Reversal of Fortune:
       Geography and Institutions in the Making of the
             Modern World Income Distribution∗
       Daron Acemoglu                                           Simon Johnson
 Department of Economics, MIT                          Sloan School of Management, MIT
                               James A. Robinson
                      Department of Political Science, Berkeley
                                           July 5, 2001


                                             Abstract
           Among countries colonized by European powers during the past 500 years those
       that were relatively rich in 1500 are now relatively poor. We document this reversal
       using data on urbanization patterns, which, we argue, proxy for economic prosperity.
       This reversal is inconsistent with a view that links economic development to geo-
       graphic factors. According to the geography view, societies that were relatively rich
       in 1500 should also be relatively rich today. In contrast, the reversal is consistent
       with the role of institutions in economic development. The expansion of European
       overseas empire starting in the 15th century led to a major change in the institu-
       tions of the societies they colonized. In fact, the European intervention appears to
       have created an “institutional reversal” among these societies, in the sense that Eu-
       ropeans were more likely to impose “extractive” institutions in densely settled and
       prosperous areas, while introducing institutions encouraging investment in regions
       that were previously poor. This institutional reversal accounts for the reversal in
       relative incomes. We provide further support for this view by documenting that the
       reversal in relative incomes took place during the 19th century, and resulted from
       societies with extractive institutions failing to take advantage of industrialization
       opportunities.

          In Progress.
   ∗
    We thank Joshua Angrist, Abhijit Banerjee, Olivier Blanchard, Jan de Vries, and seminar participants
at the All-UC History Conference at Berkeley, the conference on “Globalization and Marginalization”
in Bergen, Canadian Institute of Advanced Research, Chicago, Columbia, DELTA, MIT, Stanford, and
Yale for useful comments.

                                                   1
                                   1   Introduction

The “geography hypothesis” explains the bulk of the differences in economic performance
across countries by geographic, climatic or ecological features of the environment. Moti-
vating this view is a fact that Galbraith summarized succinctly in 1951: if “one marks a
belt a couple thousand miles in width encircling the earth at the equator one finds within
it no developed countries...”. The list of scholars who have emphasized the importance
of geographic factors is long and distinguished, and includes, inter alia, Niccol` Machi-
                                                                                 o
avelli, Charles de Montesquieu, Arnold Toynbee, Alfred Marshall, Ellsworth Huntington,
and Gunnar Myrdal. All of these authors viewed climate as a key determinant of work
effort, productivity, and ultimately, the success of nations. For example, Marshall (1890)
wrote: “vigor depends partly on race qualities: but these, so far as they can be explained
at all, seem to be chiefly due to climate” (p. 195). In a recent influential book, Jared
Diamond (1997) has argued for the importance of geographic determinants of the Ne-
olithic revolution, and linked modern prosperity to the timing of the emergence of settled
agriculture. He forcefully states that “the striking differences between the long-term his-
tories of peoples of different continents have been... [due to]... the differences in their
environments” (p. 405). Jeffrey Sachs, on the other hand, has argued for the importance
of technology, disease environment and transport costs, which are determined by phys-
ical geography and climate. For example, Sachs (2001) writes “By the start of the era
of modern economic growth, if not much earlier, temperate-zone technologies were more
productive than tropical-zone technologies...” (p. 2).
   An alternative view, instead, relates differences in economic performance to the orga-
nization of the society. Societies that provide incentives and opportunities for investment
will be richer than those who fail to do so. This view dates back at least to John Locke,
who argued for the necessity of property rights for productive activities, and to Adam
Smith, who stressed the role of “peace, easy taxes, and a tolerable administration of jus-
tice” in generating prosperity (quoted in Jones, 1981, p. 235). Recent proponents of this
approach have been economists and historians emphasizing the importance of property
rights and institutions (e.g., North and Thomas, 1973, North and Weingast, 1989, North,
1990). For this reason, we refer to this approach as the “institutions hypothesis”.
   In this paper, we attempt to distinguish between these two broad hypotheses. If geog-


                                            1
raphy is the key determinant of income differences across countries, economic performance
should be highly persistent, since geographic factors did not change much during recent
history. To the extent that other factors also matter for income, persistence will not be
perfect, but we should expect rich countries today to have been, on average, richer 100,
200, 500 or even 1000 years ago (see, e.g., Diamond, 1997). Since institutions and the way
that societies are organized are persistent as well, the institutions hypothesis also predicts
persistence in income levels. Nevertheless, if there is a major change in institutions, then
we should expect a significant change in the distribution of income across countries.
   The expansion of the European overseas empire starting at the end of the 15th century
provides an appealing “natural experiment” to distinguish between these two contrasting
predictions. In the subsequent four and a half centuries, a large part of the globe that
was previously ruled by independent polities came under European control, transforming
their social organization and institutions in fundamental ways. Despite these radical social
changes, the geography view predicts persistence in relative incomes: the same geographic,
climatic and ecological factors making countries prosperous before European colonialism
should also contribute to prosperity after the arrival of the Europeans. In contrast, if
European dominance came with a major change in the organization of these societies, the
institutions hypothesis implies that that there should not necessarily be such persistence.
   Historical and econometric evidence suggests not only that European colonialism
caused a major change in the social organization of these regions, but also that it created
an “institutional reversal”–Europeans imposed worse institutions in previously prosper-
ous areas. The main reason for the institutional reversal is that relatively poor areas were
sparsely settled, and this enabled or induced Europeans to develop settler colonies with
institutions encouraging investment and commerce by a broad cross-section of the society.
In contrast, a large population and relative prosperity made it profitable for the Euro-
peans to set up extractive institutions, with political power concentrated in the hands of
a small elite. High population density, for example, meant a large supply of labor that the
Europeans could force to work in mines or plantations, or tax heavily by taking over ex-
isting tribute systems. Furthermore, high population density and the associated diseases
often made it less attractive for Europeans to settle. Because Europeans were more likely
to set up extractive institutions in places they did not settle, high population density
also made the development of extractive institutions more likely through this channel.
The institutions hypothesis, together with the institutional reversal caused by European


                                              2
colonialism, suggests the possibility of a reversal among the European colonies: countries
that were relatively rich in 1500 should be relatively poor today.
       The major finding of this paper is that there is a reversal in relative incomes among
the former European colonies. For example, the Mughal Empire in India, the Aztec
and the Inca Empires in America were among the richest civilizations in 1500, while the
civilizations in North America, New Zealand and Australia were highly primitive. Today
the U.S., Canada, New Zealand and Australia are orders of magnitude richer than the
countries now occupying the territories of the Mughal, Aztec and Inca Empires, such
as India or Ecuador. This reversal is inconsistent with the geography hypothesis, but
consistent with the institutions hypothesis. We also show that the timing and the pattern
of the reversal is closely linked to the “institutions” imposed by the European powers.
       The obvious difficulty in our empirical investigation is lack of data on economic pros-
perity in 1500. The first contribution of our paper is to propose and use urbanization
rates as a proxy for cross-country differences in economic prosperity during preindustrial
periods. Bairoch (1988) argues that only areas with high agricultural productivity could
support large urban populations, while de Vries (1976, p.164) emphasizes the necessity
of improvements in transportation and fuel technology to provide sufficient energy sup-
plies for cities as they grow. Similarly, many economic historians subscribe to the view
that increasing urbanization is associated with economic development (see, for example,
Kuznets, 1968, de Vries, 1984, Bairoch 1988, Tilly, 1990, De Long and Shleifer, 1993).
We also present evidence that both in the time series and the cross section there is a close
association between urbanization and income per capita.1
       As an additional proxy for prosperity we use population density, on which there are
relatively more extensive data (e.g. McEvedy and Jones, 1975). Although the theoretical
relationship between population density and prosperity is more complex, it seems clear
that during the preindustrial periods only relatively prosperous areas could support dense
populations.

         Figure 1 here: GDP per capita today vs. urbanization in 1500 for ex-colonies

       With either measure, there is a negative relationship between economic prosperity in
   1
     By economic prosperity or income per capita in 1500, we do not refer to the economic or social
conditions of the masses, but to a measure of total production in the economy relative to the number
of inhabitants. Although urbanization is likely to have been associated with relatively high output per
capita, the majority of urban dwellers lived in poverty and died young because of poor sanitary conditions
(see for example Bairoch, 1988, chapter 12).

                                                    3
1500 and today. Figure 1 shows a negative relationship between percent of the population
living in towns with more than 5000 inhabitants in 1500 and income per capita today (see
below for data details). The big gap in this figure is sub-Saharan Africa for which there
are no reliable urbanization data. Figure 2 shows the same negative relationship between
population density (number of inhabitants per square km) in 1500 and income per capita
today, including data from sub-Saharan African countries. The relationships shown in
Figures 1 and 2 are robust. They are unchanged when we control for continent dummies,
the identity of the colonial power, geography-related variables, including temperature,
humidity, and distance from the equator, and when we exclude the pure settler colonies
that Alfred Crosby (1986) refers to as Neo-Europes–the U.S., New Zealand, Canada and
Australia.

  Figure 2 here: GDP per capita today vs. population density in 1500 for ex-colonies

   Some degree of caution is required in interpreting these results since both urbanization
and population density estimates for 1500 are likely to be error ridden. Yet the direct
effect of measurement error is to create an attenuation bias towards 0 in the correlation
between urbanization or population density in 1500 and income today. So measurement
error alone is unlikely to account for our results. Naturally, our findings could be explained
by errors in the estimates of urbanization or population density that are correlated with
current economic development. This possibility seems unlikely, however, since the basic
patterns we document are robust to a variety of alternative estimates, and are consistent
with qualitative assessments of historians. We therefore conclude that the data point to
a reversal among former colonies.
   While the reversal in the relative economic position of the colonies weighs strongly
against a simple geography view, one might advance a more sophisticated geography
theory, which we dub the “Northern drift hypothesis”. According to this hypothesis, the
center of gravity of economic activity has been gradually shifting to the North, or at
least away from the equator. Sachs (2001), for example, implies such a view by pointing
out that temperate-zone areas have been growing faster during the past 200 years than
tropical areas. In fact, in 1500 the tropical areas were relatively rich, and today they are
among the poorest places in the world. It can be argued that areas in the tropics had an
early advantage, but later agricultural technologies, such as the heavy plow, crop rotation
systems, use of domesticated animals and high-yield crops, have favored countries in the
temperate areas. However, the timing of the reversal in relative incomes appears to be

                                             4
inconsistent with this hypothesis. First, the reversal in relative incomes seems to be related
to population density before Europeans arrived, not to any geographic characteristics of
the area. More important, there is no compelling reason why climate and geography
should matter more for industry than for agriculture. So according to this hypothesis, the
reversal should occur when European agricultural technology spread to the colonies. Yet,
in practice the reversal occurred mostly during the 19th century, while the introduction of
European agricultural techniques, at least in North America, took place earlier. Moreover,
the reversal seems to be closely related to industrialization: ex-colonies that could take
advantage of industrial technology surged ahead, while those that failed to do so lagged
behind.
    Is the reversal related to institutions? The answer appears to be yes. We document
that differences in the institutions that the Europeans introduced statistically account for
the reversal in relative incomes from 1500 to today. Moreover, the institutions hypothesis
suggests that institutional differences should matter more when new technologies requir-
ing investments from a broad cross-section of the society become available. We therefore
expect societies with institutions of private property to take advantage of industrialization
opportunities, while societies with extractive institutions where political power is concen-
trated in the hands of the small elite fail to do so. The data support this prediction.
    We are unaware of any other work in the economics, political science or history litera-
tures that has noticed or documented this change in the distribution of economic prosper-
ity, or used the experiences of the former colonies to distinguish between the geography
and institutions hypotheses. Nevertheless, there are many country experiences consistent
with our general picture. For example, many scholars, including McNeill (1967), Jones
(1981), Kennedy (1987), Chaudhuri (1990), Wong (1997) and Pomeranz (2000), empha-
size that in 1500 the Mughal, Ottoman and Chinese Empires were highly properous,2 but
grew slowly during the next 500 years. Coatsworth (1993) and Engerman and Sokoloff
(1997, 2000) document that North America was no more developed than South America in
the early 18th century and the data presented by Eltis (1995) and Engerman and Sokoloff
(2000) suggest that Carribean islands such as Haiti and Barbados were considerably richer
than the United States during early colonial times. Our work therefore also establishes
    2
      In addition to data on urbanization and population density that we emphasize, these authors also
present extensive documentation that patterns of health, life expectancy and stature differed little in
South and East Asia from those prevailing in Europe at the same time. Reid (1988, p. 50) concludes his
survey of the evidence by noting “the relatively good health of Southeast Asians...should not surprise us
if we compare their diet, medicine, and hygiene with those of contemporary Europeans.”


                                                   5
that the pattern pointed out by Coatsworth, Eltis, and Engerman and Sokoloff applies
for comparisons of economic prosperity in 1500, i.e., before colonial times, and for the
whole sample of former European colonies. More importantly, our larger sample enables
us to investigate the robustness of this result and to investigate the relative merits of the
geography and institutions hypotheses.
       The view that the colonial experience of many countries has retarded their economic
development is put forward by many Marxist historians and dependency theorists, for ex-
ample, Williams (1944), Rodney (1972), Wallerstein (1974-1980) and Frank (1978). The
work by Engerman and Sokoloff (1997, 2000) is more closely related because, like us, they
emphasize the long-run adverse consequences of the “plantation complex” and the associ-
ated institutions in Latin America (see also work by historians Genovese, 1965, Beckford,
1972, Dunn, 1972, Curtin, 1990 and Coatsworth, 1999). Our approach differs from these
contributions in two important dimensions: first, we view European colonialism as a
“natural experiment” potentially distinguishing between the geography and institutions
hypotheses. Therefore, we emphasize not the negative effects of European colonialism
relative to what would have happened without colonialism, but the differential effect of
European colonialism in some countries vs. others: societies where colonialism led to the
establishment of good institutions prospered relative to those where colonialism imposed
extractive institutions. Second, in our theory the negative effects of colonialism do not
result from the plunder of the colonies by the Europeans or dependency as emphasized
by Williams, Rodney or Frank, but because the extractive colonial institutions stacked
the cards against industrialization.3
       Our overall interpretation of comparative development in the former colonies is related
to our previous paper, Acemoglu, Johnson and Robinson (2001), where we proposed
the disease environment at the time Europeans arrived as an instrument for European
settlements and subsequent institutional development of the former colonies, and used
this to estimate the causal effect of institutional differences on economic performance.
The current paper has a different focus and a number of innovations relative to our earlier
work. First, our focus here is on the persistence of economic performance, which we argue
can distinguish between the geography and institutions hypotheses. Second, we point out
and document the major reversal in the distribution of economic prosperity among the
   3
    Put differently, according to Williams, Rodney or Frank, countries in Central America, the Caribbean
and Africa are poor because of “too much capitalism,” whereas in our thesis, they are poor because of
“the wrong type of capitalism”.



                                                  6
former colonies. Third, our thesis here emphasizes the direct effect of population density,
and more generally prosperity at the time Europeans arrived, on the policies pursued by
the Europeans, for example, by encouraging labor-oppressive production methods, and
the takeover of existing tax and tribute systems. Finally, we suggest and document that
the interaction between the set of institutions brought by Europeans and the opportunity
to industrialize during the 19th century played a central role in the long-run development
of the former colonies.
   Our results are also relevant to the literature on the relationship between population
and growth. The recent consensus is that population density encourages the discovery and
exchange of ideas, and contributes to growth. This view goes back to Boserup (1965),
Jacobs (1967) and Kuznets (1968), and was elaborated by Simon (1977).          The recent
endogenous growth literature also emphasizes the beneficial effects of high population
through scale effects (e.g., Romer, 1986, 1990, Grossman and Helpman, 1991, Aghion
and Howitt, 1992, Kremer, 1993, Jones, 1999). Our evidence points to a major historical
episode of 500 years where high population density was detrimental to economic develop-
ment, and therefore sheds doubt on the general applicability of this recent consensus.
   The rest of the paper is organized as follows. The next section discusses the construc-
tion of urbanization and population density data, and provides evidence that these are
good proxies for economic prosperity before the modern era. In Section 3, we outline
the geography and institutions hypotheses, explain why we should expect an institutional
reversal resulting from European colonialism, and present historical and econometric evi-
dence supporting this notion. Section 4 documents the “Reversal of Fortune”–the nega-
tive relationship between economic prosperity in 1500 and income per capita today among
the former colonies. Section 5 discusses the Northern drift hypothesis, and shows that the
reversal took place during the 19th century, and was linked to industrialization, which we
interpret as evidence against this hypothesis. Section 6 documents that the institutional
reversal caused by European colonialism statistically accounts for the reversal in relative
incomes between 1500 and today, and that institutions started playing a much more im-
portant role in the age of industry. Section 7 concludes, while the Appendix describes
how urbanization and population density data are constructed.




                                            7
                      2   Urbanization and Population Density

A measure of economic prosperity in 1500 is crucial for our investigation. In this section,
we argue that urbanization and population density provide good proxies for income per
capita and/or productivity, and explain how these data are constructed. More detailed
discussion of data construction and alternative series are provided in Appendix A.

2.1    Data on Urbanization

Bairoch (1988) provides the best single collection and assessment of urbanization research.
Our base data for 1500 consist of Bairoch’s (1988) urbanization estimates augmented by
the work of Eggimann (1999). We also construct a longer time-series data for a subset
of countries to study the timing of the reversal and whether there was a similar reversal
before 1500.
      Merging the Eggimann and Bairoch series requires us to convert Eggimann’s estimates,
which are based on a minimum population threshold of 20,000, into Bairoch-equivalent
urbanization estimates, which use a minimum population threshold of 5,000. We use
a number of different methods to convert between the two sets of estimates, all with
similar results. For our base estimates, we run a regression of Bairoch estimates on
Eggimann estimates for all countries where they overlap in 1900 (the year for which
we have most Bairoch estimates for non-European countries). This regression yields
a constant of 6.6 and a coefficient of 0.67. We use these estimates below to generate
Bairoch-equivalent urbanization estimates. Alternatively, we converted the Eggimann
estimates using country-specific conversion ratios and conversion ratios at the regional
level (e.g., separately for North Africa and the Andean region) using information suggested
by Bairoch’s analysis, and also we checked the robustness of our results using a uniform
conversion rate of 2 as suggested by Davis’ and Zipf’s Laws (see Bairoch, 1988, chapter
9). We also used Bairoch’s overall assessment of urbanization for broad regions, e.g., Asia,
without the more detailed information from Eggimann, in all cases with similar results.
Details and robustness results are reported in the Appendix.
      Finally, we also constructed three alternative series without combining estimates from
different sources. One of these is based mainly on Chandler’s data, the other on Bairoch,
and the final one on Eggimann. Results using these three different measures are reported
in Table 4.
      While the data on sub-Saharan Africa are worse than for any other region, it is clear

                                              8
that urbanization before 1500 was at a higher level than in North America or Australia.
Bairoch claims that by 1500 urbanization was “well-established” in sub-Saharan Africa.4
Yet by 1900, sub-Saharan Africa certainly was less urbanized than the European settler
colonies. Because there are no detailed urbanization data for Africa, we leave this region
out of the regression analysis when we use urbanization data, though it is included in our
regressions using population density.
       Table 1 gives descriptive statistics for the key variables of interest, separately for the
whole world, for the sample of ex-colonies for which we have population density data in
1500, and for the sample of ex-colonies for which we have urbanization data in 1500.

                                            Table 1 here

2.2      Urbanization and Income

There are good reasons to presume that urbanization and income are positively related.
Kuznets (1968, p. 1) opens his book by stating:

             “we identify the economic growth of nations as a sustained increase in
         per-capita or per-worker product, most often accompanied by an increase in
         population and usually by sweeping structural changes....in the distribution
         of population between the countryside and the cities, the process of urbaniza-
         tion.”

       Bairoch (1988) gives detailed evidence that only relatively rich countries could afford
a high fraction of the population living in cities. For example, he points out that during
preindustrial periods, a large fraction of the agricultural surplus was likely to be spent
on transportation, so both relatively high agricultural surplus and a developed transport
system were necessary for large urban populations (see Bairoch 1988, chapter 1, de Vries
1976, p. 164). Bairoch writes “the existence of true urban centers presupposes not only a
surplus of agricultural produce, but also the possibility of using this surplus in trade” (p.
11). Moreover, he emphasizes that an increase in agricultural productivity almost always
tended to cause increased urbanization:
   4
    Sahelian trading cities such as Timbuktu, Gao and Djenne (all in modern Mali) were very large in
the middle ages with populations as high as 80,000. Kano (in modern Nigeria) had a population of 30,000
in the early 19th century, and Yorubaland (also in Nigeria) was highly urbanized with a dozen towns
with populations of over 20,000 and its capital Ibadan possibly had 70,000 people. For these numbers
and more, see Hopkins (1973, Ch. 2).

                                                  9
            “For while it is true that urbanization could not get underway without the
         concentration of population and the surplus of food resulting from agriculture,
         it is equally true that the emergence of agriculture set in motion forces that
         sooner or later led to the growth of cities.” (p. 94).

This view, that urbanization and income (productivity) are closely related, is shared by
many other historians (see, for example, Tilly, 1990, De Long and Shleifer, 1993, and Tilly
and Brockmans, 1994).5
       A large history literature also documents how urbanization accelerated in Europe
during periods of economic expansion (e.g., Pirenne 1956, Postan and Rich 1966, Duby
1974). For example, the period between the beginning of the 11th century and mid-14th
century is an era of rapid increase in agricultural productivity and industrial output. The
same period also witnessed a proliferation of cities. Bairoch (1988), for example, estimates
that the number of cities with more than 20,000 inhabitants increased from around 35-44
to 100-110.
       We supplement this argument by empirically investigating the link between urbaniza-
tion and income. Figures A1 and A2 in the Appendix show the time-series relationship
between urbanization and income per capita for a number of countries. In all cases,
changes in urbanization and income are highly correlated. In Table 2, we provide re-
gression evidence also pointing in the same direction. Columns 1-6 in Panel A present
cross-sectional regressions. Column 1 is for the earliest date for which we have data on
urbanization and income per capita for a large number of countries, circa 1900. The re-
gression coefficient, 0.038, is highly significant, with a standard error of 0.006. It implies
that a country with 10 percentage points higher urbanization has, on average, 46 percent
greater income per capita (38 log points). Column 2 reports a similar result using data
for 1950. Column 3 uses current data and shows that even today there is a strong rela-
tionship between income per capita and urbanization for a large sample of countries. The
coefficient is similar, 0.036, and very precisely estimated, with a standard error of 0.002.
This relationship is also shown diagramatically in Figure 3.

    Figure 3: cross-sectional graph of income per capita today vs. urbanization today
   5
    De Long and Shleifer (1993) write “The larger preindustrial cities were nodes of information, industry,
and exchange in areas where the growth of agricultural productivity and economic specialization had
advanced far enough to support them. They could not exist without a productive countryside and a
flourishing trade network. The population of Europe’s preindustrial cities is a rough indicator of economic
prosperity” (p. 675).

                                                    10
      Below, we draw a distinction between countries colonized by European powers and
those never colonized. Column 4 and 5 report the same regression separately for these
two samples. The estimates are very similar: 0.037 for the former colonies sample, and
0.033 for the rest of the countries. Finally, in column 6, we add continent dummies to the
same regression. This leads to only a slightly smaller coefficient of 0.030, with a standard
error of 0.002. This result demonstrates that the correlation between urbanization and
income per capita is not driven by differences across continents. We therefore conclude
that there is a robust cross-sectional relationship between urbanization and income per
capita.
      The second panel of the table is more speculative. Here we use estimates from Gregory
King and Paul Bairoch to construct a small panel data set of urbanization and income per
capita spanning over 200 years with sporadic observations. Columns 7-10 report regres-
sions from this dataset. Remarkably, with or without country and period dummies, the
estimates are very similar to those shown in Panel A. They indicate that a 10 percentage
point increase in urbanization is associated with an approximately 40 percent increase in
income per capita. Overall, we conclude that urbanization is a good proxy for income.

                                        Table 2 here.

2.3    Data on Population Density

The most comprehensive data on population since 1AD comes from McEvedy and Jones
(1975). They provide estimates based on censuses and published secondary sources. While
some individual country numbers have since been revised and others remain contentious
(particularly for pre-Colombian Meso-America), their estimates are still widely respected
(see, e.g., the assessment by the Bureau of the Census, www.census.gov/ipc/www/worldhis.html).
We use McEvedy and Jones (1975) for our baseline estimates, and indicate where we have
tested the effect of using alternative assumptions (e.g., lower or higher population esti-
mates for Mexico and its neighbors before the arrival of Cortes).
      We calculate population density by dividing total population by arable land (also
estimated by McEvedy and Jones). This excludes primarily desert, inland water, and
tundra (we also test for the robustness of including this area). As much as possible, we
use the land area of a country at the date which we are considering (e.g., in 1500). We
also checked the robustness of our results to changing this assumption and none of our
findings are affected by this.

