Effects of Imigration in Brazil

Download as pdf or txt
Download as pdf or txt
You are on page 1of 48

NBER WORKING PAPER SERIES

THE EFFECTS OF IMMIGRATION IN A DEVELOPING COUNTRY:


BRAZIL IN THE AGE OF MASS MIGRATION

David Escamilla-Guerrero
Andrea Papadia
Ariell Zimran

Working Paper 32083


http://www.nber.org/papers/w32083

NATIONAL BUREAU OF ECONOMIC RESEARCH


1050 Massachusetts Avenue
Cambridge, MA 02138
January 2024

For helpful comments, we thank William Collins, Martín Fernández Sánchez, David Jaeger, Aldo
Musacchio, Blanca Sanchez-Alonso, Sandra Sequeira, João Tampellini; seminar participants at
the University of Bonn, University of York, University of Newcastle, Wageningen University,
and Vanderbilt University; and conference participants at the 2021 Markets and Policy in History
Workshop, 2022 Economic History Society Conference, 2022 European Historical Economics
Society Conference, and the 2022 World Economic History Congress. Jared Katz and Víctor
Luna provided excellent research assistance. Renato Colistete kindly shared data with us.
Financial support was provided by Vanderbilt University and the University of St Andrews. The
views expressed herein are those of the authors and do not necessarily reflect the views of the
National Bureau of Economic Research.

NBER working papers are circulated for discussion and comment purposes. They have not been
peer-reviewed or been subject to the review by the NBER Board of Directors that accompanies
official NBER publications.

© 2024 by David Escamilla-Guerrero, Andrea Papadia, and Ariell Zimran. All rights reserved.
Short sections of text, not to exceed two paragraphs, may be quoted without explicit permission
provided that full credit, including © notice, is given to the source.
The Effects of Immigration in a Developing Country: Brazil in the Age of Mass Migration
David Escamilla-Guerrero, Andrea Papadia, and Ariell Zimran
NBER Working Paper No. 32083
January 2024
JEL No. F22,J61,N36,N56,O13,O15,Q15

ABSTRACT

The effects of immigration are reasonably well understood in developed countries, but they are
far more poorly understood in developing ones despite the importance of these countries as
immigrant destinations. We address this shortcoming by studying the effects of immigration to
Brazil during the Age of Mass Migration on its agricultural sector in 1920. This context benefits
from the widely recognized value of historical perspective in studies of the effects of
immigration. But unlike studies that focus on the United States to understand the effects of
migration from poor to rich countries, our context is informative of developing countries'
experience because Brazil in this period was unique among major migrant destinations as a low-
income country with a large agricultural sector and weak institutions. Instrumenting for a
municipality's immigrant share using the interaction of aggregate immigrant inflows and the
expansion of Brazil's railway network, we find that a greater immigrant share in a municipality
led to an increase in farm values. We show that the bulk of the effect of immigration can be
explained by more intense cultivation of land, which we attribute to temporary immigrants
exerting greater labor effort than natives. Finally, we find that it is unlikely that immigration's
effect on agriculture slowed Brazil's structural transformation.

David Escamilla-Guerrero Ariell Zimran


University of St Andrews Department of Economics
School of Economics and Finance Vanderbilt University
Castlecliffe 2301 Vanderbilt Place
The Scores Nashville, TN 37235
St Andrews KY16 9AZ and NBER
United Kingdom ariell.zimran@vanderbilt.edu
and IZA
drescamillag@gmail.com

Andrea Papadia
Department of Economics and Related Studies
University of York
Alcuin D Block, Campus West
York YO10 5DD
United Kingdom
andrea.papadia@york.ac.uk

A data appendix is available at http://www.nber.org/data-appendix/w32083


1 Introduction

What are the effects of immigration in a developing country? Despite the importance of developing countries

as migrant destinations (Ratha and Shaw 2007),1 hosting over 40 percent of world migrants (World Bank

2016, p. 11), the effects of immigration in these economies remain unclear. The literature on the economics of

migration focuses almost entirely on immigration to rich countries and on developing countries as immigrant

sources (Biavaschi et al. 2018; Böhme and Kups 2017; Hanson 2009; Hatton and Williamson 2002, 2005),

providing little guidance. Indeed, developing countries may attract different types of migrants or migrants

from different places of origin than more developed countries, which, together with substantial differences in

economic structure and institutions, may ultimately cause the effects of immigration to differ from those in

more developed countries.

In this paper, we study the effects of European immigration to Brazil in the period 1855–1920. Our

main focus is on Brazil’s agricultural sector in 1920, motivated by the structure of Brazil’s economy and

the role of immigrants within this economy. Brazil at this time was an overwhelmingly agrarian economy

with an emerging industrial sector. Brazilian policy makers, motivated by labor demand in the expanding

coffee sector and the decline and eventual abolition of slavery in 1888, initially encouraged immigration from

Europe to provide labor for the agricultural sector,2 and even outside of official migration subsidy programs,

immigrants largely entered the agricultural sector. Given the prominence of the agricultural sector and its

importance in the immigrant experience in Brazil, we investigate whether migration did, in fact, support

the development of this sector, which we operationalize by focusing on the value of farms across Brazilian

municipalities. We also investigate the contribution of Brazil’s agriculture-focused immigration phenomenon

to the country’s ongoing structural transformation, which enables us to address the long-standing debate

in development economics regarding the role of agricultural development in overall economic development

(e.g., Asher et al. 2022).

Our historical context is situated within the Age of Mass Migration—a period roughly 1840–1920 in which

over 50 million people migrated from Europe to the Americas (Hatton and Ward 2019; Hatton and Williamson

1998). The value to the economics of immigration from studying this period is broadly recognized,3 including

in studies of the effects of immigration (e.g., Abramitzky et al. 2023; Price, vom Lehn, and Wilson 2023;

1 It is estimated that the stock of migrants from developing countries in developing countries exceeds the stock of migrants

from developing countries in developed countries (World Bank 2016).


2 The desire to attract specifically European migrants was at least partially founded on racist beliefs of white superiority and

the widespread desire amongst Brazilian policymakers to “whiten” Brazil’s population through immigration and miscegenation
(da Costa 2000; Machado 2006; Slenes 2010).
3 See, for instance, Abramitzky and Boustan (2017), Abramitzky, Boustan, and Eriksson (2012), Collins and Zimran (2023),

Hatton and Williamson (1998), and Spitzer and Zimran (2018).

1
Tabellini 2020).4 This literature, however, focuses almost exclusively on the United States, generating

insights for migration from poor to rich countries. Brazil, however, stands apart from the United States and

the other main immigrant receiving country—Argentina.5 Its wages and per-capita GDP were substantially

lower (Bolt and van Zanden 2020; Williamson 1995), its agricultural sector was relatively larger, and its

institutions were uniquely extractive.6 These features of the Brazilian economy make Brazilian immigration

a closer historical analog of modern immigration to developing countries than is any other immigration

flow during this period. The unique and important position of the Age of Mass Migration as a source of

insights for the economics of immigration, combined with Brazil’s unique status in this context as a relatively

poor, agricultural, and undeveloped destination country, thus enables us to better understand the effects of

immigration in developing countries.

The main empirical challenge that we face is the endogeneity of immigrants’ location choice within

Brazil. To address this challenge, we adapt to the Brazilian case an identification strategy previously used

by Sequeira, Nunn, and Qian (2020) to study the effects of immigration in the United States during the Age

of Mass Migration. This strategy creates an instrumental variable for a municipality’s immigrant share of

population based on the interaction of aggregate immigrant arrivals and the development of the rail network,

which the historical evidence shows was an important method for immigrants to reach their destinations

(Holloway 1980; Lanza, Manier, and Musacchio 2023).7 Intuitively, we compare two municipalities, one of

which was linked to the railroad in the year of a large immigration inflow, and the other of which was not

linked in that year. The former, by virtue of having access to the rail network in the year of a large inflow, is

predicted to have a greater share of immigrants relative to population.8 This method enables us to control

for rail linkage directly, addressing concerns that rail was built targeting specific areas or had direct effects

on the local economy. We also show that municipalities linked during immigration booms were similar to

those linked during lulls, obviating concerns that the timing of the construction of the rail network and the

timing of immigrant inflows were directly linked (Sequeira, Nunn, and Qian 2020).

4 The value of historical perspective has also been exploited in studies of the effects of immigration in the context of the

Bracero program (e.g., Clemens, Lewis, and Postel 2018; Lee, Peri, and Yasenov 2022).
5 Brazil was the third most important of these destinations in terms of number of immigrants. Canada was a close fourth

(Ferenczi and Willcox 1929, p. 172).


6 Although both the United States and Argentina were also agricultural, Brazil’s sector was substantially larger as a share of

the labor force, and especially of the immigrant labor force—about 40 percent in 1920, as compared to about 15 percent in the
United States and 17 percent in Argentina in 1895. Moreover, in the case of the United States, the bulk of immigration provided
labor to the manufacturing sector (Lafortune, Lewis, and Tessada 2019). In Brazil, immigration was targeted to agriculture.
7 Indeed, the Hospedaria dos Imigrantes in Santos, where immigrants were lodged at arrival in São Paulo, was linked directly

to the railway network, and immigrants were often provided with tickets to reach their destinations. The same was true of the
immigrant hostels in the other major ports of arrival.
8 We advance this strategy by exploiting annual variation in the extent of the rail network rather than decadal variation of

the form used by Sequeira, Nunn, and Qian (2020).

2
Our main finding, based primarily on data that we digitized from the 1920 Brazilian census, is that

municipalities with a higher share of European immigrants in 1920 had greater farm values per hectare,

which we interpret as indicating greater development of the agricultural sector. This is true not just for the

total farm value, but also for each of its three constituent components—land, infrastructure, and tools and

machines. The finding that immigration affected not just land prices, but also the other two components is

crucial for two reasons. First it alleviates the concern that our outcome variable captures local character-

istics other than agricultural development, which would be capitalized in land values. Second, the value of

infrastructure, tools, and machines is linked to the development and productivity of the agricultural sector,

which is our ultimate object of interest. Our preferred specification indicates that a one-standard deviation

increase in a municipality’s share of European immigrants generated a 0.7-standard deviation increase in its

farm values per hectare. We show that these results are robust to using different variations of our instru-

ment and to excluding a variety of sets of municipalities that may have been systematically different from

the average municipality—those hosting immigrant colonies, large coffee producers, the earliest places linked

to the railway, and large population centers.

We also investigate the mechanism for the effect of immigration on farm values. What we find to be the

most substantial mechanism is changes in land use patterns. In particular, we find that a greater share of

European immigrants led to an increase in the share of farmland cultivated (as opposed to being left fallow

or as forest), and that this increase in the intensity of cultivation was responsible for about one quarter of the

effect of immigration. We argue that this mechanism reflects temporary immigrants exerting greater labor

effort than natives or permanent immigrants. Such a mechanism is in line with a large body of literature

showing both theoretically and empirically that temporary immigrants increase their work effort (or labor

supply) while in the destination,9 with the fact that a substantial portion of Brazilian immigration from

Europe was temporary (Hatton and Williamson 1998; Lesser 2013), with historical accounts of substantial

effort exerted by Brazilian immigrants (Cinel 1991; Florea 2023; Sánchez-Alonso 2007). Consistent with this

notion, we show that the effect of immigration on farm values was the product of immigrant labor, not immi-

grant ownership. Moreover, using farm-level data from São Paulo state and exploiting within-municipality

variation, we confirm that immigrant labor and not immigrant ownership affected farm values. These results

imply that when immigrants transitioned to ownership, indicating permanent residence, the benefits of their

presence for farm values dissipated. We also investigate, and either rule out or show to be unimportant

9 See Dustmann (1994, 2000), Dustmann, Bentolila, and Faini (1996), Dustmann and Görlach (2016), Epstein and Venturini

(2011), Galor and Stark (1991), Hill (1987), Klinthäll (2006), Kyarko and Chartouni (2017), Vijverberg and Zeager (1994), and
Wahba (2022).

3
relative to the land-use mechanism, several other mechanisms—immigrant arrivals increasing the demand

for land or providing labor that would otherwise be absent to aid in the exploitation of land; increased coffee

cultivation; and increases in the capital intensiveness of agriculture in response to immigration in the form

of an increase in the adoption of agricultural tools.

To assess the overall impact of immigration on economic development, we test whether immigration,

by facilitating agricultural development, slowed Brazil’s structural transformation. Such a test addresses a

long-standing debate about the role of agricultural development in promoting or delaying structural transfor-

mation (Asher et al. 2022), which is also present in the Brazilian historiography (Suzigan 2000, pp. 23–47).

We find that immigration did not slow, and in fact may have accelerated Brazil’s structural transformation.

In addition to a decline in the agricultural share of the labor force, we find a positive effect of immigra-

tion on the literacy of both native- and foreign-born individuals, suggestive evidence that immigration was

associated with increased industrial employment, and no evidence that immigration reduced female labor

force participation, which would have been detrimental for industrialization. We also find no evidence of an

adverse impact on institutions in the form of an increased presence of rentiers, whose presence may have

held back development, or of an inflated public sector.

This paper advances a fundamental literature within the economics of immigration—that studying the

effects of immigration on the receiving country (Bansak, Simpson, and Zavodny 2020; Borjas 2014). The near

exclusive focus of this literature on the effects of immigration in developed countries creates an important

blind spot given the importance of developing countries as destinations for immigrants. The combination

of the recognized benefit that historical perspective brings to advancing the economics of immigration and

Brazil’s unique position in the Age of Mass Migration as a major migrant destination that bears a stronger

resemblance to modern developing countries than any other destination in that period, enables us to enrich

this literature by shedding new light on the effects of immigration in developing rather than developed

countries.

This paper also contributes to the literature focusing on the economics of the Age of Mass Migration (e.g.,

Abramitzky and Boustan 2017; Hatton and Ward 2019), and more particularly to studies of the effects of

immigration in this period (e.g., Abramitzky et al. 2023; Ager et al. 2021; Cohen and Biddle 2022; Price, vom

Lehn, and Wilson 2023; Tabellini 2020).10 Just as the economics of immigration has focused on developed-
10 In addition to bringing a different perspective with its focus on Brazil, this paper provides an important insight on the
effects of immigration to the United States in the Age of Mass Migration, albeit in a counterfactual sense. Specifically, the paper
sheds light on the what the effects may have been had immigrants settled in large numbers in the US South. In reality, few
immigrants to the United States settled outside the Northeast and Midwest. But contemporaries repeatedly discussed a desire
for the labor supply that immigrants would provide in the South (Benton-Cohen 2018; Goldin 1994; US Congress 1911). Indeed,
there were a number of unsuccessful efforts to encourage Europeans to settle in the US South, and subsidized immigration—an

4
country destinations, this literature has focused largely on the United States, where immigrants arrived in

pursuit of high wages in an industrializing economy (Hatton and Williamson 1998; Lafortune, Lewis, and

Tessada 2019; Williamson 1995). Latin America and particularly Brazil, remain understudied even though

Brazil was the third-most popular overall and the second-most popular destination after the United States

in the 1890s (Ferenczi and Willcox 1929; Sánchez-Alonso 2019). Within the relatively small literature on the

Brazilian experience of the Age of Mass Migration, this paper is most closely related to Lanza, Manier, and

Musacchio (2023), who have found evidence of a linkage between the subsidized immigration program and

coffee productivity in the state of São Paulo. Beyond this, most research focuses on the long-term effects

of immigration on income and human capital accumulation in specific regions of Brazil (e.g., de Carvalho

Filho and Monasterio 2012; Klein 1995; Stolz, Baten, and Botelho 2013; Witzel de Souza 2018). This paper

provides the first causal study of the effects of immigration on Brazilian agriculture in the Age of Mass

Migration that covers the entirety of Brazil, and the whole of the immigration flow, whether subsidized or

unsubsidized, contributing to the long-standing debate about the role of immigration to the development of

the largest Latin American economy (e.g., Alston et al. 2016; Holloway 1980; Papadia 2019).

