: This table shows the effect β t of the Black Death mortality rate (%) in 1347-1352 on the percentage change in city population (%) for various
periods t. We use the main sample of 140 cities. Row 1: Main results from Table 1. Rows 2-4: Dropping the top 1, 5 and 10% changes in pop.
Rows 5-6: Dropping if mortality rates are equal to 0 or 80 respectively. Rows 7-11: Excluding the observations of the most represented countries
in the sample: France (N = 33), Germany (N = 29), Italy (N = 22), the United Kingdom (N = 19) and Spain (N = 14). Robust SE’s: * p<0.10, **
39
p<0.05, *** p<0.01. See Web Data Appendix for data sources.
TABLE 7: BLACK DEATH AND CITY GROWTH, HETEROGENEOUS EFFECTS, 1300-1600
Percentage Change in City Pop. (%) in Period t
Dependent Variable:
Col.(1)-(3): Existing Rates
Col.(4)-(6): Extrapolated Rates (140)
Mortality Rates:
1300-1600
1300-1700
1300-1750
1300-1600
1300-1700
1300-1750
(1)
(2)
(3)
(4)
(5)
(6)
Period:
Effect of Black Death Mortality Rate (%) x Dummy:
Separate regression in each row:
40
≥ Median Av. Temp. 1500-1600
≥ Median Elevation
≥ Median Cereal Suit. 25 Km
≥ Median Grazing Suit. 25 Km
1(Coast 10 Km)
1(Mediterranean Coast 10 Km)
1(Atlantic Coast 10 Km)
1(North-Baltic Coast 10 Km)
1(Rivers 10 Km)
≥ Median Longitude (Western)
≥ Median Latitude (Northern)
1(Northern Europe)
≥ Median Log City Pop. in 1300
1(Maj.RomanRd (MRR) 10 Km)
1(MRR Intersection 10 Km)
1(Any Roman Rd (ARR) 10 Km)
1(ARR Intersection 10 Km)
1(Medieval Route (MR) 10 Km)
1(MR Intersection 10 Km)
1(Market & Fair)
1(Hanseatic League)
1(University)
1(Monarchy in 1300)
1(State Capital in 1300)
1(Autonomous City in 1300)
1(Parliamentary Act. 13-1400)
1(Battle 100 Km 1300-50)
Number of observations
-0.50
-1.78*
0.23
-1.86
1.52
1.63*
9.69
5.16
-0.46
-1.41
0.62
1.45
0.82
-0.55
0.01
-0.30
-0.22
-0.22
-1.20
0.61
4.41**
0.09
1.01
0.24
1.91**
0.11
0.55
[1.28]
[1.00]
[1.09]
[2.29]
[1.06]
[0.87]
[9.10]
[3.55]
[1.24]
[1.06]
[1.35]
[1.50]
[1.33]
[0.99]
[0.96]
[1.16]
[0.95]
[1.02]
[1.20]
[1.06]
[2.14]
[1.60]
[1.00]
[1.27]
[0.90]
[1.05]
[1.13]
-4.81*
-1.73
-0.18
1.35
2.28
1.47
20.85*
15.97**
0.34
-1.18
3.65
5.65
0.94
-2.12
-0.05
-1.98
-0.49
1.27
-1.07
1.28
6.99
-1.89
2.42
2.18
0.93
0.42
1.67
[2.86]
[1.67]
[1.80]
[3.34]
[2.04]
[1.38]
[11.5]
[6.84]
[2.49]
[2.08]
[2.98]
[3.86]
[2.68]
[1.56]
[1.83]
[1.62]
[1.62]
[1.80]
[1.58]
[2.24]
[4.22]
[2.30]
[1.61]
[2.77]
[1.48]
[1.69]
[2.59]
Col.(1)-(3): 140
-7.71**
-2.22
-0.67
3.06
2.28
0.05
20.06
27.80***
0.66
-1.53
6.40*
9.85**
1.07
-3.25
0.02
-3.69
-0.67
1.48
-0.77
2.67
9.86*
-3.24
2.27
2.74
1.07
0.14
3.02
[3.46]
[2.04]
[2.15]
[5.07]
[2.45]
[1.64]
[18.2]
[9.66]
[3.13]
[2.45]
[3.59]
[4.70]
[3.35]
[2.01]
[2.28]
[2.31]
[2.01]
[2.19]
[2.11]
[2.92]
[5.82]
[2.43]
[1.94]
[3.59]
[1.81]
[2.02]
[3.23]
1.26
-1.11
0.56
-1.39
-0.20
3.29***
-1.24
-2.69
-0.60
0.56
-0.56
-0.32
2.82*
0.15
0.28
0.14
1.41
-1.04
-1.70
0.11
3.32*
2.33
2.59***