                                             11
2.4     Population Density and Income

The theoretical relationship between population density and income is more nuanced than
that between urbanization and income. With similar reasoning, it seems natural to think
that only relatively rich areas could afford dense populations (see Bairoch, 1988). This is
also in line with Malthus’s classic work. Malthus argued that high productivity increases
population by raising birth rates and lowering death rates. However, the main thrust of
Malthus’s work was on how a higher than equilibrium level of population increases death
rates and reduces birth rates to correct itself. Therefore, a high population could also be
reflecting an “excess” of population, causing low income per capita.
      To clarify the main issues, it is useful to express these relationships mathematically.
Let us denote population in country (area) j at time t by Pj (t), land area by Lj , and
the level of multifactor productivity by Aj (which is assumed to be time invariant for
simplicity). Suppose
                                   Yj (t) = Aj · Pj (t)1−θ · Lθ ,
                                                              j                           (1)

where θ ∈ (0, 1). So greater population increases output, but because of decreasing returns
to land, it does so less than proportionately. Dividing both sides of the equation by Pj (t),
we obtain
                                      yj (t) = Aj · pj (t)−θ ,                            (2)

where y is per capita output and p = P/L is population density. Next, we need an
equation linking growth of population to income per capita. Suppose that this takes the
form
                         pj (t + 1) = ρ · pj (t) + λ · (yj (t) − y ) + εj (t)
                                                                 ¯                        (3)

where y is subsistence income and ε is a random disturbance term. Equation (3) implies
      ¯
that the rate of growth of population is related to the gap between actual incomes and
the subsistence level of income, and when ρ < 1, there is also mean reversion in popula-
                                                                                  ∗
tion. As long as ρ < 1, there will be a steady-state level of income per capita, yj , and
steady-state population density, p∗ , both strictly increasing in productivity Aj . Therefore,
                                  j
cross-country variation induced by differences in productivity or technology will lead to
a positive association between population density and income.
      In contrast, consider two areas with the same productivity, Aj , but one of them has
higher population density because of differences in other factors captured by εj in equation
(3). Then, equation (2) implies that the country with higher population will have lower


                                                 12
income per capita. So there is an identification problem: it is unclear whether an area
with higher population density is in fact richer or not.
       When high population density corresponds to lower income because of transitory differ-
ences, we should observe a subsequent decline in population density in the more densely
settled areas–population there is above its long-run equilibrium level. We show be-
low that differences in population density before 1500 were highly persistent–the more
densely settled areas in 1500 were also more densely settled in 1000 or 1A.D.. This still
leaves differences in other factors, leading to permanent differences in population density,
so caution is required in interpreting population density as a proxy for income per capita.6
       The empirical evidence regarding the relationship between population density and in-
come is also less clear-cut than the relationship between urbanization and income. Figures
A3 and A4 in the appendix show that population density and income increased concur-
rently in many instances. Nevertheless, there is no similar cross-sectional relationship in
recent data, most likely because of the demographic transition–it is no longer true that
high population density is associated with high income per capita because the relation-
ship between income and the number of children has changed (e.g., Notestein, 1945, or
Cipolla, 1974).
       Despite these reservations regarding population density, we present results using pop-
ulation density, as well as urbanization, as a proxy for income per capita. This is motivated
by three considerations. First, population density data are often more extensive, so the
use of population density data is a useful check on our results using urbanization data.
Second, as argued by Bairoch, population density is closely related to urbanization, and
in fact, our measures are highly correlated. Third, variation in population density will
play an important role not only in documenting the reversal, but also in explaining it. In
any case, the relationship between population density in the past and income per capita
today parallels the relationship between urbanization in 1500 and income today.
   6
    This empirical framework also enables us to discuss another issue of interpretation. A common
current interpretation of Malthus is that population dynamics should take all countries down to the
subsistence level. This corresponds to the case in which ρ = 1 in our framework. In this case, more
productive areas (high Aj ) still have higher population densities, but all areas have the same long-run
equilibrium income per capita, yj = y . Although we could still use population density as a proxy for
                                       ¯
land productivity, it would not be a proxy for income per capita. However, we believe that the case
ρ < 1, which implies that there can be long-run differences in per capita income across countries, is more
plausible, and consistent with the evidence regarding differences in cross-country living standards before
the demographic transition.




                                                   13
                                    3   Hypotheses

In this section, we outline the two major hypotheses investigated in this paper.

3.1    The geography hypothesis

The geography hypothesis claims that geographic, climatic and ecological factors have
a first-order effect on prosperity and economic development. There are many different
versions of this hypothesis.
      Perhaps the most common is the view that climate has a direct effect on income
through its influence on work effort. This idea dates back to Machiavelli (1519) and
Montesquieu (1748). During the early 20th century, the geographer Ellsworth Huntington
(e.g. 1915, 1945) pursued this idea further, and even attempted experiments to show the
effect of climate on work effort. Both Toynbee (1934, volume 1) and Marshall (1890)
similarly emphasized the importance of climate, both on work effort, and more generally,
on productivity. One of the pioneers of development economics, Gunnar Myrdal (1968),
also placed considerable emphasis on the effect of geography on agricultural productivity.
He wrote, for example, that: “climate exerts everywhere a powerful influence on all forms
of life,” and that “serious study of the problems of underdevelopment... should take into
account the climate and its impacts on soil, vegetation, animals, humans and physical
assets– in short, on living conditions in economic development.”
      Jared Diamond has espoused a different theory of development, where the timing of
the Neolithic revolution has had a long lasting effect by determining which societies were
the first ones to develop strong armies, technology and even political and social organiza-
tion. For example, he states that: “...proximate factors behind Europe’s conquest of the
Americas were the differences in all aspects of technology. These differences stemmed ul-
timately from Eurasia’s much longer history of densely populated...” societies dependent
on food production (1997, p. 358). According to Diamond, differences in the nature and
history of food production, in turn, were due to the types of crops, domesticated animals,
and the axis of agricultural technology diffusion in different continents, all of which are
geographically determined characteristics.
      More recently, Jeffrey Sachs has argued for the importance of geography through its
effect on the disease environment, transport costs, and technology. He points out that
“Economies in tropical ecozones are nearly everywhere poor, while those in temperate
ecozones are generally rich” (2001, p. 1), and explains this by arguing that “Certain parts

                                             14
of the world are geographically favored. Geographical advantages might include access to
key natural resources, access to the coastline and sea– navigable rivers, proximity to other
successful economies, advantageous conditions for agriculture, advantageous conditions
for human health.” (2000, p. 30). He further suggests that “Tropical agriculture faces
several problems that lead to reduced productivity of perennial crops in general and of
staple food crops in particular” (2000, p. 32), and that “The burden of infectious disease
is similarly higher in the tropics than in the temperate zones” (2000, p. 32). Finally, Sachs
argues that the greater population in the temperate areas over the past centuries led to
more rapid advances in technologies appropriate for these areas relative to technologies
necessary for development in the tropics (2001, p. 3 and 2000, pp. 33-34).
      Despite important differences between these versions of the geography hypothesis, for
our purposes they all share the same persistence prediction: to the extent that geographic
factors do not change over periods of 500 years or more, countries that were relatively
rich 500 years ago should be relatively rich today. This prediction holds also when we
consider the period of European colonialism. If climate, transport costs, the timing of the
Neolithic revolution and diseases matter for income today, they should have mattered at
least as much before Europeans came into contact with the inhabitants of the colonies.

3.2    The institutions hypothesis

According to the institutions hypothesis, societies with a social organization that provides
encouragement for investment will prosper. There are also many different versions of this
view. John Locke ([1690], 1980) was perhaps the first to clearly articulate the importance
of property rights for production. Of land and productive assets, Locke wrote “...there
must of necessity be a means to appropriate them some way or other, before they can
be of any use, or at all beneficial to any particular man” (emphasis in the original, p.
10). He further argued that the main purpose of government was “the preservation of the
property of ... members of the society” (p. 47). Similarly, Adam Smith and Frederick von
Hayek, among many others, emphasized the importance of property rights for the success
of nations.
      The argument by some Marxist historians, for example Brenner (1976), on whether
the capitalist class had the social and economic power to make the transition to capitalist
agriculture is also related. Here the organization of the society, through its effect on the
organization of productive activity, determines how productive agriculture is and whether


                                             15
new technologies are adopted.
   More recently, economists and historians have emphasized the importance of institu-
tions that guarantee property rights. For example, Douglass North starts his 1990 book
by stating (p. 3): “That institutions affect the performance of economies is hardly con-
troversial,” and identifies property rights as a key aspect of good organization of society
and institutions (see also North and Thomas, 1973).
   For the purposes here, we take a “good” organization of society to correspond to
a cluster of institutions ensuring that a broad cross-section of the society has effective
property rights. We refer to this cluster as institutions of private property, and contrast
them with extractive institutions, where the majority of the population faces a high risk
of expropriation by the government, the ruling elite or other agents. Two requirements
are implicit in this definition of institutions of private property. First, institutions should
provide secure property rights, so that, as emphasized by Locke, those with productive
opportunities expect to receive returns from their investments, and are encouraged to
undertake such investments. The second requirement, which we believe is equally impor-
tant, is embedded in the emphasis on “a broad cross-section of the society” having the
opportunity to invest. A society in which a very small fraction of the population, for ex-
ample, a class of landowners, holds all the wealth and political power may not be the ideal
environment for investment, even if the property rights of this elite are secure. In such a
society, many of the agents with investment opportunities and the entrepreneurial human
capital may be those without the effective property rights protection. In particular, the
concentration of political and social power in the hands of a small elite implies that the
majority of the population does not have secure property rights, and probably risks being
held up by the powerful elite. This is also consistent with North and Weingast’s (1989,
p. 805-806) emphasis that what matters is: “... whether the state produces rules and
regulations that benefit a small elite and so provide little prospect for long-run growth,
or whether it produces rules that foster long-term growth”. Whether political power is
broad-based or concentrated in the hands of a small elite is crucial in evaluating the role
of institutions in the experiences of the Caribbean or India during colonial times, where
the property rights of a small elite were well enforced, but the majority of the population
had no civil rights or any effective property rights.
   The organization of society and institutions also persist (see, for example, the evi-
dence presented in Acemoglu, Johnson and Robinson, 2000). Therefore, the institutions


                                             16
hypothesis also suggests that, most of the time, societies that are prosperous today should
continue to be so in the future. However, if a major shock disrupts the organization of a
group of societies, we should expect much less persistence. Historical evidence suggests
that European colonialism not only disrupted the social organization of a large number
of countries, but in fact led to the establishment of extractive institutions in previously
prosperous areas and to the development of institutions of private property in previously
poor areas. Therefore, we suggest that there was an institutional reversal that resulted
from European colonialism.7 The institutions hypothesis, combined with the institutional
reversal, predicts a reversal in the relative rankings among countries affected by European
colonialism.

3.3      The institutional reversal

In this subsection, we argue that European colonialism had a fundamental effect on the
social organization of the colonies, causing an institutional reversal: previously prosperous
areas were more likely to end up with extractive institutions, while there often developed
institutions of private property in previously poor areas.
       Historical evidence is relatively unambiguous that Europeans used very different strate-
gies in different colonies and that European colonialism had a fundamental effect on the
organization of the colonial societies. While in a number of colonies such as the U.S.,
Canada, Australia, New Zealand, Hong Kong and Singapore, Europeans established in-
stitutions of private property, in many others they set up highly extractive institutions.
Examples of extraction by Europeans include the transfer of gold and silver from Latin
America in the 17th and 18th centuries and of natural resources from Africa in the 19th
and 20th centuries, the Atlantic slave trade, the plantation agriculture in the Caribbean,
Brazil and French Indochina, the rule of the British East India Company in India, and the
rule of the Dutch East India Company in Indonesia from the early 17th century on. The
distinguishing feature of the institutions that Europeans established in these countries
was a high concentration of political power in the hands of a few who used their power to
extract resources from the rest of the population. For example, the main objective of the
Spanish and the Portuguese colonization was to obtain silver, gold and other valuables
from America, and throughout they monopolized military power to enable the extraction
   7
    By institutional reversal, we simply mean that countries that were relatively prosperous before the
arrival of Europeans were more likely to end up with extractive institutions under European rule.



                                                  17
of these resources. The mining network set up for this reason was based on forced labor
and the oppression of the native population. Similarly, the British West Indies in the
17th and 18th centuries was controlled by a small group of planters (e.g., Williams, 1970,
Dunn, 1972). Political power was important to the planters in the West Indies, and to
other elites in the colonies dominated by plantation agriculture, because it enabled them
to force large masses of natives or African slaves to work for low wages.
       Europeans running the Atlantic slave trade, despite their small numbers, also had
disproportionate power in Africa. The consensus view amongst historians appears to be
that the slave trade fundamentally altered the organization of the society in Africa, leading
to state centralization and warfare as African states competed and expanded to supply the
European slavers. Manning (1990, p. 147) describes this situation as follows: “with the
allure of imported goods and the brutality of capture, slave traders broke down barriers
isolating Africans in their communities. Merchants and warlords spread the tentacles of
their influence into almost every corner of the continent. By the 19th century, much of
the continent was militarized; great kingdoms and powerful warlords rose and fell, their
fate linked to fluctuations in the slave trade.”8
       What determines whether Europeans pursued an extractive strategy or introduced
institutions of private property? And why was extraction more likely in relatively pros-
perous areas? Two factors appear crucial:

   1. The economic profitability of alternative policies: When extractive institutions were
         more profitable, Europeans were more likely to opt for them. High population
         density, by providing a supply of labor that could be forced to work in agriculture
         or mining, made the development of extractive institutions, with poltical power
         concentrated in the hands of a small elite, more profitable. For example, the presence
         of abundant Amerindian labor in the Meso-America region was conducive to the
         establishment of large plantation complexes, while the high population of West
         Africa created a profit opportunity for slave traders, supplying additional labor to
         the plantations in the Americas. The experience of the Spanish conquest around
         the La Plata river (current day Argentina) during the early 16th century gives an
         interesting example of how population density affected European colonization (see
         Lockhart and Schwartz, 1983, pp. 259-60, for details). Early in 1536, a very large
   8
   There are many detailed studies of different parts of Africa supporting this claim. See for example,
Wilks (1975) for Ghana, Low (1977) for Nigeria, Harms (1981) for the Congo/Zaire, Miller (1988) on
Angola.

                                                 18
       Spanish expedition arrived in the area, and founded the city of Buenos Aires in the
       mouth of the river Plata. The area was sparsely inhabited by non-sedentary Indians.
       The Spaniards could not enslave a sufficient number of Indians for food production.
       Starvation forced them to abandon Buenos Aires and retreat up the river to a
       post at Asuncion (current day Paraguay). This area was more densely settled by
       semi-sedentary Indians. These Indians were enslaved by the Spaniards for food
       production, and the colony of Paraguay, with relatively extractive institutions, was
       founded. In contrast, Argentina was finally colonized later, with greater European
       settlements, no forced labor, and generally better institutions.
       Other types of extractive institutions were also more profitable in densely-settled
       and prosperous areas, where there was more to be directly plundered by European
       colonists. More important, in these densely-settled areas there was an existing
       system of tax administration or tribute, and the large population made it profitable
       for the Europeans to take control of the existing tax and tribute systems and to
       increase taxes on the native population (see, e.g., Wiegersma, 1988, p. 69, on
       French policies in Vietnam, or Marshall, 1998, pp. 492-497, on British policies in
       India).

   2. Whether Europeans could settle or not: Europeans were more likely to develop in-
       stitutions of private property when they settled in large numbers, for the natural
       reason that they themselves were affected by these institutions.9 Moreover, when a
       large number of European settled, the lower strata of the settlers demanded rights
       and protection similar to, or even better than, those in the home country. This
       made the development of effective property rights for a broad cross-section section
       of the society more likely. European settlements, in turn, were affected by popu-
       lation density both directly and indirectly. Population density had a direct effect
       on settlements, since Europeans could easily settle in large numbers in sparsely in-
       habited areas. The indirect effect, on the other hand, worked through the disease
       environment. In many of the densely settled areas, there were diseases–in partic-
       ular, malaria and yellow fever–for which Europeans did not have an immunity.10
   9
     Extraction and European settlement patterns were mutually self-reinforcing. In areas areas where
extractive policies were pursued, the authorities also actively discouraged settlements by Europeans,
presumably because this would intervene with the effective extraction of resources from the locals (e.g.
Coatsworth, 1982).
  10
     Most diseases require a dense human population to act as carriers. In Acemoglu, Johnson and


                                                  19
        European settlements shaped both the type of institutions that developed and the
        structure of production. For example, while in Potos´ (Bolivia) mining employed
                                                            ı
        forced labor (Cole 1985), and in Brazil and the Caribbean sugar was produced by
        African slaves, in the U.S. and Australia mining companies employed free migrant
        labor, and sugar was grown by smallholders in Queensland, Australia (Denoon,
        1983). Consequently, in Bolivia, Brazil and the Caribbean, political institutions
        were designed to ensure the control of the laborers and slaves, while in the U.S.
        and Australia, the workers and smallholders had greater political rights (Cole 1985,
        Hughes 1988, chapter 10).

      The historical evidence is therefore consistent with our notion that European colonial-
ism, by introducing worse institutions in previously densely-settled and prosperous areas
and developing institutions of private property in previously poor regions, caused an in-
stitutional reversal. We further substantiate this view empirically in the next subsection.

3.4     Econometric evidence on institutional reversal

Table 3 shows the relationship between urbanization or population density in 1500 and
institutions using four different measures of institutions. The first two measures refer to
current institutions: protection against expropriation risk between 1985 and 1995 from
Political and Risk Services, which approximates how secure property rights are, and con-
straints on the executive in 1990 from Gurr’s Polity IV data set, which can be thought
of as a proxy for how concentrated power is in the hands of ruling groups. Columns 1-6
of Table 3 show a negative relationship between our measures of prosperity in 1500 and
current institutions.
      It is perhaps more relevant to know whether there is such an institutional reversal
during the colonial times or shortly after independence. Since the Gurr data set does
not contain information for non-independent countries, we can only look at this after
independence. Columns 7-12 show the relationship between our measures of prosperity
in 1500 and measures of early institutions. The first measure we use is constraints on the
Robinson (2000), we documented that Europeans were less likely to settle in areas infected by malaria
and yellow fever, and both of these diseases were absent in sparsely settled areas such as Australia or New
Zealand. The diseases in the New World did not initially cause high European mortality, but malaria
and yellow fever developed soon after African slaves were brought to the continent. In any case, the
correlation between population density and the European settler mortality variable we constructed in
Acemoglu, Johnson and Robinson (2000) is 0.33.



                                                    20
executive in 1900. Since colonial rule often concentrated political power in the hands of a
small elite, it seems reasonable to assign the lowest score to countries still under colonial
rule in 1900. As an alternative variable, we use constraints of the executive in the first year
of independence from the same data set, while also controlling for time since independence
as an additional covariate. Notice that when both urbanization and log population density
in 1500 are included, it is the population density variable that is significant. This supports
the interpretation that it was the differences between densely and sparsely settled areas
that was crucial in determining what type of institutions Europeans introduced (though
it may also reflect the fact that the population density variable is measured with less
measurement error). Finally, the second panel of the table includes (the absolute value
of) latitude as an additional control, showing that the institutional reversal does not
reflect some geographic pattern of institutional change.
   Figure 4 shows the negative relationship between constraints on the executive in the
first year of independence and our measures of prosperity in 1500 diagrammatically (the
effect of time since independence is controlled for by orthogonalizing both the left-hand
side and the right-hand side variables with respect to time since independence). Many
of the colonies such as Canada, the United States, New Zealand and Australia that were
relatively poor before Europeans arrived became independent with relatively good insti-
tutions, which contrasts with the experience of many Latin American, African and Asian
countries, which were relatively prosperous in 1500, but inherited extractive institutions
from colonial powers.

     Table 3: Institutions vs. urbanization in 1500 and population density in 1500
                                          Figure 4

   Overall, we interpret both the historical and the econometric evidence as supporting
the notion that European colonialism caused an institutional reversal. Therefore, the
institutions hypothesis suggests the possibility of a reversal in relative prosperity among
European colonies between 1500 and today, while the geography hypothesis predicts a
high degree of persistence during the same period.




                                             21
                               4    The Reversal of Fortune
4.1     Results with urbanization

This section presents our main results. Figure 1 in the introduction depicts the relation-
ship between urbanization 1500 and income per capita today. Table 4 reports regressions
documenting the same relationship. Column 1 is our most parsimonious specification,
regressing log income per capita in 1995 (PPP basis) on urbanization rates in 1500 for
our sample of former colonies.11 The coefficient is -0.08 with a standard error of 0.03.
This coefficient implies that a 10 percent lower urbanization in 1500 is associated with
approximately twice as high GDP per capita today (70 log points≈100 percent).12 It is
important to note that this is not simply mean reversion–i.e., richer than average coun-
tries reverting back to the mean. It is a reversal. To illustrate this, let us compare Chile
and India. The native population in Chile basically had no urbanization, while India had
an urbanization rate of at least 8.5 percent (though there are also estimates as high as 15
percent for urbanization in India, see Bairoch, 1988). The estimate in column 1 of Table
2, 0.038, for the relationship between income and urbanization implies that India at the
time was at least 33 percent richer than Chile (if we take higher estimates of urbanization
in India, this gap would be correspondingly greater). Today according to our estimate in
column 1 of Table 4, we expect Chile to be approximately 70 percent richer than India
(in practice, Chile is about 6 times as rich as India).

                                            Table 4 here.

       The second column of Table 4 excludes North African countries for which data quality
may be lower. The result is unchanged, with a coefficient of -0.11 and standard error of
0.03. Column 3 drops the Americas, which increases both the coefficient and the standard
error, but the estimate remains highly significant. Column 4 adds continent dummies,
to check whether the relationship is being driven by differences across continents. Al-
  11
     Here we limit ourselves to looking at log GDP per capita in 1995 as the dependent variable. For com-
pleteness, the top panel of Table 8 shows the relationship between urbanization in 1995 and urbanization
in 1500.
  12
     Because China was never a formal colony, we do not include it in our sample of ex-colonies. Adding
China strengthens the results further. For example, with China, the baseline estimates changes from
-0.0815 (s.e.=0.0258) to -0.0822 (s.e.=0.0255). Furthermore, our sample excludes counties that were
colonized by European powers briefly during the 20th century, such as Iran, Saudi Arabia and Syria. If
we include these observations, the results are unchanged. For example, the baseline estimate changes to
-0.074 (s.e.=0.024).



                                                   22
though continent dummies are jointly significant, the coefficient on urbanization in 1500
is unaffected: it is -0.09 with a standard error of 0.03.
    One might also be concerned that the relationship is being driven mainly by the Neo-
Europes; USA, Canada, New Zealand and Australia. These countries are settler colonies
built on lands that were inhabited by relatively undeveloped civilizations. Although the
contrast between the development experiences of these areas and the relatively advanced
civilizations of India or Central America is of independent interest, one would like to know
whether there is anything more than this contrast in the results of Table 4. In column
5, we drop these observations. The relationship is now weaker, but still negative and
statistically significant at the 6 percent level.
    Is the reversal in relative incomes related to geography? To investigate this issue,
in columns 6 and 7, we add a variety of geography-related variables, including distance
from the equator, and differences in temperature, humidity, soil quality, natural resource
endowments. These variables themselves are insignificant, and have almost no effect on
the pattern of the reversal. This indicates that the reversal is not apparently related to
geography.
    Finally, in columns 8 and 9, we add the identity of the colonial power (or legal origin),
which is emphasized by Hayek (1960) and La Porta et al. (1998) as important determi-
nants of development, and religion, stressed by Max Weber. These also have little effect
on our estimate.
    The urbanization variable used in Table 4 relies on the work by Bairoch and Eggi-
mann, and as explained in the previous section, we had to make a number of assumptions
to combine estimates from these researchers. In Table 5, we use separately data from
Bairoch and Eggimann, as well as data from Chandler, who provided the starting point
for Bairoch’s data. We repeat the regressions of Table 4 using these three alternative
series. The results are very similar to the baseline estimates reported in Table 5: in all
cases, there is a negative relationship between urbanization in 1500 and income per capita
today, and in almost all cases, this relationship is statistically significant at the 5 percent
level.

                                        Table 5 here.




                                              23
4.2    Results with population density

A big gap in our urbanization data is for sub-Saharan Africa. Bairoch (1988) refers to our
knowledge regarding Africa as: “Black Africa: An Urban History That Remains to Be
Written” (p. 52). Nevertheless, we have population density data for these countries. In
Panel A of Table 6, we regress income per capita today on population density in 1500. The
results are quite similar to those in Table 4. In all specifications, we find that countries
with higher population density in 1500 are substantially poorer today. The coefficient of
-0.38 in column 1 implies that a 10 percent higher population density in 1500 is associated
with a 4 percent lower income per capita today. For example, the area now corresponding
to Bolivia was seven times more densely settled than the area corresponding to Argentina,
so on the basis of this regression, we expect Argentina to be three times as rich as Bolivia,
which is more or less the current gap in income between these countries.

                                       Table 6 here.

      The remaining columns perform the same robustness checks as in Table 4 and 5,
and show that including geography-related variables, the identity of the colonial power,
religion variables, or dropping the Americas, the Neo-Europes, or North Africa has very
little effect on the results.
      The estimates in the top panel of Table 6 use variation in population density. This
reflects two components; differences in population and differences in arable land area.
In Panel B, we separate the effects of these two components, and find that they come
in with equal and opposite signs. This suggests that our specification with population
density alone was appropriate.
      Finally, in Panel C we estimate the same relationships as in Panel A, but using pop-
ulation density in 1000 as an instrument for population density in 1500. This is useful
since, as discussed in Section 2.4, it is differences in long-run population density that are
likely to be better proxies for income per capita. Instrumenting for population density
in 1500 with population density in 1000 isolates the long-run component of population
density differences across countries (i.e., the component of population density 1500 that is
correlated with population density in 1000). The Two-Stage Least Squares (2SLS) results
in Panel C using this instrumental variables strategy are very similar to the OLS results
in Panel A.



                                             24
4.3      Further results, robustness checks and discussion

Caution is required in interpreting the results presented in Tables 4-6. Estimates of
urbanization and population more than 500 years ago are likely to be ridden with error.
Nevertheless, the first effect of measurement error would be to create an attenuation bias
towards 0. Therefore, one might think that the negative coefficients in Tables 4-6 are, if
anything, underestimates. A more serious problem would be if errors in the urbanization
and population density estimates were not random, but correlated with current income in
some systematic way. So it is important to ensure that our results are robust to plausible
variations in data. The results with alternative urbanization estimates in Table 5 suggest
that this is not major problem. We investigate this issue further in Table 7. We use
a variety of different estimates on urbanization and population density, such as those
assigning lower urbanization and population density numbers to the Americas, North
Africa and India. The results are robust to these changes. Panel B of this table also
reports results weighted by population in 1500, with very similar results.
       Much of the variation in urbanization or population density is not at the country level,
but at the level of “civilizations”. For example, in 1500 there were many fewer separate
civilizations in the Americas, and perhaps even in Asia. For this reason, we repeat our
key regressions using variation only at the civilization level. First, in column 7 we use
only data corresponding to the areas occupied by the 7 civilizations that are identified
by Arnold Toynbee in A Study of History and are in our ex-colonies sample. In column
8, we then use an extended classification with 14 civilizations.13 The results confirm our
basic findings. With only the 7 civilizations identified by Toynbee, there is a negative, but
statistically insignificant, relationship (which is natural given the number of observations).
When we use all 14 civilizations, this relationship becomes significant at the 5 percent
level.

                                          Table 7 here.

       Finally, in the bottom panel of Table 7 we include urbanization and population density
simultaneously in these regressions. In all cases, population density is negative and highly
significant, while urbanization is insignificant. This is again consistent with the notion
  13
     The 7 civilizations identified by Toynbee which are relevant to our study, are Andean, Yucatec,
Mexican, Arabic, Hindu, Eskimos and Polynesian. We follow McNeill (1967) and add to this list, North
American Indian, Non-Andean South American Indian, Australian Aboriginal, Malaysian-Indonesian,
Filipino, Vietnam-Cambodia-Laos, Caribbean to arrive to 14 civilizations.