2 Background

2.1 The Effects of Immigration in Developing Countries

Understanding the effects of immigration on the economies of receiving countries is a fundamental goal of

the economics of immigration. Canonical studies focus on the labor-market effects of immigration, largely

in the United States after the transition to its current immigration regime in 1965 (e.g., Abramitzky and

Boustan 2017; Borjas 2003, 2014; Card 1990, 2005; Hanson 2009; Kerr and Kerr 2011), with a smaller set

focusing on the effects of immigration in other modern developed countries (e.g., Dustmann, Schönberg, and

Stuhler 2016; Manacorda, Manning, and Wadsworth 2012). This literature is enriched by studies focusing

on the effects of immigration in the United States during the Age of Mass Migration (e.g., Abramitzky et al.

2023; Ager et al. 2021; Cohen and Biddle 2022; Hatton and Williamson 1998; Price, vom Lehn, and Wilson

2023; Tabellini 2020), which exploit a variety of advantages of this historical setting, such as the long time

horizon, open borders, the ability to follow individuals over time, and differences in economic structure to

important feature of Brazilian immigration—was considered but was ultimately banned by the Foran Act in 1885. In the later
years of the Age of Mass Migration, southern and eastern European immigrants were criticized for their supposed unwillingness
to settle in the US South (Benton-Cohen 2018; Zimran 2022). As a result, what may have occurred if these large flows had
materialized, either spontaneously or if subsidized immigration had come to fruition, remains unknown. This paper sheds light
on what might have been.

5
shed light on the economics of immigration in general.11

Although these studies have enriched our understanding of the effects of immigration on receiving

economies, the focus of nearly all of them is on rich countries and largely on flows from poor to rich countries.

This is true in the case of studies of the modern United States, which draws the bulk of its migrants from

Latin America and Asia (Abramitzky and Boustan 2017). It is also true of studies of the United States in

the Age of Mass Migration. Most immigrants to the United States in this period were from countries with

substantially lower wages (Williamson 1995). Moreover, although it was more agricultural than it is now,

the US economy was far more industrialized than that of any other major origin or receiving country of the

Age of Mass Migration (Abramitzky and Boustan 2017; Hatton and Williamson 1994; Hatton and Ward

2019).12

The ultimate result of the focus on the United States in both the modern and historical literatures is that

migration to developing countries is poorly understood, particularly its effects on the receiving economies.

Indeed, with a small number of exceptions (e.g., Biavaschi et al. 2018; Gindling 2009; Lanza, Manier, and

Musacchio 2023; Özden and Wagner 2014), previous literature focuses on developing countries as the source

rather than the destination of migrants (Hanson 2009). A corollary of this, given the agricultural nature of

many developing economies, is that the effects of immigration on the agricultural sector are poorly understood

relative to other sectors (c.f., Abramitzky et al. 2023; Clemens, Lewis, and Postel 2018; Lanza, Manier, and

Musacchio 2023). Moreover, there are a number of reasons to suspect that the effects of immigration in

developing countries are different from those in developed countries. All of the factors that combine to

determine these effects—the origins and selection of immigrants, their substitutability with natives, the

sectors comprising their economies, and the duration of immigrants’ stays in the destination, among other

factors—are likely to be different. As a result, studies of the effects of immigration in a developing country

can, both in isolation and in comparison to the effects in rich countries, help to advance the empirical and

theoretical understanding of the economics of immigration.

11 Related to this is work by Clemens, Lewis, and Postel (2018) and Lee, Peri, and Yasenov (2022) on the Bracero program

(1942-1964), which was a conduit for substantial migration from Mexico to the United States. Though not perfectly comparable
to the Age of Mass Migration, this phenomenon also provides a setting of relatively unencumbered migration from which to
gain insights on the economics of migration. Also related are studies of the long-run effects of historical migration on modern
outcomes (e.g., Burchardi, Chaney, and Hassan 2019; Sequeira, Nunn, and Qian 2020).
12 Indeed, the bulk of studies examining the effects of immigration in the United States during the Age of Mass Migration

focus on the twentieth century, or even on the end of the Age of Mass Migration with the imposition of country-of-origin
quotas. Thus, although the United States was strongly agriculturally focused in the early Age of Migration, this is not the
period generally studied when focusing on the effects of immigration.

6
2.2 Brazil in the Age of Mass Migration

Over 3.5 million European immigrants entered Brazil between 1850 and 1920; Figure 1 plots the time series of

Brazilian immigration. Nearly 1.5 million of these arrived between 1888—when the government implemented

a subsidized immigration program, in large part to provide labor for coffee cultivation after the abolition

of slavery—and 1900. This made Brazil the second most popular immigrant destination in the New World

in this period, trailing only the United States (Ferenczi and Willcox 1929, p. 550).13 The largest single

group of immigrants in this period were from Italy, where immigrants were recruited for subsidized labor

contracts from the north of the country. Over 80 percent of immigrants were subsidized in the pre-1900

period, possibly reflecting the limited attractiveness of Brazil as a destination due to its relatively low wages

and living standards, the fresh memory of slavery, and the perceptions of an adverse disease environment

(Papadia 2019).

The colono contract was at the center of the subsidized immigration program, which was particularly

important in São Paulo as the epicenter of coffee production (Lanza, Manier, and Musacchio 2023).14 Under

this scheme, immigrants received free passage and lodging, and were responsible for caring for and harvesting

coffee trees for a three-year contract period. In order to reduce the incentive for return migration, the colono

contract required adult male immigrants to be accompanied by their families (Klein 1995). However, the

evidence shows that this requirement was not strictly enforced, as about 35% of all immigrants processed at

São Paulo from 1911 to 1920 were returnees (Holloway 1980, p. 56). Lesser (2013) documents that in order

to receive the subsidy, single migrants that had met during or just after the voyage reported to be a “family,”

which may help explain why only 46 percent of immigrants remained permanently in the state of São Paulo

(Sánchez-Albornoz 1986).

The immigrant experience in Brazil began at reception centers, known as Immigrant Hostels (Hospedarias

dos Imigrantes), which were present in each of the three main entry ports of Santos in the state of São Paulo,

Rio de Janeiro, and Salvador in the state of Bahia. At these hostels, immigrants could sleep, eat, and, if

necessary, receive medical attention or vaccinations after being registered. The hostels also had direct access

to railways, which facilitated the transportation of immigrants to their final destination. The São Paulo

hostel, which processed most migrants, was the venue for the matching of immigrants arriving under the

subsidy program and the coffee plantations. In general, these immigrants had one week after arrival to find a

13 European immigration to Brazil began with the establishment of agricultural colonies in southern Brazil by German and

Italian immigrants in the 1820s (Foerster 1919). There was also a wave of immigration of Americans from the US South in the
wake of the Civil War (Dawsey and Dawsey 1995) and efforts made to recruit European sharecroppers to grow coffee after Dom
Pedro II’s crackdown on the illegal Atlantic slave trade in 1850 (Barman 1999).
14 Subsidized immigrants were also recruited for mining in Minas Gerais.

7
farm on which to work, and upon signing a contract would be provided with a train ticket to their destination

(Holloway 1980; Lanza, Manier, and Musacchio 2023).

Brazilian immigration changed substantially at the beginning of the twentieth century. Concerns about

the conditions of migrants in Brazil, particularly of those residing in colonies and working on coffee planta-

tions, led to bans on subsidized migration by several European governments (Lesser 2013). Most notably, the

Italian Decreto Prinetti in 1902, which made subsidized emigration illegal, was targeted mainly at limiting

migration to Brazil.15 In part as a result of these bans, subsidized migrants no longer made up the majority

of entrants to the country, with their share falling to about 40 percent (Cameron 1931). Although these bans

did not halt mass migration to Brazil, as Figure 1 shows, they did lead to a change in the origin of migrants.

In particular, Figure 2 shows a decline in the share of immigration coming from Italy with a commensurate

increase in the share from Portugal and Spain.16 The shift in migrant origins was coupled with changes

in settlement: whereas immigrants arriving before 1900 were almost entirely directed to agricultural areas,

those entering after 1900 settled in large numbers in cities in the states of São Paulo and Rio de Janeiro

(Foerster 1919, p. 289).

Although it is well established that immigrants were more numerate and literate than the Brazilian

population (Rocha, Ferraz, and Soares 2017; Stolz, Baten, and Botelho 2013), direct evidence on the selection

of immigrants is limited. It is possible that subsidized immigrants were negatively selected, as subsidies

significantly relaxed liquidity constraints, allowing the relatively poor and unskilled to migrate (Angelucci

2015; Belot and Hatton 2012; Chiquiar and Hanson 2005; McKenzie and Rapoport 2010; Orrenius and

Zavodny 2005; Spitzer and Zimran 2018). Beyond this speculation, the few studies addressing the selection

of Italian and Portuguese migrants, the two major immigrant groups in Brazil, provide mixed evidence.17

Despite policymakers’ efforts, return migration was prevalent. Italian data on return migration, which

are available from 1905 onwards, indicate that between 50 and 75 percent of immigrants returned to Italy.18

15 Prussia also nominally prohibited emigration to Brazil as early as 1859, and similar measures were implemented for the

whole German empire from 1871 (Fausto 1999).


16 There was also a change in the main immigrant sources within Italy, with migrants before 1902 coming primarily from

the north and those after coming primarily from the south. See Online Appendix Figure A.1 for the division of immigrants by
origin in the 1920 census.
17 Fernández-Sánchez and Tortorici (2023) show that Portuguese migrants, who mostly moved to Brazil, were on average

positively selected on the basis of literacy. Determining the selection of Italians is more difficult given the highly segmented
destination choice patterns of this group (Hatton and Williamson 1998; Spitzer and Zimran 2023). But given the early dominance
of northern Italy in the migratory flow to Brazil, due in part to the fact that migration subsidies were offered exclusively to
migrants from the north of Italy (Hatton and Williamson 1998, p. 102), and that migrants from this region to the United States
were negatively selected (Spitzer and Zimran 2018), it is likely that the early migrants were negatively selected. Hatton and
Williamson (1998, p. 121) also find that farmers from Italy, who tended to be relatively poor, were more likely to travel to
Brazil than they were to the United States, possibly in search of land. However, the shift of Italian emigration to a largely
southern-Italian phenomenon at the turn of the twentieth century may have been associated with a more positive selection, as
migrants from this region who emigrated to the United States were positively selected (Spitzer and Zimran 2018, 2023).
18 We have not been able to locate similar data for the other main immigrant source country, Portugal.

8
Foreign observers widely believed that those who remained were exploited under the colono contract (Lesser

2013), fueling the bans on subsidized immigration. Recent scholarship has debated whether this was true,

providing a more nuanced story: immigrants were able to transition to land ownership after a few years

on the fazenda (Klein 1995; Lanza 2021). More specifically, Holloway (1980, p. xvi) documents that many

first-generation immigrants were likely to become owners of small and medium-sized farms. However, the

specific mechanisms through which immigrants achieved this transition have not been identified.

Altogether, Brazil’s immigration and economy bear some important similarities to those of the other

major migrant destinations—the prevalence, at least from the later nineteenth century, of migrants from

the poorer European periphery, the high rate of return migration, and, in Argentina’s case, the size of

the agricultural sector (Droller and Fiszbein 2021; Lesser 2013). But many features set it apart. Most

dramatically, there were substantial differences in development between Brazil and the other two major

migrant-receiving countries in the Americas, as shown in Figure 3: Brazil was, by a substantial margin,

the poorest of the three countries (Bolt and van Zanden 2020); indeed, Argentina was one of the 10 richest

countries in the world at the eve of World War I and had living standards similar to those of the United

States (Spruk 2019). In addition, the concentration of Brazilian immigrants in agriculture was unique

in comparison to Argentina and the United States, where most immigrants provided labor outside the

agricultural sector (Pérez 2017): in Brazil, approximately 40 percent of immigrant men worked in agriculture

in 1920 (Online Appendix Figure A.2) as opposed to about 15 percent in the United States and 17 percent in

Argentina in 1895. Subsidized migration was also uniquely important in Brazil. The United States explicitly

banned subsidized migration beginning in 1885, and although there is evidence of subsidized immigration in

Argentina, it did not have the relevance it did in Brazil (Lesser 2013). Brazil also had weaker institutions

and worse governance, reflected in low access to justice and an inefficient public sector, as well as a high

concentration of economic power and a highly unequal land distribution (Naritomi, Soares, and Assunção

2012).19 Understanding the impacts of immigration in this context can thus provide new insights beyond

those coming from studying the effects of immigration in the United States and even the more comparable

Argentina.

2.3 Existing Research on the Effects of Brazilian Immigration

Research addressing the effects of immigration in Brazil dates at least to Dean’s (1969) classic account of

the industrialization of the state of São Paulo, which had become the industrial engine of Brazil by 1920
19 While substantial inequality was present in Argentina in the early twentieth century, it was not nearly as extreme as in

other Latin American societies (Sokoloff and Engerman 2000).

9
(Palma et al. 2021). He assigns a direct role to migrants, but the crucial group of immigrants in this account

is the bourgeois immigrant—relatively wealthy and skilled, and mostly active in industry and commerce—

rather than the modal immigrant workers in agriculture. Conversely Dean (1969) argues that the bourgeois

immigrants in the agricultural sector—both those who started off as landowners and those who became

landowners at a later stage—had no detectable generalized positive effect on Brazilian local development

because of an insurmountable advantage of domestic planters, potentially arising from superior knowledge,

better access to capital, and the occupation of better land.

Among the more recent literature on the Brazilian experience of the Age of Mass Migration, Lanza,

Manier, and Musacchio (2023), the work most closely related to ours, provide the clearest evidence on

the effects of immigration. They use the same farm-level data for the state of São Paulo that we use

(though aggregated to the municipality level in their case) to study the effect of immigration as part of the

subsidized immigration program on agricultural output and capital adoption. Arguing that the assignment

of immigrants to municipalities within this program was random and comparing municipalities with a greater

immigrant share to those with a smaller share, they find that a greater share of immigrants was associated

with greater coffee output per farm and the adoption of more agricultural tools in 1920.

