American Economic Journal: Applied Economics 2017, 9(4): 105–136

https://doi.org/10.1257/app.20150532

Human Capital Persistence and Development†

By Rudi Rocha, Claudio Ferraz, and Rodrigo R. Soares*

This paper documents the persistence of human capital over time

and its association with long-term development. We exploit variation
induced by a state-sponsored settlement policy that attracted immi-
grants with higher levels of schooling to particular regions of Brazil
in the late nineteenth and early twentieth century. We show that one
century after the policy, municipalities that received settlements had
higher levels of schooling and higher income per capita. We pro-
vide evidence that long-run effects worked through higher supply of
educational inputs and shifts in the structure of occupations toward
skill-intensive sectors. (JEL I26, J22, J24, J61, N36, O15, Z13)

T he sixteenth century European colonial expansion and the ensuing migratory

waves out of the Old World permanently changed human societies. While a
number of studies document how these historical events shaped current patterns
of economic development (Acemoglu, Johnson, and Robinson 2001; Glaeser et al.
2004), a large debate remains on what mechanisms explain this persistence. On
the one hand, Acemoglu, Johnson, and Robinson (2005) suggest that persistence
occurred through institutional and political channels.1 On the other hand, a growing
set of papers suggests that culture and human capital are important channels through
which historical shocks persist through time.2
This paper documents the persistent effects of state-sponsored settlements
located in the Brazilian state of São Paulo and examines the channels behind this
persistence. After the international ban on slave trade in 1850, and in the midst of a
massive inflow of European immigrants to Brazil, immigrants with relatively more

* Rocha: Institute of Economics, Federal University of Rio de Janeiro, Av. Pasteur 250, 22290-240, Rio de
Janeiro, RJ, Brazil (email: rudi.rocha@ie.ufrj.br); Ferraz: Department of Economics, Pontifícia Universidade
Católica do Rio de Janeiro (PUC-Rio), Rua Marquês de São Vicente 225, Rio de Janeiro, RJ, 22453-900, Brazil
and BREAD (email: cferraz@econ.puc-rio.br); Soares: Columbia University, 420 W. 118th Street, New York, NY,
10027 and Sao Paulo School of Economics-FGV (email: r.soares@columbia.edu). We are grateful to three anony-
mous referees for suggestions that greatly improved the paper. We also thank Ran Abramitzky, Juliano Assunção,
Maria Silvia Bassanezi, Ciro Biderman, Adalberto Cardoso, Ernesto Dal Bó, Irineu de Carvalho Filho, Fred Finan,
Francisco Gallego, Gustavo Gonzaga, Joana Naritomi, Naércio Menezes-Filho, André Portela, Pablo Querubin,
Eustáquio Reis, Nico Voigtländer, Hans-Joachim Voth, and seminar participants at Bocconi University, Paris School
of Economics, PUC-Rio, São Paulo School of Economics—FGV, Stanford University, Universidad Carlos III de
Madrid, University of Essex, the LACEA 2010 Meeting, the SBE 2010 Meeting, and the First Economic History
Clio Lab Workshop for comments, data, and suggestions. Ferraz gratefully acknowledges financial support from
the EH Clio lab at PUC-Chile.
†
 Go to https://doi.org/10.1257/app.20150532 to visit the article page for additional materials and author
disclosure statement(s) or to comment in the online discussion forum.
1 
Dell (2010) provides evidence that extractive institutions in Peru generated persistence in the protection of
property rights that explain current differences in welfare. 
2 
See, for example, Guiso, Sapienza, and Zingales (2016); and Lowes et al. (2015) on cultural persistence and
Wantchekon, Klašnjaz, and Novta (2015); and Dittmar and Meisenzahl (2016) on the role of human capital. 

105
106 AMERICAN ECONOMIC JOURNAL: APPLIED ECONOMICS OCTOBER 2017

education were channeled into specific localities through deliberate government

policies. In the end of the nineteenth and beginning of the twentieth century, pub-
lic authorities established a number of official settlement colonies throughout the
state of São Paulo. This policy had goals involving occupation of territory and
food production, and was driven by a centralized decision at the state level. The
settlements were established typically near previously existing rural villages and
were occupied by relatively high-skill European immigrants of various nation-
alities. Simultaneously, European immigrants were also spontaneously arriving
in virtually every other area of the state. Given the ambivalent motivations, the
distribution of state-sponsored settlements ended up being roughly uniform across
the region.
We show that settlements attracted a selected type of immigrant and that, analo-
gously to the results from the literature on institutions (such as Acemoglu, Johnson,
and Robinson 2001), this selection persisted through time and was associated with
higher income per capita in the long run. But different from this literature, we exploit
variation across municipalities within the same country. Hence, our findings cannot
be explained by differences in formal political institutions. Our findings cannot be
accounted for either by the dimensions of culture and social capital that we can
observe in the data (nationality, religion, or other cultural traits possibly associated
with the incidence of entrepreneurship). In terms of proximate determinants, we
provide evidence that the local shock induced by state-sponsored settlements per-
sisted and ended up reflected on local economic development through higher supply
of schooling inputs and faster structural change.
We exploit three unique features of this historical episode to characterize the per-
sistence of the initial compositional shock through time and its effects on long-term
development. First, differently from the United States where immigrants were, on
average, less educated than the native population (see Bandiera et al. 2015), the age
of mass migration in Latin America, and especially in Brazil, brought immigrants
who were more educated than natives, creating a positive human capital shock. In
the case of state-sponsored settlements in São Paulo, this pattern was reinforced by
a particularly positive selection of immigrants in terms of education and possibly
also other unobservable skills. Second, settlements were established before Brazil’s
industrialization period. Thus, the choices of settlement location and migration deci-
sions in an agrarian economy were not correlated with expected returns from edu-
cation or entrepreneurship. And third, the areas where state-sponsored settlements
were created were not different from other areas, but for the fact that they received
different immigrants.
Our main empirical strategy compares literacy rates and other outcomes across
municipalities that received a state-sponsored settlement and municipalities that did
not receive a settlement but had similar preexisting characteristics. To measure the
effects of state-sponsored settlements, we put together data from the 1872, 1920,
1940, and 2000 Brazilian censuses, as well as a vast array of auxiliary information
from various historical sources, such as state statistical yearbooks and provincial
presidential reports. We construct variables that measure geographic, transportation,
and socioeconomic conditions within the state from the end of the nineteenth cen-
tury to the year 2000.
VOL. 9 NO. 4 ROCHA ET AL.: HUMAN CAPITAL PERSISTENCE AND DEVELOPMENT 107

We begin by documenting that in 1872, before the establishment of the first

state-sponsored settlements, localities that were to receive settlements in the future
were very similar to other areas in the state of São Paulo. We then show that the
creation of settlements attracted more educated immigrants to settlement municipal-
ities. In 1920, a few years after the establishment of the last settlements, the literacy
rate in settlement municipalities was 8 percentage points (or 27 percent) higher
than elsewhere in the state, despite an only marginally higher share of immigrants.
We focus most of our discussion throughout the analysis on educational outcomes,
but recognize that these may be associated with other unobservable differences in
individual traits.
We show that this initial shock persisted not only until 1940, but also all the way
up to 2000. Approximately one century after the establishment of many of the orig-
inal settlements, individuals living in the respective municipalities had on average
0.52 more year of schooling when compared to individuals living in other munici-
palities. This persistent difference in schooling trajectories was also followed by dif-
ferential development paths. Our estimates suggest that average income per capita
in 2000 was 15 percent higher in municipalities that had received a state-sponsored
settlement in the beginning of the century when compared to other municipalities.
We implement a series of robustness tests to make sure the results are not driven
by other factors. First, we show that all results are robust to controlling for geo-
graphic and presettlement (1872) socioeconomic characteristics. Second, the results
remain significant when we use municipality variation and control for municipal
fixed effects (when possible). The results are unchanged as well when we control for
proxies of culture and social norms associated with nationality (e.g., Italian versus
Portuguese) and religion (e.g., Catholic versus Protestant). Finally, we show that
other potential immediate effects of the settlements, such as those on property rights
or income, do not drive our results. We find no short-term (1920) effect of settle-
ments on land inequality, agricultural productivity, land prices, and wages, and no
long-term effect on entrepreneurship.
In sum, we show that the initial shock represented by the skill composition of
immigrants—along educational and possibly other unobservable dimensions—per-
sisted through time and was associated, in the long run, with higher income per
capita. In order to shed light on the persistence of this initial shock and on how
it translated into higher education and income per capita over time, we analyze
the evolution in the supply and use of educational inputs and in the structure of
employment. We show that in 1920, 1940, and 2000 enrollment rates and the num-
ber of teachers per school-aged children were higher in locations associated with
settlements. In these same locations, employment shifted away from agriculture and
toward manufacturing and, to a lesser extent, services between 1920 and 1940. In
2000, this shift in the structure of employment ended up reflected entirely on a larger
services sector. None of these differences in the use of educational inputs and in the
structure of employment were present in 1872, before the settlements were estab-
lished. We also see evidence that settlement municipalities started attracting more
educated domestic immigrants in the second half of the twentieth century.
Though we cannot identify precisely why and how these communities demanded
more investments in schooling, one might conjecture that the initially higher set
108 AMERICAN ECONOMIC JOURNAL: APPLIED ECONOMICS OCTOBER 2017

of skills was correlated with higher intrinsic values attributed to schooling or with
higher political participation. Irrespectively, human capital was associated with a
higher share of employment in manufacture already in 1920, when it did not trans-
late into higher incomes. The persistently higher human capital, nevertheless, pos-
sibly together with some previous specialization in manufacture, allowed these
communities to take better advantage of the industrialization process initiated in the
following decades. This combination, in turn, led to higher income per capita in the
long run and attracted even more skilled domestic immigrants.
Our paper speaks to three sets of literature. First, it relates to a large literature
that measures the impact of population composition and great migration episodes
on long-run economic development (e.g., Putterman and Weil 2010, Easterly and
Levine 2016). Distinct from the existing work, we use micro data from a single
country to exploit a natural experiment that distributed heterogeneous types of
immigrants across localities. By looking at a single state within Brazil, we imme-
diately hold constant the de jure macro institutions that have been the focus of a
large and influential literature (e.g., Acemoglu, Gallego, and Robinson 2014). We
also provide evidence on the specific channels through which the composition of
immigrants affected long-term development. Finally, while there is a large litera-
ture on the consequences of the Age of Mass Migration for the United States (e.g.,
Abramitzky, Boustan, and Eriksson 2014), there is very little evidence for South
America. Our paper is one of the first to provide evidence on the long-run effects of
more skilled immigrants on economic development.3
Second, the paper contributes to a growing literature that documents the per-
sistence of human capital through time. Huillery (2009) shows that colonial invest-
ments in education in West Africa had effects that persist until today: areas that
received higher investments in education in the colonial past still display higher
educational investments and higher levels of schooling. Similarly, but in a different
setting, Wantchekon, Klašnjaz, and Novta (2015) show that individuals selected to
attend school in colonial Benin, during the establishment of the first missionary
schools, benefited in terms of income and socioeconomic conditions. In addition,
in the long-run, the entire villages where historical missionary schools were estab-
lished displayed improved outcomes. Nunn (2011) also finds persistent effects of
Protestant missionary schools on educational levels in Africa. Differently from these
papers, our results address the mechanisms behind the persistence of human capital
and its correlation with income per capita.4
And third, our work sheds light on the role of skill composition as a determinant
of structural change and long-term development. A growing number of papers have
focused on the importance of education for productivity improvements through the
absorption of new technologies. Hornung (2014), for example, looks at the forced
migration of skilled Huguenots from France to Prussia in the late seventeenth century

3 
See also the related work of Droller (forthcoming) for the case of Argentina. 
4 
For a related literature that focuses on Latin America, see Summerhill (2010), de Carvalho Filho and Colistete
(2010), de Carvalho Filho and Monasterio (2012), Valencia Caicedo (2014), and Droller (forthcoming). Our paper
differs from this literature by having a clear identification strategy based on the location of state-sponsored settle-
ments and by analyzing the specific channels behind the historical persistence of human capital and its association
with long-term development. 
VOL. 9 NO. 4 ROCHA ET AL.: HUMAN CAPITAL PERSISTENCE AND DEVELOPMENT 109

and shows that firms in areas receiving immigrants experienced increased productiv-
ity. The author argues that skilled Huguenots brought industry-specific knowledge
that was complementary to new technologies. Fourie and von Fintel (2014) look at
Huguenots that left wine producing regions of France to migrate to South Africa
in this same historical episode and show, using tax records, that they became more
productive wine makers than previously established farmers and that this difference
in productivity persisted through time. Becker, Hornung, and Woessmann (2011)
also look at Prussia and use occupational data to show that historical variations in
schooling across counties, supposedly related to cultural factors, determined the
ability of the different regions to take advantage of the technological innovations
brought by the industrial revolution. Squicciarini and Voigtländer (2015) present
evidence that “upper-tail knowledge” was an important driver of city growth during
the first industrial revolution in France, mainly through increased productivity in
industrial technologies. For a more recent period and looking at cross-country data,
Ciccone and Papaioannou (2009) show that, during the 1980s and 1990s, countries
with higher initial levels of education experienced higher growth in industries inten-
sive in new technologies. Many authors also find a positive relationship between
human capital and growth exploiting cross-sectional variation within countries,
though mostly focused on contemporary data from developed regions (for example,
Glaeser, Scheinkman, and Shleifer 1995; Shapiro 2006; and Iranzo and Peri 2009).
By documenting the persistently higher use of educational inputs and the shift in
the structure of occupations towards skill-intensive sectors, we are able to present
in a single setting, and in a unified and sequential way, several results that appeared
fragmented in the previous literature. We show that the initial heterogeneity in edu-
cation persisted through time via increased investments in schooling, as in Huillery
(2009); higher educational levels—possibly also correlated with other skills useful
in manufacturing—were associated, as industrialization progressed over time, with
shifts in the structure of employment towards skill-intensive sectors, as in Becker,
Hornung, and Woessmann (2011); and together, these changes ended up reflected
on higher income per capita in the long run.
The remainder of the paper is structured as follows. Section I outlines the his-
torical background and discusses the establishment and the likely impacts of
state-sponsored settlements in the state of São Paulo. Section II presents the data
and descriptive statistics. Section III describes the empirical strategy. Section IV
presents the main econometric results. Section V discusses channels and performs
robustness exercises. Section VI closes the paper with some concluding remarks.