1.70
0.39
0.89
-0.40
[1.12]
[0.97]
[1.02]
[2.09]
[1.54]
[1.12]
[11.8]
[5.38]
[1.11]
[0.96]
[1.22]
[1.33]
[1.50]
[0.93]
[0.91]
[1.01]
[1.01]
[0.98]
[1.18]
[0.97]
[1.91]
[2.33]
[0.91]
[1.67]
[0.86]
[1.06]
[1.05]
-1.30
-0.26
-1.30
1.86
-1.89
3.44**
19.11*
-6.16
0.58
1.49
2.12
2.71
6.57*
-0.61
0.48
-0.48
1.78
1.08
-1.23
0.93
6.10
0.39
3.38**
3.34
-0.46
1.39
-0.14
[2.39]
[1.35]
[1.65]
[2.74]
[3.19]
[1.44]
[11.0]
[14.9]
[2.01]
[1.63]
[2.54]
[3.49]
[3.70]
[1.44]
[1.46]
[1.64]
[1.47]
[1.64]
[1.48]
[1.94]
[3.90]
[2.79]
[1.42]
[2.44]
[1.39]
[1.51]
[2.11]
Col.(4)-(6): 460
Notes: This table shows the interacted effects of the Black Death mortality rates in 1347-1351 and various city characteristic dummies proxying
for locational fundamentals, increasing returns and institutions. Each row represents a separate regression for each city characteristic dummy.
XXX TO BE COMPLETED XXX Robust SE’s: * p<0.10, ** p<0.05, *** p<0.01. See Web Appendix for data sources.
-3.79
-0.79
-1.71
3.58
-1.39
2.82*
22.34
3.72
0.69
1.01
5.35*
6.91*
8.15**
-1.11
0.57
-1.46
1.75
1.25
-1.03
1.85
8.78
-0.50
3.58**
4.01
-0.41
1.49
1.56
[2.80]
[1.59]
[1.88]
[4.13]
[3.40]
[1.67]
[15.6]
[15.9]
[2.47]
[1.86]
[3.04]
[4.12]
[4.01]
[1.71]
[1.74]
[2.03]
[1.70]
[1.98]
[1.91]
[2.48]
[5.41]
[3.23]
[1.63]
[3.01]
[1.54]
[1.72]
[2.51]
TABLE 8: BLACK DEATH MORTALITY RATES AND CITY GROWTH, HORSEMEN EF-
FECTS
(1) t = 1400
Regression:
(2) t = 1600
(3) t = 1700
1. Pct. Change in Pop. in 1300-t, Cities in 1300 (Obs. = 140)
2. Row 1, Extrapolated Mortality for 140 Cities (Obs. = 467)
-0.84***
-0.73***
[0.31]
[0.22]
3. Dummy if ≥ 1,000 in t, Cities Post-1300 (Obs. = 1,322)
4. Row 4, Adding Country FE (N = 17)
-0.002**
0.00200*
[0.00] -0.000
[0.00] -0.002** [0.00]
[0.00] 0.00474***[0.00] 0.00437** [0.00]
5. Row 4, Adding Controls in 1300 and Country FE (N = 17)
6. Log Pop. in t, Cities Post-1300 (Obs. = 105-894)
7. Row 6, Adding Country FE (N = 17)
8. Row 6, Adding Controls in 1300 and Country FE (N = 17)
-0.13
0.30
[0.51]
[0.50]
0.09
0.04
[0.85]
[0.75]
0.001
[0.00] 0.006**
[0.00]
-0.008
-0.0119
[0.01] 0.007***
[0.01] 0.00444
[0.00] 0.006*** [0.00]
[0.00] 0.00717** [0.00]
0.021
[0.02] 0.013***
[0.00] 0.012***
0.003
[0.00]
[0.00]
Notes: This table shows the effect β t of the Black Death mortality rate (%) in 1347-1352 on the percentage change in city population (%) for various
periods t. Rows 1-2 show the effect of the percentage change in pop. on mortality for the main sample of 140 cities and the sample of 467 when
using the extrapolated mortality rates based on 140 cities respectively. In rows 3-8, we focus on the sample of 1,322 cities that did not already exist
in 1300 but were cities at one point in the Bairoch sample (1300-1850). Rows 3-5 regress a dummy equal to one if the city emerges in year t on
mortality. Rows 6-8 regress log pop. in year t on mortality. See Web Data Appendix for data sources.
