THE COMING OF URBAN PRIMACY IN LATIN
AMERICA*
Christopher Chase-Dunn
Department of Sociology
Johns Hopkins University
Baltimore, Maryland 21218
U.S.A.
*Thanks to Carin Celebuski for help with the analysis and interpretation. This is a version of a longer paper, "Latin American Cities in the World-system," originally prepared for the Seminar on "Structural development and social problems of the Latin American city" of the Working Group on Latin American Urbanization, Tepotzlan, August 22-24, 1982. This research was supported by the National Science Foundation (SES 7825071). Published in Comparative Urban Research XI, 1-2:14-31, 1985 and as "El fenomeno de primacia de una ciudad an los sistemas urbanos latinoamericanos: su surgimiento," Pp. 27-46 in J. E. Hardoy and A. Portes (eds.) Ciudades y Sistemas Urbanos: Economia Informal y Desorden Espacial. Buenos Aires: CLACSO 1985.
Scholars of
Latin American urbanization have long noted the high degree
of
concentration in national city systems. Explanations of this
characteristic
of Latin American societies are many and various, and the
comparative
studies which have been made do not allow for firm conclusions
as to the
causes of urban primacy in Latin America. This study assembles
new data to
investigate Latin American urban primacy over the period from
1800 to 1975.
It is the first such study to systematically compare long run
patterns of
urbanization in Latin America with the urban histories of more
"developed"
national societies, those which can be said to constitute the
core areas of
the modern world-system in the twentieth century.
Urban
primacy
The empirical
distribution of city populations in a territory is a
characteristic
of any human society which contains cities, towns and villages.
The causality
of the spatial distribution of population is highly complex
and the logic
of spatial arrangements differs across different kinds of
social
systems. Furthermore, the spatial distribution of population is
itself only a
formal and superficial characteristic which does not
systematically
reflect determinant social relations (Smith, 1982). The
same may be
said, however, for many other surface-level phenomena which
social
scientists study in order to infer the logic of social systems. This
realization
encourages us to search for better measures of the social
relations
which we think are more fundamentally involved in system dynamics.
That search,
however, should not preclude us from analyzing those less-than-
ideal types
of data which are available, as long as these investigations pay
adequate
attention to the problems in inference involved.
Ideally a city-size distribution should always be studied together
with data on
the economic transactions and political dominance relations
among cities.
In practice we can obtain population data much more readily
than we can
find data on commodity flows or power relations. The choice is
whether or
not to present an analysis which infers structural relations
from an
admittedly problematic indicator. The report below is such an
analysis and
it should only be presented as preliminary. Further data-
gathering and
analysis are a prerequisite to firm conclusions. 1
Theories
of urban primacy
General
theories which seek to explain variation in the steepness or
flatness of
city size hierarchies can be categorized as economic and
political.
Following Christaller (1967), social geographers have argued
that an
integrated space economy will produce a locational distribution of
activities
such that those which can be efficiently produced in a centralized
way (because
storage and transport costs are cheap) will be located together
in a central
city, whereas those which have high transport and storage costs
will be more
dispersed over territory. This theory ignores purely
geographic
factors which alter the spatial plane such as rivers, mountain
ranges,
bodies of water, (which alter transport costs) and resources whose
locations are
determined by “nature” such as mineral deposits and land
suitable for
agriculture. Ignoring natural inputs and geographical
irregularities,
an integrated space economy will produce a distribution of
population
similar to the lognormal rank-size rule. In a lognormal rank-
size
distribution the largest city is twice as large as the second largest,
three times
larger than the third largest, and so on. Some geographers
argue that
such a space economy only operates in a competitive market
system in
which the priceSof commodities reflect their costs of production,
but
archaeological evidence shows that city systems vary around the log-
normal rule
even in societies in which price-setting markets do not playa
dominant role
in social exchange (Kowalewski, 1982). Transport costs and
storage costs
affect the economy of non-market systems, although through
different
means, and less directly than in market systems.
Another theory of variation in the steepness of city-size
hierarchies
emphasizes
the distribution of political-military power in space. Centrally
organized
political empires are thought to create more hierarchical city
systems than
networks of independent city-states or nation-states because
concentrated
political-military power allows for the concentration of wealth,
and therefore
of population. Of course the distribution of power is not
always
organized territorially, so there is considerable looseness of fit
between power
concentration and spatial hierarchies. Nevertheless,
variation in
ancient city-size distributions does correspond with the rise J
and fall of
political empires.