I.  Historical Background

A. Historical Context and the Policy of State-Sponsored Settlements

São Paulo has long been among the richest and most industrialized regions not
only in Brazil but in all of the developing world. The transition from agrarian pro-
ducer to industrial and financial center spanned several decades, culminating with its
consolidation as Brazil’s wealthiest state in the second half of the twentieth century.
This process was fueled by the nineteenth century expansion in coffee production
110 AMERICAN ECONOMIC JOURNAL: APPLIED ECONOMICS OCTOBER 2017

and the construction of an integrated railway system, which connected the seaport
of Santos to the fertile and sparsely occupied inland areas (Furtado 1959). Coffee
became then a major driver of infrastructure construction, urban development, and
accumulation of wealth, and quickly took over the agricultural frontier (Martins
1973 and Summerhill 2003).
The labor demanded during the first years of the coffee expansion was primarily
supplied by reallocation of slaves within the country. But the ban on the interna-
tional slave trade in 1850 (Eusébio de QueirÓs Law) and the abolishment of slavery
in 1888 (Lei Áurea) precluded the use of slave labor as an input in the continued
growth of the state. Already by the mid-1850s, rising popular pressure for the end of
slavery and increasing shortages of workers and food brought international immi-
gration to the center of the political agenda.5 At that time, the view that European
immigrants were more cultured and productive than afro-descendants and native
workers was widespread among the Brazilian elite (Petrone 1982). In the second
half of the nineteenth century, openness to foreign immigration in general, and to
the model of state-sponsored colonies in particular, gained increased support in leg-
islative debates (Martins 1973).
State-sponsored settlements had two main objectives: food production for the
growing urban areas and attraction of “higher” quality immigrants, through the
prospect of land ownership, to uninhabited regions. The geographic distribution of
settlements across the state reflected this dual policy motivation (Petrone 1982).
On the one hand, the role as food suppliers led to the creation of official colonies
in areas with growing demand. On the other hand, in many other cases, unclaimed
public lands used in the establishment of settlements were located in marginal or
backward areas.
The administrative procedures associated with the creation of official settlements
can be described as follows. After legislative approval, public employees searched
for locations mainly among unclaimed public lands, but also among private lands
and foreclosure opportunities. Given the alternatives available, the precise locality
was chosen based on two criteria: reasonable soil fertility and proximity to trans-
portation. Plots of land were then outlined and classified into three different types
depending on size and location within settlement. Rural plots were the largest while
suburban and urban plots were smaller and more central. The government was
responsible for building the basic infrastructure around an administrative office at
the center of the colony.
In the 1870s, the government launched a first wave of state-sponsored settlements
in São Paulo. Official settlements became a major attraction in the official Brazilian
propaganda disseminated in European cities (Petrone 1982). The plots of land
were offered by Brazilian representatives in Europe and at the Immigration Station

5 
The first experiences of state-sponsored international immigration in Brazil date back to the early nineteenth
century, when the imperial government promoted the peopling of strategic and sparsely occupied regions with
European families. The first official settlements were the colonies of Nova Friburgo (1818) in the state of Rio de
Janeiro and São Leopoldo (1824) in the state of Rio Grande do Sul. This initial wave of colonization also included
four areas in the state of São Paulo: two in the surroundings of the state capital (Santo Amaro and Parelheiros, in the
1820s) and two in the coast (Iguape and Cananéia, in the 1850s). Due to economic isolation, the initiatives in São
Paulo failed and the farthest colonies of Parelheiros, Iguape, and Cananéia were largely abandoned (Furtado 1959,
Petrone 1982, Paiva 1993, and Siriani 2005). 
VOL. 9 NO. 4 ROCHA ET AL.: HUMAN CAPITAL PERSISTENCE AND DEVELOPMENT 111

Houses, the lodgings where immigrants were first allocated to when arriving in
Brazil. Immigrants did not receive the land freely, but had to pay for it in installments
starting after the first harvest. During this initial period, some scant public support
was provided: food, seeds, agricultural tools, and payments for improvements in
infrastructure. Public employees were in charge of coordinating and regulating the
developments within official settlements. After all installments were paid for, the
settler could claim the property title for his plot. In general, a settlement was eman-
cipated from public administration once all of its plots had been entirely paid for.
The whole process seems to have been kept under serious political account-
ability. Administrative records and detailed documentation on why some areas
were selected were often presented in Annual Provincial Presidential Reports, in
Messages to the Legislative Assembly, and in other administrative documents.6
Over most of the period, state and central governments also paid for or partly sub-
sidized travel expenses, including both international and internal displacements. In
addition, through the Immigration Station Houses, local authorities coordinated the
job matching between workers and farmers, and the allocation of immigrants to
colonial settlements.
In the 1880s, with the imminence of the abolishment of slavery, openness to
immigration and the policy of colonial settlements gained renewed strength. The
political reforms that came with the 1891 constitution, resulting from the end of
the empire and the rise of the republic in 1889, decentralized immigration policy
and the management of unclaimed public lands to the states (Iotti 2001). São Paulo
then achieved prominence as the main receiving region, and immigration to the state
reached a massive scale, absorbing a major share of the total inflow of European
families arriving in Brazil. Online Appendix Figure A1 presents the cumulative
inflow of immigrants into the state from 1872 onwards, and the share of São Paulo in
the yearly inflow of immigrants to Brazil. Between 1872 and 1920, roughly 1.8 mil-
lion immigrants entered São Paulo, corresponding to more than 200 percent of the
initial population of the state in 1872 (837,354 inhabitants). This inflow amounted
to 53 percent of the total number of immigrants received by Brazil in the period, but
reached over 80 percent in specific years at the turn from the nineteenth to the twenti-
eth century. Yet only a small fraction of immigrants ended up in ­state-sponsored col-
onies. The vast majority of immigrants arriving at the Immigration Station Houses
were rapidly locked into long-term contracts offered by coffee farmers.7

6 
For example, the 1877 Annual Report of the Secretariat of Agriculture of São Paulo explains in detail the
creation of the settlement colonies of São Caetano, São Bernardo, Glória, and Santana; the 1885 Annual Provincial
Presidential Report explains the choice of locality for the settlements of Canas and Cascalho; and the 1907 Annual
Provincial Presidential Report details the procedures for the creation of the settlements of Nova Europa and Gavião
Peixoto. 
7 
Data on the number of immigrants arriving during the initial establishment of the settlements does not exist
in a systematic way, but historical sources provide information for specific cases. In particular, the 1910s Annual
Statistical Reports provide data for 12 settlements created between 1905 and 1911, for which we have the total num-
ber of arrivals between 1912 and 1920. The arrival into these 12 settlements during this period represented 1.5 per-
cent of the total number of immigrants arriving in São Paulo. It is difficult, though, to read this number as more
than a simple ballpark figure for the share of immigrants going to the settlements. First, we do not have data on the
remaining settlements. Second, entry into these 12 settlements may have been substantially higher in the first years
immediately after foundation (1905–1911) when compared to the period for which we have data (1912–1920). And
third, entry may have been lower during other periods in the 1872–1920 interval. 
112 AMERICAN ECONOMIC JOURNAL: APPLIED ECONOMICS OCTOBER 2017

The specific sequence of creation of state-sponsored settlements in São Paulo

can be outlined as follows. In the 1870s, four settlements were created around the
state capital: Santana and Glória, nowadays neighborhoods of the capital, and São
Bernardo and São Caetano, nowadays independent municipalities in the metropoli-
tan area of the capital. From the mid-1880s to the late 1890s, 13 official settlements
were created: four in the economically backward Vale do Paraíba, two in the central
area of the state, four in the expanding coffee region of the Northwest, two in the
Midwest (where the main agricultural products were cotton and sugarcane), and
one in the coastal Vale do Ribeira. In the early 1900s, nine additional colonies were
founded in still isolated areas of the Northwest.8 Finally, following a renewed inter-
est in populating lagging regions, the federal government established two official
settlements in the state: one at the southwestern border with the state of Paraná and
another at the border with the state of Rio de Janeiro (Iotti 2001).9
In total, 28 official settlements were created and consolidated from 1872 to 1920.
In terms of the 1920 administrative division of the state, settlements were distributed
across 10 percent of the existing municipalities (20 municipalities had received a
settlement by 1920). Different motivations behind the colonization policy led to a
more or less uniform distribution of official colonies across the various regions in
the state. This pattern is discussed in detail in the next session.

B. Characteristics of the State-Sponsored Settlements

Table 1 lists the state-sponsored settlements included in our analysis, present-

ing basic information on dates of foundation and—when available—emancipation
and location (in terms of the original and current political division of municipali-
ties). Online Appendix Table A1 provides further information, when available, on
the initial area, predominant immigrant nationalities, as well as the main sources
of information used. The 28 colonies were founded between 1877 and 1911, 17 of
them in the end of the nineteenth century and the remaining 11 in the first years of
the twentieth century. Emancipation came, on average, 10 years after foundation.
The size of settlements varied considerably: those created between the late 1870s
and the early 1880s had areas mostly between 1,000 and 3,000 hectares, while those
created in the 1910s were somewhat larger, roughly between 3,000 and 5,000 hect-
ares. Still, in both periods, a few settlements were created with much larger areas,
reaching over 10,000 hectares. Through time, settlements sometimes expanded geo-
graphically as the number of settlers increased. Information from the 1910s Annual
Statistical Reports of São Paulo shows that, a few years after initial operation, the
average population of settlements reached typically from 1,000 to 3,000 inhabitants.

8 
The state also created the colony of Conde do Pinhal in the coastal municipality of Ubatuba. However, the first
settlers immediately abandoned the place and the project did not take off. Official justifications for the failure are
presented in the 1907 and 1908 Annual Reports of the Secretariat of Agriculture. For this reason, this colony is not
included in the analysis. 
9 
There were a few more federal settlement attempts in Vale do Ribeira after the 1929 crisis, but these projects
were closer to regulated land occupations than to active immigration promoting initiatives (Paiva 1993). For this
reason, these cases are not included in the analysis. 
VOL. 9 NO. 4 ROCHA ET AL.: HUMAN CAPITAL PERSISTENCE AND DEVELOPMENT 113

Table 1—Basic Information on the State-Sponsored Settlements Considered in the Analysis,

São Paulo, Brazil

Settlements Year of Year of Original Current

foundation emancipation municipality municipality

Glória 1877 N/A São Paulo São Paulo

Santanna 1877 N/A São Paulo São Paulo
São Bernardo 1877 1901 São Bernardo São Bernardo
São Caetano 1877 N/A São Bernardo São Caetano
Canas 1885 1893 Lorena Canas
Cascalho 1885 1893 Limeira Cordeirópolis
Pariquera-açu 1887 1901 Iguape Pariquera-au
Senador Antônio Prado 1887 1893 Ribeirão Preto Ribeirão Preto
Ribeirão Pires 1887 1893 São Bernardo Ribeirão Pires
Rodrigo Silva 1887 1893 Porto Feliz Porto Feliz
Boa Vista 1888 1893 Jacareí Jacareí
Barão de Jundiaí 1887 1893 Jundiaí Jundiaí
Sabaúna 1889 1901 Mogi das Cruzes Mogi das Cruzes
Quiririm 1890 1893 Taubaté Taubaté
Piagui 1892 1901 Guaratinguetá Guaratinguetá
Bom Sucesso 1894 1899 Sorocaba Sorocaba
Campos Sales 1897 1903 Campinas Cosmópolis
Jorge Tibiriçá 1905 1923–24 Rio Claro Corumbataí
Nova Odessa 1905 1920–21 Campinas Nova Odessa
Gavião Peixoto 1907 1923–24 Araraquara Gavião Peixoto
Nova Paulicéia 1907 1920–21 Araraquara Gavião Peixoto
Nova Europa 1907 1920–21 Ibitinga e Araraquara Nova Europa
Bandeirantes 1908 N/A S. José do Barreiro S. José do Barreiro
Monção 1910 N/A S. Bárbara do Rio Pardo Iaras
Nova Veneza 1910 1919 Campinas Sumaré
Conde de Parnaíba 1911 1919 Mogi-Mirim Conchal
Martinho Prado Jr. 1911 1920–21 Mogi-Guaçu Mogi-Mirim
Visconde de Indaiatuba 1911 1923–24 Mogi-Mirim Mogi-Mirim

Note: N/A (Non-available) indicates that we could not find a source containing the respective information of the
settlement.

