Jewish Communities and City Growth
in Preindustrial Europe∗
Noel D. Johnson§ and Mark Koyama‡
George Mason University
February 26, 2016
Abstract
We study whether cities with Jewish communities grew faster than cities without
Jewish communities in Europe between 1100 and 1850. We match data on city
populations from Bairoch (1988) with data on the presence of a Jewish community
from the Encyclopedia Judacia. Our OLS results indicate that cities with Jewish
communities grew between 5% and 10% faster than comparable cities without Jewish
communities, but that this effect does not emerge until after 1600. To establish
causality, we create time varying instrumental variables which rely only on the
spatially extended network of Jewish communities in order to predict Jewish presence
in a given city. We also provide evidence that the Jewish city growth advantage
stemmed in part from their ability to exploit increases in market access after 1600.
Keywords: Long-Run Growth; Urbanization; Market Access; Jewish Communities; Toleration;
Religion; Little Divergence
JEL Codes: J1; N00; O1; R1; Z12
∗
We’d like to thank Eric Chaney, Larry Iannaccone, Garett Jones, Remi Jedwab, Andrea Matranga, Nathan
Nunn, Santiago Perez, Maria Petrova, Jared Rubin, Mohamed Saleh, and Melanie Meng Xue for giving us comments
and audience members at the Fall 2015 Washington Area Development and Economic History Workshop, the
Role of History in Understanding Development Conference, NES (Moscow), the 2015 Social Science History
Association Meeting (Baltimore), and Chapman University (2016). All remaining mistakes are the fault of the
Authors. We gratefully acknowledge the support of the Mercatus Center. We thank Jordan Bazak, Michael
Spizindor Watson, Megan Teague, Trey Dudley, Zhilong Ge, and Jessi Troyan for research assistance.
§
Email: noeldjohnson@mac.com. Center for Study of Public Choice, Carow Hall, MSN 1D3, 4400 University
Drive, Fairfax, VA 22030.
‡
Email: mark.koyama@gmail.com. Center for Study of Public Choice, Carow Hall, MSN 1D3, 4400 University
Drive, Fairfax, VA 22030.
1
Introduction
Numerous scholars have speculated on the role played by economically productive minority
groups, such as Protestants, Jews, and Quakers, in driving economic growth (Weber, 1968;
Sombart, 1911; Braudel, 1982; McCabe et al., 2005). Recent research has established that Jews
in Europe had high levels of human capital from medieval times onwards (Botticini and Eckstein,
2012) and played a crucial economic role in pre-World War II Eastern Europe (Acemoglu et al.,
2011). Nevertheless, despite the important role Jews played in the European economy from
the middle ages to the modern period, there has been little research exploring whether there
was a systematic link between Jewish communities and economic growth. This question is even
more important to address given our poor understanding of the links between religious tolerance,
the creation of market institutions, and perhaps the most influential economic development in
modern times—the onset of the industrial revolution.
This paper offers insight into these questions by making three contributions to our knowledge
of the role played by Jewish communities in urban development in Europe. First we combine
data from the Encyclopedia Judaica (Roth and Wigoder, 2007) and Bairoch (1988) to create
a data set of European cities with Jewish populations and urban growth between 1100 and
1850. After matching Jewish communities with Bairoch cities, we show that the presence of
a Jewish community in a premodern European city was associated with between 5% and 10%
faster growth than experienced by a comparable non-Jewish city. This finding is consistent with
the hypothesis that minorities with higher levels of human capital and access to commercial
trading networks led to faster growth.
Despite the strong theoretical and historical reasons for thinking that the presence of Jews
could be a driver of urban development, it is also highly plausible that Jewish presence might
be correlated with factors that affect growth through other channels. As such, our second
contribution is to establish a causal relationship linking Jewish presence to faster city growth. We
do this by explicitly modeling the spatial network of Jewish communities over time using tools
drawn from the market access literature (e.g. Donaldson and Hornbeck, 2016). We construct
detailed measures of historical travel cost between cities and then use these to create a timevarying index of ‘Jewish network access’ for each city in our data. We then use this index
to create instruments for Jewish presence by recalculating the Jewish network access variable
for each city j while excluding all cities within either 100km or 250km of j. The resulting
instrumental variables predict the presence of a Jewish community in j using only the extended
network of communities, which are less likely to have characteristics correlated with those of j.
The IV analysis suggests between a 15% and 40% growth advantage for Jewish cities.
1
Our final contribution is to explore the timing and channels through which Jewish presence
impacted urban development. Using flexible regressions we find that the beneficial effects of
Jewish presence on city growth emerge between the late 17th and mid 18th centuries. Our OLS
estimates suggest that between 1750 and 1850 cities with Jews grew between 12% to 20% faster
than cities without Jews. Under IV estimation the estimates suggest a growth premium at the end
of the eighteenth century of between 30% and 100%. We further show that Jewish communities
that were Sephardic grew faster as did communities with Hebrew printing presses. Finally, we
explore the interaction between market access and the presence of a Jewish community. We show
that all cities in Europe experienced significant increases in market access after 1600. However,
we find that cities with Jewish communities differentially benefited from this increase. In other
words, while all cities experienced comparable increases in market access on the extensive margin,
Jewish cities were better able to take advantage of the intensive margin and translate greater
market density into growth. To our knowledge, ours is the first paper to calculate market access
measures for European cities in the premodern period and relate it to the onset of the Great
Divergence.
Our findings have significant implications for our understanding of the drivers of economic
growth in Europe in the period leading up to the Industrial Revolution. This period saw some
parts of Europe begin to achieve higher levels of urbanization and incomes due to greater trade,
commercialization and market integration—a process economic historians label Smithian growth.
To explain this Smithian growth, previous research has argued that institutional factors, notably
constraints on the executive (De Long and Shleifer, 1993; Acemoglu et al., 2005), the printing
press (Dittmar, 2011b), and the introduction of the potato (Nunn and Qian, 2011) played major
roles. Other research has pointed to the importance of warfare in leading to higher rates of
urbanization (Rosenthal and Wong, 2011; Voigtländer and Voth, 2013; Dincecco and Onorato,
2015). We highlight another factor that helps explain differences in city growth across Europe
after 1600: the willingness or ability of a city to accommodate religious minorities.
In so doing we contribute to a growth literature interested in the role ethnic and religious
minorities play in economic development and in the conditions that support toleration for
minority groups. Hornung (2014). quantifies the importance of the Huguenots for Prussian
development Other research examines the prominent role played by the Copts in the premodern
Egyptian economy (Saleh, 2013). Relatedly Arbath and Gokmen (2015) study the role played by
Armenians in the Ottoman empire and modern Turkey. Jha (2013) and Diaz-Cayeros and Jha
(2014) demonstrates how economic complementarities between minority and majority groups in
India and Mexico respectively promoted peaceful coexistence.
2
The economic importance of Jews in the European economy from the early middle ages onwards
means that the role played by Jewish merchants, traders and financiers has been intensively
studied by historians. The most ambitious and wide-ranging studies include Sombart (1911)
and Baron (1975) and are qualitative in character. Recently, building on the classic work of
Kuznets (1960), several scholars have been putting together detailed and finely grained data
on Jewish economic attainment in eastern Europe in the early 20th century (Abramitzky and
Halaburda, 2014; Spitzer, 2015). In this paper we provide a more macro-level approach to issue
of the economic role played by Jewish in premodern Europe.
A second literature that we contribute to is that on human capital and economic development.
The link between human capital growth and modern economic growth is fairly robust, at least for
the period after 1850 (Mankiw et al., 1992; Ashraf and Galor, 2011).1 The comparatively high
levels of human capital among Jews in medieval and early modern Europe is also well established.
However, since Jews were usually small minorities among their Christian hosts, it is less evident
whether or not their high human capital could contribute to overall city growth. The argument
that elite human capital may be particularly important for transmitting knowledge and ideas has
been made recently in several papers (e.g. Squicciarini and Voigtländer, 2015; Gennaioli et al.,
2013). To the extent that we think that the channel we identify linking Jewish presence to city
growth is based on their comparatively high level of human capital our analysis is consistent
with their arguments.
Finally, our historical setting naturally leads us to ask whether cultural or religious factors play
an important in explaining the relationship between Jewish communities and urban growth. In
this respect we contribute to the growing literatures on the economics of religion and culture.
Our study is closest to Cantoni (2015) who studies the consequence of adopting Protestantism in
Germany for city growth. He finds no effect of adopting Protestantism on city growth in the
early modern period. Our analysis is also related to Becker and Woessmann (2009) who do find
an effect of Protestantism in Prussia for human capital accumulation in the nineteenth century.
Our findings also relate to a specific literature that looks at the role played by Jews in European
economic history.2 Botticini and Eckstein (2012) document the human capital advantage that
1
An older literature on British industrialization downplayed the role of human capital as measured by literacy
(see Mitch, 1999). But human capital played a crucial role in enabling Prussia to catch up to Britain in the late
nineteenth century (Becker et al., 2011) and in the economic development of the United States in the twentieth
century (Katz and Goldin, 2008). Moreover recent accounts of the Industrial Revolution in Britain now emphasis
the importance of other dimensions of human capital beyond literacy in explaining Britain’s initial economic
advantage in industrializing (Kelly et al., 2014). For theoretical foundations for why human capital became
important for growth after 1850 see Galor and Weil (2000).
2
See, in addition to cited work, earlier more qualitative studies in economic history by Roth (1961); Baron
(1975); Kahan (1986).
3
Jews possessed from the medieval period onwards. They argue that it was this comparative
advantage in occupations that required literacy or numeracy that led to their specialization
as merchants, traders, and moneylenders. In particular, Pascali (2015) examines the presence
of Jewish communities in medieval Italy and shows that financial institutions and knowledge
persisted over centuries there. Because the growth premium associated with the presence of a
Jewish community only emerged after 1600, we argue that the high level of human capital among
Jews did not have a broader effect on economic growth during the middle ages when restrictions
limited Jewish economic activity and long-distance trade remained limited and confined to a
few commodities. If there was a positive interaction between Jewish human capital and city
growth, it only emerged once institutions enabled Jews to participate more freely in the European
economy.
The structure of the rest of the paper is as follows. Section 2 outlines our hypothesis, provides
the necessary historical background for our analysis, and summarize our data. In Section 3 we
present our main results and show that they are robust. In Section 4 we describe our instrumental
variables approach and report results on the causal impact of Jewish presence. In Section 5
we investigate both the timing of the Jewish city growth advantage and the possible channels
through which this advantage emerged. We conclude in Section 6 discussing the implications our
results for our understanding of the origins of modern economic growth in Europe.
2
Historical Background, Hypotheses, and Data
2.1 Historical Background
By the middle ages, Jewish communities flourished across most of Europe. In some instances
such as in Spain and Italy these communities dated to Roman times. Elsewhere they were the
product of more recent settlement. Jews had settled in Germany in the 9th and 10th centuries
and in England from the 11th century onwards. They played an important role in trade in this
period and, over time, became increasingly involved in moneylending and banking (see Chazan,
2006, 2010). By 1100, virtually all major cities in Europe had a Jewish community.
From approximately 1200 onwards, Europe’s Jews faced increasing amounts of discrimination
(Anderson et al., 2015). Laws restricting settlement were the primary determinants of whether
a city had Jews living in it or not (Goldscheider and Zuckerman, 1984). As a result, there
was a large amount of local variation in Jewish presence. During the pre-industrial period,
England after 1655 was the only European country where Jews were free to settle where they
wished. Everywhere else Jewish settlement rights were conditional and varied unpredictably at a
local level. There was a flourishing Jewish community in Fürth in Bavavia but ‘in neighboring
4
Nuremberg a Jew could appear only in daytime and only in the company of a local inhabitant’
(Katz, 1974, 12). Even in the Netherlands, often characterized as a uniquely religiously tolerant
state, Jews were excluded from Utrecht, Gouda, and Deventer. The French monarchy in the
17th century permitted Jewish settlement in the regions it conquered from the Holy Roman
Empire and allowed Sephardic communities to settle in Bordeaux and Rouen but Jews were not
permitted to settle in Paris (Attali, 2010, 285).
2.2 Hypotheses
There are several reasons to think that Jewish communities could have had a positive effect on
economic growth in premodern Europe. We categorize these hypotheses as follows: (1) a human
capital mechanism; (2) a cultural transmission mechanism; and (3) a market access mechanism.
1. The human capital mechanism. Botticini and Eckstein (2012) document that Jews in medieval
Europe had higher levels of human capital than did Christians. In the ancient world Jews were
mostly farmers whose religious activities centered on the Temple in Jerusalem. Botticini and
Eckstein argue that Jews specialized in trade and commerce during late antiquity and in the early
middle ages because of a shift in religious doctrine following the destruction of the Temple in AD
70 that saw the rise of Rabbinical Judaism with its emphasis on mandatory male literacy. As a
result of this religious change, individuals facing a high opportunity cost to becoming literate
had a strong incentive to convert to either Christianity or Islam. The minority who remained
Jewish eventually came to specialize in long-distance trade, the wine industry, medicine, and in
providing financial services and moneylending (Botticini and Eckstein, 2012, 194).
Jews did have higher than usual human capital attainment throughout the medieval and early
modern period.3 Simon Kuznets established that this remained the case in the late 19th century
(Kuznets, 1960). Acemoglu et al. (2011) show that this held true in Eastern Europe before the
Holocaust.4 If Jewish minorities had higher human capital than their Christian neighbors for
religious reasons and if human capital is important for economic development then we should
expect cities with Jewish communities to be more successful economically.
2. The cultural transmission mechanism. A growing literature emphasizes the importance of
cultural values in shaping economic outcomes.5 This literature builds on the argument of Max
3
Higher average levels of Jewish literacy, and scientific knowledge are reflected in the extent to which Christian
society depended on Jewish doctors, merchants and moneylenders (Roth, 1953; Parkes, 1976; Israel, 1985; Kahan,
1986; Hunderet, 1987; Cohen, 1994; Shatzmiller, 1994; Hsia and Lehmann, 1995; Stacey, 1995; Lehmann, 1995;
Mundill, 2002).
4
Abramitzky and Halaburda (2014) find that Jews were not more educated than urban-non Jews. In other
words, the literacy advantage of Jews in pre-World War 2 Poland was a simple composition effect. Nevertheless,
this begs the question: why were Jews concentrated in urban sectors.
5
Contributions include Greif (2006); Guiso et al. (2006); Doepke and Zilibotti (2008); Tabellini (2008). See
5
Weber (1930) who claimed Calvinism played a role in building a spirit of capitalism. The
evidence for this hypothesis is decidedly mixed (see Becker and Woessmann, 2009; Cantoni,
2015). However, the idea that religious cultural traits can play an important role in spurring
economic growth needs to be taken seriously. McCloskey (2010), for one, has argued that it was
central to the onset of modern economic growth.