Other studies of the effects of Brazilian immigration focus in large part on the long-run effects of immigrant

human capital or of immigrant colonies and schools established by or at the behest of immigrants (de Carvalho

Filho and Colistete 2010; de Carvalho Filho and Monasterio 2012; Rocha, Ferraz, and Soares 2017; Stolz,

Baten, and Botelho 2013; Witzel de Souza 2018).20 These studies also tend to focus on only a portion of the

Brazilian immigration experience, limiting themselves to either a single source country (e.g., Witzel de Souza

2018) or region of Brazil (e.g., de Carvalho Filho and Monasterio 2012). More broadly, the understanding

of the effects of immigration in Brazil tend to focus on immigration’s connection to coffee production in São

Paulo (Holloway 1980), leaving the role of immigration in the development of other regions and sectors less

clearly understood. In particular, what remains lacking is an understanding of the effects of both subsidized

and unsubsidized immigration on contemporary outcomes covering all of Brazil and addressing directly the

potential for endogenous location choice by immigrants. This paper contributes such a study.

20 Feler, Musacchio, and Reis (2016) study the effects of immigration on banking in the 1940s and 1950s. Tang and Monteiro

(2023) also study Japanese immigration as an instrument for education.

10
3 Conceptual Framework

The expected effect of immigration on Brazil’s agricultural sector is theoretically ambiguous. The simplest

way in which immigration could impact this sector is by simply adding to the agricultural labor force. In

an environment in which the agricultural frontier was rapidly expanding, a shortage of labor could prevent

land from being cultivated; in such a situation, additional labor provided by immigration can increase the

value of land that previously could not be cultivated. Such a force would have been particularly strong

in this environment in which the valuable coffee sector was growing and demanding increasing amounts of

labor. Such an effect need not prevail, however, if, for instance, native and immigrant labor were perfectly

substitutable and natives relocated within Brazil in response to the arrival of immigrant labor.21 Similar

arguments could be made for immigrants increasing the demand for land and investing in farm capital in

the form of tools and machines where land would otherwise be uncultivated or in lower demand.

The predicted effect of immigration is complicated by the likelihood that immigrants and natives were

fundamentally different from one another, and thus not perfectly substitutable. For instance, immigrants

may have possessed certain characteristics that made them particularly productive agricultural workers, such

as specialized agricultural knowledge or greater human capital. If this were the case, a greater concentration

of immigrants would spur local agricultural development, which would ultimately be reflected in higher farm

values. Similarly, immigrant labor’s complementarity with capital may have differed from that of natives.

Importantly in the Brazilian context, immigrants may have uniquely enabled the opening of land to coffee

production.

Immigrants may also have faced different incentives than natives. For instance, temporary migrants, who

made up a substantial share of migrants to Brazil in this period, may have been incentivized to exert greater

labor effort than natives or permanent immigrants while in Brazil, which would in turn contribute to greater

land values where they settled. Such a mechanism is consistent with evidence that temporary immigrants

substitute inter-temporally, exerting greater labor effort in the destination and enjoying greater leisure at

home (Dustmann 1994, 2000; Dustmann, Bentolila, and Faini 1996; Dustmann and Görlach 2016; Epstein

and Venturini 2011; Galor and Stark 1991; Hill 1987; Holloway 1980; Klinthäll 2006; Kyarko and Chartouni

2017; Vijverberg and Zeager 1994; Wahba 2022).

The effects of immigration on sectors outside of agriculture in a largely agricultural economy are also

21 In such a situation, immigrant arrivals could lead to the increase of land values elsewhere in Brazil, if, for instance,

immigrant arrivals led natives to relocate to the frontier where labor was previously scarce. Unfortunately, no identification
strategy based on the comparison of different municipalities within Brazil can identify such an effect absent detailed data on
internal migration, which do not exist.

11
theoretically ambiguous. It has been posited that greater agricultural productivity could encourage structural

transformation (Johnston and Mellor 1961; Montero and Yang 2022; Timmer 1988), but it has also been

suggested that it can lead to specialization in primary production, discourage human and physical capital

accumulation, hinder the development of other sectors of the economy, and ultimately delay structural

change (Matsuyama 1982). The existing empirical literature is equally ambiguous (Asher et al. 2022; Bustos,

Caprettini, and Ponticelli 2016; Bustos, Garber, and Ponticelli 2020; Foster and Rosenzweig 1996; Hornbeck

and Keskin 2015), suggesting that the effect of agricultural development may be strongly context-dependent.

In the Brazilian case, one body of literature argues that industry only started to develop rapidly once the

agricultural export sector, led by coffee, was disrupted by events such as the First World War and the Great

Depression. Another, however, has argued that the development of the agricultural export sector was a

crucial precursor for the development of other sectors of the economy, particularly industry (Suzigan 2000).

The ambiguity of theory and existing empirical evidence regarding the effects of immigration on an

agricultural developing economy, and in turn on that economy’s structural transformation, implies that

these effects are ultimately empirical questions. Our analysis seeks to answer these questions, guided in our

search for potential mechanisms and effects by the theory outlined above.

4 Data

Our analysis is based on data we digitized for this project from the 1920 Population and Agricultural

Census of Brazil (Directoria Geral de Estatística 1922). The 1920 census was the first complete population

census successfully carried out in Brazil since 1872, and the first ever agricultural census covering the whole

country. While this means we do not have access to repeated data over time, the census provides a very

rich set of variables at the municipality level that capture economic, population, human capital, and labor

force characteristics.22 We also use complementary data from a variety of other sources, mainly as control

variables throughout the analysis, as well as farm-level data from the state of São Paulo to bolster some of

our main empirical results.

4.1 Outcome Variables

The 1920 census provides data on the average monetary value of farms (agricultural establishments) by mu-

nicipality in milréis—the currency of Brazil at the time—in total and broken down into three components—

22 Note that we exclude Acre from the analysis, as data for this territory are not consistently reported.

12
land, infrastructure, and tools and machinery.23 These establishments were often made up of a single plot

of land, but could also refer to multiple plots in the same municipality managed by the same person, group

of people, or organization (e.g., the government).24 Combined with information on the size of farms, these

data enable us to compute our main outcome variable—the average value of farms per hectare of land. We

consider farm values per hectare to be a measure of agricultural development, which includes any factor

making the local agricultural sector more productive and valuable. One concern with the use of farm val-

ues as a measure of agricultural development is that they include the value of land, which in turn would

capitalize local characteristics and amenities potentially unrelated to immigration. Fortunately, the other

two components of farm values—improvements and tools and machines—are strictly related to agricultural

productivity, and thus ensure we are indeed capturing the effects of migration on agricultural development.

Moreover, our control variables are designed to capture a broad range of additional factors affecting farm

values, and our identification strategy addresses the potential for endogenous location choice by immigrants.

We also collect data on a series of factors that may have affected the value of farms through their

relationship with immigration. The first set of variables covers the state of the labor force: we use population

density (inhabitants per km2 ) and agricultural employment density (workers employed in agriculture per

km2 ). As measures of land use, we use data on the share of farm land cultivated and the share of cultivated

farmland by planted crop.25 We also measure the use of tools and machines in agriculture: specifically we

use data on the share of farms with tools—plows, harrows, seeders, cultivators, harvesters, and tractors—and

machines in the form of devices employed for processing crops.26

As previous literature has shown that agricultural development, or its absence, can influence the transfor-

mation of the economy (e.g., Abadie, Gu, and Shen 2023; Asher et al. 2022; Bustos, Garber, and Ponticelli

2020; Hornbeck and Keskin 2015; Lewis 1954; Matsuyama 1982; Montero and Yang 2022; Schultz 1964;

Timmer 1988), we also examine four sets of outcome variables related to structural change. To measure hu-

man capital formation, we use data on literacy—the share of individuals who could read and write—which

23 Directoria Geral de Estatística (1922, Volume III, 1a Parte: Agricultura, pp. 298-385)
24 As defined in the census, agricultural establishments are “the whole extension of land subject to the exclusive administration
of an owner, tenant, stakeholder or administrator, who directly manages the cultivation of crops or livestock by themselves or
with the help of paid staff” (Directoria Geral de Estatística 1922, Volume III, 1a Parte: Agricultura, page 7). Land cultivated
in urban settings was excluded from the census, as were farms with an annual production worth less than 500 milréis. This
essentially means excluding establishments practicing subsistence agriculture, and focusing, instead, on commercially oriented
farms. To give some perspective, 500 milréis was approximately 107 times the average daily wage of a plough-man living in a
rural area in 1920 (Directoria Geral de Estatística 1922, Volume V, 2a Parte: Salarios, p. XXV).
25 These crops include rice, maize, wheat, beans, potatoes, manioc, cotton, sugarcane, tobacco, castor beans, coffee, cocoa,

coconut, and rubber.


26 Directoria Geral de Estatística (1922, Volume III, 3a Parte: Agricultura, p. 17–105.). A primary examples are machines

used for processing and distilling sugar in on-farm, refineries.

13
the census reports separately for immigrants and natives.27 As indicators of economic structure, we use

data on the share of workers employed in agriculture, in industry, and in the public sector. We also use

the population share of individuals living on income from property or investments (rentiers) as a measure

of the presence of local landed elites.28 To measure state capacity, we use data on public finances, including

(export) tax revenue per capita and expenditure per capita in both education and public services. Finally,

we collect data on the female to male ratio among the employed population by economic sector (agriculture

and industry).

4.2 Explanatory Variables

We use data on the total population and number of European immigrants by municipality to construct

our main explanatory variable—the population share of European immigrants by municipality.29 We focus

on European immigration for two reasons. First, the historical literature has documented that, during the

Age of Mass Migration, nearly 90 percent of the immigrant flows to Brazil came from Europe (Lesser 2013;

Sánchez-Alonso 2019). Second, the exogenous variation that our identification strategy relies on is partially

the product of shocks to European immigration.

Our identification strategy also relies on variation in railway access at the local level. We obtain these

data from Giesbrecht (2023), which allow us to identify the first station built in each municipality as well as

the year when its construction was completed—that is, the year when each municipality was linked to the

railway network. We then compute two variables with this information—the number of years that a given

municipality had been connected to the railway network by 1920 and an indicator variable for municipalities

that were not connected by 1920.

4.3 Control Variables

We collect data on a number of municipality characteristics that may have affected farm value and immigrant

settlement to include as control variables in our analysis. To capture proximity to international and domestic

markets, we create three variables—distance to the nearest port or frontier custom house, distance to the

nearest principal city, and distance to the nearest principal town.30 We also collect data on the location
27 Directoria Geral de Estatística (1922, Volume IV, 1a Parte: População, pp. 20–481)
28 Directoria Geral de Estatística (1922, Volume V, 5a Parte: População, Tomo I pp. 180–625, Tomo II pp. 6–825.)
29 Directoria Geral de Estatística (1922, Volume IV, 1a Parte: População, pp. 550-887). The computation of this variable im-

plies excluding immigrants mainly from Asia (Japan) and South America (Argentina, Paraguay, and Uruguay), who represented
about 12 percent of the immigrant population. The exact text of the immigrant data in the original source is “Popoulção es-
trangeira do estado de XX segundo a nacioladidade e o sexo, inclusivo os estrangeiros que adoptaram a nacionalidade braziliera.”
30 These features are identified using a map of Brazil created by the International Bureau of the American Republics (IBAR)

in 1905, shown in Online Appendix Figure A.3.

14
of immigrant colonies from the same source and from Gagliardi (1958) for the state of São Paulo. We also

create a battery of variables to control for differences in geographic conditions across municipalities, including

surface area, ruggedness, altitude, latitude, longitude, and the interaction of latitude and longitude. Finally,

we use data from the Global Agro-Ecological Zones project (Food and Agriculture Organization 2021) to

construct three variables that capture the suitability of land for agriculture and the adaptability of land for

the production of different crops. These variables consist of the first two principal components of suitability

for all major crops reported in the census except rubber,31 and the Herfindahl-Hirschman Index (HHI) of

suitability, capturing how concentrated land suitability is in each municipality, with lower values indicating

a greater adaptability of land for the production of different crops.

4.4 Farm-Level Data

We complement our municipal-level analysis with data from the 1904–1905 Agricultural Census of the state

of São Paulo (Secreteria da Agricultura 1906–1910). The data from this source are similar to those coming

from the 1920 census, reporting the share of cultivated farm land, the share of cultivated farm land by

planted crop, employment density (workers per hectare), and the breakdown of employment by foreign or

native birth. But these data are reported at the level of the farm rather than the municipality. In total,

we have information for over 40,000 farms across 163 municipalities. This source provides information on

land values, as well as farm ownership—that is, whether the farm was owned by a foreign- or native-born

person—and on the share of native and foreign workers.32

4.5 Summary Statistics

Online Appendix Tables A.1 and A.2 present summary statistics for all of the variables in our dataset. Figure

4 presents maps displaying the geographic distribution of our main outcome variable (panel a) and our main

explanatory variable of interest (panel b). The south and southeast are shown at increased magnification,

as these regions received the majority of migrants. A number of features of our data are readily apparent.

The first is that there is a concentration of high farm values and high immigrant shares in the vicinity of

São Paulo, in the main coffee growing area of the southeast. We will present a number of exercises verifying

that our results are not driven by this region. The second is that there was considerable variation in both

31 Specifically, this is the low input, no irrigation land suitability data for beans, cocoa, coconut, coffee, cotton, maize, potato,

rice (dry and wet), sugar, tobacco, and wheat.


32 The digitized data were kindly shared with us by Renato Colistete. See Bassanezi and Francisco (2003) (who first digitized

the data), Colistete (2015), and Luna, Klein, and Summerhill (2014) for previous uses of this source.

15
farm values and immigrant shares throughout the country.

Panel (a) of Figure 5 presents the geographic evolution of the Brazilian railway network over time—part

of the variation that contributes to our instrumental variables strategy. The expansion of the network inland

from major ports over time is clear.33 Notably, there appears to have been an important regional component

to the rail network. Rail was virtually absent from the North, and evolved into several geographically

distinct networks rather than one unified one. For this reason, and because differences across regions in

migrant settlement patterns, we posit, verify, and exploit the fact that the impact of rail on immigrant

settlement patterns may have varied by region.

5 Empirical Strategy

Our main estimating equation for the effect of immigration on farm value is given by

FarmValuei = ImmSharei + X0i ⇧ + j + "i ,

where FarmValuei is the average value per hectare of farms in municipality i, ImmSharei is the population

share of European immigrants in municipality i, j are region or state fixed effects,34 and Xi is a vector of

municipality-level covariates that we use to control for a number of factors that may influence farm value,

such as local geographic conditions—ruggedness, altitude, surface area, latitude, longitude, the interaction

between latitude and longitude, and metrics capturing the quality of land for agricultural activities—and

the linear and quadratic distance to the nearest port or custom house, principal city, and principal town.35

We also control for the years a municipality was connected to the rail network and whether it had not yet

been linked by 1920. Our main results are reported with robust standard errors, but all results are robust

to correcting for spatial correlation (Colella et al. 2019; Conley 1999), as we show in Online Appendix B.

Regardless of the richness of our controls, estimating the causal impact of immigration on farm value is

complicated by the likely endogeneity of immigrants’ destination choices.36 For instance, immigrants may

have settled in places with better land quality (beyond our ability to control for it), or where other factors,
33 The first railway station was completed in 1854. While the railway network expanded significantly into the interior starting

from the last decade of the 19th century (Figure 5), Brazil’s railway mileage in 1914 was equivalent to that of the United States
in the 1850s (Herranz-Loncán 2014; Summerhill 2003).
34 Online Appendix Figure A.4 presents a map of political divisions of Brazil, showing the country’s regions and the states

that each contains.