The most common nationalities were Italian and German, but Spanish, Portuguese,
and Brazilian, among others, were also present (Bassanezi et al. 2008).10
Detailed data on immigrants initially arriving during the establishment of the
settlements does not exist in a systematic way, but historical sources provide infor-
mation for some specific cases: the 1888 Annual Statistical Report of São Paulo
discusses in detail three settlements created in the 1870s; the 1898 Annual Report
of the Secretariat of Agriculture presents data on another 6 settlements, mostly cre-
ated in the 1880s and 1890s; and the 1910 Annual Statistical Reports mention 12
additional settlements created between 1905 and 1911. We calculate literacy rates
for these cases and compare them to the average literacy across municipalities in the
state of São Paulo at two points in time: 1890 and 1920 (using data from the respec-
tive censuses). Online Appendix Figure A2 presents this information. The average
literacy rate across these settlements was 40 percent, as compared to 12  percent
across municipalities in the state of São Paulo in 1890, and 23 percent in 1920. The
settlements surveyed in the first two reports had average literacy rates of 32 percent,
while those mentioned in the reports from the 1910s had literacy rates of 47 percent,

10 
Brazilians were allowed to purchase plots of land within settlements under certain conditions. 
114 AMERICAN ECONOMIC JOURNAL: APPLIED ECONOMICS OCTOBER 2017

indicating some improvement in the pool of immigrants over time (for the inter-
ested reader, online Appendix Figure A2 plots these numbers). Overall, the popula-
tion within settlements had substantially higher literacy rates when compared to the
remainder of the state.
While we are unaware of any study that identified this positive selection statisti-
cally and explained its origins, some hypotheses naturally arise. The state-sponsored
settlements were populated in two ways. First, Brazilian representatives in European
cities could select settlers and negotiate available plots of land. There is evidence
that representatives searched for more skilled individuals. This is illustrated by
the 1907 Provincial Presidential Report (p. 353), where it is recorded that “the
inflows of immigrants that have entered into the state, brought by the Immigration
Commissariat in Antwerp and originating from northern Europe, indicate that we
managed to find the convenient element for occupying the official colonies” (trans-
lated by the authors, emphasis added). Other examples come from the purposeful
selection of Russian families for the settlement of Nova Odessa and Belgium fam-
ilies for the settlement of Rodrigo Silva, reported, respectively, in the 1905 Annual
Report of the Secretariat of Agriculture and the 1888 Provincial Presidential Report.
Second, available plots of land within colonies were also offered at the Immigration
Station Houses. Thus, upon arrival in São Paulo, immigrants sometimes had a
choice between official settlements and job opportunities in coffee farms. As the
1906 Provincial Presidential Report (p. 43) mentions, “the immigration trends will
increase due to the official propaganda the government is disseminating abroad, and
whose efficacy has been already observed in the spontaneous arrival of immigrants
searching for the official colonies” (translated by the authors). Official selection
abroad and self-selection upon arrival may have channeled more educated immi-
grants into official settlements.
Despite the positive selection of immigrants, life conditions in the state-sponsored
settlements were not particularly good. Official documents explicitly recommended
that settlements be placed in areas close to transportation and with reasonable soil
quality.11 However, the process of choice of location was not homogeneous and,
as a result, soil quality varied considerably. On the one hand, some sources docu-
ment that engineers responsible for the settlements’ location weighed in favor of
soil fertility when deciding where to establish colonies (see Martins 1973). On the
other hand, selection of location was conditional on alternatives available among
unclaimed public lands and unused private or foreclosure lands purchased by the
government. These were areas unlikely to be highly productive or particularly suit-
able for coffee, the main crop at the time. In fact, some official reports mention
explicitly the low quality of soils in certain settlements (for example, see discussion
in Martins 1973 and the 1899 Annual Report of the Secretariat of Agriculture).
The 1910s Annual Statistical Reports of São Paulo present detailed information
on the economic conditions within 11 settlements, most of them created between

Martins (1973) details this point when describing the procedures behind the choice of location for the colony
11 

of São Caetano. The recommendation on transportation and soil fertility is explicit in many administrative reports
and presidential speeches that justified the choice of location of official settlements. Examples can be found in
the 1877 and 1885 Provincial Presidential Reports, and in the 1885 and 1892 Annual Reports of the Secretariat of
Agriculture. 
VOL. 9 NO. 4 ROCHA ET AL.: HUMAN CAPITAL PERSISTENCE AND DEVELOPMENT 115

1905 and 1911. We use this information to compare the agricultural productivity
within colonies to the average productivity in the state. Productivity is measured as
the ratio between the value of production (agricultural, extractive, and animal) and
total population. To smooth production and population fluctuations within colonies,
we calculate the average of this ratio for the years between 1915 and 1920 (when
data are available). We then compare this productivity to the per capita value of
agricultural production in the state, reported in the 1920 census (for the interested
reader, online Appendix Figure A3 plots these numbers). The annual per capita pro-
duction within colonies (median around 225 réis) was systematically lower than
the state-level production (average of 531 réis). If anything, this suggests that soil
quality may have provided relatively poor levels of subsistence within settlements.
But it is difficult to tell whether this lower productivity reflected the difficulties typ-
ically faced in maintaining the colonies in the first years immediately after occupa-
tion, or indeed lower productivity in the long run. In any case, there is no evidence
suggesting that the quality of the soil in state-sponsored settlements was above that
observed elsewhere in the state.
Examples of the challenges faced in the first years after occupation can be found
in the 1898 and the 1905 Annual Reports of the Secretariat of Agriculture, where
the cases of the colonies of Campos Salles, Nova Odessa, and Jorge Tibiriçá are
discussed. Still, relatively few immigrants left the colonies. The 1910s Annual
Statistical Reports of São Paulo provide data on arrivals and departures for 11
state-sponsored settlements (presented in online Appendix Figure A4). For each set-
tlement, we have data on cumulative arrivals, departures, and net growth from 1912
to 1918 (as shares of the 1912 population).12 The 11 settlements for which data are
available grew substantially during this period, on average by 66 percent. In 73 per-
cent of the cases, outflows were considerably low, below 20 percent of the initial
population. In the remaining 3 colonies, outflows were substantial, reaching close to
or above 100 percent in the cases of Bandeirantes and Monção. But, even in these
cases, population grew by, respectively, 29 and 242 percent. So the extreme inflows
and outflows in these two colonies seem to be related to the then recent creation of
the settlements—dated from 1908 and 1910—and to the still unstable initial occu-
pation, rather than to a dynamics typical of state-sponsored colonies.

C. Why Did State-Sponsored Settlements Matter?

In principle, the establishment of state-sponsored settlements in São Paulo could

represent a many-folded intervention. Without further historical information, one
could picture such an event as affecting human capital levels, culture, land concen-
tration, property rights, and initial agglomeration. All of these factors have received
substantial attention from specific literatures within development economics.
The evidence we present in the next sections shows that the main distinguishing
feature of the settlements was the attraction of a different pool of immigrants, with
higher educational levels and possibly also other unobservable traits. The literature

12 
This cumulative number excludes 1917, since the Annual Statistical Report of São Paulo was not available
for this year. 
116 AMERICAN ECONOMIC JOURNAL: APPLIED ECONOMICS OCTOBER 2017

on growth and development has repeatedly shown that initial levels of human cap-
ital are correlated with subsequent growth, be it across countries or across regions
or cities within a country (see, for example, Barro 1991; and Glaeser, Scheinkman,
and Shleifer 1995). The theoretical link between human capital and growth dates
back to the early endogenous growth models, such as Lucas (1988), where there are
internal (individual improvements) and external (positive externalities) effects of
human capital on productivity. Moretti (2004), for example, argues that knowledge
spillovers increase aggregate productivity over and above the direct effect of human
capital on individual wages.
Still, it may not be immediately obvious why this initial compositional shock
would have persisted over long periods of time. The literature on the history of
schooling in Brazil documents that immigrants from more educated groups actively
demanded more public investments in schooling, which were typically consolidated
through community ethnic schools (Kreutz 2000, Maschio 2005). Based on descrip-
tive historical information, this literature shows that Brazilian states that adopted
immigration as a systematic policy to attract labor experienced a process of substan-
tial expansion in schooling. According to Maschio (2005), ethnic schools benefited
not only the children of immigrants but also children of Brazilian families who pre-
viously inhabited the areas. Maschio documents the specific demands of immigrants
and respective government responses, culminating with the creation of community
schools, as in the case of some state-sponsored colonies in Paraná. It seems natural
to think that similar processes could have operated as well whenever more educated
immigrants were concentrated in other state-sponsored settlements. But it is import-
ant to keep in mind that, as mentioned in the introduction, it is possible that these
persistent differences in education were also associated with other unobservable
productive characteristics, such as entrepreneurship or previously acquired manu-
facturing skills, all of which may contribute to long-term development.
We also consider some additional hypotheses in our empirical exercise and dis-
cuss them briefly here. First, settlements may have changed the distribution of land
within settled regions. Land was partitioned into relatively small plots to be allocated
to settlers. This process could have led to a reduction in land inequality. It might also
have affected property rights, since lands used to establish colonies were sometimes
from unclaimed public areas otherwise potentially subject to dispute through vio-
lence. Both land inequality and property rights have been identified as important
determinants of long-term development (see, for example, Acemoglu, Johnson, and
Robinson 2005; Galor, Moav, and Vollrath 2009; and North 2009). Other relevant
aspects have to do with social norms, culture, and religion. As an embodied asset,
human capital can be confounded with other embodied characteristics. Settlers may
have brought not only productive skills, but also social capital, work ethics, and
trust. These factors have a long tradition in the literature on economic develop-
ment, going back as far as Weber (1930) and including more recent work such as
Landes (1999) and Fernández (2011). Finally, settlements may have represented an
initial agglomeration shock, which could then have had long-run impacts through
scale effects as suggested by the regional economics literature (see, for example,
Ciccone and Hall 1996, Baldwin and Martin 2004, and Combes et al. 2010). This
process could have been further intensified by the development of transportation
VOL. 9 NO. 4 ROCHA ET AL.: HUMAN CAPITAL PERSISTENCE AND DEVELOPMENT 117

i­ nfrastructure, as in Donaldson (forthcoming). We try to assess the relevance of each

of these hypotheses while conducting our empirical exercise.

II. Data

We use three sets of historical data. First, we build a dataset of state-sponsored

settlements in São Paulo based on historical reports. We identify, locate, and char-
acterize all official colonies created between 1870 and 1920. Second, we merge
this information with historical municipality-level data drawn from the population
censuses. Third, we complement these data with geographical variables and infor-
mation on transportation infrastructure. We describe the construction of each dataset
below.

State-Sponsored Settlements.—There are scant statistics and literature on the

state-sponsored settlements in São Paulo. In order to gather data and qualitative
information, we combine the few comprehensive analyses found in Martins (1973),
Petrone (1982), Gadelha (1982), and Iotti (2001) with case studies and historical
documentation from various sources. In particular, we make use of data from the
Annual Reports of the Secretariat of Agriculture of São Paulo, published between
1882 and 1920, and from the Annual Statistical Reports of São Paulo, published in
1888, between 1898 and 1920, and also in 1940. Additional qualitative information
on the procedures for the selection of settlers and settled regions was found in the
Public Archives of the State of São Paulo, and in the Provincial Presidential Reports
and Messages to the Legislative Assembly of São Paulo.13
These sources of information combined provide us with a list containing the loca-
tion and date of foundation of each settlement, displayed in Table 1. We make use
of this list to build our variable of interest, an indicator for settlements created in a
given municipality and year. In addition, the Annual Statistical Reports of São Paulo
contain data on demographics and socioeconomic conditions for some settlements
for selected years. Although the information for earlier settlements is scarce, life
conditions within those created in the later waves are relatively well documented.
This allows us to examine the economic and demographic patterns within settle-
ments that had not yet been emancipated by the 1910s.

Census Data.—The official settlements were created between 1877 and 1911.
Given this timing, our analysis is based on municipality-level data drawn from the
population censuses of 1872, 1920, 1940, and 2000. The 1872 census allows us to
examine preexisting socioeconomic characteristics right before the implementation
of the first settlements. The 1920 census is used to build outcome variables for the
period just after the establishment of the last colonies. The 1940 and 2000 cen-
suses allow us to analyze, respectively, the mid-term and long-term impacts of the
settlements.

