94663
GOOD, BAD, AND UGLY COLONIAL ACTIVITIES:
DO THEY MATTER FOR ECONOMIC DEVELOPMENT?
Miriam Bruhn and Francisco A. Gallego∗
Abstract—Levels of development vary widely within countries in the Amer- relied heavily on the exploitation of labor, such as mining and
icas. We argue that part of this variation has its roots in the colonial era, when
colonizers engaged in different economic activities in different regions of a
sugar production. We denote these activities to be “bad” since
country. We present evidence consistent with the view that “bad” activities ES “associate them with low levels of economic development.
(those that depended heavily on labor exploitation) led to lower economic ES’s second category includes colonial activities that did not
development today than “good” activities (those that did not rely on labor
exploitation). Our results also suggest that differences in political repre-
display economies of scale, such as the cultivation of sub-
sentation (but not in income inequality or human capital) could be the sistence crops, cattle raising, and manufacturing, that were
intermediating factor between colonial activities and current development. performed in areas with low precolonial population density.
We call these “good” colonial activities since ES argue that
activities without economies of scale led to positive long-run
I. Introduction
outcomes if they were practiced in small-scale production by
L EVELS of economic development vary widely across
countries. For example, in a sample of seventeen coun-
tries in the Americas, the richest country (the United States)
independent proprietors, which was the case in areas that did
not have a large native population. In areas with a large native
population, activities with economies of scale were typically
has fifteen times the GDP per capita of the poorest coun- performed in large-scale operations with forced labor. We
try (Honduras). Several recent papers have argued that these call these colonial activities, which relied primarily on the
large differences in economic development have their roots native population as an exploitable resource, “ugly” colonial
in history, particularly in the colonial era.1 This paper pro- activities.
vides new empirical evidence for a relationship between The reason that ES associate colonial activities that relied
colonial activities and current-day economic development heavily on the exploitation of labor with low long-run
using within-country data. In particular, we empirically test, outcomes is that areas with these activities developed an
at the subnational level, two related arguments present in the institutional environment that benefited predominantly a rel-
literature: the claim that different types of economic activi- atively small elite of slave and landowners. ES support this
ties that colonizers engaged in led to different growth paths argument with a number of summary statistics suggesting
(Engerman & Sokoloff, 1997, 2002), and the claim that colo- that countries that used to rely heavily on exploitation of
nization led to a reversal of fortunes (Acemoglu, Johnson, & labor during the colonial period extended the franchise later
Robinson, 2002). and had lower literacy rates than other countries in the early
Following Engerman and Sokoloff (1997, 2002, hence- 1900s and also had financial systems that primarily served
forth ES), our analysis starts by classifying colonial activities the elite. We add to this evidence by showing empirically that
in three categories, which we call “good,” “bad,” and “ugly” colonial activities are correlated with current-day measures
activities. ES argue that depending on factor endowments, of economic development (log GDP per capita and poverty
such as climate, geography, and precolonial population den- rates) within countries. Based on history books, we collect
sity, colonizers engaged in different types of economic data on economic activities performed during the colonial
activities in different regions that consequently led to dif- period in 345 regions in seventeen countries in the Americas.
ferent growth paths. Their first category, our bad activities, We also collect data on precolonial population density at the
includes activities that displayed economies of scale and subnational level. Each region is then assigned four dummy
variables summarizing whether it had predominantly good,
bad, ugly, or no colonial activities.
Received for publication March 20, 2009. Revision accepted for publica-
tion September 29, 2010. We collect these data at the subnational level since the dis-
∗ tinction between colonial activities is much sharper across
Bruhn: World Bank; Gallego: Pontificia Universidad Católica de Chile.
We thank Lee Alston, Hoyt Bleakley, Rómulo Chumacero, Nicolas regions within a country than it is across countries. East-
Depetris Chauvin, Irineu de Carvalho Filho, Esther Duflo, Stanley Enger-
man, José Hofer, John Londregan, Kris James Mitchener, Aldo Musacchio, erly and Levine (2003) examine at the country level whether
Jim Robinson, Jeffrey Williamson, Chris Woodruff, and seminar par- the presence of certain crops and minerals is correlated with
ticipants at Adolfo Ibáñez University, ILADES-Georgetown, the 2006 economic development. Whenever a country produced a cer-
LACEA-LAMES meetings, MIT, the 2007 NBER Summer Institute, the
2008 AEA meetings, PUC-Chile, the University of Chile, the Central Bank tain crop or mineral in 1998–1999, they code a dummy to
of Chile, Universidad Los Andes (Colombia), the 2009 World Economic be equal to 1 for that crop or mineral. This method does not
History Congress, and the 2010 New Frontiers of Latin American Eco- allow distinguishing good, bad, and ugly activities since most
nomic History Conference for useful comments. Carlos Alvarado, Kiyomi
Cadena, Felipe González, Francisco Muñoz, and Nicolás Rojas provided countries produce a range of crops and minerals. For exam-
excellent research assistance. We are also grateful to the Millenium Nuclei ple, some crops, such as maize, are produced in almost all
Research in Social Sciences, Planning Ministry (MIDEPLAN), Republic countries (68 out of 72 for maize). Within the countries in our
of Chile, for financial support. The usual disclaimer applies.
1 See Acemoglu, Johnson, and Robinson (2005) and Nunn (2009) for sample, however, it is typically the case that the types of crops
reviews of this literature. and minerals produced vary from region to region, allowing
The Review of Economics and Statistics, May 2012, 94(2): 433–461
© 2012 by the President and Fellows of Harvard College and the Massachusetts Institute of Technology
434 THE REVIEW OF ECONOMICS AND STATISTICS
Table 1.—Regional PPP GDP per Capita across the Americas
Country Observations Mean Log S.D. Minimum Maximum Ratio y max /y min
a
Argentina 24 11,706 0.553 4,578 40,450 8.84
Bolivia 9 2,715 0.395 1,245 4,223 3.39
Brazil 27 5,754 0.576 1,793 17,596 9.81
Canada 13 44,267 0.358 26,942 94,901 3.52
Chile 13 8,728 0.423 4,154 19,820 4.77
Colombia 30 5,869 0.489 2,368 22,315 9.43
Ecuador 22 5,058 0.834 1,458 26,574 18.23
El Salvador 12 3,336 0.300 2,191 5,954 2.72
Guatemala 8 3,563 0.439 2,100 8,400 4.00
Honduras 18 2,108 0.140 1,716 2,920 1.70
Mexico 32 8,818 0.461 3,664 23,069 6.30
Panama 9 4,336 0.676 1,805 12,696 7.04
Paraguay 18 4,513 0.293 2,843 7,687 2.70
Peru 24 3,984 0.570 1,287 13,295 10.33
United States 48 32,393 0.179 22,206 53,243 2.40
Uruguay 19 6,723 0.231 3,902 10,528 2.70
Venezuelab 19 5,555 0.231 3,497 9,088 2.60
a
Data for 1993. b Income data.
us to classify regions into different colonial activities. Our use a precolonial health index as a proxy for precolonial lev-
subnational analysis is also made possible by the fact that els of economic development to show that areas with bad
levels of economic development vary almost as widely across and ugly colonial activities did not have statistically differ-
states or regions within country as they do across countries ent levels of development before colonization. In fact, areas
(table 1).2 with bad and ugly colonial activities had higher measures
Our empirical evidence supports ES’s argument. We find of development before colonization than areas with good
that areas with bad colonial activities have 27.6% lower GDP or no colonial activities, although these differences are not
per capita today than areas with no activities and 27.7% lower statistically significant.
GDP per capita today than areas with good activities. Sim- Second, we follow Acemoglu et al. (2002, henceforth AJR)
ilarly, areas with ugly colonial activities have 15.9% lower in using precolonial population density as a measure of pre-
GDP per capita today than areas with no activities and 16% colonial levels of economic development, where areas with
lower GDP per capita today than areas with good activities. higher precolonial population density were more developed.
Our results also show that areas with bad colonial activities At the country level, AJR find a negative correlation between
have a 16.4% higher poverty rate than areas with no colo- precolonial population density and current GDP per capita
nial activities and a 13% higher poverty rate than areas with for countries that were colonized by European powers, sug-
good colonial activities. Ugly colonial activities are not sig- gesting a reversal of fortunes. However, in countries that were
nificantly correlated with poverty rates. Thus, overall, we not colonized, AJR find a positive correlation between pre-
find strong support for the claim that colonial activities with colonial population density and current GDP per capita. AJR
economies of scale led to low long-run levels of development. argue that the reversal of fortunes was due to the fact that
The evidence on the long-run consequences of ugly colonial Europeans settled in large numbers only in areas that had low
activities is weaker since they have a smaller negative corre- precolonial population density (in part due to differences in
lation with GDP per capita than bad colonial activities and the disease environment that were associated with precolo-
they are not correlated with current-day poverty rates. nial population density). Where Europeans settled in large
We then ask whether it is indeed the case that colonial numbers, they tended to create “neo-European,” or inclu-
activities changed the economic fortunes of certain areas. sive, institutions with strong emphasis on private property
An alternative explanation for our findings is that areas that and checks against government power. In areas where Euro-
were to have bad and ugly colonial activities already had peans did not settle in large numbers, they instead established
lower levels of economic development before colonization. “extractive states,” geared toward transferring as much as pos-
We test this alternative explanation in two ways. First, we sible of the resources of the colony to the colonizer. These
extractive states did not provide protection for private prop-
erty and had few checks and balances against government
2 Several other papers examine the effect of historical factors on within- expropriation. AJR claim that colonial institutions persisted
country variation in development, including Banerjee and Iyer (2005), over time, even after independence, and they still influence
Banerjee, Iyer, and Somanathan (2005), and Iyer (2005) for India; Ace-
moglu et al. (2008), Rosas and Mendoza (2004), and Bonet and Meisel current levels of economic development.
(2006) for Colombia; Naritomi, Soares, and Assunção (2007) for Brazil; Our within-country data mimic AJR’s findings of a reversal
Merrouche (2007) for Algeria; Huillery (2007) for French Africa; Ace- of fortunes. Our results show that in areas that had colo-
moglu et al. (2005) and Tabellini (2007) for Europe; Mitchener and McLean
(1999, 2003) and Nunn (2008a) for the United States; and Dell (2010) for nial activities, a 1 standard deviation increase in the log of
Peru. Nunn (2009) provides a review of this literature. precolonial population density is associated with about 13%
GOOD, BAD, AND UGLY COLONIAL ACTIVITIES 435
lower current GDP per capita. In areas that had no colonial AJR (2001) argue that the colonizers pursued two different
activities, a 1 standard deviation increase in the log of precolo- types of colonization strategies in the New World, depend-
nial population density is associated with about 13% higher ing on the local disease environment. The first strategy was
current GDP per capita. This supports our argument that colo- applied in areas with low settler mortality and was geared
nial activities reversed the fortunes of different areas within toward allowing a large number of Europeans to settle. This
countries. strategy involved the creation of neo-European, or inclusive,
Next, we examine the link between colonial activities and institutions with a strong emphasis on private property and
current-day economic outcomes. Both AJR and ES argue that checks against government power (exemplified by Canada
“institutions” link colonial activities to current-day economic and the United States). The second colonization strategy was
outcomes, although their definition of institutions differs. applied in areas with high settler mortality and consisted
AJR focus on the importance of secure property rights for of establishing extractive states, geared toward transferring
economic growth, whereas ES focus on the importance of as much as possible of the resources of the colony to the
economic and political inequality. An alternative theory, put colonizer. These extractive states did not provide protec-
forward by Glaeser et al. (2004), is that European settlers tion for private property and had few checks and balances
brought with them many other things, such as high human against government expropriation (exemplified by Spanish
capital, that could also explain the effect of colonial activi- and Portuguese colonization of Latin America). AJR claim
ties on current levels of development. Given data availability, that colonial institutions persisted over time, even after inde-
we test three possible mechanisms that could link colonial pendence, and they still influence current levels of economic
activities to current development: economic inequality, edu- development. They empirically show that countries with high
cation, and political representation.3 Our results indicate that settler mortality have lower measures of protection against
current income inequality and human capital are not cor- expropriation risk today and also have lower levels of GDP
related with colonial activities and thus cannot explain the per capita today.
link between colonial activities and current levels of devel- In a related paper, AJR (2002) document that among the
opment. In contrast, the results for political representation countries colonized by Europeans, the ones that were the
are mostly consistent with our reduced-form regressions in richest around 1500 are now the poorest. The authors use
terms of the signs, sizes, and significance of the effects, precolonial urbanization rates and precolonial population
suggesting that political institutions may be the intermedi- density as measures of precolonial prosperity and show that
ating factor between colonial activities and current economic both measures are negatively correlated with GDP per capita
development. today. AJR argue that this reversal of fortunes is due to the
The paper is organized as follows. Section II discusses two different colonization strategies already described. In
the theoretical background, and section III provides his- areas that were relatively poor and sparsely populated in pre-
torical examples for the theory. Section IV describes our colonial times, Europeans could settle in large numbers and
data. Section V examines what determined colonial activ- develop inclusive institutions. In areas with high precolonial
ities. Section VI analyzes the relationship between colo- population density, it was more profitable for colonizers to
nial activities and development. Section VII asks whether establish extractive institutions since they could either force
colonization did indeed reverse the fortunes of different the native population to work in mines and plantations or
regions. Finally, section VIII investigates the mediating fac- they could extract economic benefits by taking over existing
tors between colonial activities and development today, and tax and tribute systems. An additional finding of the paper
section IX concludes. confirms the hypothesis that the reversal of fortunes is due to
colonization: for countries that were not colonized, there is
II. Theoretical Background a positive instead of a negative correlation between precolo-
nial population density and current levels of GDP per capita
In recent years, many studies have investigated the ultimate (AJR, 2002).
determinants of economic development. AJR (2001, 2002, ES (1997, 2002) also develop a theory that links patterns
2005) and ES (1997, 2002) claim that levels of economic of colonization to current-day outcomes, focusing on New
development in New World countries go back to patterns of World economies. The authors argue that three local fac-
colonization. In particular, they argue that colonizers set up tor endowments determined the type of activities that the
different types of institutions in different parts of the New colonizers pursued in the New World: climate, soil, and the
World, depending on the local environment. These institu- size of the native population (labor supply). Based on these
tions have persisted over time and affect long-run levels of
economic development.4
replacing colonial powers after independence tended to maintain the same
institutional setting. As documented in Acemoglu and Robinson (2006), in
3 We do not have data on property rights institutions at the region level for some countries, the elites controlling political power were the same even
most of the countries included in this paper. well after independence. There are a number of mechanisms leading to
4 The argument that economic development depends on institutions goes inertia, as discussed in AJR (2005) and modeled in Acemoglu, Ticchi, and
back at least to North and Thomas (1973) and North (1981). There are Vindigni (2007) for the case of the emergence and persistence of inefficient
several reasons that institutions may persist over time. In fact, ruling elites states.
436 THE REVIEW OF ECONOMICS AND STATISTICS
factor endowments, ES classify New World countries into development in more detail.5 We start with the observation
three types of colonial economies. In the first category are that the economic activities that the colonizers performed
colonies that had soil and climate suitable for producing in the New World varied not only across countries but also
sugar and other highly valued crops that were most effi- within countries. For example, the southeastern United States
ciently produced on large slave plantations due to economies were dominated by large-scale cotton and rice plantations
of scale in production. These “sugar colonies” include Brazil during colonial times, while small-scale grain farmers and
and islands in the West Indies. Given the efficiency of large manufacturing workshops were predominant in the North-
plantations and the extensive use of slaves, economic and east. Similarly, most states in the northeast of Brazil grew
political power became highly concentrated in the sugar sugar during colonial times, while the states in the southeast
colonies. ES argue that this concentration of power led to engaged in grain production and cattle raising. We also note
institutions that commonly protected the privileges of the that levels of economic development vary almost as widely
elite and restricted opportunities for the broad mass of the within country in the Americas as they do across countries.
population. Table 1 shows a summary of GDP per capita (PPP) in differ-
The second category of colonial economies corresponds ent regions in seventeen countries in the Americas. For some
to a number of Spanish colonies, such as Mexico and Peru, countries, the standard deviation of GDP per capita within
that were characterized by a large native population. Eco- country is of the same order of magnitude as the standard
nomic activity in these colonies was dominated by large-scale deviation of log GDP per capita across countries, which is
estates that were to some degree based on preconquest social equal to 0.71.6
organizations in which the elites extracted tribute from the We exploit this within-country variation to study the
general population. The Spanish Crown continued this trib- relationship between different colonial activities and eco-
ute system in its colonies through so-called encomiendas that nomic development in the Americas more explicitly than
granted a small number of individuals land and claims on ES did by correlating colonial activities at the region level
labor and tribute from natives. This led to the creation of with current-day measures of economic development. Using
large-scale estates with forced labor even where the main within-country data to examine the relationship between
production activities did not display economies of scale (as colonial activities and levels of economic development has
was the case in grain agriculture). As in the sugar colonies, several advantages over using country-level data. First, it
these economies featured highly concentrated political and allows us to capture colonial activities more accurately since
economic power that translated into exclusive institutions we are not limited to using a single colonial activity as a
preserving the power of the elite. proxy for the activities performed in each country. Second,
The third category of New World colonies is composed of we obtain a much larger sample (345 regions instead of 17
the colonies on the North American mainland (Canada and countries), increasing statistical power. Moreover, the within-
the northern United States). These colonies were endowed country approach lets us examine in more detail which local
with neither an abundant native population nor a climate and conditions determined colonial activities. We can then con-
soil suitable for producing highly valued crops such as sugar. trol for these conditions when correlating colonial activities
This lack of factor endowments, combined with the existence with current-day outcomes.7
of abundant land and low capital requirements, implied that Our theoretical argument follows ES’s categorization of
most adult men operated as independent proprietors, creat- colonial economies quite closely. We classify the activities
ing a relatively egalitarian society in economic and political that colonizers preformed in different regions of the New
terms.