The literature on Latin American city systems combines these two
general
theoretical approaches with more historically specific interpretations
of the
development of Latin American societies. Some variations of
dependency
theory, especially that explicated by Frank (1969) can be read to
suggest that
peripheral areas in the world-economy will develop primate
city systems
because the largest cities will act as funnels which channel
peripheral
raw materials produced in the hinterland to core nations.
Middle-size
cities producing manufactured goods will not emerge because
they will be
unable to compete with the cheaper manufactured goods from the
core
countries.
More particular elaborations of this proposition are many and
varied.
It has been
claimed that urban primacy in Latin America developed from the
political
power which Spanish colonial municipalities exercised over their
hinterlands
(Cardoso, 1975; Portes and Walton, 1976). It has also been
claimed that
the development of large scale export agriculture explains the
rise of Latin
American urban primacy (McGreevey, 1972; Roberts, 1979).
Carol Smith
(1980) has argued that the development of a free labor market in
the largest
cities, in combination with “mercantile capitalist” class power
in secondary
cities (which restricts access to employment), results in primacy
because
migrants from rural areas concentrate in the largest cities.
Hardoy and
Langdon (1978) interpret their findings as indicating that
primacy
results from international immigration into the largest cities in
combination
with a socially produced scarcity of available rural land for
settlement.
Immigrants are then forced to congregate in the largest cities.
Before
discussing the merits of these explanations let us consider the
methods and
evidence to be examined in this research report.
The
data
My research
project has gathered data on the population sizes of the
ten largest
cities in each of 131 countries at decennial time points between
1800 and
1975. Most earlier cross-national studies of urban primacy have
studied much
shorter and more recent time periods. The slow rate of change
of urban
systems means that variation over short time periods reflects
mostly
measurement error.
Two studies of Latin American cities have analyzed data in both
the
nineteenth
and twentieth centuries. The pioneering study by McGreevey
(1971)
focussed on eight Latin American countries over the period from 1750
to 1960. A
more recent study by Hardoy and Langdon (1978) examines twenty
Latin
American countries between 1850 and 1930. The data analyzed here is
more complete
in the sense that it contains more cities at more points in
time than the
previous studies, and it compares Latin American cities to
those in the
"developed” countries of the world-system. The Latin American
countries
studied here are: Mexico, Guatemala, Honduras, El Salvador,
Nicaragua,
Costa Rica, Panama, Colombia, Venezuela, Ecuador, Peru, Bolivia,
Chile,
Paraguay, Uruguay, Brazil, Argentina, Cuba and the Dominican
Republic.
Sources of the city populations of the ten largest cities in each
of
these
nineteen Latin American countries are chiefly the population censuses
of each
country, but additional sources of population estimates such as
gazetteers
were used when census data were not available. Thus we have enough
data to
calculate a primacy index (at least four cities) for thirteen of the
countries by
1820. By 1900 we have data for all countries except Panama,
which enters
our data set in 1910. Our raw data on individual cities is
interpolated
to decennial time points by a computer program which uses
three or four
data points, and which adjusts for a change from municipal
boundary to
urban agglomeration figures. Table 1 shows the interpolated data
for Argentina
in the years 1800, 1810 and 1820. Notice that a population
for Cordoba
cannot be estimated for 1800, and since it is probably the
second or
third largest city this could have a major affect on the calculated
measures of
urban primacy. Missing data is less of a problem as we move
forward in
time. The city populations marked by an asterisk have been
estimated by
the interpolation program.
Methods
and measurements
The larger project, from which this paper is a preliminary report,
seeks
to test
propositions about the causes of national urban primacy by
comparing
changes in countries over time (see Chase-Dunn, 1981). A multivariate
model of the causes of national urban primacy
will eventually be
tested
employing economic, political-military and demographic variable
characteristics
of nations, international transactions, and changing
features of
the world-system as a whole. At this point however, only the
data on the
populations of cities is ready to be explored, and thus the
following
will report patterns observed in variation on the main dependent
variable,
national urban primacy. I shall attempt to interpret these
patterns in
the light of theoretical arguments and historical studies of
Latin
American urbanization in the existing social science literature.
