-- Project Proposal --
Connectivity:
A Key Factor in International Development
John Gerring
Department of Political Science
Boston University
[email protected]
Draft: Please do not cite without permission
At the turn of the twenty-first century, the quality of life enjoyed by persons around the world varies
enormously. While there may be a slow convergence in human development (Gerring et al. 2008;
Goesling, Firebaugh 2000; Neumayer 2003), it remains the case that one’s chances of surviving past
infancy, maintaining good health, getting educated, finding a job, and escaping poverty are largely
contingent upon what part of the planet one happens to be born in.
The markedly spatial pattern of development, both within and across countries, suggests
three broad lines of explanation. It could be that areas with high and low levels of development are
intrinsically different -- ecologically, economically, culturally, sociologically, politically, or historically
-- in ways that have enhanced/diminished their development potential. London (England) is more
developed than Bangui (Central African Republic) because London is different from Bangui. There
are many versions of this causal story, which may rest on the intrinsic properties of a land and its
climate (Hibbs, Olsson, Diamond 2004; Landes 1999), population growth in tandem with
technological innovation (Boserup 1965; Galor, Weil 1999; Goodfriend, McDermott 1995; Jones
2001; Kremer 1993), public policies and political institutions (Acemoglu, Johnson, Robinson 2004;
Glaeser, Lopez-de-Silanes, Shleifer 2004), cultural properties (Iyigun 2007; Landes 2006), and
historical contingencies (Pomeranz 2000).
A second possibility is that development is an essentially zero-sum distributional game in
which regions of the world compete. Developed regions are therefore those possessing superior
military and geo-political capacities, allowing them to monopolize technology, physical capital,
human capital, and other factors that contribute to development. London is more developed then
Bangui because London (along with London’s allies) is militarily and organizationally dominant. The
strongest and most systematic version of this story is associated with the dependency and worldsystems school of thought (Benton 1996; Bergeson 1980; Prebisch 1950; Snyder, Kick 1979; Van
Rossem 1996; Wallerstein 1974).
A final possibility is that development is the product of a region’s “connectivity.” London is
more developed than Bangui because London has greater access to the world. This, in turn, is a
product of its favorable location, transport infrastructure, and the location of adjacent populations.
Connectivity is thus understood as a structural feature of the landscape -- a product of geographic,
infrastructural, and demographic features that make it easy or difficult for individuals living in a
region to reach those outside the boundaries of that region. (A region may be any size.) We expect
that relative ease of movement will facilitate contact and that contact will, in turn, hasten
development. To motivate the analysis, let us consider briefly several reasons why this might be so.
First, connectivity facilitates the diffusion of technology, ideas, and norms. Second,
connectivity fosters the standardization of technology, ideas, norms, formal regulations, and
language across regions and this by itself (regardless of the quality of these factors) enhances
efficiencies and presumably also speeds the ongoing process of diffusion, since ideas travel more
quickly where cultural barriers are less severe. Third, connectivity reduces transaction costs, lowering
the price and increasing the variety and availability of consumer goods and the potential markets for
producer goods. Fourth, connectivity enhances the mobility of labor and capital, thus reducing
inequalities -- locally, nationally, and internationally (wherever connectivities exist) – according to
Ricardian theory. Fifth, connectivity probably strengthens the capacity of governments. It is easier
for states to perform basic functions like keeping order, collecting taxes, and surveyance, as well as
more complex policy interventions, when the population under its purview is well-connected. (We
assume that state capacity generally has a positive relationship to social and economic development.)
Sixth, connectivity enhances the access of non-market, non-governmental actors whose involvement
in the project of development is essentially voluntaristic (or at least nominally so). Insofar as
missionaries, health workers, educators, and other actors rooted in civil society have played a
2
positive role in development, this too may be related to the relative connectivity of different regions.
At the same time, connectivity reduces the power of governments, religions, businesses, and other
organizations to restrict the flow of ideas and technologies. In this respect, connectivity is inherently
anti-monopolistic, fostering a spirit of open competition across jurisdictions that are in contact with
one another. Finally, connectivity may nurture a more universalistic set of values and norms, which
may in turn play a positive role in the complex and intertwined processes of development. Regions
with high connectivity breed cosmopolitan citizens of the world, while regions with low connectivity
encourage a parochial outlook on life, love, and duty.
Several important caveats must be attached to these optimistic prognostications. First, it
should be understood that these are not invariant laws. We readily acknowledge situations in which
high connectivity has served as a detriment to short-term -- and perhaps even long-term –
development. Access to coastal areas in Africa facilitated the slave trade (Nunn, Puga 2007); access
to European explorers facilitated the spread of epidemics in the New World (Diamond 1997); and
access to the outside world today facilitates the spread of HIV/AIDS. These are important
exceptions to our argument.
Second, to the extent that our argument holds, it should not be interpreted as a monocausal
view of historical development. Connectivity matters, all else equal. But all else is not equal in the
real world, and many additional factors presumably impinge upon long-term patterns of
development. Our ambition is therefore limited to explaining some portion of the variance in this
complex set of outcomes. It should be noted that we do not test the impact of connectivity relative
to many other factors such as population, culture, institutions, policies, and technology, which are
viewed as endogenous to the theory.
Third, it must be appreciated that many of the benefits that arising from connectivity do not
manifest themselves straightaway. Indeed, the immediate effect of enhanced connectivity is often to
exacerbate latent social conflicts. (One thinks of the effect of China’s recently completed rail
connection to Tibet, which seems to have served as a stimulus for protest.) Over time, however, we
assume that the effects of connectivity will generally be positive – at least, for those who survive the
often-violent nature of initial contact.
Finally, it must be stressed that our use of the concept of development is limited to the
material dimensions of human life -- e.g., health, mortality, education, housing, employment,
income, and assets -- and does not encompass the spiritual or emotional quality of life, happiness,
liberty, empowerment, cultural integrity, or the quality of the environment.
