First Draft: November 2006
This Draft: April 2007
Determinants and Economic Consequences of Colonization:
A Global Analysis
Arhan Ertan, Louis Putterman*
Abstract
Existing research in the area of economic growth suggests that the era of colonization has
had an impact upon the modern levels of economic development of countries around the
globe. However, why some countries were colonized early, some late, and others not at
all, and what effect these differences have on current national income, has not been
studied systematically. In the first part of this paper, we show that both the occurrence
and timing of colonization can be explained by (a) differences in levels of pre-1500
development associated with different dates of transition to agriculture (Neolithic
revolution) and with the associated history of state-level polities, (b) geographic
proximity to the colonizing powers, and (c) the disease environment faced by the
colonizers. In the second part, we analyze the developmental consequences of
colonization taking the endogeneity of colonization’s occurrence and timing into
account. We find evidence that history of colonization does not have a direct affect on
recent levels of income and recent rates of economic growth. But we also find that the
share of the population that migrated from the places of greater pre-modern development
had a positive impact both on current level and growth rate of income and on the quality
of institutions in the newly settled places. Thus we conclude that the positive effect of
colonization on current development works largely through the impact on the quality of
institutions of the pre-colonial development levels of the ancestors of current
populations.
JEL Classification: O11, O13, O40, O57
Keywords: Colonization, Growth, Institutions, Pre-Modern Development, Migration.
*
Candidate for Ph.D. and Professor of Economics, respectively, Brown University, Providence, Rhode
Island 02912 USA. E-mail: [email protected], [email protected]. We would like to
thank Frank Kleibergen for econometric advice and Ashley Lester, Ross Levine and other participants at
Brown’s “macro lunch” and especially David Weil for their helpful comments.
1
Determinants and Economic Consequences of Colonization:
A Global Analysis
“The discovery of America and that of a passage to the East Indies
by the Cape of Good Hope are the two greatest and most important
events recorded in the history of mankind.”
Adam Smith, Wealth of Nations (1776)
“... the colonization of America, trade with the colonies, the
increase in the means of exchange and in commodities generally,
gave to commerce, to navigation, to industry an impulse never
before known...”
Karl Marx and Friedrich Engels,
Manifesto of the Communist Party (1848)
1. Introduction
There is little disagreement among historians that the process by which Western
European nations set sail into the Atlantic Ocean and began the conquest of its islands
and coastlines in the 15th Century, eventually coming to control vast swaths of territory in
the Americas, Africa, Asia and the Pacific, is one of the most important factors shaping
the economic contours of the modern world. The age of modern colonialism began about
1500 with the European discoveries of a sea route around Africa’s southern coast (1488)
and of America (1492) and sea power shifted from the Mediterranean to the Atlantic and
to the emerging nation-states of Portugal, Spain, the Dutch Republic, France, and
2
England. By discovery, conquest, and settlement, these nations expanded and colonized
throughout the world, spreading European institutions and culture. By the time that the
era of colonization ended, the population compositions of countries in the Americas,
Australia and New Zealand were radically transformed. Relative standards of living,
levels of human and physical capital, and also relative rates of economic growth in recent
decades, have all been influenced by the changes brought about by European
colonization.
The chief motives for establishing or winning colonies have been to get control of trade
already existing between a territory and the rest of the world; to get possession of
precious metals, gems, or raw materials; to get a market in the colony; to provide an
outlet in the colony for a surplus population; to take advantage of the cheap labor of
native peoples; and to establish naval and military bases.
Before World War II, two fifths of the world's land area and a third of its population were
in colonies, dependencies, or dominions. Another third of the world’s area had been
colonized by Europeans some time between the 15th and 19th centuries and had emerged
as independent states. While such states had emerged in the Americas long before the
1940s, they were in no sense the same entities that had existed prior to Columbus. In
some, people of European descent constituted a majority; in others, people of African or
even South Asian descent; so, it was not the once-colonized peoples who became
independent, but rather, the descendants of the colonizers and of those whom they had
imported as slave or indentured laborers. In still other cases, the resulting population was
3
a mix of these Old World with New World peoples. The population compositions of
Australia and New Zealand were also radically transformed from their pre-colonial
starting points, but like the North American countries in which descendants of Europeans
are preponderant, these countries, aptly called “Neo-Europes,” had by the early 20th
Century joined not the “developing” but the developed world. But most of the colonies
that gained independence after World War II are among the poorest nations of today, and
much of the Americas are also still counted within the “developing” world. What is
called “the Third World” or “the developing world” consists overwhelmingly of excolonies.
The impact of the colonial era has been directly or indirectly recognized in the past dozen
years by some of the most influential papers that have tried to explain the differences in
income among the world’s nations by reference to institutional and/or geographic factors.
La Porta et al. (1999) emphasize the importance of exogenous historical factors in
explaining the variation in institutional performance. They look at ethnic heterogeneity,
European origins of legal systems and religion as the factors which have the potential to
shape the institutions and government policies. Hall and Jones (1999) attribute large
differences in productivity across countries to differences in institutions and government
policies, what they call “social infrastructure”. They use location and the proportion
speaking European languages as key instruments determining the differences in “social
infrastructure”. Sokoloff and Engermann (2000) pointed out that factor endowments were
more important in explaining the level of economic success. Acemoglu et al. (2001,
2002) claim that variations in local conditions played an independent role in the
4
emergence of institutions and their impact on long-term development. They find that the
differing types of institutions, which they claim date back to the differing modes of
European colonization; have caused a divergence in the current level of economic
success.
With the exception of the Acemoglu et al.’s papers, which associate the different types of
colonization with differences in prior population density or urbanization and in disease
environment, neither the studies mentioned above, nor the others which focus more
directly on the effects of colonization, attempt to explain why some countries were
colonized and others not, or why some were European colonies by as early as the 15th
century and others were colonized only in the late 19th or early 20th centuries.
The motivation for our paper begins with the observation that the non-European world
was extremely diverse on the eve of Europe’s overseas expansion. Regions that would
eventually be colonized included places like the American and Canadian plains and
Australia which were lightly populated by pre-literate tribes or bands making their living
mainly by hunting, and other regions like central Mexico and parts of Nigeria which were
relatively densely populated, had organized states, and relied mainly on agriculture, and
also places like India and Java with still denser populations and older civilizations.
Colonization occurred in places whose climates and soils were suitable to European-style
agriculture, in places where European mortality and morbidity were similar to those at
home, and also in other places with tropical climates and disease environments so
daunting that Europeans did not see their interiors until centuries after the conquest of the
5
Americas. The non-European world in the 1490s also contained areas that were as or
more densely populated, technologically advanced, politically organized, and literate as
Europe itself: countries like Turkey, Japan, Korea and China, we’ll suggest that, not
accidentally managed to resist European annexation throughout the four plus centuries of
Europe’s colonial expansion. Therefore, we wondered whether both the occurrence and
the timing of colonization could be explained by a few key factors, and whether the same
factors might thereby help to explain differences in the way in which colonization
affected the subsequent economic development of the countries in question.
Specifically, our goal is (1) to examine whether geographic, climatic, disease, and pre15th century developmental differences can explain differences in the incidence and
timing of colonization by European powers, and (2) to see whether accounting for the
differences in occurrence and timing of colonization and for their causes changes the
conclusions one may reach regarding the effects of colonization on countries’ current
levels of development and rates of growth in recent years. Along the way, we also begin
to investigate two other related questions, which are: (3) to what degree the effects of
colonization and their differences from country to country are attributable to differences
in the transfer of European and other Old World populations to the non-European
countries and (4) if those differences in migrant contributions to population themselves
are explained by the factors similar to the ones those explain the occurrence and timing of
colonization.
6
We posit three main types of determinants of which countries got colonized and when.
These are: (1) geographic determinants: (a) the distance of the country from northwest
Europe by sea, (b) the existence of land barriers proxied by whether the country is
landlocked, and (c) latitude; (2) a biological determinant: the degree to which endemic
diseases posed a barrier to European control, (proxied by using the variable ‘malaria
ecology’); and (3) a social determinant, the level of development of the country on the
eve of the colonial era, alternately measurable by either (a) its history of state-level
polities, (b) the timing of its Neolithic revolution (transition to agriculture) or, (c) its
population density in 1500.