Several indicators of the concentration of populations in space
have
been used by
other researchers. Davis (1969) uses a ratio of the population
of the
largest city to the sum of the next three cities. Hardoy and Langdon
(1978) use
the Davis measure plus the ratio of the largest city to the total
population of
the nation. McGreevey (1971) develops an indicator which
measures the
degree to which a distribution of city populations differs from
the lognormal
rule. The lognormal rule is used here simply as a yardstick
for
comparison. A city size distribution is lognormal (a straight line when
plotted on
double log paper) when the second largest city is half the size
of the
largest, the third largest is one third the size of the largest, and
so on. A
primate city system is one in which the secondary cities are
smaller than
would be predicted by the lognormal rule, and on the other hand,
a flat or
unhierarchical city system is one in which the secondary cities
are larger
than expected according to the lognormal rule.
McGreevey's
measure has several advantages over earlier ones. It
allows for
the comparison of distributions with different numbers of cities
by correcting
for the number of cities considered.
Pamela Walters (1985) has developed an indicator of primacy
which is
based on McGreevey's, but with some improvements. It is called the
Standardized
Primacy Index (SPI). This indicator takes the value of zero (0)
when a
distribution is lognormal. It takes on positive values when a
distribution
is primate or steeper than lognormal, and negative values when
a
distribution is flatter than lognormal. McGreevey's indicator is not
signed, and
thus deviations either above or below the lognormal line are
computed
implicitly as indicating greater primacy. Walters' measure is
assigned a
sign (i.e. positive or negative) based on the sum of the
deviations
from the lognormal rule.2 Table 1 shows the observed city
populations,
the populations expected by the lognormal rule, and the
deviations
for the case of Argentina in 1800, 1810 and 1820. The main
summary
statistic analyzed in this research report is shown as the SP12,
which is a
Standardized Primacy Index calculated on all the cities for which
data are
available at a particular time point.
Findings
This research
project has plotted the decennial SPIs for the nineteen
Latin
American countries under study, and also those of seventeen countries
now
considered to be in the core of the modern world-system. First let us
examine the
results for the Latin American countries.
Figure 1
shows the SPI of Argentina plotted from 1800 to 1975. When
we focus on a
single country, such as earlier studies have done, we can try
to interpret
changes in the city-size distribution by noting the causes
of divergent
rates of growth of the cities. Vapnarsky's (1975) study of
Argentina
concluded that the development of “external orientation,” the
growth of
export agriculture, had caused the Argentine city system to
become more
primate since the early nineteenth century. Hardoy and Langdon
(1978) pay
more attention to the timing of increased primacy, noting that
European
immigration concentrating in Buenos Aires accounts for the rise in
primacy in
the last half of the nineteenth century.
The focus on single countries and the relative growth of
particular
cities is a
fascinating undertaking. Attempting an historical account
which
explains the varying fortunes of cities involves immersion in the
richness of
political and economic history of each country. Brief but
insightful
accounts of this kind are given for eight Latin American
countries in
the volume edited by Morse (1971). Arregui (1981) has under-
taken an
investigation of the linkages between the development of Peru's
city system
and its connections with the larger world-system. Now, however,
let us
examine the patterns revealed by looking at all nineteen Latin
American
countries. We shall use the telescope rather than the magnifying
glass.
Table 2 shows how the SPIs of the Latin American countries changed
between 1800
and 1975. Over the whole period for the time points 1800, 1850,
1900, 1950
and 1975 the mean SPI was 3.4, the standard deviation was 8.3
and the SPI
varied from a maximum of 16.3 to a minimum of -23.6. Remember
that this SPI
is computed on the ten largest cities in each country
(although
earlier time points may not have data on all cities --an SPI is
not computed
if there are less than four) and a score of zero means that a
city size
distribution is lognormal. The mean SPI of 3.4 indicates that
Latin
American countries do tend to have primate or steep city size
distributions.
McGreevey's (1971) earlier study of eight Latin American countries
concluded
that, though the time of the shift toward primacy varied, all
Latin
American countries, with the exception of Colombia, have become primate
by 1950.
Table 2 confirms his finding when additional countries are included.
Four countries
(Argentina, Cuba, Ecuador (1820) and Colombia) were primate
already in
1800. While there is some moving down toward lognormality, or
even flatness
on the part of some countries, the overall trend is for all
countries to
move up, but at different periods. Table 2 shows that the
main period
in which the average level of the SPI shifts toward primacy is
between 1900
and 1950. In 1800, 1850 and 1900 the average SPI is very close
to lognormal
for the Latin American countries. Between 1900 the mean SPI
jumps from
-.2 to 7.7, and in 1975 it goes even higher to 8.9.