With these caveats, the theory is plausible – perhaps even commonsensical. Indeed, the idea
of connectivity, as sketched here, incorporates insights from several closely inter-related areas of
current scholarship including: geography (Crafts, Venables 2001; Diamond 1997; Faye, McArthur,
Sachs, Snow 2000; Gallup, Sachs, Mellinger 1999; Hibbs, Olsson, Diamond 2004; Jones 1981;
Krugman, Venables 1995; Mellinger, Sachs 1999; Mellinger, Sachs, Gallup 2000; Olsson 2003;
Olsson, Hibbs 2005), transport infrastructure (Allen, Bourke,Gibson 2005; Ayogu 2007; Christiaensen,
Demery, Paternostro 2005; Coatsworth 1979; Cook et al. 2005; Escobal, Torero 2005; Fay,
Leipziger, Wdon 2005; Fedderke, Perkins, Luiz 2006; Fernald 1999; Gibson, Rozelle 2003;
Gramlich 1994; Hummels 2007; Jacoby 2000; Kessides 1993; Kim 2006; Limao, Venables 2000;
Munnell 1992; Mu, van de Walle 2007; Ravallion, Wodon 1999; Sabandar 2007; van de Walle
2002; Wilkie et al. 2000; World Bank 1994), diffusion (Barro, Sala-i-Martin 1997; Bretschger 1997;
Keller 2004; Spolaore, Wacziarg 2006), and network position (Flandreau, Jobst 2005; Hidalgo et al.
2007; Hojman, Szeidl 2006; Kali, Reyes 2006; Kim, Shin 2002).
The theory departs from extant work in three principal respects. First, we adopt a broad
conception of development, extending well beyond the usual economic outcomes (usually measured
3
by GDP per capita or trade flows). Indeed, it seems probable that the causal effect of connectivity
may be greater on human development than on economic development, at least in the contemporary
era. Second, the theory presents a broad theoretical framework, encompassing elements from many
traditions, including those listed above. It thereby articulates a more complete -- and arguably more
compelling -- account of the role of cross-district contact in creating and sustaining international
development. Finally, we hope to present a more elaborate and robust set of empirical tests featuring
a wide variety of connectivity measures (focused on local, national, and global connectivities and on
diverse transport infrastructures), both short- and long-term causal effects, individual and countrylevel outcomes, and multi-level modeling techniques.1
Measuring Connectivity
The initial unit of analysis envisioned in this study is the district (to be defined, but perhaps about
twenty-five square miles). For each district on earth, we intend to calculate various connectivity
measures by looking at the time that would be required to travel from one district to various
destinations outside that district, using modes of transport usual to that time and place. Travel-time
to various destinations thus comprises the basic metric by which connectivity will be measured.2
Destinations of concern are divided into five categories: transport routes, urbanization measures, country
measures, sub-continental measures, and global measures.
Transport routes include: (A) the nearest ocean or navigable river and (B) the nearest trunk
road (considered separately). Thus, for a given district, its connectivity scores are equivalent to the
travel-time required to reach (A) or (B). We count only ground transportation infrastructure. (Air
transport becomes a factor in development only in the late twentieth century, where it is more or
less ubiquitous with cities, and thus captured in the following set of indicators.)
Urbanization measures include: (C) the nearest city, (D) the nearest city in the country, (E)
the nearest city in the region, and (F) all global cities (mean). Again, it is the travel-time to these
destinations that determines a district’s connectivity score. (The definition of a “city” will have to
worked out; it will presumably depend upon population, though the threshold may vary across
different historical epochs.)
Country measures include: (G) all country districts (mean) and (H) all country districts
weighted by their respective population (so that connections with densely settled districts are given
greater weight). Countries are defined according to contemporary definitions and boundaries, and
include both sovereign and semisovereign states (e.g., Puerto Rico). (At a later date, we may choose
to provide historical codings of political units.)
Sub-continental measures are identical to country measures except that the territory is
defined in purely geographic terms. A sub-continent is understood to include all land areas lying
within 3,000 miles of the district being coded. Sub-continental measures are thus devised as: (I) all
districts within a sub-continent (mean) and (J) all districts within a sub-continent weighted by their
respective populations.
Global measures are similar except that in this instance no boundaries are placed upon the
analysis. Thus, global measures are devised from: (K) all global districts (mean) and (L) all global
While several similar studies have been conducted within countries or across countries, we know of no studies that
combine these two levels of analysis.
2 One might also construct a measure of connectedness based on the cost of transport, and this could be further
disaggregated into cost as a percent of GDP in a given country -- thus capturing the fact that it is easier for Americans to
travel to Kenya than for Kenyans to travel to the US. However, this requires data that are virtually impossible to gather
for eras prior to the late twentieth century. Even so, in a purely cross-sectional analysis (focused on 2000) it might be
worthwhile.
1
4
districts weighted by their respective populations. In addition, we construct a final variable (M) that
is identical to L except that the weighting principle is assigned by the connectivity of each
destination. This captures the intuition that a connection to a highly-connectivity district is more
valuable than a connection to a low-connectivity district. It might also be thought of as second-order
connectivity. (In principle, the same weighting could be assigned to connectivity at a country and
sub-continental levels; however, we do not anticipate that these would greatly alter a district’s score.)
It will be seen that these thirteen connectivity variables, all of which are listed in Table 1,
capture different aspects of the general theory. Each is therefore regarded as a sub-hypothesis, to be
independently (and perhaps jointly) tested. We anticipate that this initial list of variables will diminish
over time. It seems likely, for example, that some of these variables will be highly correlated, and
therefore empirically redundant. Even so, a great deal may be gained by looking at small variations in
empirical performance across measures that are constructed in different ways; this provides a crucial
clue into causal mechanisms that may be at work. We need to know not only whether connectivity
matters for development, but also what kinds of connectivity matter most.