                                                25
that differences in population density played a key role in the institutional reversal, and
therefore in causing the reversal in relative incomes among the colonies (though it may
also reflect measurement error in the urbanization estimates).
   Is the reversal shown in Figures 1 and 2 and Tables 4, 5 and 6 consistent with other
evidence? There seems to be little doubt from the literature on the history of civilizations,
e.g. Toynbee (1934-61), and also from recent qualitative assessments, e.g. McNeill (1967),
that 500 years ago, many parts of Asia were highly prosperous (perhaps as prosperous
as Western Europe), and civilizations in Meso-America and North Africa were relatively
developed. In contrast, there was little agriculture in most of North America, Australia
and New Zealand, at most consistent with population density of 0.1 people per square
kilometer (see the map in McEvedy and Jones p. 273, reproduced below in the Appendix).
McEvedy and Jones (1975, p. 322) describe the state of Australia at this time as “an
unchanging palaeolithic backwater”. In fact, because of the relative backwardness of these
areas, European powers did not view them as valuable colonies. Voltaire is often quoted
as referring to Canada as “few acres of snow”, and the European powers at the time paid
little attention to Canada relative to the colonies in the West Indies. In a few parts of
North America, along the East Coast and in the South-West, there was settled agriculture,
supporting a population density of approximately 0.4 people per square kilometer, but
this was certainly much less than that in the Aztec and Inca Empires, which had a fully
developed agriculture with a population density of between 1 and 3 people (or even higher)
per kilometer squared, and also much less than the corresponding numbers in Asia and
Africa.
   Overall we conclude that the evidence points to a reversal in relative incomes among
former European colonies. This reversal is inconsistent with geographic factors being
the major cause of income differences across countries. But it is consistent with the
institutions hypothesis. As argued in the previous section, historical and econometric
evidence suggests that European colonialism led to an institutional reversal, creating
worse institutions in previously prosperous areas. Therefore, if institutions are a crucial
determinant of economic performance, we should expect a reversal in relative economic
prosperity. In this light, we interpret the evidence in this section as providing support to
the institutions hypothesis against the simple geography hypothesis.




                                             26
4.4    Persistence before 1500 and Persistence among the non-colonies

Like the geography hypothesis, the institutions view predicts persistence in economic
prosperity as long as there is no major shock to the social organization of the societies
in question. It is therefore instructive to briefly look at patterns of persistence among
non-colonized countries and among our sample of ex-colonies during the period before
1500. Table 8 reports a range of relevant results.

                                       Table 8 here.

      Two important patterns emerge from this table. First, among European countries, or
all countries that were not colonized by Western Europe (including European countries),
there is persistence between 1500 and today (columns 6-9). Second, both in the sample of
non-colonized countries and in our sample of former colonies, there is persistence between
1000 and 1500, or even between 1AD and 1000 when we use log population density
(columns 1-5). These results suggest that the reversal we document in this section indeed
reflects an unusual event, and gives further support to the idea that the reversal is related
to European colonialism.

            5   The Northern Drift, The Timing of the Reversal and
                                Industrialization

The evidence presented so far weighs against the view that geography is a major deter-
minant of economic outcomes. Nevertheless, the reversal does not necessarily reject a
more sophisticated geography view, which emphasizes the Northern (or away from the
equator) shift in the center of economic gravity over time. According to this view, which
we call “the Northern drift” hypothesis, geography becomes important when it interacts
with the presence of certain technologies. For example, one can argue that tropical areas
provided the best environment for early civilizations–after all humans evolved in the
tropics and the required calorie intake is lower in warmer areas. But with the arrival of
“appropriate” technologies, temperate areas became more productive (e.g., White, 1962).
The technologies that were crucial for the rise of civilizations in temperate areas include
the heavy plow, systems of crop rotation, domesticated animals such as cattle and sheep,
and some of the high productivity European crops, such as wheat and barley. Despite the
importance of these technologies for development in the temperate areas, they have had
much less of an effect on tropical zones. Although we are not aware of any social scientist

                                            27
who has explicitly formulated this view, Jeffrey Sachs implies it in his recent paper when
he writes “Since technologies in the critical areas of agriculture, health, and related areas
could diffuse within ecological zones, but not across ecological zones, economic develop-
ment spread through the temperate zones but not through the tropical regions” (2001, p.
12, italics in the original).
       The reversal in relative incomes does not rule out an explanation for the reversal based
on the Northern drift hypothesis. Many of the colonies that subsequently became rich,
such as North America or Australia, were situated in temperate climates. The populations
in these areas did not have access to the agricultural technology in use in Europe at the
time, and the spread of this technology, which came with European colonialism, may have
enriched these areas, causing the reversal.
       The Northern drift view is plausible in the context of agricultural technology, since
different climates favor different types of crops and harvesting methods. Yet there seems
to be no compelling case that climate and geography should matter more for industry
than for agriculture. Therefore, the Northern drift hypothesis suggests that the reversal
should be associated with the spread of European agricultural technologies.14
       The timing and the nature of the reversal in relative incomes do not support the
Northern drift hypothesis, however. First, the results presented so far give no indication
that the reversal is related to any geographic characteristics (e.g., see Table 4). Perhaps
more important, European agricultural technology spread to the colonies between the
16th and 18th centuries (see for example McCusker and Menard, 1985, Chapter 3 for
North America). So the Northern drift hypothesis would suggest that the reversal should
take place during this period. In contrast, the evidence indicates that the reversal in
relative incomes is largely a 19th-century phenomenon. This is documented in Figures 5
and 6.
       Figure 5 shows that among the ex-colonies there appears to be no reversal between
1500 and 1700, while the reversal is quite apparent between 1700 and today. Figure
6 looks in detail at the evolution of urbanization in a number of countries to date the
timing of the reversal. In Figure 6a, we classify ex-colonies according to their level of
  14
    There is another line of argument which might suggest that geography matters more for industry than
for agriculture. Imagine there is more room for specialization in industry, but this requires trade, and
countries differ according to their transport costs. This may create an advantage for countries with low
transport costs once industrialization becomes possible. This argument is not very convincing, however,
since many of the previously prosperous colonies that failed to industrialize include islands such as the
Caribbean, or countries with natural ports such as those in Central America, India or Indonesia.



                                                   28
urbanization in 1500.15 The figure shows the higher level of urbanization among the
previously high urbanization colonies until 1850, and then a big surge in urbanization
among the previously low urbanization colonies starting in 1850. This is largely driven by
the U.S., but also by Canada, and Argentina. Figure 6b shows a comparison of Mexico,
India and the U.S. as three important and large colonies, reiterating the pattern in Figure
6a. Figure 6c shows the comparison of Mexico, the U.S., Brazil and Peru, this time using
data from Mitchell (1983) and (1995).16
       Finally, Figure 6d shows per capita industrial production for the U.S., Canada, New
Zealand, Australia, Brazil, Mexico, and India using data from Bairoch (1982). This figure
shows the takeoff in industrial production in the U.S., Australia, Canada and New Zealand
relative to Brazil, Mexico and India. Although it is difficult to see it in the figure, in fact
in 1750, per capita industrial production was higher in India, 7, relative to the U.S., 4
(these numbers are an index with UK industrial production per capita in 1900 normalized
to 100). Bairoch (1982) also reports that in 1750, China had industrial production per
capita twice the level of the U.S., 8. Yet, as Figure 6d shows, over the next 200 years,
there was a much larger increase in industrial production in the U.S. than in India (and
also than in China).
       This general interpretation that the reversal in relative incomes took place during the
19th century, and was linked to industrialization, is also consistent with the fragmentary
evidence we have on other measures of income per capita and industrialization. For
example, Coatsworth (1993), Eltis (1995) and Engerman and Sokoloff (1997) provide
evidence that much of Spanish America and the Caribbean was more prosperous (higher
per capita income) than British North America until 1700. The future United States rose
in per capita income during the 1700s relative to the Caribbean and South America, but
only really pulled ahead in the 100 years after 1800 (Coatsworth, 1993). McCusker and
Menard (1985) and Galenson (1996) both emphasize that productivity and income growth
  15
     Former colonies with high urbanization in 1500 are Ecuador, Mexico, Guatemala, Honduras,
Nicaragua, Panama, El Salvador, Belize, Algeria, Tunisia, Egypt, Morocco, and India. Former colonies
with low urbanization in 1500 are Argentina, Brazil, Canada, Chile, Paraguay, Uruguay, and the U.S.
The data used in this graph are from Chandler through 1861 and Bairoch for the twentieth century. Aus-
tralia and New Zealand are not included in this graph, since they were settled only in the 19th century.
Including these countries, assuming constant urbanization before European settlements, does not affect
the pattern shown in the figure.
  16
     Mitchell provides a continuous series at approximately 10 year intervals for urbanization from the
late 18th to 20th century. In contrast most of Chandler’s data end in 1850, Bairoch reports estimates
only for every 50 years, and Eggimann does not provide numbers for some key countries, such as the
U.S.. Where they overlap, Mitchell’s estimates are consistent with those from our other sources.


                                                  29
in North America before the 18th century was limited.17 Using data from Maddison
(1995), Pritchett (1997) documents the large divergence in per capita income from 1870
onwards.
       The available data also suggest that during the critical period of growth in the U.S.,
between 1840 and 1900, there was modest growth in agricultural output per capita, and
very rapid growth in industrial output per capita. Gallman (2000) reports that in 1840,
agriculture was 41 percent of GDP, while manufacturing was 17 percent of GDP. By 1900,
these numbers had changed to 19 and 31 respectively. He also reports that GNP per capita
in 1860 dollars was approximately 96.5 around 1840, and 269.1 around 1900. These
numbers imply that between 1840 and 1900, agricultural product per capita increased
roughly from 39.5 to 51.1, which is very small relative to the increase in manufacturing
output per capita, which went up from 16.4 in 1840 to 83.4 in 1900.
       Overall this evidence suggests that previously poor colonies became relatively rich
during the 19th century, and this was associated with rapid industrialization in these
areas. We interpret this pattern as evidence against the Northern drift hypothesis.18 We
next present econometric evidence that is consistent with the role of institutions in the
reversal.19

                                          Figures 5 and 6.
  17
     McCusker and Menard (1985, p. 270) argue “although there was some gains in agriculture, man-
ufacturing and shipping, improvments in technology had only a minor impact on the colonial economy
[1607-1785]....improvements in economic organization, specifically shifts away from self-sufficiency and to-
ward production for exchange, and the development of more efficient markets, were the principal sources
of growth in the colonies of British America.”
  18
     The timing and nature of the reversal also weigh against two other potential hypotheses. The first,
inspired by Williams (1944), Rodney (1972), Frank (1978) and Wallerstein (1974-1980), would be that
the Europeans extracted resources from, or more generally plundered, the formerly rich colonies, and
this plunder may have caused the reversal. Since the extraction of resources was most intense shortly
after initial colonization, this hypothesis would also suggest an early reversal. An alternative possible
thesis is that the reversal reflects the direct effect of the diseases Europeans brought to the New World
(e.g., McNeill, 1976, Crosby, 1986, Diamond, 1997). Since the remaining inhabitants of the Americas had
developed immunities to these diseases by the 17th century, this view also suggests we should observe a
relatively early timing for the reversal.
  19
     Another hypothesis with a flavor similar to the geography views would be that different paths of
development reflect the direct influence of Europeans. Places where there are more Europeans have
become richer, either because Europeans brought certain values conducive to development (e.g. Landes,
1998), or because having more Europeans confers certain benefits (e.g., through trade with Europe or
because Europeans are more productive). Since Europeans settled in greater numbers in previously
sparsely settled areas, the reversal may reflect this effect. In Acemoglu, Johnson and Robinson (2001)
we show that once the effect of institutions are taken into account appropriately, the current racial
composition of a country has no effect on income. So we do not pursue this explanation further here.


                                                   30
         6   Institutions and The Making of the Modern World Income
                                   Distribution

We have so far documented the reversal in economic prosperity among former European
colonies between 1500 and today, and argued that this is inconsistent both with the sim-
ple and sophisticated geography views. We have also linked the reversal in economic
performance to the institutional reversal among these countries caused by European colo-
nialism. The next step is to argue that this institutional reversal and the institutional
differences between these countries account for their different economic trajectories, and
to investigate the mechanism through which institutions have affected economic develop-
ment. In this section, we first show that institutional differences statistically account for
the reversal between 1500 and today. We next document that institutions started playing
a much more important role in the age of industry.

6.1    Institutions and the reversal

If the institutional reversal is the reason why there was a reversal in income levels among
the former colonies, then once we account for the role of institutions appropriately, the
reversal should disappear. That is, according to this view, the reversal documented in
Figures 1 and 2 and Tables 4-7 reflects the correlation between economic prosperity in
1500 and income today working through the intervening variable, institutions.
      How do we establish that an intervening variable X is responsible for the correlation
between variable Z and Y ? To answer this question, suppose that the true relationship
between Y , and X and Z is
                                    Y = α·X +β ·Z +ε                                    (4)

where α and β are coefficients and ε is a disturbance term. In our case, we can think of
Y as income per capita today, X as measures of institutions, and Z as population density
(or urbanization) in 1500. The variable Z is included in equation (4) either because it has
a direct effect on Y or because it has an effect through some other variables not included
in the analysis. The hypothesis we are interested in is that β = 0–that is, the only
channel through which population density or urbanization has an effect on income today
is through institutions.
      This hypothesis obviously requires that there is a statistical relationship between X
and Z. So we postulate that
                                       X = λ · Z + ν.

                                             31
To start with, suppose that ε is independent of X and Z and that v is independent of Z.
Now imagine a regression of Y on Z only (in our context, of income today on prosperity
in 1500), similar to those we reported in Tables 4-7:

                                            Y = b · Z + u1 .

As is well-known, the probability limit of the OLS estimate from this regression, b is
                                                                                  b,

                                          plimb = β + α · λ.
                                              b

So the results in the regressions of Tables 4-7 are consistent with β = 0 as long as α 6= 0
and λ 6= 0. In this case, we could be capturing the effect of Z (population density or
urbanization) on income working solely through institutions. This is the hypothesis that
we are interested in testing. Under the assumptions regarding the independence of Z from
ν and ε, and of X from ε, there is a simple way of testing this hypothesis, which is to run
an OLS regression of Y on Z and X:

                                       Y = a · X + b · Z + u2 ,                                       (5)

to obtain the estimates a and b The fact that ε in (4) is independent of both X and
                        b     b.
Z rules out omitted variable bias, so plima = α and plimb = β. Hence, a simple test of
                                          b             b
whether b = 0 is all that is required to test our hypothesis that the effect of Z is through
        b
X alone.
       In practice, there are likely to be many omitted variables, endogeneity bias because Y
has an effect on X, and attenuation bias because X is measured with error or corresponds
poorly to the real concept that is relevant to development.20 So the above procedure is
not possible. However, the exact logic applies as long as we have a valid instrument, M ,
for X, such that
                                           X = γ · M + ζ,

and M is independent of ε in (4). We can then simply estimate (5) using Two-Stage Least
Squares (2SLS) with the first-stage

                                      X = c · M + d · Z + u3 .
  20
    For example, plausibly what matters for economic development is a broad range of institutions,
whereas we only have an index for a particular type of institutions, e.g., enforcement of private property
rights.



                                                   32
Testing our hypothesis that Z has an effect on Y only through its effect on X, then
amounts to testing that the 2SLS estimates of b, e is equal to 0. Intuitively, the 2SLS pro-
                                                 b
cedure ensures that we obtain a consistent estimate of α, thus enabling us to appropriately
test whether Z has a direct effect.
       The key to the success of this strategy is obviously a good instrument for X. Here we
refer to our previous work, Acemoglu, Johnson and Robinson (2001), where we suggested
and documented that mortality rates faced by settlers are a good instrument for settle-
ments of Europeans in the colonies and the subsequent institutional development of these
countries. Table 9 reports results from this type of 2SLS tests using the log of potential
settler mortality rates as an instrument for institutional development.21

                                           Table 9 here

       Panel A reports results from regressions that enter urbanization and log population
density in 1500 as separate (exogenous) regressors in the first and the second stages, while
Panel B reports the corresponding first stages. Different columns correspond to different
institutions variables, or different specifications. For comparison, Panel C reports the
2SLS coefficient with exactly the same sample as the corresponding column, but without
urbanization or population density. We look at the same four institutions variables used
in Table 3; protection against expropriation risk from Political and Risk Services, and
constraints on the executive in 1990, in 1900 and in the first year of independence from
Gurr’s Polity IV data set. The results are consistent with our hypothesis. In all columns,
we never reject the hypothesis that urbanization 1500 or population density in 1500 have
no direct effect once we control for the effect of institutions on income per capita, and
the addition of these variables has little effect on the 2SLS estimate of the effect of
institutions on income per capita. This supports our notion that the reversal in economic
prosperity reflects the effect of early prosperity and population density working through
the institutions and policies introduced by European colonialists.

6.2     Institutions and industrialization

The historical evidence suggests that the institutional reversal–the introduction of ex-
tractive institutions in previously prosperous areas and institutions of private property in
  21
    See Acemoglu, Johnson and Robinson (2001) for details of how the settler mortality data are con-
structed and supporting evidence that it is a valid instrument for differences in institutions.



                                                33
previously poor areas–was well underway by the 19th century. So why did the reversal
take place during the 19th century?
   To answer this question, let us first reiterate that institutions of private property
require a broad cross-section of the society to have effective property rights. Imagine a
society similar to the colonies in the Caribbean during the 17th and 18th centuries where
a small elite controls all the political power. The revolution in Haiti notwithstanding, the
property rights of this elite are relatively well protected, but the rest of the population
has no effective property rights. According to our definition, this is not a society with
institutions of private property. Nevertheless, when the major investment opportunities
are in agriculture, this lack of effective private property enforcement for the majority of
the population may not matter too much, since the elite can invest in the land and employ
the rest of the population, and so will have relatively good incentives to increase output.
Now imagine the arrival of a new technology, for example, industrial opportunities. If the
elite could undertake industrial investments without losing its political power, we may
expect them to take advantage of these opportunities. However, in practice there will be
two major problems. First, the people with the entrepreneurial skills and ideas may not
be members of the elite. These potential entrepreneurs will not undertake the necessary
investments, because they do not have secure property rights. They correctly anticipate
that they will be held up once they undertake these investments because political power
still rests in the hands of the elite. Second, the elite may want to block investments in
new industrial activities, fearing the political turbulence and the threat to their political
power that new technologies will bring (see Acemoglu and Robinson, 2001).
   This reasoning suggests that whether a society has institutions of private property or
extractive institutions may matter much more when new technologies require a broad-
based economic participation. Industrialization is precisely such a process, requiring in-
vestments from a large number of agents who were not previously part of the ruling elite.
Therefore, there are natural reasons to expect that institutional differences will matter
much more during the age of industry. This is consistent with the evidence discussed in
the last subsection. We now investigate this issue in more detail.
   If this hypothesis is correct, we should expect societies with good institutions to take
better advantage of industrialization opportunities. We can test this using data on institu-
tions, industrialization and GDP from the 19th and early 20th centuries. Bairoch (1982)
presents detailed evidence on industrial output for a number of countries at a variety of


                                             34
dates, and Maddison (1995) has estimates of GDP for a larger group of countries. We take
Bairoch’s estimates of UK industrial output as a proxy for industrialization opportunities
for the whole world, since during this period the UK was the industrial leader. We then
run a panel data regression of the following form:

                        yit = δt + δi + π · Xit + φ · Xit · UKIN Dt + εit ,                   (6)

where yit is the outcome variable of interest in country i at date t. We consider industrial
output per capita and income per capita as two different measures of economic success
during the 19th century. In addition, δt ’s are a set of time effects and δi ’s denote a set
of country effects, Xit denotes the measure of institutions in country i at date t, and
U KINDt is industrial output per capita in the UK at date t. The coefficient of interest
is φ, which reflects whether there is an interaction between good institutions and industri-
alization opportunities. A positive and significant φ is interpreted as evidence in favor of
the view that countries with institutions of private property took better advantage of the
opportunities to industrialize. The parameter π measures the direct effect of institutions
on industrialization, and is evaluated at the mean value of UKIN Dt .
       The top panel of Table 10 reports regressions of equation (6) with industrial output
per capita as the left-hand side variable. Here we have data for the U.S., Canada, New
Zealand, Australia, South Africa, Brazil, Mexico, and India. The bottom panel considers
log GDP per capita as the dependent variable, and in addition, includes data also from Ar-
gentina, Bangladesh, Burma/Myanmar, Chile, Colombia, Egypt, Ethiopia, Ghana, India,
Indonesia, Ivory Coast, Kenya, Morocco, Nigeria, Pakistan, Peru, Tanzania, Venezuela,
and Zaire.22      We also have data for the following dates: 1830, 1860, 1880, 1900, 1913,
1928, 1953, and 1980 (though not for all countries for all dates).
       Column 1 reports a regression of (6) using only pre-1950 data. The interaction term, φ,
is estimated to be 0.13, and is highly significant with a standard error of 0.03. To calculate
the implication of this estimate, note that Bairoch’s estimate of total UK industrialization
rose from 16 to 115 between 1800 and 1913 (UK industrial production per capita in 1900
normalized to 100). In the meantime, the U.S. per capita production grew from 9 to
126, whereas India’s per capita industrial production fell from 6 to 2. Since the average
difference between the constraints on the executive in the U.S. and India over this period
is approximately 6, the estimate implies that the U.S. industrial output per capita should
  22
    In the top panel we did not use logs, since industrial output per capita was 0 for some of the
observations.

                                                35
have increased by 77 more than India’s, which is over half the actual difference. If instead
we use the instrumental variables (IV) estimate from column 7, 0.25, the predicted increase
turns out to be somewhat larger than the actual increase.
      In column 2, we drop the country dummies, with little effect on the results. In column
3, we use extended data through 1980, again with no effect on the results. In columns 4
                                                     ¯
and 5, we use average institutions for each country, Xi , rather than institutions at date t,
                                            ¯
so the equation becomes yit = δt + δi + φ · Xi · U KIN Dt + εit . This specification may give
more sensible results if either variations in institutions from year to year are endogenous
with respect to changes in industrialization or income, or are subject to measurement
error. Interestingly, φ is now estimated to be larger, suggesting that measurement error
is a more important problem than the endogeneity of the changes in institutions across
time periods.
    The real advantage of the specifications in columns 4 and 5 is that they allow us
                                            ¯
to instrument for the regressor of interest Xi · U KIN Dt , using the interaction between
log settler mortality (which we used in Table 9) and UK industrialization per capita,
Mi · U KINDt , to deal with the issue of endogeneity of institutions.23 Once again, in-
stitutions might differ across countries because more productive or otherwise different
countries have different institutions, and in this case, our interaction of industrialization
and institutions could be capturing the effects of these characteristics on economic perfor-
mance. To the extent that mortality rates faced by the early settlers is a good instrument
for institutions (see Acemoglu, Johnson and Robinson, 2001), the interaction between log
settler mortality and UK industrialization will be a good instrument for the interaction
between institutions and UK industrialization. In this case, the IV procedure will deal
both with the endogeneity and omitted variables biases. The IV estimates reported in
column 6 and 7 are very close to the OLS estimates in columns 4 and 5, again suggesting
                        ¯
that the endogeneity of Xi is not important.
      In columns 8 and 9, we add the interaction between latitude and industrialization.
This is useful because, if the reason why the U.S. surged ahead relative to Mexico and
India during the 19th century were its geographic advantage, our measures of institutions
might be proxying for this, incorrectly assigning the role of geography to institutions.
The results give no support to this view: the estimates of φ are affected little and remain
significant, while the interaction between industrialization and latitude is insignificant.24
 23
      We could not instrument for Xit ·U KINDt , since our potential instrument for Xit is not time varying.
 24
      It is important to note that we do not have data from Singapore, Hong Kong and Malaysia, three


                                                     36
    For completeness, column 10 reports results using European countries. This again
leads to a positive and significant estimate of φ, which is somewhat smaller than the
estimate from the former colonies sample. We conjecture that this is because there is less
variation in institutions in the European sample. The second panel of Table 10 repeats
these regressions using log GDP per capita as the left-hand side variable, and the results
are similar.25
    Overall, we interpret these results as providing support for the view that institutions
played an important role in the surge in industrialization and income of the formerly poor
areas.

                                          7    Conclusion

Among countries colonized by European powers during the past 500 years those that
were relatively rich in 1500 are now relatively poor. We document that both today and in
the past there is a close association between urbanization rates and economic prosperity.
Using both population density and urbanization as proxies for economic prosperity, we
find a robust reversal in relative incomes among the former colonies between 1500 and
today.
    We argue that this reversal is inconsistent with the geography hypothesis, which ex-
plains the bulk of the income differences we observe across countries by geographic differ-
ences, and consequently predicts a high degree of persistence in economic outcomes. In
contrast, it is quite consistent with the institutions hypothesis, combined with the fact
that European colonialism caused an “institutional reversal”, introducing worse institu-
tions in previously prosperous areas. We also show that the timing and the nature of the
reversal is not consistent with a more sophisticated geography hypothesis, the Northern
drift, which argues that the center of gravity of economic activity has been shifting to the
North (or away from the equator), and that contact with the Europeans may have led to
such a process among the colonies.
    Instead, the facts are consistent with the importance of institutions. First, the dif-
ferences in institutions (and the institutional reversal caused by European colonialism)
former colonies countries that have prospered, most probably thanks to relatively secure property rights,
and that are closer to the equator than the industrialized countries in our sample. We conjecture that
with data from these countries, institutions would be even stronger relative to latitude.
   25
      In the bottom panel, the interaction term is parameterized as Mi · ln(U KIN Dt ) since the left-hand
side variable is log of GDP per capita.



                                                   37
statistically account for reversal in relative incomes over the past 500 years. Second, we
documented that societies with institutions of private property took better advantage of
the opportunities to industrialize.
   Given the quality of data on urbanization and population density more than 500 years
ago, some degree of caution is required. Nevertheless, the broad patterns in the data seem
uncontroversial: civilizations in Meso-America and India were undoubtedly much more
urbanized and richer than those located in North America, Australia, New Zealand or the
southern cone of Latin America. The intervention of Europe reversed this pattern. This
is a first-order fact for understanding the process of economic and political development
over the past 500 years. Our empirical work and reading of historical evidence suggest
that the interaction between institutions brought by Europeans and industrialization was
essential. Much more work is still required to understand this reversal and the associated
development experiences of the former colonies fully.