35 These controls address identification concerns such as that immigrants may have settled in places with more fertile land.
36 Notably, Lanza, Manier, and Musacchio (2023) argue that the placement of immigrants by the official immigrant recruit-

ment system was random, which contributes to the validity of our OLS regressions. But these represented only a portion of
the immigrant arrivals—which also included unsubsidized immigrants and immigrants to other states—whose effects we are
interested in understanding.

16
such as better management, were available. Immigrants’ settlement patterns may also have been influenced

by local elites, whose presence could also have impacted agricultural productivity.

The direction of the resulting bias, however, is unclear. A natural concern is that our estimates would

overstate the effect of immigration if immigrants were to disproportionately settle in areas where local char-

acteristics were responsible for greater land values. But the available historical evidence suggests instead

that immigrant settlement patterns were such that our estimates might instead be biased downwards.37 In

particular, there is substantial historical evidence indicating that migrants often did not settle in the most

economically favorable locations. One reason is that migrants had very little information about the final

destination of their migration. Even more strikingly, they sometime had little say in where they eventually

settled within a state or broader region, needing to choose a destination based on labor demand during their

one-week stay at the immigrant hostel (Lanza, Manier, and Musacchio 2023). This lack of choice and the

limitations to free movement within Brazil for some time after arrival in the country, as well as frequent

maltreatment by landowners, manifested themselves in widespread discontent on the part of the migrants,

especially in the earlier phases of mass immigration (da Costa 2000; Fausto 1999). Immigrant colonies—a

Brazilian peculiarity, which saw the state cooperate with private planters to create rural settlements for

migrants—also highlight difficulties in the migrant experience in Brazil. Although research has shown that

municipalities featuring such colonies experienced faster development later on (de Carvalho Filho and Monas-

terio 2012; Rocha, Soares, and Ferraz 2017), this type of migration was not generally seen as particularly

fruitful by contemporaries. One reason was the perception that the migrant colonies were not located in

economically favorable locations (Cameron 1931).

Migrants may also have been incentivized to settle in areas predisposed to have lower land values. For

instance, some migrants may have chosen to settle in less economically dynamic areas with lower land values

(to the extent they had a choice) in exchange for the prospect of an easier access to land ownership. Historians

have argued that the prospect of land ownership was indeed a strong pull factor for migrants, especially for

the largest immigrant group, Italians (Holloway 1980). Seeking out lower land prices was likely also salient

for secondary migration following initial settlement.

Nonetheless, the potential for endogenous location choice must be addressed. To overcome the identi-

fication challenge, we implement an instrumental variable strategy based on the interaction of immigrant

inflows and the expansion of the Brazilian rail network, following Sequeira, Nunn, and Qian’s (2020) study

of the effects of immigration during the Age of Mass Migration in the United States. The intuition of this
37 Of course, measurement error in our regressor of interest could also be responsible for attenuating our estimates and would

be addressed by instrumentation.

17
instrument is to compare two otherwise identical municipalities, one of which was linked just before a year

of large immigrant inflows and the other of which was linked just after. Because of the importance of rail in

linking migrants to their destinations, they would be more likely to settle in a municipality that was linked

to rail at the time of their arrival. The municipality linked to rail just before the large immigrant inflow

would thus receive immigrants from that wave, while that linked in the following year would not. These

initial settlement patterns would then affect subsequent settlement patterns by creating migrant networks

that subsequent immigrants might follow into these destinations (Spitzer and Zimran 2023). For this reason,

even temporary migration—important in light of the high rate of return migration—would be affected by

this variation.

In particular, we construct an instrument for ImmSharei , which we refer to as AvgImmSharei , of the

form
1920
1 X
AvgImmSharei = ImmFlowt ⇥ Railwayit 1, (1)
✓i t=1855

where ImmFlowt is the immigration flow to Brazil year t, normalized by Brazil’s population size, Railwayit 1

is an indicator variable for the presence of a train station in municipality i in year t 1, and ✓i is the number

of years that municipality i had been connected to the railway network by 1920.38 This equation captures a

number of refinements that we make to Sequeira, Nunn, and Qian’s (2020) identification strategy. The first

is that we use annual data on the state of the rail network rather than data by decade.39 We also make a

slight departure by dividing by the number of years that a municipality was linked to the rail network (✓i )

rather than the number of years in the study period, though our results are robust to using Sequeira, Nunn,

and Qian’s (2020) normalization.40 Dividing by the number of years linked gives a sense of the number of

immigrants that we might expect to observe at a given point in time, which is what we will see in the 1920

census.

Given the potential for heterogeneity in the predictive power of the instrument in its first stage, we

interact the instrument with region indicators to exploit variation in the strength of first-stage identification

38 In Online Appendix C, we show that our results are robust to an alternative definition of rail connectedness in which a
municipality is considered connected to the rail network if another municipality whose centroid is within 100km of its own has
a rail station.
39 We focus on railway linkage in year t 1 rather than in year t to ensure that municipality is linked for the entire year
rather than only some, potentially small, fraction.
40 These results are available upon request.

18
across regions.41 The first-stage regression equation for our IV strategy is thus

ImmSharei = j AvgImmSharei + Xi0 ⌦ + j + ui ,

where j is a region-specific coefficient. We also show that our results are robust to exploiting variation at

the state level in Online Appendix D.42 The value of the instrument for each municipality is presented in

Figure 5(b). While the geographic extent of the non-zero instrument value matches that of the railroad (by

construction), as shown in panel (a), this map also makes clear that the variation across space is not precisely

the same as that in panel (a)—a fact arising from the incorporation of immigrant inflows in a particular year

into the creation of the instrument.

The controls most crucial to our identification are, following Sequeira, Nunn, and Qian (2020), the

number of years that a municipality was connected to the railway network by 1920 and an indicator variable

for municipalities that were not connected by 1920.43 These controls address the obvious concern that rail

linkage may have had direct effects on economic activity (Summerhill 2005), and that the location of rail

construction was not random, likely targeting areas where economic activity was or would be greater (e.g.,

Atack et al. 2010; Donaldson and Hornbeck 2016; Zimran 2020).

As in the case of Sequeira, Nunn, and Qian’s (2020) application of this instrumentation approach, per-

haps the main identification concern facing our instrument is that municipalities connected to the railway

network during an immigration boom may have been systematically different from those connected during

an immigration lull. Such differences could arise if, for example, aggregate immigrant flows increased when

locations with greater economic potential were connected to the railway network. This concern is mitigated

by the fact that the historical literature has documented that these fluctuations were influenced by a number

of global and national macroeconomic factors, including changes in the price of coffee, the increase in labor

demand due to the abolition of slavery, the implementation of the Decreto Prinetti in 1902, the First World

41 Abadie, Gu, and Shen (2023) show that limiting attention to subsamples in which the instrument does have a strong first

stage can result in bias. They also show, however, that the method that we implement, also used by Deryugina et al. (2019),
Dix-Carneiro and Kovak (2017), Jackson, Johnson, and Persico (2016), and Pascali (2017), provides more reliable estimates.
Abadie, Gu, and Shen (2023) caution that this method may be problematic where there are many different subsamples, but
our case, with five regions in Brazil, is unlikely to face such issues.
42 Our main results include a demonstration that our results are robust to controlling for state fixed effects.
43 As in Sequeira, Nunn, and Qian’s (2020) application of this identification strategy, including these controls implies that the

identification technically arises from the functional form restriction in which the control for years of rail linkage is linear whereas
the instrument is a non-linear function of a municipality’s years of rail linkage (since the year of linkage is the sole determinant
of years of linkage). But, again as in Sequeira, Nunn, and Qian’s (2020) application, the non-linearity of the instrument is
not arbitrary, but is instead the product of actual variations in Brazilian immigration rates. In addition, our normalization by
years linked to the network provides further non-linearity. The combination of the linear control for years of linkage, the control
for whether a place was ever linked, and an instrument in which the variation is determined by immigrant arrivals is simply
to ensure that the identification derives not from how long a place was linked to the rail network but only from when in the
immigration cycle it was linked. We illustrate this nonlinearity graphically in Online Appendix Figure A.6.

19
War, and global macroeconomic shocks such as the Panic of 1907 (Hatton and Williamson 1998; Sequeira,

Nunn, and Qian 2020; Spitzer 2015; Spitzer, Tortorici, and Zimran 2022). To address these concerns more

formally, we follow Sequeira, Nunn, and Qian (2020) by comparing the observable characteristics of locations

that became linked during immigration lulls to those linked during immigration booms.44 We operationalize

this test by defining booms as years with an immigrant inflow above the previous five years’ moving average

and defining lulls conversely. The booms and lulls are identified in Figure 1. Columns (1)–(3) of Table

1 present the means of the observable characteristics for the full sample and for each group of locations.

Column (4) tests for the statistical significance of differences between the groups, finding that the two sets of

municipalities were, in fact, systematically different, though these differences were, for the most part, with

respect to geographic variables, capturing the gradual rollout of the railway from the coast. But columns (5)

and (6), which repeat the analysis controlling for region and state fixed effects respectively, show that the

differences are largely explained by the linkage of different parts of the country at different times, and that

within region and states the differences are statistically significant in only a small number of cases. Thus,

the evidence does not support concerns that railway expansion may have responded to immigrant inflows or

vice versa.

6 Main Results and Robustness

Table 2 presents our main results, estimated both by OLS (panel A) and using the instrumental-variables

strategy introduced above (panel B). All variables discussed in this table are standardized, meaning that

the coefficients can be interpreted as standard deviation changes in the outcome variable induced by a

one-standard deviation increase in the share of European immigrants. Columns (1)–(4) use total farm

values per hectare as the outcome, while columns (5)–(7) divide the farm value outcome into its constituent

components—land, infrastructure, and tools and machines. Our OLS results reveal a statistically significant

positive relationship between the share of European immigrants and farm value per hectare: a one-standard

deviation increase in the immigrant share was associated with a 0.6-standard deviation increase in farm

value. Adding controls for land quality (column 2), distance to domestic and international markets (column

3), and state fixed effects (column 4) has only a negligible effect on the magnitude of the estimates and has

little impact on their precision. Dividing farm value into its constituent components shows that the effect

of immigration was realized for all three—roughly equally for land and infrastructure, and with a somewhat

44 In order to use the 1872 census for this test, we aggregate our 1920 municipalities into larger minimum comparable areas.

See the notes to Table 1 and Online Appendix Figure A.7.

20
smaller effect for tools and machines. The results for the individual components of value help to shed light on

the mechanisms for immigration’s effect. Notably, that the effect goes beyond the value of land alone suggests

that the results are not purely the product of greater demand for land or of local amenities capitalized into

the value of immobile factors (Bleakley and Rhode 2022). The effect on tools and machines also indicates an

interesting complementarity between immigration and capital. We investigate these potential mechanisms

in more detail below.

Panel B of Table 2 presents our instrumental variables estimates. This panel shows, in support of

the relevance of our instrument, that our first-stage F -statistics are nearly 5 times larger than the weak

instrument critical value of the LIML estimator with one endogenous regressor and 5 instruments, 4.84

(Bound, Jaeger, and Baker 1995; Stock and Yogo 2005).45 Our IV estimates corroborate the OLS estimates

of panel A, revealing a statistically significant positive effect of a greater share of European immigrants on

farm values in general and on each component. In our preferred specification in column (4), which controls

for state fixed effects, we find that a one-standard deviation increase in the share of immigrants yielded

approximately a 0.7-standard deviation increase in farm value.

In Table 3, we verify the robustness of our results to a number of sample restrictions.46 These sample

restrictions address concerns that should, in principle, be addressed by our instrumental variables strategy,

but which are nevertheless useful in ensuring that specific subsets of our sample are not driving our results.47

In Panel A, we exclude municipalities that contained immigrant colonies. These settlements were the product

of partnerships between private planters and national or state governments, which led to the creation of new

rural communities. This form of immigration has been associated with faster economic development as a

result of greater human capital accumulation (Rocha, Ferraz, and Soares 2017), but may also have been

built in economically unfavorable locations (Cameron 1931). Given the peculiar developmental history of

these places, their farm value could have been the product of forces other than immigration. In Panel B, we

exclude areas that were large producers of coffee in order to account for the crucial role that this commodity

played in shaping local development and attracting immigrant labor force. We define large coffee producers

as those municipalities in the top decile of agricultural land share dedicated to coffee production. Concretely,

this translates into excluding municipalities with a share of land dedicated to coffee production above 60

45 We use LIML given that the critical values for this estimator decrease as the number of excluded instruments increases,

a property that fits our setting well, but the results are very similar if we employ two stage least squares, as shown in Online
Appendix E. This is important as some of our robustness checks are based on 2SLS estimation. Our F -statistics are also greater
than thresholds for smaller numbers of instruments, which may be important given that not all of the 5 instruments may enter
into the first stage significantly.
46 We do the same for the individual components of farm value in Online Appendix F.
47 Online Appendix Figure A.5 shows which municipalities are dropped in each case.

21
percent.48 In Panel C, we exclude municipalities obtaining a railway station in the first three decades of

the rail network expansion. It is possible that municipalities connected earlier may have been systematically

different from those connected later on in a way that our controls are not be able to capture. Finally, in

Panel D we exclude large population centers—defined as municipalities in the top quartile of the population

distribution in 1920—in order to address the concern that economic centers may be significantly influencing

our results or that agriculture may not have been particularly important in these places. In all cases, we find

no reason to believe that our results were driven by any of the concerns that we address, and, in general, we

find little evidence that these concerns even influenced the magnitude of our estimates.

7 Mechanisms

Having established a large, robust causal effect of European immigration on farm value, we now examine

potential mechanisms that may explain this effect. We approach this exercise in three steps. First, in Table

4, we use our instrumental variables strategy to test whether the variables that operationalize our proposed

mechanisms were, in fact, affected by immigration. Next, in Table 5, we evaluate the impact of controlling for

this mechanism on the coefficient on the European share in our instrumental variables analysis, as in Panel

B of Table 2, with column (7) of this table showing the impact of controlling for all mechanisms. Finally, in

Table 6 we implement a Gelbach (2016) decomposition to determine the degree to which the coefficient on

the European share is changed by including the proposed mechanism as a control when all mechanisms are

included.49

7.1 Labor Force

The first mechanism that we test attributes the effects of immigration directly to its role in increasing the

size of the labor force in a particular place. That is, immigration may have increased the number of available

workers, enabling the exploitation of land, just as the official immigration program intended. A greater

population may also have increased local demand for land (though our results above for the components of

farm value suggest that the latter mechanism is unlikely to explain our results). We operationalize demand

48 We obtain similar results if we drop the top decile of municipalities by agricultural land area dedicated to coffee, or the

top decile of municipalities by volume of coffee production. The former approach, as shown in Online Appendix Figure A.5,
involves excluding the bulk of São Paulo’s coffee-producing region, which stands out in Figure 4. These results are available
upon request.
49 Note that because of technical difficulties arising from our use of multiple instruments the decomposition is performed for

the OLS analog of the regressions, not for the IV estimates, so the decomposition is not precise. But given the similarity of the
OLS and IV results, this is not a major concern.

22
for land using population density and the agricultural labor supply using agricultural employment density

(i.e., the number of agricultural workers per hectare).