5
Table 1:
Connectivity Variables: A Preliminary List
Travel-time to…
Transport route measures:
(A) Nearest ocean or navigable river
(B) Nearest trunk road
Urbanization measures:
(C) Nearest city
(D) Nearest city in the country
(E) Nearest city in the region
(F) All global cities (mean)
Country measures:
(G) All country districts (mean)
(H) All country districts, pop-weighted (mean)
Sub-continental measures:
(I) All sub-continental districts (mean)
(J) All sub-continental districts, pop-weighted (mean)
Global measures:
(K) All global districts (mean)
(L) All global districts, pop-weighted (mean)
(M) All global districts, connectivity-weighted (mean)
1
2
3
Country
(pop-weighted
district mean )
4
District
Country
(district mean )
300BC, 1820, 1950, 2000
1820, 1950, 2000
300BC, 1820, 1950, 2000
1820, 1950, 2000
1820, 1950, 2000
1820, 1950, 2000
2000
2000
300BC, 1820, 1950, 2000
1950, 2000
300BC, 1820, 1950, 2000
300BC, 1820, 1950, 2000
1820, 1950, 2000
1950, 2000
1820, 1950, 2000
300BC, 1820, 1950, 2000
1820, 1950, 2000
1950, 2000
1820, 1950, 2000
1820, 1950, 2000
2000
2000
2000
2000
1950, 2000
1950, 2000
1950, 2000
1950, 2000
1950, 2000
1950, 2000
2000
2000
300BC, 1820, 1950, 2000
1950, 2000
1820, 1950, 2000
1950, 2000
1820, 1950, 2000
1950, 2000
2000
2000
300BC, 1820, 1950, 2000
1950, 2000
300BC, 1820, 1950, 2000
300BC, 1820, 1950, 2000
1950, 2000
300BC, 1820, 1950, 2000
1820, 1950, 2000
1950, 2000
1820, 1950, 2000
2000
2000
2000
Individual
Variables may be observed at four historical points in time: a) 300 BC, b) 1820, c) 1950, and
d) 2000. These are understood as approximate dates, data to be drawn from the nearest available
year. Since data is not available for all periods, and since the relevance of some factors (e.g.,
“country”) is primarily contemporary, not all variables will be coded in all periods, as signaled in
Table 1.
Information necessary to code districts during the four periods may be understood as
geographic (the topography of the land, the existence of navigable rivers and oceans), infrastructural
(transport networks), or demographic (the location and density of human population). Because
geographic factors are relatively constant through time (with the exception of important changes like
the construction of canals), codings are easily extended back in time. Because we know something
about the changing shape and location of transport infrastructure, this too should be amenable to
historical coding, as discussed below. The density of populations across the globe is known with a
high degree of accuracy in the latter twentieth century, and historical estimates are available for 1820.
Prior to that, populations were fairly diffuse everywhere; hence, we do not bother to measure the
distribution of populations in 300 BC.
300 BC is chosen as a point of departure because it marks a time prior to the invention of an
important transportation technology, the triangular “lateen” sail. This allowed ships to tack (sail into
the wind) and thus vastly enhanced the range of sailing vessels, leading to cross-ocean voyages. Up
until this signal invention (estimated to be about 200 BC), differences in relative connectivity around
the world were largely the product of natural features of the topography.3 Travel-times over land
routes are estimated at walking speed for areas that are tractable and at half-walking speed for areas
that are “rugged” (Nunn, Puga 2006). Large bodies of water are assumed to pose insuperable
transport barriers, and thus are entirely unconnected.
The second point of measurement occurs just prior to another transport revolution -- the
advent of the railroad and steamship – which greatly diminished transport costs within and across
countries (Bairoch 1990: 142; Crafts, Venables 2001; Harley 1988). In 1820, travel-time over water
(navigable rivers, lakes, and oceans) will be calculated according to the average speed of sailing ships
at that time. Travel-times over land routes will again be estimated at walking speed for areas that are
tractable and at half-walking speed for rugged areas.
In 1950 (arguably, another watershed moment in transport technology [Dollar 2001]), traveltimes over water (river, lake, or ocean) will be calculated according to the average ocean shipping
speed at the time. Travel-times over land will be calculated using maps that locate the existence of
railroads and paved roads and estimates of the typical rate of transport over these facilities at that
time. Where no transport infrastructure is in place, travel-time will be calculated at walking speed for
areas that are tractable and half-walking speed for rugged areas. For the contemporary era (2000),
the same procedure is followed, with adjustments for changes in infrastructure and transport
technology. (We may also draw upon the infrastructure dataset developed by Canning [1998] for
codings in 1950 and 2000.)
Thus, thirteen variables (A-M) will be measured for all districts on earth (except Antarctica)
at several points in time (column 1, Table 1). Given the relatively small number of decision-rules
noted above, GIS programming should be fairly simple.
Once coded, these thirteen district-based variables can be transformed easily into
connectivity scores for larger geographic units such as the nation-state, so as to allow for countrylevel analysis. Two aggregation techniques seem appropriate. First, we average the values for all
To be sure, the presence (absence) of domesticable pack animals enhanced the carrying capacity of traders and warriors
in areas where such animals were present (Diamond 1997). At some point, we may choose to incorporate this sort of
information into our analysis.
3
districts lying within a contemporary nation-state (column 2, Table 1). Second, we apply a
population-weighted average of all districts lying within a nation-state (column 3, Table 1).
We can also use the district values for these thirteen variables to assign scores to individual
survey or census respondents -- so long as we know the location of these individuals (geographic
locations must be mapped onto districts). This provides the underlying information necessary for
individual-level variables (column 4, Table 1). Since individual-level data is generally available only
for recent years, this set of variables is observed at only one point in time, in 2000.
We recognize that many of the causal effects that are assumed to flow from physical
proximity might also be achieved by communications infrastructure. Sometimes, mail, telephone,
television, and internet networks are as important, or more important, than transport networks.