13 
It is worth mentioning that many official settlements became independent municipalities during the first half
of the twentieth century. Some of them have their histories published online. Official websites of municipalities that
have origins closely connected to the settlements also provided useful information. 
118 AMERICAN ECONOMIC JOURNAL: APPLIED ECONOMICS OCTOBER 2017

The splitting of municipalities over time generated an increasingly fragmented

administrative division. While there were 88 municipalities in 1872, the number of
municipalities reached 645 in 2000.14 Hence, we build two different samples for our
analysis. The sample used in our main exercises maintains the 1920 border defini-
tion and simply aggregates the municipalities in 1940 (270 municipalities) and 2000
(645 municipalities) to the 220 municipalities that existed in 1920.15 We also match
the original data from the 88 municipalities in 1872 to the 202 municipalities that
existed in 1920 in order to account for presettlement characteristics.16
The second dataset merges the layers of data from different census years using
the 88 municipalities from the 1872 administrative organization. In this case, the
202 municipalities from later censuses are merged to match the 1872 census bound-
aries, making the various rounds of data geographically comparable. Although the
sample is smaller, this dataset allows us to estimate our model using municipal fixed
effects.
We select a comprehensive set of variables from the 1872 census to characterize
preexisting conditions at the municipality level. This includes the shares of slaves,
foreigners, literate inhabitants, and children attending school, labor force by sector,
and population density. From the 1920 census, we use the following variables: share
of literate inhabitants, number of schools and teachers per children aged between
7 and 14, share of foreigners (total and by nationality), share of small farms (up
to 100 hectares), average value of farmland per hectare, coffee production in tons,
population density, shares of the labor force by sector, and wages in construction
and agriculture.
Both the 1940 and the 2000 censuses have educational variables similar to those
listed for 1920, as well as a larger set of demographic characteristics including the
shares of children attending school, literate individuals aged between 15–19, and,
as before, labor force by sector (though the sectors contained in the data vary across
years). Since the census micro data are available for 2000, we are able to calcu-
late average years of schooling for different cohorts born during the twentieth cen-
tury. We can also calculate years of schooling for individuals who were born in the
municipality and those who are immigrants (in this particular case, meaning not
born in the municipality). Finally, we calculate income per capita at the municipality
level for 2000 based on micro data from the census.

Data on Geography and Railroads.—We use two important sets of additional

controls in our regressions. By the late 1870s, many regions in the state of São
Paulo were still geographically isolated and had low population density. This was
particularly true for the Northwest, which was farthest away from the capital. Thus,
geographic characteristics, such as distance to the capital, latitude, and longitude are
likely to have been important determinants of the timing of the economic expansion
across São Paulo (longitude, in this case, is closely related to distance to the coast).

14 
More precisely, there were 89 municipalities in 1872. However, the independent municipality of Santo Amaro
in 1872 is considered part of the capital city, since it was incorporated by its administration in 1935. 
15 
A similar procedure has been used in the United States (see Hornbeck 2012). 
16 
Under this procedure, one municipality in 1872 may be matched to several municipalities in 1920. Thus, in
our empirical analysis, we cluster the standard errors of the regressions at the level of the original 1872 division. 
VOL. 9 NO. 4 ROCHA ET AL.: HUMAN CAPITAL PERSISTENCE AND DEVELOPMENT 119

We include these variables as controls in our analysis, as well as average e­ levation

and indicators for the types of soil present in a given municipality (the latter two
account for potential heterogeneity in land quality and agricultural productivity).
Geographic characteristics are available from Ipeadata (distance to the capital,
latitude, longitude, and elevation) and Embrapa Solos (types of soil). Finally, we
include a control for access to transportation infrastructure. Railroads were closely
associated with the expansion of the agricultural frontier and urbanization in the
state (Dean 1969). As mentioned before, the location of the state-sponsored set-
tlements was often associated with proximity to transportation. We thus include
in our analysis a control for the initial presence of the railroad system (based on
historical information on the date of construction of each train station in the state
of São Paulo).17 In the online Appendix, we provide the definition of each variable
included in our analysis.

Descriptive Statistics.—Table 2 presents summary statistics based on the admin-

istrative division from 1920, with 202 municipalities. Official settlements were
created in 20 of these municipalities. Panel A reports summary statistics for geo-
graphic variables. Since the data are restricted to the state of São Paulo, we observe
limited variation in latitude, longitude, and elevation. Latosol is the predominant
type of soil, present in 55 percent of the municipalities, while argisol is present in
37 percent of the sample. These are the main soil types in the state of São Paulo
and are both suitable for coffee production, the major agricultural crop in the late
1800s.
Panels B to E of Table 2 present characteristics of the sample by census year.
It is worth noting that average literacy rate increased by only 10 percentage points
between 1872 and 1920, while population size doubled, the average share of for-
eigners increased from 1 to 13 percent, and railroad coverage increased from 2 to
nearly 70 percent. Average literacy continued to grow in subsequent years, reaching
43 percent in 1940, though the share of children attending school still lagged at
34 percent. Illiteracy was still present up to 2000, but at that point reflecting mostly
the lower literacy rates of older cohorts. In 2000, the share of children attending
school was 96 percent, while average years of schooling reached 5.5.
Table 3 compares the 1872 characteristics of municipalities that received set-
tlements with those of other municipalities in the state. We do this comparison for
the 88 municipalities that existed in the 1872 census. Overall, demographic and
geographic variables, as well as the structure of employment, were very similar
across the two groups. Differences across municipalities that would receive settle-
ments and other municipalities are typically very small quantitatively and appear as
statistically significant for only 1 of the 16 variables considered (and, even in this
case, at the 10 percent level). In the online Appendix (Table A2), we reproduce the
comparison from Table 3 in a multivariate setting. Table A2 reports the results of
regressions estimating the “determinants of settlement presence,” where the dummy
for ­settlement is regressed on all 1872 characteristics from Table 3. In general,

17 
Data available from www.estacoesferroviarias.com.br. 
120 AMERICAN ECONOMIC JOURNAL: APPLIED ECONOMICS OCTOBER 2017

Table 2—Summary Statistics for Geographic Characteristics and for Socioeconomic

Variables by Census Years

Variables Mean SD Min Max

Panel A. Geography
Settlement 0.10 0.30 0.00 1.00
log distance to capital (km) 5.14 0.69 2.08 6.24
Latitude −22.59 0.99 −25.02 −20.13
Longitude −47.61 1.40 −51.45 −44.39
Altitude (100m) 6.02 1.87 0.01 11.98
Dummy for latosol 0.55 0.46 0.00 1.00
Dummy for argisol 0.37 0.44 0.00 1.00
Dummy for cambisol 0.11 0.30 0.00 1.00
Dummy for spondosol 0.01 0.10 0.00 1.00
Panel B. Vars in 1872
Share literate (aged 6+) 0.20 0.11 0.05 0.45
Share children attending school 0.14 0.10 0.03 0.76
Teachers/children (× 1,000) 5.07 4.38 0.00 25.51
Population (in 1,000) 11.15 7.20 1.57 41.75
Share foreigners 0.01 0.02 0.00 0.08
Share slaves 0.15 0.09 0.04 0.53
Dummy for railway 0.02 0.16 0.00 1.00
Share employment agriculture 0.59 0.10 0.35 0.91
Share employment manufacturing 0.11 0.04 0.02 0.24
Panel C. Vars in 1920
Share literate (aged 6+) 0.30 0.10 0.10 0.70
Schools/children (× 1,000) 0.39 0.38 0.00 1.86
Teachers/children (× 1,000) 9.56 6.94 0.83 46.00
Population (in 1,000) 22.6 43.1 2.9 577.6
Share foreigners 0.13 0.10 0.00 0.36
Share literate foreigners 0.44 0.16 0.00 1.00
Share small farms 0.67 0.19 0.00 0.97
Coffee production 0.03 0.04 0.00 0.24
Value of farmland 191.0 153.6 12.0 895.0
Dummy for railway 0.69 0.46 0.00 1.00
Share employment agriculture 0.78 0.13 0.07 0.96
Share employment manufacturing 0.09 0.07 0.01 0.49
Share employment services and retail 0.12 0.07 0.03 0.55
Panel D. Vars in 1940
Share literate (aged 5+) 0.43 0.13 0.14 0.84
Share children attending school 0.34 0.12 0.08 1.00
Schools/children (× 1,000) 5.63 1.84 1.61 15.42
Population (in 1,000) 37.0 102.2 3.8 1,326.3
Share employment agriculture 0.76 0.18 0.03 0.95
Share employment manufacturing 0.08 0.10 0.01 0.70
Share employment services and retail 0.16 0.11 0.03 0.77
Panel E. Vars in 2000
Share literate (aged 5+) 0.89 0.03 0.79 0.93
Share children attending school 0.96 0.02 0.86 0.99
Schools/children (× 1,000) 6.84 4.20 2.35 29.19
Teachers/children (× 1,000) 82.83 21.23 19.67 152.64
Years of schooling (aged 5+) 5.54 0.66 3.53 7.11
Population (in 1,000) 183.3 818.4 2.9 11,100
Income per capita 307.4 80.1 120.4 597.1
Share employment agriculture 0.20 0.13 0.00 0.60
Share employment manufacturing 0.25 0.09 0.10 0.55
Share employment services and retail 0.53 0.10 0.25 0.78

Notes: In all panels, the sample consists of 202 municipalities based on the 1920 municipality boundaries. Data for
geographic variables are originally from Ipeadata (distance to the capital, latitude, longitude, and elevation) and
Embrapa Solos (types of soil). Indicator for railway for each municipality and year built on data are originally from
www.estacoesferroviarias.com.br. The remainder variables in panels B to E correspond to socioeconomic charac-
teristics for each municipality and year originally available from the respective census.
VOL. 9 NO. 4 ROCHA ET AL.: HUMAN CAPITAL PERSISTENCE AND DEVELOPMENT 121

Table 3—Summary Statistics for Geographic Characteristics and for Socioeconomic Variables in
1872 for Municipalities with and without Settlements after 1872

Variables in 1872 Settlements No settlements

Mean SD Mean SD Diff. SE
log distance to capital (in km) 4.92 0.73 4.87 0.58 0.053 0.159
Latitude −22.81 0.76 −22.94 0.90 0.126 0.226
Longitude −47.14 1.49 −46.84 1.20 −0.299 0.328
Altitude (in 100m) 5.74 1.53 5.96 2.57 −0.217 0.619
Latosol (0/1) 0.53 0.38 0.44 0.44 0.091 0.111
Literacy rate 0.19 0.07 0.17 0.10 0.019 0.024
Share children attending school 0.15 0.15 0.16 0.13 −0.015 0.036
Share foreigners 0.02 0.02 0.01 0.01 0.010 0.004
Share slaves 0.20 0.10 0.17 0.10 0.025 0.026
Population density 9.72 7.49 9.47 6.88 0.249 1.817
Railway 0.11 0.32 0.01 0.12 0.091 0.054
Public administration 1.15 1.05 1.01 1.11 0.137 0.285
Legal professionals 0.95 0.57 0.86 0.94 0.087 0.227
Teachers 0.96 0.72 1.03 0.64 −0.07 0.182
Share emp. agriculture 0.61 0.10 0.59 0.12 0.014 0.029
Share emp. manufacturing 0.11 0.04 0.10 0.05 0.012 0.013
Share emp. services and retail 0.28 0.09 0.31 0.11 −0.027 0.027

Notes: This table contains information on 88 municipalities based on the map of São Paulo compatible with the
1872 census. There are 19 observations that received a settlement after 1872 (column “Settlements”) and 69 obser-
vations that did not receive a settlement after 1872 (column “No Settlements”). All variables are computed accord-
ing to the 1872 census boundaries. Data for geographic variables are originally from Ipeadata (distance to the
capital, latitude, longitude, and elevation) and Embrapa Solos (indicator for presence of latosol). Indicator for rail-
way built is from www.estacoesferroviarias.com.br. The remainder variables were built on data originally available
from the 1872 census. Share of foreigners and slaves computed over total population. Share of literate computed
over population aged 6+. Share of children attending school computed over total number of children aged 6–15.
Population density is total population per hectare. Public administration, legal profession, and teachers refer to total
number of workers in the given occupation relative to total population × 1,000. Share of workers in agriculture,
manufacturing, services, and retail is computed over total number of occupied workers.

p­ redetermined characteristics have limited, if any, predictive power for settlement

location. Once we add all variables simultaneously, only the presence of a railway
is a statistically significant correlate of the future establishment of a state-sponsored
settlement (and only at the 10 percent level).
The evidence indicates that the distribution of state-sponsored settlements across
regions of São Paulo—in terms of demographic and geographic characteristics, and
structure of economic activity—was approximately balanced. This supports the idea
that non-settlement municipalities serve as an appropriate comparison group for
settlement municipalities. Still, since some of these 1872 characteristics could be
correlated with future schooling outcomes, we control for all of them in our bench-
mark specification.