ES argue that the unequal societies in sugar colonies and
colonies dominated by encomiendas created an institutional 5 AJR’s and ES’s arguments are obviously related. It goes beyond the
environment that restricted economic growth in the long run, objective of this paper to examine the differences among them. The com-
ment by Acemoglu (2002) to ES (2002) includes a discussion on similarities
making them less prosperous today than the United States and differences among them.
and Canada. They support this argument with a number of 6 According to country-level data on PPP GDP per capita from the
summary statistics suggesting that countries that used to be International Monetary Fund’s World Economic Outlook Database (April
2010).
sugar colonies or dominated by encomiendas extended the 7 One concern with going from the cross-country level of analysis to the
franchise later than the United States and Canada and also had within-country level is that labor and capital mobility are much greater
lower literacy rates than in the United States and Canada in within country. However, our aim is to explain existing differences in eco-
nomic development across regions that have not been arbitraged away by
the early 1900s. Moreover, ES argue that the banking systems factor mobility. A caveat here is that while in a frictionless economy, labor
that developed in Latin America served primarily a wealthy and capital should move across regions to equalize incomes, this may not
elite, whereas the banking systems in the United States and be the case in an economy with institutional frictions. For example, if
property rights protection is weak in some areas of a country, capital own-
Canada were more inclusive. ES, however, do not provide ers might move their capital to areas where property rights protection is
econometric evidence for their hypotheses. stronger, exacerbating economic differences. Overall, these issues imply
In this paper, we develop AJR’s and ES’s arguments fur- that the magnitude of our estimates may not generalize to the country level.
We address this question in more detail in note 37, which compares the
ther by examining the different colonization strategies in the magnitude of our estimates to that of AJR (2002) and finds them to be
New World and their consequences for long-run economic similar.
GOOD, BAD, AND UGLY COLONIAL ACTIVITIES 437
World into four categories.8 In the first category are activities activities and regions with good colonial activities. Section
that displayed economies of scale and relied heavily on the II argues that bad colonial activities, such as the cultivation
exploitation of labor, such as mining and sugar production. of highly valued crops that were most efficiently produced
We denote these activities to be “bad” activities since ES on large slave plantations, led to extractive institutions and to
associate them with low levels of economic development in lower levels of development today. An example for this mech-
the long run. Our second category includes colonial activities anism can be found along the northeastern coast of Brazil, a
that did not display economies of scale, such as the cultiva- region that grew sugar during colonial times. Today the core
tion of subsistence crops, cattle raising, and manufacturing, of this region roughly corresponds to the states of Alagoas
which were performed in areas with low precolonial popula- and Pernambuco. These states had very unequal colonial soci-
tion density. We call these “good” colonial activities since ES eties, with a small number of plantation owners and a large
argue that activities without economies of scale led to posi- number of slaves. This inequality went along with the devel-
tive long-run outcomes if they were practiced in small-scale opment of political institutions that benefited primarily the
production by independent proprietors, which was the case elite. For example, Burns (1993) states that the typical owner
in areas that did not have a large native population. In areas of a Brazilian sugar plantation was a “patriarchal chief who
with a large native population, activities without economies ruled family, servants, slaves, and even neighbors—unless
of scale were typically performed in large-scale operations they were large estate owners like himself—with absolute
with forced labor. We call these colonial activities that relied authority. The great size of the estate, its isolation from royal
primarily on the native population as an exploitable resource officials, and the relative weakness of local bureaucrats all
“ugly” colonial activities. Fourth, some areas within a number strengthened his power. . . . The patriarch oversaw the land,
of countries were not settled by the colonizers and therefore listened to petitions, dispensed justice, and in general held
had no colonial activities. court” (p. 63).
Based on the discussion of colonial activities in this In contrast to this elite-dominated society stood the regions
section, we test the following hypotheses in this paper: in the interior and south of Brazil, such as the states of Mato
Grosso and Rio Grande do Sul, where cattle ranching was the
• Differences in current levels of development within predominant colonial activity. These states also had relatively
countries can be explained by differences in colonial low precolonial population densities and thus fall under our
activities. classification of regions with good activities. Burns (1993)
• More specifically, areas that had bad colonial activities describes colonial society in these cattle regions as follows:
(those with economies of scale, relying heavily on the “In direct contrast to the coastal owners, the cattlemen could
exploitation of labor) or ugly colonial activities (those and often did provide some variance from the more rigid
without economies of scale practiced in areas with high hierarchical system characteristic of the sugar industry. Some
precolonial population density, relying on forced labor) scholars choose to emphasize certain democratic features at
have lower levels of development today than areas that work or latent in the cattle ranching. It would seem that often
had good colonial activities or areas that had no colonial the ranch owners were closer to their workers . . . and on
activities. occasion even worked right along with them. . . . Slavery was
not a widespread institution in the cattlelands. As a means of
Both AJR and ES argue that institutions link colonial payment to his vaqueiros [cowboys], the rancher customarily
activities to current-day economic outcomes, although their shared the newborn calves with them. . . . In that way, it was
definition of institutions differs. AJR focus on the importance possible for the ambitious vaqueiro to start his own herd”
of secure property rights for economic growth, whereas ES (pp. 74–75). Burns also suggests that social and economic
focus on the importance of economic and political inequality. mobility was more feasible in the cattle industry than in other
An alternative theory, put forward by Glaeser et al. (2004), is economic activities in the colony.
that European settlers brought with them many other things, The long-run economic outcomes illustrate that colonial
such as high human capital, that could also explain the effect sugar cultivation in Alagoas and Pernambuco, a bad activity,
of colonial activities on current levels of development. In is associated with a lower level of development than cattle
section VIII, we explore different possible channels that could ranching in Mato Grosso or Rio Grande do Sul, a good activ-
link colonial activities to current-day levels of economic ity. In 2000, PPP GDP per capita in Alagoas was $2,809
development. and $3,531 in Pernambuco. In Mato Grosso and Rio Grande
do Sul, PPP GDP per capita was two to three times higher,
III. Historical Background at $6,890 and $9,059, respectively. (We use U.S. dollars
throughout the paper.)
This section illustrates the hypotheses put forward in We now discuss the difference between regions with good
section II with specific examples. We first provide an example colonial activities and those with ugly colonial activities,
that illustrates the difference between those with bad colonial using the example of textile production. In the colonial United
8 This classification is based on ES (2002) and is described in more detail States, where precolonial population densities were com-
in section IVB. paratively low, textile production was organized in many
438 THE REVIEW OF ECONOMICS AND STATISTICS
small-scale mills under property ownership (McGaw, 1994), Table 2.—Summary Statistics
corresponding to what we call a good colonial activity. In Outcome Variables Observations Mean S.D. Minimum Maximum
contrast, textile production in many Spanish colonies, where Log PPP GDP per capita 345 8.83 0.97 7.13 11.67
precolonial population densities were relatively high, was Log poverty rate 331 2.93 0.92 0.21 4.40
organized in obrajes de paño.9 Obrajes were large workshops Health Index 53 4.22 0.38 2.95 4.52
Log Gini 268 −0.74 0.16 −1.15 −0.46
that “integrated every part of the cloth production process” Log schools per child 317 −5.31 0.64 −7.29 −3.69
(Gómez-Galvarriato, 2006, p. 377). These workshops have Log literacy rate 270 −0.14 0.13 −0.76 0
been likened to modern sweatshops, and the labor force was Log seats in lower
house per voter 318 −11.42 1.19 −13.59 −8.14
based on coerced labor (for example, slavery, mita10). Obrajes Historical variables
did not exist in Spain itself and were specifically developed Good activities dummy 345 0.27 0.44 0 1
for the colonies “with the techniques and experience of Span- Bad activities dummy 345 0.21 0.41 0 1
ish masters and artisans” (Gómez-Galvarriato, 2006, p. 377). Mining dummy 345 0.12 0.33 0 1
Plantations dummy 345 0.09 0.29 0 1
Textile production in Spain was mainly organized in small Ugly activities dummy 345 0.34 0.48 0 1
shops, similar to the United States. This suggests that tex- No activities dummy 345 0.18 0.38 0 1
tile production did not display economies of scale during Log precolonial
population density 345 0.30 2.27 −6.91 5.97
colonial times. It is likely that obraje-style production of Control variables
textiles arose in the Spanish colonies only due to the avail- Average temperature 345 19.33 6.66 −9.80 29.00
ability of a large, coercible native population, resulting in Total rainfall 345 1.26 0.94 0 8.13
an ugly colonial activity. Gómez-Galvarriato (2006) claims Altitude 345 0.64 0.91 0 4.33
Landlocked dummy 345 0.56 0.50 0 1
that the obraje system had negative consequences for long-
run development. She argues that the strong dependence on
coerced labor destroyed incentives for the accumulation of
human capital among workers and increased income inequal- population from each country’s demographic census to calcu-
ity. Both of these factors contributed to the low levels of late GDP per capita. For El Salvador, Guatemala, Honduras,
industrial development in many areas in Latin America over and Paraguay, data on state-level GDP per capita come from
the nineteenth century. the national Human Development Report for each country.
For Venezuela, GDP per capita, to our knowledge, is not
IV. Data available at the region level. For this reason, we use regional
income per capita from a household survey.
We constructed a data set that covers 345 regions from Data on poverty rates come from household surveys. We
seventeen countries in the Americas. This section discusses define poverty rates according to the national definition of
general features of the data and data sources. A more detailed poverty lines. This may produce poverty rates that are not
description of the data is in the appendix. Appendix A pro- comparable across countries. To deal with this potential prob-
vides the definitions of all variables. The sources for each lem, we use the log poverty rate in our regressions and include
variable are listed in Appendix B. Summary statistics for all country fixed effects, so that the estimated effects can be inter-
variables are in table 2.11 preted as log deviations from country means. Similarly, we
use the log of GDP per capita in the regressions to allow a
A. Measures of Economic Development uniform interpretation of the effects.
Our outcome data are generally for the year 2000 (or for
The main outcome variable of our analysis is the current a year close to that if data for 2000 were not available). For
level of economic development of each modern department, some countries, such as the United States, it is a well-known
province, region, or state in the data set, measured by GDP fact that levels of economic development across regions have
per capita.12 We also use poverty rates as an alternative mea- converged quite significantly over the past few decades.13 It
sure of economic development. The data on GDP per capita would be interesting to replicate the results with earlier data
and poverty rates come from country-specific sources. We to see how they change over time, but data at the state level
obtained data on state-level GDP from local statistical agen- are not available for earlier periods for many of the countries
cies for most countries and supplemented them with data on in our sample.
9 Accordingly to Gómez-Galvarriate, obrajes were widely present in Latin
In addition to measures of current economic development,
America from the mid-sixteenth century, including places such as Puebla we also use a proxy for precolonial levels of development.
and Michoacán in Mexico; Cuzco, Cajamarca, and Huanuco in Peru; Quito This proxy is a precolonial health index that comes from the
in Ecuador; La Paz in Bolivia; and Córdoba in Argentina. Backbone of History Project (Steckel & Rose, 2002). Steckel
10 The mita system was a forced labor system that was widespread in
colonial Spanish America. Using district-level data for Peru, Dell (2010) and Rose estimate a health index that ranges from 0 to 100
provides evidence, that the mita system had a negative impact on current based on archeological data. For this paper, we match the
consumption and educational attainment. location of the archeological sites to regions within countries.
11 Our full data set is available online at http://www.economia.puc.cl/
fgallego.
12 In this paper, we use department, province, state, and region inter- 13 Iyer (2005) also shows that the effect of colonialism in India was
changeably. stronger at the time of independence than in the 1990s.
GOOD, BAD, AND UGLY COLONIAL ACTIVITIES 439
This allows us to obtain information for 53 regions in our (2002) provide a list of these crops and minerals in their
sample.14 note 11, which states that sugar, coffee, rice, and cotton all
had extensive economies of scale. Similarly, mining also had
B. Colonial Activities and Precolonial Population Density economies of scale. This list of activities is based on evidence
in Fogel (1989), Engerman (1983), and Deerr (1949–1950).
We construct three dummy variables to capture colonial We denote activities that displayed economies of scale to
activities, based on the main economic activity that colo- be bad activities, since ES associate them with bad long-run
nizers performed in a region and on precolonial population outcomes, for the reasons outlined in section II. Following
density. We identify the main economic activity that the first ES, precolonial population density does not enter into our
European settlers performed in each region using country- or classification of bad activities.16
region-specific history books.15 Given that we would like to The category “good” activities includes colonial activities
focus on the exogenous components of colonization, we iden- that did not display economies of scale, such as the cultiva-
tify the activities that were first developed in a region and not tion of subsistence crops, cattle raising, and manufacturing,
subsequent activities, which, according to the theoretical and performed in areas with low precolonial population density.17
historical arguments in the previous sections, may be endoge- We call these activities good colonial activities since ES argue
nous to the initial activities. Moreover, we consider only that activities without economies of scale led to positive long-
activities that were performed in permanent settlements— run outcomes if they were practiced in small-scale production
those that were more than just a fort or mission—even if they by independent proprietors, which they claim was the case
were small. The idea here is that an activity could not have in areas that did not have a large native population. In areas
had lasting effects if there were no settlers in a region to pass with a large native population, activities without economies
on the institutions. For example, in some regions in North of scale were typically performed in large-scale operations
America, settlers from one region traded fur and other goods with forced labor, leading to negative long-run outcomes
with natives from a different region without settling in this according to ES. We call these colonial activities that relied
different region during the colonial period. We consider these primarily on the native population as an exploitable resource
regions that were not settled by the colonizers to have had no ugly colonial activities.
colonial activities. Following this definition, 18% of regions Figure 1 summarizes our classification of activities into
in our sample had no colonial activities. good, bad, and ugly. Areas with bad colonial activities had
Our measure of precolonial population density is based activities with economies of scale, independent of the pre-
on several sources. We use precolonial population density colonial population density. Areas with good and areas with
instead of native population density during the coloniza- ugly colonial activities had activities that did not display
tion period because the second variable is probably highly economies of scale, but they differed in the availability of
endogenous to the activities developed by the colonizers, as exploitable local labor (proxied by precolonial population
suggested in several papers (Denevan, 1992). The informa- density). In our main analysis, we divide areas that had activ-
tion on this variable comes mainly from the chapters and ities without economies of scale into areas with good and
references in Denevan (1992), who provides estimates of the ugly colonial activities based on the median of precolonial
total native population for each country. For some countries, population density, with areas with above the median pre-
he also provides estimates of the native population for regions colonial population density being classified as areas with ugly
within a country. Whenever this is not the case, we comple- colonial activities. The choice of the median as the dividing
ment this information with several other sources to arrive at number is ad hoc, and we provide results using alternative
estimates of population density at the region level. (Appendix classifications based on the 75th percentile and the average
C describes in more detail how the variable was constructed.) in table A1.
In some cases, we have to impute the same value of pre- Based on this classification, we construct three dummy
colonial population density for more than one region due to variables, indicating whether a region had good, bad, or
missing information. We account for this in the empirical ugly activities, with the omitted category being no colonial
analysis by clustering at the precolonial population density activities.18 The summary statistics in table 2 show that 27%
level. Our estimated native population density ranges from
0.01 people per square kilometer in the southern regions of 16 The reason for this is that activities with economies of scale typically
Argentina and Chile to 392 in Mexico City. had revenues that were so high that they could cover the cost of importing
Taking our information on colonial activities and precolo- forced labor to a region, making the activity independent of the preexisting
local labor supply. Section V includes a longer discussion of these issues.
nial population density, we classify regions into having good, 17 Gallego (2010) also uses this classification of activities in his empirical
bad, and ugly activities following ES (2002). The category strategy.