However, for a variety of pragmatic reasons, we choose to focus on the former. Note that these two
factors are highly correlated over the long course of human history, where communications
occurred by word of mouth or written word, and thus depended upon physical connectivity. In the
twenty-first century, we are perhaps beginning to see a disconnect between transportation and
communications networks, in which the latter outstrip the former. However, in previous eras the
movement of mail as well as the installation of hard-line telephones required a strong transportation
network. Because our focus is on developments over the course of the twentieth century, we expect
that the effect of transport will be more important – or at any rate, be highly correlated with –
communications networks. Second, transportation architecture is easier to measure; we know the
general location of major roads, navigable rivers, and deep-water harbors. And because we know the
density of population settlement relative to these features, we can intuit the strength of physical
connectivity. Measuring the density of communications technology is much more difficult in the
recent era (and, as we have said, much less important in previous eras, since the latter depended
upon physical connectivity).
Testing the Causal Effects of Connectivity
The effect of connectivity on development will be tested at individual and country levels. Where
data is sufficient, these two levels will be combined with the use of hierarchical (multi-level) models.
Development outcomes of concern include quality-of-life measures (e.g., life expectancy,
infant mortality, maternal mortality, malnutrition), fertility (a key element of the demographic
transition), and income (as measured by GDP per capita for countries and assets or consumption for
individuals).4 Each of these outcomes may be independently analyzed. In addition, we propose to
subject all outcomes at the individual and all outcomes at the country level to factor analysis so as to
identify a few dimensions that may be said to represent in summary form the complex process of
development. Thus, the left-side variable in the following analyses may consist of a single outcome
measure or a composite indicator of development.
Country-level data is usually available on an annual or quasi-annual basis throughout the late
twentieth century (World Development Indicators 2007), allowing for both cross-sectional and panel
analyses from 1950 to 2004. Note that the theory supposes that connectivity will affect both longterm as well as short-term outcomes, i.e., the relative status of countries at a particular point in time
as well as the rate of improvement in development indicators over some period of time such as the
last half-century.
Individual-level data will be drawn from merged DHS surveys, along with additional surveys
and census data wherever DHS surveys are unavailable. These surveys provide a highly detailed
We may also consider analyzing several measures of political development at the national level, e.g., the six World Bank
indicators compiled by Daniel Kaufmann and colleagues.
4
8
picture of development across the contemporary world. However, there is probably not sufficient
historical data to conduct meaningful time-series analysis, given the sluggishness of our measures of
connectivity. Hierarchical (multi-level) modeling techniques will be employed in these cross-sectional
analyses so as to model individual and national-level phenomena simultaneously.
In both country- and individual-level analyses it is important to differentiate between the
effects of connectivity on economic outcomes (as measured by GDP per capita at the national level
and personal assets at the individual level) and quality-of-life outcomes (health, mortality, education,
and so forth). Since these different species of development are mutually constitutive, and not easily
modeled, this is no mean feat. One approach is to run identically specified regression models with
alternative outcomes, comparing model-fit across the models as a clue to the varying strength of the
underlying relationships. Another approach is to include economic measures as controls in the
analysis of noneconomic outcomes, and vice-versa, thereby partialing out their effects. Does
connectivity affect quality of life when controlling for income or personal assets? Does connectivity
affect income or assets when controlling for quality of life?
A larger problem encountered in all of these regression models is known generically as an
assignment (or selection) problem. In the absence of random assignment, it is difficult to ascertain
whether regressors are exogenous to the outcome and uncorrelated with possible (unmeasured)
confounders. Note that many of the connectivity measures in Table 1 are affected by infrastructure
policies. Travel-time depends upon the quality of public transport, and this in turn is a product of
investments. The problem is that infrastructural policies are not randomly assigned, and may co-vary
with other factors that also affect development outcomes. Thus, investments in infrastructure may
be concentrated in regions that are, for some intrinsic reason, suited to development. Or such
investments may be correlated with other (non-infrastructural) policies that exert an independent
effect on development outcomes. Both of these scenarios would create a spurious result; a district’s
connectivity score might be correlated with developmental performance for reasons having nothing
to do with connectivity per se.
Several approaches may be taken to this identification problem. First, we intend to lag
independent variables one or several periods behind the outcome of interest. Thus, for an outcome
measured in 2000, lags of 50, 200, and 2,300 years will be introduced. Note that all measures
constructed for 300 BC are purely geographical, and thus entirely exogenous. These may be also
employed as instruments in a two-stage least squares analysis, providing a second approach to the
identification problem. Here, we must be attentive to the potential problem of chosen instruments
that correlate with other factors that have independent effects on long-term development (through
pathways other than connectivity). For example, navigable rivers provide not only an avenue of
natural transport but also a source of water and rich topsoil for settled agriculture – the material
basis for all pre-modern civilizations. Similarly, isolated regions are likely to be characterized by poor
agricultural land with few exploitable resources, and perhaps epidemiological dangers.
Clearly, all regression analyses (whether one-stage or two-stage) must be attentive to
alternative explanations of development, particularly those which are likely to be correlated with our
connectivity measures. This includes (but is not limited to) measures of agricultural fecundity,
domesticable plants and draft animals, mineral resources, and epidemiological threats. Fortunately,
these factors have been a primary focus of recent research (see previous citations), and are easy to
incorporate with GIS data since they rest largely on the geographic features of a territory, features
that have changed relatively little over the past century and are not usually amenable to policy
intervention. In some cases, it will be necessary to reconstruct these measures so that they describe
districts rather than nation-states. Thus, we will need to integrate maps showing the quality of soil,
climate, topography, mineral resources, and other factors that can be measured at a district level.
Other factors, such as the presence of domesticable plants and animals in pre-historical times, need
9
not be measured at the district level (district-level variables may be constructed using country-level
data).5
To reprise, the core empirical strategy is to regress measures of development against
measures of connectivity, along with a wide range of controls, in one- or two-stage models. Note
that in models with purely geographic measures of connectivity (e.g., connectivity measured in 300
BC) the set of relevant controls is presumably limited to other geographic controls. All other factors
in development – e..g, technology, population, trade – may be viewed as endogenous. In models
where connectivity variables draw on infrastructure and demography, a larger and more complicated
identification problem is faced, requiring a larger set of controls and suggesting a two-stage
approach. We view these two types of models as playing a complementary role. The purely
geographic measures of connectivity should establish the plausibility of the causal argument. The
more refined – but also more endogenous – measures of connectivity should allow us to clarify
which sorts of connections matter most for development.