TABLE 9: BLACK DEATH AND CITY GROWTH, CELL-LEVEL ANALYSIS, 1300-1600
Dep.Var.:
Percentage Change
for Existing Cities
in 1300
Absolute Change
for Existing Cities in
1300
Absolute Change
for Non-Existing
Cities in 1300
Absolute Change
for All Cities
Cells:
with a City in 1300
with a City in 1300
with a City in
1300-1600
1300-1400
1300-1600
Observations
(1)
(2)
with a City in
1300-1600 but
no City in 1300
(3)
-0.90***
(0.30)
0.56
(0.78)
-0.06**
(0.03)
0.13
(0.15)
-0.00
(0.03)
0.24**
(0.11)
-0.24***
(0.08)
0.28
(0.25)
243
243
101
344
(4)
Notes: This table shows the effects for various periods (1300-1400; 1300-1600) of the average Black Death mortality rate (%) of each 1x1 degree
cell (using the spatially extrapolated mortality rates based on 139 cities) in 1347-1351 on: (1) the percentage change in city population (%) for
the existing cities in 1300 for the 243 cells with a city in 1300; (2) the absolute change in city population (inh.) for the non-existing cities in 1300
for the 243 cells with a city in 1300; (3) the absolute change in city population (inh.) for the non-existing cities in 1300 for the 101 cells with a
city at one point in 1300-1600 but with no city in 1300; and (4) the absolute change in city population (inh.) for all (existing and non-existing)
cities in 1300 for the 343 cells with a city at one point in 1300-1600. Robust SE’s: * p<0.10, ** p<0.05, *** p<0.01. See Web Appendix for data
sources.
41
Web Appendix (not for publication)
City Population Estimates
Our main source of urban population data is the Bairoch (1988) dataset of city populations. The
Bairoch dataset reports estimates for 1797 cities between 800 and 1850. We use 1792 of these cities
and 5 cities in northern Norway and Finland cannot be matched to the map that we employ to create
our geographical controls. The criterion for inclusion in the Bairoch dataset is a city population greater
than 1,000 inhabitants.
This dataset has been widely used by a range of scholars studying premodern urbanization and
economic development. We follow Bosker et al. (2013) and Voigtländer & Voth (2013b) in updating
the Bairoch dataset where a consensus of historians have provided revised estimates of the population
of a particular city, including Bruges, Paris, and London. Indeed, while the Bairoch data set is our
main source of information, Chandler (1974, 1987) is more specific in the sources used to measure
city population. This way, we can better assess the “true” population of each city in each year. For
example we prefer Chandler’s population estimates for a range of cities including Granada, Paris,
Venice, Genoa, and Milan. In our analysis we use this corrected dataset as our benchmark. We also
employ the original Bairoch dataset and the Chandler dataset in our robustness exercises.
The Baiorch dataset contains cities from the following countries that existed in 1988: Germany, Austria,
Belgium, Bulgaria, Denmark, Spain, Finland, France, United Kingdom, Greece, Hungary, Ireland,
Italy, Luxembourg, Malta, Norway, The Netherlands, Poland, Portugal, Romania, Russia, Sweden,
Switzerland, Czechoslovakia, Albania, and Yugoslavia. We drop all cities in the following Eastern
European countries as we do not possess any Black Death mortality data for them: Albania, Bulgaria,
Finland, Greece, Hungary, Romania, Russia, Malta, and Yugoslavia. We assign cities in “Czechoslovakia”
to either the Czech Republic or Slovakia based on their modern borders.
The Bairoch dataset reports city populations every century. Therefore we have population estimates
for our entire sample both for 1300 and for 1400. This provides an important benchmark for our
analysis but is far from ideal as population growth or shrinkage may have taken place between 1300
and 1348 and population recovery may have occurred between 1350 and 1400. We employ several
approaches to overcome this problem.
We collected data on estimates of pre plague population from a range of historical sources. Chandler
(1974, 1987) provides alternative estimates on city population growth including estimates for city
sizes in 1320 and 1360. Christakos et al. (2005) summarizes a wide range of historical estimates of
pre-plague populations and mortality rates. We used his data and returned to a several of the sources
in the secondary literature to provide checks on his estimates of pre-plague population including
Ziegler (1969), Russell (1972), Pounds (1973), Gottfried (1983), Nicholas (1997) and Benedictow
(2005).
Plague Mortality Estimates
We base our urban mortality data on the estimates collected by Christakos et al. (2005) which come
from a wide range of historical sources. We supplement these where possible with data from other
sources including Ziegler (1969), Gottfried (1983), and Benedictow (2005). These sources provide
an estimated mortality rate for either a city such as Florence which had an estimated mortality rate of
60% or for a region such as Navarre which had morality rate of between 60-65% (Benedictow, 2005,
42
275). This data yields estimates of mortality for 185 cities. However, not all of these cities are in the
Bairoch database in the year 1300. Therefore we have 139 cities for which we have an estimate of
plague mortality and an estimate of their population size in 1300.