While Weber did not emphasize Jewish values in particular, this idea was developed by the
controversial historical economist Werner Sombart. Sombart highlighted the role Jewish traders
played in inventing credit instruments in the middle ages and their role as financiers in the early
modern period.6 Sombart’s work is problematic; not least because he later became a National
Socialist and more recent work indicates that he both exaggerated the role Jewish traders played
in creating credit instruments and downplayed the heterogeneity within Jewish communities.
Subsequent scholars have also found a number of factual errors in his work. Nevertheless, Die
Juden und das Wirtschaftsleben (1911) develops a number of insights that are important for
our analysis.7 In particular, Sombart saw Jews as embodying a commercial ethos that made
them exceptionally successful in market society. Sombart claimed that these values–what he
called ‘the capitalist point of view’–spread to the rest of the population during the 18th and 19th
centuries. According to this view, the presence of Jews may be correlated with economic growth
because individuals in these cities were more likely to develop market-orientated cultural values.
3. The market integration mechanism.
The period 1500-1800 is seen as prelude to the onset of sustained economic growth. For this reason,
market integration in this period has been widely studied as a potential driver of urbanization
and subsequent economic growth. Within this literature numerous economists use the law of one
price and other measures of price dispersion as tests of the level of market integration (e.g Shiue
and Keller, 2007; Bateman, 2011; Chilosi et al., 2013). The consensus of this literature is that
grain markets became increasingly well integrated from the late 17th century onwards.8
Europe’s Jewish communities were a tiny portion of the continent’s population, but they were
disproportionately involved in trade and commerce; in no small part because they had cultural,
Alesina and Giuliano (2014) for a recent survey of the connection between culture and institutions.
6
Weber commented on a resemblance between Jewish cultural values and Protestantism values but he
downplayed this arguing that the role of Jewish communities was limited to ‘pariah capitalism’, a phenomenon he
termed ‘speculative’ in contrast to the Puritan ‘bourgeois organization of labour’ (Weber, 1930, 245).
7
See Davis (1997) for a balanced assessment of Sombart’s hypothesis. The English translation of the work by
Mordecai Epstein edited and shortened the passages in Sombart’s work which stressed racial factors (see Davis,
1997, 59).
8
Bateman (2011) argued that levels of market integration were stationary between the medieval period and
the onset of the industrial revolution. However, by expanding the sample of cities considered, Chilosi et al. (2013)
show that northwestern Europe had significantly more integrated markets by 1750.
6
linguistic and religious links across the continent. In Amsterdam, Portuguese Jews were heavily
involved in the Atlantic trade, particularly in sugar, tobacco and diamonds (Bloom, 1936). In
Poland they were involved in river trade with Russia, the Ottoman Empire and the Baltic. In
Germany Jews were closely associated with cattle trading (see Bell, 2008, 127-129). Sephardic
Jewish communities, in particular, became associated with international trade and with the
evolution of a ‘cross-cultural merchant network’ (Trivellato, 2004, 37). And cities like, most
famously, Livorno, settled by Sephardic Jews became centers of international trade during the
early modern period. For this reason, there are strong reasons to suspect that one channel
through which the presence of a Jewish community might benefit a city economically would be
via access to trade networks. We test this hypothesis by constructing measures of market access
based on the recent work of (Donaldson and Hornbeck, 2016). We are the first to adapt this
approach for premodern European cities in order to estimate an independent measure of market
growth and expansion in the years leading up to the Industrial Revolution. We will show that
that cities with Jewish communities benefited substantially more from greater market access
than comparable cities without Jewish populations.
2.3 Data
The measure of economic development we employ is the population of a city. This is a widely used
metric in the literature on economic development in the preindustrial period (e.g. see, De Long
and Shleifer (1993), Dittmar (2011b), Nunn and Puga (2012). Bosker et al. (2013), Dincecco and
Onorato (2015), Jebwab et al. (2015)). Given that the European economy was Malthusian in the
preindustrial period, population urban population data are an important source of disaggregated
information concerning technological change and productivity in both commerce and agriculture
(de Vries, 1976; Ashraf and Galor, 2011). Cities were centers of productive activity but they were
also disease ridden and unhealthy—urban death rates almost always exceeded rural death rates.
As such, preindustrial cities rarely grew via natural increase. To expand they had to attract
migration from the countryside by offering higher wages and greater economic opportunities.
Their ability to do so was constrained by the productivity of their surrounding agricultural land.
Increases in productive capacity were thus reflected in the growth of city populations. City
growth in the preindustrial period also reflected commercial success with many cities, such as
Genoa, Venice, Antwerp, Amsterdam, and London being, first and foremost, centers of trade
(Braudel, 1982).
We combine two main datasets for our analysis. We collect data on the presence of Jewish
communities at the cityXyear level from the Encyclopedia Judaica (2007). These are the same
data as used by Anderson et al. (2015) except we have expanded the time period covered by
fifty years so that the data extends from 1100 to 1850. There are 1,069 cities that had a Jewish
7
!
!
! !
!
!
!
(
!
!
!
!
!
!
!
!
!
((!
!
!
!
!
!
!
!
! !!
!
!
!
!
!
( (
(
(( (( (( ( ( ( (
(
(
( (
( (( (
(( (
( ( ((
(
!
!!
!
! ! !!
!
!
!
!
!!
!
!
!!
!
!
!
!
!!! !
! !
!
!
!
!
!
!
! ! !
!
!
!
!
!
!
!
!
!
!
!
!
! !
!
!
!
!!
!
! !!!
!
!!!
!
!
! ! !
!!
!
! !
!!
!
!
!
!
!
!
!
!
!
!!
!
!
!!
!
! ! ! !
!
!! ! !!
!
!
!
! !!! !! !
!! ! ! ! !
!
!
!
!
! !
!
!
!!
!
!
! !! !!!
!
!
!!
!
!! ! !!
!!
!
!!
!!
!!
!
!
! !!!
!
!! !
!
!!
!
!! !
!
!
!
!
!
!
! !
! !
!
!!
!!
!!! !
! !!
! !! !
!
! !!
!
!
!! !
!
!
! ! !!
! ! !
! !
! !
! ! ! ! ! !
!!! ! !
!
!
!
!
!
! !
!!
!
!
! !! ! ! ! ! ! !
!
!
!
!!
! !! !! ! !! ! !
! ! !!
!!
!
!
!
!
!
!
!!!! !
!
!
!
!
!
!
!
!
!
! ! !! ! !
! ! ! !!
! ! !
!! !
!!!
! !
!
!
! !!! !
!
!
!
!
!
! !!
! !!!!
!
!!
!!
!
! !!
!
!!
!
!
!! !
!! ! ! ! !
!!!
!!
!!! !
!!! !!! ! !! ! ! ! !!
!
!!
!!!!!!
! !!
!
!
!!!! !!
!!
! !
!!!!
!
!!!
!
! !!
!!!
! !! !
!
! ! ! !!
!
!!
! !
! ! ! ! ! ! ! !! !
!
!!
! ! ! !!!! ! !!!! !
!
!!
!
!
!!!!
!!
!
!! !
! !!
!! !!! ! ! ! ! !! !
!
! !
!
!
!
!!
! !!
!
!!! !
!!!!!!! ! ! !
!! !
!!
!
!!
! !
!
! !
! !!
! !
!
!
!
! !
!!!!
!
! !
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
! !
!
! !!
!! !
! !! !
!
!
!
!
!
!
!
!
!
!
!
! ! ! !
!
!
!
!!
!
!!
!
!
!
! !
! ! !
!
!
!!!
! !! ! ! !!
!
! ! ! ! !!
!
!
!
!
!
!
!
!
!
! !
! !
!!
!!
!
! !!
! !
!
!
!
!
!!
! !! ! !
! !
! !
! ! ! ! !!
!
! ! !!
!!
! ! ! !! ! ! !
!!!
! ! !
!!
! !! ! !
!
!!
!
! !
! ! !
!
!
!
! !
!
! !
!!! !
!
!
!
!
! ! !
!
! ! !! ! ! ! !
! !! !
!
!
! !
! !
! !
!
!
!!
!
! ! !
!
!
!
!! !
! !
!
! !
!
!
!
!
!
!
!
!
! !!
! !!
!
! ! !!
!
! !
!!!
! !
! !!
! !
!!
!
!
!
!
!
!
!! !
!
! !!!!
!! !
!
!
!!
!!
!
!
!
!
!!
! !!
!
!
!
!!
!! !
!
! !
!
!!
!
!
! !
!
!
! !
!
!
!
! !
!!
!
!
!
!!
!
!
!
!
!
!
!
! !
!
!
!
!
! !
!!
!!
! !
!
! ! !
!
!! ! !
!
! !
!
! ! !
!
!
!
!!
!
! ! ! !
!
! ! ! !!!!
!! ! !! !
!
!
!
! ! !!
!
!
!
!
!!
! !!! !
! ! ! ! ! ! ! !!!
!
! !!
!
! ! !
!
!!
!! !
! ! !!
!! ! !
!!
!!
!
!
!!
!! !
!
!
!
!
!!
!! ! ! !
!
!
! !
!
!! !
! !
!
!! ! ! !
!!
!!
!! !
!!
!
!
!!
!! ! !
!! !
!
!
!
!
!
!
!
!
!
!
!! ! ! !
!!
!! !
!!!! !
! !!
!
!!
!
!!! !! ! ! !
!!
!! !
! !! ! !
!
! !!
!
!
!
! !! ! !
! ! ! ! !!
!!
!!
! ! !
! ! !! !
!
!!
!
!
!!
!
!!!
! !! !
!
! !! !
!!
!
!
!
!
!
!
! !! !
!
!!
!
!!
!
!!
!!
! !
!
! ! !!
!
!! !!
!! !
!
!
!
!
!
!
! !
! !
!
!!
!! !
!!
!
!!
!
!
!
!
!
!
!!
!! ! !
!
! !
! !
!
!
!
! !!
!!
!! !
!
! !!!!
!
!
!
!
!!
!!!
! !! ! ! ! !!
!!
! ! !
!
!
!!
!!
! ! !
!
! ! !!! ! ! !
!!
!
!
!
!
!
!
! !
!
!
!!!
!
!
!!
!!! !
!
! ! ! !
!!
!!
!
!!
!!
!!
! !
!!
!
!
!! !!!
!!
!
!
!!
!
!
!! !
!
!
! !
! !
!
!
!
! !
!
!
! ! ! !!
!
!! !! !! !
! !!
!
! !
!
!!!
!
!
!
!
!! ! !
!!!
!!!
! !
!
!
!
!
!
!!!
! !! !
!!!
!
!
!
!
!
!
!
! ! !
!
!! !
!
! !
!!
! !! !! ! !
! !! ! !
! !
! !
!
!
!
!
!
!! !
!
!!!!
! ! !!
!
!!
!! !! ! !
!! ! !
!
!
!
!
!
!
!
!
!
!
!
!! !
!! !
! !!!
!!
!
! !
!!
! !!
!
! !!!!
!
!! !
!
!
!!
!! ! !
!
! ! !!
!!!
! !!! !
!
!!!!
!!
!!
!
!!
!!
!!! ! ! !!
!
!
!!
!
!
!!!
!!!!!!! !!
!!
!!
! !!
! ! ! ! !!
!!
!!
!!
!
!!
!!
!!!
!!!!
!
!!!!
!! ! ! !
!!
!!!!!
! !!!
!! !!
!
!!
!
!
!!
!!!
!!!!!!!!
!
! ! !! !! ! !
!
!
!!! !
!! !!!!
! !!!!
! !
!
!!
!
!
!! ! !
!!!!
!
!! ! ! ! ! ! !
!
!
!
!
!
!
!!!
! !!
!!
!
(
!
(
((
(
(
(
( (
(
(
(
(((
(
( (( (
(
(
(( (
((
(
(
(
(
((
(
(
(
(
((
(
( ( ( (( (( (
((
((((
( ( ( (( ( ((((( (
( (
(
(
(
(
((
( ((
((
((
(
( ( (( ( ( (
( (((((
(
(
(
(
(
(
(
(
(
(
(
(
( (( ( ( (
( ( (((
((( (
(
(
(
(( ((( (((
(
((((((
(
(
(
(
(
((((( ( (((
( ( (( ( ((
((
(( (
((
( (
(
((
(
((
(( (
(( (((( (( (
((
(
(
(( ( (( ( ((( ( ( ( (
( (( (( ((
(
(((
(
(
((( (
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
((
(
( ((( (
(
((
((((( ( ( (((((
((( ((( ((((
((( (
( (((
(
( ( (( (
((
((
(
(
(
(
(( (
(
((
((((
(
((((
( ((
( ((
((( (
((
(
( ( ((((( ( (( ( (((((
((
(((
(
(
(((( (((
(( ((( ( (((( ((
(((
(
( (
(
(
(
((
(
(
(
( (
((
((
(
( (( ((((
( ( ((((
(
(((
( (
((
(( (
(
(
(
((
(
( ( ( ((((
( ( ((
( ((
(
(
(
(
(
(( ( ( ((((((
(
((( (( ( ( ( ((
(
(
(
(
(( (
(
(
(
(
(
(
(
(
(
(
(
(
( ( ( ((( ( ((
(
(
(
(
(
(
(
(
((
(( (
( (
(
((
((
(( ( (( ( (
(( (
( ( (( (
(
(
(
(
(
(
(
(
(
(
(
(
( ((
(
(
(
(
(
(
(
(
((
(
( (
(
( ( (
(
( ((
(
((
(
((
((
( (
(( ( (
(((
(
(( ( (( (
(
(
(
((
(
( (
((
(
(
( ( ( ((
(
((((((
(
(((( (((((
((
(
(
( ((
(
(
( (
(
( ( ((
((
(( ((
(
((( (
(
(
(
(
((
( (
(
( (( (
(
(
(( ((
(
(
(
Jewish Cities
(( ( (
((( (
( ((
( ((
( ( (( (
(
(
(
( (
( (
( (( ( ((( (
(
(
(
( (( ( (
(((
(
(
(
(
(
(
(
(
(
((
(
(
((
((((
Bairoch Cities
(
((
((
(((
(( (
(
((
(
(
(
(
(
(
(
((
(
(( (
( ( ((
(
(
(
(
(((
(
(
(
((
(
(( ( (
((
((
( ((
(
(
((( (
(
(
(
(
(
(
(
(
(
(
(((
( (
(
(( (( (
((
( (
( (
((((
(
(( ( (
(( ((
(
((
(
(
((
(( (((((
( ( ( ((
(
(
(
(
((((
(
(
(
(
(
(
( ( ( ((
(
( ( (( ( (( ( (
((
(
( ( ( (((
(
((
( (( ( (( (
(( (( ((
(
(
(
(
(
((
(
(
(
(
(
( (( (
(
(
(
(
(
(
(( ((
(
(
(
(
(
((
((
(
(
((( ( (
( ((
((
( ((
(
((( (
((
(
(
(
(((((
(
( ( ((
((
(
(
(((
((
( (
(
(
(
(
!
