Pre-Colonial Institutions and Socioeconomic
Development: The Case of Latin America.
Luis Angeles and Aldo Elizaldey
January 14, 2015
Abstract
We study the e¤ects of pre-colonial institutions on present-day socioeconomic outcomes for Latin America. Our thesis is that more advanced pre-colonial political institutions relate to better socioeconomic
outcomes today - principally, but not only, through their e¤ects on the
Amerindian population. We test our thesis with a dataset of 324 subnational administrative units covering all mainland Latin American
countries. Our extensive range of controls covers factors such as geography, natural resources, colonial activities and pre-colonial characteristics - plus, crucially, country …xed e¤ects. Results strongly support
our thesis.
1
Introduction
Over the last two decades, the economics literature searching for the ultimate
drivers of the wealth and poverty of nations has given considerable attention
to the role of institutions. Taking the theoretical discussions of Douglass
North as a starting point (North 1981, 1990), the literature has progressed
mainly along empirical lines and produced an impressive array of supporting
Economics, Adam Smith Business School, University of Glasgow. Glasgow G12 8QQ,
UK. Email: [email protected].
y
Economics, Adam Smith Business School, University of Glasgow. Glasgow G12 8QQ,
UK. Email:
1
evidence: see Hall and Jones (1999) and Acemoglu et al. (2001) for two
important early contributions, and Acemoglu et al. (2005) for a review of
the literature.
As has been noted before, most of this literature focuses on the role of
colonial institutions: the institutional package that European colonial powers put in place throughout the world between the 16th and 20th centuries.
There are indeed good reasons for focusing on the impact of colonialism on
institutional development in countries outside Europe. For a start, colonialism was often a deeply disrupting process that radically modi…ed the way
societies were structured. Factors such as the content and direction of trade,
the nature of taxation, and the operation of the law were always a¤ected.
Furthermore, colonialism reached the vast majority of countries we now call
the Developing World (plus a few countries from today’s Developed World),
making it an ideal source of exogenous variation of institutional quality in
cross-country comparisons.
While the importance of colonialism is well-recognized in this context, recent research has progressively uncovered a large role for pre-colonial factors
as well. Most of this research has focused on Sub-Saharan Africa, a continent where colonialism arrived late (most of the African interior was largely
unknown to Europeans until the 1880s), did not last very long (African decolonisation took o¤ during the late 1950s), and where, with the exception
of South Africa and its neighbours, European settlement was of very limited
importance. Under these conditions, it is perhaps not surprising that precolonial factors transcend the colonial period and a¤ect African societies to
this day.
A good example of this line of research is Nunn (2008), who shows how
the intensity of the African slave trade over the pre-colonial period is negatively related to current levels of income per capita. Nunn and Wantchekon
(2011) expand on this by uncovering a relationship between the intensity of
the slave trade and present-day levels of interpersonal trust. Also of relevance, Huillery (2011) argues that the attitudes of pre-colonial African states
2
towards Europeans have an in‡uence on current development outcomes as
colonizers invested more in the areas where Africans were less hostile.
Turning to pre-colonial political institutions, Gennaioli and Rainer (2007)
and Michalopoulos and Papaioannu (2013) uncover a positive link between
the institutional characteristics of African ethnic groups, as prevalent in
the pre-colonial period, and present-day measures of socioeconomic development. Both papers measure institutions using the same categorical variable
that classi…es ethnic groups according to their level of political complexity
(a value of 0 denotes bands and tribes, while the maximum value of 4 is for
complex states). The data is taken from George Peter Murdock’s Ethnographic Atlas (1967), which codes more than 60 variables for 1267 ethnic
groups around the world1 . It is important to note that the Atlas attempts
to describe ethnic groups in the absence of foreign elements, in particular
colonial in‡uence. For the Latin American case this often implies a reliance
on historical sources in addition to direct anthropological evidence.
Gennaioli and Rainer (2007) use African nations as their unit of analysis and uncover a positive relationship between pre-colonial institutions and
measures of health, education and other public goods. The main caveat of
this approach is that national factors other than those speci…cally controlled
for, and in particular colonial and post-colonial institutions, may be behind
the uncovered statistical relationship. To address this issue Michalopoulos
and Papaioannu (2013) use ethnic groups as their unit of analysis, and show
how pre-colonial institutions have an e¤ect on present-day economic development even after controlling for all national characteristics with the use of
country …xed e¤ects. In subsequent work, Michalopoulos and Papaioannu
(2014) also show how African national institutions, that is, institutions put
in place by colonial and post-colonial forces, actually exert little or no in‡uence in areas far away from the capital city - a result that further emphasizes
the importance of pre-colonial factors.
1
The Ethnographic Atlas has been expanded and updated several times since its publication. The current version, used in this paper, is due to Gray (1999) and can be accessed
at http://intersci.ss.uci.edu/wiki/index.php/Ethnographic_Atlas
3
This paper contributes to the literature by analysing the role of precolonial institutions on present-day socioeconomic outcomes for Latin America. To the best of our knowledge, this is the …rst attempt of this kind for
the Latin American case. Because controlling for colonial and post-colonial
institutions by means of country …xed e¤ects is an important requirement,
we do not follow Gennaioli and Rainer (2007) in using the nation as the unit
of analysis. The approach of Michalopoulos and Papaioannu (2013), however, is not suitable for Latin America since a clear demarcation between the
areas inhabited by di¤erent ethnic groups is not possible there. Instead, we
take as our unit of analysis the largest administrative divisions of each country below the national level, which we refer to as "states".2 Our data thus
covers 324 states from 17 Latin American countries, and allows for the use
of country-speci…c …xed e¤ects. Among this cross-section of states we uncover a robust and statistically signi…cant relationship between pre-colonial
institutions and present-day measures of education, health, and economic
development. The relationship remains in place after controlling for each
state’s geographic characteristics, and a battery of alternative colonial and
pre-colonial determinants of economic success.
Other than the papers already mentioned, this paper also relates to a
growing literature stressing the deep historical roots of long-run development. For instance, Bockstette et al. (2002) show how regions with a longer
history of statehood in the two millennia up to the year 1950 are richer and
grow faster today. And Comin et al. (2010) demonstrate that current levels
of income per capita are strongly correlated with technological advancement
500, 2000 and even 3000 years ago. Looking beyond economic development,
the literature has also identi…ed a number of factors exerting an in‡uence
on social outcomes over the very long run: thus societies that traditionally
practiced plough-based agriculture are characterized by less equal gender
norms today (Alesina et al. 2013), having ancestors from herding communities predicts present-day homicide rates in southern areas of the United
2
The actual name given to these administrative divisions changes from country to
country: provincias in Argentina, departamentos in Bolivia, regiones in Chile, estados in
Mexico, and so on.
4
States (Grosjean, forthcoming), political independence at the city level during the Middle Ages predicts levels of social capital in present-day Italy
(Guiso et al. 2013), and German towns where Jewish pogroms took place
during the Black Death (1348-1350) were more likely to act violently against
Jews during the 20th century and vote for the Nazi party (Voigtlander and
Voth 2012). In a similar spirit, we uncover historical forces that in‡uence
socioeconomic outcomes over a span of several centuries.
The rest of the paper is organized as follows. Section 2 o¤ers some historical background in order to place the research question in its context. In
particular, it makes the case for pre-colonial institutions having an in‡uence
up to our days in Latin America. Section 3 discusses the data and presents
our empirical methodology. Section 4 presents our baseline results and extends into a number of robustness checks. Section 5, …nally, o¤ers some
concluding remarks.