We find pre-modern development as the most important factor which both decreases the
probability of being colonized and delays the date of colonization; whereas geographic
proximity to Europe plays exactly the opposite role. The role played by the disease
environment is more complicated; a more favorable disease environment causes
colonization to be earlier while its positive effect on the probability of being colonized is
not robust. All these factors jointly explain and 46% to 57% of the probability of being
colonized at all and 29% to 39% of the variation in the timing of colonization among
countries that were colonized.
Having thus explained at least some of the variance in colonization, we turn to studying
colonization’s effects on income and growth. Acemoglu et al. found that colonized
countries which were more urbanized and densely populated in 1500 had become poorer
by 1997, but Burkett et al. (1999) and Bockstette et al. (2002) reported a seemingly
7
contrary result, namely that countries with denser populations and longer-established
“advanced” civilizations, indicated by the presence of state-level polities, were growing
faster than others in the decades after 1960. Hibbs and Olsson (2005) also found that
coming early to agriculture continued to explain almost half of the variance in 1997
income levels in a global sample containing 112 countries. Chanda and Putterman
(forthcoming) investigate this “contradiction” and find evidence that Acemoglu et al.’s
“reversal of fortune” has been in the process of reversing itself since 1960. They find that
countries with earlier established agriculture and states (whether ultimately colonized or
not) were both more populous and had higher incomes in 1500, and also were growing
faster from 1960 onwards, with the period of inferior economic performance by those
countries (1500 – 1960) coinciding with the era of European expansion and standing as
an aberration within world history as a whole. Because incomes in 1960, growth after
1960, and accordingly incomes today (or in 2000) show different relationships with the
pre-1500 development level, which we find to be a major determinant of colonization, we
want to study colonization’s effect on both growth and income. We also want to
investigate the effects of colonization on the quality of institutions using the measure for
that concept as used by Knack and Keefer (1995).
For these purposes, we study simultaneously the determinants and the effects of
colonization by estimating sets of Instrumental Variables (IV) / Two Stage Least Squares
(2SLS) regressions in which the occurrence or timing of colonization is the predicted
variable in the first stage regressions and either real income in 2000 or its rate of growth
between 1960 and 2000 is the dependent variable in the second stage. We also employ
8
Three Stage Least Squares (3SLS) models in which the quality of institutions is also
included among the consequences of colonization and also among the determinants of
income and growth.
We find that colonized countries compared to the non-colonized ones, and countries
which were colonized earlier rather than later, have higher real income per capita in 2000.
But our results show that being a colony and being colonized earlier does not have a
positive impact on the growth rate of real income. Looking at the details of the results, we
argue that the positive impact of colonization on current level of income was mostly
through its mediating effect on the migration of Europeans to the colonies.
The only significant and robust determinant of economic growth in the post-colonial
world was the migration-weighted state history which implies that it is the combination
institutional, cultural and technological elements brought by the migrants with those of
the native population which matters for the growth rate of income and also for the current
quality of institutions.
Observing the crucial impact on current development of the movement of people between
the colonizing powers and the colonies during the colonization era, we also wanted to
explore the factors affecting the migration of different population groups during this
period. Our preliminary findings suggests that the types of crops produced in the places
colonized, as well as the disease environment in those places faced by the migrants were
the two important determinants of the migration during the colonial era. Another
9
interesting result was that these two variables had differential effects on the migration of
Europeans vs. Africans; types of crops were determining where the African population
migrated (was forced to migrate), while having no effect on European migration, and the
disease environment was influential in determining the places where the Europeans chose
to migrate, while it had a smaller impact on African migration.
The remainder of the paper proceeds as follows: Section 2 describes the data used.
Section 3 shows the results of the Probit and Ordinary Least Squares (OLS) regressions
which predict the occurrence and timing of colonization respectively. Section 4 reports
the results of Two-Stage Least Squares (2SLS) regressions which analyze the effects of
being colonized on current level and growth rate of income. Section 5 analyzes the
interaction of the quality of institutions and current income utilizing a Three-Stage Least
Squares (3SLS) system. Section 6 includes the regressions which explain the current
shares of population of European and Sub-Saharan African descent in colonized places.
Section 7 concludes by summarizing our main findings.
2. Data
In our sample, we have a total of 99 countries. We study 89 non-European countries
which were for some time colonized by Western European countries and 10 observations
of ones that were never colonized by Western Europeans (Afghanistan, China, Iran,
Japan, South Korea, Mongolia, Nepal, Saudi Arabia, Thailand and Turkey). We leave
out countries colonized by Russia because the nature of colonization was sufficiently
different from European overseas colonization and we treat countries that emerged from
10
the Ottoman Empire as if they were independent until ruled by France or Britain, at
which point they’re treated as colonies.
In order to analyze the determinants of colonization we need geographic, biological and
social measures. And in order to explore the subsequent effects of colonization on current
level of development we need measures of quality of institutions and income. Figure 1
summarizes each and every link we will try to analyze throughout this paper.
Figure 1: Determinants and Consequences of Colonization
 Geography
(Latitude, Climate, Location)
 Biogeography
(Disease Environment,
Agriculture)
 Social Determinants
(State History, Population
Density, Neolithic Transition)
Colonization
Quality of
Institutions
Current
Development
Migration
2.1. Determinants of Colonization
We hypothesize that 3 major factors will affect colonization of non-European regions by
European countries:
2.1.1. The Level of Development in 1500
We argue that the higher a region’s level of pre-modern development, the less likely it
was to be colonized and the later it was to be colonized, if it was colonized at all. As
measures of pre-modern development we will use:
11
A) State History of 1500:
This is an index which gives weight to; i) whether there was a government,1 ii) whether it
was home-based, iii) the fraction of the territory covered. It gives less weight to earlier
observations (5% discount per 50 years2, starting with the 1451-1500 half century, and
going back to 1 – 50 C.E.) 3.
B) Years since Transition to Agriculture:
It shows the number of centuries since the date that the first substantial group of
inhabitants makes the transition to principally depending on cultivated foods. We use a
new data set4 that is much more country-specific than Hibbs and Olsson’s (which has
values for only nine macro-regions).
C) Population Density in 1500:
As noted by Boserup, agricultural development, technological development, states, and
population density, have gone hand-in-hand throughout history. This is confirmed by a
high correlation between our three measures of pre-modern development. We use
population density in some estimates as an alternative proxy for early development. We
use the 1500 population estimates of McEvedy and Jones (1976), and then divide these
estimates by country area to find population densities.
2.1.2. Disease Environment – Malaria Ecology
1
Although every territory in the world is formally under some type of government today, in 1500 large
parts of North and South America, central and southern Africa, Australia, New Guinea, etc., had no
political institutions beyond the level of bands or tribes.
2
Bockstette et al. (2002) find their results robust to using alternate rates but use this rate due to its
somewhat better fit, as do Chanda and Putterman (forthcoming).
3
This data has been used in Chanda and Putterman (forthcoming in Scandinavian Journal of Economics)
and in some other studies and is available on a link from Putterman’s home page (Putterman, 2004).
4
This data was also developed by Louis Putterman and his research assistants and is also available on a
link from Putterman’s home page (Putterman, 2006).
12
It is commonly assumed that high mortality rates due to disease delayed the European
colonization of most of sub-Saharan Africa by centuries, whereas susceptibility of the
indigenous people to European diseases facilitated the European conquest of the
Americas.5 As a measure / proxy of the disease environment faced by the colonizers, we
have used the malaria ecology variable (Kiszewski et al., 2004). We also checked the
robustness of our regressions using a modified version of Acemoglu et al.’s settler
mortality series to control for the disease environment which was constructed by Albouy
(2004).
2.1.3. Geography
We need three types of geographic variables for our purposes:
A) Costliness of Transportation:
We tried to measure the costliness of transportation by two complementary determinants:
A1) Ocean navigation distance (pre Panama and Suez canals) from Ile D’Ouessant
(a port in France which is roughly a central point for distances from England,
Portugal, Spain, France and the Netherlands). The distances were calculated by
using Publication 151 of “National Imagery and Mapping Agency” which reports
distances between ports and junction points. This variable is used as an
explanatory variable in the regressions where we predict the timing or occurrence
of colonization and as an instrument for colonization and share of population of
5
Although the mortality of indigenous people from European diseases is agreed to have made easier the
European colonization of the Americas, we haven’t found it necessary to formally include it in our analysis;
we only include malaria ecology as a proxy for European mortality, a proxy for barriers to European
settlement.
13
European descent in the 2SLS regressions where we predict the effects of
colonization on current development. It is measured in 1000 miles.
A2) A dummy variable which is equal to one for countries which are landlocked.