                                           38
                                   8   References

Acemoglu, Daron, Simon Johnson and James A. Robinson (2001) “The Colonial
Origins of Comparative Development: An Empirical Investigation,” American Economic
Review, forthcoming.
   Acemoglu, Daron and James A. Robinson (2001) “Economic Backwardness in
Political Perspective,” Unpublished.
   Aghion, Philippe and Peter W. Howitt (1992) “A Model of Growth through
Creative Destruction,” Econometrica, 60, 323-331.
   Bairoch, Paul (1982) “International Industrialization Levels from 1750 to 1980,”
Journal of Economic History, 11, 269-333.
   Bairoch, Paul (1988) Cities and Economic Development: From the Dawn of History
to the Present, Chiacgo; University of Chicago Press.
   Bairoch, Paul, Jean Batou and Pierre Chevre (1988) La Population des villes
Europeenees de 800 a 1850: Banque de Donn´es et Analyse Sommaire des R´sultats,
                                         e                            e
Centre d’histoire ´conomique Internationale de l’Uni. de Gen`ve, Libraire Droz.
                  e                                         e
   Beckford, George L. (1972) Persistent poverty; Underdevelopment in Plantation
Economies of the Third World, New York; Oxford University Press.
   Boserup, Ester (1965) The Conditions of Agricultural Growth; the Economics of
Agrarian Change under Population Pressure, Chicago; Aldine Pub. Co.
   Brenner, Robert (1976) “Agrarian Class Structure and Economic Development in
Preindustrial Europe” Past and Present, No. 70.
   Cain, Philip J. and Anthony G. Hopkins (1993) British Imperialism: Innovation
and Expansion 1688-1914, Longman, New York.
   Chandler, Tertius (1987) Four Thousand Years of Urban Growth: An Historical
Census, Lewiston, N.Y.; St. David’s University Press.
   Chandler, Tertius and Gerald Fox (1974) Three Thousand Years of Urban Growth,
New York; Academic Press.
   Chaudhuri, Kirti N. (1990) Asia Before Europe: Economy and Civilization of the
Indian Ocean from the Rise of Islam to 1750, New York, Cambridge University Press.
   Cipolla, Carlo M. (1974) The Economic History of World Population, 6th edition,
Baltimore; Penguin.
   Coatsworth, John H. (1982) “The Limits of Colonial Absolutism: Mexico in the
18th Century,” in Karen Spalding ed. Essays in the Political, Economic and Social History


                                            39
of Latin America, University of Delaware Press, Newark DL.
   Coatsworth, John H. (1993) ”Notes on the Comparative Economic History of
Latin America and the United States,” in Walter L. Bernecker and Hans Werner Tobler
eds. Development and Underdevelopment in America: Contrasts in Economic Growth in
North and Latin America in Historical Perpsective, New York; Walter de Gruyter.
   Coatsworth, John H. (1999) “Economic and Institutional Trajectories in 19th
Century Latin America,” in John H. Coatsworth and Alan M. Taylor eds. Latin America
and the World Economy since 1800, Harvard University Press, Cambridge MA.
   Cole, Jeffrey A. (1985) The Potosi Mita, 1573-1700: Compulsory Indian Labor in
the Andes, Stanford University Press, Palo Alto.
   Crosby, Alfred (1986) Ecological Imperialism: The Biological Expansion of Europe
900-1900, Cambridge University Press, New York NY.
   Curtin, Philip D. (1990) The Rise and Fall of the Plantation Complex: Essays in
Atlantic History, Cambridge University Press, New York NY.
   De Long, Bradford J. and Andrei Shleifer (1993) “Princes and Merchants:
European City Growth before the Industrial Revolution, Journal of Law and Economics,
36, October, 671-702.
   de Vries, Jan (1976) The Economy of Europe in an Age of Crisis, 1600-1750,
Cambridge University Press.
   de Vries, Jan (1984) European Urbanization, 1500-1800, Boston; Harvard Univer-
sity Press.
   Denoon, Donald (1983) Settler Capitalism: The Dynamics of Dependent Develop-
ment in the Southern Hemisphere, Clarendon Press, Oxford UK.
   Diamond, Jared M. (1997) Guns, Germs and Steel: The Fate of Human Societies,
W.W. Norton & Co., New York NY.
   Domar, Evsey (1970) “The Causes of Slavery and Serfdom: A Hypothesis,” Journal
of Economic History, 30, 18-32.
   Duby, Georges (1974) The Early Growth of the European Economy; Warriors and
Peasants from the Seventh to the Twelfth Century, Ithaca; Cornell University Press
   Dunn, Richard S. (1972) Sugar and Slaves: The Rise of the Planter Class in the
English West Indies 1624-1713, Chapel Hill; University of North Carolina Press.
   Eggimann, Gilbert (1999) La Population des villes des Tiers-Mondes, 1500-1950,
Centre d’histoire ´conomique Internationale de l’Uni. de Gen`ve, Libraire Droz.
                  e                                         e


                                          40
   Eltis, David (1995) “The Total Product of Barbados, 1664-1701,” Journal of Eco-
nomic History, 55, 321-336.
   Engerman, Stanley L. and Kenneth L. Sokoloff (1997) “Factor Endowments,
Institutions, and Differential Paths of Growth among New World Economies,” in S.H.
Haber ed. How Latin America Fell Behind, Stanford University Press, Stanford CA.
   Engerman, Stanley L. and Kenneth L. Sokoloff (2000) “Institutions, Factor
Endowments, and Paths of Development in the New World,” Journal of Economic Per-
spectives, 3, 217-232.
   Frank, Andre Gunder (1978) Dependent Accumulation and Underdevelopment,
London; Macmillan.
   Galbraith, James K. (1951) “Conditions for Economic Change in Underdeveloped
Countries,” American Journal of Farm Economuics, 33.
   Galenson, David W. (1996) “The Settlement and Growth of the Colonies: Pop-
ulation, Labor and Economic Development,” in Stanley L. Engerman and Robert E.
Gallman eds. The Cambridge Economic History of the United States, Volume I, New
York; Cambridge University Press
   Gallman Robert E. (2000) “Economic Growth and Structural Change in the Long
Nineteenth Century,” in Stanley L. Engerman and Robert E. Gallman eds. The Cambridge
Economic History of the United States, Volume II, New York; Cambridge University Press.
   Genovese, Eugene D. (1965) The Political Economy of Slavery: Studies in the
Economy and Society of the Slave South, New York; Pantheon Books.
   Grossman, Gene M. and Elhanan Helpman (1991) Innovation and Growth in
the Global Economy, Cambridge; MIT Press.
   Halperin Donghi, Tulio (1993) The Contemporary History of Latin America,
Durham; Duke University Press.
   Harms, Robert C. (1981) River of Wealth, River of Sorrow: The Central Zaire
Basin in the era of the Slave and Ivory Trade, 1500-1891, New Haven; Yale University
Press.
   Hayek, Freidrich von (1960) The Constitution of Liberty, Chicago; University of
Chicago Press.
   Hohenberg, Paul M. and Lynn Hollen Lees (1985) The Making of Urban Eu-
rope, 1000-1950, Cambridge MA; Harvard University Press.
   Hopkins, Anthony G. (1973) An Economic History of West Africa, New York;


                                          41
AddisonWesley Longman.
   Hughes, Robert (1988) The Fatal Shore, New York, Vintage Books.
   Huntington, Ellsworth (1915) Civilization and Climate, New Haven, Yale Univer-
sity Press.
   Huntington, Ellsworth (1945) Mainsprings of Civilization, New York, J. Wiley
and Sons.
   Jacobs, Jane (1969) The Economy of Cities, New York; Random House.
   Jones, Charles I. (1998) “Population and Ideas: A Theory of Endogenous Growth,”
http://www.stanford.edu/˜chadj/pop400.pdf.
   Jones, Eric L. (1981) The European Miracle: Environments, Economies, and Geopol-
itics in the History of Europe and Asia, New York; Cambridge University Press.
   Kennedy, Paul M. (1987) The Rise and Fall of the Great Powers: Economic
Change and Military Conflict from 1500 to 2000, New York; Random House.
   Kremer, Michael (1993) “Population Growth and Technological Change: One Mil-
lion B.C. to 1990,” Quarterly Journal of Economics, 1087, 681-716.
   Kuznets, Simon (1968) Modern Economic Growth: Rate Structure and Spread,
New Haven; Yale University Press.
   Landes, David S. (1998) The Wealth and Poverty of Nations: Why Some Are So
Rich and Some So Poor, W.W. Norton & Co., New York.
   La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer and Robert
W. Vishny (1998). “Law and Finance,” Journal of Political Economy, 106, 1113-55.
   Law, Robin C.C. (1977) The Oyo Empire 1600-1836: A West Africa Imperialism
in the Era of the Atlantic Slave Trade, New York; Oxford University Press.
   Locke, John [1690] (1980) Two Treatises of Government, Indianapolis; Hackett.
   Lockhart, James and Stuart B. Schwartz (1983) Early Latin America, Cam-
bridge University Press, New York NY.
   Machiavelli, Niccol` [1519] (1987) Discourses on Livy, New York; Oxford Uni-
                      o
versity Press.
   Maddison, Angus (1995) Monitoring the World Economy, 1820-1992, Development
Centre of the Organization for Economic Cooperation and Development.
   Manning, Patrick (1988) Francophone Sub-Saharan Africa, 1880-1995, Cambridge
University Press, New York NY.
   Manning, Patrick (1990) Slavery and African Life: Occidental, Oriental and African


                                          42
Slave Trades, New York; Cambridge University Press.
   Marshall, Alfred [1890] (1949) Principles of Economics, London; Macmillan.
   Marshall, P.J. (1998) “The British in Asia: Trade to Dominion, 1700-1765,” in P.J.
Marshall ed. The Oxford History of the British Empire, Volume II The 18th Century,
Oxford University Press, New York.
   McCusker, John J. and Russell R. Menard (1985) The Economy of British
America, 1607-1785, Chapel Hill; University of North Carolina Press.
   McEvedy, Colin (1967), The Penguin Atlas of Ancient History, Penguin Books,
London.
   McEvedy, Colin (1992), The New Penguin Atlas of Medieval History, Penguin
Books, London.
   McEvedy, Colin and Richard Jones (1975), Atlas of World Population History,
Facts on File, New York.
   McNeill, William H. (1976) Plagues and Peoples, Anchor Press, Garden City NJ.
   Miller, Joseph C. (1988) Way of Death: Merchant Capitalism and the Angolan
Slave Trade, 1730-1830, Madison; University of Wisconsin Press.
   Mitchell, B.R. (1993) International Historical Statistics, The Americas 1750-1988,
Second Edition, Stockton Press, New York.
   Mitchell, B.R. (1995) International Historical Statistics, Africa, Asia & Oceania
1750-1988, Second Edition, Stockton Press, New York.
   Montesquieu, Charles de Secondat [1748] (1989) The Spirit of the Laws, Cam-
bridge University Press, New York NY.
   Myrdal, Gunnar (1968) Asian Drama; An Inquiry into the Poverty of Nations, 3
Volumes, New York; Twentieth Century Fund.
   North, Douglass C. (1990) Institutions, Institutional change, and Economic Per-
formance, New York; Cambridge University Press.
   North, Douglass C. and Robert P. Thomas (1973) The Rise of the Western
World: A New Economic History, Cambridge University Press, Cambridge UK.
   North, Douglass C. and Barry R. Weingast (1989) “Constitutions and Commit-
ment: Evolution of Institutions Governing Public Choic in Seventeenth Century England,
Journal of Economic History, 49, 803-832.
   Notestein, Frank W. (1945) “Population: The Long View,” in Theodore W.
Schultze ed. Food for the World, Chicago; University of Chicago Press.


                                            43
   Patterson, Orlando (1982) Slavery and Social Death: A Comparative Study, Cam-
bridge; Harvard University Press.
   Pirenne, Henri (1956) [1925] Medieval Cities; Their Origins and the Revival of
Trade, New York; Doubleday.
   Pomeranz, Kenneth (2000) The Great Divergence: Europe, China, and the Making
of the Modern World Economy, Princeton; Princeton University Press.
   Postan, M.M. and E.E. Rich (1966) The Cambridge Economic History of Europe:
Volume 2, Trade and Industry in the Middle Ages, Cambridge; Cambridge University
Press.
   Pritchett, Lant (1997) “Divergence, Big Time” Journal of Economic Perspectives,
11, 3-18.
   Reid, Anthony (1988) Southeast Asia in the Age of Commerce: Volume 1, The
Lands Below the Winds, New Haven, Yale University Press.
   Reid, Anthony (1993) Southeast Asia in the Age of Commerce: Volume 2, The Age
of Crisis, New Haven, Yale University Press.
   Rodney, Walter (1972) How Europe Underdeveloped Africa, London; Bogle-L’Ouverture
Publications
   Romer, Paul M. (1986) “Increasing Returns and Long-Run Growth,” Journal of
Political Economy, 94, 1002-1037.
   Romer, Paul M. (1990) “Endogenous Technical Change,” Journal of Political Econ-
omy, 98, S71-S102.
   Sachs, Jeffrey D. (2000) “Notes on a New Sociology of Economic Development”
in Lawrence E. Harrison and Samuel P. Huntington eds. Culture Matters: How Values
Shape Human Progress, New York; Basic Books.
   Sachs, Jeffrey D. (2001) “Tropical Underdevelopment,” NBER Working Paper
#8119.
   Showers, Victor (1979) World Facts and Figures, New York; Wiley.
   Simon, Julian L. (1977) The Economics of Population Growth, Princeton; Prince-
ton University Press.
   Tilly, Charles (1990) Coercion, Capital, and European States, AD 990-1990, Cam-
bridge MA; Basil Blackwell.
   Tilly, Charles and Wim P. Blockmans eds. (1994) Cities and the Rise of States
in Europe, A.D. 1000 to 1800, Boulder, Westview Press.


                                          44
   Toynbee, Arnold J. (1934-1961) A Study of History, 12 Volumes, Oxford; Oxford
University Press.
   Wallerstein Immanuel M. (1974-1980) The Modern World-System, 3 Volumes,
New York, Academic Press.
   Wiegersma, Nancy (1988) Vietnam: Peasant Land, Peasant Revolution, New York;
St. Martin’s Press.
   White, Lynn, Jr. (1962) Medieval Technology and Social Change, London: Oxford
University Press.
   Wilks, Ivor (1975) Asante in the 19th Century: The Structure and Evolution of a
Political Order, New York, Cambridge University Press.
   Williams, Eric E. (1944) Capitalism and Slavery, Chapel Hill; University of North
Carolina Press.
   Williams, Eric E. (1970) From Columbus to Castro: A History of the Caribbean,
London; Andre Deutsch.
   Wong, R. Bin (1997) China Transformed: Historical Change and the Limits of
European Experience, Ithaca; Cornell University Press.
   Young, Crawford (1994) The African Colonial State in Comparative Perspective,
Yale University Press, New Haven CT.




                                          45
                                                 Table 1
                                           Descriptive Statistics

                                                               Base Sample Base Sample
                                                   Whole      for population     for
                                                   World          density    urbanization


Log GDP per capita (PPP) in 1995                      8.3           7.8              8.5
                                                     (1.1)         (1.0)            (0.9)
Log Population Density in 1500                         1           0.59             0.29
                                                     (1.6)         (1.5)            (1.9)
Urbanization in 1500                                  7.4           6.7              6.3
                                                     (5.0)         (4.9)            (4.9)
Log Population Density in 1000                        0.6          0.11             0.11
                                                     (1.5)         (1.6)            (2.0)
Urbanization in 1000                                  4.7           4.1               4
                                                     (4.4)         (4.0)            (4.0)
Urbanization in 1995                                  53           43.5             57.5
                                                    (23.8)        (21.3)           (22.4)
European Settlements in 1900                         29.6          12.3             23.2
                                                    (41.7)        (22.9)           (28.7)
Constraint on the Executive in 1990                   3.6           3.7              4.9
                                                     (2.3)         (2.3)            (2.1)
Constraint on the Executive in 1900                   1.9            2               2.7
                                                     (1.8)         (1.9)            (2.3)
Constraint on the Executive in                       3.6           3.4              3.3
First Year of Independence                           (2.4)         (2.3)            (2.5)

Number of Observations                               162            80               41
Standard deviations are in parentheses. For detailed sources and descriptions see Appendix
Table 1.
                                                                        Table 2
                                                          Urbanization and Per Capita Income


                                                                                                                                        Cross-sectional
                                                                                                                      Cross-sectional    regression in
                                                                                                                       regression in       1995, all
                            Cross-sectional         Cross-sectional        Cross-sectional       Cross-sectional        1995, never     countries, with
                         regression in 1900, all regression in 1950, all regression in 1995,   regression in 1995,       colonized         continent
                               countries               countries            all countries      only for excolonies    countries only       dummies
                                 (1)             (2)              (3)            (4)                                       (5)               (6)
                                  Panel A: Dependent Variable is Log GDP per capita
Urbanization                    0.038           0.026            0.036          0.037                                    0.033             0.030
                               (0.006)         (0.002)          (0.002)        (0.003)                                  (0.007)           (0.002)

R-Squared                       0.69                    0.57                  0.63                   0.69                 0.34              0.68
Number of Observations           22                      128                   162                    93                   51                162


                          Using panel data set   Using panel data set   Using panel data set   Using panel data set
                             through 1900           through 1900           through 1900           through 1900
                                 (7)             (8)              (9)            (10)
                                  Panel B: Dependent Variable is Log GDP per capita
Urbanization                    0.034           0.030            0.041          0.026
                               (0.004)         (0.004)          (0.004)        (0.004)

Country Dummies                   No                     No                    Yes                    Yes

Period Dummies                    No                     Yes                    No                    Yes

R-Squared                       0.52                    0.71                  0.87                   0.96
Number of Observations           62                      62                    62                     62


GDP per capita through 1900 is from Bairoch (1978), with the addition of Gregory King for England before 1700.
Urbanization is percent of population living in towns with at least 5,000 people, from Bairoch (1988) through 1900 (see
appendix for details). GDP per capita in 1950 is from Maddison (1995); this regression uses urbanization in 1960 from the
World Bank's World Development Indicators 1999. GDP per capita and Urbanization data for 1995 are from the World
Bank's World Development Indicators 1999. Population density is total population divided by land area usable for
agriculture, both from McEvedy and Jones (1978). For detailed sources and descriptions see Appendix Table 1.
                                                                                               Table 3
                                                                           Urbanization, Population Density and Institutions
                                                                                              Dependent Variable is:
                              Average Protection Against                                                                                              Constraint on Executive in First
                             Expropriation Risk, 1985-95           Constraint on Executive in 1990        Constraint on Executive in 1900                 Year of Independence
                             (1)         (2)         (3)            (4)           (5)          (6)         (7)          (8)          (9)               (10)          (11)        (12)
                                                                             Panel A: Without Additional Controls

Urbanization                -0.11                     -0.01         -0.16                        -0.04         -0.12                    0.03            -0.14                    0.01
in 1500                    (0.04)                    (0.06)        (0.07)                       (0.10)        (0.07)                   (0.10)          (0.07)                   (0.11)

Log Population Density                   -0.37        -0.36                         -0.49        -0.40                      -0.35       -0.52                         -0.33      -0.53
in 1500                                 (0.10)       (0.15)                        (0.15)       (0.25)                     (0.12)      (0.26)                        (0.15)     (0.28)

R-Squared                   0.14        0.16          0.25          0.12           0.12         0.18          0.06         0.09         0.15           0.31          0.16        0.37
Number of Observations       42          75            42            41             84           41            43           88           43             42            85          42

                                                                                  Panel B: Controlling for Latitude

Urbanization in 1500        -0.10                    -0.003         -0.16                        -0.04         -0.10                    0.04            -0.13                    0.02
                           (0.04)                    (0.06)        (0.07)                       (0.10)        (0.07)                   (0.10)          (0.07)                   (0.11)

Log Population Density                   -0.31        -0.34                         -0.45        -0.42                      -0.27       -0.48                         -0.30      -0.53
in 1500                                 (0.10)       (0.15)                        (0.16)       (0.25)                     (0.12)      (0.25)                        (0.16)     (0.29)

Latitude                    2.85         3.53         2.57          -1.51           2.63         -1.87         5.82         4.83        5.45            1.49          2.68       1.47
                           (1.48)       (1.25)       (1.41)        (2.39)          (2.01)       (2.34)        (2.37)       (1.54)      (2.30)          (2.53)        (1.17)     (2.46)

R-Squared                   0.22         0.24         0.31          0.13            0.13         0.19          0.19        0.19         0.26            0.32         0.17        0.38
Number of Observations       42           75           42            41              84           41            43          87           43              42           84          42

Regressions use data on all former colonies for which information on urbanization and population density in 1500 is available, as explained in the text. Urbanization in 1500 is percent
of the population living in towns with 5,000 or more people, from Bairoch (1988), with supplementary sources as described in the appendix. Population density in 1500 is total
population divided by land area usable for agriculture, from McEvedy and Jones (1978). Regressions with constraint on executive in first year of independence also include time since
independence as a regressor. For detailed sources and descriptions see Appendix Table 1.
                                                              Table 4
                                Urbanization in 1500 and GDP per capita in 1995 for former European colonies
                                                   Dependent Variable is log GDP per capita (PPP basis) in 1995
                                                 without North   without the   with continent   without neo- controlling for controlling for   controlling for   controlling for
                                   base sample       Africa       Americas       dummies          Europes       latitude       geography       colonial origin      religion

                                       (1)           (2)            (3)            (4)             (5)             (6)            (7)              (8)               (9)
Urbanization in 1500                 -0.08         -0.11          -0.12          -0.09           -0.05           -0.08          -0.09             -0.07            -0.07
                                     (0.03)        (0.03)         (0.05)         (0.03)          (0.03)          (0.03)         (0.04)           (0.03)            (0.03)
Asia Dummy                                                                       -1.30
                                                                                 (0.60)
Africa Dummy                                                                     -0.50
                                                                                 (0.77)
America Dummy                                                                    -0.94
                                                                                 (0.57)
Former French Colony                                                                                                                              -0.58
                                                                                                                                                 (0.39)
Former Spanish Colony                                                                                                                             0.07
                                                                                                                                                 (0.29)
P-Value for Temperature                                                                                                         [0.24]
P-Value for Humidity                                                                                                            [0.64]
P-Value for Soil Quality                                                                                                        [0.96]
P-Value for Natural Resources                                                                                                   [0.93]

Latitude                                                                                                          1.4
                                                                                                                 (0.91)
P-value for Religion                                                                                                                                               [0.49]

R-Squared                            0.20          0.24           0.26            0.34            0.09            0.25           0.62             0.27               0.26
Number of Observations                41            37             17              41              37              41             41               41                 41
Dependent Variable: Log GDP per capita (PPP basis) in 1995, current prices (from the World Bank's World Development Indicators 1999). Base sample is
all former colonies for which we have data and excludes countries in Africa south of the Sahara. Urbanization in 1500 is percent of the population living in
towns with 5,000 or more people, from Bairoch (1988), with supplementary sources as described in the appendix. The neo-Europes are the USA, Canada,
Australia and New Zealand. The regression that controls for geography also includes a dummy for whether or not the country is landlocked; this is not
significant. The regression that controls for colonial origin (column 8) includes dummies for former French colony, Spanish colony, Portuguese colony,
Belgian colony, Italian colony, German colony, and Dutch colony. British colonies are the base case. The religion variables are percent of the population
who are Muslim, Catholic, and "other"; percent Protestant is omitted. For detailed sources and descriptions see Appendix Table 1.
                                                                        Table 5
                                                            Alternative Measures of Urbanization
                                                           Dependent Variable is log GDP per capita (PPP basis) in 1995
                                                 without North   without the   with continent   without neo- controlling for   controlling for   controlling for controlling for
                                   base sample       Africa       Americas       dummies          Europes       latitude         geography       colonial origin    religion

                                      (1)            (2)       (3)          (4)         (5)        (6)                              (7)               (8)             (9)
                                                          Panel A: Using the measure in Table 3
Urbanization in 1500                 -0.08         -0.11      -0.12       -0.09       -0.05      -0.08                            -0.09            -0.07           -0.07
                                     (0.03)        (0.03)    (0.05)       (0.03)      (0.03)     (0.03)                          (0.04)            (0.03)          (0.03)
R-Squared                            0.20          0.24       0.26        0.34        0.09        0.25                            0.62             0.27             0.26
Number of Observations                41             37        17           41          37         41                              41                41              41
                                                         Panel B: Using only Bairoch's estimates
Urbanization in 1500                 -0.14         -0.14      -0.25       -0.10       -0.10      -0.14                            -0.10            -0.13           -0.08
                                     (0.03)        (0.03)    (0.07)       (0.03)      (0.03)     (0.04)                          (0.05)            (0.03)          (0.03)
R-Squared                            0.46           0.46      0.59        0.59        0.30        0.46                            0.89             0.54             0.67
Number of Observations                30             30        10           30          27         30                              30                30              30
                                                        Panel C: Using only Eggimann's estimates
Urbanization in 1500                 -0.04         -0.04      -0.10       -0.04       -0.02      -0.04                            -0.05            -0.04           -0.03
                                     (0.02)        (0.02)    (0.05)       (0.02)      (0.02)     (0.02)                          (0.03)            (0.02)          (0.02)
R-Squared                            0.11          0.09       0.24        0.29        0.05        0.17                            0.62             0.24             0.25
Number of Observations                40             37        16           40          36         40                              40                40              40
                                                        Panel D: Using only Chandler's estimates
Urbanization in 1500                 -0.06         -0.06      -0.06       -0.07       -0.04      -0.05                            -0.04            -0.05           -0.06
                                     (0.02)        (0.02)    (0.04)       (0.02)      (0.02)     (0.02)                          (0.04)            (0.02)          (0.02)
R-Squared                            0.25          0.23       0.17        0.42        0.17        0.32                            0.84             0.44             0.57
Number of Observations                26             22        12           26          23         26                              26                26              26
Dependent Variable: Log GDP per capita (PPP basis) in 1995, current prices (from the World Bank's World Development Indicators 1999). Base sample is
all former colonies for which we have data and excludes countries in Africa south of the Sahara. Urbanization in 1500 is from alternative sources in the four
panels, as described in the text and the appendix. The neo-Europes are the USA, Canada, Australia and New Zealand. The regression that controls for
geography also includes a dummy for whether or not the country is landlocked; this is not significant. The religion variables are percent of the population
who are Muslim, Catholic, and "other"; percent Protestant is omitted. The regression that controls for colonial origin (column 8) includes dummies for
former French colony, Spanish colony, Portuguese colony, Belgian colony, Italian colony, German colony, and Dutch colony; British colonies are the base
case. For detailed sources and descriptions see Appendix Table 1.
                                                                    Table 6
                         Population Density and GDP per capita in former European colonies (including sub-Saharan Africa)

                                                        Dependent Variable is log GDP per capita (PPP basis) in 1995
                                                                                                                                          controlling
                                                                                    with                                    controlling      for      controlling
                                                        without     without the   continent   without neo-   controlling       for         colonial       for
                                          base sample   Africa       Americas     dummies       Europes      for latitude   geography       origin     religion