2
Historical overview and main hypothesis
On the face of it, it is perhaps understandable that the research e¤ort on
the role of pre-colonial institutions has so far ignored the case of Latin
America. Indeed, there is little doubt that colonialism cut much deeper
in Latin America than in Africa - or almost any other region in the world
for that matter. It is not just that colonialism lasted far longer, about
three centuries for most Latin American nations and even longer for a few
Caribbean ones. More important, only in the Americas and in some of the
Paci…c islands did the European conquest lead to a radical transformation of
the ethnic structure of the population. Su¤ering the consequences of a new
disease environment, the aboriginal population of the Americas (henceforth
Amerindians) was decimated over the hundred years or so following …rst
contact with Europeans. In its place, a society of European descendants,
mixed-race mestizos, Amerindians and Africans, these last ones imported
as slave labour, took over. Europeans moved to the Americas permanently,
and their descendants constitute the economic and political elite of most
5
Latin American nations to this day.
Under these circumstances, one would be excused to think that Latin
America constitutes a prime example of how colonial factors determine institutional structures to the detriment of everything else. Only in the Americas
can we observe colonialism wiping out not only the established institutional
order but even the very people who held that order. The rupture from the
pre-colonial world was radical, the chances for pre-colonial factors to survive
into the present-day appear, at …rst sight, very limited.
And yet, things start to look di¤erent under closer inspection. While
Amerindians remained at the fringes of the economic and political power
structures, they played a crucial role as the main source of labour, and
therefore principal factor of production, in two of the most important sectors
of the colonial economy: mining and agricultural production for the local
market.3 Amerindian labour was the main source of wealth for Spanish
settlers in the Americas, as best summarized by the aphorism "Sin Indios
no hay Indias" ("Without Indians there is no Indies") - attributed to 16th
century Spanish settlers when defending the granting of rights over Indian
labour against accusations by the Crown of excessive exploitation.
The extraction of this Amerindian labour relied on the use of aboriginal
structures of power and organization. While the growing class of mestizos
lived in towns and cities and collaborated closely with the Spanish elites,
Amerindians by and large retired to their rural communities where they lived
a separate cultural life from the rest of society. Spanish governors referred
to this network of Indian villages, where no Europeans lived permanently,
as the "Republic of Indians" - a name that reveals much about the degree of
autonomy granted to these communities in their internal a¤airs. Indian villages were compelled to pay taxes and supply tribute in the form of labour
for mining, public works and, during the early phase of the colony, agricultural production through a number of schemes such as the encomienda,
3
Agricultural production for the export market, with crops such as sugar cane, tobacco
and cotton, employed African slave labour. See Angeles (2013) for an analysis of the
factors determining the ‡ow of slaves to the Americas.
6
repartimiento or mita. The delivery of labour and taxes was organized by
local headmen and leaders, who enjoyed privileges such as the private ownership of land and exception from taxation. In this way, as James Lang put
it, "The Spanish enterprise in the New World rested on an indigenous social
order" (Lang 1975, p. 7).
After bottoming out in the early to mid-17th century, Amerindian populations started to recover all along the continent from the early 18th century
onwards (Burkholder and Johnson 1998, pp. 107-110). The rise of the large
agricultural estate (hacienda) during the late colonial period and through
the 19th century meant that many Amerindians left their communities to
…nd permanent work (and often debt bondage) in them. On the other hand,
the 20th century brought a number of revolutionary movements and leftleaning governments which aimed at redistributing land and in so doing
reversed the ‡ow of Amerindians out of their communities (most notably
the Mexican revolution of 1910 and the Bolivian revolution of 1952). These
ups and downs notwithstanding, during …ve centuries of colonial and postcolonial history the Amerindian rural community has remained a permanent
element of Latin American countries. Its existence ensured the preservation
of Amerindian languages, cultural characteristics and many institutional elements up to the present day.
The above historical overview applies to most Amerindian groups throughout the Americas - only the most remote groups such as the tropical forest
dwellers of the Amazon managed to remain outside the in‡uence of Europeans well into the 20th century. Within this large universe of Amerindian
cultures, large institutional di¤erences were in place. At the time of …rst contact with Europeans, Amerindian groups varied greatly in terms of political
structure and institutional complexity: from the multi-layered bureaucracy
administering the vast Inca Empire to the numerous small chiefdoms with
no political organization beyond the village level, passing through intermediate forms of political complexity such as the confederacies of villages and
city-states of Mesoamerica.
7
The central hypothesis of this paper is that these institutional di¤erences,
preserved throughout the colonial period in Amerindian rural communities,
continue to exert an in‡uence on socioeconomic development up to this day.
As has been shown for the case of Africa, we hypothesize that higher levels of
pre-colonial institutional development are associated with better outcomes
in areas such as education, health, and economic well-being. Several mechanisms have been advanced to explain this link in the literature. In the
context of Latin America, we would emphasize the following ones:
i) Ethnic groups with experience of large-scale political organization were
in a better position to ensure the delivery of locally-produced public
goods such as education and public health. They would also have
local forms of legal resolution that did not involve colonial or national
courts (and would for that reason be more e¢ cient).
ii) Institutionally-advanced groups found it easier to learn new techniques
and modes of production, and to integrate themselves into the colonial
and post-colonial economic system. For instance, experience with markets was much more prevalent among Aztecs and Incas than among
peoples of the Amazon.
iii) Institutionally-advanced groups were able to organize themselves and
defend their interests, including claims to land and other resources, in
front of colonial and post-colonial governments.
iv) A higher level of pre-colonial institutional development may result in
more accountability of local chiefs, in particular if some forms of political organization survived beyond the village level which would held
village leaders accountable towards their ethnic nations. This mechanism is emphasized by Gennaioli and Rainer (2007) for the case of
Africa.
The above explanations have in common that they all describe a positive link between the pre-colonial institutions of an ethnic group and the
8
subsequent socioeconomic success of that same group. In addition to them,
and of particular importance for Latin America, we may hypothesize that
pre-colonial institutions may bene…t the non-Amerindian population as well.
This would happen if higher levels of institutional development among Amerindians foster cooperation and trade between them and the rest of the population, leading ultimately to a more prosperous society. It will be useful to
keep this in mind in what follows as our unit of analysis is the sub-national
state which, in all cases, has a mixture of Amerindian and non-Amerindian
population.
3
Data and methodology
We follow Gennaioli and Rainer (2007) and Michalopoulos and Papaioannu
(2013)) in using Murdock’s (1967) index of "Jurisdictional Hierarchy beyond the local community level" as our measure of pre-colonial institutional
complexity. The variable takes discrete values between 0 and 4, where the
value represents the number of organizational levels above the local community. In practice, Murdock assigns a value of 0 to bands and tribes with
no political organization beyond the village level. Chiefdoms that comprise a few villages or a single city-state would be assigned a value of 1.
Large chiefdoms with many cities and confederations of city-states would
receive a value of 2. Finally, 3 and 4 are reserved for states with several
levels of intermediate bureaucracy between its ruler and the local community (provinces, municipalities and so on). These categories are somewhat
related to the standard classi…cation of political complexity in anthropological studies, as …rst formulated by Elman Service, which classi…es societies
into bands, tribes, chiefdoms and states (Service 1971). As discussed by
Diamond (1997), the level of political complexity is closely related to technological advancement, which is needed in order to support an ever larger
class of non-food producers.
For the Americas, the only pre-colonial group that achieves the maximum value of 4 in Murdock’s classi…cation is the Incas. Indeed, the Inca
9
Empire is well-recognized as the most sophisticated political and administrative structure developed in the Western Hemisphere before the European
conquest (Burkholder and Johnson 1998, p. 19)4 . Its northern counterpart,
the Aztec Empire of central Mexico, is only assigned a value of 2 on Murdock’s scale, together with a few other confederacies of city-states such as
the Muisca of central Colombia or the Zapotecs of southern Mexico.5 Most
other Amerindian groups are assigned a value of 0 or 1.
As in Michalopoulos and Papaioannu (2013), we chose a unit of analysis below the national level in order to control for country-speci…c factors.
