Trade and the flag:
integration and conflict in 19th and
early 20th century
deglobalization
Chris
Chase-Dunn, Anders Carlson, Chris Schmitt, Shoon Lio,
Richard
Niemeyer and Robert A. Hanneman
Institute for Research on World-Systems
University
of California-Riverside,
To be presented at the Section on Peace, War, and Social Conflict Roundtable, Fri, Aug 11 - 2:30pm - 3:25pm, Annual meeting of the American Sociological Association, Montreal. Draft v.8-9-06; 7470 words
IROWS
Working Paper #19 http://irows.ucr.edu/papers/irows19/irows19.htm
Abstract:
The density and contours of
networks of transnational and international economic integration are
hypothesized by many theorists to be causally related to the patterns of
cooperation and conflict. [1]
The usual notion is that trade creates ties of symmetrical interdependence,
which are likely to inhibit conflict. We seek to test this hypothesis in the
19th and early 20th century run-up to World War I. We examine the
relationship between the structure of conflict and the contours of trade ties
during the 19th century wave of globalization and deglobalization. How were the international trade ties related
to the patterns of conflict and alliance that emerged during World War I?
Waves of Globalization
Over the past few
decades, there has been a surge of interest in the relationship between
globalization and political conflict in the interstate system. Most of the theorists
of the global capitalism school contend that beginning with the 1960s and
1970s, the world of national economies became transformed into a transnational
and global political economy (e.g. Sklair 2001). Scholars using the world-systems perspective
contend that the world-system of capitalism has been importantly transnational
for hundreds of years and that globalization in the sense of the expansion and
intensification of large-scale intercontinental interaction networks is both an
upward trend and a cycle. There were earlier periods of rapid globalization
that were followed by periods of deglobalization in which large-scale
interactions diminished. Keynesian
national development (the global New Deal) was the predominant strategy of the
development project led by the hegemonic
Neoliberalism was
the political ideology that became hegemonic in the 1980s because the competing
core countries -
Yet, in contrast
to the global capitalism school, we argue that the old system of national
states still exists and that something like the current wave of globalization
had happened before during the decline of British hegemony in the late 19th and early 20th century.
Studies of trade globalization – the ratio of international trade to the world
GDP-- show that there was a high peak in
the 1880s, then a decline until 1900, then another small rise, and a crash in
1929, and then a rise after World War II to the present,
which is somewhat higher than the peak in 1880, but not extremely higher (see
Figure 1). Investment globalization probably followed a similar trajectory
(Chase-Dunn, et al 2002)
Figure 1: Waves of trade globalization (Chase-Dunn, Kawano and
Brewer 2000)
Global Elite Networks and International
Trade Links
This paper is
part of a larger project, the purpose of which is to study the contours of global
elite and international integration since 1840 and to study the relationship
between these contours of connection and the patterns of conflict that emerged
over the same time period. There have
been a significant number of theoretical and empirical works by political
scientists and sociologists that examine the effects of economic
interdependence and international conflict (McMillan 1997; Barbieri and
Schneider 1999; Barbieri 2002; Rosecrance and Thompson 2003; Maoz 2004, Maoz et al 2006, forthcoming). Various
liberal theories of globalization argue that economic integration should
decrease international conflict. [3]We
observe that these approaches should distinguish between horizontal connections
(of equality) and vertical connections (power-dependency relations). The latter
may be quite likely to produce conflict (Barbieri 2002; Rosecrance and Thompson
2003).
Contra these
perspectives, many observers have noted that interdependent connections have
not served to prevent major conflicts in the modern international system
(Thompson and Tucker 1997; Rosecrance and Thompson 2003). Both
Our larger project
consists of two parts. In the first we are using a world historical perspective
to examine the links between elite individuals, families and organizations
within each country with those same actors in other countries. This involves a
close reading of the histories of each country with attention to connections
with other countries (see Reifer and Chase-Dunn 2003; Barr et al 2006). The second part of our project (discussed here) uses
quantitative data on the interactions among nation-states to trace the changing
patterns of network connections since 1880.
This enables us to use the national network patterns to place the
information from our studies of elites in a world historical context, and to
study the congruence or lack thereof, between different kinds of international
connections. We also intend to examine the relationships between international
network structures and the patterns of conflictive relations that were so
evident in the first half of the twentieth century. We know that the
international system bifurcated into Allies and Axis states in the World Wars. Were
these conflict-alliance bloc structures related to the trade network? Did the
network of global trade become more factional in the years prior to the
outbreak of world war? And did these factions correspond with the emergent
conflict factions?