                                             (1)       (2)       (3)        (4)       (5)       (6)       (7)                                (8)         (9)
                                             Panel A: Log population density in 1500 as independent variable
Log Population Density in 1500              -0.38     -0.40     -0.32     -0.26      -0.31     -0.34     -0.32                              -0.32 -0.37
                                           (0.06)    (0.05)    (0.07)    (0.05)     (0.06)    (0.06)    (0.07)                             (0.06) (0.07)
Asia Dummy                                                                -0.91
                                                                         (0.55)
Africa Dummy                                                               -1.7
                                                                         (0.52)
America Dummy                                                              -0.7
                                                                         (0.52)
Former French Colony                                                                                                                        -0.48
                                                                                                                                           (0.21)
Former Spanish Colony                                                                                                                       0.26
                                                                                                                                           (0.23)

P-Value for Temperature                                                                                                      [0.32]
P-Value for Humidity                                                                                                         [0.05]
P-Value for Soil Quality                                                                                                     [0.33]
P-Value for Natural Resources                                                                                                [0.76]

Latitude                                                                                                       2.1
                                                                                                              (0.75)
P-Value for Religion                                                                                                                                   [0.74]


R-Squared                                  0.33     0.55       0.27      0.55      0.22       0.39     0.61     0.47  0.35
Number of Observations                      90       46         58        90         86        90       84       90    84
                                    Panel B: Log population and log land in 1500 as separate independent variables
Log Population in 1500                     -0.40    -0.41      -0.32     -0.28     -0.31      -0.35    -0.33    -0.34 -0.39
                                          (0.06)   (0.05)     (0.07)    (0.05)    (0.06)     (0.05)   (0.07) (0.06) (0.07)
Log Arable Land in 1500                    0.31     0.37       0.21      0.23      0.20       0.26     0.24     0.27  0.31
                                          (0.06)   (0.06)     (0.09)    (0.06)    (0.07)     (0.06)   (0.09) (0.06) (0.08)

R-Squared                        0.37      0.57        0.31     0.57       0.31      0.44      0.62     0.50    0.38
Number of Observations             90        46         58       90         86        90        84        90     84
                   Panel C: Using population density in 1000 AD as an instrument for population density in 1500 AD
Log Population Density in 1500    -0.31     -0.4       -0.15    -0.18      -0.22     -0.27     -0.24    -0.26 -0.25
                                 (0.06)    (0.06)     (0.08)   (0.07)     (0.08)    (0.06)    (0.09) (0.06) (0.08)

Number of Observations                       83           43           51           83            80            83             78            83          77
Dependent Variable: Log GDP per capita (PPP basis) in 1995, current prices (from the World Bank's World Development Indicators 1999).
Population density in 1500 is total population divided by land area usable for agriculture, from McEvedy and Jones (1978). The neo-Europes are
the USA, Canada, Australia and New Zealand. The regression that controls for geography also includes a dummy for whether or not the country is
landlocked; this is not significant. The regression that controls for colonial origin (column 8) includes dummies for French colony, Spanish colony,
Portuguese colony, Belgian colony, Italian colony, German colony, and Dutch colony; British colonies are the base case. The religion variables are
percent of the population who are Muslim, Catholic, and "other"; percent Protestant is omitted. For detailed sources and descriptions see Appendix
Table 1.
                                                                          Table 7
                                                  Robustness Checks for Urbanization and Log Population Density

                                                 Dependent Variable is log GDP per capita (PPP basis) in 1995
                                                 Assuming
                                                    lower                                                                    Using                   Alternative
                                              urbanization in               Assuming                     Using basic     augmented                  assumptions
                                               the Americas    Assuming       lower                        Toynbee         Toynbee                     for log
                                                 (9% in the      lower     urbanization Using least      definition of   definition of               population
                                                Andes and urbanization in in Indian       favorable       civilization    civilization              density (halve
                                                   Central    North Africa subcontinent combination of   (excolonies     (excolonies Using land estimates for
                                  Base Sample     America)      (10%)          (6%)      assumptions         only)           only)     area in 1995    Africa)
                                       (1)           (2)           (3)         (4)          (5)              (6)             (7)          (8)            (9)
                                                                 Panel A: Unweighted Regressions
Urbanization in 1500                 -0.08          -0.09         -0.10       -0.07        -0.11           -0.10           -0.12
                                     (0.03)         (0.03)       (0.03)      (0.03)       (0.03)           (0.06)          (0.05)
Log Population Density in 1500                                                                                                           -0.42         -0.34
                                                                                                                                         (0.06)        (0.07)

R-Squared                             0.20          0.22          0.24         0.16          0.21         0.45             0.30          0.38           0.24
Number of Observations                 41             41           41           41            41            7               14            82             82
                                                     Panel B: Regressions weighted using log population in 1500
Urbanization in 1500                 -0.07          -0.08         -0.10        -0.06         -0.10        -0.09            -0.12
                                     (0.03)         (0.03)        (0.03)       (0.03)       (0.03)        (0.07)           (0.05)
Log Population Density in 1500                                                                                                           -0.39         -0.30
                                                                                                                                         (0.06)        (0.07)
R-Squared                             0.19          0.21          0.23          0.13         0.19           0.27           0.29          0.34          0.20
Number of Observations                 40            40            40            40           40              7             14            82            82

                                      Panel C: Including both urbanization and log population density as independent variables
Urbanization in 1500                    0.04          0.04            0.02         0.04        0.02         0.07        0.07          0.01          0.01
                                      (0.03)         (0.04)          (0.03)      (0.03)       (0.04)       (0.05)      (0.05)        (0.02)       (0.02)
Log Population Density in 1500         -0.41         -0.41           -0.37        -0.40        -0.37       -0.43       -0.48         -0.43         -0.41
                                      (0.08)         (0.08)          (0.07)      (0.07)       (0.07)       (0.11)      (0.10)        (0.07)       (0.07)
R-Squared                              0.55          0.55            0.54         0.56         0.54        0.85        0.79          0.61          0.60
Number of Observations                  40            40              40           40           40           7          14            40            40
Dependent Variable: Log GDP per capita (PPP basis) in 1995, current prices (from the World Bank's World Development Indicators 1999). In our base
sample, urbanization in 1500 is percent of the population living in towns with 5,000 or more people, from Bairoch (1988), with supplementary sources as
described in the appendix. Alternative assumptions are as described at the top of each column. Population density in 1500 is total population divided by
land area usable for agriculture. For detailed sources and descriptions see Appendix Table 1.
                                                                                   Table 8
                                                             Persistence of Urbanization and Population Density

                                                                                                                                                           All countries never All countries never
                                                                                                    Using augmented                                            colonized by        colonized by
                                                                   Base Sample    Base Sample      Toynbee definition of               All countries never Western Europe,      Western Europe,
                                                 Base Sample         (without     with continent        civilization                       colonized by    without the Russian    with continent
                                  Base Sample   (without Africa)     America)       dummies          (excolonies only)   Just Europe    Western Europe            Empire            dummies
                                       (1)            (2)              (3)           (4)            (5)            (6)            (7)                             (8)                  (9)
                                                                            Panel A: Dependent Variable is Urbanization in 1995
Urbanization in 1500                  -1.80          -2.70            -1.80         -2.21          -2.70          1.30           1.10                            1.30                 1.20
                                     (0.70)         (0.87)           (1.40)        (0.76)         (2.20)         (0.40)         (0.47)                          (0.54)               (0.47)
R-Squared                             0.14           0.20             0.09          0.39           0.13           0.23           0.11                            0.14                 0.14
Number of Observations                 44             39               20            44             12             34             46                              36                   46

                                                                              Panel B: Dependent Variable is Urbanization in 1500
Urbanization in 1000                  0.94           0.86             1.50            0.87           0.82           0.46           0.42                          0.46                 0.45
                                     (0.16)         (0.14)           (0.41)          (0.13)         (0.16)         (0.21)         (0.17)                        (0.18)               (0.19)
R-Squared                             0.50           0.58             0.53            0.74           0.73           0.14           0.13                          0.17                 0.14
Number of Observations                 35             31               14              35             12             31             42                            35                   42

                                                                       Panel C: Dependent Variable is Log Population Density in 1500
Log Population Density in 1000        0.91           0.97             0.86          0.83           0.82           1.20           1.10                            1.13                 1.03
                                     (0.04)         (0.05)           (0.04)        (0.04)         (0.16)         (0.10)         (0.06)                          (0.09)               (0.06)
R-Squared                             0.89           0.92             0.90          0.92           0.70           0.83           0.85                            0.79                 0.85
Number of Observations                 80             37               55            80             12             33             60                              44                   59

                                                                       Panel D: Dependent Variable is Log Population Density in 1000
Log Population Density in 1 AD        0.94           1.10             0.86          0.89           0.98           0.82           0.82                            0.65                 0.83
                                     (0.03)         (0.02)           (0.04)        (0.03)         (0.23)         (0.04)         (0.06)                          (0.10)               (0.06)
R-Squared                             0.94           0.99             0.92          0.95           0.67           0.92           0.79                            0.54                 0.80
Number of Observations                 70             36               46            70             11             32             55                              39                   54

Urbanization is the percent of the population living in cities with more than 5,000 inhabitants (sources are described in the appendix.) Population density in 1AD, 1000 and
1500 is total population divided by land area usable for agriculture. Column 8 includes a dummy for Asia; Europe is the base case. For detailed sources and descriptions
see Appendix Table 1.
                                                                           Table 9
                                                               GDP per capita and Institutions
                                                         Dependent Variable is log GDP per capita (PPP basis) in 1995
                           Average Protection Against                                                                            Constraint on Executive in First
                           Expropriation Risk, 1985-95       Constraint on Executive in 1990   Constraint on Executive in 1900       Year of Independence
                              (1)             (2)                (3)          (4)          (5)                       (6)              (7)               (8)
                                                              Panel A: Second Stage Regressions
Institutions                  0.52           0.88               0.83         0.50         0.45                      0.51             0.37              0.46
                             (0.10)         (0.21)             (0.47)       (0.11)       (0.16)                    (0.14)           (0.12)            (0.16)

Urbanization                 -0.03                              0.03                              -0.02                              -0.02
in 1500                      (0.02)                            (0.08)                             (0.04)                            (0.03)

Log Population Density                      -0.08                            -0.10                                 -0.16                              -0.13
in 1500                                     (0.10)                          (0.10)                                 (0.10)                             (0.10)
                                                               Panel B: First Stage Regressions
Log Settler Mortality        -1.20          -0.47               -0.75        -0.88         -1.30                   -0.81             -1.80            -0.78
                             (0.23)         (0.14)             (0.44)       (0.20)         (0.46)                  (0.19)           (0.40)            (0.25)

Urbanization in 1500         -0.05                              -0.09                             -0.06                              -0.05
                             (0.04)                            (0.07)                             (0.07)                            (0.06)
Log Population Density                      -0.21                                -0.35                             -0.20                              -0.24
in 1500                                     (0.11)                               (0.15)                            (0.14)                             (0.17)

R-Squared                    0.53            0.29               0.12              0.37             0.25             0.31             0.56               0.23

Number of Observations        38          64              37            67            39           70             38                                     67
                            Panel C: Coefficient on Institutions without Urbanization or Population Density in 1500
Institutions                 0.56        0.96           0.77           0.54          0.48         0.61           0.39                                  0.52
                            (0.09)      (0.17)         (0.33)         (0.09)        (0.14)       (0.13)         (0.11)                                (0.15)


Dependent Variable: Log GDP per capita (PPP basis) in 1995, current prices, (from the World Bank's World Development Indicators
1999). The measure of institutions used in each regression is indicated at the head of each column. Regressions with constraint on the
executive in the first year of independence also include time since independence as a regressor. Urbanization in 1500 is percent of the
population living in towns with 5,000 or more people, from Bairoch (1988), with supplementary sources as described in the appendix.
Population density is calculated as total population divided by land area usable for agriculture (both from McEvedy and Jones 1978).
Constraint on the executive in 1990, 1900 and the first year of independence are all from the Polity III dataset. For detailed sources and
descriptions see Appendix Table 1.
                                                                                                                          Table 10
                                                                                                 The Interaction of UK Industrialization and Local Institutions


                                                                                                                                                       Former Colonies,    Former Colonies,       Former Colonies,    Former Colonies,
                                                                                                                                                        using average        using average         using average        using average
                                                                                                          Former Colonies,      Former Colonies, institutions for each institutions for each institutions for each institutions for each
                                                                                                            using average         using average             country,            country,               country,            country,
                                         Former Colonies,      Former Colonies,     Former Colonies,     institutions for each institutions for each instrumenting using instrumenting using instrumenting using instrumenting using           All European
                                        using only pre-1950   using only pre-1950   using data through   country, using only country, using all        settler mortality, settler mortality, all  settler mortality, settler mortality, all countries, using data
                                               data                  data             1980 (all data)       pre-1950 data               data          only pre-1950 data          data           only pre-1950 data          data              through 1950

                                               (1)                   (2)                  (3)             (4)             (5)             (6)             (7)                                           (8)                   (9)                  (10)
                                                                                            Panel A: Dependent Variable is Industrial Production Per Capita
UK Industrialization*Institutions             0.13                  0.08                 0.13            0.21            0.28            0.16           0.25                                           0.13                  0.20                  0.06
                                             (0.03)                (0.03)                (0.03)               (0.02)                (0.02)                (0.03)                (0.02)                (0.06)                (0.05)                (0.01)
Institutions                                  8.97                  9.71                 -3.36                                                                                                                                                     4.34
                                             (2.30)                (2.58)                (4.46)                                                                                                                                                   (0.96)
UK Industrialization*Latitude                                                                                                                                                                          0.30                  0.52
                                                                                                                                                                                                      (0.49)                (0.36)

R-Squared                                     0.75                  0.44                 0.74                  0.87                  0.94                  0.85                  0.94                  0.83                  0.93                  0.78

Number of Observations                         59                    59                    75                    59                    75                    59                    75                    59                   75                    138
                                                                                                     Panel B: Dependent Variable is Log GDP Per Capita
Log UK Industrialization*Institutions        0.08                   0.10                  0.06               0.13          0.12            0.15                                  0.10                  0.09                  0.07                  0.02
                                             (0.02)                (0.05)                (0.02)             (0.02)        (0.02)          (0.03)                                (0.03)                (0.05)                (0.05)                (0.01)
Institutions                                 -0.03                  0.16                  -0.08                                                                                                                                                    -0.01
                                             (0.03)                (0.04)                (0.03)                                                                                                                                                   (0.01)
Log UK Industrialization*Latitude                                                                                                                                                                      0.59                  0.27
                                                                                                                                                                                                      (0.40)                (0.38)

R-Squared                                     0.95                  0.59                 0.92                  0.96                  0.93                  0.96                  0.93                  0.96                  0.93                  0.95

Number of Observations                         79                    79                   131                    79                   131                    79                   131                    79                   131                   103

Country Dummies                               Yes                    No                   Yes                   Yes                   Yes                   Yes                   Yes                   Yes                  Yes                   Yes

Period Dummies                                Yes                   Yes                   Yes                   Yes                   Yes                   Yes                   Yes                   Yes                  Yes                   Yes

Panel regressions; country and period dummies are included as indicated at the foot of each column. Dependent variable in Panel A is industrial output per capita, from Bairoch (1982), with up to 10
observations per country, 1750-1980. Dependent variable in Panel B is log GDP per capita, from Maddison (1995). Independent variable is the UK industrial output per capita (from Bairoch 1982)
interacted with constraint on the executive from the Polity IV dataset. The main effect of institutions is evaluated at the mean value of UK idustrialization. Polity IV provides information only for
independent countries; if a country was a colony at a particular date, we assign the lowest value of constraint on the executive. For detailed sources and descriptions see Appendix Table 1.
                                10
                                     USA

                                     CAN
                                     AUS
                                           SGP
                                           HKG       Figure 1
                                           NZL
Log GDP per capita, PPP, 1995




                                     CHL
                                     ARG
                                     VEN
                                     URY
                                 9                    MYS                MEX
                                                        COL PAN
                                                            CRI
                                     BRA

                                                                  ECU   TUN
                                           DOM                BLZ PER         DZA
                                                              GTM
                                     PRY   JAM
                                           PHL         IDN                               MAR
                                 8   GUY                      SLV BOL          EGY
                                                             LKA
                                                               HND
                                                               NIC
                                                          PAK
                                                      VNM IND
                                           HTI        LAO
                                 7                           BGD

                                     0           5             10                   15         20
                                                       Urbanization in 1500
                                                  Figure 2
                                                 USA
                                                 SGP
                                          CAN    HKG
                                10         AUS
                                                               NZL

                                                                       CHL BHS
                                                                           BRB
                                                  ARG           VEN
                                                                            KNA MEX
                                                                            GAB
                                                                          MYS
Log GDP per capita, PPP, 1995




                                                    BWA          ZAF        PAN
                                                                        COL TTO
                                                                             CRI
                                                   BRA
                                                       NAM                LCA
                                                                         GRDECU               TUN
                                                                          PER
                                                                           BLZ
                                                                         DOM
                                                                         DMA              DZA
                                                                          VCT
                                                                          GTM
                                                                PRY                   JAM
                                                                SWZ         PHL      IDN    MAR
                                                         SUR    CPV BOL SLV
                                 8                       GUY                                                EGY
                                                                          AGO                    LKA
                                                                    ZWE HND
                                                                           NIC
                                                                          CMR       GIN
                                                                          COG MRTCOM CIV
                                                                                    SEN
                                                                LSO                 GHA         SDN PAK
                                                                                                     IND
                                                                                    GMB
                                                                                    TGO VNM
                                                                          CAF
                                                                         HTILAO KEN BEN
                                                                                          UGA NPL
                                                                      TCD ZAR
                                                                    ZMBMDG          BFA
                                                                                    NGA              BGD
                                                                      NER
                                                                      MLI     ERI                     BDI
                                                                    MWI MOZ                          RWA
                                                                             TZA      SLE
                                                                                            ETH
                                 6
                                     -5                              0                                            5
                                                   Log Population Density in 1500
                                                            Figure 3
                                                                                                           USA                    SGP
                                                                                                           CAN              HKG
                                10                                                                               AUS
                                                                                                                  NZL


                                                                                 BRB                             CHL
                                                                                                                   BHS
                                                                                                                     ARG
                                                                                                                  VEN URY
                                                                         MUS
                                                                   KNA              GAB
                                                                                    ZAF MYS BWA        MEX
Log GDP per capita, PPP, 1995




                                                                                    CRI PAN           COL
                                                                                                     TTO BRA
                                                                     LCA
                                                                    NAM
                                                                    GRD                       TUN
                                                                                            ECU
                                                                                BLZ       DZA DOM    PER
                                                                                                    DMA
                                                                      GTMFJI     VCT
                                                                                       PRY
                                                                                        JAM
                                                                 SWZ IDN                PHL
                                                                                      MAR
                                                                                    SUR CPV
                                 8                                    GUY      EGY
                                                                                SLV          BOL
                                                        LKA      AGO
                                                                  ZWE        HND
                                                               GIN            CMR             NIC
                                                               COMPAK        CIV
                                                                             SEN     MRT   COG
                                                            IND SDN GHA
                                                      VNM LSO GMBTGO     CAF
                                                       LAO         HTI  BEN
                                                              KEN
                                               UGA
                                             NPL     BGD
                                                  BFA TCDMDG  ZAR        NGAZMB
                                                    NER
                                           BDI      ERI      MLI
                                         RWA    MWI                 MOZ
                                                          TZA       SLE
                                                 ETH
                                 6
                                     0                                          50                                            100
                                                                     Urbanization in 1995
Residuals from Regression of Constraint on Executive in First Year


                                                                                               Figure 4A
         of Independence on Years Since Independence




                                                                     5                   USA

                                                                                        CAN         NZL
                                                                                        AUS

                                                                                       GUY        SGP
                                                                                                  JAM               MYS MMR
                                                                                                                         LKA
                                                                                                                         IND
                                                                                                  PNG                   BGD




                                                                                                                    LAO
                                                                                                                          COL HND     BOL
                                                                                                                                      ECU         MEX
                                                                                                   DOM                        CRI
                                                                                                                              SLV
                                                                     0
                                                                                                                             PAN
                                                                                                  PHL
                                                                                                                           LBY                                      MAR
                                                                                         PRY
                                                                                         CHL        HTI                               PER                EGY
                                                                                         BRA
                                                                                         ARG
                                                                                         VEN
                                                                                         URY                                    NIC
                                                                                                                                GTM
                                                                                                                                                   DZA

                                                                                                                  IDN
                                                                                                                 VNM      PAK               TUN




                                                                     -5
                                                                          -10                                   0                                              10
                                                                                Residuals from Regression of Urbanization in 1500 on Years Since Independence
First Year of Independence on Years Since Independence
                                                                                 Figure 4B
Residuals from Regression of Constraint on Executive in




                                                           5                     USA

                                                                    CAN                   NZL
                                                                    AUS
                                                                                                ZWE
                                                                          SGP BWA GUY   LSO       MYS
                                                                                                  SOM
                                                                                                    TTO       NGAMMR
                                                                                                               JAM
                                                                                                              GMB UGA
                                                                                                                         LKA
                                                                                                                        SDN IND
                                                                                                                           BGD
                                                                                                   PNG




                                                                                                    LAO KEN
                                                                                                         MRT                BDI
                                                                                                   BOL SLVCOM
                                                                                                    COL HNDECU
                                                                                                         CRI
                                                                                                               MEX
                                                                                            ZAF         DOM
                                                           0                                   MDGCOG
                                                                                         CUB          PAN
                                                                                                     PHL
                                                                                             NERCMR
                                                                                              MLI
                                                                                           ZMB AGOTZA        BEN
                                                                                                             BFA
                                                                                                             TGO
                                                                                                             SEN        LBY
                                                                                                                          AFG NPL
                                                                                                                          MAR
                                                                                 BRA          PRY CHL HTI GNB
                                                                                                          PER                        EGY
                                                                                ARG
                                                                            NAM URY         VEN          NIC
                                                                                                         GTM
                                                                                        SWZ        GAB           DZA
                                                                                                 MOZ
                                                                                                               IDN
                                                                                                TCD CAF
                                                                                              MWI   ZAR       GIN ETH
                                                                                                              GHAVNM TUN
                                                                                                              CIV
                                                                                                                            PAK
                                                                                                                           RWA




                                                          -5
                                                               -5                                         0                                5
                                                                    Residuals from Regression of Log Population Density in 1500 on Years Since
                                                                                                  Independence
                                        Figure 5A
                                                         BOL                    MAR
                       20

                                  JAM
                                                               TUN



                       15
Urbanization in 1700




                                                                     DZA

                                                         ECU

                                                                 MEX
                       10
                                                PAK
                                                BGD
                                                LKA
                                                 LBY                   EGY
                                                   PAN
                                                   SLV
                                                  GTM
                                                  HND
                                                   BLZ
                                                   NIC
                                              COL CRI
                            BRA
                                            LAO IND
                                            VNM
                            VEN                 MMR
                        5
                                                         PER
                                  PNG
                                  PHL
                                  HTI
                                  DOM
                                  NZL


                                            MYS
                                            IDN
                            URY
                            PRY
                            USA
                            GUY
                            CAN
                            CHL
                            ARG
                            AUS   SGP
                                  HKG
                        0
                            0           5           10                     15         20
                                            Urbanization in 1500
                                                       Figure 5B
                                     USA
                                                 SGP
                                                 HKG
                                10   CAN
                                     AUS
                                                 NZL
Log GDP per capita, PPP, 1995




                                     CHL

                                     ARG
                                     URY                      VEN
                                9       MYS                                       MEX
                                                                         PAN
                                                                       COL
                                                                       CRI
                                                                     BRA

                                                                                        ECU              TUN
                                                 DOMPER                   BLZ                 DZA
                                                                          GTM
                                     PRY                                                                   JAM
                                           IDN   PHL
                                                                                                                 MAR
                                8    GUY                                    EGY
                                                                          SLV                                    BOL
                                                                           LKA
                                                                          HND
                                                                          NIC

                                                                 IND       PAK
                                                               VNM
                                                 HTI
                                                                    LAO
                                7                                          BGD


                                     0                    5                 10                      15           20
                                                                    Urbanization in 1700
                                           Figure 6A
                     Urbanization in excolonies with low and high urbanization in 1500
                       (averages weighted within each group by population in 1500)
70




60




50




40




30




20




10




0
     800   1000   1200   1300       1400      1500       1600         1700   1750     1800       1850      1910      1950   1980

                                low urbanization in 1500 excolonies           high urbanization in 1500 excolonies
                                                     Figure 6B
                                             Urbanization in Mexico, India and USA, 800-1960
                                                  (from Chandler, Mitchell and the UN)

                   100


                   90


                   80


                   70
Urbanization (%)




                   60


                   50


                   40


                   30


                   20


                   10


                    0
                         800   1000   1200   1300   1400   1500      1600          1700   1750   1800   1850   1900   1920   1930
                                                                            Year

                                                                  Mexico       India      US
                                                            Figure 6C
                                                            Just the Americas (1750-1930)
                                                                   (Mitchell Series)

                    30




                    25




                    20
Urbanization Rate




                    15




                    10




                    5




                    0
                         1750   1790   1800   1810   1820   1830    1840    1850   1860   1870    1880   1890   1900   1910   1920   1930

                                                                   Mexico    USA      Brazil     Peru
                                    Figure 6D
                           Industrial Production Per Capita, UK in 1900 = 100
                                             (from Bairoch)

400



350



300



250



200



150



100



50



 0
      1750    1800         1830        1860         1880        1900            1913           1928   1953

             US      Australia     Canada      New Zealand     Brazil    Mexico        India
                                                                                            Appendix Table A1
                                                                                      Variable Definitions and Sources

                  Variable                                                                        Description                                                                                                Source
                                                                                                                                                                      World Bank, World Development Indicators, CD-Rom, 1999. Data on Surinam is from
Log GDP per capita (PPP) in 1995           Logarithm of GDP per capita, on Purchasing Power Parity Basis.                                                             the 2000 version of this same source.

Log Population Density in 1500             Logarithm of population density (total population divided by total arable land) in 1500                                    McEvedy and Jones

Log Population Density in 1000             Logarithm of population density (total population divided by total arable land) in 1000                                    McEvedy and Jones

Log Population Density in 1 AD             Logarithm of population density (total population divided by total arable land) in 1 AD                                    McEvedy and Jones

Urbanization in 1995                       Percent of population living in urban areas with a population of at least 5,000 in 1995                                    World Bank, World Development Indicators, CD-Rom, 1999.


Urbanization in 1960                       Percent of population living in urban areas with a population of at least 5,000 in 1960                                    World Bank, World Development Indicators, CD-Rom, 1999.