( ((
Figure 1: Matching the Bairoch and Jewish cities. Location of the cities in the Bairoch
dataset are solid black dots. Locations of Jewish communities are shown as circles. See text
for sources.
community at some point in these data. The majority of entries in the Encyclopedia Judaica
contain an exact date at which a community was allowed into a city. Other entries provide an
estimate of the century or decade in which a community is first known to have been established.9
Figure 1 shows the locations of these cities as open circles.
We combine the data on Jewish presence with those on city populations from Bairoch (1988).
This dataset contains all cities with populations greater than 1,000 between 800 and 1850. The
total number of cities in the dataset is 1,757. We use 1,792 of these as 5 cities in northern
Norway, Finland, and the far Western Atlantic cannot be matched to the GIS data that we
employ to create our geographical controls.10 Figure 1 shows the location of the Bairoch cities as
solid black points.
To match the Bairoch to Jewish cities we create a list of all cities in the Bairoch dataset within
15km of a Jewish city and then check each of the matched Bairoch-Jewish city pairs by hand.
Further details of the matching procedure are contained in the Appendix. We end up with two
samples. The first, which we call the main sample, consists only of Bairoch cities which we
9
Note that we focus on Jewish communities and do not have data on the presence of converted Jews in Spain
or Portugal.
10
The dropped cities are: Bergen, Trondheim, Ponta-Delgada, Falun, and Gaevle.
8
could perfectly match to the Jewish community data. The second, which we call the extended
sample, includes all Bairoch cities where we assume that if the city was not matched with a
Jewish community, then there was no community present. The combined data constitute a 10
period panel with observations in 1100, 1200, 1300, 1400, 1500, 1600, 1700, 1750, 1800, and 1850.
Our main variable of interest in both samples is ‘Jewish Community’ which is a dummy variable
equal to 1 if there was ever a Jewish community in the city during the previous time period.
Descriptive statistics for both samples are provided in tables A.1 and A.2 in Appendix A.6.
In Appendix A.6 we also provide information on how Jewish presence in European cities evolved
over time. Figures A.7 and A.8 show that before the Black Death (1347-1352), most cities in
Europe had Jewish communities. In the main sample, about 90% of cities had Jews before the
14th century. In the extended sample the respective number is about 75%. The nadir of Jewish
presence is reached in the 16th century when only about 40% of cities in the main sample had
communities. This number then gradually increases until it reaches 60% in 1850. In terms of
sample sizes, in the main sample, the fewest number of cities overall is 86 in 1200 and the highest
number of cities is 463 in 1850. The extended sample goes from a minimum of 107 cities in 1200
to 1,656 cities in 1850.11
3
The Relationship between Jewish Communities and City
Growth: Main Results
We begin by exploring the relationship between the presence of a Jewish community and
subsequent city growth over the entire period of study by estimating the following specification
which is based on the discussion in Duranton and Puga (2013):
PopGrowthi,c,t = βJewish Communityi,c,t + γlnpopi,c,t−1 +
1850
X
t=1200
0
Xi,c µt
(1)
+ δc + ηt + λc × Year + εi,c,t .
The dependent variable is the percentage growth in population of city i in modern-day country c
in period t.12 The variable of interest is the dummy Jewish Communityict which takes a value of
one if there was a Jewish community in city i during the previous century and a zero otherwise.
X is a vector of controls that measure local geography (cereal suitability, proximity to rivers,
proximity to coast) and local infrastructure (presence of university and distance to Roman road
11
Neither sample ever has all of the possible Bairoch cities (1,792) because some cities drop out of the sample
over time.
12
We calculate percentage growth as a log difference.
9
intersection).13 We allow the effect of these controls on city growth to vary by year and, as
such, we allow city i’s steady state growth rate to be a time varying function of local geography
and economic infrastructure. We also include time fixed effects: ηt in all regressions to allow
for common time-varying shocks that affect city growth (such as the Black Death of the 14th
century). Finally, in our full specifications we include modern country fixed effects along with
their interaction with a time trend so as to allow the steady state growth rate to vary with
unobserved political and geographic variables correlated with modern country boundaries (e.g.
language use, ethnic composition, etc. . . ).14
We include lag population in all specifications as city growth in premodern Europe was limited
by a fixed factor. This assumption is consistent with models of growth in a Malthusian world.
Economic historians usually assume this fixed factor is ‘land’, though a more subtle interpretation
is that transportation costs limited the amount of usable land for cities (Dittmar, 2011a; Heckscher,
1955). In our regressions we find robust evidence for convergence between small and large cities
(negative γ). A one log point increase in city population was typically associated with a reduction
of between 10% and 15% in growth rates.15
In Equation 1 we do not include city fixed effects. There is a debate in the cross-country growth
literature over the appropriateness of using fixed effects in growth regressions that also include the
lag of the level of the dependent variable as a control (Barro, 2012). While city fixed effects offer
a simple way to control for differences in unobservables across cities. We feel that in estimating
the impact of a Jewish community subsequent city growth they produce misleading estimates.
In addition to the potential bias the inclusion of such fixed effects could introduce (see Nickell
(1981)) due to including the lag of city population as a control, it is also possible that the within
city variation in growth may be correlated with Jewish migration and thus introduce greater
bias than when the FE’s are excluded (see, e.g., the discussion in Duranton and Puga (2013) p.
8). Lastly, we also do not want to identify primarily on those cites that either expel or murder
their Jewish communities as this is a major source of the time variation in Jewish community
presence (this is a particularly strong effect for the years following the Black Death in the 14th
century). Nonetheless, we present all of our baseline, IV, and robustness regressions using a
difference-in-differences specification (log level of pop as dependent variable, no lag of population
on the RHS, and city fe’s) in Appendix Tables B.3, B.4, and B.5 These results are consistent
with what we find using Equation 1.
13
Details on how these controls are constructed are contained in the Appendix.
We also experimented with allowing the country dummies to be estimated separately for each period. This
produces largely the same coefficients as in the analysis we present here, however, it also creates collinearity
which results in missing test statistics for some regressions.
15
1 log point of city population also happens to be almost exactly 1 standard deviation in the main sample.
14
10
Table 1: Jewish Communities and City Growth, 1100-1850
Dependent Variable: City Growth
Main Sample
Jewish Community
Lag Population
Year FE’s
Controls X Year FE’s
Country FE’s X Year
N
R2
Extended Sample
(1)
(2)
(3)
(4)
(5)
(6)
0.0608∗∗∗
(0.0180)
-0.103∗∗∗
(0.00922)
0.0768∗∗∗
(0.0186)
-0.113∗∗∗
(0.00987)
0.0604∗∗∗
(0.0189)
-0.109∗∗∗
(0.0101)
0.122∗∗∗
(0.0144)
-0.148∗∗∗
(0.00707)
0.148∗∗∗
(0.0150)
-0.156∗∗∗
(0.00744)
0.132∗∗∗
(0.0151)
-0.147∗∗∗
(0.00740)
Yes
No
No
2860
0.167
Yes
Yes
No
2860
0.207
Yes
Yes
Yes
2860
0.264
Yes
No
No
7377
0.172
Yes
Yes
No
7377
0.202
Yes
Yes
Yes
7377
0.258
Notes: Columns (1)-(3) use our main sample where we assign cities for which we have no
recorded Jewish presence as missing values. Columns (4)-(6) use our extended sample which
employs an alternative coding for the presence of a Jewish community that assigns cities
a zero if there is no record of a community. All specifications include year fixed effects.
Controls include cereal suitability, distance from a Roman road, and the intersection of a
Roman road, and medieval universities. Robust standard errors reported in parentheses. ∗
p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01.
We report our estimates of β in Table 1. Regressions (1) - (3) shows the result of estimating our
baseline specification for the main sample and columns (4) - (6) show results using the extended
sample. All specifications suggest that, between 1100 and 1850, cities with Jewish communities
grew faster than cities without Jews. In our preferred specification in column (3) which uses
the main sample and includes all controls the estimated effect of a Jewish community on city
population growth in the previous period is about 6%. Consistent with models of Malthusan
growth in which there is a fixed factor that limits the size of large cities, in specification (3) we
find that a one log point increase in lag city population reduces subsequent growth by about
11%. The negative sign on lag population suggests there was some sort of quasi-fixed factor of
production constraining the growth of large cities (Dittmar, 2011a). The coefficients on Jewish
Community and Lag Population are stable when we remove all controls (specification (1)) and
when we don’t include the modern country fixed effects (specification (2)).
The estimates reported in Table 1 are unbiased if the error term in Equation 1 is uncorrelated
with the Jewish community presence. As our data are non-experimental we cannot guarantee that
this is the case. As such, we now introduce an identification strategy which uses instrumental
11
variables to establish a causal relationship between Jewish communities and urban growth.
4
Access to the Network of Jewish Communities: IV Analysis
Several factors could affect Jewish settlement. On the one hand, Jews were often permitted in
certain cities by rulers because they ‘anticipated Jewish contribution to the economy’ (Chazan,
2010, 102).16 This could result in Jews settling disproportionately in unprosperous or unpromising
regions since they were intended to boost the coffers of local rulers. This was definitely the case,
for example, in Germany after the Thirty Years War (1618-1648) (Israel, 1983, 19-22).17
Another factor that influenced Jewish settlement were exclusions and expulsions. Jews were often
perceived as competitors by merchants and, thus, if a city had a powerful enough mercantile
class they were sometimes able to exclude Jews from settling. This was the case in Turin until
1424, Florence until 1437 and Milan (Roth, 1950).18 In Bologna Christian bankers succeeded
in excluding Jews until the second half of the fourteenth century (Foa, 2000, 116). Jews were
also subject to expulsion from a city or region. These could occur at the national level, as in
the infamous expulsions from England in 1290, Spain in 1492 and Portugal in 1497 or at a local
level. These local expulsions and persecutions were partially driven by economic downturns and
shocks (see Anderson et al., 2015).
If Jews were invited to settle in declining cities and expelled or excluded from prosperous ones
this would be a source of downwards bias in our OLS estimates. On the other hand, if Jews
could decide to selectively migrate to cities which were more prosperous then this could be a
source of upwards bias. There is actually little historical evidence for such positive selection – the
bulk of the qualitative literature simply suggests that Jewish traders and merchants sought to
establish communities in as many cities as possible for purposes of trade. Furthermore, as Jews
became more involved in moneylending in the middle ages, this gave them a further incentive to
16
Chazen notes: ‘the objective was to entice new Jewish settlers . . . Early sources tell us of the invitation
extended by the Duke of Flanders to Jews to settle in his domain, of the establishment of a Jewish community in
London by William the Conqueror, newly installed as king of England’ (Chazan, 2010, 6). This was also the case
in Spain during the Reconquista.
17
Israel carefully documents the revival and expansion of many German Jewish communities during this period,
concluding that ‘the terrible upheavals of the Thirty Years’ War mostly worked in favor of German and all Central
European Jewry, appreciably enhanced the Jewish role in German life, and prepared the ground fort the “Age of
the Court Jew”—the late seventeenth and early eighteenth century—the high-water mark of Jewish influence on
Central European commerce and finance’ (Israel, 1983, 30).
18
For instance: ‘it was either small or middle-sized communes (which had to call on outside financiers) or
strong governments (concerned with public order) who turned first and from choice to the Jews. In the plutocratic
towns, on the other hand, coalitions of local interests opposed to their admission were able to delay it’ (Poliakov,
1965, 58). Foa writes: ‘Not all Italian cities accepted or solicited settlement by Jews. Cities in which Christian
bankers were numerous and organized in guilds were generally hostile to Jews, in whom the former saw dangerous
competition’ (Foa, 2000, 111).
12
geographically expand as much as possible in order to smooth local shocks (see Botticini, 1997).
To assuage concerns about selection we need a source of variation in Jewish presence which
is plausibly unrelated to unobservables driving a city’s growth. We generate such a source of
variation by explicitly modeling the network of Jewish communities over time in Europe using
tools drawn from the market access literature (e.g. Donaldson and Hornbeck, 2016). We are also
inspired by how historians describe Jewish communities spreading across Europe in the medieval
period along historical trade routes (Chazan, 2006; Bell, 2008). Our strategy relies on three
assumptions:
1. A Jewish community is more likely to be established close to another Jewish community
(e.g. because of trade networks, financial relationships, cultural linkages, or other spatial
externalities).
2. ‘Close’ is defined as the least cost travel path.
3. Unobservable characteristics of the cities in which the communities are located become less
correlated the further they are from each other.
The most computational challenging task in constructing the IV’s is to create a measure of the
cost of travel between Jewish cities. To do this we begin by creating maps of Roman roads,
medieval trade routes, major rivers, and seas. Estimates from Bairoch (1988) allow us to assign
the cost of travel by each of these routes (portage is assumed to be used when there is no better
alternative). We then divide Europe into 10km x 10km grids and assign the lowest travel cost
to each grid. We apply Djikstra’s algorithm to determine the lowest cost of travel between all
3,211,264 city pairs (van Etten, 2012). Using the travel cost measures, we then create an index
showing the ‘Jewish network access’ for each city. For city j this index is defined as:
N Ajt =
X
Jewish Communityit τji−σ ,
(2)
i6=j
where Jewish Communityit is a dummy variable for city i taking a value of 1 if a Jewish community
is located in it in time period t, τji is the cost of travel between cities j and i, and σ is a trade
elasticity.19 See Appendix A.3 for more details on the construction of this index.
19
The appropriate σ depends on context. For modern, developed, economies, researchers tend to estimate
higher values. For example, Eaton and Kortum (2002) use 8.28 for OECD trade flows in 1995. Donaldson and
Hornbeck (2016) estimate an average σ = 8.22 for trade flows in the U.S. in the second half of the 19th century.
By contrast, Donaldson (2016) estimates σ = 3.8 for colonial India. Storeygard (2016) estimates the elasticity of
city economic activity with respect to transport costs across Africa and arrives at values consistently less than 1
(their preferred estimate is 0.28). Kopsidis and Wolf (2012) assume σ = 1 for their study of Prussian trade during
the Industrial Revolution. This is also the value assumed by many earlier studies of ‘market potential’ or ‘market
13
Jewish network access itself is, of course, correlated with the unobservable characteristics of the
city for which it is calculated. To overcome this we define our time varying instruments as:
Zjt = N Ajt .
(3)
i>D
Where D is the linear distance of city i from city j. In doing this we predict the presence of a
Jewish community in city j based only on the network of Jewish communities that are more
than D kilometers away. Since it is likely that any unobserved variables correlated with both the
growth of city j and Jewish presence in city j will be uncorrelated with Jewish presence in city i
that is D̂ kilometers away, then this is a potentially valid instrument.
There should, of course, be a trade-off between the relevance of the instruments and their validity
as D increases. As such we create the instruments using values for D of 0, 50, 100, 250, and
500 kilometers. In Table 2 we report the second stage results of running these regressions using
the main sample.20 As expected, for D = 500km the relevance of the instruments are quite
low (first stage F-stat = 4.68) and estimated coefficient is statistically insignificant. Somewhat
surprisingly, however, the instruments created using D = 0km also have relatively low relevance
compared to D = 50km, 100km, 250km. One plausible explanation for this is that the likelihood
of Jewish community being in a city will be less if there is another community very close by.