18 One challenge in coding these dummy variables is that it is sometimes
“bad” activities covers activities that displayed economies
the case that the colonizers engaged in various activities within the same
of scale and relied heavily on the exploitation of labor. ES region. However, the history books are typically quite clear about which
activity dominated the economy. For example, some mining regions also
14 A more detailed discussion of this index can be found in Steckel (2008) grew wheat for local consumption, but the region’s employment, infrastruc-
and Steckel et al. (2002). ture, and the administration were typically centered around mining since
15 Appendix B lists the books used for each country. this activity was export oriented and more profitable than wheat production.
440 THE REVIEW OF ECONOMICS AND STATISTICS
Figure 1.—Classification of Colonial Activities into Good, Bad, and Ugly
of all areas had good colonial activities, 21% had bad colonial which factors determined colonial activities from a histor-
activities, and 34% had ugly colonial activities. Among the ical perspective and provides empirical evidence on these
21% of areas that had bad colonial activities, 12% engaged in determinants.
mining, and the remaining 9% engaged in plantation agricul- To provide a framework for our discussion, we express
ture. Figure 2 includes a map for each country in our sample, the colonizers’ decision to settle in an area and engage in
displaying which areas had good, bad, ugly, or no colonial certain activities as the outcome of the following cost-benefit
activities. comparison,
NBir = Pi Fi (climater , soilr , elevationr , laborr , luckr , . . .)
C. Intermediating Factors and Control Variables
− Ci (laborr , transportationr , diseaser , threatsr ,
We complement the previous information with four out- luckr . . .)
come variables that could be a link between colonial activities
and current levels of economic development.19 Three of these where NBir stands for the net benefits of the different possi-
variables come from local statistical agencies or household ble colonial activities i (for example, gold mining, growing
surveys: a measure of income inequality (the Gini index), sugar, or cattle ranching) in region r . The net benefits are
schools per child, and literacy rates. The fourth variable is composed of the revenues derived from each activity, given
the number of seats in the lower house per voter (Bruhn, Gal- by the price of the product, Pi , times the quantity produced
lego, & Onorato, 2008). This variable is typically constructed with production function Fi (.), minus the costs from engag-
using information from local electoral agencies. Unfortu- ing in an activity, Ci (.). The price, the production function,
nately, these four variables are not available at the region and the cost function vary with the type of activity i. The
level for some countries, as we state in Appendix A. observable (for the econometrician) inputs into production
Finally, we also include control variables in the regressions are climate, soil, elevation, and labor. Similarly, the observ-
to account for regional differences in climate and geography. able costs of engaging in different colonial activities arise
The climate variables are average temperature and rainfall. from labor, transportation, disease, and threats. Both the pro-
The geography variables are altitude and a dummy variable duction and cost functions also depend on luck and potentially
indicating whether the region is landlocked. on other observable and unobservable factors.
We argue that in each region, the colonizers engaged in the
V. What Determined Colonial Activities? activity that had the highest net benefits according to the net
benefit function NBir . If the net benefits of all activities were
Our paper aims to provide evidence for the argument that negative in a region, the colonizers would not settle in this
the first type of economic activity that colonizers engaged region and would not engage in any activities. Section IVB
in had lasting effects on economic development. An impor- mentions that 18% of regions did not have any colonial activ-
tant question that relates to the validity of this argument ities at all. Although this model is quite simple and stylized,
and our evidence is what determined colonial activities in we feel that it captures the important elements behind the
the first place. If the determinants of colonial activities are colonizer’s decision of which economic activities to perform
correlated with current levels of economic development for in different regions of the New World.20
reasons other than the colonial activities itself, any empiri-
cal relationship between colonial activities and current levels 20 Some may object to the fact that our argument is based on a simple
of economic development may be spurious. In order to comparison of economic benefits, without considering other factors, such
examine these issues in more detail, this section discusses as ideological or religious motives. While we recognize that these factors
played a role in the colonization of the New World, we abstract from them
here and focus solely on economic motives for colonization. ES (2002)
19 We discuss the motivation for choosing these four variables in section argue that “immigrants from Europe were drawn to the New World primarily
VIII. by the prospect of improving their material welfare” (p. 54).
GOOD, BAD, AND UGLY COLONIAL ACTIVITIES 441
Figure 2.—Maps Displaying Areas with Good, Bad, and Ugly Colonial Activities
442 THE REVIEW OF ECONOMICS AND STATISTICS
Figure 2.—(Continued)
We now discuss the different elements of the net benefit appear to have been the most valuable colonial products since
function, NBir stipulated above. The first element is the price there was a large demand for them in Europe. Subsistence
or value of each product. The colonial products that had the crops, such as wheat, were mostly produced for local con-
highest value were gold and silver since they were used to sumption in the colonies. Table 3 shows the price of sugar in
mint currency. After gold and silver, cash crops, such as sugar, different colonies, compared to the price of wheat or flour.
GOOD, BAD, AND UGLY COLONIAL ACTIVITIES 443
Table 3.—Prices during Colonial Times the costs of importing labor from other areas. For example,
Price (silver grams per kilogram) most labor on sugar planations in the New World was carried
Country Year Wheat or Flour Sugar out by slaves who had been brought to the New World from
Africa. Similarly, mining in Bolivia and Mexico depended
Bolivia 1677 1.68 22.29
Chile 1634 0.26 2.94 critically on moving the native labor force from regions that
Colombia 1635 1.56 8.49 did not have mines to the regions that did have mines, such
Massachusetts 1753 2.5 8.16 as Potosí and Guanajuato (Tandeter, 2006).23 The relatively
Peru 1635 1.12 23.37
Average 1.42 13.05 low-return production of subsistence crops tended to rely on
Source: Global Price and Income History Group, UC Davis.
local labor from natives who lived in the region or from the
colonizers themselves.
An important element in the cost function of colonial
activities was the cost of transportation. These costs were
The price of sugar was 13 silver grams per kilogram on aver- particularly relevant for export goods, such as sugar, that
age, nine times as high as the price of wheat or flour (1.4 were shipped to Europe. Based on the historical evidence,
silver grams per kilogram on average). it appears that producing export goods was profitable only
Based on price alone, the most desirable colonial activity close to the sea since transportation over land was very costly
was to mine gold or silver, followed by the production of during the colonial period. However, some random elements
cash crops, such as sugar.21 The production function of gold also influenced colonial transportation costs and thereby the
or silver, however, depended on luck in the sense that these location of colonial activities. For example, in Brazil, sugar
metals could be mined only where they had been discov- in the colonial period was produced in the states along the
ered. For geological reasons, the probability of discovering northeastern seacoast. Today the Brazilian state that produces
gold, silver, or other minerals was greater in high mountains, the most sugar is São Paulo in the southeast of Brazil.24 São
such as the Andes in South America and the Sierra Madre in Paulo did not produce any sugar during the colonial period
Mexico, which is why we include elevation as one of the ele- despite having a seacoast with a port, Santos, that today is
ments in the production function.22 The production function the biggest port in Latin America.25 The reason is that the
also depends on climate and soil. While these factors seem fertile plains of São Paulo are separated from Santos by a
to have played almost no role in the production of minerals, mountain range that had to be crossed by mule during the
they were important for the production of certain crops, such colonial period, making transportation of goods very expen-
as sugar (ES, 2002). sive (today roads and railways connect the port to the rest of
A crucial factor of production for all colonial activities São Paulo).
was labor. The availability of labor and associated labor costs Transportation costs and proximity to the sea were less
varied from region to region, depending on whether natives important for the production of subsistence crops, such as
were present in a region who could be forced to work for wheat, since these were typically consumed locally. More-
the colonizers. If no natives were present in a region, the over, cattle raising for the local market had relatively low
colonizers could import forced labor from other regions or transportation costs since cattle could walk to the market to
countries at higher cost. Historical evidence suggests that the be sold. This made it possible to raise cattle in some relatively
returns to mining and cash crops were large enough to cover remote colonial areas that were far from other settlements,
such as Mato Grosso in central Brazil.
21 Two historical examples provide support for the argument that the colo-
Although transportation costs were high for gold or silver,
nizers’ preferred activity was mining, followed by producing cash crops, and
then other colonial activities. The first example, from Colombia, shows that as far as we know, conditional on gold or silver having been
the colonizers allocated labor first to gold mining and then moved it into discovered, the colonizers would always decide to engage in
plantation agriculture after mines were exhausted. The colonizers mined mining activities in a region since the revenues were so great
gold in the Pacific lowlands of the Chocó region in Colombia during the
early colonial period, relying heavily on imported slave labor. McFarlane that they outweighed even very high costs. Silver mining in
(2002) and Ocampo (1997) document that after many of the gold reserves Potosí is a telling example. Despite being one of the highest
were depleted in Chocó, owners moved their slaves to sugar plantations in
the neighboring Valle del Cauca and Cauca regions. (modern) cities in the world (with an elevation of about 4,000
Second, an example from Burns (1993) illustrates that the Portuguese meters) and being located more than 400 miles away from the
colonial administration had a greater interest in sugar production than in nearest colonial seaport, it became one of the centers of silver
the production of livestock. Burns states that “by 1600 . . . sugar yielded
more profit to Portugal than all its exotic trade with India” (p. 65) and
“as cattle breeding [in Brazil] expanded quickly in the second half of the 23 A canonical example along these lines is the decision by Viceroy Toledo
sixteenth century, the Portuguese government forbade it within ten leagues in the late sixteenth century in which “he resolved that each year more than
of the coast in order to protect the precious sugar lands” (p. 72). 13,000 indigenous workers and their families from a vast area of the Andes
22 While some mines were discovered during the colonial period, ES would travel to Potosí to work in its mines for twelve months” (Tandeter,
(2002) and Tanderet (2006) document that some mines had existed and 2006).
been used by natives prior to colonization. For example, Porco in present- 24 According to data from the Brazilian statistical institute, Instituto
day Bolivia was an Inca mine located close to Potosí, one of the largest Brasileiro de Geografia e Estatística.
silver mines in the New World that was discovered during the colonial 25 According to statistics from the Economic Commission for Latin
period (Tandeter, 2006). America.
444 THE REVIEW OF ECONOMICS AND STATISTICS
production in the New World. Potosí had up to 120,000 inhab- activities and current-day levels of economic development,
itants during colonial times (Cobb, 2006; Tandeter, 2006). we consider the lack of data on soil and threats during colo-
Silver from Potosí and supplies to Potosí were transported nial times as secondary. While soil quality likely influences
through the Andes on pack trains using mules, llamas, and, agricultural output today, GDP in agriculture as a percent-
in some cases, Indians carriers, with the help of native guides. age of total GDP is generally small.28 Moreover, we consider
Cobb (2006) documents that the Potosí-Arica pack train (cov- threats from natives or other colonizers to have no indepen-
ering a distance of about 120 miles) relied on 312 Indians and dent influence on current levels of development since these
5,000 llamas in 1603. threats disappeared over time and all areas in our sample were
Finally, our stylized model includes two additional costs of settled eventually.
settling in an area and engaging in colonial activities. The first Table 4 displays the results of regressing our colonial
of these costs is related to the risk of contracting a potentially activities dummies on the potential determinants of colonial
fatal disease such as malaria. AJR (2001) point out that the activities.29 Column 1 shows the results for our bad activities
higher the expected settler mortality, the lower the probabil- dummy. We find, consistent with our previous discussion,
ity of reaping the returns of colonial activities, discouraging that precolonial population density is not correlated with the
colonizers from settling in an area.26 The second cost arises development of bad activities: if needed, labor was imported
from threats by hostile natives or other colonizers who might from other regions or countries given the high expected prof-
attack local settlers to drive them out of a region, again lower- its. Several of the climate and geography variables predict the
ing the probability of reaping the returns of colonial activities. probability of having bad activities in a region. However, as
For example, the Mexican state of Quintana Roo, where the columns 2 and 3 suggest, the margins through which these
resort city of Cancun is located today, was not settled during variables operate are different for mining and plantations.
colonial times, in part because Spaniards who tried to set- We find that altitude is correlated in a nonlinear way with the
tle there faced frequent raids from pirates and the English in presence of mining. This mainly captures the fact that miner-
Belize. als are more likely to be found in high mountain ranges.30 Our
To summarize, our model identifies a number of factors that proxy for transportation costs (the landlocked dummy) is not
appear to have played a role in determining colonial activities: statistically significant for mining, but it is negatively corre-
availability of local labor, climate, soil, elevation, transporta- lated with the probability of having plantations. This is also
consistent with our previous discussion on the profitability of
tion costs, the disease environment, and threats from natives
these activities.
or other colonizers. We now test empirically whether these
Next, we examine what determined good and ugly colonial
factors are indeed correlated with colonial activities. We mea-
activities (columns 4 and 5). Precolonial population density is
sure the determinants of colonial activities as follows. First,
mechanically correlated with the dummy variables for good
we use precolonial population density as a proxy for the avail-
and ugly activities, since they are defined in part based on
ability of local labor. Second, we capture the local climate precolonial population density. Moreover, some of the geog-
with four variables: average temperature, average temper- raphy and climate variables show a statistically significant
ature squared, rainfall, and rainfall squared. Third, we use correlation with good and ugly activities, but these corre-
altitude and altitude squared to capture elevation.27 Fourth, lations are not robust to controlling for country dummies.
our proxy for transportation costs is a dummy indicating Finally, in column 6, we include regressions for the determi-
whether a region is landlocked. In addition, altitude may nants of not having any colonial activities. The results suggest
influence transportation costs. that areas without colonial activities correspond to relatively
We do not have a direct measure of settler mortality or dis- isolated (landlocked) regions with low precolonial population
ease risk at the region level. However, Hong (2007) shows that density.
malaria incidence at U.S. army forts in the nineteenth cen- Overall, these regression results confirm our conceptual
tury is strongly correlated with local temperature, rainfall, and historical discussion on the determinants of colonial
altitude, and an indicator for being landlocked. These vari- activities in the New World. We control for these determi-
ables are already included in our analysis and thus also act as nants in our empirical exercises below in order to make it less
proxies for disease risk. We were not able to obtain data on likely that our measured relationships between colonial activ-
soil or on threats from natives or other colonizers. However, ities and current levels of development are spurious rather
since our main concern in this section is to identify variables than causal.
that could potentially bias the relationship between colonial
28 Based on region-level data from Brazil, Chile, Colombia, Mexico, Peru,
26 On the other hand, ES (2002) state that historical migration statistics and the United States, the median percentage of GDP in argiculture is only
for British America do not support this argument since the colonies that 6.5.
had the highest mortality rates, the southern mainland and the West Indies, 29 Panel A of table 4 shows regressions without country dummies, and
attracted the great majority of European migrants (according to ES because panel B of table 4 displays regressions including country dummies. The
survivors could earn very high incomes in these colonies). results in these two panels are not qualitatively different, but we include
27 Altitude may also capture climate characteristics. Dell (2010) argues them for transparency.
that altitude is the main determinant of climate and crop choice in Peru. 30 The nonlinear effect of rainfall probably also captures this.
GOOD, BAD, AND UGLY COLONIAL ACTIVITIES 445
Table 4.—What Determines Colonial Activities?
