We are unable to create an online viewer for this document. Please download the document instead.
Immigration, Industrial Revolution and Urban Growth in the United States, 1820-1920:
Factor Endowments, Technology and Geography
Washington University in St. Louis and NBER
I thank Kei-Mu Yi and seminar participants at the 2006 Regional Science Association meetings
in Toronto, St. Louis Federal Reserve, and Philadelphia Federal Reserve for their comments.
Financial support from Washington University is gratefully acknowledged.
Industrial revolution is fundamentally linked with the rise of factories and the decline of skilled
artisans in manufacturing. Most scholars agree that factories as compared to artisan shops were
intensive in unskilled labor. Indeed, the hallmark of the early factories is the utilization of
division of labor of relatively unskilled workers. This paper explores whether the massive influx
of unskilled immigrants between 1840 and 1920, by significantly increasing the ratio of
unskilled to skilled labor endowment, contributed to the growth and spread of factory
manufacturing in the United States. The data indicate that immigration not only contributed to
the growth and spread of factories but it also contributed to the growth of cities.
The century between 1820 and 1920 defined America as a nation of immigrants or a
“melting pot.” During this century, more than 33 million people entered the ports of the United
States. Immigrants from Europe came in massive waves until the era of open immigration ended
with the passage of the 1921 Emergency Quota Act (Figure 1).1 By the end of the first three
decades of immigration, the census of 1850 finds that almost 10% of Americans was foreign-
born. The share of the foreign-born population fluctuated around 13-15% between 1860 and
1920, but immigrants and their children represented 30-40% of the white population (Figure 2).
With the passage of the Immigration Act of 1965, immigration rose steadily during the last three
decades. However, this recent wave of immigration pales in comparison to that of the earlier
waves in duration and in terms of the share of aggregate population (Figure 3).
What impact did immigration have on the American economy during the era of mass
immigration? Goldin (1994) and Hatton and Williamson (1998) find that immigrants and natives
were substitutes; immigration lowered native wages and displaced natives from the northeastern
United States where immigrants largely settled.2 Studies on immigrant assimilation generally
find that immigrants earned lower wages on arrival but provide different assessments on their
rates of assimilation. While Hannon (1982), Eichengreen and Gemery (1986) and Hanes (1996)
find that wage growth among immigrants was slower than native-born workers, Blau (1980),
Hatton (1997), and Minns (2000) find that immigrants experienced faster wage growth than
1 The movement to restrict immigration in the U.S. started in the late nineteenth century. Between 1897 and 1917,
the House and the Senate passed numerous bills on literacy tests which finally became law in 1917. When literacy
tests proved to be ineffective in curbing immigration, the Congress moved toward a quota system. See Hingham
(1955) and Goldin (1994).
2 Friedberg and Hunt (1995) provide an excellent summary of the literature, especially of works on the impact of
immigration on the second half of the twentieth century. Most studies on recent immigration find that immigrants
and natives are substitutes. Some studies such as Borjas (1999) and Borjas, Freeman and Katz (1997) find large
negative impact whereas Card (2001) finds small impact. On the other hand, Ottaviano and Peri (2005a, 2000b)
native-born and caught up to native-born level of earnings within 20 or 25 years. For the
antebellum period, Ferrie (1999) finds that immigrants were geographically, occupationally and
financially mobile.3 However, from an aggregate perspective, because American real wages rose
steadily between 1820 and 1920, many scholars such as Goldin (1994) point to the absorptive
capacity of the American economy (Figure 4).4
In this paper, I explore whether immigration had a more fundamental impact on the
American economy between 1860 and 1920. In particular, I investigate whether immigration
during this period had a significant impact on the growth and spread of factory organization in
manufacturing. Between 1820 and 1840, when factory production was still in its infancy in
America, immigration may have hindered the spread of factories as many skilled European
artisans sought refuge from the spread of European factory production by moving to America.
Immigration after 1840, however, is likely to have contributed to the growth of factories as it
significantly increased the unskilled to skilled labor endowment in America. Because factory
production utilized unskilled workers intensely, the dramatic increase in unskilled to skilled
labor endowment ratio is likely to have had a significant impact on the growth of factory
argue that immigrants and natives are complements and that immigrants had a positive impact on native wages.
3 Ferrie (1999) provides a richly detailed study of the immigrant experience by constructing a sample of immigrants
from passenger ship lists who entered through New York city in the 1840s and located in the census of populations
in 1850 and 1860. By 1850, most immigrants reached their intended destinations (New York, Ohio, and
Pennsylvania) and only 17% remained in the vicinity of New York city. Relative to the native population,
immigrants were initially disproportionately represented as skilled artisans and unskilled laborers. About a quarter of
immigrants moved downwards in occupational status, but the dominant tendency was to move upwards. The Irish,
compared to the British and Germans, were least mobile both geographically and occupationally.