In particular, we wish to control for colonial and post-colonial institutional
factors which manifest themselves mainly at the national level. We do not,
however, follow Michalopoulos and Papaioannu (2013) in using ethnic groups
as the unit of analysis since this requires a clear demarcation of the geographic boundaries of each ethnic group. While such demarcation may be
credibly established for sub-Saharan Africa, all Latin American cities and
most of its countryside are in fact shared between the population of European and mixed origin and one or more ethnic groups. Thus, while the
geographic area where a particular ethnic group is found may be established
from sources such as the "Geo-referencing of ethnic groups dataset" (Weidman et al. 2010), this would not mean, for the Latin American case, that the
area in question is inhabited exclusively by that ethnic group. Conversely,
geographic areas which are not part of an ethnic group’s "homeland" may
be characterized by a signi…cant minority from the ethnic group in question.
4
We also assign the value of 4 to the Aymaras, a large Amerindian group which was
part of the Inca empire. The Aymaras are not assigned a value of Jurisdictional Hierarchy
in Murdock (1967). Our results are not dependent on this choice.
5
This may seem surprising, given that the Aztec Empire rivalled the Incas along several dimensions such as total population, military capacity, and monumental architecture.
There were, however, important di¤erences in political structure. The Aztec rule over
central Mexico has been described as hegemonic or indirect. Conquered kingdoms remained independent in all internal matters, their rulers were typically not removed, and
representatives of the Aztec Emperor, such as provincial governors, were largely absent.
In short, territories were not remodelled according to some Imperial structure but left to
be ruled as before with the provision that tribute must be paid to the Aztec overlords.
The Aztec Empire was thus a confederacy of city-states with a vast amount of tributary
kingdoms.
10
Our strategy thus consists of using the largest sub-national administrative division of each country - what we call states - as the unit of analysis. This, however, introduces a di¢ culty which is not present in the work
of Michalopoulos and Papaioannu (2013): states contain a mixture of different Amerindian groups plus an important (in many cases dominating)
non-Amerindian population, and for that reason a single value of the index
of institutional complexity cannot be assigned to them. An alternative approach must then be followed to re‡ect the diverse institutional heritage of
each of these geographical units.
One possibility would be to follow Gennaioli and Rainer (2007) and construct a population-weighted average of the index of Jurisdictional Hierarchy
among all ethnic groups within the state, Amerindians or not. Amerindian
groups would be assigned a value from the Ethnographic Atlas, while nonAmerindian groups would have to be assigned a value based on the institutional complexity of the society they originate from. Naturally, European
descendants and mestizos would be assigned a value of 4 as their society of
origin would be the Spanish Empire. For the population of African origin
two approaches are possible. One would be to use the value of Jurisdictional
Hierarchy of the African ethnic group from which they originate. As such
detailed information is essentially unavailable, a second and perhaps superior approach would be to consider that Africans were employed as slaves
throughout the Americas, and as such forbidden to form any kind of social
or political grouping. Having been extirpated from their original societies,
Africans would have lost their original institutional heritage and acquired
those of a slave. In that case, a value of 0 would be assigned to them. Finally,
this weighted index could be used as an explanatory factor of socioeconomic
development in present times.
While potentially informative, the above approach su¤ers from a number of caveats. First, one must wonder whether putting European and
Amerindian institutions on the same scale is really appropriate. After all,
the Ethnographic Atlas explicitly states that modern Western societies are
not the focus of its study and Elman Service’s four stages of political or11
ganization were developed with pre-industrial societies in mind. The above
approach would rank the institutions of early modern Spain at the same
level as those of the Inca Empire, an assertion which can be easily called
into question. To be sure, the Inca state could rival its European counterparts in areas such as land and irrigation policy; but it was still a state with
no written records, no legal codes and hardly any constraints on the power
of the ruler. It was far more similar to the kingdoms of ancient Egypt and
Mesopotamia than to the states of early modern Europe, and one would
wish to make this distinction count in our empirical analysis.
Even more important, the construction of an index of institutional complexity among all ethnic groups con‡ates pre-colonial institutional factors
with colonial and post-colonial ones. Indeed, the resulting index would be
a weighted average of the institutional complexity brought about by Europeans with the institutional complexity in place before the conquest. As
such, its coe¢ cient could not be interpreted unequivocally as re‡ecting the
e¤ect of pre-colonial institutions on present-day outcomes. The approach
of Gennaioli and Rainer (2007), in de…nitive, is not adequate for the Latin
American case due to the importance of the non-autochthonous population.6
With the above in mind, we formulate an alternative strategy in which
pre-colonial institutional factors are kept separate from colonial and postcolonial ones. We assume that colonial and post-colonial institutions manifest themselves at the national level. In other words, all states within a given
country are a¤ected by these institutions in the same manner. This seems
reasonable as much of the colonial and post-colonial institutional package is
embedded in each countries’laws and constitution, which apply uniformly
in all sub-national units. The e¤ect of these institutions may be large, but
it can be factored out with the use of country-speci…c …xed e¤ects.
In addition to this, we posit that each subnational state is also a¤ected
by pre-colonial institutions inherited through its local Amerindian popula6
In this respect, we note that South Africa, the only African country with a large
share of European descendants in its population, is excluded from the baseline analysis in
Gennaioli and Rainer (2007).
12
tion. The e¤ect of these pre-colonial institutions on current outcomes will
depend on two magnitudes. First, the importance of Amerindians in the
local population, as measured by their population share within each state.
And second, the degree of complexity of these institutions, as measured
by the population-weighted average of Jurisdictional Hierarchy among all
Amerindian groups within the state. We refer to this last measure as the index of Pre-Colonial Institutions, with larger values denoting more advanced
institutions in pre-colonial times.
With this variable at hand, the econometric speci…cation we use to investigate the role of pre-colonial institutions in Latin America may be stated
as follows:
Ys;c =
c
+ P CIs;c + AmP ops;c + Xs;c +
s;c
(1)
In equation (1), Ys;c is an outcome variable such as a measure of schooling, health or economic well-being. Subscript s denotes sub-national states,
subscript c denotes countries, and
c
is a set of country-speci…c …xed ef-
fects. P CIs;c is the index of Pre-Colonial Institutions described above and
AmP ops;c the share of Amerindians in the total population, both de…ned
at the state level.
Finally, Xs;c is a set of variables controlling for state
characteristics such as geography, location, and a number of colonial and
pre-colonial factors potentially a¤ecting outcomes.
An important point is that AmP ops;c will probably matter for socioeconomic outcomes for reasons other than the transmission of pre-colonial
institutional characteristics. In particular, a larger share of Amerindians
may be linked with lower socioeconomic outcomes if Amerindians su¤er
from ethnic discrimination, receive less than their share of resources from
the national government, or have di¢ culty integrating into non-Amerindian
society. A good deal of contemporary evidence suggest that such is indeed the case.7 Thus, the sign and statistical signi…cance of coe¢ cient
may not be interpreted as evidence in favour or against the importance of
7
See Psacharopoulos and Patrinos (1994) for a detailed analysis.
13
pre-colonial institutions, as it re‡ects more than one socioeconomic phenomenon. A positive coe¢ cient
, on the other hand, indicates that a higher
level of pre-colonial political complexity is associated with better outcomes
after factoring out any negative e¤ect due to the importance of Amerindians
in the population. We will interpret such result as evidence of the e¤ect of
pre-colonial institutions on contemporary socioeconomic development.
The construction of the index of Pre-Colonial Institutions proceeds in
two steps. First, we must collect detailed ethnic population data at the
state level for all countries under study. Second, the ethnic groups identi…ed
in the population statistics need to be matched to the ethnic groups from
the Ethnographic Atlas, so that a value of Jurisdictional Hierarchy may be
assigned to them.