We also will eventually
use network data to compare the overall magnitude of global integration in the
nineteenth century with the magnitude and forms of integration that have
emerged since World War II.[5] The
question of global magnitudes is important because many students of
globalization have assumed that the high degree of contemporary integration of
the global capitalist class will prevent the emergence of future interimperial
rivalry and war among core states. But if there was a similarly high level of
global elite integration in the late nineteenth century this assumption may be
brought into question.
In this paper we analyze
mainly international trade relations, but international financial links are another
important dimension of global economic networks that we plan to empirically
examine in the future. Imports and exports of goods and services are much
easier to get comparable information on than flows of investment, especially
for the nineteenth century. Ideally we would like to differentiate trade flows
into goods that are more strategic and profitable vs. those that are less so.
But that is not possible on a sufficient scale for the nineteenth century.
Transnational Relations and State-centrism
The use of data on nation-states is defensible on both
theoretical and practical grounds, and should not expose us to the slings and
barbs of those who would accuse us of state-centrism. Firstly, national states
have been, and still are, important organizations within the world-system.
Transnationalism has not just arisen since 1980. There have been waves of
transnationalism and waves of nationalism since the chartered companies of the
seventeenth century organized production and distribution on a global scale. The contemporary transnational corporations
undoubtedly organize a greater portion of the total world economy than the 17th
century chartered companies did. But then and now, national states were and
remain important players on the global stage.
We may say this
without denying the perspective developed by William I. Robinson (2004) and
others on the emergence of a transnational capitalist state that reconfigures
existing national states (and international organizations) as its instruments.
Indeed, we see the emergence of a transnational state, not just in the period
since the 1980s, but since the Concert of Europe that was
The practical
reason for using data on national states is that they are the only data that
are available for most of the regions of the system over the time period that
we seek to study. As with all secondary data
analyses, we need to be chary about the ways in which the structuring of the
data by its original providers may distort our results.
Methods for Our Analysis
We adopt the
generalized strategy of measurement error modeling that is part of the
structural equations tradition. This means that instead of trying to pick the
best single empirical indicator of an underlying concept or variable, we want
to use several proxy indicators and to model the relationships among the
proxies as well as using them to estimate the true underlying values of the
variables. In practice we may not have enough data to be able to actually
employ the techniques of structural equations modeling of measurement error.
But we shall use the generalized logic of gathering multiple proxy indicators
whenever this is possible.
Because the data
are less complete in the early decades, we have a growing population of nodes
as we get closer to the present. This, and the actual changes that occurred in
country boundaries over the period studied (e.g. the break-up of the Ottoman
and Austro-Hungarian empires, etc.), mean that we have a changing set of nodes
in the network. This makes it difficult to know whether observed changes were
due to real changes in the pattern of trade ties or to the inclusion of nodes
that were formerly not included because of missing data. One approach to this
problem that we have used in earlier research is to study constant groups over time. If we find similar trends between the
constant groups and the networks that are adding (or deleting) nodes we can
infer that observed changes are not due to changes in the compared units.
Variable Construction
Trade Network Data
Much of the late 19th
century and early 20th century trade data are reported in the
country’s domestic currency, which makes cross-national comparison
impossible. There are several possible
ways to deal with this problem. One is to convert all the country currency
values into a single currency such as the pound sterling or the U.S. dollar
using currency market exchange rates.[6] There
are a number of known problems with this approach. Currency market exchange rates are set by the
competitive buying and selling of currencies during some periods, but in other
periods the exchange rates have been set by international agreements. Between
the Bretton Woods conference in 1944 and the early 1970s the U.S. dollar was
pegged to a fictitious gold standard, and other currencies were pegged to the
dollar. These regulated exchange rates can still be used to change country
currencies into dollars, but this conversion reflects a worldwide agreement to
regulate currency markets rather than a world market for money. In 1974 the
dollar and other currencies were freed to exchange in world money markets.
Another problem
is that market exchange rates reflect the activities of large currency traders,
rather than just the daily conversions of currencies carried out by people who
need to change money. The actions of currency traders are intended to make
profits by buying and selling money, and this activity does not necessarily
reflect the value of the goods and services that national economies produce.