4 Data on real wages were constructed from various sources such as Coombes (1926), Aldrich and Week’s series
from Long (1960), Rees (1961), Lebergott (1964), and Margo (2000). The nominal values were converted to real
wages using the BLS CPI from the Historical Statistics of the United States (1975). Aldrich, Weeks, Rees, and
Lebergott series are the average earnings in all manufacturing, Coombes series is the average earnings of unskilled
workers in manufacturing, and Margo series is the average earnings of common laborers in the Northeast. The ratio
of skilled to unskilled wages seems to have fluctuated without any visible trend between 1820 and 1860 based on the
ratio of wages of artisans and common laborers in the Northeast (Margo (2000)) but seems to have narrowed
between 1890 and 1940 (Goldin and Katz (1999), Goldin and Margo (1992)). Also see Williamson and Lindert
production in the United States.5
The industrial revolution, which began in England and spread to elsewhere in Europe and
to the United States, is fundamentally linked with the rise of factories and the decline of skilled
artisans in manufacturing (Mokyr (2002)). While there are many theories on the rise of factories,
this paper emphasizes the importance of the unskilled intensity of factory production.6 It is
widely believed that factories substituted skilled artisans with the division of labor of unskilled
workers who specialized in a limited number of tasks (Sokoloff (1984), Atack (1987), Goldin
and Katz (1998), Atack, Bateman and Margo (2005)).7 In the United States, in the early
industrial period (1820-1840), factories in New England utilized the unskilled labor of women
and children (Goldin and Sokoloff (1982)); however, in the second industrial revolution (1860-
1920), unskilled immigrant laborers were the dominant factory manufacturing labor force.
5 Beginning with Habakkuk (1962), there is an extensive literature on skilled-biased technology in American
history. For Habakkuk, labor scarcity rather than immigration of unskilled workers caused the early adoption of
mechanized factory production in America as compared to Britain. For Rosenberg (1972), the American system of
manufacturing was biased toward resource intensive technologies due to its abundant resources. Goldin and Katz
(1998) interpret the works of James and Skinner (1985) and Cain and Patterson (1986) as providing evidence for the
existence of technology-unskilled complementarity in the nineteenth century U.S. manufacturing. These works
indicate that physical capital, raw material and unskilled labor were complements and that they substituted for
skilled artisans. To the contrary, Williamson and Lindert (1990) argue that physical capital was a complement to
skilled rather than to unskilled labor during this period (also see Temin (1966) and David (1975)). In this paper, I
highlight the role of the relative supplies of skilled and unskilled workers as in Goldin and Katz (1998) and
Acemoglu (1998). The empirical estimation in this paper is motivated by Acemoglu’s model which predicts that an
increase in the supply of unskilled (skilled) workers increases the technologies used by unskilled (skilled) workers.
Thomas (1954) and Erickson (1957) believed that immigration contributed to the growth of factories in the U.S.
Acemoglu (1998) suggests that the increase in the supply of unskilled workers from villages and Ireland to English
cities, as documented by Williamson (1990), may have played a role in the rise of factories in England as well.
Finally, Lewis (2003, 2006) finds that immigration had an impact on the direction of American technology in the
second half of the twentieth century.
6 Mokyr (1999, 2002) provides an excellent summary of the literature on the British industrial revolution. Mokyr
(2002) examines three main classes of explanations for the rise of factories: fixed costs and physical economies of
scale, information costs and incentives and labor effort. However, he argues that the most compelling explanation for
the rise of factories is based on ideas developed by Demsetz (1988) and Becker and Murphy (1992), namely that
“division of labor is limited by the size of the knowledge set necessary to execute and operate best-practice
techniques.” While workers possessed different skill endowments, factories served as repository of technical
knowledge and reduced the costs of transmitting this knowledge to individual workers.
7 According to Atack, Bateman and Margo (2004), the factory system through division of labor shortened the period
of skill acquisition and contributed to the de-skilling of workers. Thus, the factory system was well adapted to utilize
the abundance of unskilled immigrant workers. Consistent with the de-skilling hypothesis or the intense utilization of
unskilled workers, Atack, Bateman and Margo find that average wages of firms fell with increases in firm size.