B) Latitude:
We will use the absolute value of the latitude (distance from equator) of the country
which has been used in most of the related studies mentioned above. This variable is
included as another measure of the potential for the penetration of European institutions.6
C) Climate:
This variable is constructed based on Köppen’s system of climatic classification and
taken from Hibbs and Olsson (2005). It takes four discrete values; with 3 denoting the
best climate for agriculture (Mediterranean and West Coast climates) and 0 denoting the
worst (tropical dry) 7.
2.2. Determinants and Indicators of Current Development
We have utilized several new variables from newly constructed data sets to be used as the
determinants of current development as well as the standard variables used in other
studies as the indicators of current level of development.
6
Easterly and Levine (2003) interpret absolute value of latitude as an “objective measure of tropics”.
The climatic zones are ordered in ascending value according to how favorable conditions are to
agriculture: 1 = dry tropical or tundra and ice, Köppen classification B and E. 2 = wet tropical, Köppen
classification A. 3 = temperate humid subtropical and temperate continental, Köppen classification Cfa,
Cwa and D. 4 = dry hot summers and wet winters, Köppen classification Csa, Csb, Cfb and Cfc, which are
particularly favorable to annual heavy grasses.
7
14
2.2.1. Shares of Population of European and Sub-Saharan African Descent
We have calculated the percentages of the population of European and Sub-Saharan
African descent. This is from a new data set being assembled by Louis Putterman using
multiple sources. It shows the proportions of the ancestors of each country’s current
population estimated to be living, as of 1500, in each possible country of origin (as
defined by present-day borders).
2.2.2. Migration-Adjusted State History
We have constructed this variable by multiplying the current shares of population from
each country of origin by the state history as of 1500 of the origin country. We intend to
measure the cumulative effect of the “culture” brought by the people currently living in a
particular country, with culture indexed by this early development indicator. In this way,
the U.S., for instance, is today assigned a migration-adjusted state history value in which
the early development measures for countries such as England, Germany, Ireland and
Italy predominate (with smaller weights for African countries, and so forth), while
Guatemala is assigned a migration-adjusted state history value in which the early
development measure of its indigenous inhabitants is dominant, and Haiti’s migrationadjusted state history is one in which the state histories of its population’s mainly African
ancestors are dominant. Countries that have experienced much less migration during the
last 500 years, such as Turkey, Korea and China, display little or no difference in value
for the migration-adjusted versus the (unadjusted) state history variable.
15
2.2.3. Real GDP per Capita in 1960 and 2000
Purchasing power parity adjusted Gross Domestic Product per capita data is available in
Penn World Tables version 6.2. Using this data, we also calculated the annual average
growth rate of income between 1960 and 2000.
Table 1: Descriptive Statistics
Variable
Observations
Mean
98
94
94
99
99
96
97
89
99
99
99
99
99
99
Real GDP per capita in 2000
Real GDP per capita in 1960
Quality of Institutions
State History 1500
Migration Adjusted State History
Log(Population Density 1500)
Years since Transition to Agriculture
Date of Colonization
Abs(Latitude)
Climate
Navigation Distance
Malaria Ecology
Share of Population of European Descent
Share of Population of Sub-Saharan
African Descent
Min
Max
6273.72
616.62
0.5432
0.2762
0.4148
0.4598
4073
1771
18.6302
1.2828
6526
5.6874
0.1551
Standard
Deviation
7260.99
590.60
0.1614
0.3275
0.2877
1.5886
2482
154
11.8492
0.9040
3787
8.2942
0.2678
359.15
87.75
0.2250
0
0
-5.8861
400
1462
0.2280
0
960
0
0
34364.50
3224.03
0.9860
1
1
3.8275
10500
1936
47.4930
3
15321
32.2030
0.9189
0.0557
0.1646
0
0.9489
Note: Descriptive statistics (mean and standard deviation) of the variables used in our empirical analysis.
2.2.4. Quality of Institutions
This is an index (measured on a scale from zero to one) of government anti-diversion
policies created using data assembled by ‘Political Risk Services’ which rates 130
countries according to 24 categories. We follow Knack and Keefer (1995) in using the
equal weighted average of five of these categories for years 1986 to 1995. Two of the
categories relate to the government’s role in protection against private diversion [(i) law
and order, (ii) bureaucratic quality] and three categories relate to the government’s
possible role as a diverter [(i) government corruption, (ii) risk of expropriation, (iii) risk
of government repudiation of contracts].
16
3. Explaining the Timing and the Occurrence of Colonization
We begin our analysis by first exploring the determinants of the date and occurrence of
colonization. For that purpose we estimate two kinds of models:
a) Probit models, where the dependent variable is 1 if a country was ever colonized
by Europeans and 0 if not.
b) OLS models that include only countries that were eventually colonized, where the
dependent variable is the year of colonization.
Table 2: Determinants of Being Colonized
Malaria Ecology
State History 1500
First Agriculture
(years before present)
Log (Population
Density 1500)
Navigation Distance
to Europe
Landlocked
0.0025
[0.2950]
-0.0402***
[0.0001]
-0.0024**
[0.0196]
-0.0109
[0.4223]
Abs(Latitude)
Pseudo R2
# of Observations
0.4603
99
0.0000
[0.9526]
-0.0342***
[0.0000]
-0.0030**
[0.0141]
-0.0133
[0.2460]
-0.0011**
[0.0253]
0.5724
99
0.0015
[0.3864]
-0.0001
[0.9700]
-0.0005***
[0.0002]
-0.0006***
[0.0001]
-0.0033***
[0.0054]
-0.0243*
[0.0816]
0.4780
97
-0.0047***
[0.0078]
-0.0406*
[0.0910]
-0.0015*
[0.0818]
0.5405
97
0.0046
[0.1026]
0.0013
[0.6976]
-0.0056**
[0.0275]
-0.0035**
[0.0138]
-0.0057
[0.6258]
-0.0135***
[0.0037]
-0.0056**
[0.0237]
-0.0300
[0.2655]
-0.0031***
[0.0045]
0.5576
96
0.4770
96
Note: Results of Probit Regressions where the dependent variable is the dummy for being colonized.
Marginal effects rather than the coefficient estimates are reported. The numbers in parentheses are the pvalues calculated using robust standard errors. The constant is omitted from the table.
Table 2 reports the results of the Probit regressions. These suggest that pre-modern
development (measured by either one of the three alternative measures – state history,
first agriculture and population density in 1500) is an important factor which decreases
the probability of being colonized. Navigation distance to Europe and the absolute value
of latitude also seem to be important in determining the chances of being colonized,
17
whereas being landlocked is significant only when we use first agriculture as a measure
of pre-modern development and disease environment (as measured by malaria ecology8)
do not appear as a significant explanatory factor in any version of our estimations.
Table 3: Determinants of the Date of Colonization
Malaria Ecology
State History 1500
First Agriculture
(years before present)
Log (Population
Density 1500)
Navigation Distance
to Europe
Landlocked
0.0051***
[0.0000]
0.0938***
[0.0013]
0.0041*
[0.0727]
0.0419*
[0.0641]
Abs(Latitude)
R2
# of Observations
0.3502
89
0.0061***
[0.0000]
0.0801***
[0.0041]
0.0054**
[0.0148]
0.0427*
[0.0688]
0.0018**
[0.0258]
0.3864
89
0.0052***
[0.0000]
0.0062***
[0.0000]
0.0012***
[0.0000]
0.0011***
[0.0002]
0.0041*
[0.0854]
0.0459**
[0.0419]
0.3590
87
0.0054**
[0.0176]
0.0451*
[0.0543]
0.0017**
[0.0342]
0.3934
87
0.0046***
[0.0000]
0.0061***
[0.0000]
0.0068
[0.1664]
0.0046
[0.1058]
0.0378**
[0.0463]
0.0123
[0.0113]**
0.0061**
[0.0197]
0.0419**
[0.0453]
0.0029***
[0.0011]
0.3814
86
0.2923
86
Note: Results of OLS Regressions where the dependent variable is the year of colonization. The numbers in
parentheses are the p-values calculated using robust standard errors. The constant is omitted from the table.
When we look at the results of the OLS regressions reported at Table 3, where the
dependent variable is the logarithm of the year of colonization, we again see that premodern development emerges as an important factor, in this case significantly delaying
the date of being colonized. Similar to the results of the Probit regressions, the absolute
value of latitude and distance to Europe have positive and significant coefficients in all
versions but now disease environment and the landlocked dummy also appear as
important factors which delay the date of colonization.