These mechanisms, however, are not supported by the data. In columns (1) and (2) of Table 4, we find

no statistically significant relationship between immigration and either population density or agricultural

employment density. Indeed, the point estimates are negative, contrary to the effect of immigration on

density that would be required under such a mechanism. Consistent with this finding, although column (2)

of Table 5 shows that both measures of the availability of labor were positively related to the value of farms,

the European share coefficient is only minimally changed by the introduction of the labor force variables.

This result is confirmed by the Gelbach (2016) decomposition in Table 6, which shows, both for farm values

in total and for each individual component, that the inclusion of labor force variables has a relatively small

(in comparison to the other mechanisms that we explore below) impact on the coefficient on the European

share of population with virtually all of the effect operating through the value of land. These results rule out

what is perhaps the most obvious mechanism for the effect of migration on farm prices, showing instead that

the effect of immigration is connected to some peculiar characteristic of the immigrant labor force rather

than merely its presence.

7.2 Land Use

The next set of mechanisms that we investigate concerns the use of land—specifically cultivation intensity

and the crop mix. We capture cultivation intensity with the share of farmland cultivated (as opposed to

being left fallow or forest). We measure the crop mix using two variables—the share of cultivated land

dedicated to coffee and the share of cultivated land dedicated to other cash crops (cotton, cocoa, sugarcane,

tobacco, castor beans, and rubber).50 The hypothesis is that some characteristic of the immigrants enabled

a more intensive production of coffee, and more generally may have either incentivized or enabled a change

in the crop mix. Immigrants may have also enabled greater utilization of the available resources, leading,

for example, to more intense cultivation of land. There is historical evidence pointing in this direction, as

immigrants under the subsidy program were able to, and derived a substantial share of their incomes, from

using the land between rows of coffee trees to cultivate other crops (Holloway 1980, p. 30).

Column (3) of Table 4 shows that a greater share of Europeans increased the share of farm land cultivated.

Columns (4) and (5) show that a greater immigrant share led to a greater share of land devoted to coffee

production, but had no impact on the share of land devoted to other cash crops. Columns (3) and (4) of

50 The residual category is “staple” crops (rice, maize, wheat, beans, potatoes, manioc, and coconut).

23
Table 5 control for the cultivated share of farms and for the crop mix, respectively. These factors were both

associated with greater farm values, and their inclusion leads to a substantial decline in the magnitude of

the European share coefficient, which constitutes evidence that these are mechanisms through which the

effect of the immigration share passed. This is confirmed by the Gelbach (2016) decomposition in Table

6, which shows that it was the cultivated share of farms that was responsible for the bulk of the reduction

in the effect of the share of Europeans on farm values. The production of coffee, while responsible for a

non-negligible change in the immigrant share coefficient, particularly with respect to the value of tools and

machines, appears to have been a less important mechanism despite the links between immigration and coffee

production.

Based on these results, we conclude that European immigrants enabled a greater share of farm land to be

cultivated, raising the value of the farm. As shown above, however, this was not the product of immigrants

providing labor where it was absent. Instead, it appears that immigrants provided some peculiar benefit

that was unique to them. We argue that this peculiar characteristic of immigrants was that they exerted

greater labor effort than native workers, and in particular that temporary immigrants exerted greater labor

effort than natives or permanent immigrants. A substantial literature has shown, both theoretically and

empirically, that when immigration is temporary, immigrants substitute inter-temporally, exerting greater

labor effort in the destination while enjoying greater leisure when returning home.51 Given the high rates of

return migration to Europe from Brazil, it is plausible that this inter-temporal substitution operated in our

case. The historical literature has also highlighted that migrants were reputed for their labor effort (Cinel

1991; Florea 2023; Holloway 1980; Sánchez-Alonso 2007). The historical literature also provides evidence of

a potentially related mechanism in which permanent immigrants who did not yet own land but who sought

to transition into landownership would have been incentivized to increase their labor effort until they had

saved sufficiently to acquire land. Indeed, the historical literature has highlighted that one of the incentives

driving immigrants in exerting labor effort was the possibility of eventually owning land (Holloway 1980;

Lesser 2013).

Table 7 presents evidence substantiating this view based on the logic that, if the effect of immigrants on

farm values came from temporary migrants, then permanent immigrants in the form of foreign-born land

owners should not exhibit the same effect. Columns (1) and (2) use the same municipality-level data from

the 1920 Census as in our analysis above. Column (1) shows that a greater share of European immigrants

51 In particular, see Dustmann (1994, 2000), Dustmann, Bentolila, and Faini (1996), Dustmann and Görlach (2016), Epstein

and Venturini (2011), Galor and Stark (1991), Hill (1987), Klinthäll (2006), Kyarko and Chartouni (2017), Vijverberg and
Zeager (1994), and Wahba (2022).

24
led to a greater share of farms that were foreign owned. But in column (2), we find no evidence that foreign

ownership was associated with greater farm values; indeed, the point estimate for the coefficient on the share

of farms owned by the foreign born is negative.52

The 1920 census, however, does not enable us to explicitly separate native- and foreign-born agricultural

workers (it permits only the observation of the share of Europeans in the whole population). The 1904–

1905 Agricultural Census of the State of São Paulo, however, makes such a distinction, and also provides

information on land values, the nationality of farm owners, and number of foreign and domestic workers for

over 40,000 farms across 163 municipalities (Secreteria da Agricultura 1906–1910). Moreover, because the

data are reported at the level of the farm rather than the municipality, we can include municipality fixed

effects in order to exploit only within-municipality variation. Column (3) of Table 7 substantiates our main

result using the farm-level data, showing that a greater share of foreign-born workers was associated with

greater land values. Column (4) shows that foreign ownership had no such association. Column (5) includes

both variables in the regression. While we continue to find that a greater share of foreign-born workers

was associated with greater land values, the effect is completely dissipated for farms owned by a foreign-

born individual—the coefficient on foreign ownership is statistically significant and negative. Moreover,

the magnitude of the estimate is such that foreign ownership almost entirely eliminates the benefit of a

one-standard deviation increase in the share of foreign workers.

Altogether, the results of Table 7 are consistent with immigrants exerting greater labor effort as long as

they were temporary immigrants, but reverting away from this exertion upon purchasing land and becoming

a permanent immigrant.

7.3 Tools and Machines

We also consider the adoption of agricultural tools and machines as a potential mechanism for the effect of

immigration. We build on our finding that immigration increased all components of farm value, including

that of tools and machinery and on Lanza, Manier, and Musacchio’s (2023) findings that immigration

under the subsidized immigration program in São Paulo state led to greater adoption of agricultural tools

and machines. We operationalize this mechanism using the share of farms in the municipality that had

agricultural tools or machines. Column (6) of Table 4 shows, consistent with our results in column (7) of

Table 2, that immigration led to a greater share of farms using tools while column (7) shows that that

immigration had no effect on the share of farms using machines. Such a relationship can be rationalized by
52 We perform this exercise while recognizing the potential endogeneity of the share of farms owed by the foreign born in

addition to the European share of population.

25
considering tools as complementary to labor and machines as substitutes (Lanza, Manier, and Musacchio

2023). In Table 5, we find that farms that adopted agricultural tools were indeed more valuable, though

the effect is less than 0.2 standard deviations, and that the size of the European migrants share coefficient

is reduced by around 0.12 standard deviations when it is included. Notably, columns (2) and (3) of Table

6 show that the tool mechanism goes beyond the mechanical channel of increasing the tools and machines

component of farm values—tools are also responsible for reducing the coefficient on the immigrant share for

the land and infrastructure components of value as well. The magnitude of the impact is about one-third

that of the share of farms cultivated, and so we view the tools mechanism as second order in comparison to

it.

8 Implications for Development and Structural Change

Our analysis has established a robust and quantitatively large positive effect of European immigration on

the agricultural sector in Brazil. As outlined in section 3 above, there are a variety of possible implication of

this effect for the rest of the Brazilian economy and its transition to industrialization and sustained economic

growth beyond the primary sector.

An important caveat is that we can only answer this question in the short run, since we observe our

outcomes shortly after the end of mass migration. In principle, development and structural transformation

may have continued along the same lines or changed direction in unpredictable ways in the decades after

our outcomes are observed, for instance due to a lag in the manifestation of the effects of migration through

channels like human capital accumulation. At the same time, our short-term analysis has the advantage of

not being influenced by subsequent events, such as the introduction of import substitution policies and the

many regime changes experienced by Brazil from the 1930s until the 1980s, which had their own, sometimes

substantial, economic consequences (Ferraz, Finan, and Martinez-Bravo 2022).

In Table 8, we implement our IV strategy as above, but focus on a variety of outcomes that capture

differences in human capital formation (panel A), economic structure (panel B), state capacity (panel C),

and female labor force participation (panel D). For each set of variables, we present two specifications—

one including the whole sample and another in which we exclude large population centers (as above, these

are municipalities in the top quartile of population) in order to ensure that places that may not have

been particularly reliant on agriculture to begin with do not drive our results. On the whole, we find no

evidence that migration negatively affected other parts of the economy, or that it slowed down structural

26
transformation. In fact there are indications that the reverse may have been true.

Human capital formation has previously been positively linked to immigration in Brazil (de Carvalho

Filho and Monasterio 2012; Rocha, Ferraz, and Soares 2017; Witzel de Souza 2018), although not in a causal

framework such as ours. We capture human capital by analyzing the literacy rates of the whole population,

of natives, of females, and of children between the ages of 7 and 14. These variables are intended to capture

spillovers from migrants to the rest of the population. We find that European immigration had a strong

positive effect on all of our literacy measures: a one-standard deviation increase in the share of migrants

in the population led to a 0.7–0.8-standard deviation increase in each of the measures of literacy. The

positive link between immigration and human capital found in previous work is thus confirmed. This is not

surprising: immigrants had greater human capital to begin with and they were often successful in lobbying

for the creation of schools (de Carvalho Filho and Colistete 2010). These schools were presumably one of

the channels through which the provision of education improved at the local level, leading to higher literacy

for the overall population.

Next, we focus on employment shares by sector, and specifically on the share of the labor force employed

in agriculture and in industry. These measures give an indication of the degree to which immigration aided

or slowed down structural change. We also study the effect of immigration on two variables capturing the

presence and influence of local landed elites. The first is the share of population made up of rentiers—people

relying on returns to wealth as their main source of income. The second is the share of the population

employed in public administration, capturing the tendency for local oligarchs to exert their influence by

placing loyalists in public administration (Graham 1990). Indeed, many parts of Brazil were characterized

by a plantation-based agricultural sector and connected oligarchic system known as coronelismo, in which

local landed elites offered votes in exchange for aid, employment, and protection, in a classic example of

clientelism and patronage (Nunes Leal 1977; Woodard 2005).53 The influence of landed elites is connected

with underdevelopment in the Brazilian context, as the rapid growth following the military coup of 1964,

which weakened these elites’ grip on power, demonstrates (Ferraz, Finan, and Martinez-Bravo 2022).54

We find that municipalities with more migrants had a lower share of workers employed in agriculture,

even when excluding large urban centers.55 We also find suggestive evidence that a greater immigrant

53 This phenomenon was particularly severe in rural areas, but the situation was only marginally different in coastal cities

and other urban centers, where other power groups, such as merchants and professionals, exerted influence alongside traditional
elites
54 It is important to note for our discussion of state capacity below that neither the oligarchs nor their clients in public

administration were keen to increase public revenues, given that they were mainly interested in power and authority, not
direct embezzlement or corruption involving public funds (Abreu and Lago 2001; Graham 1990). This ensures that oligarchic
entrenchment and fiscal capacity remain distinct in our analysis.
55 Note that the outcome in this case is the share of agricultural workers out of the labor force. The analysis in Tables 4, 5,

27
share led to a higher share of workers employed in industry, as indicated by the positive coefficient for

industrial employment, though the result is not statistically significant. In any case, we find no evidence

that immigration reduced industrial employment, which would indicate a slowing of structural change. In

fact, our results on the whole point to a valuable agricultural sector that demanded less, not more, labor.

This supports the view of agricultural development as a driver of overall economic development. We also find

no relationship between immigration and rentiers or public administration employees. Thus, immigration,

despite its positive effect on the agricultural sector, did not create a larger class of people living off their

landed property or an inflated public sector.

We focus next on state capacity, which is a firmly established contributor to economic development.56 We

concentrate specifically on the ability of local governments to collect taxes (i.e., fiscal capacity) and on their

provision of public goods. We measure fiscal capacity using total tax revenues per capita, and tax revenues

per capita excluding export taxes. We exclude export taxes because trade taxes are generally considered

an indicator of low fiscal development, as opposed to direct taxes, and they are also strongly related to the

importance of the agricultural sector in the Brazilian setting, making them a poor measure of development

beyond primary production. We measure public good provision using per capita expenditure on education

and on public services.57

We find that European immigration was strongly positively associated with greater fiscal capacity, par-

ticularly when export taxes are excluded, and when major urban centers are left out of the sample. We

also find suggestive evidence of higher per capita expenditure on public services in response to immigration,

although these estimates are not uniformly statistically significant. Altogether, these results indicate greater

state capacity in places with greater immigration.

Finally, we analyze the effect of immigration on the sex ratio, measured as women per man. We examine

this ratio in the population as a whole, in total employment, and in two key economic sectors—agriculture

and industry. Migration during the Age of Mass Migration was male dominated (Hatton and Williamson

1998), leading to potentially severely skewed sex ratios in locations where more migrants settled. Skewed

sex ratios, in turn, can have adverse consequences for attitudes towards women in the workplace and female

labor market participation (Grosjean and Khattar 2019). The economic changes arising from immigration

may have also reduced employment opportunities for women, who, in the Brazilian context, found gainful

and 6 above focused on the ratio of agricultural employment to area.


56 See Acemoglu (2005), Bardhan (2016), Becker et al. (2022), Besley and Persson (2010), Dincecco and Katz (2016), Dincecco

and Prado (2012), Dittmar and Meisenzahl (2020), Epstein (2000), Hoffman (2015), and Johnson and Koyama (2017).
57 Public services comprise public hygiene, assistance and aid, police, justice and electoral services, infrastructure and sewers,

and public lights. The residual categories are the costs of running the local public administration and expenses related to public
debt. For more information on local finances in Brazil in this period see Papadia (2019).

28
employment in the industrial and service sectors more often than in the primary sector (Pena 1981). In

turn, reduced opportunities for women and a generally lower labor market participation may have negatively

effected economic development.

As expected, we find a more male-skewed sex ratio in municipalities with more immigrants. The effect

is large both for the overall sample and when we exclude major population centers. But this male-skewed

sex ratio in the population does not translate into a statistically significantly male-skewed sex ratio in either

overall employment or employment in agriculture and industry. The lack of such an effect suggests the

presence of countervailing forces, which led to women compensating for the lower presence in the population

with greater labor force participation.

9 Conclusion

In this paper, we aim to better understand the effects of immigration on the receiving economy in the context

of a developing-country destination. As a fundamental component of the economics of immigration (Bansak,

Simpson, and Zavodny 2020; Borjas 2014), the effects of immigration have been widely studied, but almost

exclusively in developed-country contexts, where the effects may be differ from those in developing countries.