The pace and the skill composition of immigrants differed greatly between the early and
late industrial period in the United States (Tables 1 and 2). In the early period of industrialization
between 1820 and 1840, the pace of immigration was modest and most of the immigrants were
skilled artisans and were relatively wealthy. In the transition period between the early to late
industrialization, the rate of immigration rose dramatically and a great majority of immigrants
were unskilled farmers, laborers and servants. Although the pace of immigration fluctuated and
the sources of immigrants shifted from northwestern to central and southeastern Europe by the
second industrial revolution, a majority of immigrants remained unskilled workers. Thus,
immigration between 1846 and 1920 significantly increased the unskilled to skilled labor
endowment ratio in the United States.
One major issue is whether factory jobs “pulled” immigrants to the United States or
whether immigrants endogenously changed the direction of American technology toward factory
organization of production. My principal identification strategy rests on the exogeneity or the
“push” factor of immigration between 1847-1854, a period which marked the first major wave of
mass immigration. Many scholars agree that the most important cause of immigration during this
period was the potato famine in Ireland and in other European countries (Ó Gráda and O’Rourke
(1997), Cohn (2000), Hatton and Williamson (2005)). The potato famine, caused by p. infestans,
a fungus-like disease that turns the potato into inedible black mush, reduced the acreage of
potato in Ireland from 2.1 million acres in 1845 to a mere 0.3 million in 1847 causing massive
deaths and emigration (Ó Gráda (1999), Mokyr (1985)).
The instruments used in this paper are the share of foreign-born population in 1850 and
the growth in the foreign-born population between 1850 and 1860.8 Because a large share of
8 The growth in the foreign-born population between 1850 and 1860 is likely to capture the large inflow of German
immigration prior to the antebellum period occurred during the famine period, the share of
foreign-born population in 1850 should be highly correlated with the share of immigrants
induced by the potato famine. The total numbers of immigrants between 1846 and 1850 equaled
that of the entire period between 1820 and 1845; in addition, the Irish foreign-born represented
almost 43% of all foreign-born population in 1850 (Gibson and Lennon (1999)).9 The growth in
the foreign-born population between 1850 and 1860 should also be correlated with the famine-
induced immigration as the number of immigrants between 1851 and 1854 were one and a half
times greater than the number between 1855 and 1860.
Data on the occupation of immigrants during the mid-nineteenth century provide little
evidence for the proposition that these first wave of immigrants was “pulled” by factory jobs in
the United States. Cohn (1992), based on data from passenger lists from ships which arrived in
New York between 1836-1853, finds that a majority of English immigrants were unskilled and
less than 5% of them possessed prior industrial experience.10 Relative to the English population,
immigrants were over-represented as farmers and laborers but under-represented in almost all
other occupational categories (Table 3). For a sample of immigrants who came in the 1840s,
Ferrie (1999) finds that a majority of the Irish came as unskilled laborers, the British as unskilled
immigration between 1852 and 1854 sparked by the political repression following the unsuccessful 1848 revolution
(Atack and Passell (1994)). In addition, as compared to studies such as Altonji and Card (1991) and others which use
historic shares of foreign-born as instruments, the historic level of share of foreign-born in 1850 and the growth of
foreign-born population between 1850 and 1860 are much more likely to be exogenous.
9 It is important to note that immigration data between 1820 and 1860 were incomplete and subject to both under-
and over-enumeration. No data were collected for immigrants arriving from Canada, Mexico and Pacific ports and
the data included transients bound for territories outside of the U.S. and double counted merchants and visitors who
made more than one return voyage from Europe. Of these factors, the most important factor was the under-
enumeration of immigration flow through Canada. Between 1810-1839, a large share of immigrants to the United
States arrived at St. Lawrence ports (see McClelland and Zeckhauser (1982)). Before 1865, about half of UK
immigrants, especially the poor, may have come through Canada because the fare to Canadian maritime ports was
less than half of that to American ports (Ferrie (1999)).
10 There is considerable anecdotal evidence that most skilled immigrants were artisans rather than factory managers,
mechanics or operatives. As the rise of factory production in Europe displaced artisans in Europe, they moved to
America. Thus, the arrival of skilled artisans may have delayed the onset of industrial revolution in American cities.
However, as factory production gained momentum in the United States, there is evidence that skilled artisans
laborers and skilled artisans and Germans as farmers and skilled artisans.
While the long-run ebb and flow of immigration was due to the combination of “push”
and “pull” factors, one additional exogenous factor which significantly increased the share of
unskilled immigrants is related to major advances in transportation.11 First, advances in internal
transportation due to railroads provided easy access to major ports for most European
populations (Hatton and Williamson (2005)). Second, the advances in steamship technology
made the trans-Atlantic travel shorter, safer, and easier to get in and out of secondary ports in the
Mediterranean (Cohn (2005), Keeling (1999)). Third, the passenger costs relative to per capita
income fell significantly between 1820 and 1860 (Galenson (1984)).