8
We also tried estimates substituting the logarithm of settler mortality data for malaria ecology as a proxy
for disease environment and found similar results.
18
Both the results of the Probit and the OLS regressions are robust to using alternative
measures of pre-modern development and also to the inclusion of the absolute value of
latitude. By these results, we can conclude that the set of countries that were colonized
and their dates of colonization were not random but determined by the factors discussed
above. In the next section, we want to analyze the effects of the colonization on current
development taking the dependence of colonization on those factors into account.
4. Explaining the Effects of Colonization on Current Development
In this section, we investigate the effect of colonization on current level of income taking
the endogeneity of colonization into account in a 2SLS regression system. In the first
stage, we use state history of 1500 and navigation distance to Europe as instruments to
predict either the date or the occurrence of colonization (we also try absolute value of
latitude as the third instrument), and then in the second stage we estimate the effect of
colonization on current level of income (real GDP per capita in 2000) controlling for the
direct effects of disease environment and some other geographic factors.
Specifically, we will estimate the following equations:
First Stage:
Y = ao + a1X1 + a2X2 + v
(1)
Second Stage:
Z = bo + b1Y + b2X2 + e
(2)
where Y is either the dummy for being colonized or the logarithm of the date of
colonization, X1 is the set of instruments, X2 is the set of exogenous variables, Z is the
either logarithm of real per capita income in 2000 or the annual average growth rate of
real per capita income between 1960 and 2000, and v and e are the error terms. For the
19
parameter estimates of this 2SLS equation system to be meaningful we need the
following three conditions for the instruments X1:
a) Instruments X1 should be significantly correlated with the endogenous variable Y,
that is, E(Y.X1) ≠ 0,
b) Instruments X1 should be uncorrelated with the error terms of the first stage
regression v; that is, E(v.X1) = 0,
c) Instruments X1 should not be directly effecting the dependent variable Z; that is
E(Z.X1) = 0.
Therefore, in order for state history of 1500 and navigation distance to Europe to be valid
instruments, we should clean away the direct effect of these variables on current income,
so that those variables chosen as instruments for colonization will predict only the date or
the occurrence of colonization in the first stage regressions but will not have any direct
effect on current income. For that purpose, we constructed two new variables to be used
as regressors in the second stage regressions:
a) Migration-Adjusted State History
The early development levels of the ancestors of a country’s current population may
affect the country’s present level of development, but the early development levels in
question can vary appreciably from those of the country’s residents in 1500, due to
subsequent migration. By including this variable, we can control for the direct effect of
pre-modern development on current level of income while retaining the use of
(unadjusted) state history of 1500 as an instrument.
20
b) Air Distances:
We have used the minimum of the air distances of individual countries to Berlin, New
York and Tokyo in order to control for the direct effect of being close to “trade centers”
or “countries at the world technology frontier” on current level income. Thanks to the
difference between air and sea distance, to the fact that the navigation distances used to
predict colonization are pre Panama and Suez Canal distances, and to the three-centered
approach used for current distance, we can control for potential locational advantages
today while preserving navigation distance of 1500 as an instrument.
Other exogenous variables (X2) which we included in the second stage are: dummy for
being landlocked, malaria ecology, absolute latitude and share of the population of
European descent9. Table 3 reports the results of the 2SLS regressions, where the
dependent variable Y in the first stage is the dummy for being colonized.10
4.1. Effect of Colonization on Current Level of Income
Looking only at the first column of Table 4, one can conclude that being colonized is
good for current income. But when we control for the share of population of European
descent in columns 2 and 4, the coefficient of colonization first becomes insignificant and
then becomes negative and significant when we use absolute latitude as another
instrument. Another variable which positively affects current income is migrationadjusted state history, which has a significant coefficient until we use share of population
9
We experimented also with shares of sub-Saharan African and other descent groups but found only the
European descent variable to be significant in most estimates.
10
Although our dependent variable in the first stage regressions is a dummy, according to Wooldridge
(2002, page 84), first stage of 2SLS should be performed by OLS even when the variable to be
instrumented is binary, discrete, or limited in range.
21
of European descent as another control variable, which is also treated as endogenous and
predicted by the same set of instruments that were used for the dummy of being
colonized.
Table 4: Effect of Being Colonized on Current Level of Income
Panel A – First Stage Regressions (Predicting Being Colonized)
State History 1500
-0.3327*** -0.3110** -0.3327*** -0.3110**
[0.0090]
[0.0107]
[0.0090]
[0.0107]
Navigation Distance to Europe
-0.0270*** -0.0299*** -0.0270*** -0.0299***
[0.0006]
[0.0001]
[0.0006]
[0.0001]
Abs(Latitude)
-0.0080***
-0.0080***
[0.0021]
[0.0021]
Probability of F-statistic
0.0000
0.0000
0.0000
0.0000
Panel B – Second Stage Regressions (Predicting real GDP p.c. 2000)
Colonization Dummy
1.9095**
0.7921
-2.6060
-1.3936*
[0.0187]
[0.1888]
[0.1808]
[0.0954]
Migration-Adjusted State History 1.5845*** 1.2003***
0.0907
0.4948
[0.0014]
[0.0040]
[0.9101]
[0.2972]
Minimum Air Distance
-0.0066
-0.0037
-0.0008
-0.0026
[0.2679]
[0.4746]
[0.9177]
[0.6365]
Malaria Ecology
-0.0543*** -0.0511***
-0.0027
-0.0145
[0.0001]
[0.0000]
[0.9139]
[0.3408]
Landlocked
-0.4477
-0.6389** -0.7568** -0.6498**
[0.1248]
[0.0110]
[0.0378]
[0.0161]
Share of Population of
3.5917*** 2.8138***
European Descent
[0.0083]
[0.0000]
# of Observations
98
98
98
98
Note: 2SLS Regressions where the dependent variable in first stage is the dummy for being colonized and
the dependent variable in the second stage is logarithm of real GDP per capita in year 2000. When the
variable ‘share of population of European descent’ is used in the regressions, it has been treated as
endogenous, and the same set of instruments is used for it. The numbers in parentheses are p-values. The
constants are omitted from the table, as are the coefficients of second-stage regressors in the first stage
regressions. All versions of first stage equations passed tests for over-identifying restrictions11, the presence
of weak instruments12 and also the tests for endogeneity13.
11
Our regressions passed Sargan's (1958) and Basmann's (1960) tests of over-identifying restrictions, so
that the excluded instruments are valid instruments, i.e., they are uncorrelated with the error term and
correctly excluded from the estimated equation.
12
Our regressions passed both the Anderson canonical correlations likelihood-ratio test and the CraggDonald chi-squared test for the presence of weak instruments; i.e. first stage equation is not only weakly
identified.
13
Our regressions rejected both the Durbin-Wu-Hausman chi-square test and the Wu-Hausman T-square
test about the endogeneity of the first stage dependent variable where the null hypothesis is that an OLS
estimator of the same equation would yield consistent estimates; that is, any endogeneity among the
regressors would not have deleterious effects on OLS estimates. See Davidson and MacKinnon (1993, 237240) and Wooldridge (2000, 483-484) for more details of these test procedures.
22
These results in part reflects the fact that colonization and European migration raised the
income levels of the countries like North America and Australia, above what would
otherwise be predicted, by transferring older European societal capabilities to those
places which had low values of state history of 1500. Note again that once these factors
are controlled for, colonization per se is not a significant determinant of income. Another
interesting finding is that malaria ecology has a negative impact on current level of
income through its negative impact on European migration. Almost all of these findings
are robust to the inclusion of the absolute value of latitude as another instrument in the
first stage.
Next we turn to the analysis of the effects of the date of colonization on current income.
Table 5 reports 2SLS regressions similar to those in Table 4. The only difference is that
the dependent variable in the first stage is the logarithm of the year of colonization
instead of the dummy of being colonized.
The results reported in Table 5 are telling us a story similar to the one told by the results
of Table 4. Being colonized later decreases current level of income but the positive effect
of being colonized earlier disappears when we control for the share of population of
European descent, which has a positive and significant coefficient when we use absolute
value of latitude as an instrument in the first stage. Migration-adjusted state history is
another related factor which has a positive impact on current level of income in all
versions except one. In column 4, where we control for the share of the population of
European descent and include absolute value of latitude as an instrument, malaria
23
ecology also becomes significant with a negative coefficient, as was the case in the
results of the regressions reported in Table 4. Among our instruments, state history of
1500 and absolute latitude are always significant but navigation distance sometimes loses
its significance in the first stage; but together all of the first stage regressions as such are
highly significant, as indicated by the F statistics of the first stage regressions.