Bad Activities Mining Plantations Good Activities Ugly Activities No Activities
(1) (2) (3) (4) (5) (6)
A: Without Country Dummies
Log precolonial population density 0.009 0.005 0.004 −0.068∗∗∗ 0.112∗∗∗ −0.053∗∗∗
(0.012) (0.008) (0.010) (0.015) (0.021) (0.011)
Average temperature 0.023∗∗ 0.008 0.014∗∗ 0.016 0.024∗∗∗ −0.062∗∗∗
(0.009) (0.005) (0.007) (0.010) (0.009) (0.012)
Average temperature squared −0.001 −0.000 −0.000 −0.000 −0.001 0.002∗∗∗
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
Total rainfall 0.007 −0.063 0.070∗ 0.046 0.005 −0.058
(0.053) (0.040) (0.036) (0.048) (0.052) (0.041)
Total rainfall squared 0.011 0.022∗∗∗ −0.011∗∗ −0.008 −0.008 0.006
(0.007) (0.006) (0.005) (0.007) (0.008) (0.006)
Altitude −0.017 −0.017 −0.001 −0.208∗∗∗ 0.142∗ 0.083∗
(0.079) (0.060) (0.058) (0.070) (0.085) (0.048)
Altitude squared 0.049∗∗ 0.047∗∗ 0.002 0.039∗∗ −0.045∗∗ −0.043∗∗∗
(0.023) (0.019) (0.013) (0.017) (0.022) (0.013)
Landlocked dummy −0.127∗∗∗ −0.028 −0.099∗∗ 0.125∗∗ −0.093∗ 0.095∗∗
(0.048) (0.032) (0.040) (0.057) (0.052) (0.041)
R2 0.124 0.162 0.055 0.226 0.357 0.277
B: With Country Dummies
Log precolonial population density 0.013 −0.018 0.030∗∗ −0.027 0.096∗∗∗ −0.082∗∗∗
(0.017) (0.013) (0.013) (0.022) (0.020) (0.019)
Average temperature 0.028∗∗ 0.009 0.018∗ 0.015 −0.001 −0.041∗∗∗
(0.013) (0.007) (0.010) (0.016) (0.007) (0.016)
Average temperature squared −0.000 −0.000 −0.000 −0.000 −0.000 0.001∗∗
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
Total rainfall −0.052 −0.119∗∗∗ 0.067∗ 0.124∗∗ −0.031 −0.041
(0.058) (0.045) (0.039) (0.051) (0.055) (0.044)
Total rainfall squared 0.018∗∗ 0.028∗∗∗ −0.010∗ −0.015∗∗ −0.006 0.004
(0.007) (0.006) (0.005) (0.007) (0.008) (0.006)
Altitude −0.011 −0.078 0.067 −0.105 0.074 0.041
(0.085) (0.062) (0.065) (0.085) (0.082) (0.059)
Altitude squared 0.054∗∗ 0.065∗∗∗ −0.011 0.007 −0.032 −0.029∗
(0.024) (0.019) (0.016) (0.023) (0.023) (0.016)
Landlocked dummy −0.115∗∗ −0.023 −0.092∗∗ 0.069 −0.080 0.126∗∗
(0.057) (0.045) (0.043) (0.067) (0.054) (0.051)
R2 0.215 0.261 0.196 0.324 0.444 0.318
Observations 345 345 345 345 345 345
Robust standard errors (clustered at precolonial population density level) in brackets. Significance levels: ∗ 10%, ∗∗ 5%, ∗∗∗ 1%.
VI. The Effects of Historical Factors on Development population density for more than one region due to missing
information.
Section II argues that areas with bad and ugly colonial According to the hypotheses stated in section II, the coef-
activities should have lower levels of economic develop- ficient on bad and ugly colonial activities should be negative:
ment today than areas with good or no colonial activities. We we expect areas with bad and ugly colonial activities to have
test this hypothesis by running the following reduced-form lower levels of development than areas with no activities.
regression, Moreover, testing for the equality of the coefficients on the
good and bad activities dummies, as well as on the good and
Yrc = Zrc α + Xrc β + ηc + erc , (1) ugly activities dummies, should also show that the coeffi-
cients on bad and ugly activities are statistically significantly
where c refers to country, r stands for region, Y is a measure lower than the coefficient on good activities.
of development, Z is a vector of historical variables, X is a Table 5 displays the regression results for log GDP per
vector of current-day control variables, η is a country fixed capita (PPP) as the outcome variable. While the coefficients
effect, and e is the error term. on the good, bad, and ugly colonial activities dummies repre-
The set of historical variables, Z, includes our good, sent the differences in GDP per capita today relative to areas
bad, and ugly colonial activities dummies, with no activities with no colonial activities, the bottom panel of table 5 shows
being the omitted category, as well as precolonial popula- the statistical differences between the coefficients on good,
tion density. The vector X consists of climate and geography bad, and ugly colonial activities relative to each other. Col-
controls. We cluster standard errors at the precolonial popu- umn 1 includes only good, bad, and ugly colonial activities
lation density level since, as discussed in section IV, there are dummies, without any control variables. As predicted, areas
some cases where we impute the same value of precolonial with bad activities have statistically significantly lower GDP
446 THE REVIEW OF ECONOMICS AND STATISTICS
Table 5.—Colonial Activities and Current GDP per Capita
Dependent Variable: Log PPP GDP per Capita
(1) (2) (3) (4) (5) (6)
Good activities dummy −0.061 −0.019 0.018 0.001 0.004
(0.101) (0.091) (0.081) (0.076) (0.077)
Bad activities dummy −0.392∗∗∗ −0.286∗∗∗ −0.293∗∗∗ −0.276∗∗∗
(0.099) (0.087) (0.085) (0.082)
Ugly activities dummy −0.263∗∗∗ −0.123 −0.138∗ −0.159∗∗ −0.166∗∗
(0.099) (0.090) (0.084) (0.079) (0.079)
Log precolonial population density −0.078∗∗∗ −0.059∗∗ −0.056∗∗ −0.052∗∗ −0.053∗∗
(0.024) (0.024) (0.022) (0.020) (0.020)
Plantations dummy −0.231∗∗
(0.100)
Mining dummy −0.318∗∗
(0.097)
Average temperature 0.002 −0.002 −0.002
(0.016) (0.015) (0.015)
Average temperature squared −0.000 −0.001 −0.001
(0.000) (0.000) (0.000)
Total rainfall −0.211∗∗ −0.198∗∗ −0.207∗∗
(0.084) (0.082) (0.083)
Total rainfall squared 0.026 0.022 0.024
(0.020) (0.020) (0.020)
Altitude (per km) −0.050 −0.055
(0.107) (0.108)
Altitude squared −0.031 −0.028
(0.035) (0.035)
Landlocked dummy −0.106∗ −0.104
(0.063) (0.064)
Observations 345 345 345 345 345 345
R2 0.806 0.803 0.811 0.819 0.829 0.830
Coefficient Bad − Good −0.332 −0.267 −0.311 −0.277
F -test: Good = Bad p-value (0.002) (0.008) (0.002) (0.005)
Coefficient Ugly − Good −0.203 −0.104 −0.156 −0.160 −0.170
F -test: Good = Ugly p-value (0.061) (0.336) (0.115) (0.084) (0.075)
Coefficient Bad − Ugly −0.129 −0.163 −0.154 −0.117
F -test: Bad = Ugly p-value (0.071) (0.025) (0.032) (0.082)
Coefficient Plantations − Mining −0.087
F -test: Plantations = Mining p-value (0.450)
Robust standard errors (clustered at precolonial population density level) in brackets. Regressions include country fixed effects. Significance levels: ∗ 10%, ∗∗ 5%, ∗∗∗ 1%.
per capita today than areas with no colonial activities and control variables that are statistically significant are rainfall
areas with good colonial activities (by 39.2% and 33.2%, and the landlocked dummy.31 In our preferred specification,
respectively). Similarly, areas with ugly colonial activities in column 5, areas with bad colonial activities have 27.6%
have 26.3% lower GDP per capita today than areas with no lower GDP per capita today than areas with no activities and
colonial activities and 20.3% lower GDP per capita today 27.7% lower GDP per capita today than areas with good activ-
than areas with good colonial activities. ities. Similarly, areas with ugly colonial activities have 15.9%
We now add precolonial population density to the regres- lower GDP per capita today than areas with no activities and
sion. It is important to include this variable since precolonial 16% lower GDP per capita today than areas with good activ-
population density is a predictor of colonial activities. Col- ities. An additional finding that is not directly discussed in
umn 2 displays a regression with precolonial population our theoretical arguments is that areas with good colonial
density as the only regressor, showing that it is negatively activities are not significantly different from areas with no
and significantly correlated with current GDP per capita. Col- activities in terms of current GDP per capita. Moreover, bad
umn 3 includes all historical variables together, and columns colonial activities are associated with lower GDP per capita
4 and 5 add the set of climate and geography controls to the today than ugly colonial activities.32
regression step by step. Column 4 includes climate variables:
average yearly temperature and total rainfall, and both of
these variables squared. Column 5 includes climate controls, 31 The landlocked dummy works as a proxy for transportation costs that
plus altitude, altitude squared, and a dummy for whether the could generate a number of negative effects on trade and development (see
Frankel & Romer, 1999; Irwin & Tervio, 2000; and Spolaore & Wacziarg,
region is landlocked. 2005).
The coefficients on the bad and ugly dummies shrink in 32 In a robustness exercise that we do not report in paper, we add an
magnitude once precolonial population density is included in interaction of our bad activities dummy with a dummy for being in a region
with precolonial population density above the median and find that, as
the regression, but they are relatively robust to the inclusion of expected given our theoretical discussion, this interaction is not statistically
the climate and geography control variables. In fact, the only different from 0.
GOOD, BAD, AND UGLY COLONIAL ACTIVITIES 447
The coefficient on log precolonial population density is Table 6 considers poverty rates as an alternative measure
statistically significant in all regressions. In column 5, the of economic development. The data set for poverty rates
coefficient on log precolonial population density is (−0.052), is slightly smaller than for GDP per capita since data on
implying that going from the 25th percentile in log precolo- poverty rates are not available for eight Colombian regions,
nial population density (−0.91) to the 75th percentile (2.10) two Honduran regions, and one Argentinean region. Simi-
is associated with 15.5% lower GDP per capita. A 1 standard lar to table 5, table 6 first shows the relationship between
deviation increase in precolonial population density is associ- poverty rates and our colonial activities dummies. Here,
ated with an 11.7% decrease in current percapita GDP. There only the coefficient on the bad colonial activities dummies
are several possible explanations for finding a direct nega- is positive and statistically significant. In the next column,
tive effect of precolonial population density on current levels we display the correlation between the log of poverty rate
of development that is independent of colonial activities. and precolonial population density alone. Finally, we include
First, greater precolonial population density could be associ- all historical variables in the same regression and add cli-
ated with stronger precolonial tax systems that the colonizers mate and geography control variables to the regression. All
could take over and exploit, independent of colonial activi- columns unambiguously show that current poverty rates are
ties. Second, in areas with low precolonial population density, positively correlated with precolonial population density. The
Europeans tended to be more likely to settle in large num- coefficients imply that going from the 25th percentile in log
bers and to introduce institutions that encouraged investment precolonial population density (−0.97) to the 75th percentile
(Acemoglu, 2002; AJR, 2002).33 (2.10) is associated with a 14% higher poverty rate. Our pre-
The last column of table 5 displays a regression that splits ferred specification in column 5 shows that areas with bad
our bad colonial activities dummy among plantation and min- colonial activities have a 16.4% higher poverty rate than areas
ing dummies to check whether bundling them together in with no colonial activities and a 13% higher poverty rate than
one category is supported by the data. Both dummies show areas with good colonial activities. The results for ugly colo-
a negative and statistically significant correlation with cur- nial activities are different from those in table 5. Unlike for
rent GDP per capita. Although the plantations dummy has a GDP per capita, we do not find a statistically significant cor-
larger coefficient in absolute terms, the F -test at the bottom relation between ugly colonial activities and current poverty
of table 5 suggests that we cannot reject the null hypothesis rates.35
that the coefficients on the mining and plantations dummies Overall, the evidence in this section supports the theoreti-
are not statistically different. Therefore, in the remainder of cal argument that bad colonial activities led to lower long-run
the paper, we continue grouping them into a single category levels of economic development than good colonial activi-
(bad activities).34 ties or no colonial activities. The corresponding evidence for
ugly colonial activities is mixed, suggesting that ugly colo-
33 A negative coefficient on precolonial population density in this regres-
nial activities may have been less detrimental for long-run
sion could also be due to convergence if the precolonial production function
was Y = L (1−α) N α , where L stands for land and N stands for labor and if
economic development than bad colonial activities. However,
there were a convergence process of the form log(Yt /Nt ) − log(Y0 /N0 ) = ugly colonial activities are associated with significantly lower
−λ log(Y0 /N0 ), where t stands for the current period, 0 for the past, and GDP per capita today.
λ is a function of the speed of convergence. In this setup, we get that
log(Yt /Nt ) = −λ(1 − α) log(N0 /A0 ). However, simulations using this pro-
cess and a time span of 400 years show that, at the speed of convergence
that Barro and Sala-i-Martin (1999) find for U.S. states from 1880 to 1980,
namely, 1.74% per year, the coefficient on precolonial population density
would have to be between 100 and 1,000 times smaller in absolute value than
what we find, depending on the assumption of the share of labor in produc- activities. It is now equal to 13.7 (versus 15.9) and is statistically significant
tion. Thus, a conventional convergence model cannot explain the magnitude only at the 21.9% level (possibly due to the smaller sample size). Overall,
of our coefficient. Moreover, notice that the absolute value of our estimate the results in table A1 imply that our main estimates are not driven by any
of the effect of precolonization population density on current development country in particular and do not seem to be affected by data imputation in
is probably biased downward, given that our population density variable is the precolonial population density variable.
measured with error. Table A2 includes estimates in which we divide our good and ugly activ-
34 In order to examine the robustness of the results, panel A of table A1 ities dummies using three different cutoffs for defining high population
displays seventeen different runs of the regression in column 5 of table 5. density: the median (0.30, the results we report in the main text), the aver-
Each row corresponds to this regression with a different country excluded age (0.50), and the 75th percentile (2.10) of the distribution of population
from the sample. The bottom part of panel A of table A1 includes sum- density. Results show that the coefficients on our variables of interest are
mary statistics for the seventeen coefficients. The estimated coefficients fairly robust to these different definitions.
on the bad and ugly colonial activities dummies, as well as the coeffi- We also experimented with adding other control variables to the regres-
cient on precolonization population density, are fairly robust to excluding sion to examine the robustness of our results. For instance, we added the
singly countries from the sample. The coefficients on the good activities share of the current population that is indigenous. This does not change our
dummy vary more, but the dummy is never statistically significant. Panel results, although the variable is negatively correlated with GDP per capita.
B of table A1 focuses, on the robustness of the precolonial population vari- The interactions between good and bad colonial activities and the share of
able. It excludes from the sample all the countries for which more than the population that is native are negative and statistically different from 0.
50% of regions have data on this variable imputed from other countries However, the magnitude of our main effects does not change much, sug-
(Guatemala, Honduras, Panama, Paraguay, Uruguay, and Venezuela). The gesting that the interaction effects are not economically relevant. Results
results are mostly consistent with the results for the complete sample: the available on request.
only coefficient that drops in magnitude slightly is that on ugly colonial 35 The correlation between GDP per capita and poverty rates is −0.46.
448 THE REVIEW OF ECONOMICS AND STATISTICS
Table 6.—Colonial Activities and Current Poverty Rate
Dependent Variable: Log PPP GDP per Capita
(1) (2) (3) (4) (5)
Good activities dummy 0.091 0.054 0.001 0.035
(0.108) (0.096) (0.080) (0.070)
Bad activities dummy 0.297∗∗∗ 0.202∗∗ 0.194∗∗ 0.164∗
(0.110) (0.098) (0.090) (0.094)
Ugly activities dummy 0.095 −0.029 −0.026 0.001
(0.118) (0.110) (0.100) (0.106)
Log precolonial population density 0.054∗∗ 0.053∗ 0.052∗∗ 0.046∗∗
(0.027) (0.027) (0.025) (0.023)
Average temperature −0.000 0.015
(0.038) (0.031)
Average temperature squared 0.000 0.001
(0.001) (0.001)
Total rainfall 0.267∗∗∗ 0.252∗∗∗
(0.081) (0.078)
Total rainfall squared −0.028∗ −0.023
(0.015) (0.014)
Altitude (per km) 0.028
(0.118)
Altitude squared 0.059∗
(0.033)
Landlocked dummy 0.175∗∗
(0.076)
Observations 331 331 331 331 331
R2 0.766 0.762 0.770 0.790 0.818
Coefficient Bad − Good 0.206 0.148 0.193 0.130
F -test: Good = Bad p-value (0.028) (0.117) (0.031) (0.096)
Coefficient Ugly − Good 0.003 −0.084 −0.027 −0.033
F -test: Good = Ugly p-value (0.975) (0.447) (0.786) (0.703)
Coefficient Bad − Ugly 0.203 0.232 0.220 0.163
F -test: Bad = Ugly p-value (0.017) (0.007) (0.010) (0.033)
Robust standard errors (clustered at precolonial population density level) in brackets. Regressions include country fixed effects. Significance levels: ∗ 10%, ∗∗ 5%, ∗∗∗ 1%.