My second identification strategy utilizes the fact that most immigrants entered the
United States through New York and used domestic transportation networks to reach their
intended destinations. Between 1850 and 1914, 70% of immigrants arrived via the ports in New
York (Keeling (1999)); and many immigrants moved immediately from the port of entry to
internal destinations (Ferrie (1999)). Distance from New York city and access to water
transportation in 1850 are likely to capture the influence of transportation costs on immigrant
settlement patterns. It is important to note that New York city became the dominant mercantile
port and later the port of entry for immigrants long before it became an industrial city (Albion
(1939)). However, distance from New York city and access to water transportation in 1850 may
also be correlated with access to markets for manufacturers.
To determine whether immigrants contributed to the growth and spread of unskilled-
biased technology embodied in factory-assembly production, I use data of manufacturing firms
drawn from the manuscripts of the decennial censuses between 1850 and 1880 which have been
became managers and foreman of factories suggesting complementarities between immigrant artisans and laborers.
merged with county-level information from the censuses of populations from the same respective
years.12 Data analysis suggests that immigration between 1850 and 1920 may have had a
fundamental impact on the direction of American technology. The data reveal that firms in
counties with a higher share of foreign-born were much more likely to be organized as factories
and were generally larger. In addition, firms in these counties were also more productive and
were likely to pay higher average wages to their workers. Standard tests of the instruments
indicate their general validity and that the 2SLS estimates are generally similar to those of
ordinary regression estimates.
The paper is organized as follows. Section II presents the model, empirical strategy and
data analysis on the impact of immigration on unskilled biased technology as embodied in
factories. Section III examines the immigrant diversity of occupations by nationality. Section IV
studies immigration and urbanization. Section V concludes.
II. Endowments and Technology
Industrial development in the United States exhibited three major production
technologies: artisan shops (1790-1820), factory-assembly (1820-1920), and factory-continuous
(1920- ). Prior to the industrial revolution, skilled artisans produced the entire product with the
help of apprentices and family members. With the industrial revolution, factories hired numerous
unskilled workers who specialized in few tasks based on division of labor and few skilled
workers who operated machines and supervised workers. In the first half of the twentieth
century, however, the factory-continuous method began to replace the factory-assembly system
11 See Thomas (1954) and Hatton and Williamson (1998).
12 The firm-level manuscript data were constructed by Jeremy Atack, Fred Bateman and Thomas Weiss; Michael
Haines generously provided the merged data set. For discussions on sampling criteria and other pertinent concerns
related to the firm-level data, see Atack and Bateman (1999), Atack, Bateman and Margo (2004, 2005), and Kim
in a number of industries (Goldin and Katz (1998), Jerome (1924)). Unlike the earlier factories,
the new factory-continuous system was intensive in skilled rather than unskilled workers. With
the advent of electric motors, mechanization replaced the division of labor of unskilled workers.
Whereas Goldin and Katz (1998) examine the shift in production from factory-assembly
to the factory-continuous methods, I examine the shift from artisan shop to the earlier factory
system based on unskilled division of labor.13 The empirics of this paper is motivated by
Acemoglu’s (1998) model of endogenous technological change. The model assumes that there
are two types of workers, skilled (H) and unskilled (L), who supply labor inelastically. The
consumption good is produced from two complementary production process, one using skilled
and the other using unskilled. Firm level technology, Ah or Al, is determined by technology
employed by the firm where skilled and unskilled workers are assumed to use different
technologies. The main result of Acemoglu’s model is summarized by the following equation:
Ah/Al = f(p, H/L) where p=ph/pl. In this model, an increase in the relative supply of a skill type
will lead to an improvement in the technologies that uses that particular skill type. Thus, an
exogenous increase in the supply of skilled (unskilled) workers will lead to an improvement in
technologies which utilize skilled (unskilled) workers.14
In this paper, I assume that the adoption of the factory method of production signifies the
increase in the use of unskilled intensive technology and also assume that the share of foreign-
born population (FB/(FB+NB)) is a useful measure of the relative supply of unskilled to skilled
13 Goldin and Katz (1998) suggests that the growth in the supply of skilled labor due to the sudden growth in high
school educated workers may have fueled skilled-biased technology toward the factory-continuous batch production.
In addition, the supply of unskilled workers fell dramatically during this period as immigration slowed to a trickle.
14 The intuition of the model is as follows: when there are more unskilled workers, then the market for unskilled-
complementary technologies, such as factories, is larger. As a result, more resources will be devoted to the invention
of unskilled-complementary technologies. In this model, Acemoglu (1998) shows that the impact on wages is
dynamic. Initially, the shift in supply of unskilled workers will lower unskilled wages; however, as the “directed