Table 5: Effect of the Date of Colonization on Current Level of Income
Panel A – First Stage Regressions (Predicting Year of Colonization)
State History 1500
0.1615*** 0.1632*** 0.1615*** 0.1632***
[0.0000]
[0.0000]
[0.0000]
[0.0000]
Navigation Distance to Europe
0.0033
0.0058**
0.0033
0.0058**
[0.2022]
[0.0286]
[0.2023]
[0.0286]
Abs(Latitude)
0.0025***
0.0025***
[0.0031]
[0.0031]
Probability of F-statistic
0.0000
0.0000
0.0000
0.0000
Panel B – Second Stage Regressions (Predicting real GDP p.c. 2000)
Log(Colonization Year)
-11.0215*** -5.5550** -48.3601
3.7405
[0.0010]
[0.0122]
[0.6613]
[0.2474]
Migration-Adjusted State History 1.5468*** 1.3253***
3.5486
0.7601*
[0.0024]
[0.0009]
[0.5608]
[0.0651]
Minimum Air Distance
0.0068
0.0032
0.0391
-0.0059
[0.3291]
[0.5485]
[0.6887]
[0.3048]
Malaria Ecology
-0.0012
-0.0261*
0.0822
-0.0351**
[0.9537]
[0.0910]
[0.7464]
[0.0217]
Landlocked
-0.1059
-0.3172
0.5749
-0.3825
[0.7570]
[0.2324]
[0.8004]
[0.1400]
Share of Population of
-7.6617
2.9545***
European Descent
[0.7340]
[0.0002]
# of Observations
88
88
88
88
Note: 2SLS Regressions where the dependent variable in first stage is logarithm of the year of being
colonized and the dependent variable in the second stage is logarithm of real GDP per capita in year 2000.
When the variable ‘share of population of European descent’ is used in the regressions, it has been treated
as endogenous, and the same set of instruments is used for it. The numbers in parentheses are p-values. The
constants are omitted from the table, as are the coefficients of second-stage regressors in the first stage
regressions. All versions of first stage equations passed tests for over-identifying restrictions, the presence
of weak instruments and also tests for endogeneity.
24
4.2. Effect of Colonization on Recent Growth Rate of Income
We are also interested in analyzing the effects of colonization on growth rates of per
capita income. We choose the period 1960 – 2000 for two reasons. First, good income
estimates become available for most countries only after 1960. Second, Chanda and
Putterman (forthcoming) suggest that a regime change occurred some time after the end
of the Second World War, when Europe’s control over most of Africa and Asia was
relaxed and a modern “race for growth” under such institutions as the World Bank, IMF,
and GATT began. As mentioned, they find that countries with higher levels of early or
pre-modern development began to be “high performers” in this period. Unlike their study,
ours uses the migration-adjusted state history variable which controls not only for the
advanced early development of countries in Asia and the Middle East but also for the
advanced early development transferred to countries like the U.S. and Australia by
European migration.
Specifically, we’re interested in income growth after 1960, the date which we considered
as the start of the post-colonial period. We conjecture that countries with higher shares of
population from the regions with greater pre-modern development (higher values of ‘state
history’) in 1500 (the date we consider as the beginning of colonization era) should grow
faster during this post-colonial period. Table 6 and Table 7 report the results of 2SLS
regressions where the dependent variable in the first stage is either the dummy for being
colonized or the date of colonization, and the dependent variable in the second stage is
the average annual growth rate of real GDP per capita between 1960 and 2000. In
addition to our previous controls in the regressions of current level of income, the growth
25
regressions also include the logarithm of the initial year’s real GDP per capita, in order to
control for convergence to a steady-state. We do not want to include the other standard
control variables used frequently in other cross-country studies, such as the average
investment share and average years of schooling, since we are not interested in explaining
the components of growth associated with human capital and physical capital
accumulation. What matters for our purposes is the overall effect of all of these dynamics
on the growth rate of income.
The results of the 2SLS growth regressions suggest that being colonized either does not
play any role or it has a negative effect in determining the growth rate of real GDP per
capita after 1960. The results support the idea of convergence since the coefficient of the
logarithm of initial year’s real GDP per capita has a significant negative coefficient in all
versions. Our proxy for disease environment, malaria ecology, seems to have no impact
on recent economic growth. And another important finding is that our migration adjusted
state history variable has a positive impact on current economic development until we
include share of population of European descent which is treated as endogenous and is
significant when absolute latitude is used as another instrument. These results are similar
to the findings of the above regressions where the dependent variable was the current
level of income.
We also estimate similar 2SLS growth regressions where the dependent variable in first
stage is the logarithm of the year of colonization instead of the dummy for being
colonized. The results of these regressions are reported in Table 7, which suggest that the
timing of colonization has no effect on the growth rate of real GDP. This result is similar
26
to our finding in the previous sets of regressions where we found no effect of being
colonized on recent growth rate of income.
Table 6: Effect of Being Colonized on Recent Growth Rate of Income
Panel A – First Stage Regressions (Predicting Being Colonized)
State History 1500
-0.3578*** -0.3389*** -0.3578*** -0.3389***
[0.0059]
[0.0060]
[0.0059]
[0.0060]
Navigation Distance to Europe
-0.0279*** -0.0308*** -0.0279*** -0.0308***
[0.0006]
[0.0001]
[0.0006]
[0.0001]
Abs(Latitude)
-0.0085***
-0.0085***
[0.0011]
[0.0011]
Probability of F-statistic
0.0000
0.0000
0.0000
0.0000
Panel B – Second Stage Regressions (Predicting Income Growth, 1960 – 2000)
Colonization Dummy
-0.1968
-0.1797
-1.9648
-0.5405*
[0.2682]
[0.2429]
[0.1225]
[0.0505]
Migration-Adjusted State History 0.3144*** 0.3205***
-0.2905
0.1987
[0.0064]
[0.0038]
[0.5758]
[0.1994]
Minimum Air Distance
-0.0007
-0.0007
0.0015
-0.0006
[0.6233]
[0.5918]
[0.7352]
[0.7450]
Malaria Ecology
-0.0044
-0.0045
0.0135
0.0003
[0.1811]
[0.1707]
[0.3760]
[0.9589]
Landlocked
-0.1028
-0.1002
-0.3308
-0.1428
[0.1408]
[0.1420]
[0.1928]
[0.1140]
Log(Real GDP per capita in 1960) -0.0794** -0.0799**
-0.4069* -0.1843***
[0.0200]
[0.0184]
[0.0862]
[0.0094]
Share of Population of
2.0356
0.6156*
European Descent
[0.1272]
[0.0613]
# of Observations
94
94
94
94
Note: 2SLS Regressions where the dependent variable in first stage is the dummy for being colonized and
the dependent variable in the second stage is the annual growth rate of real GDP per capita between 1960
and 2000. When the variable ‘share of population of European descent’ is used in the regressions, it has
been treated as endogenous, and the same set of instruments is used for it. The numbers in parentheses are
p-values. The constants are omitted from the table, as are the coefficients of second-stage regressors in the
first stage regressions. All versions of first stage equations passed tests for over-identifying restrictions, the
presence of weak instruments and also tests for endogeneity.
The results do not provide much support for convergence (coefficient of GDP per capita
of initial year is negative and significant only in one version of the regressions). But
again, the variable which takes into account both migration since the beginning of
colonization and pre-modern development of the places where the people of the countries
27
in our sample were living in year 1500 has a positive and significant coefficient in 3 out
of 4 versions above, which supports our conjecture that in the post-colonial world, what
is important for economic growth is not being colonized or an earlier date of colonization
but having higher shares of population from places where the pre-colonial development
was greater. Another different result is that for the sample of colonized countries, the
share of population of European descent itself do not seem to play a significant role in
determining recent growth rates of income, when it is treated as endogenous.