For the …rst step, we accessed the population statistics collected by the
di¤erent Latin American national statistical agencies by means of surveys
and censuses. These data sources are not in a common format and had to be
accessed individually, often by contacting the di¤erent statistical agencies
directly. It is worth noting that in all cases the information on the ethnic
group of the person being surveyed is based on a subjective assessment by
the person herself or, in some cases, the interviewer. Typically, a question
such as "To which ethnic group do you belong?" is asked. The question
cannot be answered based on biological or genetic criteria, but on cultural
ones. Thus, answers to the question will vary over time and are sensible to
the speci…c wording being used.
The resulting dataset gives the population shares of all Amerindian
groups residing in each state for all Latin American countries in South
America, Central America and Mexico - a total of 17 countries.8 We do not
8
For Brazil the population shares of di¤erent Amerindian groups is only available at the
level of regions (groups of 3 to 9 states). We assign to each Brazilian state the population
shares of the region it belongs to. For Argentina the data is available at the state level
but gives only a partial breakdown, with the population of only the main Amerindian
groups of each state given. We complete the missing data for Argentina using national
totals for each group and assumptions about the distribution of each group outside the
14
include Latin American countries from the Caribbean as the Amerindian
population of these countries disappeared completely following the colonial
conquest. Table 1 presents the list of the 17 Latin American countries under
study together with their total population (column 2) and total Amerindian
population as a share of total population (column 3).
[Table 1]
As table 1 shows, the share of Amerindians in total population for Latin
America as a whole is estimated at 5%. This average hides large di¤erences,
as Amerindians represent from 0.2% of the population in El Salvador to 41%
of the population in Bolivia. Moreover, the variability in the importance of
the Amerindian population is reproduced within each Latin American nation. Consider as an example Chile, where 4.6% of the population identi…es
itself as Amerindian. The ethnic composition of Chilean sub-national states
varies from 0.8% for the region of Coquimbo to 23.4% for the region of Araucania. Table 1 also reports in its last column the number of sub-national
states within each country, which varies from 7 for Costa Rica to 33 for
Colombia.
With the required population statistics at hand, we matched the ethnic
groups listed under the di¤erent censuses with the ethnic groups described in
the Ethnographic Atlas. This is not always a straightforward process as some
ethnic groups are listed under two di¤erent names in these two sources. We
used a diversity of additional material in order to make sure that as many
ethnic groups as possible were matched - please refer to table A1 in the
Appendix for details. In de…nitive, we were able to match 102 Amerindian
ethnic groups from our census data to the Atlas.
The Amerindian groups for which a value of the index of Jurisdictional
states where they are most numerous. For Uruguay we do not have data on di¤erent
Amerindian groups, only the population share of all Amerindians in each state. However,
we are able to compute the index of Pre-Colonial Institutions for Uruguayan states as we
know that all Amerindian groups in Uruguay have a value of zero in Murdock’s measure
of Jurisdictional Hierarchy. For all other 14 countries we have a complete dataset giving
population shares for all Amerindian groups at the state level.
15
Hierarchy could be assigned represent 71% of the total Amerindian population of Latin America - albeit this percentage varies signi…cantly from
country to country (see column 4 in table 1). The fact that almost 30% of
the Amerindian population could not be matched is to be expected given
that the Ethnographic Atlas does not o¤er an exhaustive list of all groups
but rather a survey of the groups for which anthropological work is available. For the Amerindian groups that could not be matched, we assign the
minimum value of Jurisdictional Hierarchy under the assumption that small
and less organized groups were more likely to be overlooked by anthropologists. The assumption is supported by the fact that all groups with a value
of Jurisdictional Hierarchy of 1 or higher were matched to our data. As a
robustness check, we also experiment assigning these groups a value equal
to the average of all other Amerindian groups within the same state.
Turning to the rest of our dataset, we consider a total of eight alternative
dependent variables and eleven control variables in our baseline analysis, all
de…ned at the state level. Table 2 reports summary statistics for these
variables, while their sources and precise de…nitions can be found in table
A2 in the Appendix. Extensions to our baseline analysis will call for the
introduction of additional controls, and these are discussed in detail when
used in the next section.
[Table 2]
4
4.1
Empirical analysis
Baseline results
We begin our statistical analysis by considering di¤erent versions of equation (1) where, initially, we do not control for country-speci…c …xed e¤ects
and do not include state characteristics other than the index of Pre-Colonial
Institutions and the share of Amerindians in the total population. Other
factors are added progressively in order to appreciate their e¤ect on our coe¢ cient of interest. The dependent variable we select for this initial analysis
16
is the percentage of the population who …nished secondary schooling, in
logarithmic form.
The …rst column of table 3 thus reports the results of our most simple
regression. As expected, the coe¢ cient on the share of Amerindians in total
population is negative and statistically signi…cant, indicating that states
with a larger Amerindian population tend to have lower levels of secondary
education. A number of mechanisms may explain this correlation: the central
government may invest less in the education of areas where Amerindians
live, Amerindians may …nd schooling more di¢ cult as it is not o¤ered in the
language they speak at home, or Amerindian children may feel ethnically
discriminated at school. While this negative correlation between Amerindian
population and educational achievement is not new, table 1 also uncovers
a previously unexplored and positive e¤ect related to the characteristics of
the Amerindian population. The coe¢ cient on Pre-Colonial Institutions
is positive and statistically signi…cant at the 1% level: states where the
Amerindian population had higher levels of institutional complexity tend to
be characterized by higher educational achievement today.
The next two columns of table 3 start to control for national characteristics by adding a rule of law index (column 2) or national GDP per capita
in logarithmic form (column 3). In both cases these national characteristics
have the expected positive sign, but in both cases the coe¢ cient for PreColonial Institutions changes only marginally and its magnitude actually increases. Column 4 then controls for all country-speci…c characteristics with
the inclusion of …xed e¤ects. As discussed previously, we expect …xed e¤ects
to control not just for the overall economic development of each country but
also for all institutional factors which manifest themselves at the national
level - in other words for colonial and post-colonial institutions. As country …xed e¤ects absorb a sizeable share of the variation in the dependent
variable (the R2 coe¢ cient increasing from 0.45 to 0.77), the coe¢ cient on
Pre-Colonial Institutions falls in magnitude from 0.162 in the …rst column
to 0.089 in the fourth one - but remains statistically signi…cant at the 1%
level. Thus, pre-colonial institutions predict socioeconomic development in
17
Latin America even after removing all between-country variation. As …xed
e¤ects are particularly important in this context all subsequent regressions
in the paper will include them.
Columns 5 to 7 add a battery of state characteristics whose omission so
far may well be producing a bias in our estimates. First, in column 5, we
control for the population density of each state - as the provision of education may be more costly in less densely settled territories. Next, in column
6, we include a vast array of geographic and locational characteristics. These
control for aspects such as the state’s climate (latitude, altitude, temperature), its size (land area), its capacity to pro…t from maritime transportation
(distance to the coast, dummy for landlocked states) and its capacity to in‡uence the main seat of power (distance to the capital). Finally, column
7 also controls for the presence of natural resources, with dummy variables
indicating the existence of oil or gas, gold or silver, and any other types of
mines.
The main result of these three columns is that the coe¢ cient on PreColonial Institutions remains statistically signi…cant at the 1% level in all
cases and its magnitude is not much a¤ected. When all controls are included
in column 7, the coe¢ cient takes a value of 0.085. As our dependent variable
is measured in logarithmic form, this coe¢ cient indicates that an increase in
the average level of institutional complexity of the Amerindian population
by 1 unit is associated with an increase in secondary education achievement
of 8.5%. This is a large e¤ect: passing from a pre-colonial population of
tribesmen to one of multi-city chiefdoms (2 units) would lead to an increase
of 17%.