This might be especially the case if the skills of Jews (e.g. banking or trade) are substitutes
when they are spatially close to one another, but become more complementary with distance (up
to a point). The instruments setting D = 50, 100, 250 kilometers are extremely good predictors
of Jewish community presence with first stage F-stats of 134, 95, and 27 respectively. Figure A.5
illustrates the very robust correlation between the D = 100km instruments and the likelihood of
a Jewish community.
Considering the trade-off between relevance and validity, we prefer the IV regressions using
D = 100km and D = 250km that are reported in columns (3) and (4). The coefficient in Column
(3) is statistically significant at the 5% level and equal to 0.185 which is consistent in size with the
estimates we obtain from using our OLS regressions using the extended sample and considerably
larger than the coefficients we obtain from using the main sample in Table 1. It suggests that
over the entire sample average Jewish city growth was 18.5% greater than for non-Jewish cities.
access’ (Harris, 1954). Since our study covers relatively underdeveloped markets in Europes between 1100 and
1850, we follow Storeygard (2016) and Kopsidis and Wolf (2012) by setting σ = 1 which is lower than what is
preferred for studies of more developed economies, but higher than what it is estimated as for underdeveloped
regions in Africa today.
20
Running the regressions using the extended sample makes little sense considering that, mechanically, the
instruments will lose relevance as the only difference between the main sample and the extended sample is the
addition of cities that, by definition, never have a Jewish community.
14
Table 2: Jewish Communities and and City Growth, 2nd Stage IV Analysis, 1100-1850
Dependent Variable: City Growth
Main Sample
(1)
All Cities
(2)
>50km
(3)
>100km
(4)
>250km
(5)
>500km
Jewish Community
0.169
(0.107)
0.147*
(0.0850)
0.185**
(0.0937)
0.418**
(0.181)
-0.929
(0.620)
Year FE’s
Controls X Year FE’s
Country FE’s X Year
N
First Stage F-stat
Yes
Yes
Yes
2860
55.69
Yes
Yes
Yes
2860
133.95
Yes
Yes
Yes
2860
95.27
Yes
Yes
Yes
2860
27.40
Yes
Yes
Yes
2860
4.68
Notes: This table presents our 2nd stage IV estimates using the main sample. Column 1
uses our simple Jewish Network Access measure. Columns 2-5 use our instruments where
we exclude cities within a 50, 100, 250, and 500 km radius respectively. Our preferred
specifications are in Columns (3) and (4). Robust standard errors reported in parentheses.
∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01.
In Column (4), using D = 250km, the coefficient blows up to 0.418 and retains a significance
level of 5%.
It is natural to ask whether the magnitude of these effects is plausible? Dittmar (2011b) finds
that between 1500 and 1600, European cities where printing presses were established in the 1400s
grew 60% faster than otherwise similar cities. Nunn and Qian (2011) finds that a one percent
increase in potato suitability increased city size by 0.5% implying that an area which was twice
as good for growing potatoes would have an 50% boast to population size. Given these findings
in other papers, ours estimates of between a 19% and 41% growth advantage for Jewish cities
seems plausible. One interpretation of the difference in size between our OLS and IV estimates
is that there is downwards bias in our baseline OLS estimates. This might be driven by selection
(whereby Jews were permitted, or forced, to settle in cities that were declining economically) or
attenuation bias (if the Encyclopedia Judaica is missing smaller Jewish settlements, for example).
Alternatively, the IV estimates may overstate the impact of Jewish communities if they are
systematically correlated with drivers of growth. One possible candidate would be access to low
cost transportation (lower τ in equation 2). In this case, our IV regressions would be identifying
the effect of being well situated in the network of Jewish cities which engage in commerce or
banking activities. However, it is reassuring to note that Jewish network access is uncorrelated
15
with market access (Figure A.6). It is also suggestive that the IV coefficients increase so
dramatically from D = 100km to D = 250km. An interpretation is that the complementarity of
Jewish cities in production increased with distance before eventually tailing off. This is consistent
with our discussion of Jews as merchants and bankers who played a vital role in creating and
sustaining long distance trade in the early-modern urban network. This interpretation is also be
consistent with our findings concerning market access and Jewish city growth in Section 5 where
we discuss mechanisms.
4.1 Robustness
A natural concern with our results might be that they are driven by a purely mechanistic
relationship between the presence of a Jewish community and the size of a city’s population.
This would be a major issue if Jewish communities were large in size as, for example, they were
in Eastern Europe in the 19th and early 20th centuries. However, this was not the case during
the medieval or early modern period. Jews made up only a small proportion of the population
of the countries where they lived. At a national level, estimates exist for the end of the 18th
century when there were approximately 175,000 Jews in Germany, 100,000 in Hungary, 70,000 in
the Austrian Empire, 40,000 in France, and 50,000 in the Netherlands and Britain. This meant
that Jews were approximately: 2.5% of the population of the Netherlands, around 1% of the
Habsburg empire, 0.6% of the population of Germany, 0.3 % of the population of Britain and
0.16% of the population of France.21
Furthermore, individual Jewish communities remained small throughout this period. There were
only a small number of exceptions to this generalization. One of the biggest communities was in
Amsterdam where the size of the Ashkenazim community was approximately 5,000 in 1674 or
2.5% of the total city population. It grew rapidly to 22,000 by 1795 or approximately 10% of the
population but this was exceptional. The largest community in Germany was Frankfurt with a
population of 3,000 in 1610. Prague also had a large Jewish community 6,000 in 1600 and over
11,500 by 1702 (Bell, 2008, 36). At its peak the Jewish population of Venice numbered 4,800.
But the vast majority of Jewish communities were much smaller.22
In Table 3 we check the robustness of our main results by running regressions using alternative
specifications and samples. For each robustness check we report the coefficient on the Jewish
Community dummy variable using the main sample, the extended sample, under IV regression
with D = 100km and under IV regression with D = 250km. All specifications include our
21
Authors’ calculations. Populations of Jewish communities are from Katz (1974). Population estimates for
the leading European countries at the end of the 18th century are from Maddison (Maddison).
22
Most German and Italian communities numbered in the hundreds. When Hanover permitted Jewish
resettlement in the 17th century it allowed in 7 Jewish families.
16
controls interacted with year as well as year and modern country fixed effects unless otherwise
noted (same controls as in Columns (3) and (6) in Table 1).
In Columns (1)-(5) we interact our year fixed effects with additional time invariant controls
that could influence city growth. Nunn and Qian (2011) find that areas with high suitability
for the potato experienced more rapid urban growth after 1700 as a result of the Columbian
exchange. With this in mind in Column (1) of Table 3 we control for the interaction between
potato suitability and year fixed effects. It’s also possible that cities at higher altitudes may have
grown less quickly. Thus, in Column (2), we include a control for the city’s elevation. Recent
research has established that levels of anti-Semitism in Europe varied at a local level (Voigtländer
and Voth, 2012). As this might affect how Jews interacted with Christian populations, we
explicitly control for expulsions of Jewish communities in Column (3) and the total number of
pogroms or expulsions experienced by a community using data from Anderson et al. (2015) in
Column (4). The Black Death was a major shock to urban development in late medieval Europe
which also was strongly correlated with Jewish persecution (Voigtländer and Voth, 2012). To
control for this shock we use city-level mortality data from Jebwab et al. (2015) interacted with
year fixed effects. Our main results are robust to the inclusion of all of these variables with
coefficients retaining their size and significance. The only exception being that the 100 km IV
loses significance when controlling for latitude.
17
Table 3: Jewish Communities and and City Growth, Robustness, 1100-1850
βOLS Main
βOLS Extended
βIV 100km
βIV 250km
N Main
N Extended
18
βOLS Main
βOLS Extended
βIV 100km
βIV 250km
N Main
N Extended
(1)
Potato
(2)
Elevation
(3)
Expulsions
(4)
Persecutions
(5)
BD
Mortality
(6)
Years
Jewish
(7)
Pop >
4,000
(8)
City FE’s
(9)
D-in-D
(levels)
(10)
Cluster
Country
0.0674∗∗∗
(0.0189)
0.134∗∗∗
(0.0150)
0.187∗∗
(0.0915)
0.408∗∗
(0.178)
2860
7377
0.0522∗∗∗
(0.0188)
0.126∗∗∗
(0.0151)
0.119
(0.0961)
0.335∗
(0.187)
2860
7377
0.0693∗∗∗
(0.0201)
0.143∗∗∗
(0.0163)
0.214∗
(0.111)
0.487∗∗
(0.226)
2860
7377
0.0660∗∗∗
(0.0205)
0.144∗∗∗
(0.0167)
0.197∗
(0.119)
0.481∗
(0.246)
2860
7377
0.0566∗∗∗
(0.0193)
0.134∗∗∗
(0.0151)
0.171∗
(0.0990)
0.408∗∗
(0.194)
2860
7377
0.0632∗∗∗
(0.0199)
0.137∗∗∗
(0.0163)
0.205∗
(0.105)
0.501∗∗
(0.235)
2860
7377
0.0744∗∗∗
(0.0194)
0.139∗∗∗
(0.0152)
0.228∗∗
(0.0993)
0.456∗∗
(0.185)
2579
6261
0.124∗∗∗
(0.0272)
0.113∗∗∗
(0.0274)
0.220∗∗∗
(0.0708)
0.251∗∗∗
(0.0937)
2860
7377
0.328∗∗∗
(0.0491)
0.288∗∗∗
(0.0448)
0.552∗∗∗
(0.0800)
0.674∗∗∗
(0.114)
3278
9075
0.0604∗∗
(0.0267)
0.132∗∗∗
(0.0211)
0.185∗
(0.102)
0.418
(0.322)
2860
7377
(11)
Cities
1300
(12)
Cities
1400
(13)
Cities
1500
(14)
Cities
1600
(15)
Cities
1700
(16)
Drop UK
(17)
Drop
France
(18)
Drop
Germany
(19)
Drop Italy
(20)
Drop
Spain
0.0398
(0.0330)
0.0723**
(0.0290)
0.180
(0.146)
0.472*
(0.242)
1094
1435
0.0698***
(0.0236)
0.105***
(0.0195)
0.155
(0.113)
0.346**
(0.176)
1969
3451
0.0667***
(0.0219)
0.103***
(0.0179)
0.120
(0.107)
0.362*
(0.194)
2137
4027
0.0666***
(0.0202)
0.116***
(0.0166)
0.117
(0.0953)
0.329*
(0.170)
2467
4987
0.0620***
(0.0197)
0.123***
(0.0160)
0.154
(0.0941)
0.404**
(0.173)
2639
6003
0.0327*
(0.0197)
0.115***
(0.0154)
0.250**
(0.114)
0.789**
(0.320)
2607
6675
0.0657***
(0.0215)
0.139***
(0.0169)
0.314**
(0.159)
0.959
(0.602)
2305
6092
0.0771***
(0.0222)
0.145***
(0.0182)
0.185**
(0.0916)
0.310**
(0.141)
2190
6174
0.0561***
(0.0217)
0.136***
(0.0171)
0.0994
(0.108)
0.262
(0.191)
2380
5817
0.0602***
(0.0189)
0.129***
(0.0153)
0.137
(0.0940)
0.316*
(0.175)
2482
6349
Notes: In Columns (1)-(5) we interact our year fixed effects with a range of additional time invariant controls: potato
suitability (Col. 1); elevation (Col. 2); expulsions (Col 3.); pogroms or expulsions (Col. 4); Black Death mortality (Col. 5);
Years of Jewish presence (Col. 6). In Column 7 we drop cities with less than 4,000 population; Column 8 reports a diff-in-diff
specification in growth rates with city and year fixed effects; Column 9 reports a diff-in-diff specification in levels. In Column
10 we cluster our standard errors at the modern country level. In Columns 11-15 we include only cities that existed in either
1300, 1400, 1500, 1600, or 1700 respectively. Columns 15-20 drop cities from the largest modern countries. Robust standard
errors in parentheses. ∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01.
In Columns (6)-(10) we experiment with different ways of defining our specification and main
variables. Our primary variable of interest is a dummy variable for Jewish presence. However, in
many cases Jewish communities either arrive or disappear in the middle of a century. Rather
than code these cities as 1, in Column (6) we redefine our treatment variable to be the proportion
of years in the previous period that a Jewish community was in the city. This does not have a
significant effect on our estimates.
If the Encyclopedia Judaica is less likely to include an entry for smaller cities, then it is possible
that our estimates suffer from attenuation bias and will be too small. As such, in Column (7) we
drop any city with less than 4,000 people. This accounts for 25% of the Bairoch data. When we
do this the OLS coefficient using the main sample does increase slightly, but the estimates using
the extended sample and the IV’s are unaffected.
In Columns (8) and (9) we switch from using modern country dummies to city fixed effects.
This results in true difference-in-differences specifications in which we are identifying the effect
of Jewish communities on growth using only the within city variation in Jewish presence and
population changes. Column (8) uses growth as the dependent variable. A discussed above, a
concern with this specification is Nickell bias since we also include the lag of city population as
a control variable. Another potential concern with using just the within city variation is that
Jews may be more likely to move to a city if it grew faster in the previous period. Thus, the
OLS coefficient estimates may be biased upwards. The estimates under OLS do increase in size
for the main sample, but are smaller once we appropriately instrument for Jewish presence. In
Column (9) we use the log level of city population as the dependent variable and do not include
the lag of population as a control. All the coefficients significantly increase in size and retain
their significance under this specification.23 In Column (10) we cluster our standard errors at
the modern country level so as to account for spatial correlation across cities. When we do this
the OLS and the 100km IV estimates retain their size and significance. The 250km IV retains its
size but becomes statistically insignificant.
Our baseline analysis uses an unbalanced panel in which many cities are entering and exiting
over the 750 year period we study. In Columns (11)-(15) restrict our sample to only cities that
23
We investigated why these coefficients increase so much using both non-parametric analysis and by allowing
the coefficient on Jewish Community to vary by century. The increase stems in part because the period of the
Black Death (1347-1352) now contributes a great deal more to the story. Cities that expelled or killed their Jewish
populations grew much faster than those that did not (or could not because they had no Jewish population).
We are currently studying the relationship between Jewish persecution and Black Death mortality rates in more
detail in another paper (Jebwab et al., 2016) in which we find cities experiencing higher mortality rates were less
likely to persecute their Jewish community. The results we find in Section 5 below suggesting that under our
preferred specification the Jewish city growth advantage began after 1600 is also present under the diff-in-diff
specifications.
19
existed in 1300, 1400, 1500, 1600, or 1700 respectively. With the exception of 1300, when the
sample size is very small, the OLS estimates are stable across these samples. In addition the
250km IV is stable over all samples, including 1300. The 100km IV retains its size, but loses
significance, though as the sample size increases from 1300 to 1700 it goes from being very
insignificant (p-value = 0.22 in 1300), to being just barely insignificant (p-value = 0.11 in 1700),
suggesting that much of this is driven by lack of power in the regressions.