Let us look
more closely at the period between 1900 and 1970. Table 3
shows the
decennial means and standard deviations of the SPIs for the nine-
teen Latin
American countries, and it also shows the ten year differences
in the means,
as well as indicating the statistical significance of the ten
year
differences from zero. Table 3 shows that the mean SPI has risen in
every decade
of the twentieth century, and the standard deviations have
decreased, indicating
a convergence. The biggest shifts occurred in the
decades of
the twenties, the thirties and the forties, with the change
occurring
during the forties attaining statistical significance in its
differing
from zero.3
The
city systems of core countries
The
specification and testing of a causal model of urban primacy is
beyond the
scope of this present exercise. This would require the measurement
of variables
thought to account for primacy over the period being studied.
Such model
testing is the long run goal of our research project, but for now
we may simply
compare Latin American city systems with those found in
countries now
occupying core positions in the larger world-economy.
Table 4 shows
the distribution of SPIs for seventeen countries which
were in the
core by 1930.4 As with Table 2 (Latin American countries) the
times covered
are 1800, 1850, 1900, 1950 and 1975. The first striking thing
to note in
comparing Tables 2 and 4 is that the core countries do not
shift toward
primacy over time. The mean SPI is generally close to log-
normal. only
for 1950 does it climb higher than 2, and by 1975 five of the
seventeen
core countries have very flat city size distributions. Recall
that none of
the Latin American countries are flat, and only Colombia is
near
lognormal in 1975.
Table 5 shows the periods in which change occurred within the city
systems of
the core. The right side of the table shows those countries who
shift SPI
categories between the years designated. It can be seen that the
biggest concentrations
of changes are of two kinds. In the period between
1850 and 1900
six countries shifted toward steeper city size distributions,
and of these
all but France had yet to enter the core of the world-system.
From 1900 to
1950 five countries moved toward less hierarchical city
systems: the
USA, the United Kingdom, Canada, the Netherlands and Belgium.
The U.K.,
Netherlands and Belgium are old core countries. The U.S. entered
the core in
the 1870s or 1880s and began the rapid flattening of its city
system in
1930, while Canada (which entered the core in the 1920s) moved
toward a
flatter distribution after 1910, and again after 1940.
As with the Latin American countries, the core countries deserve
individual
attention to the timing and extent of changes in their city
systems, but
here we want to focus on the broad outlines of core patterns
in order to
compare them with Latin America.
Comparing
core and Latin American city systems
When we
compare Tables 2 and 4 we can observe that the core mean SPI
for all time
points is .7 while the Latin American mean for all time points
is more
primate, 3.4. A test for the statistical significance of the
difference
between these two means is significant at the .04 level. This
may be
interpreted as weak evidence in support of the hypothesis that
position in
the world-economy is related to urban primacy.
However, when
we look at the patterns over time and compare core and
Latin
American means at different time points a more refined picture emerges.
Between 1900 and 1950 both core and Latin American SPI means rise,
but the
increase is
very small for the core and very large for Latin America. And
between 1950
and 1975 the mean SPI for core countries decreases, indicating
the
flattening of city size distributions, while the mean for Latin American
countries
continues to rise substantially. Statistical tests for the
difference
between core and Latin American SPI means do not reach
significance
for 1800, 1850 or 1900. But for 1950 the difference is
significant
at the .003 level and for 1975 at the .0002 level.
Thus these
comparisons show that the divergence in urban primacy
between the
core and periphery emerges in the twentieth century, between
1900 and
1950, and continues to increase until 1975. Since the overall
core distribution
does not change much, except for a small trend toward
flattening in
the 1950-75 period, we may conclude that most of the change
is due to the
growth of primacy in the 1920s, 30s and especially the 1940s
in Latin
America.
From comparing Tables 2 and 4 it may be suggested that SPls are
more
stable in
core countries than in Latin America. When we compare changes
over time in
Latin America and the core, and test the hypothesis that there
is no
difference between these changes, no significant differences are found
for the
1800-1850, and the 1850-1900 periods. However, for the 1900 to 1950
period, the
core SPI changes .9 while the Latin American SPI changes 7.9
points, and
this difference is statistically significant at the .01 level.
From 1950 to 1975 the core declines -1.9 points while the Latin
American
countries
increase 1.2 points, and this difference is significant at the
-.01 level.