There is considerable variation in mortality rates. In Milan the morality rate was just 15%. It
was also estimated to be only 10% in Alsace, Lorraine, and Bohemia (Gottfried, 1983). However,
Cambridgeshire had an estimated mortality rate of between 53-70%.
The data reported by Christakos et al. (2005) are based on four types of information. Approximately
64% of our sample are based on numerical estimates of the numbers killed during the plague. Christakos et al. (2005) systematically analyze each datapoint, modifying widely inaccurate numbers on
the basis of other sources and pieces of information.
The second source of information are descriptions of the impact of the plague. Chroniclers used
descriptions such as “completely depopulated” or “close to being depopulated” in which case we code
plague mortality as equal to 80%. Similarly if a city is described as “partially spared’ or experiencing
‘minimal’ mortality we assign an estimate of 10% mortality. This data correspond to approximately
18% of our main sample.
The third source of information comes from estimates of desertions. Christakos et al. (2005, 143)
show that there is a linear relationship between mortality rates and desertion rates and that the on
average desertion rates were around 20% higher than mortality rates. Following this discussion in our
baseline analysis, we reconstruct mortality rates from desertion rates by using a ratio of 1.2.
The fourth source of data are information of mortality rates among the clergy. The clergy keep excellent
records during the Black Death. However, death rates among the clergy may have systematically
differed from those among the general population. For example, it is possible that death rates among
the clergy may have been disproportionately high as they had a duty of care to the sick. On the other
hand, the clergy may have been better fed on average than the general population in which case
their death rates may have been lower. Christakos et al. (2005) argue that clergy mortality rates are
unlikely to have been much greater than those of the general population. We assume a relationship of
1.08 between morality rates for the clergy and those of the ordinary population, implying that the
clergy experienced an 8% higher death rate. As there are only 5 cities for which we rely on clergy
mortality estimates, this does not affect any of our results.
There are several sources of potential bias in this data The first source of potential basis is reliance on
medieval chroniclers who form the basis of many of the estimates in the Christakos et al. (2005) data.
Chroniclers were notoriously innumerate. The numbers they reported for medieval battles are known
to exaggerate, sometimes by an order of magnitude. We know that some of the figures recorded
by chroniclers are also exaggerations. For example, Agnolo di Tura thought that 52,000 individuals
died in Siena but historians regard this as implausible as the population of Siena was no higher than
60,000 (Gottfried, 1983, 45). Indeed historians of the late nineteenth and early twentieth century
were inclined regard the estimates of death tolls during the Black Death as uniformly exaggerated.
Modern research, however, has largely confirmed the estimates of medieval chroniclers. Therefore,
modern historians tend to regard chroniclers estimates as highly informative. They are no doubt
subject to error but this is classical measurement error and there is no reason to suspect any source of
systematic bias.
A second source of potential bias is heaping. It is well known that in societies with low numeracy
43
individuals tend to report their age to the nearest five or ten digits (Baten et al., 2013, e.g.). Similarly,
there is a tendency of contemporary observes to choose round numbers such as one-half or one-quarter
when reporting Black Death mortality. We control for this source of bias in our regression analysis.
A third source of possible bias is that many of our estimates are based on tax data. Tax records are
used in additional to and in instead of chroniclers estimates because there was much greater incentive
to be accurate. However, tax records only record households wealthy enough to be assed for taxation.
The poor were not counted or assessed. Furthermore, as Benedictow notes ‘[a]nother problem arises
from the fact that the proportion of the poor and destitute that were unable to pay taxes or rents was
dramatically reduced in the wake of the Black Death. In the small city of Albi in southern France, the
recorded population of the population that was too poor to pay taxes diminished from 43 per cent in
1343 to 28 per cent in 1357’ (Benedictow, 2005, 263). This places a source of downwards bias on
estimates of plague death especially as there is evidence for supermortality of the poor during the
Black Death.
A fourth source of potential bias occurs when the underlying estimates of plague mortality come from
data on the death of the clergy. It is not clear what the direction of this bias is. On the one hand,
the clergy may have been somewhat better educated and better fed than the rest of the population,
leading to a lower mortality rate. On the other hand, the duty of the clergy lay in looking award the
sick and this exposed them to infection and may having given rise to a higher than average mortality
rate.
Rural morality rates were at least as high as urban mortality rates. Benedictow (2005) provides a
range of estimates for some but not all parts of Europe. Christakos et al. (2005) also report scattered
estimates of rural mortality.
Extrapolated Mortality Data
In order to extend our analysis to cities for which we don’t have explicit mortality rates, we use spatial
analysis to impute the missing values. Our assumptions in doing this are that (1) there exist some
underlying causes of mortality rates which are unobserved, (2) these causes have a large random
component (i.e. are external to our model of subsequent city growth), (3) these causes are also
spatially correlated. For example, it is widely acknowledged that fleas living off of rat populations were
a primary vector for the plague. It is highly plausible that a latent variable measuring the suitability
of a city’s surrounding region for sustaining large rat populations satisfies the three criteria laid out
above.