One final empirical problem must be dealt with. In recent decades, sub-Saharan Africa has
been devastated by the HIV/AIDS epidemic. The regional concentration of this disease – which,
thus far, has not had a substantial effect on mortality rates in other parts of the world – is accidental
with respect to the theoretical interests of this study. To be sure, the diffusion of HIV/AIDS bears
directly on the theory of connectivity, since more connected districts are presumably at higher risk.
HIV/AIDS thus offers a good example of a negative connectivity effect, and in this respect is
similar to previous epidemics (e.g., smallpox, cholera). However, at the present time the effects of
HIV/AIDS are heavily concentrated on one region of the world – an accidental feature of the
disease that has no bearing on our theory. Its effects on human welfare are so enormous that it
cannot be absorbed as part of the error term. Two approaches might be taken to address this
problem. The first option would be to endogenize it by including a measure of HIV prevalence as a
regressor in each analysis. (This would probably have to be a country-level indicator, since
disaggregated analyses are not yet available, to our knowlege.) A second option would be to measure
developmental outcomes of interest in eras prior to the spread of the disease – i.e., to back-date the
last observations in each regression to the mid-to-late 1980s.
In addition to the “global modeling” approach it may also be possible to exploit “natural
experiments” that have taken place via immigration history across the globe. I am thinking of
situations in which an immigrant group settles in several territories within a new country, where
those territories exemplify high and low connectivity, and – crucially -- where the settlement
decision is the product of exogenous factors (e.g., family or village connections) having no bearing
on development outcomes of interest, and thus where ceteris paribus conditions across groups is
plausibly maintained. Further research on this possibility seems warranted.
Alternative Theories of Long-Run Development, Briefly Considered
Technology and technical knowledge (which we subsume under the rubric of technology) is
generally acknowledged as a critical factor in long-run development (Landes 1969; Mokyr 1992).
While most of the literature on this subject is focused on economic development it is easy to see
how technology might also contribute to human development (e.g., through improved knowledge
about the sources of public health, the availability of contraceptives, medicine, and so forth). Indeed,
technology is a critical factor in our speculative discussion of causal mechanisms (see above).
Unfortunately, technology is difficult to measure prior to the late twentieth century (and even now is
rather problematic). Its precise role in development thus resists empirical testing (but see Comin,
5
For examples of this kind of analysis at the country level see Escobal and Torero (2005), Ravallion (2005).
10
Easterly, Gong 2006). In lieu of direct tests, we make two critical assumptions about the
development and spread of technology. First, we assume that innovation is fostered by connectivity,
areas with high connectivity being more likely to serve as incubators for new ideas. Second, we
consider technological development to be primarily a product of diffusion rather than original
creation (Barro, Sala-i-Martin 1997; Keller 2004). Few inventions are truly independent of one
another, and the diffusionist feature becomes even more evident in contemporary eras of world
history. Our measures of connectivity should account for the likelihood that a given district will
benefit from the knowledge of humankind, regardless of where that knowledge was originally
created. Thus, technology is looked upon as largely endogenous to the theoretical framework.
Population is also regarded as a crucial factor in long-run development, since it may
stimulate innovation, political development, and other intermediate causal factors (Boserup 1965;
Galor, Weil 1999; Goodfriend, McDermott 1995; Jones 2001; Kremer 1993). While we do not
dispute the general argument, we view demographics as secondary to geography. It is no accident
that populations in some parts of the world have grown, while others have stagnated, over the
millennia of human history prior to the industrial revolution. Long-run demographic variation may
be accounted for by some combination of intrinsic (non-relational) and extrinsic (relational) features
of the landscape. While our theoretical interest is in the latter, our empirical challenge is to measure
both aspects adequately such that long-run population growth is accounted for with strictly
exogenous variables. If this can be accomplished, then population may be regarded as an
endogenous feature of the narrative.
Commercial exchange of goods and services is a central mechanism of contact across human
populations. As such, trade and trade policy serves as important causal pathway in any story of longterm development that hinges upon connectivity. Importantly, we assume the effect of trade on
development is best understood as a long-run relationship; short-term effects, in our view, are much
more difficult to theorize and to test. Thus we do not view cross-country regressions of trade
openness and growth during the postwar era (e.g., Yanikkaya 2003) as offering strong support for
or against the connectivity thesis. In any case, trade flows are viewed as largely endogenous to a
district’s (or a country’s) level of connectivity (Frankel, Romer 1999), and thus are not incorporated
explicitly into the analysis (except perhaps as a probe into causal mechanisms).
Objectives and Prospects
The objectives of this project may now be re-stated in more specific terms. We hope, first of all, to
provide a reasonably comprehensive map of human welfare as it varies spatially across the globe in
the contemporary era (based on individual-level data drawn from DHS surveys and other surveys
and censuses). In particular, we want to examine the variation that occurs (we presume) as one
moves a) from major transport arteries to more remote areas, b) from densely settled (“urban”)
regions to diffusely settled regions, and – more generally – c) from highly connected regions to
regions of low connectivity. These are descriptive questions, and perhaps best approached visually
(with maps of the world, color-coded to reveal possible causal processes).