Table 7: Effect of the Date of Colonization on Recent Growth Rate of Income
Panel A – First Stage Regressions (Predicting Year of Colonization)
State History 1500
0.1487*** 0.1530*** 0.1487*** 0.1530***
[0.0001]
[0.0000]
[0.0001]
[0.0000]
Navigation Distance to Europe
0.0040
0.0064**
0.0040
0.0064**
[0.1400]
[0.0219]
[0.1400]
[0.0219]
Abs(Latitude)
0.0023***
0.0023***
[0.0077]
[0.0077]
Probability of F-statistic
0.0000
0.0000
0.0000
0.0000
Panel B – Second Stage Regressions (Predicting Income Growth, 1960 – 2000)
Log(Colonization Year)
-1.0877
-0.444
-25.6394
-0.051
[0.1935]
[0.4818]
[0.6200]
[0.9576]
Migration-Adjusted State History 0.3881*** 0.3639***
1.6023
0.3413***
[0.0003]
[0.0002]
[0.5486]
[0.0018]
Minimum Air Distance
0.0007
0.0001
0.0268
-0.0003
[0.6630]
[0.9178]
[0.6318]
[0.8351]
Malaria Ecology
-0.0006
-0.0032
0.0622
-0.0038
[0.8888]
[0.4131]
[0.6482]
[0.3639]
Landlocked
0.0136
-0.002
0.7328
-0.0149
[0.8467]
[0.9750]
[0.6482]
[0.8354]
Log(Real GDP per capita in 1960) -0.0748*
-0.0581
0.7153
-0.0862
[0.0752]
[0.1207]
[0.6710]
[0.1740]
Share of Population of
-6.4499
0.1731
European Descent
[0.6320]
[0.5734]
# of Observations
84
84
84
84
Note: 2SLS Regressions where the dependent variable in first stage is logarithm of the year of being
colonized and the dependent variable is annual growth rate of real GDP per capita between 1960 and 2000.
When the variable ‘share of population of European descent’ is used in the regressions, it has been treated
as endogenous, and the same set of instruments is used for it. The numbers in parentheses are p-values. The
constants are omitted from the table, as are the coefficients of second-stage regressors in the first stage
regressions. All versions of first stage equations passed tests for over-identifying restrictions, the presence
of weak instruments and also tests for endogeneity.
28
The robust and maybe the most important message of all these 2SLS regressions where
we considered the effects of both the occurrence and the date of colonization both on the
level and growth rate of current income is that more people from places which had
greater pre-modern development is associated with higher level of income and faster
economic growth. A parallel finding is that the positive impact of being colonized and
being colonized earlier on current level of income (note that these variables do not have
any effect on growth rate of income) is due to the migration of the Europeans to the
colonized places. But when we consider the post-colonial-era growth rate of income, the
positive effect of migration adjusted state history dominates the positive effect of the
greater share of population of European descent. This may be because countries that were
relatively advanced and had long state histories before the colonial era, like China, India,
and Korea, were growing at least as fast as countries that benefited from European
migration, like the U.S., Australia, and Argentina, during the 1960 – 2000 period. In
order to explore the channels through which pre-modern development and people of
European descent affect current level and growth rate of income, in Section 5 we will try
to analyze the role of institutions in this story.
5. Analyzing the Role of Institutions
In this section, we will utilize a 3SLS regression system in order to explore both the
effects of colonization on quality of institutions and also the subsequent effects of the
institutions on current level of income. Specifically, we estimate two 2SLS regression
systems simultaneously. In both regression systems, either the probability of being
colonized or the year of colonization is predicted in the first stage. In the second stage
29
regression of the first of the two systems, our measure of quality of institutions is the
dependent variable, and in the second stage regression of the second system, the
logarithm of real GDP per capita is the dependent variable. Estimating both systems
simultaneously, in a 3SLS regression system, is equivalent to estimating the institutions
and the income equations as seemingly unrelated regressions. Note, however, that we
impose a degree of hierarchy on this system: we include quality of institutions in the
equation for income, but not the converse, thus assuming that colonization may affect
both quality of institutions and current income, institutions can affect current income, but
current income does not affect anything else in the system. The strength of this
methodology is that it not only takes into account the endogeneity of the occurrence or
the date of colonization, but also solves the problem of omitted variable bias by using the
correlation matrix of the error terms of two 2SLS regressions to weight the explanatory
variables.
Table 8 reports the results of the 3SLS regressions in which we consider the effects of
being colonized on quality of institutions and current level of income simultaneously. We
observe that migration-adjusted state history does not have a positive impact on the
quality of institutions but the share of population of European descent has a significant
positive impact when it is included in the regressions, which suggests that it was the
European settlers who constructed ‘good’ institutions in the colonized places. But when
we look at Panel C of Table 8 we find that share of the population of European descent
has no direct effect on current level of income, other than through the channel of quality
30
of institutions. Migration adjusted state history also has a direct positive and significant
effect on the current level of income in both versions of the regressions.
In line with our 2SLS regression results, when we consider their impact on the level of
income, the dummy for being colonized has a positive and significant effect only when
share of population of European descent is not included. Another interesting finding is
that malaria ecology has a negative and significant effect on level of income but has no
effect on quality of institutions in both versions of 3SLS regressions. Overall, we can
argue that these results provide support for the 2SLS results for current level of income
by suggesting that higher quality institutions constructed by Europeans were one of the
possible channels for positive influence of colonization on current development.
Next, we analyze the impact of the date of colonization on quality of institutions and
current level of income, in a similar 3SLS regression system. The results of these 3SLS
regressions are reported in Table 9. These results are mostly parallel to the findings of
previous 3SLS regressions and those of the 2SLS regressions where the dependent
variable of the first stage regression was the dummy of being colonized.
The share of the population of European descent has a positive impact on the level of
income only through the quality of institutions and probably explains why earlier
colonization is associated with better quality of institutions but not with higher income
(only in the version in which the variable share of population of European descent is
included) while migration-adjusted state history shows positive and significant effect on
31
only income, in both specifications. As before, malaria ecology has a significant negative
coefficient in the income equation only.
Table 8: Effect of Being Colonized on Institutions and Income
Panel A – First Stage Regressions (Predicting Being Colonized)
State History 1500
-0.3251***
-0.3251***
[0.0080]
[0.0080]
Navigation Distance to Europe
-0.0297***
-0.0297***
[0.0000]
[0.0000]
Abs(Latitude)
-0.0079***
-0.0079***
[0.0024]
[0.0024]
Probability of F-statistic
0.0000
0.0000
Panel B – Second Stage Regressions (Dependent Variable: Quality of Institutions)
Colonization Dummy
-0.1013
-0.5365***
[0.3517]
[0.0028]
Migration-Adjusted State History
0.0848
-0.0559
[0.2410]
[0.5741]
Minimum Air Distance
0.0007
0.0010
[0.4816]
[0.3928]
Malaria Ecology
-0.0029
0.0043
[0.1651]
[0.1869]
Landlocked
-0.0569
-0.0426
[0.1903]
[0.4439]
Share of Population of
0.5666***
European Descent
[0.0001]
Panel C – Second Stage Regressions (Dependent Variable: Log(real GDP p.c. 2000)
Quality of Institutions
6.2489***
4.5188**
[0.0000]
[0.0168]
Colonization Dummy
1.2671***
0.8173
[0.0052]
[0.4718]
Migration-Adjusted State History
0.7038**
0.7617**
[0.0205]
[0.0120]
Minimum Air Distance
-0.0072*
-0.0058
[0.0668]
[0.1451]
Malaria Ecology
-0.0334***
-0.0338***
[0.0001]
[0.0059]
Landlocked
-0.2325
-0.3219*
[0.2052]
[0.0707]
Share of Population of
0.3575
European Descent
[0.7562]
# of Observations
93
93
Note: 3SLS Regressions where the dependent variable in first stage is the dummy for being colonized and
in the second stage the dependent variables are Quality of Institutions and Real GDP per capita in year
2000. When the variable ‘share of population of European descent’ is used in the regressions, it has been
treated as endogenous, and the same set of instruments is used for it. The numbers in parentheses are pvalues. The constants are omitted from the table, as are the coefficients of second-stage regressors in the
first stage regressions.