This is why economists have tried to devise a better method for converting
currencies into a single comparable metric that is based on purchasing power in
different countries (Kravis Heston and Summers 1982). These so-called purchasing
power parity (PPP) conversion ratios are not available for the 19th
century and the whole approach has been savaged by critics (e.g. Korzeniewicz,
Stach and Patil 2004).
Another method of
making country currency values comparable is to compute a percentage using a
denominator in the same metric units (country currencies). We have the total value of exports and
imports for each country in country currency, so we could compute the
percentage of the country’s trade with a particular other trade partner. This puts the numbers into a comparable
metric: percentages. But this is not a good solution to the problem for our
purposes. It does eliminate the need to use exchange rates, but at the cost of
computing a variable that will not be useful for our purpose of examining the relative
importance of a particular trade link in the context of the larger world trade
network. Knowing that the imports of
(did we use imports, exports, both or what?)
Construction of Conflict Dyads
The
countries considered in this analysis are those that fought in World War I, those
core countries of
To quantify
the intensity of conflict between two states (dyads), we relied upon three
separate indicators: 1) the Correlates of War data set compiled by Singer and
Small measuring the number of battle deaths experienced by each country in the
WWI, and 2) the Barbieri conflict data set consisting of two ordinal measures of conflict during WWI: one
representing the level of aggression country A displayed towards country B, and
the other representing the level of aggression country B displayed towards
country A.
Interestingly, each of these data sets exhibited complimentary
weaknesses. The Correlates of War data
is useful after a country goes to war because it demonstrates how
“intensely” the country was committed to fighting as a function of the number
of its dead, but the data says nothing of the level of conflict between
countries before they go to war.
In a similar fashion, the Barbieri data does an excellent job of
demonstrating the “ramping up” processes leading up to WW1, but after the fact
it is useless in distinguishing various levels of commitment to the war once it
has begun. Further, both data sets in
isolation demonstrated very high skewness and kurtosis, making interpretation
of correlation coefficients problematic.
To
remedy both of these problems we constructed a standardized index of conflict
intensity that combined all three measures.
This was carried out by transforming the “raw” values of each data set
into standardized scores using SPSS, and summing the result. At this point we realized that by
constructing the index in this fashion, we had inadvertently reduced the
contribution of the Correlates of War data.
What was once a very large difference in intensity between “a
militarized shared border,” and, “a combined war dead of over one million
soldiers,” had now been reduced to a one or two point index difference. Also, one country’s decision to enter into the
war as an ally of another is an indicator level of (low) conflict intensity
that was not being taken into account. So we modified our conflict indicator by
doubling the weight of the contribution of the battle deaths, and also coding
for whether or not a state was an ally of another.
So as to make neutrality during WWI represent zero
conflict between a pair of states, the index was scaled so that a value of –3.16
equated to war ally, 0 equated to neutral and a value of 16 equated to the highest level of conflict intensity. We do not have a measure that takes into
account various degrees of “war ally,” so the index jumps from –3.14 to 0, and
then ramps up incrementally to more than 16.
It should also be noted that although this final measure of conflict
does still display minor skewness and kurtosis, it is by far the best in this
regard when compared to the Correlates of War and Barbieri indicators (see Table
1).
|
|
N |
Minimum |
Maximum |
Mean |
Std. Deviation |
Skewness |
Kurtosis |
|
|
|
|
|
|
|
|
|
| Level
of Conflict |
552 |
-3.14 |
16.09 |
.6315 |
3.14 |
2.023 |
4.8 |
| COW
|
552 |
0 |
3500000 |
208681.88 |
602579.10 |
3.143 |
9.584 |
| Barbieri Conflict Measure |
552 |
0 |
20 |
1.04 |
3.998 |
3.747 |
12.386 |
Dyadic Correlations between Conflict and Trade in 1880-1913
After constructing the conflict index, a Pearson r
test was used to determine the correlation between levels of trade for eight
time periods leading up to World War I and the intensity of conflict between
combatants during the war. The results
are shown in Table 2:
|
|
Intensity
of Conflict |
|
Amount
of Trade 1913 2-tailed Significance |
.231 .000*** |
|
Amount
of Trade 1912 2-tailed Significance |
.232 .000*** |
|
Amount
of Trade 1911 2-tailed Significance |
.233 .000*** |
|
Amount
of Trade 1910 2-tailed Significance |
.235 .000*** |
|
Amount
of Trade 1905 2-tailed Significance |
.173 .000*** |
|
Amount
of Trade 1900 2-tailed Significance |
.103 .016* |
|
Amount
of Trade 1895 2-tailed Significance |
.110** .010 |
|
Amount
of Trade 1890 2-tailed Significance |
.101* .017 |
|
Amount
of Trade 1885 2-tailed Significance |
.088* .040 |
|
Amount
of Trade 1880 2-tailed Significance |
.044 .299 |
|
*** Significant at .001 level ** Significant
at .01 level *
Significant at .05 level |
|
Table 2: Dyadic Correlations between Conflict and Trade in 1880-1913
As indicated
by the table, the amount of imports one country received from another had a
significant positive correlation with the level of conflict experienced within
the dyad during WWI. This was the case
in each of the above years, except 1880, which was also positive.