III.  Empirical Strategy

The paper focuses on the effect of state-sponsored settlements on the evolution of

educational outcomes over time and on long-term development. Therefore, our main
empirical strategy compares schooling and other outcomes across municipalities
that received a state-sponsored settlement and municipalities that did not receive a
122 AMERICAN ECONOMIC JOURNAL: APPLIED ECONOMICS OCTOBER 2017

settlement but had similar pre-1920 characteristics. We examine the short, middle,
and long-run effects of the settlements using municipality-level data from the 1920,
1940, and 2000 censuses and the following regression:

​yi​​​  =  α  +  β ​Si​​​  + ​X​  i′ ​ ​  Γ + ​W​  ′ 

(1) ​ i​ ​  Π   + ​ϵ​i​​​,

where ​y​ ​i​​​is the outcome of interest in municipality ​i​measured in 1920 for the short-
run effects, 1940 for the middle-run effects, and 2000 for the long-run effects; ​S​ ​i​​​ is
a variable indicating whether the municipality received a state-sponsored settlement
between 1872 and 1920; and ​​ϵi​ ​​​is an error term. The regression also includes ​​X​i​​​  ,
a vector of geographic controls (distance to the capital city, latitude, longitude—
which is similar to distance to the coast in this context—, altitude, and dummies
for the presence of four different types of soil—argisol, cambisol, spondosol, and
latosol), and ​W ​ ​i​​​a vector of socioeconomic characteristics measured at the baseline
in 1872 (discussed in detail in the presentation of the results). The regression is
estimated with the 1920 administrative division, which contains 202 municipalities.
Because these municipalities were originally part of the less fragmented 1872 divi-
sion and we control for predetermined characteristics based on this administrative
organization, we report standard errors that are clustered at the more aggregated
1872 division. Our main regressions focus on schooling outcomes, but we use iden-
tical specifications in complementary exercises to analyze whether there were other
dimensions of heterogeneity across municipalities that received and did not receive
settlements in 1920, immediately after their initial establishment; long-term effects
on income per capita; whether other factors, such as nationality or religion of immi-
grants, intervened in the relationship between settlements and long-term outcomes;
and the evolution of educational inputs, enrollment rates, and structure of employ-
ment over time.
One potential concern in relation to the specification from equation (1) is that
there might be unobservable characteristics that could have affected the decision to
place settlements in specific locations and that were also correlated with patterns of
skilled migration. If this were true, we would observe individuals with more edu-
cation located in municipalities with settlements, but this correlation would have
been driven by unobserved characteristics that might directly affect long-term edu-
cational outcomes. One way to control for unobservable factors that are fixed over
time is to use panel data and control for municipal fixed effects. We use an alterna-
tive dataset based on the 1872 municipality boundaries and pool data from the dif-
ferent censuses to run specifications that estimate year-specific differences between
municipalities with and without settlements, relative to the base year of 1872. This
specification includes municipality and year fixed effects, and each ​β​is interpreted
as the average difference in the outcome variable between settlement and non-set-
tlement municipalities in a given year relative to 1872 (this approach is similar to
that adopted by Hornbeck 2010 and Hornbeck and Naidu 2014). The specific model
that we estimate is

​y​it​​  =  α  + ​β​t​​ ​S​i​​  + ​γ​t​​ ​X​i​​  + ​π​t​​ ​W​i​​  + ​η​i​​  + ​μ​t​​  + ​ϵ​it​​,​

(2) ​
VOL. 9 NO. 4 ROCHA ET AL.: HUMAN CAPITAL PERSISTENCE AND DEVELOPMENT 123

where ​i​indexes a municipality and t​​indicates the census year. The variables ​η​ i​ ​​​ and​​
μ​t​​​represent municipality and year fixed effects, and ​ϵ​ it​ ​​​is an error term. Here, ​X
​ ​i​​​ is
a vector of geographic controls and ​W ​ ​i​​​a vector of socioeconomic characteristics
measured at the baseline in 1872, both of which are interacted with year effects to
allow for differential trends across settlement and non-settlement municipalities.
The sample follows the 1872 administrative division and contains 88 municipalities.
This exercise is limited to the analysis of literacy rates, the only schooling outcome
that can be compared across the different censuses.

IV. Results

A. The Educational Effects of State-Sponsored Settlements

We start the analysis by examining the short-term effects of state-sponsored set-

tlements on education and other local characteristics, including the presence of
foreigners, population density, land distribution, and agricultural productivity in
coffee. The first three columns of Table 4, panel A, report the results from estimat-
ing equation (1) using the literacy rate in 1920 as the dependent variable. In col-
umn 1, we report the bivariate relationship between literacy rate and the settlement
indicator; in column 2, we control for geographic characteristics. The coefficients
on the settlement variable in columns 1 and 2 are very similar and show that 1920
literacy rates were higher in municipalities that received a state-sponsored settle-
ment. The estimate represents a 10 percentage point difference in the literacy rate
between settlement and ­non-settlement municipalities (or a 33 percent difference,
given the average of 0.30 in 1920). In column 3, we control for the 1872 socioeco-
nomic characteristics. The estimated coefficient on the settlements drops to 8 per-
centage points, but remains highly significant. Thus, while the share of slaves and
the presence of a railway in 1872 were correlated with literacy rates in 1920, our
results are robust to controlling for these initial differences across municipalities.
The relationship we uncover between settlements and literacy rates is not driven by
geographic characteristics or previously existing socioeconomic conditions associ-
ated with these areas.
We next investigate whether the presence of state-sponsored settlements also
affected the composition of immigrants or other local characteristics that might have
had an effect on long-term development and, indirectly, on educational outcomes.
In Table 4, columns 4–5 of panel A, and 1–6 of panel B report a series of results
from estimating equation (1) using a large set of 1920 dependent variables that, in
principle, could have been affected by the settlements. All regressions use the same
specification as column 3 of panel A, controlling for geographic and pre-1872 socio-
economic characteristics.
Columns 4 and 5 of panel A in Table 4 report the initial effect of the settlements on
the presence of foreigners and the percentage of literate foreigners. Municipalities
with settlements had 2.5 percentage points more foreigners when compared to
municipalities without settlements (average of 13 percent in 1920). Conditioning
on the presence of foreigners, column 5 shows that the settlements increased the
percentage of literate foreigners in 5 percentage points (an increase of 11 percent
124 AMERICAN ECONOMIC JOURNAL: APPLIED ECONOMICS OCTOBER 2017

Table 4—The Short-Term and Medium-Term Effects of Settlements on Literacy Rates

Panel A
Literacy rate 1920 Foreigners 1920
No Control for Control for Share Share
controls geography charact. 1872 foreigners literate
(1) (2) (3) (4) (5)

Settlement 0.104 0.102 0.081 0.025 0.051

[0.030] [0.028] [0.027] [0.011] [0.028]
Observations 202 202 202 202 172
Adj. R2 0.095 0.310 0.382 0.598 0.260
Geography No Yes Yes Yes Yes
Charact. 1872 No No Yes Yes Yes

Panel B
Other economic and demographic characteristics in 1920
Population Share Coffee ln land log wages log wages
density small farms product. prices construc. agric.
(1) (2) (3) (4) (5) (6)
Settlement 14.22 0.046 −0.008 0.185 0.040 0.031
[15.06] [0.029] [0.007] [0.144] [0.062] [0.056]
Observations 202 202 202 202 148 125
Adj. R2 0.244 0.180 0.202 0.468 0.373 0.553
Geography Yes Yes Yes Yes Yes Yes
Charact. 1872 Yes Yes Yes Yes Yes Yes

Notes: Robust standard errors are in brackets, clustered at the 1872 census boundaries. All columns report the
results from OLS regressions where the dependent variable is the municipality literacy rate in 1920 (share of liter-
ate individuals) for columns 1–3, panel A, and the variable listed in the top of the column from columns 4–5, panel
A, and 1–6 in panel B. The variable settlement is a dummy that equals one if the municipality received at least one
state-sponsored settlement before 1920. Geographic controls include: distance to the capital, latitude, longitude,
elevation, and indicators for different types of soil—latosol, argisol, cambisol, and spondosol. Characteristics in
1872 include presence of railway, share of foreigners, share of slaves, share of literate population, share of children
attending school, population density, total number of workers in public administration and legal professions relative
to total population, share of workers in agriculture, manufacturing, services, and retail computed over total num-
ber of occupied workers. All regressions estimated for 202 municipalities are based on the 1920 census boundaries.

over a base of 44 percent). Thus, settlements brought slightly more foreigners to the
respective areas, but particularly more educated foreigners.18
In columns 1 to 6 of panel B, we examine other potential effects of the settle-
ments associated with alternative theoretical channels that have been highlighted
in the literature, such as land inequality, population agglomeration, and agricultural
productivity. We find no evidence that state-sponsored settlements affected popula-
tion density, the percentage of small farms (proxy for land inequality), or agricul-
tural productivity in the short run (proxied by coffee productivity). We also find no
significant effect on land prices, which could capture investments in rural infrastruc-
ture, nor on wages of either construction or agricultural workers. Despite the fact

18 
These results should be interpreted with caution. In the 1920 census, the question that allows identification of
foreigners is related to nationality, not locality of birth as in most current censuses. So it is possible that foreign born
individuals who by 1920 had acquired Brazilian citizenship would be identified as Brazilian nationals rather than
foreigners. Without specific assumptions about the identification of these individuals—and its difference across
settlements and non-settlement municipalities—it is difficult to tell the type of bias that would be implied by this
measurement error. 
VOL. 9 NO. 4 ROCHA ET AL.: HUMAN CAPITAL PERSISTENCE AND DEVELOPMENT 125

Table 5—The Medium-Term and Long-Term Effects of Settlements on Schooling and Income

Schooling 1940 Schooling 2000 Income 2000

Dep. variable: Literacy Literacy Literacy Literacy Schooling Income
(age 5+) (age 15–19) (age 5+) (age 15–19) (age 5+) per capita
(1) (2) (3) (4) (5) (6)
Settlement 0.107 0.135 0.015 0.001 0.519 0.153
[0.034] [0.040] [0.005] [0.002] [0.138] [0.054]

Geography Yes Yes Yes Yes Yes Yes

Charact. 1872 Yes Yes Yes Yes Yes Yes
Observations 202 202 202 202 202 202
Adj. R2 0.405 0.406 0.361 0.089 0.365 0.443

Notes: Robust standard errors are in brackets, clustered at the 1872 census boundaries. All columns report the
results from OLS regressions where the dependent variable is listed on the top of the column. In columns 1–4, the
dependent variables refer to literacy rates computed for individuals aged within the range listed respectively in each
column. In column 5, the dependent variable is measured in completed years of schooling. All columns include geo-
graphic controls (distance to the capital, latitude, longitude, elevation, and indicators for different types of soil—
latosol, argisol, cambisol, and spondosol), presence of railway, demographic and economic characteristics in 1872
(share of foreigners, share of slaves, share of literate population, share of children attending school, population
density, total number of workers in public administration and legal professions relative to total population, share of
workers in agriculture, manufacturing, services, and retail computed over total number of occupied workers). All
variables are computed according to the 1920 census boundaries.

that settlements attracted a selected pool of immigrants, this seems to have had no
immediate effect on any form of income or on the value of land. These results are
consistent with the idea that state-sponsored settlements induced a positive selection
of foreigners to localities where the higher level of human capital and skills had no
immediate economic return, at a moment when coffee production was still the dom-
inant and most profitable economic activity (Dean 1969).
As a robustness test of the previous results, Table A3 in the online Appendix
shows the same set of regressions using a continuous variable measuring the inten-
sity of treatment: the share of foreigners in the initial settlement relative to the exist-
ing population in the municipality in 1872. This is a very noisy statistic for the size
of settlements, since initial population is not always measured at the same moment.
In addition, we do not have this information for some of the settlements and there-
fore lose some observations under this specification. Despite these limitations, all
specifications show that an increase in the relative size of the settlement was associ-
ated with an increase in the literacy rate in 1920, while little effect is found in other
dimensions.

B. The Medium-Term and Long-Term Effects of State-Sponsored Settlements

We next document the persistent effects of the establishment of state-sponsored

settlements. Table 5 shows results from estimating equation (1) for literacy rates in
1940 (medium-term effects) and literacy rate, years of schooling, and income per
capita in 2000 (years of schooling are not available in the 1920 and 1940 censuses).
We use two measures of literacy rate, one for the entire population (above age 5)
and another for those aged 15 to 19. The second measure captures the effect on
individuals who were not directly affected by the policies implemented before 1920.
126 AMERICAN ECONOMIC JOURNAL: APPLIED ECONOMICS OCTOBER 2017

Column 1 of Table 5 shows that, in 1940, municipalities that had a settlement

established between 1872 and 1920 had literacy rates 10 percentage points higher
when compared to municipalities that did not have a settlement (this represents a
23 percent difference based on the average literacy rate of 0.43 in 1940). This effect
is 13 percentage points if we focus on individuals aged 15–19 (column 2), sug-
gesting that the effect is not driven solely by the older cohorts who arrived with the
establishment of the settlements or soon thereafter. The fact that individuals aged
15–19 were more educated in 1940 suggests persistence of policies at the munici-
pality level or intergenerational transmission of human capital.
In columns 3 to 5 of Table 5, we show the long-run effects of state-sponsored set-
tlements on education using data from the 2000 census. By the year of 2000, roughly
90 percent of São Paulo’s population was literate, with relatively little variation across
municipalities. Because literacy at that point was close to universal for younger cohorts,
we find a small effect of settlements on overall literacy (column 3) and no effect for
individuals aged 15–19 (column 4). So we also look at the effect of the settlements on
years of schooling in 2000. We find that municipalities that received a state-sponsored
settlement before 1920 had 0.52 more year of schooling when compared to other
municipalities (column 5). Given the average years of schooling of 5.5 in 2000, this
represents a 10 percent increase in long-run educational attainment. Finally, in column
6 of Table 5, we examine whether the establishment of settlements and the persistent
increase in human capital also induced a differential development path. Our estimates
show that income per capita in 2000 was 15 percent higher in municipalities that
received a state-sponsored settlement when compared to other municipalities.
One potential drawback of the estimates presented above is that there may be
unobservable characteristics that affect the decision of location of settlements and
that are also correlated with patterns of skilled migration. In Table 6, we present
results from an alternative specification that controls for unobservable factors that
are fixed over time. We construct a panel of municipalities covering 1872, 1920, and
1940, use municipal fixed effects, and allow coefficients capturing the effect of set-
tlements to vary over time. Different from the sample used in Tables 4 and 5, these
regressions use the 1872 map with the initial 88 municipalities. For 1920 and 1940,
the municipalities are aggregated into the original 88 municipalities. We use as a
dependent variable the literacy rate, which is the only educational outcome that can
be compared across censuses. We restrict the analysis to 1872, 1920, and 1940, since
by 2000, differences across municipalities were not reflected anymore on literacy
but on completed years of schooling (see Table 5).
The coefficients displayed in column 1 suggest that in 1920, there was a 4.4 per-
centage point increase in literacy rates relative to the base year of 1872 in localities
with settlements. This effect increased to 9.7 percentage points in 1940. To assess
whether the results are robust to trends in initial characteristics that could be cor-
related with the choice of location of the settlements, column 2 presents results with
year effects interacted with baseline geographic characteristics, and column 3 shows
the results including year effects interacted with both geographic and socioeconomic
characteristics. While the effect in 1920 remains similar and close to 4 percentage
points, the effect in 1940 decreases a bit, reaching 6.8 percentage points in the more
saturated specification in column 3, but remains statistically significant. Overall,
VOL. 9 NO. 4 ROCHA ET AL.: HUMAN CAPITAL PERSISTENCE AND DEVELOPMENT 127