32
Table 9: Effect of Date of Colonization on Institutions and Income
Panel A – First Stage Regressions (Predicting Log(Colonization Year))
State History 1500
0.1493***
0.1493***
[0.0000]
[0.0000]
Navigation Distance to Europe
0.0063
0.0063
[0.0230]
[0.0230]
Abs(Latitude)
0.0023
0.0023
[0.0071]
[0.0071]
Probability of F-statistic
0.0000
0.0000
Panel B – Second Stage Regressions (Dependent Variable: Quality of Institutions)
Log(Colonization Year)
-0.6345
1.2138**
[0.1713]
[0.0388]
Migration-Adjusted State History
0.1350*
0.0259
[0.0723]
[0.7088]
Minimum Air Distance
0.0015
-0.0004
[0.1562]
[0.6719]
Malaria Ecology
-0.0002
-0.0024
[0.9409]
[0.3702]
Landlocked
-0.0137
-0.0288
[0.7890]
[0.5142]
Share of Population of
0.5646***
European Descent
[0.0000]
Panel C – Second Stage Regressions (Dependent Variable: Log(real GDP p.c. 2000)
Quality of Institutions
5.7922***
4.2071***
[0.0000]
[0.0003]
Log(Colonization Year)
-2.1015
-1.1033
[0.1671]
[0.6709]
Migration-Adjusted State History
0.5228**
0.6185**
[0.0389]
[0.0166]
Minimum Air Distance
-0.0048
-0.0045
[0.1725]
[0.2385]
Malaria Ecology
-0.0229**
-0.0257**
[0.0186]
[0.0136]
Landlocked
-0.2364
-0.2744*
[0.1433]
[0.0994]
Share of Population of
0.6122
European Descent
[0.4565]
# of Observations
84
84
Note: 3SLS Regressions where the dependent variable in first stage is the date of colonization and in the
second stage the dependent variables are Quality of Institutions and Real GDP per capita in year 2000. The
numbers in parentheses are p-values. The constants are omitted from the table as are the coefficients on
second-stage regressors, which are also entered in the first stage regressions. When the variable ‘share of
population of European descent’ is used in the regressions, it has been treated as endogenous, and the same
set of instruments is used for it.
How past colonization and its timing affects today’s income when we properly account
for the endogeneity of those factors and for other controls is the main question motivating
33
both this and the last part of our analysis. The 3SLS results, which account for the effects
of colonization and migration via the channel of institutions, are similar to the 2SLS
results that both being colonized and being colonized late appear not to have significant
affects on income14.
6. Impacts of Colonization on International Migration
One of the our most important findings appears to be that the era of colonization has
impacts on current level of income but our results suggest that most of these effects are
due to the ‘better’ quality institutions which were brought by Europeans who migrated to
some of the colonies. Another important and robust finding is that migration-adjusted
state history appears to have a significant positive effect on both the recent growth rate of
income, and on its current level.
Since the determinants of both the current level and the growth rate of income is related
with migration that occurred largely as a consequence of the colonization era, we decided
to explore which factors might explain the proportions of people descended from
European and African migrants, whose movement from Old to New Worlds accounts for
the largest shifts in population associated with the colonization and its aftermath.
Specifically we will look at the effects of; (a) level of pre-modern development, (b)
geography (latitude and climate), and (c) crops produced (using dummies for wheat and
sugarcane producing regions). Inspired by Landes (1998), Sokoloff and Engermann
14
We do not have 3SLS regressions with growth rate of income being the dependent variable since the data
we have for ‘quality of institutions” is for years between 1986 – 1995; therefore it would be inappropriate
to use it as an explanatory variable for growth rate of income due to reverse causality problem.
34
(2000) and Easterly and Levine (2003), we included controls for whether wheat and sugar
cane are grown, in the OLS regressions15 reported in both Table 10 and Table 11 where
the dependent variables are the share of population of European and that of Sub-Saharan
African (predominantly involuntary, i.e. slave) migrant descent, respectively.
6.1. Impacts of Colonization on European Migration
The results reported in Table 10 tell us that Europeans migrated to places where disease
environment was better, population density was lower and which were less developed (as
proxied by state history 1500). Given the finding from our previous section that share of
European descent population has a positive influence on current level of income through
the channel of better quality institutions, the results of these regressions confirm the
“reversal of fortune” idea of AJR but also brings a new explanation for that observation.
Table 10: Share of Migrant Population of European Descent
Malaria Ecology
State History 1500
-0.0112***
[0.0000]
-0.4737***
[0.0000]
-0.0130***
[0.0000]
-0.4554***
[0.0000]
Log(Population Density 1500)
Wheat Producer Dummy
Sugarcane Producer Dummy
Abs(Latitude)
0.0579
[0.3434]
0.0718
[0.2950]
0.0113***
[0.0002]
Climate
R2
# of Observations
0.5349
66
0.1200*
[0.0506]
-0.0124
[0.8796]
0.0559
[0.1593]
0.4289
66
-0.0096***
[0.0005]
-0.0099***
[0.0008]
-0.0980***
[0.0000]
0.0008
[0.9902]
0.1312
[0.1891]
0.0049*
[0.0787]
-0.1069***
[0.0000]
0.0268
[0.6824]
0.1065
[0.2744]
0.5652
64
0.0312
[0.4280]
0.5509
64
Note: Results of the OLS Regressions where the dependent variable is share of migrant population of
European descent. Constant is omitted from the table and the numbers in parentheses are the p-values
which are computed using robust standard errors.
15
Sample sizes are smaller now, due to data availability.
35
Not only disease environment but also the pre-modern development level of the places
intended to be colonized determined the intensity of European migration to those places.
Europeans could not settle in large numbers in places with long-established denselysettled civilizations, like India or China, and their population proportion would also
remain lower in the more developed and populous parts of the New World, such as
Mexico. Although malaria ecology and pre-modern development have similar effects on
European migration, pre-modern development is associated with higher growth rate of
income in the post-colonial era while disease environment continues to play its negative
role even on current levels and growth rates of income. But our results suggest that,
positive effect of pre-modern development (weighted by population shares) dominates
both direct and indirect (through lower European migration which has an influence on the
quality of institutions) negative effects of disease environment.
6.2. Impacts of Colonization on Sub-Saharan African Migration
Table 11 reports the results of the OLS regressions where the dependent variable is the
share of population having Sub-Saharan African ancestry. These results show that disease
environment and pre-modern development plays the same role with respect to the
migration of Sub-Saharan African migrants as they did for European migrants, although
for both variables, now the coefficients are a little bit smaller. Neither the absolute value
of latitude nor a climate better for agriculture seems to play any role in determining the
migration of Sub-Saharan Africans.
36
We find that the role of these crops is quite different for African versus European
populations: Sub-Saharan Africans ended up in larger numbers in places where the soil
and climate were more suitable for sugarcane productions (to be used as slaves in
sugarcane plantations) and less suitable for wheat production. Since we found that current
level of income and quality of institutions are positively associated with migrationadjusted state history and proportion of population descended from Europeans, and since
most of the Sub-Saharan countries had low levels of state history, one can argue that the
factors which gave rise to plantation economies with higher proportions of population
being descended from African slaves are at least indirectly accounting for lower quality
institutions and hence lower incomes today.
Table 11: Share of Migrant Population of Sub-Saharan African Descent
Malaria Ecology
State History 1500
-0.0065***
[0.0077]
-0.1218***
[0.0094]
-0.0077***
[0.0081]
-0.1173**
[0.0107]
Log(Population Density 1500)
Wheat Producer Dummy
Sugarcane Producer Dummy
Abs(Latitude)
-0.1385**
[0.0258]
0.1006**
[0.0198]
0.0024
[0.1429]
Climate
R2
# of Observations
0.2633
66
-0.1190**
[0.0286]
0.0600**
[0.0468]
-0.0174
[0.2121]
0.2517
66
-0.0064***
[0.0089]
-0.0074***
[0.0095]
-0.0115
[0.1840]
-0.1701**
[0.0183]
0.1245**
[0.0173]
0.0021
[0.2525]
-0.0165**
[0.0304]
-0.1519**
[0.0160]
0.0876**
[0.0246]
0.2661
64
-0.0221
[0.1135]
0.2626
64
Note: Results of the OLS Regressions where the dependent variable is share of migrant population of SubSaharan African descent. Constant is omitted from the table and the numbers in parentheses are the pvalues which are computed using robust standard errors.
37
7. Summary of Findings and Concluding Remarks
In this paper, first we have examined the determinants and then tried to explain the
developmental consequences of colonization. In doing that, we analyzed separately the
roles of different geographic and historical determinants, first on the colonization
process, and then on current level and growth rate of income. Table 12 summarizes our
findings from different regressions specifications and methodologies. In the final section
of the paper, we tried to explore briefly which factors might explain the proportions of
people descended from European and African migrants.