In Columns (15)-(20) we systematically drop the cities of the largest modern countries from our
sample. One concern with our result that the emergence of the Jewish city growth premium
only emerges after 1600 may be that it is driven by the precocious growth in Britain in early
years of the Industrial Revolution. However, in Column (16) we drop cities in the UK and our
coefficients are stable (under OLS there is a slight decline in size, under IV a slight increase).
The most notable effect we find is that when we drop Italy the OLS coefficients are robust, but
the IV estimates become statistically insignificant. We do not know why this happens, though
we can speculate that perhaps the active participation of Jews in banking in Italy meant that
the network there mattered more than in other regions of Europe, and that this network effect is
contributing a great deal to the power of our IV estimators (Pascali, 2015). However, it also well
known that Jews were also heavily involved in banking activities in the Holy Roman Empire and
the IV coefficients are robust to dropping modern German cities.
5
The Relationship between Jewish Communities and City
Growth: Timing and Mechanisms
5.1 Flexible Estimates
The regressions we report in Tables 1 and 2 suggest that cities with Jewish communities grew
faster on average between 1100 and 1850. However, we would like to know if this growth
advantage varied over time as this will help us identify possible explanations for it. We therefore
estimate a flexible version of Specification 1 in which we allow the estimated coefficient on Jewish
Community βt to vary by time period.24
In Table 4 we report the effect of a city possessing a Jewish Community in each year of the sample
based on our regression results. The 12th century is always the omitted period. Columns (1), (2),
and (3) report OLS estimates using the main sample. With the exception of the coefficient on the
13th century when we don’t include any controls, these regressions tell a consistent story. There
was no relationship between Jewish presence and city growth for the entire period up until 1750
P1850
The regression specification we use is: PopGrowthi,c,t = t=1200 βt Jewish Communityi,c,t + γlnpopi,c,t−1 +
P1850
0
t=1200 Xi,c µt + δc + ηt + λc × Year + εi,c,t where all variables are defined the same as in equation 1.
24
20
at which time Jewish city growth diverges sharply from the rest of the sample. The coefficients
for the 17th century and the period from 1700 to 1750 are positive but imprecisely estimated.
When we estimate the regressions using 2SLS and our Jewish network IV’s a slightly different
story emerges. The 100k IV suggests that Jewish city divergence begins in the 17th century
and, consistent with the IV coefficients in the reduced form regressions, the size of the Jewish
city growth advantage is larger than under OLS. The 250k IV yields a similar story, but now
divergence appears in the 16th century and the coefficients are even bigger. Taken together, the
flexible IV estimates suggest divergence around 1600 or so with the Jewish city growth advantage
ranging from 32% to 230% depending on the period. One puzzling result from the IV regressions
is that the coefficient for the period 1700-1750 is inconsistent with those preceding and following
it. After having checked the data for outliers or obvious explanations we are uncertain why this
is the case.
In Columns (6)-(8) we report the coefficients from running the flexible specification using the
extended sample of all cities in the Bairoch database. These regressions tell a very clear story.
Cities with Jews always possessed a growth premium over non-Jewish cities of between 5% and
10% (though this is likely to have been to selection). Beginning in the 17th century, however, this
growth premium steadily increases from about 10% to about 25% in 1850. The period 1700-1750
is, again, an outlier, though in this case it exhibits a positive and statistically significant growth
advantage for Jewish cities.
Figures 2 and 3 plot the coefficients along with their 95% confidence intervals for each year using
the results from Columns (3) and (8) of Table 4. Overall, two facts emerge from the flexible
regressions. First, before 1600 there is little evidence for a Jewish growth premium. Second,
depending on the sample and estimator used there appears to have been a divergence between
the growth rates of Jewish and non-Jewish cities sometime between 1600 and 1750 which persists
up until the end of the sample in 1850. These results are extremely consistent with historical
accounts which view the period between 1300 and 1600 as one of crisis and decline for Europe’s
Jewish population whereas the period after 1600 was one of economic and demographic expansion
(Braudel, 1949; Israel, 1985).
Moreover, these findings appear inconsistent with a pure human capital story as Jews had higher
human capital than Christians throughout the medieval and early modern period. Rather it
suggests that something else changed after around 1600-1700 that made the human capital
and skills of Jews more complementary to economic growth. We turn to explore some of these
mechanisms in the next section.
21
Table 4: Jewish communities and City Growth: Flexible Regressions
Table 4: Jewish Communities and City Growth, Flexible Regressions, 1100-1850
Dependent Variable: Log City Growth
Main Sample
Extended Sample
22
(1)
OLS
(2)
OLS
(3)
OLS
(4)
IV100k
(5)
IV250k
(6)
OLS
(7)
OLS
(8)
OLS
Jewish Community X 1300
-0.3369*
(0.2024)
-0.1532
(0.1987)
-0.0567
(0.2143)
-3.2141
(11.7598)
-1.4142
(2.5892)
0.0559
(0.0919)
0.1055
(0.0964)
0.1437
(0.0986)
Jewish Community X 1400
-0.0071
(0.0828)
-0.0139
(0.0770)
0.0026
(0.0808)
0.0353
(0.5126)
0.2277
(0.4131)
0.0763*
(0.0446)
0.0613
(0.0470)
0.0522
(0.0487)
Jewish Community X 1500
0.0119
(0.0645)
0.0029
(0.0686)
0.0673
(0.0688)
-0.0746
(0.3540)
0.0924
(0.2492)
0.0061
(0.0423)
0.0574
(0.0450)
0.0637
(0.0457)
Jewish Community X 1600
-0.0199
(0.0456)
-0.0215
(0.0456)
-0.0051
(0.0462)
0.3513
(0.2626)
0.9643***
(0.3330)
0.0631*
(0.0372)
0.0694*
(0.0384)
0.0607
(0.0381)
Jewish Community X 1700
-0.0063
(0.0639)
0.0438
(0.0621)
0.0448
(0.0600)
0.8775***
(0.2466)
2.3432***
(0.8897)
0.0936
(0.0589)
0.1689***
(0.0570)
0.1564***
(0.0565)
Jewish Community X 1750
0.0470
(0.0356)
0.0528
(0.0374)
0.0272
(0.0379)
-0.0066
(0.1361)
0.2373
(0.3451)
0.0701**
(0.0320)
0.0893***
(0.0332)
0.0735**
(0.0327)
Jewish Community X 1800
0.1232***
(0.0327)
0.1603***
(0.0347)
0.1165***
(0.0343)
0.3239**
(0.1364)
0.6837*
(0.3616)
0.1589***
(0.0280)
0.1894***
(0.0290)
0.1684***
(0.0278)
Jewish Community X 1850
0.2157***
(0.0367)
0.2297***
(0.0430)
0.1431***
(0.0396)
0.5910***
(0.1967)
1.2091**
(0.5098)
0.2735***
(0.0296)
0.2903**
(0.0323)
0.2481***
(0.0295)
Yes
No
No
2860
Yes
Yes
No
2860
Yes
Yes
Yes
2860
Yes
Yes
Yes
2860
Yes
Yes
Yes
2860
Yes
No
No
2860
Yes
Yes
No
2860
Yes
Yes
Yes
2860
Year FE's
Controls X Year FE's
Country FE's X Year
N
Notes: This table presents our flexible specification where we allow the effect of a Jewish community on city growth to vary
over time. Columns 1-5 focus on our main sample. Columns 1-3 present our OLS estimates by century (the omitted century is
the 13th century). Column 1 employs year fixed effects; Column 2 interacts year fixed effects with our controls; Column 3
include modern country fixed interacted with the year. Columns 4-5 use our preferred instruments for Jewish presence in
conjunction with fixed effects interacted with our controls and with modern country fixed effects. Columns 6-7 repeat our
OLS estimates for the extended sample. * p < 0.10, ** p < 0.05, *** p < 0.01.
.3
.4
.2
0
.1
E(Jewish City Growth | X)
.2
0
-.2
E(Jewish City Growth | X)
-.1
-.4
1300
1400
1500
1600
1700
1800
1300
Year
Jewish City Growth
1400
1500
1600
1700
1800
Year
95% CI
Jewish City Growth
Figure 2: The effect of a Jewish community on city growth over time. This
Figure plots coefficients obtain from Table 3, (Col. 3.) for the main sample.
95% CI
Figure 3: The effect of a Jewish community on city growth over time. This
Figure plots coefficients obtain from Table 3, (Col. 8) for the extended sample.
5.2 Mechanisms
We now consider some of the mechanisms linking the presence of a Jewish community with
more rapid economic growth. We identify four potentially important factors which could explain
the Jewish city growth advantage: Whether the Jewish community was Sephardic, whether the
community had a Jewish-run printing press, whether it was a ‘Port Jew’ community, and whether
there was a relationship between the density of markets and the performance of Jewish cities.
We run regressions where we interact proxies for each of these mechanisms with our main Jewish
city dummy variable. We report these regressions in Table 5. In what follows, we discuss each of
these potential mechanisms in greater detail and interpret the results of the regressions.
Sephardic Communities A sizable historical literature associates the expansion of Jewish
trade and economic activity after 1600 with the Sephardic Jewish communities who migrated
from Spain and Portugal after 1492 and settled across Europe in cities such as Amsterdam,
Bordeaux, Hamburg, Livorno and London during the 16th and 17th centuries (see Braudel, 1949;
Israel, 1985; Trivellato, 2009).
Israel (2005, 11) argues that the Sephardic diaspora created ‘a new phenomenon . . . a new type
of Jewish commercial system’ that was based not on local markets, or on trade in agricultural
products but on the transportation of luxury goods over long distances. Similarly, Trivellato
(2009) argues that the Sephardic diaspora was remarkable for its ‘geographical breadth’ and
23
‘stability’ and that [‘o]ther branches of the Jewish diaspora could not count on the same
geographical dispersion or interconnectedness’ (Trivellato, 2009, 149).25 They formed a network
of interconnected merchants tied together ‘on the basis of implicit contracts with blood-kin and
in-law’. Sephardic merchants formed long-lasting partnerships and employed long distance agency
relationships relying on both formal courts and on reputation-based mechanisms of the kind
that Greif (1992, 2006) studied in the context of the Geniza records. Community organizations
strove to uphold the collective reputation of local merchants, excommunicating members found
guilty of trading in counterfeit coins or goods or acting in such a way that would “discredit the
commerce of the Jewish nation” (Trivellato, 2009, 166).
Sephardic merchants were involved in a variety of mercantile activities. The Sephardic or
Portuguese community in Amsterdam was heavily involved in the silk trade (until it was closed
to them in the 1650s), in sugar, and in the trade with both the Levant and with the Portuguese
colonies in the Americans and in Asia (see Bloom, 1936). They were a significant presence in
the colonial trade between the British West Indies and the England from the 1650s onwards (?).
Sephardic Jews came to play a similarly important role elsewhere in Europe, in the Venetian
economy, for example, where they imported Spanish wool and Spanish American dyestuff for the
Italian textiles industry (Fusaro, 2015, 261). Livorno, in particular, grew in importance as an
entrepôt for trade with the Levant; it was the fastest growing port in Italy in the 17th century
(Trivellato, 2009, 71).
We create a time varying dummy variable equal to one if a Jewish community was known to
be Sephardic based on the information contained in Beinart (1992) and checked with Roth and
Wigoder (2007). We report the results of running these regressions in Table 5 Columns (1) and
(2).26 Regardless or whether we control for modern country dummies, the coefficients are large
and statistically significant. The estimates suggest that cities with Sephardic communities grew
about 18% or 19% more quickly than non-Sephardic communities (which could be either Jewish
and non-Sephardic or have no Jewish presence whatsoever).
25
There was a widespread perception in Amsterdam and in other cities such as Hamburg that the Sephardic
Jews were significantly more prosperous and entrepreneurial than were Ashkenazi Jews. For example, Bloom
writes: ‘Unlike their Sephardic brethren the Ashkenazic Jews, because of different background and tradition, were
not concerned with secular matters but were deeply engrossed in the study of the Talmud.’ He notes that though
‘the Ashkenazic community by dint of sober industry and thrift had acquired a certain degree of prosperity . . . It
is self evident that, as compared with the Sephardim, the Ashkenazic group was poor indeed” (Bloom, 1936).
26
For our regressions on Sephardic Community and Port City we report only the coefficient on the direct effect
of Sephardic or Port on city growth since these dummies are perfectly collinear with their interaction with Jewish
Community (All Sephardic or Port Jew cities are, by definition, also Jewish). As explained in the text, this is not
the case with Hebrew Press.
24
Hebrew Printing Press To explore evidence of a cultural mechanism linking the presence of
a Jewish community to economic growth we collected data on the existence of Jewish printing
presses based on information contained in Beinart (1992) and ?. The historical literature suggests
that the existence of a Hebrew printing was a measure of cultural interaction between Jews
and Christians (see Burnett, 1998). A Hebrew printing press either meant there was a Jewish
community in the city had the status and freedom to print their own religious literature or it
signaled the presence of Christian Hebraists who were interested in studying Hebrew literature.27
We create a time varying dummy variable equal to one if a city had a Hebrew Printing press
in that year. In Table Mechanisms Columns (3) and (4) we report the effect of a city having a
Hebrew Printing press on population growth. The coefficients suggest a growth premium similar
to Sephardic cities. We interpret this large effect as being consistent with evidence for a cultural
transmission channel and with the view that the benefits of Jewish presence were greatest where
they were able to interact with their Christian neighbors.
Port Cities and Port Jews We also consider the interaction between the existence of a
Jewish community and a dummy for whether or not that city is on the coast in order to test
whether communities of so-called Port Jews had a greater impact on city growth. Historians refer
to ‘port Jews’ to describe those maritime Jewish communities that flourished in early modern
cities such as Livorno, Trieste, Amsterdam, and London (Cesarani, 2001; Dubin, 2001, 2006;
Sorkin, 2001). These communities tended to offer great religious freedom to Jewish communities.
The estimates in Columns (5) and (6) suggest no difference in growth rates between cities with
Jewish communities that were coastal and those that were not.
Market Access The historical literature points to the importance of Jewish trading and
financial networks. But, while economic historians have conducted numerous studies of market
integration during the early modern period, with a few exceptions these have focused on the grain
trade with little systematic study of other markets due to data limitations.28 Jewish merchants
in medieval and early modern Europe, however, did not play a prominent role in the grain trade
but, rather, were involved in the transport of diamonds, sugar, silks, tobacco, and other luxury
products in addition to playing a large role in banking and finance. Therefore, rather than
looking at grain markets, we explore a more general measure of market integration based on
market access.
27
Thus there are some cities which acquired Hebrew printing presses but did not have Jewish communities
such as London prior to 1655.
28
Studies include (Bateman, 2011; Chilosi et al., 2013; Clark, 2015). One study of financial market integration
in premodern Europe is Boerner and Volckart (2011).