Our study is situated in the context of Brazil in the Age of Mass Migration, enabling us to benefit from

the broad utility of studying this period for the economics of immigration and specifically from Brazil’s

unique position in this context as a destination that more closely parallels migration to modern developing

countries than to developed ones. Focusing on Brazil’s agricultural sector, we find that immigration led to

an increase in farm values per hectare, coming primarily from an increase in the share of land cultivated—

a mechanism that we, in turn, attribute to the exertion of greater labor effort by temporary immigrants

relative to natives and permanent immigrants. We find no evidence that this agricultural development

slowed Brazil’s structural transformation; indeed we find suggestive evidence that it may have accelerated

it by fostering human capital accumulation, reducing agricultural employment, and increasing female labor

market participation. Thus, immigrants enabled this developing country to exploit its natural resource

endowments, and may have eventually set it on the course of broad-based economic development.

29
References
Abadie, Alberto, Jiaying Gu, and Shu Shen (2023). “Instrumental Variables Estimation with First-Stage Heterogene-
ity.” Journal of Econometrics Forthcoming.
Abramitzky, Ran, Philipp Ager, Leah Platt Boustan, Elior Cohen, and Casper W. Hansen (2023). “The Effects of
Immigration on Local Labor Markets: Lessons from the 1920s Border Closure.” American Economic Journal:
Applied Economics 15:1, pp. 164–191.
Abramitzky, Ran and Leah Platt Boustan (2017). “Immigration in American Economic History.” Journal of Economic
Literature 55:4, pp. 1311–1345.
Abramitzky, Ran, Leah Platt Boustan, and Katherine Eriksson (2012). “Europe’s Tired, Poor, Huddled Masses: Self-
Selection and Economic Outcomes in the Age of Mass Migration.” American Economic Review 102:5, pp. 1832–
1856.
Abreu, Marcelo de Paiva and Luis A. Corrêa do Lago (2001). “Property Rights and the Fiscal and Financial Systems
in Brazil: Colonial Heritage and the Imperial Period.” In Transferring Wealth & Power from the Old to the New
World, Monetary and Fiscal Institutions in the 17th through the 19th Centuries. Michael D. Bordo and Roberto
Cortés-Conde (ed.). Studies in Macroeconomic History Series. Cambridge: Cambridge University Press.
Acemoglu, Daron (2005). “Politics and Economics in Weak and Strong States.” Journal of Monetary Economics 52:7,
pp. 1199–1226.
Ager, Philipp, James J. Feigenbaum, Casper W. Hansen, and Hui Ren Tan (2021). “How the Other Half Died:
Immigration and Mortality in US Cities.” Review of Economic Studies Forthcoming.
Alston, Lee J., Marcus André Melo, Bernardo Mueller, and Carlos Pereira (2016). Brazil in Transition: Beliefs,
Leadership, and Institutional Change. Princeton: Princeton University Press.
Angelucci, Manuela (2015). “Migration and Financial Constraints: Evidence from Mexico.” Review of Economics and
Statistics 97:1, pp. 224–228.
Asher, Sam, Alison Campion, Douglas Gollin, and Paul Novosad (2022). “The Long-Run Development Impacts of
Agricultural Productivity Gains: Evidence from Irrigation Canals in India.” Mimeo., Imperial College London.
Atack, Jeremy, Fred Bateman, Michael R. Haines, and Robert A. Margo (2010). “Did Railroads Induce or Follow
Economic Growth? Urbanization and Population Growth in the American Midwest, 1850–1860.” Social Science
History 34:2, pp. 171–197.
Bansak, Cynthia, Nicole Simpson, and Madeline Zavodny (2020). The Economics of Immigration. New York: Rout-
ledge.
Bardhan, Pranab (2016). “State and Development: The Need for a Reappraisal of the Current Literature.” Journal
of Economic Literature 54:3, pp. 862–892.
Barman, Roderick J. (1999). Citizen Emperor: Pedro II and the Making of Brazil, 1825–1891. Stanford: Stanford
University Press.
Barro, Robert J. and José F. Ursúa (2008). “Macroeconomic Crises since 1870.” Brookings Papers on Economic
Activity, Economic Studies Program 39:1, pp. 255–350.
Bassanezi, Maria Silvia C. Beozzo and Priscila M. S. Bergamo Francisco, eds. (2003). Estado de São Paulo. Estatística
Agrícola e Zootécnica, 1904-1905. Campinas: Nepo/Unicamp.
Becker, Sascha O., Andreas Ferrara, Eric Melander, and Luigi Pascali (2022). “Wars, Taxation and Representation:
Evidence from Five Centuries of German History.” Mimeo., University of Pittsburgh.
Belot, Michèle V. K. and Timothy J. Hatton (2012). “Immigrant Selection in the OECD.” Scandinavian Journal of
Economics 114:4, pp. 1105–1128.
Benton-Cohen, Katherine (2018). Inventing the Immigration Problem: The Dillingham Commission and its Legacy.
Cambridge: Harvard University Press.
Bértola, Luis and José Antonio Ocampo (2012). The Economic Development of Latin America since Independence.
Oxford: Oxford University Press.
Besley, Timothy and Torsten Persson (2010). “State Capacity, Conflict and Development.” Econometrica 78:1, pp. 1–
34.
Biavaschi, Costanza, Giovanni Facchini, Anna Maria Mayda, and Mariapia Mendola (2018). “South-South Migration
and the Labor Market: Evidence from South Africa.” Journal of Economic Geography 18, pp. 823–853.
Bleakley, Hoyt and Paul W. Rhode (2022). “The Economic Effects of American Slavery: Tests at the Border.” Mimeo.,
University of Michigan.
Böhme, Marcus H. and Sarah Kups (2017). “The Economic Effects of Labour Immigration in Developing Countries:
A Literature Review.” OECD Development Centre Working Paper 335.

30
Bolt, Jutta and Jan Luiten van Zanden (2020). “Maddison Style Estimates of the Evolution of the World Economy.
A New 2020 Update.” Maddison-Project Working Paper WP-15.
Borjas, George J. (2003). “The Labor Demand Curve Is Downward Sloping: Reexamining the Impact of Immigration
on the Labor Market.” Quarterly Journal of Economics 118:4, pp. 1335–1374.
——— (2014). Immigration Economics. Cambridge: Harvard University Press.
Bound, John, David A. Jaeger, and Regina M. Baker (1995). “Problems with Instrumental Variables Estimation
when the Correlation between the Instruments and the Endogenous Explanatory Variable is Weak.” Journal of
the American Statistical Association 90:430, pp. 443–450.
Burchardi, Konrad, Thomas Chaney, and Tarek Hassan (2019). “Migrants, Ancestors, and Foreign Investments.”
Review of Economic Studies 86:4, pp. 1448–1486.
Bustos, Paula, Bruno Caprettini, and Jacopo Ponticelli (2016). “Agricultural Productivity and Structural Transfor-
mation: Evidence from Brazil.” American Economic Review 106:6, pp. 1320–1365.
Bustos, Paula, Gabriel Garber, and Jacopo Ponticelli (2020). “Capital Accumulation and Structural Transformation.”
Quarterly Journal of Economics 135:2, pp. 1037–1094.
Cameron, C. R. (1931). “Colonization of Immigrants in Brazil.” Monthly Labour Review 33:4, pp. 36–46.
Card, David (1990). “The Impact of the Mariel Boatlift on the Miami Labor Market.” Industrial and Labor Relations
Review 43:2, pp. 245–257.
——— (2005). “Is the New Immigration Really So Bad?” Economic Journal 115, F300–F323.
Chiquiar, Daniel and Gordon H. Hanson (2005). “International Migration, Self-Selection and the Distribution of
Wages: Evidence from Mexico and the United States.” Journal of Political Economy 113:2, pp. 239–281.
Cinel, Dino (1991). The National Integration of Italian Return Migration, 1870-1929. Cambridge University Press.
Clemens, Michael A., Ethan G. Lewis, and Hannah M. Postel (2018). “Immigration Restrictions as Active Labor
Market Policy: Evidence from the Mexican Bracero Exclusion.” American Economic Review 108:6, pp. 1468–
1487.
Cohen, Elior and Jeff Biddle (2022). “Immigration Disruptions and the Wages of Unskilled Labor in the 1920s.”
Federal Reserve Bank of Kansas City Working Paper 22-12.
Colella, Fabrizio, Rafael Lalive, Seyhun Orcan Sakalli, and Mathias Thoenig (2019). “Inference with Arbitrary Clus-
tering.” IZA Discussion Paper 12584.
Colistete, Renato Perim (2015). “Regiões e Especialização na Agricultura Cafeeira: São Paulo no Início do Século
XX.” Revista Brasileira de Economia 69:3, pp. 331–354.
Collins, William J. and Ariell Zimran (2023). “Working Their Way Up? US Immigrants’ Changing Labor Market
Assimilation in the Age of Mass Migration.” American Economic Journal: Applied Economics 15:3, pp. 238–269.
Conley, Timothy G. (1999). “GMM Estimation with Cross Sectional Dependence.” Journal of Econometrics 92:1,
pp. 1–45.
da Costa, Emilia Viotti (2000). The Brazilian Empire: Myths & Histories. Chapel Hill: University of North Carolina
Press.
Dawsey, Cyrus B. and James M. Dawsey (1995). “Leaving: The Context of the Southern Emigration to Brazil.” In
The Confederados: Old South Immigrants in Brazil. Cyrus B. Dawsey and James M. Dawsey (ed.). Tuscaloosa:
University of Alabama Press. Chap. 1.
de Carvalho Filho, Ireneu and Renato P. Colistete (2010). “Education Performance: Was It All Determined 100 Years
Ago? Evidence from São Paulo, Brazil.” MPRA Paper 24494.
de Carvalho Filho, Irineu and Leonardo Monasterio (2012). “Immigration and the Origins of Regional Inequality:
Government-Sponsored Migration to Southern Brazil before World War I.” Regional Science and Urban Economics
42, pp. 794–807.
Dean, Warren (1969). The Industrialization of São Paulo (1880-1945). Austin: Universoty of Texas Press.
Deryugina, Tatyana, Garth Heutel, Nolan H. Miller, David Molitor, and Julian Reif (2019). “The Mortality and
Medical Costs of Air Pollution: Evidence from Changes in Wind Direction.” American Economic Review 109:12,
pp. 4178–4219.
Dincecco, Mark and Gabriel Katz (2016). “State Capacity and Long-Run Economic Performance.” Economic Journal
126:590, pp. 189–218.
Dincecco, Mark and Mauricio Prado (2012). “Warfare, Fiscal Capacity, and Performance.” Journal of Economic
Growth 17:3, pp. 171–203.
Directoria Geral de Estatística (1908). Boletim Commemorativo da Esposição Nacional de 1908. Rio de Janeiro:
Typographia de Estatística.
——— (1922). Recensamento do Brazil Realizado em 1 de Setembro de 1920. Rio de Janeiro: Officina da Estatística.

31
Dittmar, Jeremiah E. and Ralf R. Meisenzahl (2020). “Public Goods Institutions, Human Capital, and Growth:
Evidence from Germany.” Review of Economic Studies 87:2, pp. 959–996.
Dix-Carneiro, Rafael and Brian K. Kovak (2017). “Trade Liberalization and Regional Dynamics.” American Economic
Review 107:10, pp. 2908–2946.
Donaldson, Dave and Richard Hornbeck (2016). “Railroads and American Economic Growth: A ‘Market Access’
Approach.” Quarterly Journal of Economics 131:2, pp. 799–858.
Droller, Federico and Martin Fiszbein (2021). “Staple products, linkages, and development: Evidence from Argentina.”
The Journal of Economic History 81:3, pp. 723–762.
Dustmann, Christian (1994). “Differences in the Labour Market Behaviour Between Temporary and Permanent
Migrant Women.” CEPR Discussion Paper 947.
——— (2000). “Temporary Migration and Economic Assimilation.” IZA Discussion Paper 186.
Dustmann, Christian, Samuel Bentolila, and Riccardo Faini (1996). “Return Migration: The European Experience.”
Economic Policy 11, pp. 213–250.
Dustmann, Christian and Joseph-Simon Görlach (2016). “The Economics of Temporary Migrations.” Journal of
Economic Literature 54:1, pp. 98–136.
Dustmann, Christian, Uta Schönberg, and Jan Stuhler (2016). “Labor Supply Shocks, Native Wages, and the Adjust-
ment of Local Employment.” Quarterly Journal of Economics 132:1, pp. 435–483.
Epstein, Gil S. and Alessandra Venturini (2011). “The Impact of Worker Effort on Public Sentiment Toward Tempo-
rary Migrants.” Research in Labor Economics 33, pp. 239–261.
Epstein, Stephan R. (2000). Freedom and Growth: The Rise of States and Markets in Europe, 1300-1750. London:
Routeledge.
Fausto, Boris (1999). A Concise History of Brazil. Cambridge: Cambridge University Press.
Feler, Leo, Aldo Musacchio, and Eustaquio Reis (2016). “Immigration and Banking Development in Brazil.” Mimeo.,
Brandeis University.
Ferenczi, Imre and Walter F. Willcox (1929). International Migrations. New York: National Bureau of Economic
Research.
Fernández-Sánchez, Martín and Gaspare Tortorici (2023). “Male and Female Self-Selection during the Portuguese
Mass Migration, 1885–1930.” Explorations in Economic History Forthcoming.
Ferraz, Claudio, Frederico Finan, and Monica Martinez-Bravo (2022). “Political Power, Elite Control, and Long-Run
Development: Evidence from Brazil.” NBER Working Paper 27456.
Florea, Cristina (2023). “Frontiers of Civilization in the Age of Mass Migration from Eastern Europe.” Past & Present
258:1, pp. 115–150.
Foerster, Robert F. (1919). The Italian Emigration of Our Times. New York: Russell & Russell.
Food and Agriculture Organization (2021). “Agro-Ecological Suitability and Potential Yields.” Global Agro-Ecological
Zones (GAEZ) v4.
Foster, Andrew D. and Mark R. Rosenzweig (1996). “Technical Change and Human-Capital Returns and Investments:
Evidence from the Green Revolution.” American Economic Review 86:4, pp. 931–953.
Gagliardi, Fabio (1958). Localização da Colonização Oficial em São Paulo. Digital Archive of the Museu da Imigração
do Estado de São Paulo.
Galor, Oded and Oded Stark (1991). “The Probability of Return Migration, Migrants’ Work Effort, and Migrants’
Performance.” Journal of Development Economics 35, pp. 399–405.
Gelbach, Jonah B. (2016). “When Do Covariates Matter? And Which Ones, And How Much?” Journal of Labor
Economics 34:2, pp. 275–543.
Giesbrecht, Ralph Mennucci (2023). “Estações Ferroviárias do Brasil.” http://www.estacoesferroviarias.com.br/
index.html, accessed April 25, 2023.
Gindling, T. H. (2009). “South-South Migration: The Impact of Nicaraguan Immigrants on Earnings, Inequality and
Poverty in Costa Rica.” World Development 37:1, pp. 116–126.
Goldin, Claudia (1994). “The Political Economy of Immigration Restriction in the United States, 1890 to 1921.” In
The Regulated Economy: A Historical Approach to Political Economy. Claudia Goldin and Gary D. Libecap (ed.).
Chicago: University of Chicago Press, pp. 223–258.
Graham, Richard (1990). Patronage and Politics in Nineteenth-Century Brazil. Stanford: Stanford University Press.
Grosjean, Pauline and Rose Khattar (2019). “It’s Raining Men! Hallelujah? The Long-Run Consequences of Male-
Biased Sex Ratios.” Review of Economic Studies 86:2, pp. 723–754.
Hanson, Gordon H. (2009). “The Economic Consequences of the International Migration of Labor.” Annual Review
of Economics 1, pp. 179–208.