Turning to the state-level control variables, population density has a positive and statistically signi…cant e¤ect on education in all regressions where
it is included - in accordance with our priors. On a similar vein, altitude has
also a negative e¤ect on educational levels as it may increase transportation costs. Interestingly, however, distance to the capital is associated with
higher levels of education. Finally, a tropical climate correlates with lower
18
levels of education, an e¤ect which is captured by the positive coe¢ cient on
absolute latitude.
The results of table 3 may be reproduced over a large array of socioeconomic indicators, as shown in table 4. This table takes as its baseline the
regression reported in the last column of table 3, which controls for country
…xed e¤ects and all state characteristics considered so far, and changes the
dependent variable. We consider one indicator of public health (infant mortality rate), three indicators of education (percent of the population who
completed primary education, who completed secondary education, and average years of schooling), two indicators of economic well-being (percent of
the population with access to drinking water, percent with access to electricity), and two indicators of overall economic development (GDP per capita
and poverty rates). Remarkably, our index of Pre-Colonial Institutions has
a positive and statistically signi…cant e¤ect on all of them - in all but one
case at the 1% level of signi…cance.9
The e¤ect of Pre-Colonial Institutions is not only statistically signi…cant,
the magnitude of the e¤ect is also large. Since all dependent variables are
used in logarithmic form, coe¢ cients may be interpreted directly as semielasticities. Remarkably, the largest e¤ects are observed for our measures of
overall economic development. A 1-unit increase in the index of Pre-Colonial
Institutions is associated with a 20% increase in GDP per capita and a 13%
decrease in the poverty rate. The e¤ects for all other dependent variables
are in the 3-8% range for a 1-unit increase, in all cases a sizeable change.
4.2
Controlling for colonial activities
While our results so far control for a vast array of geographic and locational
factors at the state level, and for institutional factors at the national level,
other historical factors not yet considered could be a source of bias. In particular, pre-colonial institutional development may be related to the colonial
9
By "positive e¤ect" we mean an improvement in socioeconomic development. Thus,
the coe¢ cient is positive for all measures of education, economic well-being and GDP per
capita, and negative for infant mortality and poverty rates.
19
activities put in place following the conquest. Indeed, the territories of the
most advanced pre-colonial civilizations - central and southern Mexico, the
Andes - were also the source of most Amerindian labour. The availability of
this labour made possible a range of economic activities during the colony,
most notably silver mining and the agricultural latifundia. If these activities
then have an e¤ect on present-day outcomes, pre-colonial institutions would
be correlated with socioeconomic development but the causal mechanism
would work through the colonizing process.
To test for this alternative explanation we take advantage of the work of
Bruhn and Gallego (2012), who investigate the role of colonial activities on
economic development in Latin America using states as the unit of analysis.
They classify states into four mutually exclusive groups according to the
main economic activity taking place in their territory during the colonial
period. These four groups are:
a) Mining. In particular the gold mines of Brazil, the silver mines of Mexico,
Peru and Bolivia, and the associated mines producing mercury for the
process of silver extraction through amalgamation.
b) Plantations. Places dedicated to the cultivation of high-value cash crops
for the export market, in particular sugar, tobacco and cotton. Plantations relied essentially on slave labour.
c) Other colonial activities. Places where the dominant economic activity
was agricultural production for the local markets (from Amerindian
lands or from latifundia) and industry.
d) No colonial activities. Places where the colonial state had marginal or
no in‡uence, like remote parts of the Amazonian rainforest and the
extreme south of Argentina and Chile.10
10
Bruhn and Gallego (2012) don’t use these four groups in their analysis. Instead, they
combine the information on the type of economic activity in each state with data on precolonial population density to produce a classi…cation into three types of colonial activities
which they refer to as "bad", "good" and "ugly". We don’t follow their approach as it
incorporates value judgements as to what is believed to be "good" or "bad" (let alone
20
We incorporate dummy variables for the …rst three types of economic
activities, leaving the case of no colonial activities as our excluded category.
Results are reported in table 5, where all regressions control for country
…xed e¤ects and each state’s population density, geography, location and
natural resources.
Table 5 is strongly supportive of our thesis. Indeed, the index of PreColonial Institutions continues to have a positive and statistically signi…cant
e¤ect on the eight dependent variables we consider. The size of the coef…cients is not much a¤ected with respect to table 4, indicating that the
relationship between pre-colonial institutional development and present-day
outcomes is not mediated by the type of economic activity put in place
during the colony.
Turning to the e¤ects of colonial activities on present-day outcomes, table 5 gives us a mixed picture. The e¤ect seems clearest on overall measures
of economic development, as states associated with mining and plantation
agriculture have lower levels of GDP per capita and higher poverty rates than
states left untouched by the colonial economy. The e¤ect is also present for
states that developed other colonial activities, although the estimated coe¢ cients are smaller and less statistically signi…cant. This is in line with Bruhn
and Gallego (2012), who base most of their analysis on the e¤ects on GDP
per capita. For other measures of socioeconomic development, however, the
evidence is less conclusive. Areas where slave-based plantations were located
are indeed characterized by higher levels of infant mortality and lower levels
of education, and the relationship is statistically signi…cant. But no further
statistically signi…cant e¤ects are estimated for areas formerly dedicated to
mining or other colonial activities. Overall, however, colonial activities do
play a role in determining current development but their consideration does
not diminish the importance of pre-colonial institutions.
"ugly"). The classi…cation of colonial activities into mining, plantations, and others is, we
belive, much less likely to be a¤ected by our own beliefs.
21
4.3
Controlling for other pre-colonial characteristics
If the results so far clearly point towards a persistent role of pre-colonial
institutions on current socioeconomic development, one may still argue that
pre-colonial features other than institutional complexity may explain our
results. As we mentioned brie‡y, institutional complexity usually correlates
with economic development, and it is possible that richer pre-colonial societies were able to adapt better and take advantage of the new colonial
environment simply because of their wealth. Furthermore, the Ethnographic
Atlas provides a large array of cultural and economic practices of the societies it surveys. We are therefore in a position to control for a number
of pre-colonial characteristics other than the complexity of their political
structure - and we do so in what follows.
We start with overall economic development in pre-colonial times. Clearly,
measures of income per head are not available for this time period in the
Americas, but we may follow much of the relevant literature and rely on
estimates of population density as a proxy for overall economic development
(see, for instance, Acemoglu et al. 2002). The data on pre-colonial population density at the state level comes from Bruhn and Gallego (2012), and
table 6 adds this variable as an additional control to the regressions reported
in table 5.
Once again, the results are fully consistent with the thesis of this paper. The coe¢ cient of Pre-Colonial Institutions is hardly a¤ected by the
inclusion of this variable and remains statistically signi…cant for all dependent variables. Pre-colonial population density has a statistically signi…cant
in‡uence on years of schooling, secondary education, GDP per capita and
poverty rates; but not on the other four dependent variables we consider.
Its e¤ect is negative, in accordance with the "reversals of fortune" thesis
of Acemoglu et al. (2002), whereby richer pre-colonial areas end up being
poorer after the colonial process.
In table 7 we take an additional step and control for nine social and economic characteristics of pre-colonial societies other than their institutional
22
complexity. These characteristics are the fraction of the population dedicated to gathering, hunting, …shing and agriculture; their typical pattern
of settlement (from fully nomadic to compact and permanent settlements);
their degree of class strati…cation; a dummy for the existence of slavery; a
dummy for the existence of elections in determining leader succession and,
…nally, a dummy for the existence of inheritance rules for property (see table
A3 in the Appendix for detailed de…nitions). For each of them, we proceed
as in the construction of our index of Pre-Colonial Institutions: we calculate
the population-weighted average among all Amerindian groups present in
the state.11 The …rst four measures, all relating to the economic activity of
the population, are included simultaneously in column 2.12 All other variables are included separately in the remaining columns of the table. The
regressions also control for the di¤erent colonial activities as in table 5 and
for pre-colonial population density as in table 6, besides all the state-speci…c
characteristics and country …xed e¤ects that have been included all along.