25
Table 5: Mechanism regressions
Dependent Variable: Log City Growth
Baseline Sample
Sephardic
(1)
(2)
(3)
(4)
Sephardic
City
Sephardic
City
Hebrew
Press
Hebrew
Press
(5)
(6)
Port Jewish Port Jewish
City
City
(7)
(8)
Market
Access
Market
Access
0.1772*** 0.1900***
(0.0410)
(0.0373)
HebrewPress
0.2426*** 0.2182***
(0.0540)
(0.0500)
26
PortCity
0.0000
(0.0626)
-0.0953
(0.0628)
MarketAccess Non-Jewish
0.0061*** 0.0062***
(0.0022)
(0.0020)
MarketAccess Jewish
0.0141*** 0.0087***
(0.0030)
(0.0028)
Year FE's
Controls X Year FE's
Country FE's X Year
N
R-sq
Yes
Yes
No
2860
0.215
Yes
Yes
Yes
2860
0.271
Yes
Yes
No
2860
0.216
Yes
Yes
Yes
2860
0.271
Yes
Yes
No
2860
0.272
Yes
Yes
Yes
2860
0.271
Yes
Yes
No
2860
0.218
Yes
Yes
Yes
2860
0.267
Notes This table reports the coefficient on the interaction term for each of our mechanism regressions. Robust Standard
errors in parentheses. * p < 0.10, ** p < 0.05, *** p < 0.01.
Studying market access has a long tradition in economic geography and trade with one of the
first modern studies done by Harris (1954). The market access approach emphasizes that a
region’s level of economic development should be positively related to the density of development
surrounding it. This harkens back to Marshall’s three centripetal forces that lead to increasing
returns and, as a result, urban agglomeration: (1) forward and backwards linkages, (2) thick
markets, and (3) knowledge spillovers (Marshall, 1890; Fujita et al., 2001). The concept of
market access plays a large role in a growing literature that focuses on the welfare effects of
extended network of markets surrounding a region and how this interacts with transportation and
other transaction costs (see, e.g., Eaton and Kortum (2002); Donaldson and Hornbeck (2016);
Storeygard (2016); Donaldson (2016)).
Our measure of market access is virtually identical to our calculation of Jewish Network Access
in Equation 2, however, instead weighting the index by a dummy for Jewish Community, we
weight by the populations of surrounding cities. We calculate market access for city j as
P
MAjt = i6=j Nit τji−σ , where Nit is simply the time-varying measure of city i’s population from
the Bairoch dataset.29
There are two potential ways that market access could differentially impact the growth of cities
with Jewish communities. First, density of potential economic activity could increase more
surrounding Jewish communities than for other cities. This could occur for reasons having nothing
to do with Jewish presence, but nonetheless, cities that happen to have a Jewish community
might benefit from it. We say this sort of growth stems from Jews benefiting from the extensive
margin of market access. Alternatively, there could be something about Jewish communities
that makes them, intrinsically, more able to take advantage of market potential, regardless of
the level of market access for their city. We say this sort of growth stems from cities with Jewish
communities benefiting on the intensive margin of access.
In Figure 4 we plot the average value for market access for Jewish and non-Jewish cities over our
study period. Two facts are made clear from this figure. First, there is no difference in average
market access between Jewish and non-Jewish cities for the entire period. If anything, non-Jewish
cities experienced greater market density after 1700. This suggests that a growth advantage
on the extensive market access margin cannot explain the Jewish city growth advantage we
identify. The second fact evident from Figure 4 is the increasing value of market access for all
cities. Between 1300 and 1600 market access goes from about 5 to 10 on average. Between 1600
and 1850, however, it increases on average from about 10 to 25. In other words, market access
increases for all cities at an increasing rate after 1600 or so. This opens the possibility that the
29
Consistent with our discussion of the Jewish Network Variable in Section 4 we set σ = 1.
27
25
Market Access for Jewish and Non-Jewish Cities
10
15
20
5
1300
1400
1500
Year
1600
Jewish City MA
1700
1800
Non-Jewish City MA
Figure 4
Jewish city growth advantage may have stemmed from their ability to better take advantage of
this growth in market density.
We test this possibility in Columns (7) and (8) of Table 5 by interacting the Jewish Community
dummy variable with our measure of market access. We then report both the marginal effect of
market access for both non-Jewish and Jewish cities. In Column (7) when we don’t control for
modern country fixed effects, the estimates suggest that cities with Jewish communities were
more than twice as good at translating gains in market access into growth than were cities that
did not have Jewish communities. When we focus on just the within modern country variation
in Column (8) this estimate shrinks, but city with Jewish communities still growth substantially
faster.
These findings are consistent with the argument made by numerous historians that Jewish trading
and finance networks help to knit together the European economy, particularly in the period 1650
to 1800 (Israel, 1985). Israel, for example, notes that the importance of Jewish merchants and
trade lay not in ‘any important innovations’ or in a particularly distinctive capitalistic outlook as
Sombart maintained: ‘the techniques of Jews commerce and finance differed not a jot from other
commerce and finance (Israel, 1985, 222). Rather the distinctive and important characteristic of
Jewish merchants was their access to a wide network of merchants and financiers.30 Examining
30
For instance: ‘The key factor which imparted a certain import to the post-1570 Jewish role’ he writes ‘was
28
.2
Difference in Growth Predicted by Market Access
.05
.1
.15
0
1300
1400
1500
Year
Jewish vs Non-Jewish
1600
1700
1800
Jewish vs Non-Jewish (w/in country)
Figure 5: Predicted growth difference due to increases in market access between cities with
Jewish communities and cities without Jewish communities.
the letters of two Sephardic Jewish merchants from Livorno, Trivellato (2009) found of the
nearly 14,000 letters exchanged, a considerable proportion involved Amsterdam, London, Aleppo,
Marseilles, Lisbon as well as Venice, Genoa and Florence (Trivellato, 2009, 195-196). Our analysis
substantiates this qualitative evidence: Jewish merchants had access to a commercial network
that extended beyond Europe and allowed them to transport information and resources across
long distances.
In Figure 5 we put together what we know about the growth in market access for Jewish and
non-Jewish cities on the extensive margin with what the estimates in Table 5 suggest about the
ability of Jewish cities to take advantage of the intensive margin. We multiply the yearly values
of market access by the coefficients in Table 5 to arrive at the predicted city growth stemming
from increases in market access for both groups. Then we plot the difference in these city growths
in the Figure. The lower line shows predicted gains in growth based on the estimates in Column
(8) which use only within country variation. These suggest that gains in market access can
account for about a 5% difference in Jewish city growth by the end of the period. When we use
the simultaneous penetration during the sixteenth century of both Ashkenazi and Sephardi Jews, as well as of the
Marranos living in Portugal and the Portuguese empire, into maritime and overland long-distance transit trades
linking the Levant to Italy, Poland with the Levant, Poland with Germany, and Portugal and the Portuguese
empire with northern Europe . . . This entrenched position in so many crucial but distant markets proved a factor
of great potency’ (Israel, 1985, 222-223).
29
the estimates in Column (7) we get much higher predicted growth differences. Our calculations
suggest that the Jewish city ability to take advantage of market density may have accounted for
an increase in population growth from 10% in 1600 to almost 20% in 1850.
6
Conclusion
This paper studies the relationship between the presence of a Jewish community in a city and
that city’s population growth in pre-industrial Europe. We find cities with Jewish communities
grew faster in the preindustrial period by between 5% to 15%. We develop an IV strategy based
on modeling the spatial network of Jewish communities. These IV estimates suggests that the
presence of a Jewish community indeed had a causal impact on subsequent city growth.
Our analysis of a flexible specification suggests that the Jewish city growth advantage is driven
by the post-1600 period. There is little discernible impact of a Jewish community on city growth
in the middle ages. This analysis is not consistent with a simple human capital story. Jews had
higher human capital than their Christian counterparts throughout the middle ages but this did
not result in notably faster economic growth. One reason for this was that, in the middle ages
these skills were exploited by political elites who, for example, often licensed and taxed Jewish
moneylending (Koyama, 2010). The net effect was that the presence of a Jewish community
did not translate into economic growth—at least as measured by city growth—in the medieval
period.
This story changed in the post-1600 period. After this date we do find a growth effect associated
with the presence of a Jewish community. In investigating potential channels linking the presence
of a Jewish community to city growth we find indirect evidence that this effect was driven by
Jewish merchant networks. This result provides support for the accounts of historians which
have emphasized the important role played by Jewish traders in 17th and 18th century Europe
(such as ?Israel, 1985; Trivellato, 2009).
Our research setting has a number of advantages. It enables us to study the economic consequences
of religious toleration at a disaggregated level over a long span of time. However, we also face
a number of limitations imposed on us by the nature of the data available and it is important
to acknowledge these caveats. Not all Jews in medieval and early modern Europe were traders
and merchants and the occupational distribution of Jews likely varied from place to place. This
was particularly true after 1500 as Jewish communities moved eastwards (having being expelled
from much of western Europe), they also became more economically diversified.31 In Eastern
Europe they were the predominant ‘capitalist class’ (Katz, 2000, 45). But they were not solely
31
See Katz (2000, 38-62).
30
concentrated in moneylending and frequently worked in a range of occupations (e.g. Penslar,
2001, 19). Compared to Western Europe, the division of labor was less extensive and there were
fewer possibilities for specialization so they worked as peddlers, landlords, artisans, inn keepers
and tax collectors. The fact that their economic role varied over time and across space suggests
that the presence of the Jews in a community need not have the same effect in all places and at
all times. Data limitations mean that we cannot assess the actual occupations of the Jews in
each city in our database. Nor are there systematic data on the size of most Jewish communities
in the middle ages.
It is also important to note that we focus on the preindustrial period as this was when economic
development was largely driven by Smithian growth—trade and the expansion of the market—and
that we do not consider the period of sustained modern economic growth after 1850. Historians
have shown that Jewish traders and bankers played a significant role in the commercial economy
of the medieval and early modern period, but they did not play an important role in developing
the technologies which are most associated with the Industrial Revolution.32 Nonetheless, an
implication of our research is that minorities like Armenians, Quakers and the overseas Chinese
may play an important role in market formation and in a development context.
32
Bairoch (1999) conducted a preliminary investigation of this topic. He studied several sources including the
Biographical Dictionary of the History of Technology and The Timetables of Technology. In the former he was
able to identify 57 out 2,160 or 2.6% of inventors with Jewish names and in the later he could only find 11 (1.2%)
certainly Jewish and 40 (4.4%) possibly Jewish inventors. This led him to conclude that during the industrial
revolution era: ‘the share of the Jews contributing to technological innovations is lower than their share in the
total population . . . not only were there few Jews among the innovators of technology nor were they outstanding
in their accomplishments’ (Bairoch, 1999, 132–133). For a discussion of why Jews did not play an important role
in the First Industrial Revolution see Rubinstein (1999).
31
References
Abramitzky, R. and H. Halaburda (2014). Were jews in interwar poland more educated? memo.
Acemoglu, D., T. A. Hassan, and J. A. Robinson (2011). Social structure and development: A legacy of
the Holocaust in Russia. The Quarterly Journal of Economics 126 (2), 895–946.
Acemoglu, D., S. Johnson, and J. Robinson (2005, June). The rise of Europe: Atlantic trade, institutional
change, and economic growth. American Economic Review 95 (3), 546–579.
Alesina, A. and P. Giuliano (2014, September). Culture and institutions. Working Paper.
Anderson, R. W., N. D. Johnson, and M. Koyama (2015). Jewish persecutions and weather shocks,
1100-1800. Economic Journal .
Arbath, C. E. and G. Gokmen (2015, March). Minorities and long-run development: Persistence of
Armenian and Greek influence in turkey. memo.
Ashraf, Q. and O. Galor (2011, August). Dynamics and stagnation in the Malthusian epoch. American
Economic Review 101 (5), 2003–41.
Attali, J. (2010). The Economic History of the Jewish People. Portland: ESKA Publishing Washington.
Bairoch, P. (1988). Cites and Economic Development, from the dawn of history to the present. Chicago:
University of Chicago Press. translated by Christopher Braider.
Bairoch, P. (1999). The jews, the english industrial revolution, technological innovation and the sciences.
In E. Brezis and P. Temin (Eds.), Elites, Minorities and Economic Growth, Amsterdam, pp. 115–128.
Elsevier.
Baron, S. W. (1975). Economic History of the Jews. Jerusalem: Keter Publishing House. edited by
Nachum Gross.
Barro, R. J. (2012, August). Convergence and Modernization Revisited. NBER Working Papers 18295,
National Bureau of Economic Research, Inc.
Bateman, V. N. (2011). The evolution of markets in early modern Europe, 1350–1800: a study of wheat
prices. The Economic History Review 64 (2), 447–471.
Becker, S. O., E. Hornung, and L. Woessmann (2011, July). Education and Catch-Up in the Industrial
Revolution. American Economic Journal: Macroeconomics 3 (3), 92–126.
Becker, S. O. and L. Woessmann (2009, May). Was Weber wrong? a human capital theory of Protestant
economic history. The Quarterly Journal of Economics 124 (2), 531–596.
Beinart, H. (1992). Atlas of Medieval Jewish History. London: Hendrickson Publishers.
Bell, D. P. (2008). Jews in the Early Modern World. Lanham: Rowman & Littlefield Publishers INC.
Bloom, H. I. (1936). The Economic Activities of the Jews of Amsterdam in the Seventeenth and Eighteenth
Centuries. Port Washginton: Kennikat Press.
Boerner, L. and O. Volckart (2011, January). The utility of a common coinage: Currency unions
and the integration of money markets in late Medieval Central Europe. Explorations in Economic
History 48 (1), 53–65.
Bosker, M., E. Buringh, and J. L. van Zanden (2013). From Baghdad to London: unravelling urban
development in Europe and the Arab world 800-1800. Review of Economics and Statistics Forthcoming.
Botticini, M. (1997). New evidence on Jews in Tuscany, CA. 1310–1435: The ‘friends & family’ connection
again. Zakhor. Rivista di Storia degli Ebrei d’Italia 1, 77–93.
Botticini, M. and Z. Eckstein (2012). The Chosen Few. Princeton, New Jersey: Princeton University
Press.
Braudel, F. (1972, 1949). The Mediterranean, and the Mediterranean world in the age of Phillip II, Vol.
II. Berkeley: University of California Press. Translated by Sian Reynolds.
Braudel, F. (1979, 1982). The Wheels of Commerce. London: Phoenix Press.
Burnett, S. G. (1998). The Regulation of Hebrew Printing in Germany, 1555-1630: Confessional Politics
and the Limits of Jewish Toleration. In M. Reinhart and T. Robisheaux (Eds.), Infinite Boundaries:
Order, Disorder, and Reorder in Early Modern German Culture, Kirksville, MO, pp. 329–348. Truman
32
State University Press.