32
Hatton, Timothy and Jeffrey Williamson (1994). “What Drove Mass Migrations from Europe in the Late Nineteenth
Century?” Population and Development Review 20:3, pp. 533–559.
Hatton, Timothy J. and Zachary Ward (2019). “International Migration in the Atlantic Economy 1850–1940.” In
Handbook of Cliometrics. Claude Diebolt and Michael Haupert (ed.). Springer.
Hatton, Timothy J. and Jeffrey G. Williamson (1998). The Age of Mass Migration: Causes and Economic Impact.
New York: Oxford University Press.
——— (2002). “What Fundamentals Drive World Migration?” NBER Working Paper 9159.
——— (2005). Global Migration and the World Economy: Two Centuries of Policy and Performance. Cambridge:
MIT Press.
Herranz-Loncán, Alfonso (2014). “Transport Technology and Economic Expansion: The Growth Contribution of
Railways in Latin America Before 1914.” Journal of Iberian and Latin American Economic History 32:1, pp. 13–
45.
Hill, John K. (1987). “Immigrant Decisions Concerning Duration of Stay and Migratory Frequency.” Journal of
Development Economics 25, pp. 221–234.
Hoffman, Philip T. (2015). “What Do States Do? Politics and Economic History.” Journal of Economic History 75:2,
pp. 303–332.
Holloway, Thomas H. (1980). Immigrants on the Land. Coffee and Society in São Paulo, 1886-1934. Chapel Hill:
University of North Carolina Press.
Hornbeck, Richard and Pinar Keskin (2015). “Does Agriculture Generate Local Economic Spillovers? Short-Run and
Long-Run Evidence from the Ogallala Aquifer.” American Economic Journal: Economic Policy 7:2, pp. 192–213.
Instituto Brasileiro de Geografia e Estatística (1954). Anuário Estatístico do Brasil. Rio de Janeiro: Conselho Nacional
de Estatística.
——— (2011). “Evolução da Divisão Territorial do Brasil: 1872-2010.” https://maps.princeton.edu/catalog/
stanford-sf351hx7032, accessed April 24, 2023. Rio de Janeiro.
Jackson, C. Kirabo, Rucker C. Johnson, and Claudia Persico (2016). “The Effects of School Spending on Educa-
tional and Economic Outcomes: Evidence from School Finance Reforms.” Quarterly Journal of Economics 131:1,
pp. 157–218.
Johnson, Noel D. and Mark Koyama (2017). “States and Economic Growth: Capacity and Constraints.” Explorations
in Economic History 64:2, pp. 1–20.
Johnston, Bruce F. and John W. Mellor (1961). “The Role of Agriculture in Economic Development.” American
Economic Review 51:4, pp. 566–593.
Kerr, Sari Pekkala and William R. Kerr (2011). “Economic Impacts of Immigration: A Survey.” Finnish Economic
Papers, Finnish Economic Association 24:1, pp. 1–32.
Klein, Herbert S. (1995). “European and Asian Migration to Brazil.” In The Cambridge Survey of World Migration.
Robin Cohen (ed.). Cambridge: Cambridge University Press, pp. 208–214.
Klinthäll, Martin (2006). “Immigration, Integration and Return Migration.” International Symposium on Interna-
tiopnal Migration and Development.
Kyarko, Yaw and Carole Chartouni (2017). “Managed International Labor Mobility as Contribution to Economic De-
velopment and Growth: A Brief Literature Review.” Summary Paper to the World Bank Project on “Afghanistan:
Managed International Labor Mobility as Contribution to Economic Development and Growth”.
Lafortune, Jeanne, Ethan G. Lewis, and José Tessada (2019). “People and Machines: A Look at the Evolving Relation-
ship between Capital and Skill in Manufacturing, 1860–1930, Using Immigration Shocks.” Review of Economics
and Statistics 101:1, pp. 30–43.
Lanza, André Luiz (2021). “De Braços para a Lavoura a Proprietários Rurais: Imigrantes e Aceso à Terra em São
Paulo, 1886–1920.” PhD thesis. Universidade de São Paulo.
Lanza, André Luiz, Manas Manier, and Aldo Musacchio (2023). “European Immigration and Agricultural Productivity
in São Paulo, Brazil, 1898–1920.” Mimeo., Brandeis University.
Lee, Jongkwan, Giovanni Peri, and Vasil Yasenov (2022). “The Labor Market Effects of Mexican Repatriations:
Longitudinal Evidence from the 1930s.” Journal of Public Economics 205, 104558.
Lesser, Jeffrey (2013). Immigration, Ethnicity, and National Identity in Brazil, 1808 to the Present. Cambridge
University Press.
Lewis, Arthur W. (1954). “Economic Development with Unlimited Supplies of Labour.” Manchester School 22,
pp. 139–191.
Luna, Francisco Vidal, Herbert S. Klein, and William R. Summerhill (2014). “A Agricultura Paulista em 1905.”
Estudos Econômicos 44:1, pp. 153–184.

33
Machado, Maria Helena Pereira Toledo (2006). “From Slave Rebels to Strikebreakers: The Quilombo of Jabaquara
and the Problem of Citizenship in Late-Nineteenth-Century Brazil.” Hispanic American Historical Review 86:2,
pp. 247–274.
Manacorda, Marco, Alan Manning, and Jonathan Wadsworth (2012). “The Impact of Immigration on the Structure
of Wages: Theory and Evidence from Britain.” Journal of the European Economic Association 10:1, pp. 120–151.
Matsuyama, Kiminori (1982). “Agricultural Productivity, Comparative Advantage and Economic Growth.” Journal
of Economic Theory 58, pp. 317–334.
McKenzie, David and Hillel Rapoport (2010). “Self-Selection Patterns in Mexico-US Migration: The Role of Migration
Networks.” Review of Economics and Statistics 92:4, pp. 811–821.
Montero, Eduardo and Dean Yang (2022). “Religious Festivals and Economic Development: Evidence from the Timing
of Mexican Saint Day Festivals.” American Economic Review 112:10, pp. 3176–3214.
Naritomi, Joana, Rodrigo R Soares, and Juliano J Assunção (2012). “Institutional development and colonial heritage
within Brazil.” The Journal of Economic History 72:2, pp. 393–422.
Nunes Leal, Victor (1977). Coronelismo: The Municipality and Representative Government in Brazil. Cambridge:
Cambridge University Press.
Orrenius, Pia M. and Madeline Zavodny (2005). “Self-Selection among Undocumented Immigrants from Mexico.”
Journal of Development Economics 78, pp. 215–240.
Özden, Çağlar and Mathis Wagner (2014). “Immigrants versus Natives? Displacement and Job Creation.” World
Bank Policy Research Working Paper 6900.
Palma, Nuno, Andrea Papadia, Thales Pereira, and Leonardo Weller (2021). “Slavery and Development in Nineteenth
Century Brazil.” Capitalism: A Journal of History and Economics 2:2, pp. 372–426.
Papadia, Andrea (2019). “Slaves, Migrants and Development in Brazil, 1872–1923.” EUI Working Paper MWP
2019/05.
Pascali, Luigi (2017). “The Winds of Change: Maritime Technology, Trade, and Economic Development.” American
Economic Review 107:9, pp. 2821–2854.
Pena, Maria Valéria Junho (1981). Mulheres e Trabalhadoras : Presença Feminina na Constituição do Sistema Fabril.
Rio de Janeiro: Paz e Terra.
Pérez, Santiago (2017). “The (South) American Dream: Mobility and Economic Outcomes of First- and Second-
Generation Immigrants in Nineteenth-Century Argentina.” Journal of Economic History 77:4, pp. 971–1006.
Prados de la Escosura, Leandro (2009). “Lost Decades? Economic Performance in Post-Independence Latin America.”
Journal of Latin America Studies 41, pp. 279–307.
Price, Joseph, Christian vom Lehn, and Riley Wilson (2023). “The Winners and Losers of Immigration: Evidence
from Linked Historical Data.” NBER Working Paper 27156.
Ratha, Dilip and William Shaw (2007). “South-South Migration and Remittances.” World Bank Working Paper 102.
Rocha, Rudi, Claudio Ferraz, and Rodrigo R. Soares (2017). “Human Capital Persistence and Development.” American
Economic Journal: Applied Economics 9:4, pp. 105–136.
Rocha, Rudi, Rodrigo R. Soares, and Claudio Ferraz (2017). “Human Capital Persistence and Development.” American
Economic Journal: Applied Economics 9:4, pp. 105–136.
Sánchez-Albornoz, Nicolás (1986). “The Population of Latin America, 1850–1930.” In. The Cambridge History of
Latin America. Leslie Bethell (ed.). Vol. 4. The Cambridge History of Latin America. Cambridge University
Press, pp. 121–150.
Sánchez-Alonso, Blanca (2007). “The Other Europeans: Immigration into Latin America and the International Labour
Market (1870–1930).” Journal of Iberian and Latin American Economic History 25:3, pp. 395–426.
——— (2019). “The Age of Mass Migration in Latin America.” Economic History Review 72:1, pp. 3–31.
Schultz, Theodore W. (1964). Transforming Traditional Agriculture. New Haven: Yale University Press.
Secreteria da Agricultura, Commercio e Obras Publicas do Estado de São Paulo (1906–1910). Estatistica Agricola e
Zootechnica do Estado de São Paulo no Anno Agricola de 1904-1905. São Paulo: Typ. Brazil—Carlos Gerke &
Rothschild.
Sequeira, Sandra, Nathan Nunn, and Nancy Qian (2020). “Immigrants and the Making of America.” Review of
Economic Studies 87, pp. 382–419.
Slenes, Robert W. (2010). “Brazil.” In The Oxford Handbook of Slavery in the Americas. Mark M. Smith and Robert
L. Paquette (ed.). Oxford University Press.
Sokoloff, Kenneth L and Stanley L Engerman (2000). “History Lessons: Institutions, Factor Endowments, and Paths
of Development in the New World.” Journal of Economic Perspectives 14:3, pp. 217–232.
Spitzer, Yannay (2015). “The Dynamics of Mass Migration: Estimating the Effect of Income Differences on Migration
in a Dynamic Model with Diffusion.” Mimeo., The Hebrew University of Jerusalem.

34
Spitzer, Yannay, Gaspare Tortorici, and Ariell Zimran (2022). “International Migration Responses to Modern Europe’s
Most Destructive Earthquake: Messina and Reggio Calabria, 1908.” NBER Working Paper 27506.
Spitzer, Yannay and Ariell Zimran (2018). “Migrant Self-Selection: Anthropometric Evidence from the Mass Migration
of Italians to the United States, 1907–1925.” Journal of Development Economics 134, pp. 226–247.
——— (2023). “Like an Ink Blot on Paper: Testing the Diffusion Hypothesis of Mass Migration, Italy 1876–1920.”
NBER Working Paper 30847.
Spruk, Rok (2019). “The Rise and Fall of Argentina.” Latin American Economic Review 28:16.
Stock, James H. and Motohiro Yogo (2005). “Testing for Weak Instruments in Linear IV Regression.” In Identification
and Inference for Econometric Models. D. W. K. Andres (ed.). New York: Cambridge University Press, pp. 80–
108.
Stolz, Yvonne, Joerg Baten, and Tarcísio Botelho (2013). “Growth Effects of Nineteenth-Century Mass Migrations:
‘Fome Zero’ for Brazil?” European Review of Economic History 17, pp. 95–121.
Summerhill, William R (2003). Order Against Progress: Government, Foreign Investment, and Railroads in Brazil,
1854-1913. Stanford University Press.
——— (2005). “Big Social Savings in a Small Laggard Economy: Railroad-Led Growth in Brazil.” Journal of Economic
History 65:1, pp. 72–102.
Sutch, Richard (2006). “National Income and Product.” In Historical Statistics of the United States: Earliest Time to
the Present. M. R. Haines et al S. B. Carter S. S. Gartner (ed.). Vol. III. New York: Cambridge University Press,
pp. 23–25.
Suzigan, Wilson (2000). Indústria Brasileira: Origem e Desenvolvimento. Sã Paulo: Editora Hucitec, Editora da
Unicamp.
Tabellini, Marco (2020). “Gifts of the Immigrants, Woes of the Natives: Lessons from the Age of Mass Migration.”
Review of Economic Studies 87, pp. 454–486.
Tang, John P. and Luis Carvalho Monteiro (2023). “Literacy and Occupational Mobility: Evidence from Brazil.”
Mimeo., University of Melbourne.
Timmer, Peter C. (1988). “The Agricultural Transformation.” In Handbook of Development Economics. Hollis Chenery
and T.N. Srinivasan (ed.). Vol. 1, pp. 275–331.
US Congress (1911). Reports of the Immigration Commission: Statements and Recommendations Submitted By So-
cieties and Organizations Interested in the Subject of Immigration. Vol. 41. Washington: Government Printing
Office.
Vijverberg, Wim P. M. and Lester A. Zeager (1994). “Comparing Earnings Profiles in Urban Areas of an LDC:
Rural-to-Urban Migrants vs. Native Workers.” Journal of Development Economics 45, pp. 177–199.
Wahba, Jackline (2022). “The Economics of Return Migration.” In Handbook of Return Migration. Russell King and
Katie Kuschminder (ed.). Northampton: Edward Elgar. Chap. 2, pp. 24–37.
Williamson, Jeffrey G. (1995). “The Evolution of Global Labor Markets since 1830: Background Evidence and Hy-
potheses.” Explorations in Economic History 32, pp. 141–196.
Witzel de Souza, Bruno Gabriel (2018). “Immigration and the Path Dependence of Education: The Case of German-
Speakers in São Paulo, Brazil (1840–1920).” Economic History Review 71:2, pp. 506–539.
Woodard, James P. (2005). “Coronelismo in Theory and Practice: Evidence, Analysis and Argument from São Paulo.”
Luso-Brazilian Review 42:1, pp. 99–117.
World Bank (2016). Migration and Remittances Factbook 2016. 3rd edition. Washington, DC: World Bank.
Zimran, Ariell (2020). “Transportation and Health in the Antebellum United States 1820–1847.” Journal of Economic
History 80:3, pp. 670–709.
——— (2022). “US Immigrants’ Secondary Migration and Geographic Assimilation in the Age of Mass Migration.”
Explorations in Economic History 85, 101457.

35
Figures

Figure 1: Immigrant arrivals to Brazil by year, and immigration lulls and booms
Note: Shaded areas are immigration booms, defined as years with an immigrant flow above the previous five years’ moving
average. The remaining periods are lulls. Gaps in the population data are linearly interpolated.

Source: Directoria Geral de Estatística (1908) and Instituto Brasileiro de Geografia e Estatística (1954) for aggregate im-
migration numbers. Bolt and van Zanden (2020) and Instituto Brasileiro de Geografia e Estatística (1954) for population
numbers.

1
Other

.8
Share of Brazilian Immigration

Spanish
.6

.4

Portuguese
.2
Italians

0
1850 1860 1870 1880 1890 1900 1910 1920
Year

Figure 2: Distribution of origin countries for Brazilian immigration


Note: This graph shows the share of immigration to Brazil coming from each source country. The color key is as follows:
Italians (red), Portuguese (orange), Spanish (yellow), Japanese (green), Germans (blue), Russians (purple), and other (gray).