The dependent variable is the percentage of the population with secondary
education.
As table 7 makes clear, the inclusion of these additional pre-colonial
characteristics does not challenge the importance of pre-colonial institutions.
In all regressions the coe¢ cient on the index of Pre-Colonial Institutions
remains positive and statistically signi…cant at the 1% level. The magnitude
of the coe¢ cient is remarkably consistent, ‡uctuating closely around the
value of 0.075 in all but one case (column 2, where the coe¢ cient equals
0.099). Thus, the coe¢ cient is usually very similar to what is obtained
before any of these pre-colonial characteristics is controlled for (…rst column
of table 7). Accordingly, our thesis regarding the importance of pre-colonial
11
For the Amerindian groups that could not be matched to the Atlas we assign a value
equal to the average value of all other groups within the state. This is di¤erent from
what we did for the index of Pre-Colonial Institutions, as the variables considered here
are not necessarily related to social complexity, and could not be assumed to take the
lowest possible value. Uruguay is excluded from table 7 as we don’t have enough data to
calculate these additional variables for it.
12
These four variables do not sum up to 1, as a fraction of the population may be
counted in more than one of them, and sometimes in none of them.
23
institutions comes out reinforced.
Most of the additional pre-colonial characteristics considered turn out to
have no statistically signi…cant e¤ect on education. The exception relates to
the occupational variables, where the percentage of the population employed
in …shing is positively linked to education while the percentage employed
in agriculture comes out with a negative e¤ect. We may speculate that
Amerindian groups dedicated to farming were more likely to be exploited
as latifundia workers during the colony, and that …shing activity proxys for
proximity to rivers and the lower transportation costs they bring along.
The exercise of table 7 may be reproduced using the other seven dependent variables considered previously. While we do not report these results
for conciseness, we have carried them out and the importance of pre-colonial
institutions is never challenged. The sign and statistical signi…cance of precolonial institutions carries through for all seven alternative outcome variables and in all the speci…cations considered in table 7 (results are available
upon request).
We conclude our analysis in this section by noting that we have also run
all our regressions so far using a di¤erent assumption for the Amerindian
groups in our data that could not be matched to a group from the Atlas.
Instead of assigning them a value of 0 for the Jurisdictional Hierarchy index, we assign them a value equal to the average value of all matched groups
within the state. The results of this exercise continue to be strongly supportive of our thesis, as pre-colonial institutions continue to exert a positive and
statistically signi…cant e¤ect on most outcome variables and speci…cations
(results are available upon request).
4.4
Comparing rural and urban regions
As a …nal piece of evidence, we have gathered data allowing us to run separate regressions for the rural and urban regions of Latin American subnational states. While most of our data is not available at this level of
24
disaggregation, we were able to …nd separate values for the rural and urban
region of each state for four dependent variables (primary education, secondary education, access to drinking water, access to electricity) and for the
ethnic composition of the population, which allows us to calculate the percentage of Amerindians in the total population and to construct our index
of Pre-Colonial Institutions. All other control variables may be used in the
analysis, but their values do not change between the rural and urban area
of any given state13 .
The interest of this exercise is that, as we discussed previously, the importance of pre-colonial institutions ought to be far more marked in rural
areas. Amerindians may be numerous in urban areas, but by migrating to
them they enter a process of cultural assimilation within the dominant mestizo society. Amerindians no longer rely on their pre-colonial institutions
once they leave their rural communities, as a di¤erent set of institutional
arrangements is imposed upon them. If our hypothesis is correct, we should
…nd that the positive relationship between pre-colonial institutions and socioeconomic development is stronger among rural areas.
And indeed, the results clearly support this prior. Table 8 reproduces the
regressions of table 2, where the set of control variables includes country …xed
e¤ects, population density, and measures of natural resources, geography
and location. We consider the four dependent variables mentioned above,
and for each case run separate regressions using all rural areas or all urban
areas. As it turns out, the e¤ect of Pre-Colonial Institutions is positive and
statistically signi…cant for the four cases covering rural areas, but only for
two of the four cases covering urban areas (plus an additional case which
is marginally signi…cant). More important, the coe¢ cient on Pre-Colonial
Institutions is always far larger for the case of rural areas. For instance,
it takes a value of 0.090 for rural areas when the dependent variable is
secondary education as opposed to 0.017 for the corresponding regression
13
Argentina is omitted for this exercise, as there is no information on the distribution
of its Amerindian population between urban and rural areas.
25
using urban areas. When compared to the urban e¤ect, the e¤ect of precolonial institutions in rural areas is four times larger for primary education,
…ve to six times larger for secondary education and access to drinking water,
and as much as twenty times larger for access to electricity.
5
Concluding remarks
If one thing has been learned from the last two decades of research on economic development over the very long run it is that the past cannot be
easily cast aside. Every society builds on the successes and mistakes of its
predecessors, and inherits a set of rules and institutions that are usually
modi…ed only gradually. While this seems obviously true for the "winners"
of economic history, the European nations that colonized the world, it is
also the case for the "losers", those nations being colonized. What came out
of the colonizing process throughout the world was not a mirror image of
European society but a new reality where pre-colonial culture and institutions survived, often below a layer of o¢ cial or dominant culture. Of course,
these two layers interacted and modi…ed each other, but both of them ought
to be considered in the study of today’s developing countries.
This paper brings support to the above assertions, and adds to the substantive evidence already in place for the case of Africa. As our empirical
results show, Latin American pre-colonial institutions - and more precisely
the degree of political complexity - are powerful predictors of a large array
of measures of socioeconomic development. Several aspects render our evidence particularly convincing. First, our results are obtained controlling
for country …xed e¤ects; thus factoring out institutional factors playing a
role at the national level. Second, we introduce not just standard controls
for present-day factors such as geography and natural resources, but also
consider additional historical forces such as the type of economic activity
in place during the colony and the economic and social pro…le of the precolonial societies (besides political complexity). Finally, we show how the
in‡uence of pre-colonial institutions is far stronger in rural areas, which is
26
in accordance with the historical account we give for the transmission of
pre-colonial factors.
The present paper, together with the literature it contributes to, enhances our understanding of how developing countries got to where they are
now. Understanding this is important in its own right, but also increases
the chances of making the right decisions when considering where they head
to in the future.
27
References
Acemoglu, D., Johnson, S. and Robinson, J. A, 2001, The colonial origins
of comparative development: an empirical investigation, American Economic
Review 91 (5), 1369-1401.
Acemoglu, D., Johnson, S. and Robinson, J. A, 2002, Reversals of Fortune: Geography and Institutions in the making of the modern world income
distribution, Quarterly Journal of Economics.
Acemoglu, D., Johnson, S. and Robinson, J. A, 2005, Institutions as a
fundamental cause of long-run growth, in (eds.): Aghion, P. and Durlauf, S.
N., Handbook of Economic Growth, volume 1A, Elsevier.
Alesina, A., Giuliano, P. and Nunn, N., 2013, On the origins of gender
roles: women and the plough, Quarterly Journal of Economics 128 (2), 469530.
Angeles, L., 2013, On the Causes of the African Slave Trade, Kyklos 66
(1), 1-26.
Bockstette, V., Chanda, A. and Putterman, L., 2002, States and markets: the advantage of an early start, Journal of Economic Growth 7, 347369.
Bruhn, M. and Gallego, F. A., 2012, Good, bad, and ugly economic
activities: do they matter for economic development?, Review of Economics
and Statistics 94 (2), 433-461.
Burkholder, M. A. and Johnson, L. L., 1998, Colonial Latin America,
Third Edition, Oxford: Oxford University Press.