Urbanization in 1900                       Percent of population living in urban areas with a population of at least 5,000 in 1900                                    Bairoch and supplemental sources, as described in the appendix.

Urbanization in1500                        Percent of population living in urban areas with a population of at least 5,000 in 1500                                    Bairoch and supplemental sources, as described in the appendix.

Urbanization in 1000                       Percent of population living in urban areas with a population of at least 5,000 in 1000                                    Bairoch and supplemental sources, as described in the appendix.


European settlements in 1900               Percent of population that was European or of European descent in 1900. Ranges from 0 to 0.99 in our base sample. McEvedy and Jones (1978) and other sources listed in Appendix Table 5 of AJR (2000).

European settlements in 1800             Percent of population that was European or of European descent in 1800. Ranges from 0 to 0.99 in our base sample. McEvedy and Jones (1978) and other sources listed in Appendix Table 5 of AJR (2000).
                                                                                                                                                           Dataset obtained directly from Political Risk Services, September 1999. This data was
Average Protection against Expropriation Risk of expropriation of private foreign investment by government, from 0 to 10, where a higher score means less  previously used by Knack and Keefer (1995) and was organized in electronic form by
Risk, 1985-95                            risk. We calculated the mean value for the scores in all years from 1985 to 1995.                                 the IRIS Center (University of Maryland). The original compilers of this data are

                                           A seven category scale, from 1 to 7, with a higher score indicating more constraints. Score of 1 indicates unlimited
                                           authority; score of 3 indicates slight to moderate limitations; score of 5 indicates substantial limitations; score of 7 Polity III dataset, downloaded from Inter-University Consortium for Political and Social
Constraint on Executive in 1990            indicates executive parity or subordination. Scores of 2, 4, and 6 g
                                                        g y        ,            ,         g                    indicate intermediate values.                        Research. Variable described in Gurr 1997.
                                           authority; score of 3 indicates slight to moderate limitations; score of 5 indicates substantial limitations; score of 7 Polity III dataset, downloaded from Inter-University Consortium for Political and Social
Constraint on Executive in 1970            indicates executive parity or subordination. Scores of 2, 4, and 6 indicate intermediate values. Set equal to 1 if       Research. Variable described in Gurr 1997.
                                           From 1 to 7. Coding as for Constraint on Executive in 1990 and 1970. Set equal to 1 if country was not independent at Polity III dataset, downloaded from Inter-University Consortium for Political and Social
Constraint on Executive in 1900            that date.                                                                                                               Research. Variable described in Gurr 1997.


Constraint on Executive in first year of   From 1 to 7. Coding as for Constraint on Executive in 1970. Date of independence is first year that country appears in Polity III dataset, downloaded from Inter-University Consortium for Political and Social
independence                               Polity III dataset.                                                                                                    Research. Variable described in Gurr 1997.
Percent of European descent 1975           Percent of population that was European or of European descent in 1975. Ranges from 0 to 1 in our base sample.
                                                    g          p p                g                         y p        g                            (             McEvedy and Jones (1978).
                                           95 for countries formed more recently). The four classifications are: Roman Catholic, Protestant, Muslim, and
Religion Variables                         "other".                                                                                                                    La Porta et al (1999)
                                           Dummy variable indicating whether country was a British, French, German, Spanish, Italian, Belgian, Dutch or
Colonial Dummies                           Portuguese colony.                                                                                                          La Porta et al (1999)
                                           Temperature variables are average temperature, minimum monthly high, maximum monthly high, minimum monthly
Temperature Variables                      low, and maximum monthly low, all in centigrade.                                                                            Parker (1997)
                                           Humidity variables are morning minimum, morning maximum, afternoon minimum, and afternoon maximum, all in
Humidity Variables                         percent.                                                                                                                    Parker (1997)
                                           Measures of soil quality/climate are steppe (low latitude), desert (low latitude), steppe (middle latitude), desert (middle
Soil Quality                               latitude), of steppe wasteland, desert dry of world gold reserves
                                           Measuresdrynatural resources are: percent winter, and highland. today, percent of world iron reserves today, percent Parker (1997)
Natural Resources                          of world zinc reserves today, number of minerals present in country, and oil resources (thousands of barrels per           Parker (1997)
Dummy for Landlocked                       Dummy variable equal to 1 if country does not adjoin the sea.                                                              Parker (1997)
Malaria in 1994                            Malaria in 1994 is percent of people living in area where falciparum malaria is endemic                                    Gallup and Sachs 1998.

Latitude                                   Absolute value of the latitude of the country, scaled to take values between 0 and 1, where 0 is the equator.              La Porta et al (1999)
Log Mortality                              Log of estimated settler mortality. From 1.7 to 6.2.                                                                       AJR (2000)
                                    Appendix Table 2
                       Abbreviation used in Log population density   Urbanization in 1500 (base
       Country name            graphs                     in 1500             sample estimate)
Former European Colonies (in our base sample)
Argentina                       ARG                         -2.21                          0.0
Australia                       AUS                         -3.65                          0.0
Bangladesh                      BGD                          3.17                          8.5
Belize                          BLZ                          0.43                          9.2
Bolivia                         BOL                         -0.19                         10.6
Brazil                          BRA                         -2.13                          0.0
Canada                          CAN                         -3.83                          0.0
Chile                           CHL                         -0.22                          0.0
Colombia                        COL                         -0.04                          7.9
Costa Rica                      CRI                          0.43                          9.2
Dominican Re                    DOM                          0.38                          3.0
Algeria                         DZA                          1.95                         14.0
Ecuador                         ECU                          0.77                         10.6
Egypt                           EGY                          4.61                         14.6
Guatemala                       GTM                          0.43                          9.2
Guyana                          GUY                         -1.55                          0.0
Hong Kong                       HKG                         -2.44                          3.0
Honduras                        HND                          0.43                          9.2
Haiti                           HTI                          0.28                          3.0
Indonesia                       IDN                          1.45                          7.3
India                           IND                          3.17                          8.5
Jamaica                         JAM                          1.53                          3.0
Laos                            LAO                          0.55                          7.3
Sri Lanka                       LKA                          2.74                          8.5
Morocco                         MAR                          2.21                         17.8
Mexico                          MEX                          0.96                         13.0
Malaysia                        MYS                          0.20                          7.3
Nicaragua                       NIC                          0.43                          9.2
New Zealand                     NZL                         -0.99                          3.0
Pakistan                        PAK                          3.17                          8.5
Panama                          PAN                          0.43                          9.2
Peru                            PER                          0.45                         10.5
Philippines                     PHL                          0.52                          3.0
Paraguay                        PRY                         -0.69                          0.0
Singapore                       SGP                         -2.44                          3.0
El Salvador                     SLV                          0.43                          9.2
Tunisia                         TUN                          2.46                         12.3
Uruguary                        URY                         -2.21                          0.0
USA                             USA                         -2.44                          0.0
Venezuela                       VEN                         -0.82                          0.0
Vietnam                         VNM                          1.82                          7.3
                                             Appendix A
                                         Urbanization Estimates

1. General
Sources
        We use four main sets of sources for information on urbanization. The first set of sources are
publications by Paul Bairoch and his associates. The second are publications by Tertius Chandler,
particularly Chandler (1988). The third are urbanization estimates that are more focussed on particular
places and time periods, but which fill important gaps in the first two sources or supplement their data in
a useful way. Finally, Mitchell (1993 and 1995) provides good urbanization data, but only for about
1750 to the present.
        Bairoch (1988) provides the best single collection and assessment of urbanization research.1
This volume reviewed evidence on urbanization from around the world since the beginning of recorded
time and we rely primarily on his assessment of plausible urbanization estimates throughout our analysis.
Even when we are forced to use specific numbers from another source, we check these estimates for
consistency with Bairoch’s various qualitative assessments (as documented in this appendix.)
        Bairoch (1988) argues that the correct threshold for measuring urbanization in general is 5,000
inhabitants. There are some exceptions, but in general this level of urban population indicates that
people are engaged in non-agricultural activities. On average over the past 200 hundred years
urbanization and industrial employment have moved in step (Bairoch, 1988, Table 29.2).
         After the publication of Bairoch (1988), Bairoch and his associates continued to develop more
detailed country-level data.2 Two publications are particularly useful. Bairoch, Batou and Chevre
(1988) provide a comprehensive dataset on European urbanization since 800 AD. We use this primarily
as a check on the European estimates in Bairoch (1988). We find no major revisions, but there is much
more detail and documentation. There is also more detailed data on Eastern Europe and the Balkans.
        The second useful publication is data on non-European urban populations collected by Gilbert
Eggimann and published by Librarie Droz as “La Population des villes des tiers-mondes, 1500-1950” on
the web at http://histsociale.isuisse.com/histsociale/ (see also www.droz.org) since fall 1999. The
Scientific Editors of this publication were Paul Bairoch and Jean-Claude Toutain. Eggimann draws
heavily on Chandler (1987) for information on early cities, and Showers (1979) for more recent data. He
also makes extensive use of country-specific sources that are well-documented in his bibliography. The
main drawback is that Eggimann’s data only covers cities that crossed the threshold of 20,000
inhabitants at some point between 1500 and 1950.3 Eggimann also does not cover the former colonies
that have become rich (the US, Canada, Australia and New Zealand).
        The second main source of urban population estimates is work by Tertius Chandler. Bairoch
(1988) reports that he used Chandler and Fox (1974) as the basis of his first estimates.4 Independently of

1
  The original French publication was in 1985.
2
  Professor Bairoch sadly passed away in 1999. We understand that a more complete global urbanization
data was under preparation at the time of his death, but have not been able to locate it. We rely instead
on official publications of his Center of International Economic History which moved to the University
of Lausanne in September 1999. Many of its publications are available through Librairie Droz (Geneva),
under the direction of Jean Batou and Bouda Etemad. For helpful communications on these issues, we
thank Professor Etemad.
3
  Eggimann originally intended to look at cities with populations above 10,000, but this proved to be
much more work. See the section on his website under “Pourquoi avoir abandonné la limite de 10'000
pour se limiter à 20'000?”
4
  Bairoch (1988) writes, “Indeed, had it not been for the publication of the investigations of Chandler and
Fox (1974), I would never have done the research reported here, at least not in its present audacious
form” (p.519). Bairoch explains that his task was to enlarge the list of cities and using more recent


                                                     i
Bairoch’s later work, Chandler (1987) extended and refined these estimates. Chandler’s data is
attractive because it is more detailed for non-European countries than what has been published by
Bairoch’s group (with the exception of Eggimann, who only covers countries that are now part of the
“Third World”). In particular, Chandler reports numbers by individual city since 800 A.D. as much as
possible. He also has information on cities before 800 A.D., although this data is less comprehensive.
Eggimann (1987) cites Chandler as a source for some countries.
         However, there are three limitations of the Chandler data that make us prefer Bairoch’s estimates
(augmented with Eggimann’s data). First, Chandler’s minimum city size for Europe and the Americas is
20,000 (although he has estimates for some years with a lower threshold). For the reasons outlined
above, this is not as satisfactory as Bairoch’s 5,000 inhabitants threshold. Second, for Asia, Chandler’s
minimum city size is 40,000, making it difficult to compare precisely with his own estimates for Europe
and the Americas. For example, this leads to a much lower (and rather implausible) estimate of
urbanization in India and China compared with what he finds for Europe and the Americas, and
compared with the assessments in Bairoch (1988). Third, Chandler’s data stops in 1850 (1861 for the
Americas). Given the importance for our analysis of the urbanization that came with industrialization,
we prefer to use data that is consistent into the twentieth century. Nevertheless, Chandler (1987)
provides a very useful check on the Bairoch-school, particularly because it was an important source for
that work. Chandler (1987) also provides usefully detailed urbanization estimates across almost all
countries.
         Our third set of urbanization estimates comes from scholars who have looked a urbanization in
particular places or periods. De Vries (1984) provides a detailed set of estimates for urbanization in
Europe, with a minimum city size of 10,000. We use this series as a check on the Bairoch-school, and
find some differences. However, the differences do not affect our main results (e.g., concerning the
persistence of prosperity within Europe and the absence of a reversal of fortune within Europe since
1500). Hohenberg and Lees (1985) also assess urbanization in Europe over the past 1000 years.
McEvedy (1967, 1992) reports approximate urbanization estimates for Europe and the Middle East from
the neolithic revolution to 1000 AD. He also provides a useful series of maps that show how the
development of large urban centers related to trade flows. We use these numbers in our assessment of
the persistence of urbanization.

Calculating Urbanization
         Bairoch generally provides an estimate of urbanization (i.e., percent of the total population
living in urban areas). However, most other scholars of non-European urbanization, including
Eggimann, Chandler and Fox, and Chandler, provide only estimates of total urban population (in cities
about their minimum population threshold.) To convert these into urbanization estimates, we use the
population figures in McEvedy and Jones (175). We also use the land area and population estimates in
McEvedy and Jones (1975) to calculate population density.5
         We report our urbanization estimates in three sections. First, we explain in detail our baseline
estimates for 1500, including both our main sources (usually from Bairoch-Eggimann) and supporting



sources. “This project was conducted relatively in depth for the following regions and periods: Europe
from 800 to 1800 (ten periods), Latin America from 1600 to 1950 (nine periods), Africa from 1900 to
1950 (five periods), and the larger countries of Asia from 1500 to 1950 (nine periods).
5
  The McEvedy and Jones work is nearly 25 years old but as far as we know it has not been superceded
as a single comparable set of estimates across all countries for the past 2000 years. Some estimates are
still controversial (e.g., on pre-European America, but McEvedy and Jones are rather conservative and
higher population estimates for that region would help our results) and estimates for Africa must be
treated with particular caution (as McEvedy and Jones acknowledge.) In a recent assessment, the US
Bureau of the Census (see the main text for the precise reference) showed that McEvedy and Jones’s
aggregate estimates compare favorably with more recent estimates. Using the McEvedy and Jones data
is standard in the economics literature; see for example Kremer (1993).


                                                    ii
evidence (usually from Chandler). Second, we provide evidence on the development of urbanization
from 1500. Third, we provide evidence on urbanization before 1500.

2. Base Estimates for 1500
          Our base data consists of Bairoch’s (1988) urbanization augmented by the work of Eggimann
(1999). Merging these two series requires us to convert Eggimann’s estimates, based on a minimum
population threshold of 20,000, into Bairoch-equivalent urbanization estimates, based on a minimum
population threshold of 5,000. It is not appealing to look only at towns above the threshold of 20,000
people, as this would ignore even substantial urbanization levels. For example, England before 1500 had
only one city with population over 20,000, but an urbanization rate of 8%; see Chandler 1987, p.19,
Bairoch 1988, p.179, and Bairoch, Batou and Chevre 1988, pp.32-35.
          To convert between the two sets of estimates, we run a regression of Bairoch estimates on
Eggimann estimates for all countries where they overlap in 1900 (the year for which we have most
Bairoch estimates for non-European countries).6 There are thirteen countries for which we have good
overlapping data. This regression yields a constant of 6.6 and a coefficient of 0.67. We use these
estimates below to generate Bairoch-equivalent urbanization estimates.
          We have also checked the robustness of our results using alternative methods of converting
Eggimann estimates into Bairoch-equivalent numbers. We have calculated country-specific conversion
ratios and conversion ratios at the regional level (e.g., North Africa and the Andean region separately).
We have also used Bairoch’s overall assessment of urbanization for broad regions, e.g., Asia, without the
more detailed information from Eggimann. None of these alternative methods affect any of our main
results.
          We provide urbanization estimates for all countries that now have a population over 500,000.
Most countries with a population under 500,000 either lack reliable data or are not fully independent
(e.g., this is true of many small islands that are nominally independent in the Caribbean or the Pacific) or
both.

Europe in 1500
         The best available urbanization estimates are those for Europe. After more than 40 years of
detailed quantitative studies, there has been remarkable convergence in the opinion of scholars about the
extent of urbanization in the late Middle Ages.
         Bairoch (1988, table 11.2, p.179) reports urbanization in Europe in 1500, using a population of
5,000 as the minimum threshold. These estimates, together with his numbers for 1300 and 1700 are
reproduced in Appendix Table 3. Through 1800 this data is in the form of a range of estimates and we
use the mid-point of this range in our base data.7
         Bairoch, Batou and Chevre (1988) provide estimates for the population of every European town
that had “at some time between 800 and 1800, 5,000 or more inhabitants” (p.ix). We use these estimates
where Bairoch (1988) does not provide data, and as a check on Bairoch (1988). However, because
Bairoch, Batou and Chevre (1988) do not provide any non-European data, we prefer to use Bairoch
(1988) as the basis for our comparative analysis.
         One important question is which urbanization estimate to use when borders have changed. For
example, for the Baltic states (Latvia, Lithuania and Estonia) should we use their urbanization or the
urbanization of Russia-in-Europe (arguably the relevant geopolitical entity in 1500)? We use the
political entity selected by Bairoch and his associates to report urbanization numbers. We also report
alternative estimates wherever available. None of these alternative estimates affect our main results.


6
  Alternatively, two sets of estimates indicate that a rough and conservative conversion factor is to
multiply the Eggimann urbanization estimates by two. We have constructed an alternative set of
estimates using this method and none of our main results were affected.
7
  Slightly different estimates are reported in Bairoch, Batou and Chevre (1988), but using these does not
affect our results.


                                                    iii
        We have direct estimates from Bairoch (1988) for France (9-12%), Germany (7-9%), Belgium
(30-45%), the Netherlands (20-26%), Portugal (11-13%), Spain (10-16%), and Switzerland (6-8%).8
Land area has not changed significantly, except that Germany lost some of its relatively less urbanized
eastern area (i.e., the area occupied by Germany today almost certainly had a higher urbanization in 1500
than did the Germany of 1500). For Britain we use Bairoch’s (1988, p.179) estimate for England (7-
9%).9
        We use the Scandinavia estimate (6.5%) for Sweden, Norway, Denmark and Finland. This is
probably high for Finland.10 We use the Russia estimate (4.5%) for Russia, Ukraine, Belarus, Estonia,
Latvia and Lithuania.11 We use the Austria-Hungary (6.5%) estimate for Austria, Slovakia, Hungary and
the Czech Republic.12
        Bairoch (1988) does not provide an estimate for Romania or Poland. We use the estimates of
3% (Romania) and 6% (Poland) urbanization from Bairoch, Batou and Chevre (p.259).
        From Bairoch, Batou and Chevre, we know the urban population of Greece in 1500 was 35,000.
The population from McEvedy and Jones (p.113) was 1m. So we use an urbanization rate of 3.5% in our
base sample.
        From Bairoch, Batou and Chevre (p.13), the urban population of Bulgaria in 1500 was 97,000.
The population from McEvedy and Jones (p.113) was 800,000. We use an urbanization rate of 12.1% in
our base sample. Note that this was by far the most urbanized country in the Balkans (with the exception
of Turkey-in-Europe.)
        From Bairoch, Batou and Chevre (p.3), the urban population of Albania in 1500 was 5,000. The
population from McEvedy and Jones (p.113) was 200,000. We therefore use an urbanization rate of
2.5% in our base sample.
        From Bairoch, Batou and Chevre (p.69), the urban population of the countries that later
constituted Yugoslavia was 112,000. The population from McEvedy and Jones (p.113) was 2.25m. We
use an urbanization rate of 5% for Macedonia, Croatia, and Slovenia. Note that Bosnia-Hercegovina and
Serbia are not in our dataset as we do not have data for GDP per capita in 1995.13


8
  The estimates from Bairoch, Batou and Chevre (1988, p.259) are quite similar: France, 8.8%;
Germany, 8.2%; Belgium, 28%; the Netherlands, 29.5%; Portugal, 15.0%; Spain, 18.4%; and
Switzerland, 6.8%.
9
  Urbanization in the UK as a whole was almost certainly lower; Bairoch, Batou and Chevre give an
estimate of 4.6%. We do not have reliable urbanization estimates for Ireland. Using a threshold of
10,000 inhabitants, de Vries (1984, Appendix 3) puts the urban population of England and Wales at
80,000 in 1500, with Scotland at 13,000, and Ireland at 0. Bairoch, Batou and Chevre (p.39) give a total
Irish urban population of 19,000. For 1500, McEvedy and Jones (p.43) give the population of England
and Wales as 3.75m, Scotland as 500,000, and Ireland as 800,000. The de Vries urban population
numbers would therefore imply urbanization of 2.1% in England and Wales and 2.6% in Scotland. With
their lower threshold, Bairoch, Batou and Chevre (1988, p.259) give total urban population for the
United Kingdom as 230,000; assuming this includes Ireland implies urbanization of 4.6%. Their implied
Irish urbanization rate is 2.4%, but only Dublin had more than 5,000 inhabitants.
10
   De Vries 1984 p.270 does not report any towns with a population over 10,000 between 1500 and 1800.
Bairoch, Batou and Chevre report a total urban population in 1500 of only 2,000 (in Turku). McEvedy
and Jones (p.53) put the total Finnish population in 1500 at 100,000, implying urbanization of 2%.
11
   Bairoch, Batou and Chevre (1988, p.259) report urbanization for “Russie d’Europe” as 5.4% in 1500.
12
   Bairoch, Batou and Chevre (1988, p.259) provide an urbanization estimate of 4.8% for the aggregate
of “Autriche-Hongrie-Tchéc.”
13
   Bairoch, Batou and Chevre (1988, p.259) put total urbanization in the Balkans at 11.3%. This average
estimate is puzzling and it is not clear to which countries they are referring. By their own numbers, total
urbanization in Greece, Bulgaria, Albania and future Yugoslavia was 249,000. McEvedy and Jones
(p.113) give total population in the Balkans as 4.5m (excluding 300,000 people in Turkey-in-Europe),
which gives urbanization of 5.5%.


                                                    iv
The Americas in 1500
          There is some doubt about the exact level of urbanization in the Americas before the Europeans
arrived, but the relative distribution of urbanization across North and South America is clear. To the
north of Mexico, there was relatively little urbanization. In Mexico, other parts of central America and
northern/western South America, urbanization was relatively high (probably average urbanization was
just a little lower than in Asia at this time, but we cannot be sure).
          For Latin America, Bairoch (1988, Table 24.1, p.389) reports the distribution of cities with
populations of 20,000 or more. In 1500, there were 20 such cities in the Northern Andes (Bolivia,
Colombia, Equador and Peru), and 10 in Mexico. There were no cities above this threshold size in the
Caribbean, Brazil or the “Temperate Regions” (Argentina, Chile, and Uruguay). He further estimates
that there were 4-7 cities with populations of more than 50,000 and “25-30 (perhaps even 40)” with
populations of 20,000-50,000 (Bairoch 1988, p.66). There were two main concentrations of urban
population: the Aztecs, Tairona, Tarascans, and Zapotecs (roughly in the area now Mexico); and the
Incas, Mayas, Chibchas, and Cakchiquels (Central and north-western South America.) Urbanization
may have been as high as 10-13% in some places, but Bairoch’s assessment is that it was more likely
around 7% on average. For the Chimu region of the Andes, Bairoch cites favorably an estimate as high
as 14% (Bairoch 1988, p.66). For the Andean region as a whole, Bairoch estimates the low end for
urbanization was 2-3%; and the high end was 10-13%.
          For Colombia in 1500, Eggimann’s estimates imply urbanization of 2%. As a Bairoch-
equivalent urbanization measure, this would be 7.94%. We use this in our baseline data.
          For Ecuador in 1500, Eggimann estimates urbanization as 15%. Part of this urban population
should be ascribed to the area occupied by modern-day Bolivia. Ecuador and Bolivia both had 6% in
Eggimann’s numbers, if we ascribed urbanization evenly (probably there was more in Ecuador). We use
an estimate of 10.62% urbanization for both in our baseline data.
          For Peru in 1500, Eggimann estimates urbanization was 5.8%. This is converted to a Bairoch-
equivalent urbanization level of 10.49%. We use this in our baseline data.
          For Guatemala in 1500, Eggimann estimates 3.8% (deflating the urban population using whole
population of Central America). Clearly part of this should be ascribed to other parts of Central
America. Our baseline data has an estimate of 9.2%, which seems quite reasonable for Costa Rica,
Honduras, Panama, Nicaragua, Belize, El Salvador, and Guatemala on average (it might have been lower
in Costa Rica and Panama and higher in Guatemala, but there is no way to be sure.)
          For Mexico in 1500, Eggimann estimates urbanization was 12.3%. Our baseline data uses the
Bairoch-equivalent urbanization estimate of 14.84%. According to Bairoch, p.389, Mexico had half the
number of cities with at least 20,000 population as the northern Andes (10 vs. 20) in 1500. The
population in both areas was about the same according to McEvedy and Jones, around 5m. When
Europeans arrived in 1519, Tenochtitlan (now Mexico City) probably had a population in the range of
150-200,000 and lay in the middle of a heavily urbanized region with up to 400,000 people living in
smaller cities. “During this same period, there were only four European cities this size; and on the
Iberian peninsula, the largest cities at that time, Granada and Lisbon, had around 70,000 inhabitants
each” (p.63).
          Despite his extensive discussion of urbanization in South America, Bairoch makes no mention
of pre-European urbanization in Argentina or Brazil or other parts of South America. For Brazil in 1500,
Eggimann estimates 0.1% urbanization (just 1,000 people in a town). We assign a value of zero in our
baseline data. For other countries in South America with low levels of urbanization we assume the same
value. Thus we use 0% for Argentina, Chile, Uruguay, Paraguay and Guyana.
          While we also do not know the precise level of urbanization in North America, Bairoch (1988) is
clear that this was much lower than in Central and South America. He refers to the overall situation as
“Culture without Cities” (p.68). There was probably some urbanization, for example the Anasazi people
may have had cities with populations of 10,000 or more between the eleventh and fourteenth centuries,
but most of the southeastern and eastern United States and Canada was much less urbanized (with only
the Huron-Iroquois having even slash-and-burn agriculture). He estimates that population density north
of the Rio Grande before the Europeans was only 0.2 people per square kilometer or 0.4 people per



                                                   v
square kilometer if we exclude all “truly uninhabitable” areas (p.69). “In any event, the displacement of
villages caused by the practice of slash-and-burn cultivation considerably reduced the likelihood of the
creation of urban centers by preventing or impeding any permanent large-scale settlements (p.69).” In
his assessment, there were no “genuine cities” (p.69).
         For the Americas, Chandler uses a minimum city size of 20,000 inhabitants. According to
Chandler (1987), in 1500 urbanization was 6.5% in Mexico, 2.5% in Peru, 5% in Ecuador, 2% in
Colombia. His urbanization estimate for Guatemala is 19.6%, but this is presumably case where the
urban population should be divided by a total population that is larger than the number of people who
were living then within the borders of Guatemala now. Chandler does not report separate data for
Bolivia, but again this population was almost certainly part of the urban areas in Ecuador. He does not
report any urbanization for Canada, Brazil, Paraguay, Uruguay, Argentina or Chile.14
         We have very little information on urbanization in the Caribbean. Cuba, the Dominican
Republic, Haiti and Jamaica all probably had some urbanization, although clearly they had less
urbanization than Central America. We bring these countries into the analysis only when we use
population density data.