Cantoni, D. (2015). The Economic Effects of the Protestant Reformation: Testing the Weber Hypothesis
in the German Land. Journal of the European Economic Association Forthcomng.
Cesarani, D. (2001). Port Jews: Concepts, cases and questions. Jewish Culture and History 4 (2), 1–11.
Chazan, R. (2006). The Jews of Western Christendom. Cambridge: Cambridge University Press.
Chazan, R. (2010). Reassessing Jewish Life in Medieval Europe. Cambridge: Cambridge University
Press.
Chilosi, D., T. E. Murphy, R. Studer, and A. C. TunâĹŽßer (2013). Europe’s many integrations:
Geography and grain markets, 1620–1913. Explorations in Economic History 50 (1), 46 – 68.
Clark, G. (2015, september). Markets before economic growth: the grain market of medieval England.
Cliometrica, Journal of Historical Economics and Econometric History 9 (3), 265–287.
Cohen, M. (1994). Under Crescent and Cross. Princeton, New Jersey: Princeton University Press.
Davis, N. Z. (1997). Religion and Capitalism Once Again? Jewish Merchant Culture in the Seventeenth
Century. Representations (59), 56–84.
De Long, J. B. and A. Shleifer (1993, October). Princes and merchants: European city growth before
the industrial revolution. Journal of Law & Economics 36 (2), 671–702.
de Vries, J. (1976). Economy of Europe in an Age of Crisis. Cambridge, U.K: Cambridge UP.
Diaz-Cayeros, A. and S. Jha (2014, January). Can contract failures foster ethnic assimilation? evidence
from cochineal in Mexico. memo.
Dincecco, M. and M. G. Onorato (2015, January). Military conflict and the rise of urban Europe.
Dittmar, J. (2011a, August). Cities, markets, and growth: The emergence of Zipf’s Law.
Dittmar, J. E. (2011b). Information technology and economic change: The impact of the printing press.
The Quarterly Journal of Economics 126 (3), 1133–1172.
Doepke, M. and F. Zilibotti (2008, 05). Occupational choice and the spirit of capitalism. The Quarterly
Journal of Economics 123 (2), 747–793.
Donaldson, D. (2016). Railroads of the Raj: Estimating the impact of transportation infrastructure.
American Economic Review .
Donaldson, D. and R. Hornbeck (2016). Railroads and american economic growth: A “mark access”
approach. Quarterly Journal of Economics Forthcoming.
Dubin, L. (2001). Researching Port Jews and Port Jewries: Triest and Beyond. Jewish Culture and
History 4 (2), 47–58.
Dubin, L. (2006). Introduction: Port Jews in the Atlantic world Jewish History. Jewish History 20,
117–127.
Duranton, G. and D. Puga (2013, May). The growth of cities. Memo.
Eaton, J. and S. Kortum (2002). Technology, geography, and trade. Econometrica 70 (5), 1741–1779.
Foa, A. (2000). The Jews of Europe after the Black Death. Berkeley: University of California Press.
Fujita, M., P. R. Krugman, and A. Venables (2001). The spatial economy: Cities, regions, and
international trade. MIT press.
Fusaro, M. (2015). Political Economies of Empire in Early Modern Mediterranean. Cambridge: Cambridge
University Press.
Galor, O. and D. N. Weil (2000). Population, technology, and growth: From Malthusian stagnation to
the demographic transition and beyond. American Economic Review 90, 806–828.
Gennaioli, N., R. LaPorta, F. L. de Silanes, and A. Shleifer (2013). Human capital and regional
development. Quarterly Journal of Economics 128, 105–164.
Goldscheider, C. and A. S. Zuckerman (1984). The Transformation of the Jews. Chicago: University of
Chicago Press.
Greif, A. (1992, May). Institutions and international trade: Lessons from the commercial revolution.
American Economic Review 82 (2), 128–33.
Greif, A. (2006). Institutions and the Path to the Modern Economy. Cambridge, U.K.: Cambridge
33
University Press.
Guiso, L., P. Sapienza, and L. Zingales (2006). Does culture affect economic outcomes? The Journal of
Economic Perspectives 20 (2), 23–48.
Harris, C. D. (1954). The, market as a factor in the localization of industry in the United States. Annals
of the association of American geographers 44 (4), 315–348.
Heckscher, E. F. (1955). Mercantilism, Volume II. London: George Allen & Unwin LTD. translated by
E.F. Soderlund.
Hornung, E. (2014, January). Immigration and the Diffusion of Technology: The Huguenot Diaspora in
Prussia. American Economic Review 104 (1), 84–122.
Hsia, R. P.-C. and H. Lehmann (Eds.) (1995). In and Out of the Ghetto: Jewish-Gentile Relations in
Late Medieval and Early Modern Germany, Cambridge. Cambridge University Press.
Hunderet, G. D. (1987). The role of the Jews in commerce in early modern Poland-Lithuania. The
Journal of European Economic History 16 (2), 245–277.
Israel, J. (1985). European Jewry in the Age of Mercantilism, 1550–1750. Oxford: Oxford University
Press.
Israel, J. (2005). Disporas jewish and non-jewish in the world maritime empires. In I. B. McCabe,
G. Harlaftis, and I. P. Minoglou (Eds.), Dispora Entrepreneurial Networks: Four Centuries of History,
New York, pp. 3–27. Berg.
Israel, J. I. (1983). Central European Jewry during the Thirty Years’ War. Central European History 16 (1),
3–30.
Jarvis, A., H. Reuter, A. Nelson, and E. Guevara (2008). Hole-filled seamless srtm data v4. International
Centre for Tropical Agriculture (CIAT).
Jebwab, R., N. D. Johnson, and M. Koyama (2015, September). Bones, bacteria, and break points: The
heterogeneous effects of the Black Death and long-run growth. memo.
Jebwab, R., N. D. Johnson, and M. Koyama (2016). The Black Death, labor scarcity and Jewish
persecutions.
Jha, S. (2013). Trade, complementaries and religious tolerance: evidence from India. American Journal
of Political Science 107 (4).
Kahan, A. (1986). Essays in Jewish Social and Economic History. Chicago: University of Chicago Press.
Katz, J. (1974). Out of the Ghetto. Cambridge, MA.: Harvard University Press.
Katz, J. (2000). Tradition and Crisis. New York: Syracuse University Press. translated by Bermard
Dov Cooperman.
Katz, L. F. and C. Goldin (2008). The Race Between Education and Technology. Cambridge, MA.:
Harvard University Press.
Kelly, M., J. Mokyr, and C. ï. Grda (2014, 08). Precocious Albion: A New Interpretation of the
British Industrial Revolution. Annual Review of Economics 6 (1), 363–389.
Kopsidis, M. and N. Wolf (2012). Agricultural productivity across Prussia during the industrial revolution:
a Thünen perspective. The Journal of Economic History 72 (03), 634–670.
Koyama, M. (2010, December). The political economy of expulsion: The regulation of Jewish moneylending in medieval England. Constitutional Political Economy 21 (4), 374–406.
Kuznets, S. (1960). Economic structure and life of the Jews. In L. Finkelstein (Ed.), The Jews: Their
History, Culture, and Religion, Philadelphia, pp. 1597–1666. Jewish Publication Society of America.
Lehmann, H. (1995). The Jewish minority and the Christian majority in early modern central Europe.
In R. P.-C. Hsia and H. Lehmann (Eds.), In and Out of the Ghetto: Jewish-Gentile Relations in Late
Medieval and Early Modern Germany, Cambridge, pp. 305–311. Cambridge University Press.
Maddison, A.
Mankiw, N. G., D. Romer, and D. Weil (1992). A contribution to the empirics of economic growth.
Quarterly Journal of Economics 107 (2), 407–437.
Marshall, A. (1890). Principles of Economics. Macmillan.
34
McCabe, I. B., G. Harlaftis, and I. P. Minoglou (Eds.) (2005). Dispora Entrepreneurial Networks: Four
Centuries of History, New York. Berg.
McCloskey, D. N. (2010). Bourgeois Dignity: Why Economics Can’t Explain the Modern World. Chicago:
University of Chicago Press.
Mitch, D. (1999). The role of skill and human capital in the “British” industrial revolution. In J. Mokyr
(Ed.), The British Industrial Revolution: An Economic Perspective, Boulder, Colorado, pp. 241–279.
Westview Press.
Mundill, R. R. (2002). Christian and Jewish lending patterns and financial dealings during the twelfth
and thireenth centuries. In P. Schofield and N. J. Mayhew (Eds.), Credit and Debt in Medieval England
c. 1180–c.1350, Oxford, pp. 42–67. Oxbow Books.
Nickell, S. (1981). Biases in dynamic models with fixed effects. Econometrica: Journal of the Econometric
Society, 1417–1426.
Nunn, N. and D. Puga (2012, February). Ruggedness: The blessing of bad geography in Africa. The
Review of Economics and Statistics 94 (1), 20–36.
Nunn, N. and N. 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 (2012). Euratlas.
Parkes, J. (1976). The Jew in the Medieval Community. Hermon Press.
Pascali, L. (2015). Banks and development: Jewish communities in the Italian Renaissance and current
economic performance. Review of Economics and Statistics Forthcoming.
Penslar, D. J. (2001). Shylock’s Children. Berkeley: University of California Press.
Poliakov, L. (1977, 1965). Jewish Bankers and the Holy See. London: Routledge & Kegan Paul.
translated by Miriam Kochan.
Rosenthal, J.-L. and R. B. Wong (2011). Before and beyond divergence: the politics of economic change
in China and Europe. Cambridge, MA: Harvard University Press.
Roth, C. (1950). Genoese Jews in the Thirteenth Century. Speculum 25 (2), 190–197.
Roth, C. (1953). The qualification of Jewish physicians in the middle ages. Speculum 28 (4), pp. 834–843.
Roth, C. (1961). The economic history of the Jews. The Economic History Review 14, 131–135.
Roth, C. and G. Wigoder (Eds.) (2007). Encyclopedia Judaica (2nd edition ed.). Macmillan Reference.
Rubinstein, W. D. (1999). The Weber Thesis and the Jews. In E. Brezis and P. Temin (Eds.), Elites,
Minorities and Economic Growth, Amsterdam, pp. 129–136. Elsevier.
Saleh, M. (2013). On the road to heaven: Self-selection, religion, and socioeconomic status. TSE Working
Paper, No. 13-428.
Shatzmiller, J. (1994). Jews, Medicine, and Medieval Society. Berkeley and Los Angeles: University of
California Press.
Shepherd, W. R. (1923). Historical Atlas. New York: Henry Holt and Company.
Shiue, C. H. and W. Keller (2007, September). Markets in China and Europe on the eve of the industrial
revolution. American Economic Review 97 (4), 1189–1216.
Sombart, W. (1962, 1911). The Jews and Modern Capitalism. New York: Collier Books. translated by
M. Epstein.
Sorkin, D. (2001). Port Jews and the three regions of emancipation. Jewish Culture and History 4 (2),
31–46.
Spitzer, Y. (2015, April). Pale in Comparison The Economic Ecology of the Jews as a Rural Service
Minority. memo.
Squicciarini, M. P. and N. Voigtländer (2015). Human capital and industrialization: Evidence from the
Age of Enlightenment. Quarterly Journal of Economics Forthcoming.
Stacey, R. C. (1995). Jewish lending and the medieval English economy. In R. H. Britnell and B. M.
Campbell (Eds.), A Commercialising Economy, England 1086 to c. 1300, Manchester, pp. 7–27.
Manchester University Press.
35
Storeygard, A. (2016). Farther on down the road: transport costs, trade and urban growth in sub-Sharan
Africa. Review of Economic Studies.
Tabellini, G. (2008, 04-05). Presidential address institutions and culture. Journal of the European
Economic Association 6 (2-3), 255–294.
Trivellato, F. (2004). The Port Jews of Livorno and their Global Networks of Trade in the Early Modern
Period. Jewish Culture and History 7 (1-2), 31–48.
Trivellato, F. (2009). The Familiarity of Strangers. New Haven: Yale University Press.
van Etten, J. (2012). gdistance: distances and routes on geographical grids. r package version 1.1-4.
Voigtländer, N. and H.-J. Voth (2012). Persecution perpetuated: The medieval origins of Anti-Semitic
violence in Nazi Germany. Quarterly Journal of Economics 127 (3), 1–54.
Voigtländer, N. and H.-J. Voth (2013). The three horsemen of riches: Plague, war, and urbanization in
early modern Europe. Review of Economic Studies 80, 774–811.
Weber, M. (1922, 1968). Economy and Society. New York: Bedminster.
Weber, M. (1930). The Protestant Ethic and the Spirit of Capitalism. London, U.K.: Allen and Unwin.
36
A
Supplementary Data Appendix (Not for Publication)
A.1 Jewish Communities Data
Our dataset includes all cities that are recorded as having a permanent Jewish population in the period
under consideration. As we note in the text, we explicitly code whether or not a Jewish community is
present for every between 1100 and 1800. We do this on the basis of the information contained in the
Encyclopedia Judaica. This is the same data as is used in Anderson et al. (2015). The Encyclopedia
contains information that allows us to create a database of 1,069 cities that had a Jewish presence at
some point from 1000 to 1850.
For most communities the Encyclopedia explicitly mentions the date when a Jewish community is first
recorded. For Florence a Jewish community was officially established in 1437. Alternatively, in other
cases the Encyclopedia mentions the first date at for which we know for certain that there is a Jewish
community. For example, in the entry for Trier, the Encyclopedia notes that ‘The first definitive evidence
for the presence of a Jewish community dates from 1066, when the Jews were saved from an attempted
expulsion on the part of Archbishop Eberhard through his sudden death at the altar’. The entry for
Burgus gives 974 as the date for which we know there was a Jewish community. In other cases, the data
of first entry is a rough estimate. The York entry records that Jews settled in the middle of the twelfth
century. We therefore code a community as present from 1150 onwards.
We only include communities that are labelled as Jewish by the Encyclopedia. As a result we do not
track the populations of converted Jews in Spain and Portugal after 1492 and 1497.
A.2 City Population Data
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 and Voth (2013) 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.
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.
As the relationship between Jewish communities and their hosts was qualitatively different in the
Ottoman Empire and in Russia we drop all cities iAlbania, Bulgaria, Greece, Hungary, Romania, Russia,
Malta, and Yugoslavia. We also drop cities Finland as they have no Jewish populations during the
medieval or early modern period.
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
37
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.
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.
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 (2012).
A.3 Routes Used for Travel Cost Calculations
In order to construct our instruments based on Jewish Network Access and our Market Access variable,
we needed to calculate τ , the historical cost of travel between cities. We allow for four different types
of transportation: seas, rivers, roads, and portage. The data on river locations are from Nussli (2012)
and are illustrated in Figure A.1. The data on Roman roads are from the Digital Atlas of Roman and
Medieval Civilizations. It is available from:
http://darmc.harvard.edu/icb/icb.do?keyword=k40248pageid=icb.page601659
The Roman road network is illustrated in Figure A.2 below.