Table 12: Summary of Findings
Estimation
Method
Dependent ►
Variable
Independent
Variable ▼
Being Colonized
Earlier
Colonization
Pre-Modern
Development
Geographic
Proximity to
Europe
Disease
Quality of
Institutions
OLS
OLS
Being
Colonized
Earlier
Colonization
2SLS
3SLS
Real Income
in 2000
2SLS
3SLS
Growth Rate of
Real Income
1960-2000
Quality of
Institutions
+
+ (*)
0
+
+
+
0
+
– (*)
– (*)
+ (**)
+ (**)
+ (**)
+
+ (*)
+
0
+
0
0
–
– (*)
– (*)
– (**)
–
0
+ (*)
Note: Table shows the sign of the coefficient of the corresponding variable when it was significant in the
corresponding specification and 0 means the variable was not significant. The symbols in parentheses: *
means that the significance of the coefficient was robust to inclusion of additional control variables and **
implies that it was robust in both specifications of the regressions where colonization dummy or the year of
colonization were the dependent variables in the first stage.
In the first part of this study, we have found that pre-modern development was an
important determinant of both the probability of being colonized and also the date to be
colonized; it was less likely for a place to be colonized early or colonized at all which had
38
greater state history or population density in 1500 or longer experience with agricultural
production. The role played by geographic and climatic proximity to Europe was exactly
the same. It is interesting to note that disease environment did not affect the probability of
being colonized while its effect is significant in delaying the date of colonization. This
finding accords with the fact that the disease environment slowed but did not prevent
European penetration into sub-Saharan Africa.
In the second part of the paper, we employed instrumental variable regressions in which
we endogenized colonization and we found that both the colonization itself and some of
its determinants are important in explaining current levels and growth rates of income.
Our results suggest that the being colonized (and colonized earlier) did not have a direct
impact on current level of income and its recent growth rate. We argue that the positive
impact of colonization on current development was mostly through its mediating effect
on the migration of Europeans to the colonies. Because of this observation, in the final
section of the paper, we also tried to explain the factors affecting the migration of
different population groups during the colonization and found the types of crops
produced as well as the disease environment faced by the migrants as the two important
determinants of who migrated to where but these two variables had differential effects on
the migration of Europeans vs. Africans. Further research along these lines may
contribute to the ongoing debate among the economists about the primacy of institutions
vs. geography in terms of their effects on current development.
39
The most significant determinant of economic growth in the post-colonial world was the
migration-adjusted state history which implies that institutional, cultural and
technological elements, whether brought by migrants or sustained by original inhabitants,
determine the growth rate of income and also the current quality of institutions.
Observing differential effects of colonization on current level vs. growth rate of income
implies that dynamics of development are different between the pre-colonial and postcolonial world, and that’s why the roles played by the same factors differ across these
two periods. One might think about it in the framework of a formal growth model in
which two different steady states exist under different conditions.
40
References
Acemoglu, D., Johnson, S., Robinson, J. (2001). “The colonial origins of comparative
development: An empirical investigation”, American Economic Review 91 (5), 1369–
1401.
Acemoglu, D., Johnson, S., Robinson, J. (2002). “Reversal of fortunes: Geography and
institutions in the making of the modern world income distribution”, Quarterly Journal of
Economics 117 (4), 1231–1294.
Albouy, David (2004). “The Colonial Origins of Comparative Development: An
Investigation of the Settler Mortality Data”, unpublished paper, University of California,
Berkeley.
Bertocchi, Graziella and Fabio Canova (2002). “Did Colonization Matter for Growth? An
Empirical Exploration into the Historical Causes of Africa’s Underdevelopment”,
European Economic Review, XLVI, 1851-1871.
Bockstette, Valerie, Areendam Chanda and Louis Putterman (2002). “States and Markets:
The Advantage of an Early Start,”, Journal of Economic Growth 7: 347-69.
Boserup, Ester (1965). “Conditions of Agricultural Growth: The Economics of
Agricultural Change Under Population Pressure”, New York: Aldine.
Burkett, John P., Catherine Humblet and Louis Putterman (1999). “Pre-Industrial and
Post-War Economic Development: Is There a Link?”, Economic Development and
Cultural Change, 47 (3): 471-95.
Chanda, Areendam and Louis Putterman (forthcoming). “Early Starts, Reversals and
Catch-up in the Process of Economic Development,”, Scandinavian Journal of
Economics.
Davidson, R. and J. G. MacKinnon (1993). “Estimation and Inference in Econometrics”,
2nd ed. New York: Oxford University Press.
Diamond, J. (1997). “Guns, Germs and Steel: The Fates of Human Societies.” Norton,
New York.
Easterly, W., Levine, R. (2003). “Tropics, Germs, and Crops: How Endowments
Influence Economic Development.”, Journal of Monetary Economics 50, 3–39.
Galor, O., Weil, D. (2000). “Population, technology, and growth: From the Malthusian
regime to the demographic transition and beyond.”, American Economic Review 90 (4),
806–828.
41
Galor, O., Moav, O. (2002). “Natural selection and the origin of economic growth.”,
Quarterly Journal of Economics 117, 1133–1192.
Glaeser, Edward L., Rafael La Porta, Florencio Lopez-de-Silanes and, Andrei Shleifer,
(2004). “Do Institutions Cause Growth?”, National Bureau of Economic Research
Working Paper 10568.
Goodfriend, M., McDermott, J. (1995). “Early Development”, American Economic
Review 85, 116–133.
Grier, Robin, (1999). “Colonial Legacies and Economic Growth”, Public Choice 98, 317335.
Hall, R., Jones, C. (1999). “Why do some countries produce so much more output than
others?”, Quarterly Journal of Economics 114 (1), 83–116.
Hausman, J. (1978). “Specification Tests in Econometrics”, Econometrica 46(6): 12511271.
Hibbs, Douglas A., and Ola Olsson, 2004, “Geography, biogeography, and why some
countries are rich and others are poor”, Proceedings of the National Academy of Sciences
101: 3715-3720.
Hibbs, Douglas A., and Ola Olsson (2005). “Biogeography and Long-Run Economic
Development”, European Economic Review, 49: 909-938.
Kiszewski A., A. Mellinger, P. Malaney. A.Spielman, S.Ehrlich and J.D. Sachs (2004).
“A Global Index of the Stability of Malaria Transmission Based on the Intrinsic
Properties of Anopheline Mosquito Vectors”, American Journal of Tropical Medicine
and Hygiene (forthcoming).
Knack, S. and P. Keefer (1995). “Institutions and Economic Performance: Cross-Country
Tests Using Alternative Institutional Measures”, Economics and Politics 7(3): 207225.
Kremer, M. (1993). “Population growth and technological change: One million B.C. to
1990”, Quarterly Journal of Economics 108 (3), 681–716.
La Porta, Lopez-de-Silanes, Shleifer, Vishny (1999). “The Quality of Government”, The
Journal of Law, Economics and Organization, 15 (1), 222-279.
La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer, and Robert Vishny
(1998). “Law and Finance”, Journal of Political Economy 106(6), 1113-1155.
Landes, David S. (1998). The Wealth and Poverty of Nations: Why Some are so Rich and
Some so Poor. New York: Norton.
42
Mauro, P. (1995). “Corruption and Growth”, Quarterly Journal of Economics, 110, 681712.
McEvedy, C. and R. Jones (1978). Atlas of World Population History, Viking Press.
North, D. (1981). Structure and Change in Economic History. New York: Norton.
North, Douglas C. (1990). Institutions, Institutional Change and Economic Performance,
New York: Cambridge University Press.
North, Summerhill and Weingast (2000). “Order, Disorder and Economic Change: Latin
America versus North America,” in: Mesquita and Root, eds., Governing for Prosperity.
New Haven: Yale University Press.
Putterman, Louis (2004). “State Antiquity Index Version 3”, available at
http://www.econ.brown.edu/fac/Louis_Putterman/.
Putterman, Louis (2006). “World Migration Matrix, 1500 – 2000”, available at
http://www.econ.brown.edu/fac/Louis_Putterman/.
Putterman, Louis, with Cary Anne Trainor (2006). “Agricultural Transition Year Country
Data Set”, available at http://www.econ.brown.edu/fac/Louis_Putterman.
Rodrik, D., Subramanian, A., Trebbi, F. (2002). “Institutions Rule: The Primacy of
Institutions over Geography and Integration in Economic Development.”, Working Paper
No. 9305, NBER, Cambridge, MA.
Sokoloff, Kenneth and Stanley Engermann (2000). “Institutions, Factor Endowments and
Paths to Development in the New World,” Journal of Economic Perspectives 14: 217232.
Wooldridge, J. (2002). “Introductory Econometrics: A Modern Approach”, 2nd ed. New
York: South-Western College Publishing.
Wu, D. M. (1973). “Alternative Tests of Independence Between Stochastic Regressors
and Disturbances”, Econometrica 41(4): 733-750.
43