The main tool of causal analysis is a series of regression models, with countries and/or
individuals as units of analysis. First, and most generally, we want to test whether connectivity
matters to development and whether the impact (if any) is significant. Second, we want to test
whether the strength of this relationship has changed significantly over the past two centuries. Does
connectivity increase, or diminish, in predictive power from 1820 to 2000? Third, we want to test
which of the manifold measures of connectivity demonstrate the strongest empirical relationship to
development. Are the effects of connectivity realized because of proximity to a) an urban area, b) a
transport hub, c) a dominant political unit such as a country, d) a larger region, or e) the rest of the
11
world? Fourth, we want to test whether these various types of connectivity have differential effects
over different time-periods. Is the force of connectivity on development increasingly global, and
decreasingly national and regional? Fifth, we want to test a variety of different development
outcomes – economic, demographic, and quality-of-life -- to see which are most sensitive to levels
(and types) of connectivity. Does connectivity have a greater impact on human development than on
economic development in the contemporary era? Sixth, we want to test whether certain regions of
the world are more sensitive to variations in connectivity. (This can be accomplished by conducting
split-sample regression tests or by constructing interactive variables in the full, global sample.)
Seventh, we want to test whether connectivity has a redistributive effect (transferring wealth to
poorer areas on the periphery or allowing poor persons to migrate to wealthier areas at the core),
leading to a more equal distribution of income. Finally, want to piece together the results from this
wide range of tests in order to shed light on the vexing question of causal mechanisms. Why and
how do connectivity factors affect varying developmental outcomes?
In order to fulfill these ambitious goals, the research must move forward on two tracks. The
first track involves the construction of a GIS database with which the key travel-time variables will
be calculated. In order to do this, several things need to happen. First, further research needs to be
carried out on travel-times usual to varying terrains and different transport types so that the
assumptions underlying each connectivity variable can be verified. Second, further coding issues
must be resolved, such as how to code travel-time between two districts that are impossible to reach,
given extant technology. Third, the locations of cities in the four time-periods need to be
determined, and entered into the database. Fourth, estimates of population density across the world
in 1950 and 2000 must be identified (presumably from historical atlases) and entered into the
database.
The second track involves the construction of a standard rectangular database (using Stata or
some other software) that integrates variables from DHS and other relevant surveys from around
the world. (Not sure what sort of software will handle such a large number of observations;
fortunately, the number of variables will not be overwhelming.) Relevant variables include: a)
development outcomes (health, malnutrition, et al.), b) the location of the respondent (so that the
individual’s district and country can be coded), c) all other factors that might affect the individual’s
level of development. Next, country-level variables of relevance to the analysis must be merged into
this dataset. Finally, the travel-time variables constructed with GIS software must be merged. At this
point, the data collection process will be complete and analysis may begin.
Data sources
CEPII: http://www.cepii.fr/anglaisgraph/bdd/distances.htm
12
References
Acemoglu, Daron; Simon Johnson; James A. Robinson. 2001. “Colonial Origins of Comparative
Development: An Empirical Investigation.” American Economic Review 91:5, 1369-1401.
Acemoglu, Daron; Simon Johnson; James A. Robinson. 2004. “Institutions as the Fundamental
Cause of Long-Run Growth.” NBER Working Paper 10481.
Allen, Bryant; R. Michael Bourke; John Gibson. 2005. “Poor Rural Places in Papua New Guinea.”
Asia Pacific Viewpoint 46:2 (August) 201-17.
Ayogu, Melvin. 2007. “Infrastructure and Economic Development in Africa: A Review.” Journal of
African Ecomomies 16, suppl. 1, 75-126.
Bairoch, Paul. 1990. “The Impact of Crop Yields, Agricultural Productivity and Transport Costs on
Urban Growth between 1820 and 1910.” In A.D. van der Woude, A. Hayami, and Jan de Vries
(eds), Urbanization in History (Oxford: Clarendon Books).
Barro, Robert J.; Xavier Sala-I-Martin. 1997. “Technological Diffusion, Convergence, and Growth.”
Journal of Economic Growth 2, 1-27.
Benton, Lauren. 1996. “From the World-Systems Perspective to Institutional World History:
Culture and Economy in Global Theory.” Journal of World History 7:2, 261-95.
Bergesen, Albert (ed). 1980. Studies of the Modern World-System. New York: Academic Press.
Bloom, David E.; Jeffrey Sachs. 1998. “Geography, Demography, and Economic Growth in Africa.”
Brookings Papers in Economic Activity 2, 207-95. [NS; downloadable]
Boserup, Ester. 1965. The Conditions of Agricultural Growth: The Economics of Agrarian Change under
Population Pressure. Chicago: Aldine.
Bretschger, Lucas. 1997. “International Trade, Knowledge Diffusion, and Growth.”
International Trade Journal 11:3, 327-48.
Canning, David. 1998. “A Database of World Stocks of Infrastructure, 1950-95.” World Bank
Economic Review 12:3 (September) 529-47.
Carneiro, Robert L. 1970. “A Theory of the Origin of the State.” Science 169, 733-38.
Christiaensen, Luc; Lionel Demery; Stafano Paternostro. 2005. “Reforms, Remoteness and Risk in
Africa: Understanding Inequality and Poverty During the 1990s.” In Ravi Kanbur, Anthony J.
Venables (eds), Spatial Inequality and Development (Oxford: Oxford University Press).
Clark, Gordon L.; Meric S. Gertler; Maryann P. Feldman (eds). 2000. The Oxford Handbook of
Economic Geography. Oxford: Oxford University Press.
Coatsworth, John H. 1979. “Indispensable Railroads in a Backward Economy: The Case of Mexico.”
Journal of Economic History 39:4 (December) 939-60.
Coleman, James S. 1988. “Social Capital in the Creation of Human Capital.” American Journal of
Sociology 94 (supplement) S95-S120.
Comin, Diego; William Easterly; Erick Gong. 2006. “Was the Wealth of Nations Determined in
1000 B.C.?”
Cook, Cynthia C.; Tyrrell Duncan; Somchai Jitsuchon; Anil Sharma; Wu Guobao. 2005. Assessing the
Impact of Transport and Energy Infrastructure on Poverty Reduction. Asian Development Bank.
Crafts, Nicholas; Anthony J. Venables. 2001. “Globalization in History: A Geographical
Perspective.” Unpublished manuscript, London School of Economics
Dollar, David. 2001. “Globalization, Inequality, and Poverty since 1980.” Washington: Development
Research Group, World Bank.