Note that from 1900 to 1950 the core changes much less than
Latin
America, supporting the hypothesis of core stability. But from 1950
to 1975 the
core changes more than Latin America and the changes occur in
opposite
directions. Latin American countries continue to become more
primate,
while the core countries are, on the average, moving toward less
hierarchical
city systems.
Discussion
and conclusions
We have found patterns in the above one-variable analysis which
indicate
that indeed
there are differences between Latin American and core country
city-size
distributions. In both the core and in Latin America we can find
primate and
flat city-size distributions, and in both zones of the world-
economy we
can find instances where a country moves toward greater primacy
or greater
flatness. Statistically significant differences in the city-size
distributions
do not emerge until the twentieth century, and are largely due
to the major
shift toward primacy which occurs in Latin America in the
1920s, 1930s
and 1940s.
The relationship between
position in the world economy and urban primacy
is not a
simple one. While all Latin American countries except Colombia
are primate
by 1950, all core countries are not lognormal or flat. Indeed,
in 1975 ten
of the seventeen core countries may be described as primate.
two as
lognormal, and five as flat. Trends, however, show that while Latin
American city
systems continue to increase their levels of primacy, all of
the core
countries, with the exception of Belgium, are either stable or
becoming
increasingly less hierarchical in the 1950-1975 period.
How ought different theories of the causes of urban primacy
account
for the
patterns found? Generally we may classify theoretical approaches
to city-size
distributions as those which focus on the distribution of
institutionalized
power over space, and those which focus on the economically
efficient location
of production and services in space. Some have claimed
that Latin
American countries exhibit greater degrees of primacy because of
the heritage
of colonial institutions which located political power in the
administrative
cities which functioned as extractors of surplus from the
countryside.
Cardoso (1975) makes an argument of this sort in his
discussion of
Latin American urban politics.
While such arguments may be useful in explaining other
institutional
differences
between core and Latin American societies, they do not help
explain the
city-size patterns we find. Latin American city systems only
became
significantly more primate than those in the core in the twentieth
century. It
is possible, however, that another version of the political
power approach
is more compatible with the results. Frank (1979) has argued
that
peripheral countries engage in more autocentric development during
periods when
core-periphery trade is reduced or when core countries are
fighting
among themselves. More autonomous state formation in the periphery
is often
accompanied by import-substitution which locates new industries in
urban areas.
When this is further encouraged by national states much of
the new
investment is located in the already largest and most politically
central
cities, thus increasing urban primacy, although other authors
(Arregui,
1981) have argued just the opposite: that such periods will see
a trend
toward the growth of middle-size cities specialized in manufacturing.
The 1930s and 1940s saw the biggest increases in Latin American
primacy.
In the 1930s
core-periphery trade was reduced as the world-economy stagnated.
Prices of
traditional peripheral exports plumetted, and Latin American
countries
were not able to buy manufactured goods from core countries. The
rise of
populist governments in several Latin American countries encouraged
a more
self-reliant approach to development in this period and import
substitution
was encouraged by state policies (Cardoso and Faletto, 1979).
In the 1940s
the demand for peripheral raw materials returned as a result of
the war needs
of core states, but manufactured goods from the core were still
in short
supply due to the war. Further investment in manufacturing and
heavy
industry occurred in many countries, and this tended to be located in
the already
large cities.
Industrialization and the growth of more capital intensive
production
does not
necessarily cause urban primacy. In Britain the industrial
revolution
and attendant production for the world market led to a reduction
in the level
of British primacy between 1801 and 1851 as Liverpool,
Manchester,
Birmingham and Glasgow grew more rapidly than London. The
industrialization
which occurred in the United States in the nineteenth
century did
not much flatten the already lognormal city-size distribution,
but neither
did it cause primacy.
Does industrialization result in primacy in Latin American but not
in
core
countries, and if this is so what accounts for it? First let us look
at Latin
American countries to see if the early industrializers increase
primacy
before the later industrializers. A more sophisticated test of this
hypothesis
will be possible when we have time series on the economic
development
and international trade of Latin American countries, but for now
let us make a
rough overview. In Brazil , Mexico, Argentina and Chile, import
substitution
industrialization begins earlier than in other Latin American
countries.
Three of these countries become more primate during the period
of import
substitution. Mexico and Chile slowly increase their levels of
primacy from
1900 to 1975. Brazil moves toward two-city primacy in the
1940s but the
steepness of its overall city size distribution declines
somewhat
after 1950. Argentina does not conform at all to the hypothesis
that import
substitution industrialization causes increased primacy.