Table 6—Mid-Term and Long-Term Effects of Settlements on Human Capital:

Panel-Data Specifications

Dependent variable
Literacy rate

(1) (2) (3)

Settlement ×
​ ​ 1920 0.044 0.043 0.041
[0.024] [0.024] [0.024]
Settlement ×
​ ​ 1940 0.097 0.086 0.068
[0.037] [0.031] [0.031]

Municipality fixed effect Yes Yes Yes

Year fixed effect Yes Yes Yes
Geography times year effects No Yes Yes
Demographics times year effects No No Yes
Observations 264 264 264
Municipalities 88 88 88
R2 0.748 0.799 0.810

Notes: Robust standard errors are in brackets. In all columns, the dependent variable is the liter-
acy rate (share of literate individuals). Dependent and independent variables for 1872, 1920, and
1940 are computed according to the 1872 census boundaries. All specifications include munici-
pality fixed effects and year fixed effects. Column 2 adds interaction terms between year of cen-
sus and geographic variables (distance to the capital, latitude, longitude, elevation, and indicators
for different types of soil—latosol, argisol, cambisol, and spondosol). Column 3 adds interaction
terms between year of census and demographic variables in 1872 (share of foreigners, share of
slaves, and share of children attending school).

these results are consistent with the cross-sectional estimates presented in Tables 4
and 5 showing that state-sponsored settlements led to a significant increase in liter-
acy rates in the short and medium run.

V.  Mechanisms of Persistence

A. School Inputs

Tables 4 to 6 show that state-sponsored settlements represented a positive shock

to localities and the effects of the shock lasted until 2000. We next document how
this initial shock persisted through time. Table 7 presents a series of cross-sectional
specifications for 1920, 1940, and 2000 where we regress three different measures
of schooling inputs on the settlement indicator. Because the specific data collected
by the census vary over time, the three measures are not available for all years. In
column 1 of Table 7, we show that settlement municipalities had more schools per
child in 1920 than non-settlement municipalities. The effect on number of schools,
however, was only significant in the short run. Column 2 of Table 7, however, shows
that settlement municipalities had more teachers per child in 1920 and that this
effect persisted until 2000 (coefficient 11.2 in column 2, panel C). Finally, in col-
umn 3 of Table 7, we show that, consistently with the positive shock to the number
of schools and teachers, a larger percentage of children in settlement municipalities
were enrolled in school. Similarly to literacy, we find a stronger effect on the per-
centage of children enrolled in school in 1940, since by 2000, basic education was
128 AMERICAN ECONOMIC JOURNAL: APPLIED ECONOMICS OCTOBER 2017

Table 7—The Effects of Settlements on School Inputs and School Attendance

Dependent variable Schools per Teachers per Share children

school-aged child school-aged child attending school
(1) (2) (3)
Panel A. 1920 census
Settlement 0.222 5.053 NA
[0.108] [1.874]
R2 0.090 0.373

Panel B. 1940 census

Settlement −0.251 NA 0.080
[0.294] [0.025]
R2 0.133 0.327

Panel C. 2000 census

Settlement −0.622 11.235 0.006
[0.721] [4.601] [0.003]
R2 0.54 0.045 0.354

Geographic controls Yes Yes Yes

Characteristics 1872 Yes Yes Yes
Observations 202 202 202

Notes: Robust standard errors are in brackets, clustered at the 1872 census boundaries. In all columns and panels,
the variable settlement is a dummy that equals one if the municipality received at least one state sponsored settle-
ment before 1920. Dependent variables are defined at the top of each column, and were built on data originally
from the 1920 census (panel A), 1940 census (panel B), and the 2000 census (panel C). All dependent variables
computed relative to total population of children aged 7–14 years old. In columns 1–2, dependent variables are
computed as the total number of schools (1) or teachers (2) per school-aged children × 1,000. All specifications
include geographic controls (distance to the capital, latitude, longitude, elevation, and indicators for different types
of soil—latosol, argisol, cambisol, and spondosol) and controls for the presence of railway and other characteristics
in 1872 (share of foreigners, share of slaves, share of literate population, share of children attending school, pop-
ulation density, total number of workers in public administration and legal professions relative to total population,
share of workers in agriculture, manufacturing, services, and retail computed over total number of occupied work-
ers). All variables computed according to the 1920 census boundaries.

close to universal in the state of São Paulo (data on school enrollment were not
available from the 1920 census).

B. Structural Transformation

We next provide evidence that settlements induced a structural transformation in

economic activity and, in the long run, attracted more educated immigrants from
inside Brazil. The long-term growth in income per capita is likely to have been a
consequence of both the increase in the education level of the native population and
the attraction of more educated domestic immigrants. In Table 8, we assess specif-
ically how the increased schooling translated into higher income per capita in the
long run. In particular, we analyze whether the higher level of education allowed
individuals to explore more skill-intensive economic activities. Panel A of Table 8
shows that, already in 1920, municipalities with settlements had lower employment
shares in agriculture and higher shares in manufacturing and services. At this early
point, coffee production was still the leading and more profitable economic activity
in the state of São Paulo and accounted for close to 70 percent of total Brazilian
exports (Abreu and Bevilaqua 1996). So, as Table 4 indicates, this initial shift in
VOL. 9 NO. 4 ROCHA ET AL.: HUMAN CAPITAL PERSISTENCE AND DEVELOPMENT 129

Table 8—The Long-Run Effects of Settlements on Employment and Structural Transformation

Dependent variable Share emp. Share emp. Share emp.

agriculture manufacturing retail and services
(1) (2) (3)
Panel A. 1920 census
Settlement −0.121 0.072 0.048
[0.039] [0.024] [0.017]
Adjusted R2 0.457 0.390 0.503

Panel B. 1940 census

Settlement −0.165 0.092 0.073
[0.046] [0.027] [0.028]
Adjusted R2 0.496 0.472 0.396

Panel C. 2000 census

Settlement −0.053 0.006 0.048
[0.019] [0.016] [0.017]
Adjusted R2 0.391 0.315 0.280

Observations 202 202 202

Geographic controls Yes Yes Yes
Characteristics 1872 Yes Yes Yes

Notes: Robust standard errors are in brackets, clustered at the 1872 census boundaries. Dependent variables are
defined at the top of each column, and were built on data originally from the 1920 census (panel A), the 1940 census
(panel B), and the 2000 census (panel C). All dependent variables refer to the number of workers employed in the
given industry relative to the total number of occupied individuals in the municipality. In all columns and panels, the
variable settlement is a dummy that equals one if the municipality received at least one state-sponsored settlement
before 1920. All specifications include geographic controls (distance to the capital, latitude, longitude, elevation,
and indicators for different types of soil—latosol, argisol, cambisol, and spondosol) and controls for the presence
of railway and other characteristics in 1872 (share of foreigners, share of slaves, share of literate population, share
of children attending school, population density, total number of workers in public administration and legal profes-
sions relative to total population, share of workers in agriculture, manufacturing, services, and retail computed over
total number of occupied workers). All variables are computed according to the 1920 census boundaries.

employment toward manufacturing and services was not reflected on higher urban-
ization, higher wages, or higher value of land. It was simply a direct result of the
different composition of immigrants arriving at these locations. Manufacturing was
then still incipient and marginal within the local economy, mostly supplying basic
goods demanded by the coffee economy, such as bricks, tiles, cement, glass, and
plumbing (Dean 1969).
The process of structural transformation deepened and became an economic
advantage between 1920 and 1940, when industrialization took off in the state and
the capital previously accumulated with coffee production was invested in manufac-
turing (Dean 1969). As panel B of Table 8 shows, by 1940 there was a 9 percentage
point difference in the share of employment in manufacturing—and a 7 percentage
point difference in services—across municipalities that had received a settlement
before 1920 and other municipalities. Finally, while by 2000 the whole state of São
Paulo had gone through a process of structural transformation, municipalities that
had received a settlement before 1920 still had 5 percentage points more employ-
ment in services—and less in agriculture—when compared to other areas in the
state. Interestingly, the higher share of employment in services in 2000 was mainly
concentrated in high skill sectors, such as banking and finance, education, health,
and retail (see online Appendix Table A4).
130 AMERICAN ECONOMIC JOURNAL: APPLIED ECONOMICS OCTOBER 2017

C. Agglomeration

Once the process of structural transformation and income growth took place, one
should expect settlement municipalities to become poles of attraction for immi-
grants from inside Brazil. If this was indeed the case, long-term growth would be
partly attributable as well to agglomeration effects.19
In Table 9, we examine the effect of state-sponsored settlements on years of
schooling by cohort of birth and place of birth using data from the 2000 census. We
use data from the census and split the sample between individuals born in the respec-
tive municipality and those born in other localities. We then estimate the effects of
the state-sponsored settlements across cohorts for both types of individuals. For pur-
poses of comparison, in panel A of Table 9 we show the results for individuals born
in the municipality, which corroborate the findings from Tables 3 and 4. In column 1
of Table 9, where we consider all cohorts simultaneously, the coefficient is substan-
tially larger than that from Table 5, indicating that most of the quantitative effect of
settlements on schooling is indeed driven by individuals who were educated in the
municipalities. Individuals born in settlement municipalities have, on average, 0.88
more year of schooling when compared to individuals born in non-settlement munic-
ipalities. Looking at birth cohorts from 1920–1929 to 1960–1969 (columns 2 to 6 of
Table 9), which were likely to have already completed their schooling by 2000, we
see a positive and statistically significant effect that increases over time. Individuals
born in settlement municipalities between 1960 and 1969 had on average 1 more
year of schooling when compared to individuals born elsewhere. The analogous
number for the cohort born between 1920 and 1929, soon after the establishment
of the settlements, was 0.66 year of schooling. But comparing the point estimates
with the mean of the dependent variable—displayed in the table—one can see that
the proportional effect of settlements on years of schooling declined over time, from
24 percent for the 1920–1929 cohort to 15 percent for the 1960–1969 cohort.
To assess the agglomeration effects induced by the settlements, in panel B of
Table 9, we look at how they affected the schooling level of domestic immigrants
(now meaning individuals who were not born in the municipality where they lived in
2000, rather than foreign born). We estimate the effect of the settlements on years of
schooling across different cohorts for individuals who were not born in the munici-
palities where they live. Two hypotheses underlie this exercise and our interpretation
of its results: first, individuals not born in the settlement municipality where they
live are more likely to have been born in municipalities without settlements; and
second, older cohorts of individuals not born in the municipality where they live are
more likely to have migrated to this municipality longer ago in the past (as com-
pared to younger cohorts of immigrants).
Looking across all cohorts, in column 1 of Table 9, we see that immigrants to
settlement municipalities have, on average, 0.54 more year of schooling when

19 
For instance, Ciccone and Hall (1996), Ciccone (2002), and Combes et al. (2010) show that employment den-
sity positively affects labor productivity at the regional level in the United States, Europe, and France, respectively.
See also Rosenthal and Strange (2004) and Duranton and Puga (2004) for reviews of the empirical evidence and a
discussion of the likely mechanisms. 
VOL. 9 NO. 4 ROCHA ET AL.: HUMAN CAPITAL PERSISTENCE AND DEVELOPMENT 131

Table 9—Persistent Effects of Settlements on Years of Schooling, by Cohorts of Individuals

Dep. variable All cohorts 1920–1929 1930–1939 1940–1949 1950–1959 1960–1969

Years of schooling (1) (2) (3) (4) (5) (6)
Panel A. Individuals born in the municipality
Settlement 0.883 0.655 0.759 0.941 0.927 1.022
[0.229] [0.236] [0.225] [0.277] [0.303] [0.200]
Mean dep. var. 6.63 2.73 3.54 4.64 6.10 7.04
Adj. R2 0.453 0.334 0.415 0.397 0.384 0.451