VII. Did Colonization Reverse Fortunes? Using our precolonial measure of economic development,
the health index, and our current-day measure of economic
Our argument postulates that colonial activities changed development, GDP per capita, we construct a difference-in-
the economic fortunes of certain areas. Thus, before colo- difference exercise in the following way. First, we standardize
nization, areas where bad and ugly colonial activities were to the health index by subtracting its sample average from each
take place should not have been worse off than other areas, observation and dividing it by its standard deviation. We stan-
ceteris paribus. If those areas were worse off even before dardize log current GDP per capita in the same way for each
colonization, then there must be something other than colo- of the 53 regions for which we have data on the precolonial
nization patterns that explains these differences. We would health index. Then we run the following regression:36
thus like to verify that bad and ugly colonial activities were
not negatively correlated with economic development before
colonization. This check is, however, not easily done since Srct = Zrc × α + Zrc × Postt × αpost + Xrc β + Postt × γ
there are no measures of precolonial GDP per capita or + ηc + erct , (2)
other conventional measures of development at the region
level. where S is the standardized measure of development in region
To get a proxy measure of economic development, we use r in country c at time t . Time t takes two values, before or
a health index, which is available for 53 regions in seven of after colonization; therefore, the variable Post is a dummy
the seventeen countries in the full sample: Brazil, Canada, that takes the value of 1 when t = after colonization. The
Chile, Ecuador, Mexico, Peru, and the United States. The variables Z, X, and η are defined as in equation (1). Thus,
health index was calculated based on skeletons found in each α captures the effect of the vector of historical variables on
region. These skeletons often come from different centuries. development before colonization, and αp captures the change
To control for possible differences in the quality of the data in this effect after colonization.
arising from the age of the skeletons, we add the variable
“year” to the health index regression. “Year” is the aver- 36 This idea of using standardized outcomes is based on the estimation
age of all the estimated years in which the found bodies of mean standardized treatment effects in the policy evaluation literature
lived. (Kling et al., 2004).
GOOD, BAD, AND UGLY COLONIAL ACTIVITIES 449
Table 7.—Colonial Activities and Precolonial Development Table 8.—Reversal of Fortunes
Dependent Variable: Dependent Variable:
Normalized Measure of Log PPP GDP per Capita
Economic Development
(1) (2)
Good activities dummy 0.193 ∗∗
(0.333) Log precolonial population density −0.065 0.061∗∗
Good activities dummy × Post −0.790 (0.028) (0.028)
(0.608) Observations 284 61
Bad activities dummy 0.472 R2 0.773 0.932
(0.323) Regions in sample With colonial Without colonial
Bad activities dummy × Post −1.406∗∗ activities activities
(0.614) Robust standard errors (clustered at precolonial population density level) in brackets. Regressions include
Ugly activities dummy 0.307 country fixed effects, as well as climate and geography controls. Significance levels: ∗ 10%, ∗∗ 5%, ∗∗∗ 1%.
(0.358)
Ugly activities dummy × Post −0.832
(0.629)
Log precolonial population density 0.135∗ precolonial population density were more developed. If col-
(0.069)
Log precolonial population density × Post −0.191
onization reversed fortunes, then we should not see a negative
(0.135) correlation between precolonial population density in areas
Observations 106 that did not have any colonial activities. Table 8 shows the
R2 0.718 correlation between precolonial population density and cur-
Coefficient Bad − Good 0.280
F -test: Good = Bad p-value (0.482) rent GDP per capita separately for areas that had colonial
Coefficient Ugly − Good 0.114 activities and areas that did not have colonial activities. The
F -test: Good = Ugly p-value (0.798) results illustrate that while population density has a negative
Coefficient Bad − Ugly 0.166
F -test: Bad = Ugly p-value (0.331) and significant relationship with current GDP per capita in
Robust standard errors (clustered at precolonial population density level) are in parentheses. The outcome areas with colonial activities, the same correlation is positive
variable is composed of a normalized health index for the precolonial period and normalized GDP per
capita for the postcolonial period. The post dummy refers to postcolonization. The regression includes
and statistically significant for areas without colonial activi-
the uninteracted post dummy, country fixed effects, climate and geography controls, as well as a variable ties. For areas with colonial activities, a 1 standard deviation
denoting the year for which the health index is observed, to control for differences in the quality of the
index. Significance levels: ∗ 10%, ∗∗ 5%, ∗∗∗ 1%. increase in the log of precolonial population density is associ-
ated with about a 13% lower current GDP per capita. For areas
without colonial activities, a 1 standard deviation increase in
Table 7 shows the results of this difference-in-difference the log of precolonial population density is associated with
regression. Before colonization, areas with bad colonial activ- about a 13% higher GDP per capita. These results mimic
ities and those with ugly colonial activities did not have the cross-country results in AJR (2002). AJR find a reversal
statistically different levels of economic development from of fortunes only for countries that were colonized by Euro-
areas with good colonial activities and areas with no colonial pean powers, not for countries that were not colonized (AJR,
activities. In fact, the coefficients on the activities dummies 2002).37 Our within-country finding of no reversal of fortunes
suggest that areas with bad and areas with ugly colonial activ- in areas without colonial activities supports the argument that
ities had higher levels of economic development than areas colonization changed the fortunes of colonized areas.
with good colonial activities and areas with no colonial activ-
ities before colonization, although these coefficients are not VIII. History and Development: Looking inside the
statistically significant. The interactions of the colonial activ- Black Box
ities with the postcolonization dummy show that areas with
bad colonial activities and areas with ugly colonial activities This section examines several potential mechanisms that
saw a decline in economic development after colonization may account for the effect on history on current levels of
(for ugly colonial activities, this differences is statistically development.38 We consider three potential mechanisms:
significant at the 19.3% level in this sample). Table 7 also
shows that precolonial population density was associated 37 AJR’s cross-country results and our within-country results provide
with higher levels of economic development before coloniza- an opportunity to assess whether the magnitude of the effect of colonial
tion, but that this correlation was reversed after colonization activities on current levels of development differs when measured within
(this difference is statistically significant at the 16.4% level). countries vs. across countries. As mentioned in note 7, the within-country
results may be different due to larger factor mobility. We used AJR’s data
Although the results in table 7 are not precisely estimated, set to compute that for countries in the Americas, a 1 standard deviation
possibly because the sample is quite small, they suggest that increase in precolonial population density decreases GDP per capita by
colonization reversed the economic fortunes of different areas about 0.22 standard deviations when controlling for similar variables as
we do. The comparable elasticity in our within-country data is 0.18 (after
in the New World. removing country fixed effects from the standard deviations). The fact that
Another way of testing the hypothesis that colonization the two numbers are quite close suggests that within-country findings may
reversed economic fortunes is to follow AJR (2002) in using also apply across countries.
38 Our approach here is similar to the one used by Nunn (2008b, sec. VI),
precolonial population density as a proxy for precolonial where he studies the mechanisms through which the past slave trade could
levels of economic development, where areas with higher be affecting current development.
450 THE REVIEW OF ECONOMICS AND STATISTICS
Table 9.—Possible Channels Linking Colonial Activities to Current Levels of Development
Dependent Variable
Log GDP Log Gini Log Schools Log Literacy Log Seats in Lower
per Capita Index per Child Rate House per Voter
(1) (2) (3) (4) (5)
A: Regression Results
Good activities dummy 0.001 0.001 0.073 −0.006 −0.014
(0.076) (0.014) (0.071) (0.010) (0.085)
Bad activities dummy −0.276∗∗∗ 0.017 −0.011 −0.023 −0.273∗∗
(0.082) (0.018) (0.085) (0.015) (0.113)
Ugly activities dummy −0.159∗∗ −0.005 −0.075 −0.016 −0.321∗∗
(0.079) (0.020) (0.096) (0.016) (0.132)
Log precolonial population density −0.052∗∗ 0.000 −0.014 −0.001 −0.070∗∗
(0.020) (0.007) (0.018) (0.003) (0.028)
Observations 345 268 317 270 318
R2 0.829 0.738 0.713 0.646 0.861
Coefficient Bad − Good −0.277 0.016 −0.084 −0.017 −0.259
F -test: Good = Bad p-value (0.005) (0.354) (0.161) (0.355) (0.013)
Coefficient Ugly − Good −0.160 −0.006 −0.148 −0.010 −0.307
F -test: Good = Ugly p-value (0.084) (0.761) (0.037) (0.571) (0.017)
Coefficient Bad − Ugly −0.117 0.022 0.064 −0.007 0.048
F -test: Bad = Ugly p-value (0.082) (0.221) (0.336) (0.734) (0.503)
B: Standardized Effects
Good activities dummy 0.000 0.002 0.051 −0.022 −0.005
Bad activities dummy −0.117 0.044 −0.007 −0.072 −0.094
Ugly activities dummy −0.078 −0.017 −0.056 −0.058 −0.129
Log precolonial population density −0.120 0.002 −0.050 −0.012 −0.134
Robust standard errors (clustered at precolonial population density level) in brackets. Regressions in panel A include country fixed effects, as well as climate and geography controls. The number of observations
varies, depending on data availability. Standardized effects in panel B show the standard deviation change in the dependent variable for a 1 standard deviation change in the colonial activities variable. Significance
levels: ∗ 10%, ∗∗ 5%, ∗∗∗ 1%.
income inequality, human capital, and political represen- Carneiro (2007) document that enforcement of labor regula-
tation. Our theoretical discussion suggests that extractive tion varies widely across cities within Brazil and that areas
colonial activities went along with the formation of an eco- with stricter enforcement have higher unemployment. Laeven
nomic and political elite. As a result, society came to be and Woodruff (2007) exploit the fact that state laws and also
dominated by relatively few individuals, making it difficult legal enforcement differ from state to state in Mexico to show
for others to prosper and acquire human and physical capital. that average firm size is larger in states with more effective
Based on this argument, colonial activities could be correlated legal systems. In order to verify whether differences in institu-
with income inequality, such that areas with bad or ugly colo- tions, or regulations, are a possible channel, we would need a
nial activities and high population density are more unequal measure of institutions at the subnational level. To our knowl-
today, which could imply that these areas have lower lev- edge, such a measure does not yet exist for the set of countries
els of development today. Similarly, education levels could in our analysis. Some of the countries, such as Brazil and
also be lower in areas with unfavorable colonial activities, Mexico, have some measures or proxies for institutions at
which could lead to lower levels of development. We discuss the state level. However, these measures differ from country
the third possible channel, political representation, in more to country, and the coverage within country is often limited.
detail below. For the other three channels mentioned above, we have
Ideally, we would also like to examine institutions, such data at the subnational level. Panel A of table 9 displays the
as property rights, as a channel. The previous literature has results of regressing proxies for each of the three channels
often referred to institutions as constraints on the govern- on our colonial variables. The regressions include all con-
ment and security of property rights and has argued that these trol variables. Column 1 of table 9 reproduces the results for
particular institutions drive economic growth at the country GDP per capita from column 5 of table 5, as a benchmark.
level. At the subnational level, constraints on the govern- Panel B of table 9 shows the size of the standardized effects
ment may be less important, since these constraints typically of the colonials variables on each dependent variable (that
relate to the central government only, and policies and reg- is, by how many standard deviations the dependent variable
ulations may be more important. Clearly regions within a increases with a 1 standard deviation increase in the colo-
country are subject to many common policies and regulations nial variable). Panel B also displays the same standardized
set by the central government. However, in many countries, effects of colonial activities on GDP per capita, correspond-
regions also set their own local policies. Moreover, the way ing to the regression results in column 5 of table 5. In order
in which de jure national policies and regulations are applied to evaluate whether each potential mechanism could account
and enforced locally often varies, implying that de facto for the effect of history on development, we consider three
institutions could be different. For example, Almeida and criteria:
GOOD, BAD, AND UGLY COLONIAL ACTIVITIES 451
1. The estimated coefficients on the colonial variables in Next, we study the potential role of political representation
each column are statistically significant, and their sign in column 5 of table 9. Our dependent variable is the log of the
is consistent with our theoretical predictions.39 ratio of seats in the lower house to the total population in each
2. The size of the effect of each colonial variable is region.41 Our rationale for using this variable relates to Bruhn,
economically significant; the colonial variables have Gallego, and Onorato’s (2010) ongoing research on legisla-
nontrivial effects on the channel variables that could be tive malapportionment. Legislative malapportionment results
consistent with the size of our reduced-form estimates when the number of seats per voter is unequally distributed
of the effects of colonial variables on development. across regions within a country, leading to an overrepresen-
3. The regressions show a differential effect of good ver- tation of some regions and an underrepresentation of other
sus bad and of good versus ugly colonial activities on regions. Bruhn, Gallego, and Onorato (2010) develop a the-
the channel variables. oretical argument and provide corresponding evidence that
political elites used legislative malapportionment to secure
The first possible channel is that extractive colonial activ- their political power after transition to democracy. That is,
ities led to higher inequality, which led to lower GDP per the paper finds that areas that used to vote for parties associ-
capita, implying that the correlation of inequality and colonial ated with military dictators before transition to democracy are
activities should have the opposite sign from the correlation overrepresented today. These areas continue to elect the same
of economic development and colonial activities. Column parties as before transition to democracy. The paper also finds
2 in table 9 shows a regression of the log Gini index on that areas that are overrepresented in the lower house receive
colonial activities. Areas with bad and ugly colonial activ- higher transfers from the central government and have a lower
ities do not have statistically different levels of inequality degree of political competition, both of which can influence
today from areas with good or areas with no colonial activ- economic outcomes.
ities. Precolonial population density is also not significantly We believe that the link to colonial activities and colonial
correlated with inequality today. Overall, current income elites could be as follows. Areas with bad and ugly colonial
inequality does not seem to be a relevant mechanism for activities tended to be wealthier during the colonial period
explaining the effects of colonial activities on current levels than other areas (since they produced the highest-value prod-
of development.40 ucts). They also had the strongest elites capturing this wealth
The second possible link between colonial activities and and unequal societies, probably with a small middle class.
current economic outcomes we study is human capital. We When military dictatorships developed in Latin America, the
use two proxies for education in each region: the log of military and associated parties were often supported by the
schools per child and the log literacy rate. The first vari- middle class (see, for example, Burns, 1993), possibly in an
able measures the current availability of schools, and the effort to break the political power of colonial elites. This
second is a proxy for the current stock of human capital of then might have led to political overrepresentation of areas
the population. Columns 3 and 4 in table 9 present the results. with good and no colonial activities (through the mechanism
None of the colonial variables show any statistically signifi- described in Bruhn et al., 2010) after transition to democracy
cant correlation with our human capital variables. Therefore, relative to areas with bad and ugly colonial activities. This
current schooling and education do not seem to explain the overrepresentation could have resulted in increased economic
correlation between colonial activities and current levels of benefits for areas with good and no colonial activities (such
development. as higher transfers from the central government) that could
explain why areas with bad and ugly colonial activities are
39 Formally, this criterion works as follows. In each case, we estimate the less economically developed today.
following equation, The results in column 5 of table 9 show a pattern that
Mrc = Zrc αI + Xrc βI + θc + εrc , is quite similar to our regression results with current GDP
where M is a measure of a potential mechanism. This regression includes per capita as the dependent variable: bad and ugly colo-
the vector of historical variables, Z, and control variables, X, as well as a nial activities, as well as precolonial population density,
country fixed effect, θ. We then assess whether the variable M could explain are negatively correlated with political representation, and
the effects of colonial activities on development by verifying whether
the standardized effects are big. Areas with good colonial
∂Y activities are not statistically different from areas with no
sign(αI ) = sign(α) × sign ,
∂I activities. Comparing areas with good and bad colonial activ-
where ∂ Y
is the theoretical partial effect of variable I on economic devel- ities suggests that areas with bad activities are less politically
∂I
opment (Y ). For instance, ES argue that more inequality leads to lower
Y
represented today, and this difference is statistically signif-
levels of development, implying that ∂ ∂I
< 0. Therefore, the correlation icant (p-value of 0.01). Similarly, areas with ugly colonial
of inequality and colonial activities should have the opposite sign from
the correlation of economic development and colonial activities, such that activities are less politically represented than areas with good
sign(αI ) = −sign(α).
40 Nunn (2008a) and Acemoglu et al. (2008) find evidence against the
argument that economic inequality is the reason that colonial activities and 41 We use lower house representatives because the composition of this
historical variable influence current levels of development, using data from chamber is typically less distorted by geopolitical factors than the upper
the U.S. and Colombia, respectively. house (see Bruhn et al., 2010, for a more detailed discussion).