Duflo, Esther. 2001. “Schooling and Labor Market Consequences of School Construction in
Indonesia: Evidence from an Unusual Policy Experiment.” American Economic Review 91:4, 795813.
13
Dyson, Tim. 2001. “A Partial Theory of World Development: The Neglected Role of the
Demographic Transition in the Shaping of Modern Society.” International Journal of Population
Geography 7, 67-90.
Easterly, William; Jo Ritzen; Michael Woolcock. 2001. “Social Cohesion, Institutions, and Growth.”
MS, World Bank.
Englebert, Pierre. 2000. State Legitimacy and Development in Africa. Boulder: Lynne Rienner.
Escobal, Javier; Maximo Torero. 2005. “Adverse Geography and Differences in Welfare in Peru.” In
Ravi Kanbur, Anthony J. Venables (eds), Spatial Inequality and Development (Oxford: Oxford
University Press).
Fay, Marianne; Danny Leipziger; Quentin Wodon. 2005. “Achieving Child-Health-Related
Millennium Development Goals: The Role of Infrastructure.” World Development 33:8, 1267-84.
Faye, Michael L.; John W. McArthur; Jeffrey D. Sachs; Thomas Snow. 2004. “The Challenges Facing
Landlocked Developing Countries.” Journal of Human Development 5:1 (March) 31-68.
Fedderke, J.W.; P. Perkins; J.M. Luiz. 2006. “Infrastructural Investment in Long-run Economic
Growth: South Africa 1875-2001.” World Development 34:6, 1037-59.
Fernald, John G. 1999. “Roads to Prosperity? Assessing the Link between Public Capital and
Productivity.” American Economic Review89:3 (June) 619-38.
Flandreau, Marc; Clemens Jobst. 2005. “The Ties That Divide: A Network Analysis Of The
International Monetary System, 1890-1910.” Kennedy School of Government.
www.wcfia.harvard.edu/seminars/pegroup/FlandreauJobst.pdf.
Frankel, Jeffrey A.; David Romer. 1999. “Does Trade Cause Growth?” American Economic Review
89:3, 379-98.
Galor, Oded; David N. Weil. 1999. “From Malthusian Stagnation to Modern Growth.” American
Economic Review 89:2 (May) 150-4.
Gallup, John Luke; Jeffrey D. Sachs; Andrew Mellinger. 1999. “Geography and Economic
Development.” International Regional Science Review 22:2 (August) 179-232.
Gerring, John; Strom C. Thacker; Ruben Enikolopof; Julian Arevalo. 2008. “Public Health
Performance: A Model-Based Approach.”
Gibson, John; Scott Rozelle. 2003. “Poverty and Access to Roads in Papua New Guinea.” Economic
Development and Cultural Change 52:1, 159-85.
Glaeser, Edward L.; Florencio Lopez-de-Silanes; Andrei Shleifer. 2004. “Do Institutions Cause
Growth?” Journal of Economic Growth 9, 271-303.
Goesling, Brian; Glenn Firebaugh. 2000. “The Trend in International Health Inequality.” Population
and Development Review 30:1 (March) 131-46.
Goodfriend, Marvin; John McDermott. 1995. “Early Development.” American Economic Review 85:1
(March) 116-33.
Gramlich, Edward M. 1994. “Infrastructure Investment: A Review Essay.” Journal of Economic
Literature 32:3 (September) 1176-96.
Harley, C. Knick. 1988. “Ocean Freight Rates, and Productivity, 1740-1913: The Primacy of
Mechanical Invention Reaffirmed.” Journal of Economic History 48, 851-76.
Hibbs, Douglas A.; Ola Olsson; Jared Diamond. 2004. “Geography, Biogeography and Why Some
Countries are Rich and Others Poor.” Proceedings of the National Academy of Sciences 101, 3715-20.
Hidalgo, C.A.; B. Klinger; A.-L. Barabasi; R. Hausmann. 2007. “The Product Space Conditions the
Development of Nations.” Science 317 (July 27) 482-7.
Hojman, Daniel; Adam Szeidl. 2006. “Core and Periphery in Endogenous Networks.” Faculty
Research Working Papers Series, RWP06-022, Harvard University, Kennedy School of
Government.
14
Howe, John; Peter Richards (eds). 1984. Rural Roads and Poverty Alleviation. London: International
Labour Organization.
Hummels, David. 2007. “Transportation Costs and International Trade in the Second Era of
Globalization.” Journal of Economic Perspectives 21:3 (Summer) 131-54.
Iyigun, Murat. 2007. “Monotheism (From a Sociopolitical and Economic Perspective).”
Jacoby, Hanan G. 2000. “Access to Markets and the Benefits of Rural Roads.” The Economic Journal
110:465 (July) 713-37.
Jones, Charles I. 2001. “Was an Industrial Revolution Inevitable? Economic Growth over the very
Long Run.” Advances in Macroeconomics 1:2.
Jones, Eric L. 1981. The European Miracle: Environments, Economics and Geopolitics in the
History of Europe and Asia, 2d ed. Cambridge: Cambridge University Press.
Kali, Raja; Javier Reyes. 2006. “The Architecture of Globalization: A Network Approach to
International Economic Integration.” Journal of International Business Studies (forthcoming).
Keller, Wolfgang. 2004. “International Technology Diffusion.” Journal of Economic Literature xlii
(September) 752-82.
Kessides, Christine. 1993. “The Contributions of Infrastructure to Economic Development: A
Review of Experience and Policy Implications.” World Bank Discussion Paper #213.
Washington: World Bank.
Kim, Byoungki. 2006. “Infrastructure Development for the Economic Development in Developing
Countries: Lessons from Korea and Japan.” GSICS Working Paper Series. Graduate School of
International Cooperation Studies, Kobe University.