Panel B. Individuals not born in the municipality

Settlement 0.544 0.332 0.285 0.534 0.541 0.765
[0.147] [0.174] [0.184] [0.191] [0.198] [0.177]
Mean dep. var. 5.79 3.02 3.61 4.56 5.85 6.51
Adj. R2 0.329 0.275 0.272 0.265 0.214 0.202

Observations 202 202 202 202 202 202

Geography Yes Yes Yes Yes Yes Yes
Charact. 1872 Yes Yes Yes Yes Yes Yes

Notes: Robust standard errors are in brackets, clustered at the 1872 census boundaries. In all columns, the variable
settlement is a dummy that equals one if the municipality received at least one state-sponsored settlement before
1920. Dependent variables are average years of schooling for each cohort of individuals born in the municipality,
as defined at the top of each column, and built on micro data originally from the 2000 census. All specifications
include geographic controls (distance to the capital, latitude, longitude, elevation, and indicators for different types
of soil—latosol, argisol, cambisol, and spondosol) and controls for the presence of railway and other characteristics
in 1872 (share of foreigners, share of slaves, share of literate population, share of children attending school, pop-
ulation density, total number of workers in public administration and legal professions relative to total population,
share of workers in agriculture, manufacturing, services, and retail computed over total number of occupied work-
ers). All variables are computed according to the 1920 census boundaries.

c­ ompared to immigrants to other municipalities. Columns 2 to 6 of Table 9 present

the effect of the settlements on the characteristics of immigrants across cohorts.
Older cohorts of immigrants, born between 1920 and 1939, were not significantly
different between settlement and non-settlement municipalities. But starting with
the cohorts born between 1940 and 1949, immigrants to settlement municipali-
ties are significantly more educated than immigrants to other locations. This effect
increases in magnitude for younger cohorts and continues to be strongly signifi-
cant. Though substantially smaller than the effect detected for natives in panel A of
Table 9, the relative difference between natives and immigrants in settlement munic-
ipalities shrinks in magnitude for the later cohorts, as can be seen by comparing the
analogous columns. In addition, contrary to the results from panel A, the relative
size of the coefficient (as compared to the mean of the dependent variable) does
not decline with time. This pattern suggests that the initial shock of state-sponsored
settlements also attracted, over time, more educated individuals from within Brazil.
We interpret this evidence to be consistent with agglomeration effects that increase
income per capita in the long run.

D. Culture and Religion

Up to this point, we have documented that state-sponsored settlements brought

positively selected foreigners to some municipalities in the state of São Paulo. These
foreigners were more literate, but they might as well have had other traits that were
132 AMERICAN ECONOMIC JOURNAL: APPLIED ECONOMICS OCTOBER 2017

important for long-run economic development. Specific skills, culture, or social

norms associated with these groups could also be related to schooling outcomes
and development. Although we have limited ways to measure these characteristics,
we have information on nationalities and religion in the 1920 census. In Table 10,
we reestimate the regressions presented in Table 5 controlling for the nationality
and religion of immigrants. We control for the 1920 percentage of foreigners by
major national groups—German, Spaniard, Italian, Portuguese, and Japanese—and
for the 1920 percentage of Catholic individuals. The coefficients on state-sponsored
settlements remain almost unaffected, despite the fact that the percentage of some
nationalities and Catholics appear as positively associated with literacy rates and
income per capita. The results suggest that the long-term effects of the presence of
state-sponsored settlements were not driven by cultural factors associated with the
nationality or religion of immigrants. It remains true, nevertheless, that other unob-
servable productive attributes correlated with education, such as previously accumu-
lated manufacturing skills, could have been a contributing factor.20

VI.  Concluding Remarks

In the end of the nineteenth and beginning of the twentieth century, public author-
ities established a number of official settlement colonies throughout the state of São
Paulo, Brazil. Simultaneously, European immigrants were also arriving in virtually
every other area of the state. We exploit a key feature of this episode to document
the persistence of human capital through time and its relationship with long-term
development: state-sponsored settlements were not too different from other areas in
the state, but for the fact that they received a different pool of immigrants.
We follow more than 120 years of Brazilian history to characterize the initial
conditions in the state of São Paulo in the second half of the nineteenth century and
the dynamics of the impact of state-sponsored settlements through time. We use data
from the 1872, 1920, 1940, and 2000 Brazilian censuses, as well as a vast array of
auxiliary information from various historical sources. The results show that, in 1872,
before the establishment of any settlement, areas that were to receive settlements in
the future were very similar to other areas in the state. In 1920, immediately after
the initial establishment of the settlements, these areas had populations with literacy
rates 8 percentage points higher than elsewhere, despite having only marginally
higher shares of immigrants. At that moment, areas that had received settlements
were not noticeably different from other areas in terms of density, concentration of
land, agricultural productivity, price of land, or wages. Still, in 1940 and 2000, long
after the settlements had officially ceased to exist, their populations continued to be
more educated. In 2000, areas corresponding to previous ­settlements had income per

20 
It is worth mentioning that we do not find differences in the incidence of entrepreneurship across settlement
and non-settlement municipalities when looking at data from the 2000 census (either by cohort or average). So
results do not seem to be driven by unobservable traits—cultural or otherwise—associated with entrepreneurship.
These results are presented in Table A5 in the online Appendix. Also, if we include an additional control for the
political influence of agricultural groups in 1905, results remain unchanged in Table 10, indicating that the initial
distribution of political power across agriculture and manufacturing does not seem to be a relevant mechanism
behind our results. 
VOL. 9 NO. 4 ROCHA ET AL.: HUMAN CAPITAL PERSISTENCE AND DEVELOPMENT 133

Table 10—The Effects of Settlements on Literacy Rates, Conditional on National Identities and
Religion

Dependent variable Literacy rate Literacy rate Years of Income per

in 1920 in 1940 schooling 2000 capita 2000
(1) (2) (3) (4)
Settlement 0.091 0.121 0.521 0.182
[0.026] [0.032] [0.149] [0.054]

Share of Germans 0.206 0.182 0.061 −0.320

[0.080] [0.100] [0.528] [0.240]
Share of Spaniards 0.066 0.113 0.602 0.208
[0.066] [0.066] [0.356] [0.149]
Share of Italians 0.129 0.211 0.540 0.394
[0.055] [0.062] [0.332] [0.155]
Share of Portuguese 0.104 0.242 1.081 0.291
[0.059] [0.086] [0.378] [0.134]
Share of Japanese 0.020 0.034 0.462 0.078
[0.070] [0.077] [0.411] [0.184]
Share of Catholics 0.081 0.064 0.128 −0.013
[0.024] [0.039] [0.307] [0.091]

Geography Yes Yes Yes Yes

Charact. 1872 Yes Yes Yes Yes
Observations 202 202 202 202
Adj. R2 0.478 0.435 0.368 0.47

Notes: Robust standard errors are in brackets, clustered at the 1872 census boundaries. In columns 1 and 2, the
dependent variable is literacy rate in 1920 (1940) built on data originally from the 1920 (1940) census. In col-
umns 3 and 4, the dependent variable is, respectively, average years of schooling and income per capita, both built
on data from the 2000 census. In all columns, the variable settlement is a dummy that equals one if the municipal-
ity received at least one state-sponsored settlement before 1920. Columns 1–4 add the shares of individuals from
different nationalities computed as the total number of individuals from the given nationality relative to total pop-
ulation in the municipality in 1920, as well as the share of Catholics computed as the number of Catholics relative
to total population in 1920. Data on nationalities and number of Catholics are originally from the 1920 census. All
specifications include geographic controls (distance to the capital, latitude, longitude, elevation, and indicators for
different types of soil—latosol, argisol, cambisol, and spondosol) and controls for the presence of railway and other
characteristics in 1872 (share of foreigners, share of slaves, share of literate population, share of children attending
school, population density, total number of workers in public administration and legal professions relative to total
population, share of workers in agriculture, manufacturing, services, and retail computed over total number of occu-
pied workers). All variables are computed according to the 1920 census boundaries.

capita 15 percent higher than the remainder of the state. We also show that, through-
out the twentieth century, enrollment rates and number of teachers per school-
aged children were higher in locations associated with settlements. In these same
locations over this period, employment shifted away from agriculture and toward
manufacturing and services. None of these differences in the use of educational
inputs and in the structure of employment were present before the settlements were
established.
The evidence suggests that local communities that developed from historical set-
tlements demanded higher educational investments and, through time, shifted eco-
nomic activity to skill-intensive sectors. The pattern we uncover is consistent with
the idea that state-sponsored settlements brought more skilled foreigners—along
educational and possibly other unobservable dimensions—to localities where ini-
tially these skills had no immediate economic return. At that point, coffee produc-
tion—intensive in unskilled agricultural workers—was still the dominant and most
134 AMERICAN ECONOMIC JOURNAL: APPLIED ECONOMICS OCTOBER 2017

profitable economic activity. But as soon as industrialization took hold, these skills
started representing an economic advantage that ended up materializing in increased
income per capita in the long run.
Our results document persistence in educational investments and outcomes over a
long period of time within the context of homogenous institutions represented by the
Brazilian state of São Paulo. Even in this setting, we are able to show that early dif-
ferences in human capital—possibly also associated with individual heterogeneity
along other unobservable dimensions—persisted through time and were correlated
with long-run differences in income per capita.

REFERENCES

Abramitzky, Ran, Leah Platt Boustan, and Katherine Eriksson. 2014. “A Nation of Immigrants:
Assimilation and Economic Outcomes in the Age of Mass Migration.” Journal of Political Econ-
omy 122 (3): 467–506.
Abreu, Marcelo de P., and Alfonso S. Bevilaqua. 1996. “Brazil as an export economy, 1880–1930.”
Pontificia Universidade Católica do Rio de Janeiro (PUC-Rio) Working Paper 363.
Acemoglu, Daron, Francisco A. Gallego, and James A. Robinson. 2014. “Institutions, Human Capital,
and Development.” Annual Review of Economics 6: 875–912.
Acemoglu, Daron, Simon Johnson, and James A. Robinson. 2001. “The Colonial Origins of Compar-
ative Development: An Empirical Investigation.” American Economic Review 91 (5): 1369–1401.
Acemoglu, Daron, Simon Johnson, and James A. Robinson. 2005. “Institutions as a Fundamen-
tal Cause of Long-Run Growth.” In Handbook of Economic Growth, Vol. 1A, edited by Philippe
Aghion and Steven N. Durlauf, 385–472. Amsterdam: North-Holland.
Baldwin, Richard E., and Philippe Martin. 2004. “Agglomeration and Regional Growth.” In Handbook
of Regional and Urban Economics, Vol. 4, edited by J. Vernon Henderson and Jacques-François
Thisse, 2671–2711. Amsterdam: North-Holland.
Bandiera, Oriana, Myra Mohnen, Imran Rasul, and Martina Viarengo. 2015. “Nation-Building
Through Compulsory Schooling During the Age of Mass Migration.” http://sticerd.lse.ac.uk/dps/
eopp/eopp57.pdf.
Barro, Robert J. 1991. “Economic Growth in a Cross Section of Countries.” Quarterly Journal of Eco-
nomics 106 (2): 407–43.
Bassanezi, Maria Silva C. Beozzo, Ana Silvia Volpi Scott, Carlos de Almeida Prado Bacellar, and
Oswaldo Mário Serra Truzzi. 2008. Atlas da Imigração Internacional em São Paulo, 1850–1950.
São Paulo: Editora UNESP.
Becker, Sascha O., Erik Hornung, and Ludger Woessmann. 2011. “Education and Catch-Up in the
Industrial Revolution.” American Economic Journal: Macroeconomics 3 (3): 92–126.
Ciccone, Antonio. 2002. “Agglomeration effects in Europe.” European Economic Review 46 (2): 213–27.
Ciccone, Antonio, and Robert E. Hall. 1996. “Productivity and the Density of Economic Activity.”
American Economic Review 86 (1): 54–70.
Ciccone, Antonio, and Elias Papaioannou. 2009. “Human Capital, the Structure of Production, and
Growth.” Review of Economics and Statistics 91 (1): 66–82.
Combes, Pierre-Philippe, Gilles Duranton, Laurent Gobillon, and Sébastien Roux. 2010. “Estimating
Agglomeration Economies with History, Geology, and Worker Effects.” In Agglomeration Econom-
ics, edited by Edward L. Glaeser, 15–66. Chicago: University of Chicago Press.
Dean, Warren. 1969. The Industrialization of São Paulo, 1880–1945. Austin: University of Texas Press.
de Carvalho Filho, Irineu, and Renato P. Colistete. 2010. “Education Performance: Was It All Deter-
mined 100 Years Ago? Evidence From São Paulo, Brazil.” Munich Personal RePEc Archive
(MPRA) Paper 24494.
de Carvalho Filho, Irineu, and Leonardo Monasterio. 2012. “Immigration and the origins of regional
inequality: Government-sponsored European migration to southern Brazil before World War I.”
Regional Science and Urban Economics 42 (5): 794–807.
Dell, Melissa. 2010. “The Persistent Effects of Peru’s Mining Mita.” Econometrica 78 (6): 1863–1903.
Dittmar, Jeremiah, and Ralf R. Meisenzahl. 2016. “State Capacity and Public Goods: Institutional
Change, Human Capital and Growth in Early Modern Germany.” Centre for Economic Perfor-
mance (CEP) Discussion Paper 1418.
VOL. 9 NO. 4 ROCHA ET AL.: HUMAN CAPITAL PERSISTENCE AND DEVELOPMENT 135