452 THE REVIEW OF ECONOMICS AND STATISTICS
colonial activities (with a p-value of 0.02). Overall, the results We examine three potential channels that could be the
for political representation are consistent with this channel link between colonial activities and current level of develop-
being a link between colonial activities and current levels of ment: inequality, human capital, and political representation.
development. The previous literature has argued that two of these vari-
All in all, the regressions in table 9 indicate that only the ables, income inequality and human capital, are correlated
correlations between colonial activities and political repre- with colonial characteristics at the country level. Our within-
sentation are fully consistent with the correlations between country regression, however, does not show a correlation
colonial activities and current levels of development. The between these two variables and colonial activities. But we
other two mechanisms we examine, income inequality and do find that a variable that has been less emphasized in the
human capital, do not seem to explain the effects of colonial literature, political representation, is highly correlated with
activities on current levels of economic development. colonial activities. Areas with high precolonial population
density and areas with bad activities are underrepresented
in the lower house today. This could lead to lower levels of
IX. Conclusion development since underrepresented areas have been shown
to receive fewer transfers and to be subject to less political
This paper shows that within-country differences in lev- competition. We are not able to investigate whether differ-
els of economic development in the Americas are correlated ences in regulations or other subnational institutions are also
with the type of activity performed in each region during correlated with colonial activities due to a lack of consistent
colonial times. In particular, it provides evidence that areas subnational data. More research is needed to study the effect
that were suitable for the activities that displayed economies of colonial activities on these types of institutions.
of scale and relied heavily on the exploitation of labor, such
as mining and plantation agriculture, have lower levels of
economic development today. The estimated effects are eco- REFERENCES
nomically relevant. Our estimates imply areas that had bad
Acemoglu, Daron, “Comments to Factor Endowments, Inequality, and
colonial activities (mining and plantation agriculture) have Paths of Development among New World Economies,” Economía
27.6% lower GDP per capita today than areas that had no 3:1 (2002), 94–107.
colonial activities within the same country. Areas with bad Acemoglu, Daron, Maria Angelica Bautista, Pablo Querubin, and James A.
Robinson, “Economic and Political Inequality in Development: The
colonial activities also have 27.7% lower GDP per capita Case of Cundinamarca, Colombia,” in E. Helpman (Ed.), Institutions
than areas that had good colonial activities (activities with- and Economic Performance (Cambridge, MA: Harvard University
out economies of scale, such as growing staple crops and Press, 2008).
Acemoglu, Daron, Simon Johnson, and James A. Robinson, “The Colonial
textile production that were performed in areas with low pre- Origins of Comparative Development: An Empirical Investigation,”
colonial population density). Our results also show that ugly American Economic Review 91 (2001), 1369–1401.
colonial activities (activities without economies of scale that ——— “Reversal of Fortune: Geography and Institutions in the Making
of the Modern World Income Distribution,” Quarterly Journal of
were performed in areas with high precolonial population Economics 117 (2002), 1232–1294.
density and therefore typically relied on the exploitation of ——— “Institutions as the Fundamental Cause of Long-Run Growth”
local labor) have 15.9% lower GDP per capita today than (pp. 385–472), in Philippe Aghion and Stephen Durlauf (Eds.),
Handbook of Economic Growth (Amsterdam: Elsevier, 2005).
areas with no activities and 16% lower GDP per capita today Acemoglu, Daron, and James A. Robinson, “De Facto Political Power and
than areas with good activities. Institutional Persistence,” American Economic Review 96 (2006),
These results lend support to ES’s (1997, 2002) argument 325–330.
Acemoglu, Daron, David Ticchi, and Andrea Vindigni, “Emergence and
that the type of colonial activity performed in a region mat- Persistence of Inefficient States” (2007), http://econ-www.mit.edu/
tered greatly for long-run development paths. More broadly, faculty/acemoglu/paper.
our results extend the theoretical and empirical findings of the Almeida, Rita, and Pedro M. Carneiro, “Inequality and Employment in a
Dual Economy: Enforcement of Labor Regulation in Brazil,” IZA
recent literature that investigates the effects of colonial char- discussion paper no. 3094 (2007).
acteristics on development at the country level. The fact that Andrews, Charles M., “Colonial Commerce,” American Historical Review
we find within-country correlations between colonial activ- 20 (1914), 43–63.
Armani, Alberto, Ciudad de Dios y Ciudad del Sol (El “Estado” jesuita
ities and current levels of development suggests that it is de los guaraníes (1609–1768)) (Mexico City, Mexico: Fondo de
not only the identity (nationality) of the colonizers that mat- Cultura Económica, 1988).
ters for subsequent development, as argued by some papers. Arze Aguirre, Rene, Breve Historia de Bolivia, Vol. 2 (La Colonia Sucre,
Bolivia: Universidad Andina Simón Bolívar).
The identity of the colonizer varies across countries, but we Banerjee, Abhijit, and Lakshmi Iyer, “History, Institutions and Economic
control for country effects. Moreover, our paper provides Performance: the Legacy of Colonial Land Tenure Systems in India,”
within-country evidence for a reversal of fortunes due to col- American Economic Review 95 (2005), 1190–1213.
Banerjee, Abhijit, Lakshmi Iyer, and Rohini Somanathan, “History, Social
onization that AJR (2002) find at the country level. In areas Divisions and Public Goods in Rural India,” Journal of the European
that had no colonial activities, precolonial population density Economic Association 3 (2005), 639–647.
(a proxy for precolonial levels of development) is positively Barro, R. J., and X. Sala-i-Martin, Economic Growth (Cambridge, MA:
MIT Press, 1999).
correlated with current GDP per capita. In areas with colonial Bauza, Francisco, Historia de la dominación española en el Uruguay
activities, this correlation is negative. (Montevideo: A. Barreiro y Ramos, 1895).
GOOD, BAD, AND UGLY COLONIAL ACTIVITIES 453
Bethell, Leslie (Ed.), Colonial Brazil (Cambridge: Cambridge University Gómez-Galvarriato, Aurora, “Premodern Manufacturing” (pp. 357–394), in
Press, 1987). Victor Bulmer-Thomas, John H. Coatsworth, and Roberto Cortés-
Bonet, Jaime, and Adolfo Meisel Roca, “El Legado Colonial como Deter- Conde (Eds.), The Cambridge Economic History of Latin Amer-
minante del Ingreso per Cápita Departamental en Colombia,” Banco ica, Vol. 1: The Colonial Era and the Short Nineteenth Century
de la República, Sucursal Cartagena Documentos de Trabajo Sobre (Cambridge: Cambridge University Press, 2006).
Economía Regional no. 75 (2006). Grassi, B., A. Pasten, and Jorge Armoa, “Análisis de la tendencia de
Brown, Craig (Ed.), The Illustrated History of Canada, 4th ed. (Bolton, la temperatura del aire en el Paraguay,” unpublished nanuscript,
ON: Key Porter Books, 2003). Universidad Nacional de Asunción (2004).
Brown, J., A Brief History of Argentina (New York: Facts on File, 2003). ——— “Un análisis del comportamiento de la precipitación en el
Bruhn, Miriam, Francisco Gallego, and Massimiliano Onorato, “Malappor- Paraguay,” unpublished manuscript, Universidad Nacional de Asun-
tionment and Economic Development: An Economic Assessment,” ción (2005).
unpublished manuscript, Catholic University of Chile (2008). Hamnett, Brian R., A Concise History of Mexico (Cambridge: Cambridge
——— “Legislative Reapportionment and Institutional Persistence,” World University Press, 1999).
Bank Policy working paper no. 5467 (2010). Hong, Sok Chul, “The Burden of Early Exposure to Malaria in the United
Bucheli M., and W. Cabela, “Perfil demografico y socioeconomico de States, 1850–1860: Malnutrition and Immune Disorders,” Journal
la poblacion uruguaya según su ascendencia racial” (Encuesta of Economic History 67 (2007), 1001–1035.
Nacional de Hogares Ampliada, 2006). Huillery, E., “Colonization, Institutions and Development in the Former
Burns, Bradford E., A History of Brazil (New York: Columbia University French West Africa,” unpublished manuscript (2007).
Irwin, Douglas, and Marko Tervio, “Does Trade Raise Income? Evidence
Press, 1993).
from the Twentieth Century” NBER working paper no. 7745 (2000).
Calvo, Alfredo, Analisis de la Situacion Socioeconomica del Pais (Organi- Iyer, Lakshmi, “The Long-Term Impact of Colonial Rule: Evidence from
zación Panamericana de la Salud, 2000). India,” Harvard Business School working paper no. 05-041 (2005).
CEPAL, Población indígena y afroecuatoriana en Ecuador: Diagnóstico Jiménez, Alfredo, Antropología histórica: La audiencia de Guatemala en
sociodemográfico a partir del censo de 2001 (CEPAL, 2005). el siglo XVI (Sevilla: Universidad de Sevilla, 1997).
Cobb, Gwendolin B., “Supply and Transportation for the Potosí Mines, Kellerman, Michael, “Power Resources Theory and Inequality in the
1545–1640,” in James Gerbera and Lei Guang (Eds.), Agriculture Canadian Provinces,” mimeograph, Harvard University (2005).
and Rural Connections in the Pacific, 1500–1900 (Farnham, UK: Klein, Herbert, A Concise History of Bolivia (Cambridge: Cambridge
Ashgate Publishing, 2006). University Press, 2003).
Collier, Simon, and William F. Sater, A History of Chile, 1808–2002, 2nd Kling, J. R., J. B. Liebman, L. F. Katz, and L. Sanbonmatsu, “Moving to
ed. (Cambridge: Cambridge University Press, 2004). Opportunity and Tranquility: Neighborhood Effects on Adult Eco-
Cook, Noble David, Demographic Collapse: Indian Perú, 1520–1620 nomic Self-Sufficiency and Health from a Randomized Housing
(Cambridge: Cambridge University Press, 1981). Voucher Experiment,” Princeton University, Department of Eco-
Cooper, John M., “The Araucanians” (pp. 687–760), in Julian H. Stew- nomics, Industrial Relations Section, working paper no. 481 (2004).
ard (Ed.), Handbook of South American Indians (Washington, DC: Knight, Alan, Mexico: The Colonial Era (Cambridge: Cambridge
Smithsonian Institution, 1946). University Press, 2002).
Cumberland, Charles C., Mexico: The Struggle for Modernity (New York: Laeven, Luc, and Christopher Woodruff, “The Quality of the Legal System
Oxford University Press, 1968). and Firm Size,” this review 89 (2007), 601–614.
Deerr, Noel, The History of Sugar (London: Chapman and Hall, 1949– Lombardi, John V., Venezuela: The Search for Order, the Dream of
1950). Progress (New York: Oxford University Press, 1982).
Dell, Melissa, “The Persistent Effects of Peru’s Mining Mita,” unpublished Lugones, Leopoldo, El Imperio Jesuítico (Buenos Aires: Hyspamérica
manuscript, MIT (2010). Ediciones, 1985).
Denevan, William M. (Ed.), The Native Population of the Americas in 1492, McCusker, John J., and Russell R. Menard, The Economy of British Amer-
2nd ed. (Madison: University of Wisconsin Press, 1992). ica, 1607–1789 (Chapel Hill, NC: University of North Carolina
Dobyns, Henry F., and Paul L. Doughty, Peru: A Cultural History (New Press, 1985).
York: Oxford University Press, 1976). McFarlane, Anthony, Colombia before Independence: Economy, Soci-
Easterly, William, and Ross Levine, “Tropics, Germs, and Crops: How ety, and Politics under Bourbon Rule (Cambridge: Cambridge
Endowments Influence Economic Development,” Journal of Mone- University Press, 2002).
tary Economics 50:1 (2003), 3–39. McGaw, Judith A., Early American Technology: Making and Doing Things
Eccles, William J., France in America (New York: Harper & Row, 1972). from the Colonial Era to 1850 (Chapel Hill: University of North
Engerman, Stanley L., “Contract Labor, Sugar and Technology in the Carolina Press, 1994).
Nineteenth Century,” Journal of Economic History 43 (1983), Merrouche, Ouarda, “The Long Term Impact of French Settlement on
635–659. Education in Algeria,” unpublished manuscript, 2007.
Engerman, Stanley L., and Kenneth L. Sokoloff, “Factor Endowments, Ministerio de Economia y Finanzas, Republica de Panama, “La pobreza en
Institutions, and Differential Paths of Growth among New World Panama. Encuesta de Niveles de Vida 2003.” mimeograph (2005).
Economies” (pp. 41–109), in Stephen H. Haber (Ed.), How Latin Mitchener, K. J., and I. McLean, “U.S. Economic Growth and Convergence,
1880–1980,” Journal of Economic History 59 (1999), 1016–1042.
America Fell Behind (Stanford, CA: Stanford University Press,
——— “The Productivity of U.S. States since 1880,” Journal of Economic
1997). Growth 8:1 (2003), 73–114.
——— “Factor Endowments, Inequality, and Paths of Development among Naritomi, Joana, Rodrigo Soares, and Juliano J. Assunção, “Rent Seeking
New World Economies,” Economia 3:1 (2002), 41–109. and the Unveiling of ‘De Facto’ Institutions: Development and
Fisher, John R., Government and Society in Colonial Peru: The Intendant Colonial Heritage within Brazil,” unpublished manuscript (2007).
System 1784–1814 (London: University of London, 1970). North, Douglas C., Structure and Change in Economic History (NewYork:
Fogel, Robert William, Without Consent or Contract (New York: Norton, Norton, 1981).
1989). North, Douglas C., and Robert P. Thomas, The Rise of the Western World:
Frankel, Jeffrey, and David Romer, “Does Trade Cause Growth?” American A New Economic History (Cambridge: Cambridge University Press,
Economic Review 89 (1999), 379–399. 1973).
Gallego, Francisco, “Historical Origins of Schooling: The Role of Democ- Nunn, Nathan, “The Long-Term Effects of Africa’s Slave Trades,”
racy and Political Decentralization,” this review 92 (2010), 228– Quarterly Journal of Economics 123:1 (2008a), 139–176.
243. ——— “Slavery, Inequality, and Economic Development in the Americas:
Gerhard, Peter, The Southeast Frontier of New Spain (Princeton, NJ: An Examination of the Engerman-Sokoloff Hypothesis,” in E. Help-
Princeton University Press, 1979). man (Ed.), Institutions and Economic Performance (Cambridge,
Glaeser, Edward L., Rafael LaPorta, Florencio Lopes-de-Silanes, and MA: Harvard University Press, 2008b).
Andrei Shleifer, “Do Institutions Cause Growth?” Journal of Eco- ——— “The Importance of History for Economic Development,” Annual
nomic Growth 9 (2004), 271–303. Review of Economics 1 (2009).
454 THE REVIEW OF ECONOMICS AND STATISTICS
Ocampo, José Antonio (Ed.), Historia económica de Colombia (Bogotá: Torres-Rivas, Edelberto, Historia general de Centroamérica: El régimen
Presidencia de la República, 1997). colonial (1524–1750) (Madrid: Sociedad Estatal Quinto, 1993).
Ots y Capdequí, J. M., Las instituciones del nuevo reino de Granada al Ubelaker, Douglas H., “The Sources and Methodology for Mooney’s
tiempo de la independencia (Madrid: Consejo Superior de Investi- Estimates of North American Indian Populations” (pp. 243–289),
gaciones Científicas, Instituto Gonzalo Fernández de Oviedo, 1958). in William M. Denevan (Ed.), The Native Population of the Amer-
Padre Juan de Velasco, Padre Juan de Velasco, S.I. Biblioteca ecuatoriana icas in 1492, 2nd ed. (Madison: University of Wisconsin Press,
mínima (Puebla, México: J. M. Cajica, 1960). 1992).
Peñaloza Cordero, Luis, Nueva Historia Economica de Bolivia (La Paz, Villamarín, Juan, and Judith Villamarín, “Chiefdoms: The Prevalence and
Bolivia, 1981). Persistence of ‘Señoríos Naturales,’ 1400 to European Conquest”
Powers, Karen V., Andean Journeys: Migration, Ethnogenesis and the (pp. 577–656), in Frank Salomon and Stuart B. Schwartz (Eds.),
State in Colonial Quito (Albuquerque: University of New Mexico The Cambridge History of the Native Peoples of the Americas,
Press, 1995). Volume 3: South America (Cambridge: Cambridge University Press,
Pyle, Jane, “A Reexamination of Aboriginal Population Estimates for 1999).
Argentina” (pp. 181–204), in William M. Denevan (Ed.), The Native Webre, Stephen, La sociedad colonial en Guatemala: Estudios regionales
Population of the Americas in 1492, 2nd ed. (Madison: University y locales, ed. Stephen Webre, trans. Margarita Cruz de Drake
of Wisconsin Press, 1992). (Antigua Guatemala: Centro de Investigaciones Regionales de
Reyes, Oscar Efren, Breve historia general del Ecuador (Quito: s.n., 1965). Mesoamérica, 1989).
Rivarola Paoli, Juan Bautista, La Economía Colonial (Asunción, Paraguay: Zabre, Alfonso Teja, Guide to the History of Mexico: A Modern
Editora Litocolor, 1986). Interpretation (Austin, TX: Pemberton Press, 1969).