Comin, D., Easterly, W. and Gong, E., 2010, Was the wealth of nations
determined in 1000 BC?, American Economic Journal: Macroeconomics 2,
65-97.
Diamond, J., Guns, germs, and steel, London: W. W. Norton & Company.
Gennaioli, N and Rainer, I., 2007, The modern impact of precolonial
centralisation in Africa, Journal of Economic Growth 12, 185-234.
28
Gray, J. P., 1999, A corrected Ethnographic Atlas, World Cultures Journal 10 (1), 24-85.
Grosjean, P., forthcoming, A History of Violence: The Culture of Honor
and Homicide in the US South, Journal of the European Economic Association.
Guiso, L., Sapienza, P. and Zingales, L., 2013, Long-term Persistence,
EIEF working paper 23/13.
Hall, R. E. and Jones, C. I., 1999, Why do some countries produce so
much more output per worker than others?, Quarterly Journal of Economics
114, 83-116.
Huillery, E., 2011, The impact of European settlement within French
West Africa: Did pre-colonial prosperous areas fall behind?, Journal of
African Economies 20 (2), 263-311.
Lang, J., 1975, Conquest and Commerce: Spain and England in the
Americas. New York: Academic Press.
Michalopoulos, S. and Papaioannou, E., 2013, Pre-colonial ethnic institutions and contemporary African development, Econometrica 81 (1), 113152.
Michalopoulos, S. and Papaioannou, E., 2014, National institutions and
subnational development in Africa, Quarterly Journal of Economics 129,
151-213.
Murdock, G. P., 1967, Ethnographic Atlas: A Summary, Ethnology 6
(2), 109-236.
North, D. C., 1981, Structure and change in economic history, New York:
W. W. Norton & Co.
North, D. C., 1990, Institutions, institutional change, and economic performance, Cambridge: Cambridge University Press.
Nunn, N., 2008, The long-term e¤ects of Africa’s slave trades, Quarterly
Journal of Economics 123 (1), 139-176.
Nunn, N. and Wantchekon, L., 2011, The slave trade and the origins of
mistrust in Africa, American Economic Review 101 (7), 3221-3252.
Psacharopoulos, G. and Patrinos, H. A., 1994, Indigenous People and
Poverty in Latin America, Washington: The World Bank.
29
Service, E. R., 1971, Primitive social organization: an evolutionary perspective, 2nd edition, New York: Random House.
Voigtlander, N. and Voth, H.-J., 2012, Persecution perpetuated: the medieval origins of anti-semitic violence in Nazi Germany, Quarterly Journal
of Economics 127 (3), 1339-1392.
Weidmann, N. B., Rod, J. K. and Cederman, L.-E., 2010, Representing
ethnic groups in space: A new dataset, Journal of Peace Research 47 (4),
491-499.
30
Table 1
Latin American countries and their Amerindian population
Total populationa
Amerindian
population as % of
total population
Amerindian population
matched to Ethnographic Atlas
as % of total Amerindian
population
Number of
states
Argentina
Bolivia
Brazil
Chile
Colombia
Costa Rica
Ecuador
El Salvador
Guatemala
Honduras
Mexico
Nicaragua
Panama
Paraguay
Peru
Uruguay
Venezuela
40,117,096
10,059,856
190,755,799
15,116,385
41,174,853
4,301,712
14,451,115
57,44,113
11,237,196
6,076,885
103,263,388
5,483,447
3,405,813
5,163,198
27,412,157
3,285,877
27,225,775
0.028
0.41
0.004
0.046
0.034
0.016
0.07
0.002
0.39
0.063
0.057
0.08
0.12
0.017
0.15
0.02
0.028
0.25
0.84
0.38
0.96
0.47
0.26
0.42
0.15
0.62
0.96
0.77
0.27
0.29
0.55
0.96
1.0
0.79
24
9
27
13
33
7
24
14
8
18
32
17
12
18
25
19
24
TOTAL
514,274,665
0.05
0.71
324
Country
a
: Total population on the year in which the data was collected. All censuses take place between 2001 and 2012.
Table 2
Descriptive Statistics
Observations
Mean
Std. Dev.
Min
Max
Infant Mortality Rates
324
20.6
10.8
1.4
56.4
Years of Schooling
319
5.98
2.13
1.03
11.45
Primary School
324
0.81
0.16
0.29
0.96
Secondary School
324
0.41
0.15
0.05
0.75
Drinking Water
324
0.86
0.17
0.05
0.99
Electricity
324
0.84
0.19
0.03
0.99
Log GDP per capita
310
8.42
0.66
7.13
10.61
Log Poverty rates
272
3.01
0.98
0.21
4.4
324
0.57
1.05
0
3.99
Rule of Law
324
-0.54
0.58
-1.4
1.3
Countries’ Log PPP GDP per capita
324
8.96
0.41
8.13
9.48
Share of Total Ethnic Population
324
0.11
0.19
0.00001
0.96
Population Density
324
394.9
3407.41
0.13
58706.88
Latitude
324
16.02
10.73
0.015
54.33
Altitude (km.)
324
0.68
0.92
0
4.33
Temperature (Celsius)
319
20.72
5.28
4.7
27.77
Land area (sq. km.)
324
63815.63
151115.3
44
1559162
Landlocked dummy
324
0.54
0.49
0
1
Distance to capital (km.)
324
464.08
477.69
0
2559.34
Inverse distance to coast
320
0.89
0.1
0.54
0.99
Oil & Gas dummy
324
0.16
0.36
0
1
Gold & Silver dummy
324
0.12
0.32
0
1
Other mines dummy
324
0.23
0.42
0
1
Dependent variables:
Main regressor of interest :
Index of Pre-colonial Institutions
Controls variables (national level):
Controls variables (state level):
Table 3
Baseline results
Dependent variable: Percent of the population having completed Secondary education (in logs)
(1)
(2)
(3)
(4)
(5)
(6)
(7)
Pre-Colonial Institutions
0.162***
[0.0187]
0.166***
[0.0193]
0.169***
[0.0190]
0.0894***
[0.0199]
0.0751***
[0.0216]
0.0869***
[0.0208]
0.0851***
[0.0211]
Share of Amerindian population
-0.832***
[0.160]
-0.804***
[0.166]
-0.535***
[0.182]
-0.746***
[0.113]
-0.632***
[0.129]
-0.602***
[0.133]
-0.588***
[0.133]
0.0279***
[0.00936]
0.0504***
[0.0131]
0.0517***
[0.0135]
Latitude
0.0113***
[0.00215]
0.0115***
[0.00213]
Altitude (km.)
-0.0371**
[0.0181]
-0.0434**
[0.0182]
Temperature (Celsius)
-0.00163
[0.00395]
-0.000877
[0.00402]
Land area (sq. km.)
4.24e-08
[7.01e-08]
1.06e-08
[7.35e-08]
Landlocked dummy
-0.0625
[0.0396]
-0.0585
[0.0397]
Distance to capital (km.)
8.17e-05***
[3.08e-05]
8.05e-05**
[3.11e-05]
Inverse distance to coast
-0.845***
[0.247]
-0.848***
[0.252]
Rule of Law (country level)
0.101***
[0.0328]
GDP per capita (country level)
0.623***
[0.0549]
Log population density
Oil & Gas dummy
0.00279
[0.0256]
Gold & Silver dummy
0.0578
[0.0481]
Other mines dummy
0.0113
[0.0288]
Country Fixed Effects
Observations
Adjusted R-squared
NO
NO
NO
YES
YES
YES
YES
324
0.157
324
0.170
324
0.448
324
0.767
324
0.775
319
0.789
319
0.789
Notes: Robust standard errors are in brackets. *** p<0.01, ** p<0.05, * p<0.1
Table 4
Baseline results with 8 different dependent variables
Dependent variable:
Infant
Mortality
Years of
Schooling
Primary
education
Secondary
education
Drinking
water
Electricity
Log GDP
per capita
Poverty rates
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
-0.0578**
[0.0254]
0.0573***
[0.0183]
0.0265***
[0.00874]
0.0851***
[0.0211]
0.0532***
[0.0169]
0.0832***
[0.0309]
0.204***
[0.0702]
-0.135***
[0.0351]
Share of Amerindian pop.