Donaldson, Dave. Forthcoming. “Railroads of the Raj: Estimating the Impact of Transportation Infra-
structure.” American Economic Review.
Droller, Federico. Forthcoming. “Migration, Population Composition and Long Run Economic Devel-
opment: Evidence from Settlements in the Pampas.” Economic Journal.
Duranton, Gilles, and Diego Puga. 2004. “Micro-Foundations of Urban Agglomeration Economies.”
In Handbook of Regional and Urban Economics, Vol. 4, edited by J. Vernon Henderson and
J­ acques-François Thisse, 2063–2117. Amsterdam: North-Holland.
Easterly, William, and Ross Levine. 2016. “The European origins of economic development.” Journal
of Economic Growth 21 (3): 225–57.
Fernández, Raquel. 2011. “Does Culture Matter?” In Handbook of Social Economics, Vol. 1, edited
by Jess Benhabib, Alberto Bisin, and Matthew O. Jackson, 481–510. Amsterdam: North-Holland.
Fourie, Johan, and Dieter von Fintel. 2014. “Settler skills and colonial development: The Huguenot
wine-makers in eighteenth-century Dutch South Africa.” Economic History Review 67 (4): 932–63.
Furtado, Celso. 1959. Formação Econômica do Brasil. São Paulo: Companhia Editora Nacional.
Gadelha, Regina Maria A. F. 1982. “Os Núcleos Coloniais e o Processo de Acumulação Cafeeira
(1850–1920): Contribuicões ao Estudo da Colonização de São Paulo.” PhD diss. Universidade de
São Paulo.
Galor, Oded, Omer Moav, and Dietrich Vollrath. 2009. “Inequality in Landownership, the Emergence
of Human-Capital Promoting Institutions, and the Great Divergence.” Review of Economic Studies
76 (1): 143–79.
Glaeser, Edward L., José A. Scheinkman, and Andrei Shleifer. 1995. “Economic growth in a
­cross-section of cities.” Journal of Monetary Economics 36 (1): 117–43.
Glaeser, Edward L., Rafael La Porta, Florencio Lopez-de-Silanes, and Andrei Shleifer. 2004. “Do
Institutions Cause Growth?” Journal of Economic Growth 9 (3): 271–303.
Guiso, Luigi, Paola Sapienza, and Luigi Zingales. 2016. “Long-Term Persistence.” Journal of the Euro-
pean Economic Association 14 (6): 1401–36.
Hornbeck, Richard. 2010. “Barbed Wire: Property Rights and Agricultural Development.” Quarterly
Journal of Economics 125 (2): 767–810.
Hornbeck, Richard. 2012. “The Enduring Impact of the American Dust Bowl: Short- and Long-Run
Adjustments to Environmental Catastrophe.” American Economic Review 102 (4): 1477–1507.
Hornbeck, Richard, and Suresh Naidu. 2014. “When the Levee Breaks: Black Migration and Eco-
nomic Development in the American South.” American Economic Review 104 (3): 963–90.
Hornung, Erik. 2014. “Immigration and the Diffusion of Technology: The Huguenot Diaspora in Prus-
sia.” American Economic Review 104 (1): 84–122.
Huillery, Elise. 2009. “History Matters: The Long-Term Impact of Colonial Public Investments in
French West Africa.” American Economic Journal: Applied Economics 1 (2): 176–215.
Iotti, Luiza Horn. 2001. Imigração e Colonização: Legislação de 1747–1915. Caxias do Sul: Editora
da Universidade de Caxias do Sul.
Iranzo, Susana, and Giovanni Peri. 2009. “Schooling Externalities, Technology, and Productivity:
Theory and Evidence from U.S. States.” Review of Economics and Statistics 91 (2): 420–31.
Kreutz, Lúcio. 2000. “Escolas comunitárias de imigrantes no Brasil: Instâncias de coordenação e estru-
turas de apoio.” Revista Brasileira de Educação 15: 159–76.
Landes, David S. 1999. The Wealth and Poverty of Nations: Why Some Are So Rich and Some So Poor.
New York: W. W. Norton and Company.
Lowes, Sara, Nathan Nunn, James A. Robinson, and Jonathan Weigel. 2015. “The Evolution of Cul-
ture and Institutions: Evidence from the Kuba Kingdom.” National Bureau of Economic Research
(NBER) Working Paper 21798.
Lucas, Robert E., Jr. 1988. “On the mechanics of economic development.” Journal of Monetary Eco-
nomics 22 (1): 3–42.
Martins, José de Souza. 1973. A Imigração e a Crise do Brasil Agrário. São Paulo: Pioneira.
Maschio, Elaine Cátia Falcade. 2005. “A Experiência Escolar entre Imigrantes Italianos no Paraná.”
http://www.histedbr.fe.unicamp.br/acer_histedbr/jornada/jornada6/trabalhos/602/602.pdf.
Moretti, Enrico. 2004. “Human Capital Externalities in Cities.” In Handbook of Regional and Urban
Economics, Vol. 4, edited by J. Vernon Henderson and Jacques-François Thisse, 2243–91. Amster-
dam: North-Holland.
North, Douglass C. 2009. Institutions, Institutional Change and Economic Performance. Political
Economy of Institutions and Decisions. Cambridge: Cambridge University Press.
Nunn, N. 2011. “Christians in Colonial Africa.” http://www.econ.yale.edu/~egcenter/Nunn_Paper.pdf.
Paiva, Odair da Cruz. 1993. “Intervenção Governamental e Reordenação Fundiária.” Master’s diss.
Universidade Estadual de Campinas.
136 AMERICAN ECONOMIC JOURNAL: APPLIED ECONOMICS OCTOBER 2017

Petrone, Maria Thereza Schorer. 1982. O Imigrante e a Pequena Propriedade, 1824–1930. São Paulo:
Brasiliense.
Putterman, Louis, and David N. Weil. 2010. “Post-1500 Population Flows and the Long-Run Deter-
minants of Economic Growth and Inequality.” Quarterly Journal of Economics 125 (4): 1627–82.
Rocha, Rudi, Claudio Ferraz, and Rodrigo R. Soares. 2017. “Human Capital Persistence and Devel-
opment: Dataset.” American Economic Journal: Applied Economics. https://doi.org/10.1257/
app.20150532.
Rosenthal, Stuart S., and William C. Strange. 2004. “Evidence on the Nature and Sources of Agglom-
eration Economies.” In Handbook of Regional and Urban Economics, Vol. 4, edited by J. Vernon
Henderson and Jacques-François Thisse, 2119–71. Amsterdam: North-Holland.
Shapiro, Jesse M. 2006. “Smart Cities: Quality of Life, Productivity, and the Growth Effects of Human
Capital.” Review of Economics and Statistics 88 (2): 324–35.
Siriani, Sílvia Cristina Lambert. 2005. “Os descaminhos da Imigração alemã para São Paulo no seculo
XIX—aspectos políticos.” Almanack Braziliense 1 (2): 91–100.
Squicciarini, Mara P., and Nico Voigtländer. 2015. “Human Capital and Industrialization: Evidence
from the Age of Enlightenment.” Quarterly Journal of Economics 130 (4): 1825–83.
Summerhill, William R., III. 2003. Order against Progress: Government, Foreign Investment, and
Railroads in Brazil, 1854–1913. Stanford: Stanford University Press.
Summerhill, William R., III. 2010. “Colonial Institutions, Slavery, Inequality, and Development: Evi-
dence from São Paulo, Brazil.” Munich Personal RePEc Archive (MPRA) Working Paper 22162.
Valencia Caicedo, Felipe. 2014. “The Mission: Economic Persistence, Human Capital Transmission
and Culture in South America.” http://sites.bu.edu/neudc/files/2014/10/paper_320.pdf.
Wantchekon, Leonard, Marko Klašnjaz, and Natalija Novta. 2015. “Education and Human Capital
Externalities: Evidence from Colonial Benin.” Quarterly Journal of Economics 130 (2): 703–57.
Weber, Max. 1930. The Protestant Ethic and the Spirit of Capitalism. London: Allen and Unwin.
This article has been cited by:

1. Matthias Blum, Karl‐Peter Krauss, Dmytro Myeshkov. 2022. Human capital transfer of German‐
speaking migrants in eastern Europe, 1780s–1820s. The Economic History Review 75:3, 703-738.
[Crossref]
2. Maria Micaela Sviatschi. 2022. Spreading Gangs: Exporting US Criminal Capital to El Salvador.
American Economic Review 112:6, 1985-2024. [Abstract] [View PDF article] [PDF with links]
3. Gulbaram Kulzhanova, Zhuldizay Kulzhanova, Larisa Efimovа. 2022. Human Capital and
Sustainable Development in a Religious Context. Journal of Ethnic and Cultural Studies 8:3,
169-182. [Crossref]
4. Christoph Eder. 2022. Missing Men: Second World War Casualties and Structural Change.
Economica 89:354, 437-460. [Crossref]
5. Bruno Gabriel Witzel de Souza. 2022. Precedents of mass migration: Policies, occupations, and
the sorting of foreigners in São Paulo, Brazil (1872). Economic History of Developing Regions 37:1,
1-26. [Crossref]
6. Thomas Kang, Anders Nilsson. The Role of Education in Modernization Drives in Brazil and
in Sweden 173-209. [Crossref]
7. Miguel Angel García-Arias, Alfredo Tolón-Becerra, Xavier Lastra-Bravo, Úrsula Torres-Parejo.
2021. The out-migration of young people from a region of the “Empty Spain”: Between a
constant slump cycle and a pending innovation spiral. Journal of Rural Studies 87, 314-326.
[Crossref]
8. Zhe Wang, Huan Yang, Xinning Zhang. 2021. History matters: the effects of Chinese ports from
170 years ago on entrepreneurship today. Regional Studies 55:4, 630-644. [Crossref]
9. Lyudmila Makarova, Elena Tkach, Irina Kudryavtseva, Anna Seliverstova, Natalya Vinogradova.
2021. Life quality and human capital within reproduction system: regional aspect. SHS Web of
Conferences 110, 02005. [Crossref]
10. Philipp Ehrl, Leonardo Monasterio. 2021. Spatial skill concentration agglomeration economies.
Journal of Regional Science 61:1, 140-161. [Crossref]
11. François Seyler, Arthur Silve. 2021. Frontier planters, immigrants, and the abolition of slavery in
Brazil. SSRN Electronic Journal . [Crossref]
12. Yi Chen, Ziying Fan, Xiaomin Gu, Li-An Zhou. 2020. Arrival of Young Talent: The Send-
Down Movement and Rural Education in China. American Economic Review 110:11, 3393-3430.
[Abstract] [View PDF article] [PDF with links]
13. Felipe González. 2020. Immigration and human capital: consequences of a nineteenth century
settlement policy. Cliometrica 14:3, 443-477. [Crossref]
14. Bruno Barsanetti. 2020. Road Endpoints and City Sizes. SSRN Electronic Journal . [Crossref]
15. Philipp Ehrl, Leonardo Monasterio. 2019. Skill concentration and persistence in Brazil. Regional
Studies 53:11, 1544-1554. [Crossref]
16. Johannes C. Buggle, Steven Nafziger. 2019. The Slow Road from Serfdom: Labor Coercion and
Long-Run Development in the Former Russian Empire. The Review of Economics and Statistics
67, 1-17. [Crossref]
17. Maria Micaela Sviatschi. 2019. US Criminal Deportations and Human Capital in Central
America. AEA Papers and Proceedings 109, 239-242. [Abstract] [View PDF article] [PDF with
links]
18. Felipe Valencia Caicedo. Missionaries in Latin America and Asia: A First Global Mass Education
Wave 61-97. [Crossref]
19. Bruno Gabriel Witzel de Souza. Organizations Matter: German Schools and Educational
Performance Amid Brazilian Coffee Plantations (1840–1940) 219-250. [Crossref]
20. Amanda Guimbeau, Nidhiya Menon, Aldo Musacchio. 2019. The Brazilian Bombshell? The
Long-Term Impact of the 1918 Influenza Pandemic the South American Way. SSRN Electronic
Journal . [Crossref]
21. Yi Chen, Ziying Fan, Xiaomin Gu, Li-An Zhou. 2018. Arrival of Young Talents: Send-Down
Movement and Rural Education in China. SSRN Electronic Journal . [Crossref]
22. Patrick Testa. 2018. The Economic Legacy of Expulsion: Lessons from Postwar Czechoslovakia.
SSRN Electronic Journal . [Crossref]
23. Sandra Sequeira, Nathan Nunn, Nancy Qian. 2017. Migrants and the Making of America:
The Short- and Long-Run Effects of Immigration During the Age of Mass Migration. SSRN
Electronic Journal . [Crossref]
24. Anna B. Faria. 2017. Coffee, Immigrants and Growth in Brazil. SSRN Electronic Journal .
[Crossref]
25. Federico Droller. 2016. Migration, Population Composition and Long Run Economic
Development: Evidence from Settlements in the Pampas. SSRN Electronic Journal . [Crossref]