In order to impute the missing mortality rates we create a two-dimensional surface of predicted plague
mortality using an inverse distance weighted function of known mortality rates. For every point on
the surface a predicted mortality rate is generated using the closest 15 cities within an approximately
1,000 km radius circle around the point.27 For a point on the surface, x, with unknown mortality
the influence of city, i, with known mortality diminishes with its distance from x according to the
weights used. These weights are determined by a parameter, p ≥ 0, referred to as power. As the
power decreases, the influence of more distant points increases. If p = 0, then all points receive equal
weight in determining all other points on the map. The influence of more distant points decreases
exponentially as p increases.
27
If there are fewer than 15 cities in this radius, the a minimum of 10 are used. Our predictions are robust to
experimenting with different values of the maximum and minimum number of cities allowed.
44
a. 139 City Sample
b. 185 City Sample
FIGURE 9: Predicted vs. measured mortality rates at the optimal power. The 139 city sample
consists of cities with reported mortality rates that existed in 1300. The 185 city sample consists of
all cities with reported Black Death mortality rates.
To create our mortality estimates we choose an optimal p using cross-validation techniques. The
procedure begins by choosing some power, p̄. Then, using the sample of n cities with known mortality
rates, we create a predicted mortality rate surface using all of the cities except for city j. We then
predict the mortality rate for city j as m̂ j using our mortality surface and create it’s residual as (m̂ j − m j )
where m j is the known mortality rate for city j. This procedure is then repeated to create predicted
mortality rates and residuals for all the remaining cities.
r We then calculate the Root Mean Square
Pn
(m̂ −m )2
i
i
i=1
Error (RMSE) of the residuals created as RMSE(p̄) =
, where i indexes the cities with
n
known mortality rates. This procedure is repeated for a large number of choices of p and then the
optimal power is chosen as the one which minimizes the RMSE.
We generate optimal mortality surfaces using several different city samples. Our baseline sample
consists of cities that existed in 1300 for which mortality rates are reported in the historical literature.
There are 139 of these cities and the cross-validation exercise chooses an optimal power for creating
the mortality surface as 1.02 (RMSE=14.25). Panel A of Figure 9 shows the relationship between the
measured and predicted mortality rates for this 139 city sample. Figure 3 shows the mortality surface
generated using the 139 city sample along with the location of all cities (not just those with mortality
rates) for which we have population data in 1300.
We also generate an optimal mortality surface using the sample of 185 cities for which we have an
estimate of Black Death mortality. The most likely reason the extra 46 cities included in this sample
may not have been recorded in Bairoch for 1300 is because their populations were too small (less
than 5,000) in the 13th century. Nonetheless, they provide valuable information on mortality rates.
45
The cross-validation analysis suggests an optimal power of 1 for these cities, which yields a RMSE of
14.10. Panel B of Figure 9 shows the measured versus predicted values using the optimal power.
Some of the mortality rates for the 139 and 185 city samples are reported with greater accuracy than
others. For example, often, the rate is given as ‘about half’ or ‘at least half’ of the population, which
we then code as 50%. This can be seen in the Figure 9 as the heaping around 0.50. In other cases,
the mortality rates are reported as for ‘clergy’ as opposed to the general population. As a robustness
check on our mortality predictions, we generate subsamples of the 139 and 185 city samples in
which we drop any vague or ambiguous mortality reports. This generates a sample of 89 cities that
existed in 1300 for which we have mortality rates and another sample of 111 cities consisting of all
unambiguously reported death rates. For the 89 city sample, the optimal power used is 1.13 and this
yields a RMSE of 13.89. For the 111 city sample, the optimal power is 1.35 and this yields a RMSE of
13.51. It is reassuring that largest sample consisting of precisely reported death rates also yields the
most accurate predictions from its mortality surface.
Control Variables
We employ a wide range of geographic, economic and institutional city-level controls variables. We
categorize our controls into three categories: physical geography controls; economic geography
controls’ and institutional controls.
Physical Geographical Controls
Distance to the Coast and Major Rivers
We create a variable to measure distance to the coast and major rivers in meters using ArcGIS. We
base these distances on the 1300 shape file downloaded from Nussli (2011). All maps are projected
into World Mercator WGS 1984.