Source: Directoria Geral de Estatística (1908) and Instituto Brasileiro de Geografia e Estatística (1954).

36
10000
8000

Real GDP per capita in 2011$


6000

4000

2000

1860 1880 1900 1920


Year

Brazil Argentina
United States

Figure 3: Real GDP per capita in major immigrant destinations


Source: Brazil: Barro and Ursúa (2008), Bolt and van Zanden (2020), and Prados de la Escosura (2009); US: Bolt and van
Zanden (2020), Prados de la Escosura (2009), and Sutch (2006); Argentina: Bértola and Ocampo (2012), Bolt and van Zanden
(2020), and Prados de la Escosura (2009).

37
(a) Farm value per hectare

(b) Population share of European immigrants

Figure 4: Farm values and immigration


Source: Directoria Geral de Estatística (1922) and Instituto Brasileiro de Geografia e Estatística (2011)

38
(a) Expansion of Brazil’s railway network (1850–1920)

(b) Predicted share of European immigrants

Figure 5: Geographic distribution of the railway and the instrument


Source: Directoria Geral de Estatística (1922), Giesbrecht (2023), and Instituto Brasileiro de Geografia e Estatística (2011).

39
Tables

Table 1: Balance test for observables: rail connection during immigration booms and lulls

Full sample Boom Lull Difference Cond. Diff. Cond. Diff.


Regions States
(1) (2) (3) (4) (5) (6)
Euro share 1872 0.01 0.02 0.01 -0.01 -0.01 -0.00
Population 1872 (1,000s) 16.41 20.79 17.66 -3.12 -3.41 -1.27
Agricultural employment 1872 0.31 0.32 0.29 -0.03⇤ -0.03⇤ -0.03⇤⇤
Justice work. 1872 0.24 0.23 0.29 0.06 0.05 0.04
Slave share 1872 0.14 0.17 0.15 -0.02 -0.00 0.00
White share 1872 0.38 0.43 0.41 -0.02 0.00 0.01
Literacy 1872 0.15 0.15 0.16 0.01 0.01 0.01
School attendance 1872 0.13 0.12 0.14 0.01 0.01 0.01
Distance (port/custom house) 249.70 221.06 186.05 -35.01 -3.86 1.35
Distance (city) 50.15 43.97 33.18 -10.80⇤⇤ -8.22⇤ -6.59
Ruggedness 100.95 120.53 100.40 -20.13⇤⇤ -11.35 -6.37
Altitude 401.77 495.58 404.27 -91.30⇤⇤ -23.27 -7.50
Latitude -14.25 -18.42 -16.44 1.98⇤ 0.02 -0.16
Longitude -43.58 -44.37 -43.02 1.35⇤ 0.00 -0.05
Land quality -0.01 -0.31 -0.30 0.01 -0.16 -0.11
Observations 605 168 77 245 245 245
Notes: *** p<0.01, ** p<0.05, * p<0.1. Column (1) shows the average value of each variable for the whole sample (including
municipalities never linked to the railroad). Columns (2) and (3) show the averages for the two groups of municipalities—those
connected during a boom in aggregate immigration and those connected during a lull. Booms and lulls are defined and illustrated
in Figure 1. Column (4) illustrates the results of a equality of means test; columns (5) and (6) do the same conditioning on
regional and state fixed effects respectively. For this exercise we use minimum comparable areas rather than municipalities. This
is necessary to link data across census years, given the creation and suppression of municipalities over time. Online Appendix
Figure A.7 provides an illustration of minimum comparable areas and municipalities across the Brazilian territory created to
link the 1872 and 1920 census data.

40
Table 2: European immigration and farm value

Farm Value per hectare Land Infrastructure Tools & Machines


(1) (2) (3) (4) (5) (6) (7)
Panel A: OLS
Share Europeans 0.586*** 0.598*** 0.595*** 0.605*** 0.568*** 0.599*** 0.428***
(0.0571) (0.0579) (0.0570) (0.0663) (0.0659) (0.0658) (0.0548)
Observations 1,289 1,289 1,289 1,289 1,289 1,289 1,289
R2 0.541 0.557 0.575 0.595 0.604 0.463 0.404
Panel B: IV
Share Europeans 0.690*** 0.670*** 0.735*** 0.739*** 0.730*** 0.573*** 0.744***
(0.151) (0.116) (0.116) (0.118) (0.120) (0.139) (0.123)
Observations 1,289 1,289 1,289 1,289 1,289 1,289 1,289
1st stage F-stat 17.01 22.38 22.79 22.46 22.46 22.46 22.46
Railway years 3 3 3 3 3 3 3
No rail 3 3 3 3 3 3 3
Geo Controls 3 3 3 3 3 3 3
Region FE 3 3 3
Land adaptability 3 3 3 3 3 3
Land quality 3 3 3 3 3 3
Dom market access 3 3 3 3 3
Int market access 3 3 3 3 3
State FE 3 3 3 3

Notes: *** p<0.01, ** p<0.05, * p<0.1. Robust standard errors in parentheses. The IV regressions are estimated using the
LIML estimator with the predicted share of Europeans interacted with region indicators as the instruments for the actual share
of Europeans. Sample includes all municipalities. The unit of observation is a municipality. All variables are standardized to
have mean zero and standard deviation one.

41
Table 3: European immigration and farm value, robustness tests

Farm Value per hectare


(1) (2) (3) (4) (5) (6) (7) (8)
Panel A: no immigrant colonies Panel B: no large coffee producers
Share Europeans 0.549*** 0.606*** 0.667*** 0.656*** 0.961** 0.713*** 0.823*** 0.677***
(0.149) (0.124) (0.123) (0.122) (0.425) (0.213) (0.212) (0.154)
Observations 1,159 1,159 1,159 1,159 1,157 1,157 1,157 1,157
1st stage F-stat 17.60 21.39 21.53 22.14 7.98 10.61 10.89 13.25
Panel C: no early railway connections Panel D: no large population centers
Share Europeans 0.501** 0.566*** 0.663*** 0.626*** 0.781*** 0.710*** 0.787*** 0.748***
(0.204) (0.146) (0.153) (0.163) (0.250) (0.157) (0.169) (0.146)

Observations 1,111 1,111 1,111 1,111 967 967 967 967


1st stage F-stat 13.68 17.02 16.85 20.58 11.68 14.42 14.42 16.39
Railway controls 3 3 3 3 3 3 3 3
Geo controls 3 3 3 3 3 3 3 3
Region FE 3 3 3 3 3 3
Land controls 3 3 3 3 3 3
Market access controls 3 3 3 3
State FE 3 3

Notes: *** p<0.01, ** p<0.05, * p<0.1. Robust standard errors in parentheses. All regressions are by instrumental variables
estimated using the LIML estimator with the predicted share of Europeans interacted with region indicators as the instruments
for the actual share of Europeans. Sample includes all municipalities except those indicated in the panel heading. The unit of
observation is a municipality. All variables are standardized to have mean zero and standard deviation one.

Table 4: Potential mechanisms and European immigration

Population Agricultural Cultivated Coffee Cash crops Share farms Share farms
density emp. density share of farms share share (no coffee) with tools with machines
(1) (2) (3) (4) (5) (6) (7)
Share Europeans -0.235 -0.115 0.817*** 1.019*** 0.0225 0.698*** 0.0418
(0.181) (0.170) (0.181) (0.202) (0.118) (0.111) (0.143)

Observations 1,296 1,295 1,287 1,285 1,285 1,289 1,289


1st stage F-stat 22.51 22.54 22.45 22.42 22.42 22.46 22.46
All controls 3 3 3 3 3 3 3

Notes: *** p<0.01, ** p<0.05, * p<0.1. Robust standard errors in parentheses. All regressions are by instrumental variables
estimated using the LIML estimator with the predicted share of Europeans interacted with region indicators as the instruments
for the actual share of Europeans. Sample includes all municipalities. The unit of observation is a municipality. All variables
are standardized to have mean zero and standard deviation one. “All controls” includes all controls in column (4) of Table 2.

42
Table 5: Farm value and potential mechanisms

Farm Value per hectare


(1) (2) (3) (4) (5) (6)
Share Europeans 0.764*** 0.770*** 0.549*** 0.586*** 0.643*** 0.341***
(0.106) (0.0967) (0.127) (0.124) (0.127) (0.131)

Panel A: Labor force


Pop density 0.106** 0.223***
(0.0519) (0.0650)
Agr. emp. density 0.167*** -0.0290
(0.0456) (0.0407)

Panel B: Land use


Cultivated share of farms 0.384*** 0.393***
(0.0589) (0.0597)
Coffee share 0.218*** 0.0519
(0.0421) (0.0319)
Other cash crops share 0.125*** 0.0900***
(0.0240) (0.0200)

Panel C: Tools & Machines


Share farms with tools 0.173*** 0.174***
(0.0474) (0.0390)

Share farms with machines 0.0207 0.00249


(0.0182) (0.0159)
Observations 1,231 1,231 1,231 1,231 1,231 1,231
1st stage F-stat 25.13 27.55 17.81 17.75 18.60 12.50
All controls 3 3 3 3 3 3
Notes: *** p<0.01, ** p<0.05, * p<0.1. Robust standard errors in parentheses. All regressions are by instrumental variables
estimated using the LIML estimator with the predicted share of Europeans interacted with region indicators as the instruments
for the actual share of Europeans. Sample includes all municipalities. The unit of observation is a municipality. All variables
are standardized to have mean zero and standard deviation one. “All controls” includes all controls in column (4) of Table 2.

43
Table 6: Gelbach decomposition for the European share coefficient

Outcome
Farm Land Infrastructure Tools & machines
(1) (2) (3) (4)
Mechanism
Panel A: Labor force
Pop density 0.058* 0.049* 0.078 0.015
(0.032) (0.027) (0.050) (0.012)
Agr. emp. density -0.005 -0.002 -0.014 0.003
(0.005) (0.005) (0.010) (0.006)
Total labor force 0.053* 0.048** 0.064 0.018*
(0.029) (0.024) (0.042) (0.010)
Panel B: Land use
Cultivated share of farms 0.182*** 0.173*** 0.178*** 0.114***
(0.041) (0.041) (0.039) (0.030)
Coffee share 0.015** 0.016** 0.004 0.034**
(0.008) (0.008) (0.008) (0.012)
Other cash crops share -0.004 -0.004* -0.002 -0.004
(0.002) (0.003) (0.002) (0.003)
Total land use 0.193*** 0.184*** 0.179*** 0.144***
(0.039) (0.038) (0.037) (0.029)
Panel C: Tools & Machines
Share of farms with tools 0.060*** 0.058*** 0.046*** 0.083***
(0.012) (0.012) (0.011) (0.016)
Share of farms with machines 0.000 -0.000 -0.000 0.002
(0.000) (0.000) (0.000) (0.007)
Total tools & machines 0.060*** 0.058*** 0.046*** 0.085***
(0.012) (0.012) (0.011) (0.018)
All mechanisms 0.309*** 0.285*** 0.316*** 0.246***
(0.042) (0.40) (0.054) (0.035)
Notes: *** p<0.01, ** p<0.05, * p<0.1. Sample includes all municipalities. The unit of observation is a municipality. All
variables are standardized to have mean zero and standard deviation one. This table presents the results of a Gelbach (2016)
decomposition for the coefficient in columns (4)–(7) of panel A of Table 2. These estimates are by OLS.

44
Table 7: Farm values and foreign ownership

Municipality sample Farm-level sample


Foreign-owned farms Farm value per hectare Land value per hectare
(1) (2) (3) (4) (5)
Share Europeans 0.812*** 0.884** Share foreign workers 0.159*** 0.199***
(0.0685) (0.356) (0.0241) (0.0254)

Foreign-owned farms -0.289 Foreign owner 0.0656 -0.177***


(0.268) (0.0533) (0.0606)
Observations 1,289 1,289 40,693 40,693 40,693
1st stage F-stat 22.46 9.81 R2 0.395 0.380 0.398
All controls 3 3
Municipality FE 3 3 3

Notes: *** p<0.01, ** p<0.05, * p<0.1. Robust standard errors in parentheses. Regressions in columns (1) and (2) are
by instrumental variables estimated using the LIML estimator with the predicted share of Europeans interacted with region
indicators as the instruments for the actual share of Europeans in a sample including all municipalities where the unit of
observation is a municipality. Regressions in columns (3)–(5) are by OLS using the São Paulo farm-level data. All variables are
standardized to have mean zero and standard deviation one. “All controls” includes all controls in column (4) of Table 2.

45
Table 8: Implications for development and structural change

Panel A: Human capital formation (literacy)


Population Brazilians Females Children
(1) (2) (3) (4) (5) (6) (7) (8)
Share Europeans 0.808*** 0.812*** 0.761*** 0.772*** 0.735*** 0.720*** 0.753*** 0.766***
(0.181) (0.168) (0.181) (0.169) (0.187) (0.159) (0.162) (0.149)
Observations 1,296 972 1,296 972 1,296 972 1,296 972
1st stage F-stat 22.51 16.37 22.51 16.37 22.51 16.37 22.51 16.37
Panel B: Economic structure
Employed Employed Employed
Agriculture Industry Rentiers Pub. admin
Share Europeans -0.418** -0.373** 0.279 0.184 0.0969 -0.00726 -0.186 -0.0216
(0.185) (0.190) (0.191) (0.178) (0.174) (0.219) (0.170) (0.171)
Observations 1,295 971 1,295 971 1,295 971 1,295 971
1st stage F-stat 22.54 16.41 22.54 16.41 22.54 16.41 22.54 16.41
Panel C: State capacity
Tax Tax no exports Public education Public services
Share Europeans 0.553* 0.789*** 0.751** 0.884*** 0.162 0.0167 0.233 0.282**
(0.326) (0.235) (0.324) (0.273) (0.143) (0.120) (0.204) (0.130)
Observations 1,237 924 1,097 817 1,031 768 1,031 768
1st stage F-stat 21.55 15.75 20.28 15.35 16.89 12.89 16.89 12.89
Panel D: Female participation in labor markets (female to male ratio)
Population Employed Agriculture Industry
Share Europeans -0.459*** -0.521*** -0.195 -0.0594 -0.243 -0.0267 0.274 0.205
(0.131) (0.169) (0.144) (0.160) (0.180) (0.189) (0.178) (0.240)
Observations 1,296 972 1,295 971 1,295 971 1,295 971
1st stage F-stat 22.51 16.37 22.54 16.41 22.54 16.41 22.54 16.41
All controls 3 3 3 3 3 3 3 3

Notes: *** p<0.01, ** p<0.05, * p<0.1. Robust standard errors in parentheses. All regressions are by instrumental variables
estimated using the LIML estimator with the predicted share of Europeans interacted with region indicators as the instruments
for the actual share of Europeans. Sample in columns (1), (3), (5), and (7) includes all municipalities. Columns (2), (4), (6),
and (8) exclude municipalities in the top 25 percentiles of population. The unit of observation is a municipality. All variables
are standardized to have mean zero and standard deviation one. “All controls” includes all controls in column (4) of Table 2.

46

You might also like