The Pacific Region in 1500
         For Australia, Bairoch’s assessment is that “agriculture and urbanization emerged only in the
nineteenth century, brought by the European immigrants” (p.69). The pre-European population existed
as hunter-gatherers with no towns. We use a value of 0% in our baseline data.
         There probably was some agriculture in Western Polynesia (p.70). In Bairoch’s assessment, this
level of agriculture generally supports a minimal level of urbanization at around 3%. Bairoch does not
discuss pre-European New Zealand explicitly, but other sources suggest that the culture was similar to
Western Polynesia. We therefore assign an urbanization level of 3%.

Africa in 1500
        There is reasonably good urbanization data for North Africa. For Tunisia in 1900, Eggimann
estimates urban population as 305,000 out of 1.5m; urbanization of 20.3%. In 1500, Eggimann estimates
urban population as 65,000 out of 800,000; urbanization of 8.1%. We use the Bairoch-equivalent
urbanization estimate of 12.3% in our baseline data.
        For Libya in 1900, Eggimann estimates urban population as 57,000 out of 800,000; urbanization
of 7.1%. In 1500, Eggimann estimates urban population as 16,000 out of 500,000; urbanization of 3.2%.
We use the Bairoch-equivalent urbanization estimate of 8.74% in our baseline data.
        For Morocco in 1900, Eggimann estimates urban population as 433,000 out of 5m; urbanization
of 8.7%. In 1500, Eggimann estimates urban population as 250,000 out of 1.5m; urbanization of 16.7%.
We use the Bairoch-equivalent urbanization estimate of 17.79% in our baseline data.
        For Algeria in 1900, Eggimann estimates urban population as 499,000 out of 5m; urbanization
of 10%. In 1500, Eggimann estimates urbanization as 165,000 out of 1.5m; urbanization of 11%. We
use the Bairoch-equivalent urbanization estimate of 13.97% in our baseline data.
        For Egypt in 1900, Eggimann estimates total urban population was 1.603m, while McEvedy and
Jones put total population at 10m. For 1900, Eggimann’s estimate is 16% and Bairoch’s estimate is
19%, giving a ratio of 1.19. Urbanization in 1500, according to Eggimann, was 475,000, while
population was 4m. In 1500, Eggimann’s estimate is 11.9%, implying a Bairoch-equivalent estimate of
14.2%. We use the Bairoch-equivalent urbanization estimate of 14.57% in our baseline data.
        Including Egypt in 1900 the population of North Africa was 22.3m. Urban population from
Eggimann was 2.9m. Implied urbanization for all of North Africa was therefore 13%. Bairoch (p.429)
estimates North African urbanization in 1900 as 16%. In 1900 North Africa without Egypt had a

14
  Chandler suggests there was some urbanization in the area now occupied by the United States; he
reports an urban population of 20,000 in Nanih Waiya. With a total population of 800,000 (from
McEvedy and Jones), this implies urbanization of 2.5%. However, this seems to have been an isolated
number. He does not report any urbanization for 1400 or 1600 or any other date before 1750.


                                                   vi
population of 12.3m. Urban population for this area according to Eggimann was 1.294m, which implies
an average urbanization level of 10.5%. In 1500, the population of the same area was 4.3m, while urban
population according Eggimann was 0.496m. This would imply an urbanization rate of 11.5%.
         In 1500, the total population of North Africa was 8.3m (McEvedy and Jones p.221, p.224 and
p.227). Urban population according to Eggimann was 0.971m, implying an urbanization for all of North
Africa of 11.7%.
         While the data on sub-Saharan Africa is worse than for any other region, it is clear that
urbanization before 1500 was at a higher level than North America or Australia. By 1500 urbanization
was “well-established” in sub-Saharan Africa (Bairoch 1988, p.55). In fact, there are much older records
from Arab explorers regarding African cities. “In summary, there were sizable cities in Black Africa by
1000 B.C., if not earlier” (Bairoch 1988, p.56).
         Leading examples of urbanization around the time of 1500 are to be found in the kingdoms of
Ghana, Songhai, Benin, Congo, Zimbabwe and the Yoruba states (of which Benin was one) (pp.56-60).
Around 1500, the city of Benin had a population in the range of 60-70,000 and was a “well-ordered
urban center with a system of water conduits and a sizable artisanry working at an advanced technical
level” (p.58). The same kingdom had about 10 other cities at this time. At the time the Portuguese first
made contact with the kingdom of the Congo (in 1484), the capital city probably had a population in the
range of 40-60,000 (p.58).
         There were also a number of cities that were closely connected with or even founded by Islamic
traders from further north. Gao, Gober, Jenné, Kano, Kazargamu, Timbuktu and Zaria all fall into this
category (p.61). “Around the beginning of the sixteenth century, these cities had populations ranging
from between twenty-five thousand and seventy thousand, with an average somewhere on the order of
forty thousand” (p.61).
         Bairoch estimates that urbanization on average in sub-Saharan Africa was 3% in 1900 (p.430).
For sub-Saharan Africa in 1800, Bairoch (1988, p.393) estimates the total urban population was 3-4
million, of which Europeans were between 30,000 and 120,000. For 1900, he puts urbanization in the
range of 2-5% (p. 413-414, in particular Table 26.1). His discussion makes it clear that there was
substantial variation within sub-Saharan Africa. West Africa, West-Central Africa (i.e., around the
Congo) and East Africa (around Zanzibar and Kenya) were clearly more urbanized than Southern Africa
(with the exception of Zimbabwe, but the decline of this civilization can be placed around 1450).
         Given the weakness and incompleteness of data for sub-Saharan Africa, we do not include any
estimates in our baseline dataset.

Asia in 1500
         Bairoch estimates that average urbanization in Asia in 1900 was 9%, but he argues that this was
higher in the eighteenth century before European colonization (p.430). He also argues there was more
homogeneity of urbanization within Asia than within Africa or America (pp.430-431).
         Bairoch (1988, table 25.1, p.407) reports urbanization in Asia in 1900 as follows.
China, 7-9%; India, 9-11%; Indonesia, 5-8%; Philippines, 14-16%; Vietnam, 6-9%; Iran, 12-15%;
Korea, 8-11%; Thailand, 8-10%; Turkey, 15-18%; and Asia as a whole, 8-10%. Bairoch expresses the
general view that urbanization had not changed much in the years before 1900. Bairoch (1988, p.406)
estimates that “Around 1700, as well as around 1930, 10-12% of the population of Asia lived in cities.”
He further estimates that because urbanization fell in China, the level of urbanization had dropped to 8-
10% by the start of the nineteenth century.
         Bairoch emphasizes that there is a long-standing history of urbanization throughout Asia. In
fact, prior to 1500, Asia was almost consistently among the most urbanized places in the world. By 1900
this had changed. For example, Bairoch 1988, p.350, states, “Between 100 B.C. and A.D. 1200 there
were always one or two Asian cities among the three largest in the world. And between 1200 and 1820
there were invariably six or seven Asian cities (even excluding the Middle East) among the ten largest in
the world. By contrast, there were only three around 1875, and none at all in 1900.”
         Overall, “Around 1500 the world appears to have had some fifty to sixty cities with populations
or more than 100,000, and all but four lay in the regions destined to become the Third World of today”



                                                   vii
(p.436). Most of those cities were in Asia, with perhaps 2-3 in the Americas, and 3-4 in North Africa. In
contrast, there were only 4 cities in Europe with more than 100,000 in population in 1500” (Bairoch
1988, Table 27.3, p.437).15
         For China in 1900, Eggimann provides an urbanization estimate of 2.95% and Bairoch directly
estimates 7.75%. For 1500, Eggimann gives 2.18%, and we use a Bairoch-equivalent urbanization
estimate of 8.06% in our baseline data. This is a little low, as Bairoch says, p.403, that urbanization in
China in the sixteenth century was 11-14% (falling to 6-7.5% by 1850 but rising slightly to 7-8.5% by
1900.) According to Chandler, urbanization in China in 1500 was 2%.
         For China, Bairoch favors an estimate of urbanization by the twelfth century around 10-13%,
taking the population threshold of 5,000 for a city (p.353). This was 1-2 points higher than Europe
(excluding Russia) at the same time and “close to the maximum reached in Europe (minus Russia) at any
time before the Industrial Revolution.” He justifies this level of urbanization partly by reference to the
relatively high levels of agricultural productivity in China at that time (p.354). At the beginning of the
sixteenth century, the level of urbanization in China was probably 11-14% (Bairoch 1988, p.356). By
the middle of the nineteenth century, urbanization was probably 6-7.5% (Bairoch 1988, p.357.)
         Regarding India, Bairoch (1988, p.350) writes, “By 1300, probably five cities, and perhaps six to
ten, had surpassed [100,000 population]. It may be recalled that during this same period, with a
population of some 80-100 million, just about the same as India’s, Europe also had five cities with
populations of more than 100,000.” According to Bairoch, urbanization in India in 1300 was roughly at
the same level as in Europe. By the early sixteenth century the largest city in India, Vijayanagar, had a
population of 500,000 while Paris, the largest city in Europe, had a population of only 200-250,000.16
         For 1500, Eggimann’s estimate for India, Pakistan and Bangladesh combined is urban
population 2.323m. With a population of 80.6m, this implies urbanization was 2.9%. We convert to our
baseline Bairoch-equivalent urbanization estimate of 8.54% for India. We use this also for Pakistan,
Bangladesh and Sri Lanka.17
         For India in 1900, Eggimann’s estimate is 6.9% and Bairoch’s estimate is 10%. But Bairoch and
Eggimann overlap with an estimate of urban population closer to 1500. In 1700, Eggimann reports an
urban population of 2.879m; McEvedy and Jones report a total population of India of, implying
urbanization of 2.2%. For India, Bairoch cites favorably the estimate of urbanization around 15% in
1600 (but indicates this may be on the high end) (p.351).
         The midpoint of Bairoch’s estimated urbanization in this year is 12%. Bairoch’s assessment is
that urbanization in India declined after 1500.
         Bairoch reports direct estimates of urbanization in India from census figures as follows (p.400):
in 1700, 11-13%; in 1800, 9-11.5%; in 1850, 7.5-9.5; in 1900, 9-10; in 1940, 14-16%. He argues
forcefully that the evidence of deindustrialization in the nineteenth century points to a decline of
urbanization.
         According to Bairoch’s assessment, China and India probably had about the same level of
urbanization in 1500, although India was at the top of an urbanization cycle and China was at the
bottom. According to Chandler, urbanization in India in 1500 was 1.8%.
         At the start of the eighteenth century, Bairoch argues that Japan had urbanization of 11-14%.
Bairoch, pp.45-46: urbanization in Japan dates only from about 800 AD, but large cities developed fast;
largest city was larger than in Europe at this time. In the sixteenth century, Bairoch p.46 says
urbanization was “certainly less than 10% and probably on the order of 5-8%”. We use a rate of 6.5% in
our baseline estimates. Eggimann does not have an estimate for Japan as it is not in the “Third World.”

15
   Bairoch (1988) doesn’t name those four European cities. According to Chandler, there were five:
Paris had a population of 185,000, Venice had 115,000, Naples had 114,000, and Adrianople had
127,000. The largest city in Europe was Constantinople, with 200,000 inhabitants.
16
   Chandler’s estimate for Vijayanagar is XX and for Paris is YY…
17
   We have to estimate the population of India from McEvedy and Jones’s estimate of population on the
subcontinent. We use the population ratios of 1945 (check), when India was 80.6% of the total
subcontinent population. 160x0.806=128.96m


                                                   viii
According to Chandler, urbanization in Japan was as high as 5% in 800 but fell to a low of 0.2% in 1500;
it was 4.8% in 1600.
         Urbanization in Japan was likely 14-15% by 1800 and continued to rise (Bairoch 1988, p.360).
Note that none of this increase was due to international trade, as Japan was cut off from the outside
world at this time.
         Urbanization in Byzantium (present-day Turkey) was “certainly more than 8-9%, and probably
around 10-13%” for several hundred years after 1000 A.D. (Bairoch 1988, p.370). For Turkey in 1900,
Eggimann’s estimate or urbanization is 12.9 and Bairoch’s estimate is 16.5%, giving a ratio of 1.28. For
1500, Eggimann’s estimate is 7.4%, which implies a Bairoch-equivalent urbanization estimate of
11.56%. We use this in our baseline data. According to Chandler, in 1500 urbanization in Turkey was
5.9%.18
         For Thailand in 1900, Eggimann’s urbanization estimate is 6.25% and Bairoch’s estimate is 9%.
For 1500, Eggimann’s estimate is 6.3, implying a Bairoch-equivalent estimate of 10.82%, which we use
in our baseline estimates. According to Chandler, urbanization in Thailand in 1500 was 6.3%.
         For Korea in 1900, Eggimann’s urbanization estimate is 4.8% (note that we combine north and
south Korea’s urban population for this calculation) and Bairoch’s estimate is 9.5. In 1500, Eggimann’s
estimate is 3.1%, implying a Bairoch-equivalent estimate of 8.68%. According to Chandler,
urbanization in Korea in 1500 was 4.1%.
         For Indonesia in 1900, Eggimann’s urbanization estimate is 2.28 and Bairoch’s estimate is 6.5,
giving a ratio of 2.85. For 1500, Eggimann does not have an estimate. From Eggimann, the urban
population in Malaysia in 1500 was 80,000. The population of Malaysia was only 400,000, which
would imply 20% urbanization. However, making a distinction between Malaysia and Indonesia at that
time is quite artificial.19 The population of Indonesia was 7,750,000, giving a total population of 8.15m,
and an Eggimann urbanization estimate of 1%. In our baseline data we use the Bairoch-equivalent
estimate of 7.27% for both Malaysia and Indonesia. According to Chandler, urbanization in Indonesia
and Malaysia (using combined urban populations and combined total populations) was 0.5% in 1500.
         Eggimann reports the urban population in Laos in 1500 as 40,000. The population was 400,000,
giving an urbanization rate of 10%. But it seems more reasonable to ascribe part of this urbanization also
to Vietnam which, in 1500, had a population of 4m. This implies urbanization of 1%. Converting this to
Bairoch-equivalent urbanization gives 7.27%, which we use in our baseline estimate. According to
Chandler, urbanization in Laos was 10% in 1500, but he does not give any urban population after that
date.
         For Vietnam in 1900, Eggimann’s estimate is 2.9% and Bairoch’s estimate is 7.5. There is no
data from Eggimann for 1500. According to Chandler, urbanization in Vietnam in 1500 was 2%.
         According to Chandler, urbanization in Cambodia was 3.1% in 1400; he does not give an
estimate for 1500.
         According to Eggimann, in 1500 there was an urban population of 110,000 in Burma/Myanmar.
With a population of 4m, this implies an urbanization rate of 2.8%. Converted into Bairoch-equivalent
urbanization, this gives 8.48% and we use this in our baseline estimates. According to Chandler,
urbanization in Burma was 3.5% in 1500.
         We assign an urbanization value of 3% to the Philippines, based on the qualitative assessment by
Bairoch (1988). On p.48: Bairoch says of the Philippines, “before the arrival of the Spanish colonizers,
there were no genuine cities.”
         There is not much information on Singapore and Hong Kong. We know that neither area was
developed and neither was even a minor port. Hong Kong had some small fishing villages. We assign
3% urbanization to Singapore and Hong Kong (Bairoch’s estimate of urbanization when there is settled
agriculture), but check our results carefully under alternative assumptions. Both Singapore and Hong

18
   The Ottoman Turks took Constantinople in 1453 and over the next one hundred years developed a
major international city. By 1600 the population was 700,000 (up from 200,000 in 1500, according to
Chandler) and according to Eggimann urbanization reached 12%.
19
   Cite material from A History of Malaysia here


                                                    ix
Kong were built up by the Europeans, i.e., the Europeans did not take over an existing port but rather
started from scratch.

The Middle East in 1500
        The Middle East was one of the first regions of the world to be urbanized (see the section on
urbanization before 1500 below.) Around the year 1000 A.D., the Muslim world had urbanization
around 10-13%, compared with only 8-9% in Europe (excluding Russia and Spain) (Bairoch 1988,
p.375). In 1500, however, the region was probably no more urbanized than Western Europe on average
and most likely less urbanized than the most advanced parts of Europe and Asia.
        For Iran in 1900, Eggimann’s urbanization estimate is 10% and Bairoch’s estimate is 13.5. For
1500, Eggimann’s estimate is 11.95, giving a Bairoch-equivalent urbanization estimate for our baseline
estimates of 14.61%. According to Chandler, urbanization in Iran was 11.1% in 1500.
        According to Eggimann, the urban population in Iraq was only 12,000 in 1500. Total population
was 1m (McEvedy and Jones, p.151), giving urbanization of 1.2%. We use the Bairoch-equivalent
urbanization estimate of 7.4%. According to Chandler, urbanization in Iraq was 9% in 1400 and 5% in
1700, but he does not provide numbers for 1500 or 1600 (i.e., no city was above his minimum
threshold).
        According to Eggimann, the urban population in Saudi Arabia in 1500 was 34,000 and in 1600
was 80,000. McEvedy and Jones (p.147) put the population of “The Interior” at 2m in 1500 and perhaps
2.1m in 1600. Eggimann does not report any urbanization numbers for this year for the Gulf, Oman or
Yemen (although he does for later years). In comparison, McEvedy and Jones put the population in
1500 at 100,000 on the Persian Gulf Coast, 2.25m in Yemen, and 200,000 in Oman. This implies an
Eggimann urbanization estimate of 1.7% if we use just the population of the Interior, 1.5% if we use the
population of the Interior plus the Gulf and Oman, and 0.7% if we include also the population of Yemen.
We use 1.5% as our Eggimann estimate, which implies Bairoch-equivalent urbanization of 7.6%.
According to Chandler, urbanization in Saudi Arabia in 1600 was 1%. He does not provide numbers for
1500.
        According to Eggimann, in Syria the urban population was 137,000 in 1500. The population of
Syria and the Lebanon combined was 1.5m in 1500 (McEvedy and Jones, p.139), most of which was in
Syria. This implies an urbanization rate of 9.1%, which converts into a Bairoch-equivalent urbanization
estimate of 12.7%. According to Chandler, urbanization in Syria/Palestine was 6.4% in 1500.
        According to Chandler, urbanization in Central Asia was 4.9% in 1500 (using the population
numbers for Turkestan from McEvedy and Jones.) This converts into a baseline Bairoch-equivalent
urbanization estimate of 9.9%.
        Eggimann does not report any urbanization numbers for Lebanon or Jordan in 1500.


3. Urbanization from 1500 to 2000
         There has been a great deal of scholarly work on urbanization in Europe since 1500 (to be
completed using de Vries, BBC). The pattern of urbanization in 1500 in Eastern Europe fits the
discussion of the situation in 1800 from de Vries (1984), Appendix 4. This suggests there was
persistence within Eastern Europe.
         In Europe as a whole, urbanization did not increase from 1700 to 1800 (Bairoch 1988). For the
Americas, Bairoch (1988) comments as follows “Around 1700 North America north of the Rio Grande
was still very thinly populated, there being some 250,000 Europeans, some 30,000 blacks, and perhaps
one to three million native Americans. But in contrast to what was the case south of the Rio Grande, the
white newcomers did not confront indigenous urban cultures… [W]hile certain of the cultures of
northern North America had attained a high degree of development in art and in a number of techniques,
they remained on the whole essentially rural and even, in most cases, preagrarian.” (p.221, emphasis
added). For 1800, Bairoch (1988, p.222) estimates that urbanization in North America was only 5-6%,
compared with 12% in Europe. Around the same time, urbanization in Latin America was around 12-
14% (p.222).



                                                    x
By 1900 North America was more urbanized than Europe.
         In 1800, Bairoch (1988, p.389) estimates that Latin America was the most highly urbanized part
of the world, with urbanization around 13-16%, i.e., 2-3 points higher than Europe (excluding Russia).
Bairoch 1998, Chapter 21 is about the link between urbanization and economic development during the
Industrial Revolution.
         According to Bairoch, table 27.1, p.428, urbanization in the Third World was 8.5% of total
population in 1800, 9.1% in 1900, and 15.7% in 1950. In comparison, in Europe (without Russia) it was
12.1% in 1800 (down slightly from 12.3% in 1700), 37.9% in 1900, and 50.7% in 1950 (Table 13.2,
p.216.)
         Bairoch p.460-461: urbanization and growth in industrial employment proceeded together until
at least 1900. On China, p.476, “Progress in agriculture stimulated urban growth, and declines in
agriculture prompted declines, if not in the cities themselves, at least in the level of urbanization.”
         Urbanization in North America occurred relatively late and quite quickly. Chandler (1988) has
the most detailed information through 1861 and also makes a comparison with 1775 for almost all
countries in the Americas. In 1775 there was a total population of about 4m in what became the United
States. The total urban population in cities above 10,000 inhabitants was 89,000 (34,000 in Philadelphia,
24,000 in New York, 19,000 in Boston and 12,000 in Charleston), implying an urbanization rate of 2.2%
(Chandler 1987, p.52.)20
         In 1800, according to Chandler, the urban population was 192,000 out of a total population of
6m, implying urbanization of 3.2%.21 A gap was already opening up between the north and the south of
the US. In the North (defined by the states that stayed with the union in the civil war), urbanization was
7.1% while urbanization in the South was around 0.8% by Chandler’s figures.22
         In 1850, total urbanization, using a threshold of 18,000 inhabitants from Chandler, was 19.8%.
In 1861, urbanization in the North was up to 31.5% and while in the South it was no more than 5%.23 At
the same time, urbanization in Canada was 12.9%.
         In contrast, in Central and South America, urbanization was relatively high in the mid-1700s but
in most countries it either did not increase as rapidly as in the US or actually fell in the next 100 years.
In 1775, Chandler (1987, p.52) reports that the urban population of Mexico was 351,000, implying an
urbanization rate of 7%. In 1800, urbanization was still 7% and in 1850 it was only 7.5%. By 1861,
urbanization was no more than 8.6%. In Peru, Chandler (p.52) reports an urban population in1775 of
165,000, implying an urbanization rate of 11%. By 1800, however, his numbers imply that urbanization
was down to 9.3% and by 1850 it was down to 6.6%. In 1861, urbanization in Peru was no more than
8.7%. In Brazil, urbanization in 1775 was 7.3%, while in 1861 it was 7.4%. In Venezuela, urbanization
was 5.2% in 1775 but only 7.1% in 1861. Urbanization in Guatemala and Haiti seems to have stayed
around 2% throughout this period.
         Urbanization fell in the former Inca empire. Urbanization in Bolivia was 9.1% in 1775 but no
more than 7.7% in 1861. In Ecuador, urbanization was 15.8% in 1775, but no more than 10.4% in 1861.
In Colombia, urbanization was 5.2% in 1775 but only 4.1% in 1861. In Cuba, urbanization in 1775 was
24.6%; in 1861 it was 23.7%. In Jamaica, urbanization was 9% in 1800 but only 8.3% in 1850. There
was no appreciable increase in urbanization over the same time period in Martinique or Barbados.
         The notable exceptions are in the southern cone, where urbanization was relatively low in the
mid-eighteenth century but then grew rapidly. In Chile, urbanization was 2.5% in 1775, but up to 11.1%


20
   In 1700 Boston had a population of 6-7,000 (Bairoch 1988, p.296). In 1750 Boston was the only city
in North America that had a population of 20,000 (Chandler 1987).
21
   Even with his lower threshold for urbanization (5,000 vs. 18,000 in this case for Chandler), Bairoch
estimates that urbanization was just 5% (Bairoch 1988, p.221).
22
   For 1820, Bairoch (1988, p.297) reports that 6% of the US population lived in towns with at least
5,000 inhabitants. By Bairoch’s figures, in 1850 urbanization was up to 13-14%. Bairoch emphasizes
that urbanization grew particularly rapidly between 1820 and 1870.
23
   These North vs. South calculations use the population numbers of 1850.


                                                     xi
in 1861. In Argentina, urbanization was 8% in 1775 and up to 12.5% in 1861. In 1861, urbanization in
Paraguay was 9.6% and 51% in Uruguay.
        Coatsworth, not using urbanization estimates, comes to the same conclusion that we do about
when Latin America fell behind the USA.


4. Urbanization in 1000 AD
         The most comprehensive cross-continent data on urbanization in 1000 AD is from Chandler
(1987). Eggimann’s (1999) data starts only in 1500; de Vries also starts only in 1500. Bairoch provides
useful some qualitative assessments and quantitative estimates from 1300. Bairoch, Batou and Chevre
(BBC, 1988) provide detailed European urban population estimates by city from 800 AD, although they
put more reliance in these numbers from 1200/1300. As we want to compare urbanization in 1000 with
urbanization in 1500 (on the Bairoch basis), we use the BBC numbers where they are available. The
BBC estimates are consistently higher than those of Chandler. We convert urban population numbers
into urbanization using total population from McEvedy and Jones (1978).
         The other Chandler estimates we convert into BBC equivalents using coefficients from a
regression of the 11 overlapping data points for 1000 AD (all European). The coefficient on the
Chandler estimate is 1.13 (significant at the 1% level) and the constant is 1.03. Note that given the result
of this regression, our adjustment is essentially the same as applying Davis-Zipf’s law for low
urbanization countries (although a smaller adjustment for high urbanization countries).
         In this part of the appendix we focus on urbanization in 1000. However, to assess whether
urbanization in that year was at all unusual, as well as to give a sense of persistence of urbanization over
shorter periods, we also report urbanization for 800, 1200, 1300, and 1400 (if we do not report a
particular year, this indicates that the data is missing.) We should emphasize that reporting these
numbers (and using one decimal place) does not imply that we necessarily think these estimates are
accurate to within a few percentage points. Rather we argue that the broad pattern of urbanization is
clear from these estimates and we can discern (in conjunction with qualitative evidence) to what extent
urbanization has changed over periods of time such as 500 years.