Soil Suitability
Our soil suitability data are from the FAO Global Agro-Ecological Zones (GAEZ) dataset as described
in Fischer et al. (2002). We use these in preference to the Ramankutty et al. (2002) as the latter does
not have full coverage for all of western Europe (it omits Sicily for example). We use the GAEZ’s
overall cereal suitability data assuming low inputs and rain-fed irrigation. We extract the average soil
suitability within 25, 50, and 100 kilometer radius circles around each city. Overall cereal suitability is
scaled from 1-9 where 1 is best, 8 is unsuitable and 9 is water (seas and oceans are treated as missing
values).
Elevation
City elevation data come from Jarvis et al. (2008) which is available at
http://srtm.csi.cgiar.org
This data reports elevation in meters. The spatial resolution between 1 and 3 arc-seconds. Where
there is missing data we have supplemented it using wikipedia.
46
Grazing Suitability
In the literature on long-run economic growth in Europe emphasizes the importance of pastoral
farming. Broadberry (2013) emphasizes that regions that adopted pastoral farming engaged in higher
value-added production that was also more capital intensive. Voigtländer & Voth (2013a) argue that
there was an interaction between the Black Death and pastoral farming which led to lower fertility
rates and increased human capital accumulation. We control for the potential suitability of a region
surrounding a city for pastoral farming with a variable measuring grazing suitability. This variable
come from Erb et al. (2007) who create land use measures at a resolution of 5 arc minute cells, or
approximately 10 km X 10 km at the equator. The record how land is used in each cell in 2000. The
five categories they code for are: cropland, grazing, forestry, urban, and areas without land use. Their
grazing category is calculated as a residual after accounting for the percentage of area taken up by
the other four uses. As part of this analysis they also generate a variable measuring the suitability of
each cell for grazing (as opposed to actual present-day use). The suitability measure is created by
first separating grazing land into three categories based on cover: ‘high suitability of cultivated and
managed areas, medium suitability of grazing land found under tree cover, and low suitability if shrub
cover or sparse vegetation is detected in remote sensing’ (Erb et al., 2007, 199). The then further
subdivide the first two of these categories into areas with a net primary productivity of Carbon per
meter squared is greater than 200 grams and those in which it is less than 200 grams. This results in
five categories which they regroup into four categories with 1 = most suitable and 4 = least suitable.
There is a fifth category which is ‘no grazing’ which we re-code as 5. We then extract the average
suitability of the region around a city for grazing using circles of 25, 50, and 100 km’s.
Temperature
We use temperature data from Luterbacher et al. (2004). They reconstruct seasonal European
temperatures since 1500 using proxy data from ice cores, tree rings, and written records. The data
cover 0.5◦ X 0.5◦ grids which is approximately 50km X 50 km at European latitudes. The data extend
from 25◦ W to 40◦ E and 35◦ N to 70◦ N which includes all of the cities in our extended Bairoch sample.
We extract the growing season (summer) temperature for each of our cities during the 16th century
as this is the closest century to the Black Death period for which we have data.
Economic Geographical Controls
Roman Romans
Data on Roman roads is provided by the Digital Atlas of Roman and Medieval Civilizations. It is available
from:
http://darmc.harvard.edu/icb/icb.do?keyword=k40248pageid=icb.page601659
We use this shape file to create two distances: (1) distance to all Roman roads and (20 distance to
‘major’ Roman roads. Since major settlements often formed along the intersection of the road network.
we also create a variable for distance to Roman road intersection using ArcGIS.
47
Medieval Universities
Bosker et al. (2013) provides data on the presence of medieval universities for European cities with
populations greater than 10,000 (at some point between 800 and 1800. We consulted Wikipedia and
other sources to find evidence of medieval universities in European cities with smaller populations.
There are five medieval universities missing from the list in Bosker et al. (2013) Angers, Greifswald,
Ingolstadt, Tuebingen, and Uppsala. However, as none of these universities were established prior to
the Black Death do not include them in our analysis.
Medieval Fairs
Medieval fairs were important locations of economic activity in the medieval period. Fairs provided
central meeting points for trade to take place—enabling a greater degree of economic specialization
than would otherwise have been possible (Spufford, 2002). The major fairs were consequently
economic centers, particularly in the period before the Black Death.28 Proximity to a major fair might
therefore give a city an economic advantage. As such we control for whether or not a city contained
an important fair or its proximity to a fair.
We obtain data on the location of important medieval fairs from two sources. The first source is
Shepherd (1923). The second source we use is the Digital Atlas of Roman and Medieval Civilizations:
http://darmc.harvard.edu/icb/icb.do?keyword=k40248&pageid=icb.page188865
The original source for this information is: Ditchburn, David and MacLean, Simon (eds.) 2007, Atlas
of Medieval Europe, 2nd edn, London and New York, p. 158.
We drop the following fairs as they cannot be matched with cities in the Bairoch dataset: Stamford, St
Ives, Bergen op Zoom, Mesen, Bar-sur-Eabue and Lagny.