Argentina is
very primate during the whole twentieth century, with primacy
declining
slightly between 1910 and 1930 (see Figure 1 above).
The Dominican Republic, Bolivia and Paraguay have sharp increases
toward
primacy between 1900 and 1910. Between 1910 and 1920 Honduras,
Panama and
again the Dominican Republic increase primacy. In the twenties,
El Salvador,
Bolivia, Nicaragua, Peru and Panama have sharp increases, and
in the
thirties El Salvador, the Dominican Republic, Ecuador, and Nicaragua
do the same.
These shifts do not support the hypothesis that import
substitution
industrialization is a major cause of increased primacy. These
countries
experienced only minor shifts toward import substitution during
the earlier
decades of this century. Neither do these facts support the
hypothesis of
El-Shakhs (1972) that primacy increases with middle levels
of economic
development.
Again, more rigorous analysis of the city-size data in combination
with
measures of
other variables is required. But at this point it seems unlikely
that the
shift toward import substitution industrialization and later export
industrialization
(Gereffi and Evans, 1981) can explain much about increases
in primacy.
The striking thing is that almost all Latin American countries
increase
their levels of primacy in the twentieth century and so we may
suspect that
contextual world-system properties which affect all countries
over this
period may be responsible. Secular changes in the core-periphery
division of
labor do not affect all countries equally however, and in order
to test this
kind of proposition we must be much more specific about the
kinds of
institutional change that is meant.
Comparison of Latin America with other peripheral areas will shed
more
light on the
possible explanations for increasing urban primacy, although
this is
beyond the purview of this paper. It may be that certain kinds of
foreign
investment increase primacy, or that the emergence of increasing
ties among
peripheral countries accounts for the relatively rapid growth
of the
largest cities. Students of the system of world cities have noted
that a
certain standardization has occurred such that every “international"
city must
have certain amenities (an airport, banks, hotels, etc.). Thus
mere
participation in the world-economy requires smaller and less developed
countries to
have at least one large city. This requirement will have little
effect on the
more integrated and elaborated city systems of the core, but
in small
peripheral countries it may account for increased primacy.
Endnotes
1. This is a
preliminary report from a project which is studying the
evolution of
the world city system and national city systems from 1800 to
1980. The
major hypothesis being examined is that a nation-states' location
and
trajectory in the larger world-system is an important determinant of
features of
its city system. In order to complete this project additional
data must be
gathered on the location of national societies in the world-
economy and
interstate system.
2. A
confusing situation may arise with the SPI when both positive and
negative
deviations appear in the distribution, that is, when some cities
are smaller
and others are larger than expected according to the lognormal
rule. This
occurs in the situation of what is called "two-city primacy,"
when a
country has two very large cities and no middle-size cities. Thus
the second
city is larger than expected and the others are smaller. The
sign of the
SPI is determined by the sum of these deviations and thus small
changes in
the relative size of cities can cause a large SPI to radically
change its
sign if the sum of the positive and negative deviations is around
zero. In
situations of this kind (two-city primacy) the SPI is an unreliable
indicator of
the overall steepness or flatness of a city size hierarchy (a
situation we
call "sign-flip") and thus we calculate other measures. One
is the least
squares slope of the overall distribution, which is set equal
to -1 when
the distribution is lognormal and thus varies around -1 (see
Table 1). The
other we label the SPI 3, an SPI index calculated ignoring
the largest
city. In the analysis contained in this paper, situations of
two-city
primacy have been handled by utilizing the SPI 3. This tells us
the
:steepness or flatness of the overall distribution of city sizes ignoring
the deviation
due to the existence of two similarly large cities.
We also
calculate a "three city SPI” which is based solely on the three
largest
cities. This allows us to focus on the relations among the largest
cities,
excluding complications due to the addition of the rest of the
distribution.
The combined use of these available indicators is made
necessary by
the inability of any single statistic to capture all the
information
available in a complex distribution.
3. Another
way of studying change in SPIs is to test the hypothesis that
two sample
means are really the same. The null hypothesis is that the
means are
really drawn from the same population and that sampling error alone
could account
for an apparent difference. When we perform such a comparison
of means for
the decades between 1900 and 1970, the years when Latin American
SPIs were
changing the most, none of the decade means are different from one
another at a
statistically significant level (less than .05). Indeed only
two twenty
year periods exhibit statistically significant differences between
means
(1930-50 and 1940-60) and these both contain the decade of World War
II when SPI
change was greatest. When we consider 30 year lags we begin to
more
frequently find mean differences which are statistically significant.