Robles, Marcos, “Pobreza y Gasto Publico en Educacion en Paraguay,”
Economia y Sociedad 3 (2001).
Rock, David, Argentina, 1516–1987: From Spanish Colonization to APPENDIX A
Alfonsin (Berkeley: University of California Press, 1987).
Rodríguez Becerra, Salvador, Encomienda y conquista: Los inicios Variable Definitions
de la colonización en Guatemala (Sevilla: Publicaciones de la
Universidad de Sevilla, 1977). PPP GDP per capita: Gross state product for each state divided by the
Rosas, Andres, and Juan Mendoza, “The Economic Effects of Geogra- contemporaneous population of that state and converted to PPP values
phy: Colombia as a Case Study,” Universidad Javeriana Bogotá using the 2000 value from the World Development Indicators. Due
documentos de economía 003584 (2004). to data limitations, the data for Venezuela correspond to household
Rubio, Julián María, Exploración y conquista del río de La Plata: Siglos income.
XVI y XVII (Barcelona: Salvat, 1942). Poverty rate: Percentage of the population living below the poverty
Sanders, William T., “The Population of the Mexican Symbiotic Region, line, according to each country’s definition of the poverty line.
the Basin of Mexico, and the Teotihuacán Valley in the Sixteenth Health index: The health index measures the quality-adjusted-life-
Century” (pp. 85–151), in William M. Denevan (Ed.), The Native years (QALY) based on the health status attributed to skeletal remains,
Population of the Americas in 1492, 2nd ed. (Madison: University which display chronic health conditions and infections. The health
of Wisconsin Press, 1992). index is adjusted for the age distribution of the population and is a
Serrano Bravo, Claudio, “Historia de la minería andina en Bolivia (siglos simple average of seven health indicators: stature, hypoplasias, ane-
XVI–XX),” unpublished manuscript, UNESCO (2004). mia, dental health (teeth and abscesses), infections, degenerative joint
Spolaore, Enrico, and Romain Wacziarg, “Borders and Growth,” Journal disease, and trauma.
of Economic Growth 10 (2005), 331–386. Precolonial population density: The ratio of the estimated precolonial
Steckel, Richard H., “Biological Measures of the Standard of Living,” population to the area of modern states.
Journal of Economic Perspectives 22:1 (2008), 129–152. Colonial activities: Predominant economic activity performed during
Steckel, Richard H., and Jerome C. Rose (Eds.), The Backbone of History: the colony in the region that matches the current-day state.
Health and Nutrition in the Western Hemisphere (Cambridge: Average temperature: Average yearly temperature (◦ C).
Cambridge University Press, 2002). Total rainfall: Total yearly rainfall (mm).
Steckel, Richard H., Jerome C. Rose, Clark Spencer Larsen, and Phillip Altitude: Elevation of capital city of the state (km).
L. Walker, “Skeletal Health in the Western Hemisphere from Landlocked dummy: This dummy is equal to 1 if the state does not
4000 BC to the Present,” Evolutionary Anthropology 11 (2002) have a seacoast.
142–155. Gini index: Gini measure of income inequality for households (not
Tabellini, Guido, “Culture and Institutions: Economic Development in available for Honduras, Panama, Paraguay, and Uruguay).
the Regions of Europe” (2007), http://www.igier.uni-bocconi.it/ Schools per child: Total number of schools divided by the school-age
whos.php?vedi=327&tbn=albero&id_doc=177. population of a region (not available for Panama and Uruguay).
Tandeter, Enrique, “The Mining Industry” (pp. 357–394), in Victor Literacy rates: Percentage of the population that is literate (not
Bulmer-Thomas, John H. Coatsworth, and Roberto Cortés-Conde available for El Salvador, Honduras, Panama, and Paraguay).
(Eds.), The Cambridge Economic History of Latin America, Volume Ratio of seats to total voting population: The ratio of lower house
1: The Colonial Era and the Short Nineteenth Century (Cambridge: representatives of a region to the total number of people eligible to
Cambridge University Press, 2006). vote (not available for Panama, Peru, and Uruguay).
GOOD, BAD, AND UGLY COLONIAL ACTIVITIES 455
Table Appendix
Table A1.—Colonial Activities and Current GDP per Capita: Robustness
Coefficient on
Good Activities Bad Activities Ugly Activities Log Precolonial Population Density Observations R2
A: Excluded country
Argentina 0.047 −0.275∗∗∗ −0.160∗∗ −0.036 321 0.822
(0.081) (0.084) (0.080) (0.022)
Bolivia 0.004 −0.278∗∗∗ −0.167∗∗ −0.052∗∗ 336 0.812
(0.078) (0.083) (0.080) (0.020)
Brazil 0.024 −0.278∗∗∗ −0.158∗∗ −0.053∗∗ 318 0.832
(0.076) (0.082) (0.077) (0.021)
Canada 0.014 −0.287∗∗∗ −0.162∗∗ −0.055∗∗∗ 332 0.782
(0.081) (0.082) (0.079) (0.021)
Chile 0.004 −0.264∗∗∗ −0.142∗ −0.055∗∗ 332 0.818
(0.075) (0.083) (0.079) (0.022)
Colombia −0.017 −0.289∗∗∗ −0.120 −0.048∗∗ 315 0.851
(0.066) (0.085) (0.081) (0.020)
Ecuador −0.054 −0.208∗∗∗ −0.130∗ −0.039∗∗ 323 0.845
(0.066) (0.077) (0.076) (0.020)
El Salvador 0.001 −0.276∗∗∗ −0.158∗∗ −0.052∗∗∗ 333 0.813
(0.076) (0.082) (0.079) (0.020)
Guatemala 0.000 −0.277∗∗∗ −0.160∗∗ −0.052∗∗ 337 0.816
(0.076) (0.082) (0.079) (0.020)
Honduras 0.003 −0.289∗∗∗ −0.146∗ −0.052∗∗ 327 0.801
(0.076) (0.084) (0.079) (0.021)
Mexico 0.007 −0.267∗∗∗ −0.152∗ −0.063∗∗∗ 313 0.829
(0.076) (0.084) (0.083) (0.020)
Panama −0.014 −0.288∗∗∗ −0.171∗∗ −0.045∗∗ 336 0.821
(0.076) (0.082) (0.082) (0.020)
Paraguay 0.023 −0.256∗∗∗ −0.142 −0.053∗∗∗ 327 0.816
(0.082) (0.086) (0.086) (0.021)
Peru −0.009 −0.252∗∗∗ −0.156∗ −0.050∗∗ 321 0.820
(0.077) (0.085) (0.081) (0.022)
United States −0.031 −0.376∗∗∗ −0.240∗∗∗ −0.050∗∗ 297 0.697
(0.101) (0.104) (0.093) (0.022)
Uruguay 0.001 −0.263∗∗∗ −0.140 −0.062∗∗ 326 0.817
(0.084) (0.086) (0.086) (0.025)
Venezuela 0.006 −0.304∗∗∗ −0.186∗∗ −0.047∗∗ 326 0.817
(0.081) (0.088) (0.086) (0.021)
None 0.001 −0.276∗∗∗ −0.159∗∗ −0.052∗∗ 345 0.815
(0.076) (0.082) (0.079) (0.020)
Average 0.001 −0.278 −0.158 −0.051 326 0.812
Median 0.003 −0.277 −0.158 −0.052 327 0.817
Maximum 0.047 −0.208 −0.120 −0.036 345 0.851
Minimum −0.054 −0.376 −0.240 −0.063 297 0.697
B: Excluding countries with low data quality 0.024 −0.289∗∗∗ −0.137 −0.055∗∗ 254 0.810
(0.100) (0.103) (0.111) (0.027)
The dependent variable is log GDP per capita. Panel B excludes all the countries for which more than 50% of regions have data on this variable imputed from other countries (Guatemala, Honduras, Panama, Paraguay,
Uruguay, and Venezuela). Significance levels: ∗ 10%, ∗∗ 5%, ∗∗∗ 1%.
Table A2.—Colonial Activities and Current GDP per Capita: Different Definitions of Good and Ugly Activities
Dependent Variable: Log PPP GDP per Capita
Cutoff of Precolonial Median Average 75th Percentile
Population Density (1) (2) (3)
Good activities dummy 0.001 0.003 −0.035
(0.076) (0.078) (0.067)
Bad activities dummy −0.276∗∗∗ −0.271∗∗∗ −0.237∗∗∗
(0.082) (0.083) (0.077)
Ugly activities dummy −0.159∗∗ −0.147∗ −0.127
(0.079) (0.078) (0.094)
Log precolonial −0.052∗∗ −0.053∗∗∗ −0.059∗∗∗
population density (0.020) (0.020) (0.020)
Observations 345 345 345
R2 0.829 0.829 0.828
Coefficient Bad − Good −0.277 −0.274 −0.202
F -test: Good = Bad p-value (0.005) (0.007) (0.006)
Coefficient Ugly − Good −0.160 −0.150 −0.092
F -test: Good = Ugly p-value (0.084) (0.117) (0.301)
Coefficient Bad − Ugly −0.117 −0.124 −0.110
F -test: Bad = Ugly p-value (0.082) (0.060) (0.209)
Robust standard errors (clustered at precolonial population density level) in brackets. Regressions include country fixed effects, as well as climate and geography controls. Significance levels: ∗ 10%, ∗∗ 5%, ∗∗∗ 1%.
Appendix B.—Data Sources
456
Variable Argentina Bolivia Brazil Canada Chile Colombia
GDP INDEC: Dirección de Instituto Nacional de IBGE: Contas Regionais Statistics Canada Central Bank of Chile DANE: Cuentas
Cuentas Nacionales: Estadísticas de Departamentales
PBG por provincia y Bolivia: PIB
sector de actividad departamental
económica
Population INDEC: Censo Nacional Instituto Nacional de IBGE: Censo Statistics Canada MIDEPLAN projections DNP projections–2000
de Población, Estadísticas de Demográfico 2000 based on 2002 census
Hogares y Viviendas Bolivia: MECOVI
2001 1999
Poverty rate INDEC-EPH: May 2001 Instituto Nacional de http://tabnet.datasus Canadian Council on MIDEPLAN: 2000 SISD
Estadísticas de .gov.br/cgi/idb2004/ Social Development CASEN Survey
Bolivia: MECOVI b05uf.htm
1999
Health index Backbone of History Project (Steckel & Rose, 2002)
Colonial acitvities J. Brown (2003), Rock Peñaloza (1981), Arze Bethell (1987), C. Brown (2003) Collier and Sater (2004) McFarlane (1992),
(1987) Aguirre (1996), Klein Burns (1993) Ocampo (1997)
(2003), Serrano
(2004)
Precolonial population density Own calculations from Own calculations from Own calculations from Own calculations from Own calculations from Own calculations from
Pyle (1992) Denevan (1992) Denevan (1992) Denevan (1992) and Denevan (1992) Denevan (2002),
“Canada Natives Ocampo (1997),
People 1630” of the and Villamarín and
National Atlas of Villamarín (1999)
Canada
Temperature Servicio Metereológico Servicio Nacional de IBGE - Annuário National Climate Data Dirección Metereológica IDEAM
Nacional Meteorología e estatístico do Brazil and Information de Chile
Hidrología Archive
Rainfall Servicio Metereológico Servicio Nacional de IBGE - Annuário National Climate Data Dirección Metereológica IDEAM
Nacional Meteorología e estatístico do Brazil and Information de Chile
Hidrología Archive
Altitude Global Gazetteer Version 2.1 (www.fallingrain.com)
THE REVIEW OF ECONOMICS AND STATISTICS
GINI index Own calculations from Calvo, Alfredo (2000). IBGE: Censo Kellerman (2005) Own calculations from SISD
1998 EPH Análisis de la Demográfico 2000 2000 CASEN
Situación
Socioeconómica del
País. Organización
Panamericana de la
Salud
Number of schools INDEC: Estadísticas de Instituto Nacional de Sinopse estatística da Statistics Canada MINEDUC: Directorio Ministerio de Edu-
Educación y Ciencias Estadísticas de educação básica: de Establecimientos cación Nacional:
Bolivia: Estadisticas Censo escolar 2001 Educacionales 2000 Oficina Asesora
Departamentales de Brasília, DF: INEP, de de Planeación
Bolivia 2002 y Finanzas,
Estadísticas
Educativas
Appendix B.—(Continued)
Variable Argentina Bolivia Brazil Canada Chile Colombia
Literacy rate INDEC: Censo Nacional Instituto Nacional de IBGE: Censo 2001 Census MIDEPLAN: 2000 SISD
de Población, Hogares Estadísticas de Bolivia: Demográfico 2000 CASEN Survey
y Viviendas 2001 Estadísticas
Departamentales de
Bolivia 2005
Political representation Cámara Nacional Corte Nacional Electoral Tribunal Superior Elections Canada Tribunal Calificador de Consejo Nacional
Electoral Eleitoral Elecciones Electoral
Variable Ecuador El Salvador Guatemala Honduras Mexico Panama
GDP Banco Central del Informe del Desarrollo Informe del Desarrollo Informe del Desarrollo INEGI: Producto Interno Dirección de Estadís-
Ecuador: Cuentas Humano El Salvador Humano Guatemala Humano Honduras Bruto por Entidad ticas y Censos, PIB
Provinciales (2005) (2002) (2002) Federativa Provincial
Population Instituto Nacional de Dirección General de Informe del Desarrollo Informe del Desarrollo INEGI: Censo General de Dirección de Estadísti-
Estadísticas y Censos, Estadísticas y Censos, Humano Guatemala Humano Guatemala Población y Vivienda cas y Censos, 2000
2001 Census projections 2005 (2002) (2002) 2000 Census
Poverty rate Informe sobre Desarrollo Compendio Estadistico Informe Nacional de Informe sobre Desarrollo SEDESOL “La pobreza en Panama.
Humano, Ecuador Desarrollo Humano Humano 2002. PNUD Encuesta de Niveles
2001. PNUD Guatemala (2005) de Vida 2003.” Min-
isterio de Economia y
Finanzas, República de
Panamá. 2005
Health index Backbone of History Project (Steckel & Rose, 2002)
Colonial acitvities Reyes (1965); Padre Juan Rodríguez Becerra Webre (1989); Jiménez Torrer-Rivas (1993); Cumberland (1968), Ots y Capdequí (1958)
de Velasco (1960) (1977); Torrer-Rivas (1997) Jiménez (1997) Gerhard (1979),
(1993) Hamnett (1999),
Knight (2002), Zabre
(1969)
Precolonial population Own calculations from Own calculations from Own calculations from Own calculations from Own calculations from Own calculations from
density Denevan (1992) Denevan (1992) Denevan (1992) Denevan (2002) Denevan (2002) and Denevan (2002)
Sanders (1992)
GOOD, BAD, AND UGLY COLONIAL ACTIVITIES
Temperature Instituto Nacional de Servicio Nacional Instituto Nacional De Servicio Metereológico INEGI Dirección de
Meteorología e de Estudios Sismología, Nacional Meteorología
Hidrología Territoriales—Perfiles Vulcanología,
Climatológicos Meteorología E
Hidrología
Rainfall Instituto Nacional de Servicio Nacional Instituto Nacional De Servicio Metereológico INEGI Dirección de
Meteorología e de Estudios Sismología, Nacional Meteorología
Hidrología Territoriales—Perfiles Vulcanología,
Climatológicos Meteorología E
Hidrología
Altitude Global Gazetteer Version 2.1 (www.fallingrain.com)
GINI index Informe sobre Desarrollo Informe sobre Desarrollo Informe Nacional de — Own calculations from
Humano, Ecuador Humano, El Salvador Desarrollo Humano 2000 ENE
2001, PNUD 2005, PNUD Guatemala (2003)
457
Appendix B.—(Continued)
458
Variable Argentina Bolivia Brazil Canada Chile Colombia
Schools per children Ministerio de Educación, Ministerio de Educación, Ministerio de Informe de Progreso INEGI: Anuario de
Censo Nacional de Censo Matricular, Educación—Boletín Educativo, Honduras, Estadísticas por
Instituciones 1999–2000 Estadístico, 2001 2005, PREAL Entidad Federativa,
Educativas 2003
Literacy rate Informe sobre Desarrollo PREAL. Informe de Informe del Desarrollo Informe del Desarrollo INEGI: Censo General de Informe Nacional de
Humano, Ecuador Progreso Educativo, El Humano Guatemala Humano Honduras Población y Vivienda, Desarrollo Humano,
2001, PNUD Salvador, 2002 (2002) (2002) 2000 2002, PNUD
Political representation Consejo Nacional — Tribunal Supremo Tribunal Supremo Mexico Electoral, Tribunal Electoral
Electoral Electoral Electoral 1970–2003 Banamex
CD
Variable Paraguay Peru United States Uruguay Venezuela
GDP Atlas de Desarrollo INEI: Dirección Nacional BEA: Gross domestic Anuario Diario El País, Own calculations from
Humano Paraguay, de Cuentas Nacionales: product by state 2001 1998 EHM (household
2005 PBI por departamento income)
Population Dirección General de INEI U.S. Census Bureau Instituto Nacional de INE
Estadísticas, Encuestas Estadística de Uruguay
y Censos
Poverty rate Robles (2001) INEI State and Metropolitan Desarrollo Humano en INE
Area Data Book, Uruguay 2001. PNUD
1997–1998
Health index Backbone of History Project (Steckel & Rose, 2002)
Colonial acitvities Lugones (1985), Rivarola Fisher (1970), Dobyns Andrews (1914), Eccles Bauza (1895); Rubio Lombardi (1982)
(1986), Armani (1988) and Doughty (1976) (1972), McCusker and María (1942)
Menard (1985)
Precolonial population Own calculations from Own calculations from Own calculations from Own calculations from Own calculations from
density Denevan (1992) Denevan (2002) and Ubelaker (2002) Denevan (1992) Denevan (1992)
Cook (1981)
Temperature Grassi, Pasten, and Armoa INEI http://www.met.utah Wikipedia.org INE
(2004) .edu/jhorel/html/wx/
climo.html
THE REVIEW OF ECONOMICS AND STATISTICS
Rainfall Grassi et al. (2005) INEI http://www.met.utah.edu/ Wikipedia.org INE
jhorel/html/wx/climo.html
Altitude Global Gazetteer Version 2.1 (www.fallingrain.com)
GINI index — Own calculations from U.S. Census Bureau, — Own calculations from
2000 ENAHO Table S4 1998 EHM
Schools per kids La Educación en Cifras INEI U.S. Department of — INE
2000. Ministerio de Education, National
Educación y Cultura, Center for Education
DPEI Statistics
Literacy rate — INEI — Desarrollo Humano en INE: 2001 Census
Uruguay 1999, PNUD
Political representation Tribunal Supremo de — House Election Statistics Corte Electoral Consejo Nacional
Justicia Electoral Electoral
GOOD, BAD, AND UGLY COLONIAL ACTIVITIES 459
APPENDIX C (1992) for the total population estimates and on and the map “Canada
Natives People 1630” of the National Atlas of Canada (5th edition) for
Precolonial Population Density the distribution of native people across regions. This map provides infor-
mation on the location and population of native settlements around 1630.