Trade Routes
We use Shepherd (1923) to obtain the path of major medieval land trade routes. We use ArcGIS
to create a shape file that allows us to measure distance to major medieval land trade route or the
intersection of two major trade routes. Figure 10 depicts this data. We create a dummy variable that
takes the value of 1 if a city is within 10 kilometers of trade route or intersection.
Institutional Controls
Bosker et al. (2013) provides data on the presence of bishoprics, parliaments and participatory urban
governments for European cities with populations greater than 10,000 (at some point between 800
and 1800. This gives us a rich set of geographic and institutional control variables for 677 of the 1797
cities in our sample. Full details on these controls are provided in the appendix to Bosker et al. (2013).
We briefly summarize some of the most important controls here.
Bosker et al. (2013) collect data on capital cities from McEvedy & Jones (1978); information on
universities from the ninth edition of the Encyclopedia Britannica; data on communes from Lexikon
des Mittelalters in addition to other historical sources; data on parliamentary activity from Zanden
28
Fairs have been studied by North et al. (2009) and Edwards & Ogilvie (2012).
48
FIGURE 10: Major Medieval Trade Routes
!
!
!
!
!
!
!
!
!
!
!
! !
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!
! !
!
!
!
! !
!
!
!
!
!
!
!
!!
! !
!
!
! !! !
!
! !
!
!
!!
!
!
! !
!! !!
! ! !!!
!
!!! !!!
!
!! ! ! !
!
!
!
!
!
!
!
! ! !!
! !
!
!
!
!!
!
!
!
!
!! !
!
!
!!
! !
!
!
!
!
!
!
!
!
!
! !
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
! !
!
! !
!
! !
! !
!
!
! !
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
! !
!
!
!
!
!
!
!
!
!
!
! !
!
!
!
! !
!
!
!
! ! ! !! ! ! ! !
! !
!
!
! !! !
!
! !
!
!!
!
!
!!
!
!!
!
!
!
!
!!
!
!
!
! !
!
! !
!!
!!
!!
!
!
!
! ! ! !
!
!
!
!
! !! !
!
!
!
!
!
! !
!
!
! ! !
!!
!
!
!
! !
!
!
!!
! !
!
!
!
!
!
!
!
! !
!
!
!!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!!
!
!
! !
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
! !
!
!
Medieval Trade Routes
!
All Cities Existing in 1300
!
!
!
!
!
!
!
!
!
!
! !
!
!
!
!!
!
!
!
! !
!
!
! !
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
! !!
! !!!!!!! ! ! !!
!!!!!!!! !!
!!
!
!!!
! ! !!
! !!!
!
! ! !
! !!
!
!
!
!
!
!
Source: Shepherd (1923)
et al. (2012).
Hanseatic League
We document whether or not a city was a member of the Hanseatic League. We do this by matching
where possible the Bairoch city data with available lists of cities which belonged to the League. We
include only cities which were members of the League and do not included cities with Hansa trading
posts or Hansa communities. Our main source is Dollinger (1970).
We are able to match 59 cities from Dollinger to Bairoch cities. The locations we did not match are
Alhen, Alfeld, Allendforf, Altena, Arnsberg, Attendorn, Balve, Belecke, Blankenstien, Boholt, Bockenem,
Bodefeld, Borgentreich, Borken, Brakel, Brackerfeld, Brilon, Buxtehude, Cosefeld, Dorsten, Doseborg,
Drolshagen, Duderstadt, Dulmen, Elburg, Eversberg, Freiohl, Frustenau, Gardelegen, Geseke, Gottingen, Grevenstein, Grieth, Gronau, Hachen, Hagen, Haltern, Hattem, Hirschberg, Horde, Husten,
Iburg, Kallenhardt, Kamen, Langenscheid, Legmo, Ludenscheid, Lunen, Melle, Menden, Muhlhuasen,
Nehim, Nejmegen, Neuenrade, Neustadt, Nieheim, Northeim, Oldenzzal, Olpe, Ommen, Osterburg,
Osterode, Peckelsheim, Plettenerg, Quakenbruck, Ratingen, Recklinghuasen, Ruthen, Schwerte, Seehuasen, Stavoren, Sundern, Tangermunde, Telgte, Ulzen, Uslar, Vorden, Vreden, Warendorf, Warstein,
Wattenscheid, Werben, Werl, Werne, Westohfen, Wetter, Wiedenbruck, Zaltbommel.