Kim, Sangmoon; Eui-Hang Shin. 2002. “A Longitudinal Analysis of Globalization and
Regionalization in International Trade: A Social Network Approach.” Social Forces 81:2
(December) 445-68.
Kremer, Michael. 1993. “Population Growth and Technological Change: One Million B.C. to 1990.”
Quarterly Journal of Economics 108:3 (August) 681-716.
Krugman, Paul; Anthony J. Venables. 1995. “Globalization and the Inequality of Nations.” The
Quarterly Journal of Economics 110:4 (November) 857-80.
Landes, David S. 1969. The Unbound Prometheus: Technological Change and Industrial Development in Western
Europe from 1750 to the Present. Cambridge: Cambridge University Press.
Landes, David S. 1999. The Wealth and Poverty of Nations: Why Some Are So Rich and Some So
Poor. New York: W.W. Norton.
Landes, David S. 2006. “Why Europe and the West? Why Not China?” Journal of Economic
Perspectives 20:2 (Spring) 3-22.
Levine, Ruth (ed). 2005. Millions Saved: Proven Successes in Global Health. Washington: Center for
Global Development.
Limao, Nuno; Anthony J. Venables. 2000. “Infrastructure, Geographical Disadvantage, Transport
Costs and Trade.”
Mellinger, Andrew D.; Jeffrey D. Sachs. 1999. “Climate, Water Navigability, and Economic
Development.” CID Working Paper No. 24 (September).
Mellinger, Andrew D.; Jeffrey D. Sachs; John Gallup. 2000. “Climate, Coastal Proximity, and
Development.” In Gordon L. Clark, Meric S. Gertler, & Maryann P. Feldman (eds), The Oxford
Handbook of Economic Geography (Oxford: Oxford University Press) 169-94.
Miguel, Edward; Michael Kremer. 2000. “Worms: Identifying Impacts on Education and Health in
the Presence of Treatment Externalities.” Econmetrica 72:1 (January) 193-217.
Mokyr, Joel. 1992. The Lever of Riches: Technological Creativity and Economic Progress. Oxford: Oxford
University Press.
15
Mu, Ren; Dominique van de Walle. 2007. “Rural Roads and Poor Area Development in Vietnam.”
Policy Research Working Paper 4340. World Bank.
Munnell, Alicia H. 1992. “Policy Watch: Infrastructure Investment and Economic Growth.” Journal
of Economic Perspectives 6:4 (Autumn) 189-98.
Neumayer, Eric. 2003. “Beyond Income: Convergence in Living Standards, Big Time.” Structural
Change and Economic Dynamics 14:3 (September) 275-96.
North, Douglass C.; Robert Paul Thomas. 1973. The Rise of the Western World. Cambridge: Cambridge
University Press.
Nunn, Nathan; Diego Puga. 2007. “Ruggedness: The Blessing of Bad Geography in Africa.”
Unpublished manuscript.
Olsson, Ola. 2003. “Geography and Institutions: A Review of Plausible and Implausible
Linkages.” Working Papers in Economics No. 106.
Olsson, Ola; Douglas A. Hibbs, Jr. 2005. “Biogeography and Long-Run Economic Development.”
European Economic Review 49, 909-38.
Pomeranz, Kenneth. 2000. The Great Divergence: Euyrope, China, and the Making of the Modern World
Economy. Princeton: Princeton University Press.
Prebisch, Raul. 1950. “The Economic Development of Latin America and Its Principal
Problems.” Staff report for the United Nations. Lake Success, NY: United Nations.
Pritchett, Lant H.; Lawrence H. Summers. 1996. “Wealthier is Healthier.” Journal of Human Resources
31:4 (Fall) 841-868.
Putnam, Robert D. 1993. Making Democracy Work: Civic Traditions in Modern Italy. Princeton:
Princeton University Press.
Ravallion, Martin. 2005. “Externalities in Rural Development: Evidence for China.” In Ravi Kanbur,
Anthony J. Venables (eds), Spatial Inequality and Development (Oxford: Oxford University Press).
Ravallion, Martin; Quentin Wodon. 1999. “Poor Areas, or only Poor People?” Journal of Regional
Science 39:4, 689-711.
Ricardo, David. 1821. On the Principles of Political Economy and Taxation.
Sabandar, William. 2007. “Transport and the Rural Economy: Institutions and Institutional Change
in Ambeso Village, Indonesia.” Asia Pacific Viewpoint 48:2 (August) 200-18.
Sachs, Jeffrey D.; Andrew Warner. 1997. “Fundamental Sources of Long-run Growth.” American
Economic Review 87:2, 184-8.
Snyder, David; Edward L. Kick. 1979. “Structural Position in the World System and Economic
Growth, 1955-1970: A Multiple-Network Analysis of Transnational Interactions.” American
Journal of Sociology 84:5 (March) 1096-1126.
Spolaore, Enrico, Romain Wacziarg. 2006. “The Diffusion of Development.” NBER Working Paper
Series # 12153.
van de Walle, Dominique. 2002. “Choosing Rural Road Investments to Help Reduce Poverty.”
World Development 30:4,
Van Rossem, Ronan. 1996. “The World System Paradigm as General Theory of Development: A
Cross-National Test.” American Sociological Review 61:3 (June) 508-27.
Wallerstein, Immanuel. 1974. The Modern World-System. Capitalist Agriculture and the Origins
of the European World Economy in the Sixteenth Century. New York: Academic Press.
Wasserman, Stanley; Katherine Faust. 1994. Social Network Analysis: Methods and Applications.
Cambridge: Cambridge University Press.
Woolcock, Michael. 1998. “Social Capital and Economic Development: Toward a Theoretical
Synthesis and Policy Framework.” Theory and Society 27:2, 151-208.
16
World Bank. 1994. World Development Report 1994: Infrastructure for Development. Washington: World
Bank.
Yanikkaya, Halit. 2003. “Trade Openness and Economic Growth: A Cross-Country Empirical
Investigation.” Journal of Development Economics 72, 57 – 89.
17