This appendix describes in detail how we construct the precolonial pop- So we use data in the map to compute an initial estimate by region and
ulation density variable. We use data from several sources to estimate next adjust the map estimates so as to match Denevan’s estimates for the
precolonial population density at the state level. The main sources of infor- macroregions of eastern (54,200) and central Canada (50,950).
mation are region-specific chapters in Denevan (1992) and references cited
in that book. This section presents the main sources for each country and C5. Chile
explains the assumptions we used to impute population estimates for the
different regions of each country. In each case, we adjust the estimated size In the case of Chile, there are no detailed estimates of population by
of the native population in each country to match the numbers presented in state. Instead, there is some information on the location of several native
Denevan (1992, table 00.1). groups, except for the Mapuche people. In this case, Cooper (1946), quoted
in Denevan (1992), estimates a precontact population of the Mapuche people
C1. Argentina of between 500,000 and 1.5 million, and we use the mean point of 1 million.
We also know that these people were located between the fifth and the tenth
The only source of information we use corresponds to Pyle (1992), a regions. So we estimate a precontact population density of 4.7. For the other
chapter in Denevan (1992). This paper includes several estimates of the regions in the country, we know the location of other people and take the
native population for different regions of Argentina. We take the average estimates of population density for these tribes in neighboring countries. In
of the number of natives in each region as our estimate of the denominator. particular, we know that about half of the modern first region was populated
In addition, using maps from the same paper, we allocate different tribes or by tribes linked to the Inca empire. So we use half of the estimate we have
groups to the different modern states. As some of the Argentinean regions for the Tacna region in Peru, which is equal to 1.3. For the second region, we
identified in Pyle (1992) correspond to clusters of more than one modern know it was just sparely unpopulated so we use an estimate of 0.1 (similar
Argentinean state, we estimated population density for the regions presented to the estimate used by Denevan, 1992, for other sparely populated regions
in Pyle (1992) and impute the same population density for all the states in the in Latin America). The third region was populated in part by the Diaguita
same region. In particular, the regions that include more than one state are: people, who also lived in the Catamarca region in Argentina. So we use
(a) Buenos Aires and Capital Federal, and (2) Chubut, La Pampa, Neuquén, half of the estimate for 0.13 for the region and 0.1 for the remainder area
Río Negro, Santa Cruz, and Tierra del Fuego. of the region. The fourth region was populated by the Diaguita people,
so we use in this case the same estimate as for Catamarca, equal to 0.17.
Finally, the peoples living to the south of the tenth region were basically the
C2. Bolivia same as those living in the Argentinean Patagonia, so we assume the same
population density, equal to 0.01 people per square kilometer.
The information for Bolivia comes from Denevan (1992) for the east of
the country. We also use estimates for the South Sierra derived from Cook
(1981), implying a population density of 17.3 people per square kilometer C6. Colombia
for the South Sierra. In addition, Denevan (1992) presents his preferred
estimated population figures for different regions of northeastern Bolivia: We take the information on total precontact population for Colombia
floodplain (14.6 people per square kilometer), lowland savanna, mainly from Denevan (1992, table 00.1). He estimates a total population of 3 mil-
Mojos (2.0), Santa Cruz area (1.8), upland forest (1.2), lowland forest (0.2), lion people. Using information from Ocampo (1997) and Villamarín and
and superhumid upland forest (0.1). Using estimates for the area of each Villamarín (1999), we estimate population densities for eight regions: east-
state belonging to each region, we estimate population density in each state. ern Cordillera (13 people per square kilometer), Cauca Valley (9.2), the
Caribbean coast (2.8), Upper Magdalena (4.9), Lower Magdalena (4.3),
Pasto (7.7), and Llanos (1.3). In the case of the Amazonas region, we use
C3. Brazil estimates for the Brazilian amazonas from Denevan (1992), which are equal
to 0.2 people per square kilometer. Next, we classify each modern state
The main source of information is Denevan (1992). Denevan presents in one of these regions according to the Colombian maps of the Perry-
estimated population density at time of contact for different habitats in Castañeda Library Map Collection of the University of Texas. Finally, for
Greater Amazonia, which includes most of the Brazilian states except for the San Andrés, Providencia, and Santa Catalina islands, we use population
portions of the coastal states in the south (Paraná, Rio Grande do Sul, density for the Caribbean islands from Denevan (1992).
Santa Catarina, and São Paulo). The habitats (estimated population density
at contact time) considered by Denevan are central coast (9.5 people per
square kilometer), floodplain (14.6), lowland–Amazon Basin (0.2), man- C7. Ecuador
grove coasts (4.75),42 and upland and central savannas (0.5). Using these
estimates, we classify each Brazilian state in each habitat and estimate pop- Estimates for Ecuador are very sparse, and we apply estimates for neigh-
ulation density for the states. In the cases that a state has more than one boring countries and complement them with some information available in
habitat, we use a weighted average considering the different habitats. In Powers (1995) for the coastal regions. We classify each state into the fol-
order to identify the habitats of the different regions, we use information lowing regions: central Andes (for which we use an estimated population
from the Natural Vegetation Map from the Perry-Castañeda Library Map density of 12.1 people per square kilometer, which is the average for similar
Collection of the University of Texas. regions in Colombia and Peru); coast (for which use estimates from Viera
For the southern states we also use information from Denevan (1992, Powers that range from 1 to 2 people per square kilometer); upland forest
table 00.1) on the total population for southern coastal Brazil combined (1.2, from Denevan); and east (0.7 from similar regions in Colombia and
(which implies a population density of four people per square kilometer) Perú). Using estimates for the area of each state belonging to each region,
with the previous information on the density for the different habitats of the we estimate population density in each state.
Greater Amazonia. Finally, we impute the population density of the state
of Goias to the Federal District (Brasilia). C8. El Salvador
C4. Canada Denevan (1992) argues that population in Central America was mainly
located in the plain regions close to the Pacific coast, “where there were rich
The information for British Columbia comes from Denevan (1992), volcanic soils from Guatemala to Costa Rica, and also in Panamá.” Thus,
equivalent to 85,800. For the other regions of Canada, we rely on Denevan for all Central American countries, we keep this stylized fact in mind in
order to assign populations to different regions. In addition, Denevan gives
an estimate of the total population living in El Salvador before contact with
42 For mangrove coasts, Denevan (1992) states “Probably considerably colonizers of about 500,000. Thus, we classify all states in two regions:
less than 9.5 per square kilometer.” We use 50% of 9.5. coast and mountains. In the case of population density for mountains, we
460 THE REVIEW OF ECONOMICS AND STATISTICS
use 0.01 people per square kilometer, and for the coastal regions, we use a in Argentina, Bolivia, and Brazil as benchmarks to estimate population den-
population density of 39.3 people per square kilometer, so that we generate sity in different regions. In particular, for Alto Paraguay, we use the average
a total population of 500,000. As for other countries, using estimates for the population density of Santa Cruz (Bolivia) and Matto Grosso do Sul (Brazil).
area of each state belonging to each region, we estimate population density For Alto Paraná and Caaguazú, we use the estimated population density for
in each state. the interior areas of neighboring Paraná (Brazil). For Amambay, we just
the estimate from Matto Grosso so Sul (Brazil). For Asunción, Central,
and Cordillera, we use weighted averages of the Argentinian regions of
C9. Guatemala Corrientes and Formosa. For Boquerón, we use population density from
the Chaco region in Argentina. For Caazapá and Guairá we use the sim-
As for El Salvador, we take advantage of the estimate of the total pop- ple average of the estimates for Alto Paraná and Misiones (Argentina). For
ulation from Denevan (1992). In this case Denevan gives an estimate of 2 Canindeyú, we also use estimates for Alto Paraná, but in this case, we take
million. To distribute this population in the states, we proceed as follows. the simple average with population density for Matto Grosso do Sul. For
First, we consider the state of Petén and parts of the Norte and Norocci- Concepción, we take the simple average of Matto Grosso do Sul and Chaco.
dente states. For these states, we use a population density of 5.63 people per For Itapúa, we use the simple average of Rio Grande do Sul (Brazil) and
square kilometer, which corresponds to the simple average of the popula- Misiones (Argentina). For Misiones, we use the average of the Argentinean
tion density of the Mexican state of Campeche and the estimated population states of Corrientes and Misiones. For Ñeembucú, we use a weighted aver-
density for Belize. Second, we assign a population density of 23.60 to all age of estimates for Formosa and Chaco in Argentina. For Paraguarí, we use
areas on the coast (which correspond to parts of the states of Central, Suror- the average of estimates for Misiones (Paraguay) and Central. For Presidente
iente, and Suroccidente), where the value 23.60 is our estimate for the state Hayes, we apply the estimates from Formosa (Argentina). And, finally, for
of Ahuachapán in Salvador. This leaves us with a total estimated popula- San Pedro, we take the average of Presidente Hayes and Canindeyú. All
tion of about 500,000 people. The remaining population corresponds to the these estimates imply a population density of 0.9, similar to those implied
highlands, which were populated by Mayan tribes (such as the Cakchiquel in Denevan’s calculations.
and the Quiché). Thus, we assign 29.05 people per square kilometer to these
areas, so as to arrive at the total population estimated in Denevan.
C14. Peru
C10. Honduras The information for Peru comes from Cook (1981) for most of the regions
in the country and from Denevan (1992) for the east of the country. In par-
As for El Salvador and Guatemala, we take advantage of the estimate of ticular, Cook (1981) presents his preferred estimated population figures for
the total population from Denevan (1992), which is 750,000 for both Hon- six Peruvian regions: north coast, central coast, south coast, north sierra,
duras and Belize. We assume a similar population density in both areas and central sierra, and south sierra. From Denevan (1992), we estimate the pop-
therefore get a total estimated population of 622,843 people for Honduras. ulation density for six regions located in the east of the country: Amazonas
To distribute this population across the states, we proceed as follows. First, (50% of the area), Loreto, Madre de Dios, Puno (50% of the area), and
we consider the coastal states of Choluteca and Valle and parts of the state of Ucayali.
El Paraíso. For these areas, we apply a population density of 17.70, which
corresponds to the simple average of the coastal states of La Unión and
Morazán in El Salvador. This leaves us with a total estimated population of
about 220,000 people. The remaining population corresponds to the eastern C15. United States
regiones of the country, which were populated by several peoples, such as
the Lencas. Thus, we apply an estimate of 8.19 people per square kilometer The raw information on the native population of the United States comes
to these areas, to get the total population estimated in Denevan. from Ubelaker (1992). This paper presents information on the native pop-
ulation of all the tribes in the United States and the location of these tribes
(see map 8.1, p. 244). Using this information, we assign each tribe to the
C11. Mexico modern U.S. states and in this way estimate precontact population densities.
In some cases, it was impossible to estimate population densities for spe-
Estimates for Central Mexico come from Sanders (1992), in particu- cific states because some tribes lived in more than one state, so we present
lar for Mexico, DF, Hidalgo, Puebla, Tlaxcala, Tamaulipas, and Morelos. population density estimates for groups of modern states. This is the case
In addition, Denevan (1992) presents population estimates for the follow- for Arizona and New Mexico; Delaware and New Jersey; Rhode Island
ing regions: Baja California Norte and Sur; Campeche, Quintana Roo, and and Massachusetts; Maryland and Washington, DC; and Virginia and West
Yucatán; Chiapas; Chihuahua, Durango, Sinaloa, and Sonora; Coahuila de Virginia.
Zaragoza and Nuevo León; Colima; and Tabasco. In the cases in which a
region includes more than one state, we impute the same population den-
sity for each region. As in all the other cases, we adjust the population C16. Uruguay
estimates so as to match the total estimate for Mexico from Denevan (1992,
table 00.1). First, we consider a number of regions for which there was no evidence of
being settled by natives. The states of Artigas, Flores, Florida, Lavallejana,
Montevideo, Rivera, Canelones, Maldonado, and San José fall into this cat-
C12. Panama egory. We assign a population density of 0.01 people per square kilometer to
all these states. Next, we consider regions in which there was some evidence
As for all other Central American countries, we take advantage of the of settlements by some peoples, such as the Gueonas, Chaná, Bohan, and
information that coastal areas were more densely populated. In this case, Charrua. These states are Cerro Largo, Colonia, Paysandú, Rocha, Salto,
we use a population density of 0.01 people per square kilometer in the Tacuarembo, and Treinta y Tres, and we assign them a population density
mountain areas. For the coastal regions, we apply a population density of of 0.05 people per square kilometer. Finally, the remaining three states of
30.88 people per square kilometer, so that we generate the total population Durazno, Soriano, Río Negro were more heavily settled by peoples such as
of 1 million estimated by Denevan (1992). Using estimates for the area of the Yaros, Chaná, and Charruas, and we assign them a weighted average of
each state belonging to each region, we estimate population density in each population density estimated for Entre Ríos (Argentina), where the weights
state. are increasing in the area closer to this region.
C13. Paraguay C17. Venezuela
Estimates of the total population for Paraguay, Uruguay, and the south Denevan (1992) presents estimates for the total precontact population
of Brazil in Denevan (1992) imply a population density of 0.9 people per of Venezuela and gives precontact population densities for the Orinoco
square kilometer. We use this estimate and estimates for neighboring regions llanos (1.3 people per square kilometer), Amazon Basin (0.2), and Guiana
GOOD, BAD, AND UGLY COLONIAL ACTIVITIES 461
Highlands (less than 0.5 people per square kilometer, we use 0.4). In order for the same habitat in Colombia; and (c) the Caribbean (the Dependencias
to get estimates for the other regions of Venezuela, first we use estimates Federales region): we use estimates from Denevan for the Caribbean islands.
available from other countries with similar habitats and native groups in the Finally, we estimate population density for the coastal ranges and the eastern
region (in particular, from north and east Colombia and the Caribbean) in Andes by choosing a precontact population density that matches the total
the following way: (a) the Caribbean coast: we use estimates for the same population of about 1 million people for Venezuela, as presented in Denevan
habitat in the Colombian Caribbean coast; (b) the Selva: we use estimates (1992, table 00.1).