For an alternative list of cities with Hansa affliations, we also consult the list in:
https://en.wikipedia.org/wiki/Hanseatic_League
We are able to match the majority of these Hansa cities. The exceptions are as follows. We do not match
Demmin, Darlowo, Falsterbo Muster, Pasewalk as they not in our dataset as it they were presumably
49
too small during the middle ages to make it into the Bairoch dataset. As the Bairoch dataset did not
include cities in the then Soviet Union, we do not include the following cities which are in modern
Latvia, Estonia, Belorussia, or Russia: Riga, Reval, Doporat, Kanunaas, Köingsberg, Novgorod, Pleskau,
and Polotsk.
Political Boundaries in 1300
The shape files provided by Nussli (2011) report political boundaries in Europe for every century. We
use the shapefile in Nussli (2011) to obtain political boundaries for Europe in 1300. We then assign
each city in the Bairoch dataset to its political boundary in 1300.
Monarchy and Republic Variables
We further code whether or not a city had a Republican government. Our definition of a republic
is a narrow one. It does not encompass cities that had some measure of self-government but owed
ultimate fealty to a higher sovereign (such as London or many of the free imperial cities of the Holy
Roman Empire). It includes the Republic of Florence, the Republic of Lucca, the Republic of Siena, the
Republic of Pisa, the Republic of Venice.
Finally, we code a city as belonging to a monarchy in 1300 if it it belonged to the Kingdom of Bohemia,
the Kingdom of Denmark, the Crown of Castile, the Kingdom of France, the Kingdom of Norway,
the Kingdom of England, the Kingdom of Sicily in Naples, the Kingdom of Granada, the Kingdom of
Scotland, the Kingdom of Hungary, the Kingdom of Sicily, the Kingdom of Galicia-Volhynia, the Crown
of Aragon, the Kingdom of Portugal, the Kingdom of Majorca, the Kingdom of Sweden.
Nussli (2011) does not provide separate political boundaries for all of the small states that comprised
the Holy Roman Empire coding themselves instead as ‘small states of the Holy Roman Empire’. We
code these as a small states in our main analysis. For robustness we also code all cities within the
borders of the Holy Roman Empire as belonging to a large state.
Commune and Autonomous Cites
Bosker et al. (2013) provide information on the existence of communes for a subset of the cities in
the Bairoch dataset. Bosker et al. (2013) create a variable“commune” that takes a value of 1 if there
is indication of the presence of a local urban participative organization that decided on local urban
affairs.
Stasavage (2014) provided us with data on 169 cities that where autonomous at some point between
1000 and 1800. We utilize the variable for 1300-1400. There are xx autonomous cities in our dataset.
Stasavage (2014) defines autonomous cities in the following terms:
‘I have defined an “autonomous city” as being one in which there is clear evidence that
such institutions of self-governance existed, and in addition there is also clear evidence of
exercise of prerogatives in at least one of the policy areas referred to above. In the absence
of such evidence the default is to code a city as non-autonomous (6).
As Stasavage (2014) notes, his definition of city autonomy is stricter than the definition of commune
used by Bosker et al. (2013).
50
Warfare data
Warfare was a leading contributor to the fourteenth century crisis and therefore affected population
trends during the period of analysis (Voigtländer & Voth, 2013b, see). As it is impossible to obtain
precise numbers on excess mortality due to warfare for the medieval period, we follow a recent
scholarship in collecting data on the location of conflicts. This allows to create a control variable
which reflects variation in the intensity of warfare over time and space.
As our main source we use Wikipedia’s list of all battles that took place between 1300 and 1600.
https://en.wikipedia.org/wiki/List_of_battles_1301-1800
This is a highly reliable source for the most important battles of the period. We are not concerned
about sample selection here as Wikipedia’s coverage of European history is extensive; battles not listed
on Wikipedia are likely to have been extremely small.
For each battle we assign a geo-coordinate based on either the location of the battle or the location of
the nearest town or city mentioned in the entry. We exclude naval battles and conflicts which cannot
be located (such battles were typically extremely minor).
Since not all battles were the same, we also collect data on the intensive margin of conflict by recording
information on the total number of participants in each battle. We use this metric for several reasons.
The most important is it captures the size of forces involved in a battle which is a good proxy for the
amount of devastation in terms of disruption of food supplies and disease that warfare imposed on
the nearby population. Additionally, the size of the armies involved in battle is more widely available
and recorded with less noise than are estimates of battlefield casualties.
These estimates are not ideal. The location of a battle may not account for the all of the devastation
that armies might wreck on the countryside of a city. Nevertheless, given data limitations it is the best
way to account for the affects of major conflicts on city populations in the middle ages.
51
FIGURE 11: The Severity of the Black Death by Month: Testaments and Vacant Benefices
This figure plots data for testaments from Siena, Perugia, Rome, Areezzo and Florence for the year
1348 and vacant benefices in Barcelona. Both measures are proxies for overall mortality. They
indicate that mortality peaked in the summer months during the outbreak of 1348 in Italy. Source
Cohn03|.
52