Europe in 1000
         According to Chandler, France had an urban population of 95,000 in 1000, when total
population was 5 million (McEvedy and Jones), implying an urbanization rate of 1.5%. Before and after
this date urbanization was fairly constant, although showing a slow and steady rise towards 1500. This
fits with what we know about the gradual spread of prosperity in North-West Europe at this time.
According to Chandler, the French urbanization rate in 1.8% in 800, 1.5% in 1000, 4% in 1200, 4.2% in
1300 and 6.1% in 1400. BBC put total urban population in 1000 at 255,000, implying urbanization of
5%. We use the BBC estimate.
         The total Italian urban population was 362,000 in 1000 (Chandler), implying urbanization of
7.2% (McEvedy and Jones give total population as 5 million). Italian urbanization was 5.7% in 800,
10.1% in 1200, 8.2% in 1300, and 10.9% in 1400. BBC put total urban population in 1000 at 623,000,
implying an urbanization rate of 12%. We use the BBC estimate.
         In Spain, urbanization peaked at 19.2% in 1000 using Chandler’s data, primarily due to the large
(but short-lived) size of Cordoba. Total urban population was 766,000, of which 450,000 are estimated
to have lived in Cordoba (total population was 4 million, from McEvedy and Jones). Urbanization was
7.9% in 800, 8.8% in 1000, 7.2% in 1300 and 7.7% in 1400. BBC estimate total urban population in
1000 was 970,000, which gives urbanization of 24%. We use the BBC estimate.
         In Portugal, the urban population (just Lisbon according to Chandler) was 15,000 in 1000. With
total population estimated at 600,000, this implies urbanization was 2.5%. We have no data on
urbanization in 800. Urbanization was 4.4% in 1200, 2.8% in 1300, and 7.8% in 1400. BBC estimate
total urban population in 1000 was 15,000, giving the same estimate as Chandler for urbanization. We
use the BBC estimate.




                                                    xii
         Chandler has no data on urbanization for the Netherlands before 1300. In 1300 urbanization was
2.5% and in 1400 it was 5.25%. BBC do not have any data for the urban population in 1000; for 1200
they estimate the urban population was 12,000. Our estimate is explained in the next paragraph.
         For Belgium, the first Chandler data is for 1200, showing urbanization of 7.2%. In 1300
urbanization was 9.8%, and in 1400 it was 29.5%, with 70,000 people in Ghent and 60,000 people in
Bruges (most of the prosperity was based on the production and export of woolen cloth). BBC put the
urban population in 1000 at 34,000, implying urbanization of 9% (McEvedy and Jones, p.63, put total
population at 400,000). Most of this urban population was on the border with the Netherlands, which
had a total population of 300,000. Therefore if we apply this urban population to Belgium plus the
Netherlands, it would give an average urbanization rate of 4.9%. We use this average BBC estimate for
both countries.
         Chandler’s work implied that German urbanization was 2.8% in 1000 (111,000 urban population
and total population, from McEvedy and Jones, of 4 million.) Urbanization was 0.6% in 800, 5.3% in
1200, 4% in 1300, and 4.3% in 1400. BBC put total urban population at 209,000, implying urbanization
of 5.2%. We use the BBC estimate.
         For Austria, Chandler provides no data before 1500. BBC also provides no data. We use an
estimate of zero urbanization, which is probably too low.
         BBC (who were based in Geneva) estimate the urban population of Switzerland in 1000 AD as
just 1000 people. McEvedy and Jones, p.87, put the total population at 300,000 at this date, implying
urbanization of 0.3%. We use this estimate. Chandler does not have any relevant data.
         For Britain, including Ireland, the total urban population according to Chandler was 15,000.
Using the population for the British Isles minus Ireland (2 million, from McEvedy and Jones p.49), gives
urbanization of 0.8%. Urbanization was 1.7% in 1200, 1.9% in 1300, and 1.4% in 1400. BBC estimate
the total urban population for the British Isles minus Ireland was 97,000 (total population for this area
was 1.7 million), implying urbanization of 5.7%.
         BBC estimate the urban population of Ireland was 19,000 in 1000 AD. McEvedy and Jones put
total population in 1000 AD at 300,000, implying urbanization of 6.3%. We use this estimate. Chandler
does not give any data for Ireland at this time.
         For Bulgaria, Chandler estimates that total urban population was 80,000 in 1000. With total
population of 800,000, urbanization was 10%. Urbanization is estimated to have been 6.8% in 800,
5.5% in 1200, 5.5% in 1300 and 1400 in 4.2%. BBC estimate the urban population was 70,000,
implying urbanization of 8.8%.
         For Greece, Chandler estimates total urban population was 40,000 in 1000 AD. With total
population of 1 million, this implies urbanization of 4%. Urbanization was 6.9% in 800, 2.5% in 1200,
6% in 1300 and 7.7% in 1400. BBC estimate the urban population was 55,000, implying urbanization of
5.5%.
         For Denmark, Chandler estimates the urban population was 15,000. With total population
estimated at 400,000 (McEvedy and Jones), the implied urbanization estimate is therefore 3.8%.
Urbanization was 3.8% in 1200, 4.4% in 1300 and 5.2% in 1400. BBC do not have data for Denmark in
1000. We use the BBC-equivalent Chandler estimate, which is 5.3%.
         We use the population for Czechoslovakia, which was 1.25 million according to McEvedy and
Jones, p.85 (this may give lower urbanization numbers than is reasonable, because Slovakia had lower
urbanization and was not so closely linked to the Czech Republic at this time.) Chandler’s estimate of
urban population for 1000 AD is 15,000 (all in Prague). With total population of 1.25m, this implies
urbanization of 1.2%. Urbanization in 1200 was 1.1%, urbanization in 1300 was 1.3% and urbanization
in 1400 was 3.8%. BBC report urban population for Czechoslovakia in 1000 AD of 10,000, giving
urbanization of 0.8%. We use the BBC estimate for the Czech Republic and Slovakia.
         Chandler estimates total urban population in Hungary as 15,000 in 1000 AD. McEvedy and
Jones put total population at 500,000, implying urbanization of 3%.
Urbanization was 4.1% in 1200 and 1.2% in 1300. We do not have data for 1400. BBC put the total
urban population at 2,000 in 1000AD, implying urbanization of 0.4%. We use the BBC estimate.




                                                  xiii
        In Poland, Chandler estimates the total urban population was 30,000 in 1000 AD, implying
urbanization of 2.4% (McEvedy and Jones estimate the total population was 1.25m). In 1200
urbanization was 1.3%, for 1300 we do not have data and in 1400 it was 0.7%. BBC have no data on
urbanization in Poland in 1000. We use the BBC-equivalent Chandler estimate, which is 3.74%.
        In the area now comprising Russia, Ukraine and Moldova, Chandler estimates the urban
population was 88,000 (of which 45,000 in Kiev). We deflate this using the population of Russia-in-
Europe (McEvedy and Jones, p.79), which was 4 million, giving a total urbanization level of 2.2%.
Urbanization was 0.6% in 800, 1.4% in 1200, 3% in 1300 and 2.8% in 1400. BBC put urban population
in 1000 AD for this area at 152,000, implying urbanization of 3.8%. We use this for Russia, Ukraine
and Moldova.
        Chandler does not have urbanization estimates for Switzerland, Estonia, Lithuania, Latvia,
Romania, Serbia, Macedonia, other parts of Yugoslavia, Albania, Finland, Norway or Sweden in 1000
AD.
        BBC does not have estimates for these countries, with one exception. BBC estimates urban
population in the former Yugoslavia was 55,000. McEvedy and Jones, p.113, estimate that the total
population living in this area was 1.25m, which implies that urbanization was 4.4%. We use the BBC
estimate for all countries that emerged from the former Yugoslavia.
        Where both Chandler and BBC do not give data, and there is no nearby urban population in
neighboring countries, we assume urbanization was zero. The exception is the Baltics, which may or
may not reasonably be considered part of Russia in 1000 AD. As these are small countries, we prefer to
leave their data as missing.
        We use an estimate of zero urbanization for Finland, Norway, Sweden, and Romania. This
seems quite consistent with the evidence from a variety of sources, for example de Vries.

Asia in 1000
         Chandler’s estimates of urbanization in Asia are almost certainly underestimates because he uses
a minimum threshold of 40,000 for 1000 AD. We use the same conversion factor (from the regression
of BBC on Chandler) for our base estimates. All our estimates for Asia are BBC-equivalent urbanization
rates formed by thus converting Chandler’s data.
         For the area now occupied by Turkey in 1000 (the empire of Byzantium), Chandler estimates
total urban population was 455,000. From McEvedy and Jones, using data for Turkey-in-Asia and
Turkey-in-Europe, we have total population of 7.3 million. This implies an urbanization rate of 6.2%.
Urbanization from Chandler was 4.6% in 800, 4.5% in 1200, 3% in 1300, and 4.8% in 1400. We use a
BBC-equivalent estimate of 8%.
         For Iraq, Chandler estimates the total urban population of Iraq to have been 175,000 in 1000
AD. McEvedy and Jones estimate total population to have been 2m, implying urbanization of 8.8%.
Urbanization was 36.8% in 800 (when Baghdad was the capital of a large empire), 6.7% in 1200, 4% in
1300 and 9% in 1400. We use a BBC-equivalent estimate of 11%.
         For Iran, urban population, according to Chandler, in 1000 AD was 427,000. Total population,
according to McEvedy and Jones, was 4.5m, implying urbanization of 9.5%. Urbanization in 800 was
4.5%, 5.8% in 1200, 8.6% in 1300, and 9.4% in 1400. We use a BBC-equivalent estimate of 11.8%.
         Chandler estimates that the urban population of Saudi Arabia in 1000 was 110,000. Given
McEvedy and Jones’ population estimate of 5m, this implies urbanization of 2.4%. Urbanization was
1.6% in 800 and we do not have any other information before 1600. We use a BBC-equivalent estimate
of 3.7%.
         To assess urbanization in Central Asia we use the estimate of population for Russian Turkestan,
from McEvedy and Jones, p.163 – 2.5 million. Chandler estimates the urban population of this area was
270,000 (Samarkand, Bokhara and Balasaghun), implying urbanization of 7.6%. Urbanization was 7.3%
in 800, 5.9% in 1200, and 5.2% in 1400. Eggimann does not have data for Central Asia in 1500, so we
use Chandler’s data (urban population of 110,000 in a total population of 3.5 million), which implies
3.1% urbanization. We use a BBC-equivalent estimate of 9.6% for both Kazakhstan and Uzbekistan in
1000 and 4.5% in 1500.



                                                   xiv
         The urban population of Syria, Lebanon and Palestine was 140,000 in 1000 AD, according to
Chandler. The total population of this same area (McEvedy and Jones, p.139 and p.143) was 2m,
implying urbanization of 7%. Urbanization was 4.5% in 800, 7.8% in 1200, 2.8% in 1300 and 7.1% in
1400. We use a BBC-equivalent estimate of 8.9%.
         The urban population of Pakistan, India and Bangladesh was 627,000 in 1000 AD, while its total
population was 77 million (McEvedy and Jones, p.185), giving urbanization of 0.8%. Urbanization was
0.8% in 800, 0.8% in 1200, 0.8% in 1300, and 1.2% in 1400. These estimates are almost certainly much
too low. For example, they imply that urbanization in 1500 was only 1.8%, whereas the evidence
suggests it was much higher – probably at least 8%. Chandler’s method seems to be particularly biased
for larger countries (see also the problems with China below). We use his numbers here, however, as we
are primarily interested in the change between 1000 AD and 1500 AD using the same methodology.
         To assess urbanization in Indonesia and Malaysia we use the estimate of the total population of
the Malay Archipelago, from McEvedy and Jones, p.199. Chandler provides no data for 1000 AD. For
800 AD, he estimates total urban population was 60,000. With total population of 3.5 million, this
implied urbanization of 1.7%.
In 1300, urbanization was 0.7% and in 1400 it was 1.1%. We use the 800 AD estimate and convert this
to a BBC equivalent number of 3%.
         Chandler estimates the urban population of Cambodia was 14,000 in 1000 AD. McEvedy and
Jones estimate total population was 1 million in the same year, implying urbanization of 14%.
Urbanization was 5% in 800 AD, 13.6% in 1200, 7.5% in 1300 and 3.1% in 1300. We use a BBC-
equivalent estimate of 16.9.
         The area occupied by “China Proper” (excluding Turkestan and Tibet) contained 60 million
people in 1000 AD (McEvedy and Jones, p.171). Chandler estimates the total urban population in this
area was 1.504 million, implying urbanization of 2.5%. Measured in this way, urbanization was 2.9% in
800, 1.4% in 1200, 2.4% in 1300, and 2.8% in 1400. We use a BBC-equivalent urbanization estimate of
3.9%.
         Note that this is almost certainly an underestimate, as in the case of India. Given that the urban
population of China in 1500 is underestimated for the same reason (both Eggimann and Chandler count
only large cities), this should still allow us to check for persistence, i.e., if urbanization either fell or rose
dramatically, it would probably be picked up in these numbers. If anything, the use of a relatively high
threshold – like 40,000 inhabitants – should exaggerate the fluctuations in urban population. We should
exercise care, however, when comparing large countries such as China and India with smaller countries.
         Chandler estimates that the urban population of Korea was 110,000 in 1000 AD. McEvedy and
Jones put total population on that date at 2.5 million, implying urbanization of 4.4%. Urbanization was
2% in 800 and 2.9% in 1400. We use a BBC-equivalent urbanization estimate of 6%.
         The urban population of Japan in 1000 AD was, according to Chandler, 175,000. Total
population from McEvedy and Jones, was 4.5 million, implying urbanization of 3.9%. Urbanization was
5% in 800, 3.7% in 1200, 2.5% in 1300, and 1.5% in 1400. Japan is one the few countries in which
urbanization in 1500 (0.2%) was much lower than in 1000 AD – the size of this discrepancy was
temporary and largely the result of civil war. We use a BBC-equivalent estimate of 5.4%.
         The urban population of Burma/Myanmar in 1000 AD was 40,000 (Chandler). The total
population was 2 million (McEvedy and Jones), implying urbanization of 2%. Urbanization was 7.3% in
1200 and 3.1% in 1400. We use a BBC-equivalent estimate of 3.3%. Note that while we have
urbanization data on Myanmar in 1000 and 1500 but not data on GDP per capita today. This country
therefore appears in graphs and regressions of urbanization in 1500 on urbanization in 1000 but not in
our base sample.
         Chandler estimates the urban population of Afghanistan was 60,000, while McEvedy and Jones
put total population at 2.25 million in 1000 AD. This puts the “pure Chandler” urbanization estimate at
2.7% and we use a BBC-equivalent estimate of 4.1%. No other information on urbanization before 1500
is available.




                                                       xv
        Urbanization in Sri Lanka was 7.5% in 1200 – urban population of 75,000 from Chandler and
total population of 1 million from McEvedy and Jones. No other data on urbanization before 1500 is
available. We treat this as missing data for 1000 AD.
        Urbanization in Georgia was 8% in 1200 (urban population of 80,000 and total population of 1
million). No other data is available on urbanization before 1500. We treat this as missing data for 1000
AD.
        We have no data on urbanization in Vietnam until 1300. At that date, the urban population,
according to Chandler was 40,000. McEvedy and Jones’ total population estimate of 1 million implies
an urbanization estimate of 4%. We treat this as missing data for 1000 AD.
        We also have no data on urbanization in Thailand until 1300. At that date, the urban population
(Chandler) was 40,000 and the total population was 1.5 million, implying urbanization of 2.7%.
Urbanization was 2.3% in 1400. We treat this as missing data for 1000 AD.
        We have no data on urbanization in Nepal or the Philippines on or around 1000 AD. We have no
information about urbanization in Laos in or around 1000 AD. We treat this as missing data for 1000
AD.
        We assume urbanization in Hong Kong and Singapore was 3% in 1000, following the same
reasoning as in our previous appendix on 1500.

America in 1000
         For Mexico, Chandler reports an urban population in 1000 AD of 170,000. McEvedy and Jones
put total population at 3 million, implying urbanization of 5.7%. Urbanization was 6% in 800, 3.1% in
1200, 2.8% in 1300 and 4.9% in 1400. We convert Chandler’s estimate to a BBC-equivalent estimate of
7.5% and use this.
         For Peru, Chandler’s urban population estimate in 1000 AD is 40,000. McEvedy and Jones put
total population at 1.5 million, implying urbanization of 2.7%. Calculated in this way, urbanization was
1.6% in 800, 2.3% in 1200, 3.45 in 1300 and 5.3% in 1400. We use a BBC-equivalent estimate of 4.1%.
         Chandler estimates that the urban population of Bolivia was 20,000 in 800 and 40,000 in 1200,
but he does not give an estimate for 1000 AD. McEvedy and Jones put population at 500,000 in 800 and
800,000 in 1200, implying urbanization of 4% and 5% respectively. An estimate of 3% Chandler-
equivalent urbanization in 1000 therefore seems reasonable. We use a BBC-equivalent estimate of
4.4%.
         The earliest urbanization data we have for Ecuador is 1200, for which date Chandler estimates
an urban population of 40,000. McEvedy and Jones suggest that total population was half a million,
which would imply that urbanization was 8%. We use this estimate also for 1000 AD, converting it to a
BBC-equivalent estimate of 10.1%. Urbanization was 16% (using Chandler’s measure) in 1300, and
18% in 1400.
         According to Chandler, the area now consisting of central America (particularly Guatemala) had
an urban population of 100,000 in 800 (at the height of the Mayan civilization) and 45,000 in 1200.
McEvedy and Jones put the population of Central America at 500,000 and 600,000 respectively at those
dates, implying urbanization of 20% at the earlier date and 7.5% at the later date. Unfortunately, we do
not have data on urbanization in 1000. To be on the safe side, given that we know the Mayan
civilization collapsed suddenly before 1000, we urbanization from 1200. We convert this to a BBC-
equivalent estimate of 9.5.
         Chandler does not report urbanization data for Argentina, Brazil, Chile, Canada, Paraguay,
Uruguay or the United States before 1500. For all these countries, it is safe to assume that urbanization
was essentially zero in 1000. Chandler also does not report data for any part of the Caribbean,
Colombia, Guyana, or Venezuela. We have reason to think that urbanization may have been above zero
in some of these cases, but we assume zero urbanization for the sake of consistency (and because it goes
against our hypothesis.)




                                                   xvi
Africa in 1000
         As is the case for 1500, the data on North Africa is reasonable and probably not subject to
greater error than other non-European data. We have information on urbanization in 1000 for all North
African countries, except Libya.
         For Egypt in 1000, Chandler reports an urban population of 285,000. McEvedy and Jones
estimate the total population was 5 million at this date, implying urbanization of 5.7%. Urbanization,
calculated from these sources, was 3.9% in 800, 10.6% in 1200, 13.5% in 1300, and 14.5% in 1400. We
use a BBC-equivalent estimate of 7.5%.
         For Tunisia, Chandler puts the urban population in 1000 at 60,000 and McEvedy and Jones
estimate total population was 1 million. We convert this estimate of urbanization, 6%, into a BBC-
equivalent estimate of 7.8%. Urbanization, in Chandler units, was 12.5% in 800, 6% in 1200, 7% in
1300 and 8.8% in 1400.
         Chandler estimates the urban population of Algeria was 20,000 in 1000 AD. McEvedy and
Jones put total population at 2 million, implying urbanization of 1%. Urbanization was 1.6% in 800, 2%
in 1200, 7% in 1300 and 9.1% in 1400. We use a BBC-equivalent estimate of 2.2%.
         For Morocco, McEvedy and Jones put total population 1000 at 2 million and Chandler puts the
urban population at 100,000, implying urbanization of 5%. The same combination of sources gives
urbanization of 3.25 in 800, 23% in 1200, 17.3% in 1300, and 14.8% in 1400. We use a BBC-equivalent
estimate of 6.7%.
         As previously discussed, the data on urbanization in sub-Saharan Africa is fragmentary, of
dubious value and almost certainly underestimates with biases that are hard to assess. The data on 1000
AD is worse than for 1500. However, there is some relevant information on the persistence of
urbanization (assuming similar biases in estimates for 1000 and 1500). All the information reported here
comes from combining Chandler and McEvedy and Jones. As we do not calculate urbanization in 1500,
we do not convert into BBC-equivalent measures (and this information on sub-Saharan Africa is not
used in our graphs or regressions.)
         For the Sudan, our sources imply urbanization of 0.8% in 800, 1.8% in 1000, 1.7% in 1200,
1.4% in 1300, 0.75 in 1400, and 0.5% in 1500. For Ethiopia, we have no data on urbanization in 1000
AD, but estimates of 2.5% for 800, 1.3% for 1200, 1.3% for 1300, 1.7% for 1400 and 2.7% for 1500.
For Mali, urbanization was 4% in 800, 2% in 1000, 2.6% in 1200, 6.6% in 1300, 6.2% in 1400, and
7.5% in 1500. For Zimbabwe, urbanization is estimated to have been 5% in 800, 4% in 1000, 3.3% in
1200, 3.6% in 1300, 3.1% in 1400 and 2% in 1500. For coastal West Africa (using the population
estimates for West Africa, from McEvedy and Jones, p.243), with most of the urbanization in and around
Nigeria, we have urbanization in 800 and 1000 of 0.4%, 1.1% in 1200, 1.3% in 1300, 1.8% in 1400 and
2.65 in 1500.
         We have no other urbanization data for African countries on or around 1000 AD.

The Pacific Region in 1000
        We assume zero urbanization in 1000 for Australia, New Zealand. This does not seem
controversial. There is no sign of significant urban structures in Australia before 1800. In New Zealand,
there may have been some very small scale urbanization, but even if we increased the urbanization
estimate to 3% (as for 1500), this would not change any of our results.

Missing Data for 1000
         Given the state of knowledge about urbanization, we have less data on 1000 AD than we do on
1500 AD. In particular, reliable data about the Caribbean, Papua New Guinea and Philippines are
missing. Sri Lanka, Vietnam, Laos, Libya are also missing data. However, for all of these countries,
there is no reason for believe there was a big change in urbanization rates between 1000 and 1500.




                                                  xvii
                                             Appendix Figure A1
                             Log GNP per capita       Urbanization                                            Log GNP per capita        Urbanization
                      7                                                     50                         7                                                      60




                                                                            40
Log GNP per capita




                                                                                 Log GNP per capita
                     6.5                                                                               6                                                      40




                                                                                    Urbanization




                                                                                                                                                                   Urbanization
                                                                            30

                                                                                                       5
                      6                                                                                                                                       20
                                                                            20



                                                                                                       4
                                                                            10
                     5.5                                                                                                                                      0
                           1800           1850             1900      1950                                   1200       1400         1600        1800   2000
                                                   date                                                                             date
                                              Germany                                                                              Britain
                             Log GNP per capita       Urbanization                                            Log GNP per capita        Urbanization

                                                                            40                        6.5                                                     25
                     6.5



                                                                                                                                                              20
Log GNP per capita




                                                                                 Log GNP per capita
                      6                                                     30      Urbanization       6




                                                                                                                                                                   Urbanization
                                                                                                                                                              15


                     5.5                                                    20                        5.5
                                                                                                                                                              10




                      5                                                     10                         5                                                      5
                           1750       1800         1850       1900   1950                                   1800           1850              1900      1950
                                                   date                                                                             date
                                                  France                                                                           Sweden
                                             Appendix Figure A2
                             Log GNP per capita        Urbanization                                             Log GNP per capita        Urbanization
                     7.5                                                                                5.8                                                       14
                                                                              40



                      7                                                                                 5.6                                                       12
Log GNP per capita




                                                                                   Log GNP per capita
                                                                              30




                                                                                      Urbanization




                                                                                                                                                                       Urbanization
                     6.5                                                      20                        5.4                                                       10




                      6                                                       10                        5.2                                                       8




                     5.5                                                      0                          5                                                        6
                           1700                   1800                1900                                    1750       1800          1850         1900   1950
                                                    date                                                                               date
                                                   US                                                                                Russia
                             Log GNP per capita        Urbanization                                             Log GNP per capita        Urbanization

                     5.6                                                      18                         7
                                                                                                                                                                  40
Log GNP per capita




                                                                                   Log GNP per capita
                     5.4                                                                                6.5                                                       30
                                                                                      Urbanization




                                                                                                                                                                       Urbanization
                                                                              16


                     5.2                                                                                 6                                                        20




                                                                                                                                                                  10
                      5                                                       14                        5.5
                           1750       1800         1850        1900    1950                                   1800           1850                1900      1950
                                                   date                                                                                date
                                                  Japan                                                                              Australia
                                                   Appendix Figure A3
                             Log GNP per capita       Log Population Density                                            Log GNP per capita        Log Population Density
                      7                                                          5                               7                                                           6




                                                                                       Log Population Density




                                                                                                                                                                                   Log Population Density
Log GNP per capita




                                                                                        Log GNP per capita
                                                                                                                 6                                                           5
                     6.5

                                                                                 4.5


                                                                                                                 5                                                           4
                      6




                                                                                 4                               4                                                           3
                     5.5
                           1800           1850             1900           1950                                        1200       1400         1600        1800        2000
                                                   date                                                                                       date
                                              Germany                                                                                        Britain
                             Log GNP per capita       Log Population Density                                            Log GNP per capita        Log Population Density

                     6.5                                                         4.3                            6.5                                                          6




                                                                                       Log Population Density




                                                                                                                                                                                   Log Population Density
Log GNP per capita




                                                                                        Log GNP per capita
                                                                                 4.2
                      6                                                                                          6                                                           5.5


                                                                                 4.1

                                                                                                                                                                             5
                     5.5                                                                                        5.5
                                                                                 4



                                                                                                                                                                             4.5
                      5                                                          3.9                             5
                           1750       1800         1850       1900        1950                                        1800           1850              1900           1950
                                                   date                                                                                       date
                                                  France                                                                                     Sweden
                                             Appendix Figure A4
                             Log GNP per capita        Log Population Density                                            Log GNP per capita         Log Population Density
                     7.5                                                                                         5.8
                                                                                                                                                                               3
                                                                                  2




                                                                                        Log Population Density




                                                                                                                                                                                     Log Population Density
                      7                                                                                          5.6
Log GNP per capita




                                                                                         Log GNP per capita
                                                                                                                                                                               2.5


                     6.5                                                                                         5.4
                                                                                  0


                                                                                                                                                                               2
                      6                                                                                          5.2




                     5.5                                                          -2                              5                                                            1.5
                           1700                   1800                   1900                                          1750       1800          1850         1900       1950
                                                    date                                                                                        date
                                                   US                                                                                         Russia
                             Log GNP per capita        Log Population Density                                            Log GNP per capita         Log Population Density

                     5.6                                                                                          7                                                            0
                                                                                  6.6




                                                                                        Log Population Density




                                                                                                                                                                                     Log Population Density
                                                                                                                                                                               -1
Log GNP per capita




                                                                                         Log GNP per capita
                     5.4                                                          6.4                            6.5


                                                                                                                                                                               -2


                     5.2                                                          6.2                             6
                                                                                                                                                                               -3




                      5                                                           6                              5.5                                                           -4
                           1750       1800         1850        1900        1950                                        1800           1850                1900          1950
                                                   date                                                                                          date
                                                  Japan                                                                                       Australia