Controlling for Size:
Partial Correlation between Trade and Conflict
Given that a portion of the intensity of conflict index is
measured in battle deaths, it is advisable to control for the size of the
population of the countries involved.
Population dyads were created as a control variable using 1913
population data compiled by the
Correlates of War Project and the
|
Partial Correlation Controlling for Population |
|
|
|
Intensity
of Conflict |
|
Amount
of Trade 1913 2-tailed Significance |
.2289 .000*** |
|
Amount
of Trade 1912 2-tailed Significance |
.2302 .000*** |
|
Amount
of Trade 1911 2-tailed Significance |
.2304 .000*** |
|
Amount
of Trade 1910 2-tailed Significance |
.2327 .000*** |
|
Amount
of Trade 1905 2-tailed Significance |
.1703 .000*** |
|
Amount
of Trade 1900 2-tailed Significance |
.1012 .018* |
|
Amount
of Trade 1895 2-tailed Significance |
.1081 .011 |
|
Amount
of Trade 1890 2-tailed Significance |
.0994* .020 |
|
Amount
of Trade 1885 2-tailed Significance |
.0864** .043 |
|
Amount
of Trade 1880 2-tailed Significance |
.0425 .319 |
|
*** Significant at .001 level ** Significant
at .01 level *
Significant at .05 level |
|
Table 3: Dyadic Correlations between Conflict and Trade in 1880-1913 controlling
for population size
Although controlling for population size reduced the strength of
the positive correlation between level of trade and conflict by small amount,
the relationship once again remains significant in all years except 1880. Thus
our new analysis of dyadic correlations using an improved measure of conflict
confirms earlier results by Barbieri (2002) that show a significant positive
relationship between trade connections and the emergence of conflict. But does
this relationship hold up when we examine the whole network of interaction.
Analysis of dyads cannot take account of indirect connections but formal
network analysis can examine the structure of the whole system and look for
cliques or factions in the system. Are there strong subgroups in the trade
structure and, if so, do these correspond with the conflict factions that
emerged in World War I?
Network Analysis of Trade and Conflict
We used UCINet to produce comparable square matrices of our conflict and trade datasets for purposes of formal network analysis. A square matrix is produced by UCINet for purposes of formal network analysis. Network analysis is superior to dyadic correlation analysis because it allows the whole structure of a network to be analyzed including all the direct and indirect links and non-links. This makes it possible to identify cliques or factions within a network and to examine the centrality or peripherality of network nodes. The nodes in this analysis are countries.
The
conflict matrix contains the values for each pair of countries computed for
our level of conflict indicator described above. This is then used to produce
Figure 2 by means of specifying a cutting point in the distribution of dyad
values. For Figure 2 we used ________________________________.
Figure 2: Level of Conflict Network for World War I. (Country names are Correlates of War abbreviations.)
Compare Figure 2 with the list of the
blocks in World War I in Table 4.
|
Allies (Entente) |
Central Powers |
Neutrals |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Balkans (YUG) |
|
|
Table 4: Conflict Blocs in World War I
The
Central Powers in the middle of Figure 2 are not linked by conflict ties with
one another (except for something between
(insert the graphic of trade network in
1880 here and compare it with the next figure)
Figure
3 shows the network structure of trade in 1913 just before the outbreak of
World War I. The cutting point we used for the trade network graphic is
______________.