Although the
sample is small, this confirms earlier concerns that studying
changes in
primacy over short periods reveals mainly measurement error, and
thus the slow
rate of change of national city systems requires a research
design with
longer time lags (Chase-Dunn, 1982).
4. Of these
seventeen core countries, five (United Kingdom, France,
Netherlands,
Belgium and Austria) were already in the core in 1800. Four
entered the
core in the nineteenth century (United States, Italy, Germany, and
Switzerland),
and eight entered the core in the twentieth century (Japan,
Sweden,
Denmark, Finland, Norway, Australia, Canada and the Soviet Union).
References
Arregui,
Patricia
1981 "Dependence and primacy: trends and
fluctuations in Peruvian
urban history,11 Ph.D. prospectus,
Department of Social
Relations, Johns Hopkins University.
Cardoso,
Fernando H.
1975 “The city and politics," pp.
157-190 in Jorge Hardoy (ed.)
Urbanization in Latin America.
Garden City, N.Y. : Anchor
Chase-Dunn,
Christopher
1982 “World division of labor and the
development of city-systems:
a longitudinal study,'1 Comparative
Research 9,3:3-9 (Winter).
Forthcoming
“The system of world cities, 800-1975,” in Michael Timberlake
(ed.) Urbanization in the
World-Economy. New York: Academic
Press.
Christaller,
Walter
1967 (1933)
Central Places in Southern Germany. Translated by
C. W. Baskin. Englewood Cliffs,
N.J.: Prentice-Hall.
Davis,
Kingsley
1969 World Urbanization 1950-1970, Volume 2.
Institute of
International Studies, University
of California, Berkeley.
El-Shakhs, Salah
1972 “IDevelopment, primacy and systems of
cities,” Journal of
Developing Areas 7(October): 11-36.
Frank, Andre
Gunder
1969 Latin America: Underdevelopment or
Revolution? New York:
Monthly Review Press.
Gereffi, Gary
and Peter Evans
1981 “Transnational corporations, dependent
development, and state
policy in the semiperiphery: a
comparison of Brazil and
Mexico.” Latin American Research
Review 16, 3:31-64.
Hardoy, Jorge
and Maria Elena Langdon
1978 “Analisis estadistico preliminar de la
Urbanizacion de
America Latina entre 1850 y 1930,” Revista
Paraguaya de
Sociologia 42/43: 115-173
(May-December).
Kowalewski,
Stephen
1982 “The evolution of primate regional
systems,” Comparative Urban
Research 9, 1:60-78.
McGreevey,
William P.
1971 “A statistical analysis of primacy and
lognormality in the
size distribution of Latin
American cities, 1750-1960,” in
Richard M. Morse (ed.) The
Urban Development of Latin America
1750-1920. Center for Latin
American Studies, Stanford
University, Palo Alto, California.
Morse,
Richard M. (ed.) with Michael Conniff and John Wibel
1971 The Urban Development of Latin America
1750-1920. Stanford
University: Center for Latin
American Studies.
Portes,
Alejandro
1976 “The economy and ecology of urban
poverty,” Pp. 7-69 in
Alejandro Portes and John Walton,
Urban Latin America: The
Political Condition from Above
and Below. Austin: University
of Texas Press.
Roberts,
Bryan
1979 Cities of Peasants. Beverly Hills:
Sage.
Smith, Carol
A.
1980 “On urban primacy, export dependency
and class struggle in
peripheral regions of world
capitalism," Typescript,
Department of Anthropology, Duke
University.
1982 "Placing formal geographical
models into cultural contexts:
the anthropological study of
urban systems,” Comparative
Urban Research 9, 1:50-59.
Vapnarsky,
Cesar A.
1975 "The Argentine system of cities:
primacy and rank-size rule,"
Pp. 369-390 in Jorge Hardoy (ed.)
Urbanization in Latin
America. Garden City, N.Y.:
Anchor.
Walters,
Pamela B.
1985 “Systems
of cities and urban primacy: problems of definition
and measurement,” in Michael
Timberlake (ed.) Urbanization in
the World-Economy. New York:
Academic Press.