We complement the Roman road data with data medieval trade routes from Shepherd (1923) shown in
Figure A.3. These data are especially helpful given that the Roman road coverage did not extend into
the North-Eastern part of Europe.
Figure A.4 shows the rasterized travel cost grid along with all the Bairoch cities. The cell size in the
raster is 10km x 10km.
38
Figure A.1: Major rivers in Europe.
Figure A.2: Major Roman Roads in Europe. Source: Digital Atlas of Roman and Medieval
Civilizations.
39
Figure A.3: Medieval Trade Routes in Europe. Source: Shepherd (1923).
40
41
Figure A.4: Bairoch cities and the least cost travel raster. Raster grid size is 10km x 10km. 1 = highest cost travel (portage
over land).
A.4 Relevance of the IV
.25
0
-.25
-.5
E(Jewish Presence Dummy | X)
.5
Figure A.5 shows the conditional non-parametric relationship between the 100k IV and Jewish community
presence. To make the figure, we ran the first stage regression with all controls while leaving out, first,
the instrument and then the Jewish presence dummy variable. We calculated the residuals from both
these regressions and then made the plot using the lpoly command in stata.
-1
-.75
-.5
-.25
0
.25
.5
.75
1
E(IV100km | X)
95% CI
Lpoly of IV100km vs Jewish Presence Dumy
kernel = epanechnikov, degree = 0, bandwidth = .2, pwidth = .29
Figure A.5: This figure depicts the relationship between our 100 km IV and the probability
that a city has a Jewish community. It establishes the relevance of our instrument.
A.5 The Correlation Between Market Access and Jewish Network Access
Figure A.6 shows the correlation between our measure of Jewish Network Access and Market Access.
This correlation is -0.037. The low correlation is not surprising given that we weight by a dummy
variable for Jewish community presence when creating the Jewish Network Access measure, whereas we
weight by city population when calculating Market Access.
42
20
15
10
Jewish Network Access IV (>100km)
5
0
0
50
100
150
Market access
200
250
300
Figure A.6: The relationship between Jewish network access and market access. Correlation
is -0.037 (p=0.0126). Source: see text.
1
A.6 Descriptive Statistics
(127)
.8
(257)
(285)
.6
(463)
(456)
(404)
1600
1700
.2
.4
(439)
(355)
0
Percent Cities with Jewish Communities
(86)
1200
1300
1400
1500
1750
1800
1850
Figure A.7: Main Sample: Proportion of cities with a Jewish community by year. Numbers
in parentheses are total number of cities in sample in that year. Source: see text.
43
1
.8
.6
(169)
.4
(467)
(564)
.2
(764)
(1,031)
(1,272)
1700
1750
(1,347)
(1,656)
1800
1850
0
Percent Cities with Jewish Communities
(107)
1200
1300
1400
1500
1600
Figure A.8: Extended Sample: Proportion of cities with a Jewish community by year.
Numbers in parentheses are total number of cities in sample in that year. Source: see text.
44
Table A.1: Descriptive Statistics: Base Sample
Variable
Jewish Community
City Growth
Lag Population
Cereal Suitability
D Rivers
D Seas
University
D Roman Road
overall
between
within
overall
between
within
overall
between
within
overall
between
within
overall
between
within
overall
between
within
overall
between
within
overall
between
within
Mean
Std. Dev.
Min
Max
Observations
0.5297203
0.4992032
0.3651457
0.3775996
0.4772435
0.2569539
0.4283768
1.020099
0.7983598
0.5723905
0.740273
0.7573301
4.37e-16
89.35061
92.00672
9.93e-15
144.6529
146.6066
0
0.2478603
0.21542
0
221.6714
225.2452
2.07e-14
0
0
-0.3591686
-2.70805
-0.5815754
-2.450224 2
0
0
-1.15797
1.9
1.9
4.542345
0.0163904
0.0163904
70.98345
0
0
160.627
0
0
0.0657343
0.1994566
0.1994566
127.2062
1
1
1.40472
2.564949
1.540445
0.406291
6.854354
5.54567
5.067735
7.278689
7.278689
4.542345
623.6542
623.6542
70.98345
621.5079
621.5079
160.627
1
1
0.0657343
1986.023
1986.023
127.2062
N = 2860
n = 493
T-bar = 5.80
N = 2860
n = 493
T-bar = 5.80
N = 2860
n = 493
T-bar = 5.80
N = 2860
n = 493
T-bar = 5.80
N = 2860
n = 493
T-bar = 5.80
N = 2860
n = 493
T-bar = 5.80
N = 2860
n = 493
T-bar = 5.80
N = 2860
n = 493
T-bar = 5.80
0.1993713
2.245025
4.542345
70.98345
160.627
0.0657343
127.2062
Notes: See text for sources.
45
Table A.2: Descriptive Statistics: Extended Sample
Variable
Jewish Community
Growth
Lag Population
Cereal Suitability
D Rivers
D Seas
University
D Roman Road
overall
between
within
overall
between
within
overall
between
within
overall
between
within
overall
between
within
overall
between
within
overall
between
within
overall
between
within
Mean
Std. Dev.
Min
Max
Observations
0.205368
0.4039977
0.3089845
0.2350868
0.4706206
0.2988284
0.4061791
0.9058235
0.6431225
0.5069051
0.8338525
0.840744
3.76e-16
131.1171
130.7078
1.36e-14
133.7657
129.6932
0
0.1576024
0.117636
0
272.5372
253.898
2.85e-14
0
0
-0.6835209
-2.70805
-1.504077 2
-2.438038
0
0
-1.545625
1.4375
1.4375
4.576107
0.0163904
0.0163904
111.9918
0
0
123.2865
0
0
0.0254846
0.1197585
0.1197585
149.2134
1
1
1.080368
2.564949
0.302585
2.418477
6.854354
5.54567
4.68008
7.84127
7.84127
4.576107
699.7518
699.7518
111.9918
651.5474
651.5474
123.2865
1
1
0.0254846
2042.151
2042.151
149.2134
N = 7377
n = 1711
T-bar = 4.312
N = 7377
n = 1711
T-bar = 4.312
N = 7377
n = 1711
T-bar = 4.312
N = 7377
n = 1711
T-bar = 4.312
N = 7377
n = 1711
T-bar = 4.312
N = 7377
n = 1711
T-bar = 4.312
N = 7377
n = 1711
T-bar = 4.312
N = 7377
n = 1711
T-bar = 4.312
0.2115573
1.85737
4.576107
111.9918
123.2865
0.0254840 6
149.2134
Notes: See text for sources.
46
B
Additional Tables
In this section we present our main results using a difference-in-differences specification. That is, we
estimate a version of Equation 1 where we use log level of population as the dependent variable and
include city fixed effects as controls. As this is no longer a Solow-type growth regression, we do not
include the lag of population as a control variable. We also do include the country fixed effects since
these would simply be absorbed by the city fixed effects.
In Table B.3 we report our results for the 1400-1850 sample. We obtain results that are qualitatively
similar to those obtained in the baseline. The coefficients we obtain increase in magnitude when we
focus on the period after the Black Death. B.4
In B.5 we present a range robustness checks for our 1400-1850 sample. We report our beta coefficient on
Jewish presence across specifications that control for the potato suitability (col 1.) , elevation (col. 2),
persecutions of Jews (cols. 3 and 4), Black Death mortality rates (col. 5), years of Jewish presence (col.
5), a specification where we use the unadjusted Bairoch cities populations (col. 7), a specification which
uses growth as the dependent variable (cols. 8-9), and the lag of population (col. 10). We also show
that are results are robust to using only those cities that exist in the dataset in 1400, 1500, 1600, and
1700 (cols. 11-14 respectively) and to dropping the largest countries (col. 16-20).
Table B.3: Jewish Communities and City Growth, 1400-1850
Dependent Variable: City Growth
Main Sample
Jewish Community
Year FE’s
Controls X Year FE’s
City FE’s
N
R2
Extended Sample
(1)
(2)
(3)
(4)
(5)
(6)
0.173∗∗∗
(0.0400)
Yes
No
No
2898
0.0887
0.271∗∗∗
(0.0397)
Yes
Yes
No
2898
0.200
0.340∗∗∗
(0.0456)
Yes
Yes
Yes
2898
0.469
0.623∗∗∗
(0.0310)
Yes
No
No
8440
0.154
0.639∗∗∗
(0.0306)
Yes
Yes
No
8440
0.232
0.293∗∗∗
(0.0428)
Yes
Yes
Yes
8440
0.442
Notes: Columns (1)-(3) use our main sample where we assign cities for which we have no
recorded Jewish presence as missing values. Columns (4)-(6) use our extended sample which
employs an alternative coding for the presence of a Jewish community that assigns cities
a zero if there is no record of a community. All specifications include year fixed effects.
Controls include cereal suitability, distance from a Roman road, and the intersection of a
Roman road, and medieval universities. Robust standard errors reported in parentheses. ∗
p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01.
47
Table B.4: Jewish Communities and and City Growth, 2nd Stage IV Analysis, 1400-1850
Dependent Variable: City Growth
Main Sample
(1)
All Cities
(2)
>50km
(3)
>100km
(4)
>250km
(5)
>500km
Jewish Community
0.700∗∗∗
(0.0986)
0.574∗∗∗
(0.0765)
0.572∗∗∗
(0.0844)
0.573∗∗∗
(0.106)
0.347
(0.258)
Year FE’s
Controls X Year FE’s
City FE’s
N
First Stage F-stat
Yes
Yes
Yes
2889
55.69
Yes
Yes
Yes
2889
133.95
Yes
Yes
Yes
2889
95.27
Yes
Yes
Yes
2889
27.40
Yes
Yes
Yes
2889
4.68
Notes: This table presents our 2nd stage IV estimates using the main sample. Column 1
uses our simple Jewish Network Access measure. Columns 2-5 use our instruments where
we exclude cities within a 50, 100, 250, and 500 km radius respectively. Our preferred
specifications are in Columns (3) and (4). Robust standard errors reported in parentheses.
∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01.
48
Table B.5: Jewish Communities and and City Growth, Robustness, 1100-1850
βOLS Main
βOLS Extended
βIV 100km
βIV 250km
N Main
N Extended
49
βOLS Main
βOLS Extended
βIV 100km
βIV 250km
N Main
N Extended
(1)
Potato
(2)
Elevation
(3)
Expulsions
(4)
Persecutions
(5)
Black
Death
Mortality
(6)
Years
Jewish
(7)
Unadjusted
Bairoch
(8)
Solow
(base)
(9)
Solow (extended)
0.347∗∗∗
(0.0459)
0.301∗∗∗
(0.0429)
0.495∗∗∗
(0.0893)
0.614∗∗∗
(0.108)
2889
8440
0.317∗∗∗
(0.0446)
0.273∗∗∗
(0.0417)
0.690∗∗∗
(0.107)
0.472∗∗∗
(0.113)
2889
8440
0.389∗∗∗
(0.0503)
0.364∗∗∗
(0.0490)
0.580∗∗∗
(0.0986)
0.677∗∗∗
(0.131)
2889
8440
0.371∗∗∗
(0.0507)
0.356∗∗∗
(0.0503)
0.00631∗∗∗
0.343∗∗∗
(0.0459)
0.301∗∗∗
(0.0429)
0.580∗∗∗
(0.0986)
0.556∗∗∗
(0.131)
2889
8440
0.00411∗∗∗
(0.000518)
0.00359∗∗∗
(0.000483)
0.00631∗∗∗
(0.000949)
0.00618∗∗∗
(0.00118)
2889
8440
0.348∗∗∗
(0.0457)
0.300∗∗∗
(0.0429)
0.589∗∗∗
(0.0850)
0.595∗∗∗
(0.107)
2889
8440
0.0826∗∗∗
(0.0188)
0.151∗∗∗
(0.0154)
0.217∗∗∗
(0.0497)
0.404∗∗∗
(0.0754)
2647
7101
0.158∗∗∗
(0.0290)
0.146∗∗∗
(0.0292)
0.283∗∗∗
(0.0755)
0.280∗∗∗
(0.0946)
2602
7101
0.324∗∗
(0.136)
0.390∗∗∗
(0.107)
0.572∗∗
(0.258)
0.573∗∗
(0.281)
2889
8440
(11)
Cities
1400
(12)
Cities
1500
(13)
Cities
1600
(14)
Cities
1700
(15)
Drop UK
(16)
Drop
Netherlands
(17)
Drop
France
(18)
Drop
Germany
(19)
Drop Italy
(20)
Drop
Spain
0.362∗∗∗
(0.0595)
0.315∗∗∗
(0.0558)
0.490∗∗∗
(0.0986)
0.486∗∗∗
(0.120)
1757
3178
0.343∗∗∗
(0.0542)
0.297∗∗∗
(0.0504)
0.450∗∗∗
(0.0962)
0.450∗∗∗
(0.121)
1953
3850
0.348∗∗∗
(0.0503)
0.295∗∗∗
(0.0468)
0.482∗∗∗
(0.0834)
0.458∗∗∗
(0.104)
2349
5002
0.351∗∗∗
(0.0486)
0.305∗∗∗
(0.0455)
0.551∗∗∗
(0.0834)
0.551∗∗∗
(0.104)
2564
6272
0.202∗∗∗
(0.0421)
0.169∗∗∗
(0.0375)
0.112
(0.121)
0.0106
(0.193)
2643
7636
0.340∗∗∗
(0.0456)
0.293∗∗∗
(0.0428)
0.572∗∗∗
(0.0844)
0.573∗∗∗
(0.106)
2889
8440
0.383∗∗∗
(0.0540)
0.331∗∗∗
(0.0500)
0.728∗∗∗
(0.0848)
0.748∗∗∗
(0.103)
2340
6960
0.415∗∗∗
(0.0530)
0.339∗∗∗
(0.0501)
0.621∗∗∗
(0.0822)
0.601∗∗∗
(0.101)
2191
7111
0.329∗∗∗
(0.0495)
0.301∗∗∗
(0.0477)
0.503∗∗∗
(0.0921)
0.463∗∗∗
(0.113)
2422
6641
0.306∗∗∗
(0.0441)
0.269∗∗∗
(0.0424)
0.566∗∗∗
(0.0956)
0.571∗∗∗
(0.122)
2518
7239
0.690∗∗∗
(0.145)
2889
8440
(10)
Lag Population
Notes: In Columns (1)-(5) we interact our year fixed effects with a range of additional time invariant controls: potato
suitability (Col. 1); temperature (Col. 2); latitude (Col 3.); pogroms or expulsions (Col. 4); Black Death mortality (Col.
5); Years of Jewish presence (Col. 6). In Column 7 we use the unadjusted Bairoch dataset. Column 8 reports a Solow
specification with initial population size. Column 10 uses lag population as in our baseline. In Columns 11-15 we include only
cities that existed in either 1300, 1400, 1500, 1600, or 1700 respectively. Columns 15-20 drop cities from the largest modern
countries. Robust standard errors in parentheses. ∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01.