Log population density
Controls for geography,
location and natural resources
Country fixed effects
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
Observations
Adjusted R-squared
319
0.787
318
0.815
319
0.875
319
0.789
319
0.496
319
0.652
309
0.603
272
0.839
Pre-Colonial Institutions
Controls included:
Notes: All dependent variables are in logs. Robust standard errors are in brackets. *** p<0.01, ** p<0.05, * p<0.1
Table 5
Controlling for colonial activities
Dependent variable:
Infant
Mortality
Years of
Schooling
Primary
education
Secondary
education
Drinking
water
Electricity
GDP per
capita
Poverty rates
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pre-Colonial Institutions
-0.0594**
[0.0260]
0.0454***
[0.0165]
0.0178***
[0.00586]
0.0704***
[0.0184]
0.0466**
[0.0182]
0.0561***
[0.0164]
0.205***
[0.0719]
-0.138***
[0.0364]
Other colonial activities
0.0387
[0.0554]
-0.0339
[0.0387]
-0.0101
[0.0139]
-0.0493
[0.0393]
-0.0529
[0.0457]
-0.0285
[0.0205]
-0.200*
[0.103]
0.145
[0.0963]
Mining colonial activities
0.0163
[0.0658]
-0.00837
[0.0478]
0.0134
[0.0162]
0.00348
[0.0523]
-0.0590
[0.0465]
-0.0120
[0.0404]
-0.352***
[0.117]
0.249**
[0.115]
Plantation colonial activities
0.155**
[0.0770]
-0.0960*
[0.0574]
-0.0183
[0.0171]
-0.143***
[0.0544]
-0.0508
[0.0558]
-0.0674
[0.0411]
-0.402***
[0.152]
0.345**
[0.147]
Share of Amerindian pop.
Log population density
Controls for geography,
location and natural resources
Country fixed effects
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
Observations
Adjusted R-squared
282
0.773
281
0.816
282
0.914
282
0.813
282
0.464
282
0.737
282
0.579
272
0.842
Controls included:
Notes: All dependent variables are in logs. Robust standard errors are in brackets. *** p<0.01, ** p<0.05, * p<0.1
Table 6
Controlling for pre-colonial characteristics: population density
Dependent variable:
Infant
Mortality
Years of
Schooling
Primary
education
Secondary
education
Drinking
water
Electricit
y
GDP per
capita
Poverty
rates
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pre-Colonial Institutions
-0.0587**
[0.0249]
0.0466***
[0.0167]
0.0183***
[0.00573]
0.0726***
[0.0200]
0.0458***
[0.0175]
0.0559**
[0.0231]
0.214***
[0.0766]
-0.143***
[0.0425]
Pre-colonial population density
-0.00666
[0.0140]
-0.0109*
[0.00632]
-0.00496
[0.00332]
-0.0210***
[0.00798]
0.00834
[0.00834]
0.00197
[0.00566]
-0.0869***
[0.0254]
0.0460**
[0.0219]
Other colonial activities
0.0441
[0.0644]
-0.0252
[0.0424]
-0.00611
[0.0145]
-0.0323
[0.0415]
-0.0597
[0.0461]
-0.0301
[0.0257]
-0.129
[0.0988]
0.108
[0.103]
Mining colonial activities
0.0198
[0.0727]
-0.00274
[0.0501]
0.0160
[0.0152]
0.0145
[0.0632]
-0.0634
[0.0467]
-0.0131
[0.0479]
-0.306***
[0.111]
0.225*
[0.116]
Plantation colonial activities
0.164*
[0.0870]
-0.0807
[0.0571]
-0.0112
[0.0184]
-0.114**
[0.0537]
-0.0626
[0.0610]
-0.0702
[0.0427]
-0.279*
[0.146]
0.281*
[0.145]
Share of Amerindian pop.
Log population density
Controls for geography,
location and natural resources
Country fixed effects
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
Observations
Adjusted R-squared
282
0.773
281
0.816
282
0.915
282
0.816
282
0.463
282
0.736
282
0.603
272
0.845
Controls included:
Notes: All dependent variables are in logs. Robust standard errors are in brackets. *** p<0.01, ** p<0.05, * p<0.1
Table 7
Controlling for pre-colonial characteristics: socioeconomic factors
Dependent variable: Percent of the population having completed Secondary education (in logs)
(1)
(2)
(3)
(4)
(5)
(6)
(7)
Pre-Colonial Institutions
0.0735***
[0.0201]
0.0991***
[0.0200]
0.0743***
[0.0204]
0.0778***
[0.0259]
0.0740***
[0.0199]
0.0747***
[0.0218]
0.0751***
[0.0211]
Pre-colonial population density
-0.0294***
[0.0103]
-0.0273***
[0.00999]
-0.0301***
[0.0104]
-0.0293***
[0.0104]
-0.0292***
[0.0105]
-0.0303***
[0.0104]
-0.0292***
[0.0103]
Other colonial activities
-0.0397
[0.0479]
-0.0512
[0.0506]
-0.0423
[0.0484]
-0.0420
[0.0490]
-0.0428
[0.0495]
-0.0414
[0.0482]
-0.0410
[0.0484]
Mining colonial activities
0.00786
[0.0654]
-0.00610
[0.0618]
0.00832
[0.0653]
0.00495
[0.0615]
0.00415
[0.0689]
0.00502
[0.0643]
0.00732
[0.0660]
Plantation colonial activities
-0.113*
[0.0600]
-0.115**
[0.0583]
-0.122*
[0.0618]
-0.118*
[0.0619]
-0.116*
[0.0615]
-0.118*
[0.0606]
-0.116*
[0.0604]
Population employed in:
Gathering
0.150
[0.416]
Hunting
0.360
[0.442]
Fishing
0.615**
[0.277]
Agriculture
-0.468***
[0.135]
Settlement pattern
-0.00343
[0.00335]
Class Stratification
-0.00804
[0.0193]
Slavery
-0.0208
[0.0566]
Elections
-0.122
[0.0971]
Property rights
-0.0367
[0.0478]
Controls included:
Share of Amerindian pop.
Log population density
Controls for geography, location
and natural resources
Country fixed effects
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
Observations
Adjusted R-squared
263
0.811
263
0.822
263
0.810
263
0.810
263
0.810
263
0.811
263
0.810
Notes: Robust standard errors are in brackets. *** p<0.01, ** p<0.05, * p<0.1
Table 8
Contrasting rural and urban areas
Dependent variable:
Pre-Colonial Institutions
Primary education
rural
urban
(1)
(2)
Secondary education
rural
urban
(3)
(4)
Drinking water
rural
urban
(5)
(6)
Electricity
rural
urban
(7)
(8)
0.0228***
[0.00845]
0.00577***
[0.00221]
0.0899***
[0.0250]
0.0167**
[0.00718]
0.0800***
[0.0245]
0.0150
[0.0103]
0.108***
[0.0328]
0.00504*
[0.00289]
Share of Amerindian pop.
Log population density
Controls for geography,
location and natural resources
Country fixed effects
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
Observations
Adjusted R-squared
293
0.880
292
0.933
293
0.813
292
0.885
293
0.610
292
0.224
293
0.685
292
0.615
Controls included:
Notes: Robust standard errors in brackets. *** p<0.01, ** p<0.05, * p<0.1