Appendix
to "Trade Globalization since 1795: waves of integration in the world-system"
American Sociological Review February 2000, Millennial Symposium.
Christopher Chase-Dunn
Yukio Kawano
Benjamin Brewer
Department of Sociology
Johns Hopkins University
Baltimore, MD. 21218 USA
chriscd@jhu.edu
v. 8-31-99

Table of Contents

Link to Data: tradglob.xls

Figure A1: Number of Countries with Data on Openness, 1795-1995

Figure A2: World Exports Per Capita in constant 1990 U.S. dollars.

Section 1: Weighting by GDP

Equation 3: Weighted Average Openness (GDP)

Figure A3: Average Openness Trade Globalization Weighted by GDP compared with Population Weighting

Section 2: Checking the New Measure of Trade Globalization

Figure A4: World Totals and Average Openness Measures, 1950-1992

Figure A5: Average Openness of Constant Groups of Countries (5 year moving averages)

Table A1: Pearson’s r correlation coefficients of subgroup scores with our estimate of world trade globalization that uses all data.

Figure A6: Average Unweighted Openness Scores for Core , Periphery and Semiperiphery Groups, 1908-1995

Table A2: Country list showing categorization of Core, Peripheral and Semiperipheral Countries

Table A3: T-test of mean differences between Core and Periphery Average Openness Scores (Unweighted), 1908-1995

Table A4: T-test of mean differences between Core and Periphery Average Openness Scores (Unweighted), 1908-1949

Figure A7: Imports/GDP for the United States

Figure A8: Average Openness for the U.S., Great Britain and France

Figure A9: Average Openness for the Group of 7.

Figure A10: Average Openness for the Group of 14.

Figure A11: Average Openness for the group of 24.

Figure A12: Core Country Shares of World GDP

Figure A13: Trade and Investment Globalization


The number of countries for which we have data on trade openness that were used to estimate world trade globalization is graphed in Figure A1.

Figure A1: Number of Countries with Data on Openness, 1795-1995

Figure A2: World Exports Per Capita in constant 1990 U.S. dollars.

Source: Maddison (1995: 226, 239)



Section 1: Weighting by GDP

A reviewer suggested that economic size (GDP) might be a better weight than population size for the purpose of estimating a characteristic of the world economy. Weighting by economic size requires the use of country GDPs that have been converted into constant U.S. dollars, and so this reintroduces the problematic assumptions about exchange and inflation rates. Also the GDP figures in constant U.S. dollars from Maddison (1995:180-192) are available for only 56 countries, and for time points that are widely spread in the 19th century. Nevertheless, we interpolated these to produce yearly estimates and used these to calculate trade globalization weighted by GDP using the same weighting method used with population, i.e.

Equation 3

Note that in Equation 3 Imports and the country GDPs used to calculate the openness ratios are in local country currencies, while the GDP figures used to calculate the weights for each country are in U.S. dollars. Exchange and inflation assumptions may not be quite as problematic here since we are only using the exchange and inflation rates to calculate weights for the openness scores, not the openness ratios themselves. This weighting using economic size resulted in a trajectory of estimated trade globalization that is substantially similar to that produced by using population weights. The Pearson’s r correlation coefficient for these two series is .91. Figure A3 displays the trajectory of trade globalization weighted by GDP and compares it with the same measure weighted by population.

Figure A3: Average Openness Trade Globalization Weighted by GDP (56 countries) compared with Population Weighting

 

Section 2: Checking the new measure of Trade Globalization

The biggest contribution reported in this paper is our new and improved measure of trade globalization, and thus before we interpret our results we will examine the new measure more closely. There are two techniques we can use to check that our "sample" of countries with available data is providing a reliable estimate of world-wide trade globalization. The first is to compare our estimated level of trade globalization based on Average Openness with the World Total approach based on Maddison’s (1995) estimates. For the period after 1950 Maddison has yearly data on GDP and the problems of the World Totals approach are probably reduced, because exchange and inflation rates are more reliably known. Thus we can use Maddison’s (1995: 227,239) data to see how our Average Openness measures compare after 1950.

Figure A4: World Totals and Average Openness Measures, 1950-1992

Figure A4 indicates that the Average Openness measures may overestimate the levels of trade globalization. Especially the unweighted Average Openness does this, probably because the small countries with high levels of openness are over-weighted and raise the level of the estimates. The weighted Average Openness is a much closer estimate of the real levels of trade globalization in this period.

The other big difference between the two series is in 1913 when the unweighted series spikes to a high of .32. This is because Switzerland and Belgium enter the data for that year only, with high openness values (.48 and .77), and also because of the very high openness scores of the Netherlands during this period (1.39). The Netherlands has always been a big trading state, but this very high score could be some kind of error. The weighted series does not show such a large spike because these are all relatively small countries.

Another technique we can use to examine the errors due to missing data is to select different subgroups of countries and hold these groups constant over time and then to compare the groups to see if they are revealing similar temporal sequences and similar levels of openness. Our overall measure of average openness trade globalization gradually adds cases, and so the patterns we observe could be caused by the addition of cases. If we find that constant subgroups exhibit patterns similar to those found for the whole data set we may be comforted regarding the proposition that our restricted sample earlier in time is not a bad estimate of the true world level of trade globalization, though this will not at all be certain. The openness values in the constant groups are weighted by the ratio of the country population to the average population of the group at each time point.

We can determine the effects of adding cases by comparing the overall measure with groups of countries in which the cases are held constant over time, shown in Figure A5.

Figure A5: Average Openness of Constant Groups of Countries (5 year moving averages)

Figure A5 graphs the weighted Average Openness values for six groups of countries, with the groups held constant over time so that changing country composition does not affect the averages. The trajectories get shorter as we add countries.

The first "group" is the United States, the only country for which we have data for the whole period from 1795 to 1995. The second group is composed of the U.S., the United Kingdom and France with average scores beginning in 1830. The third group, beginning in 1861, adds to these Australia, Denmark, Italy and Sweden (seven countries). The fourth group (fourteen countries), beginning in 1905, adds Cuba, Spain, India, Japan, Mexico, the Netherlands and Taiwan. The fifth group (24 countries) begins in 1927 and adds Austria, Canada, Colombia, Greece, Guatemala, Honduras, Hungary, Indonesia, South Africa and Zimbabwe. The sixth group (50 countries), begins in 1950 and the seventh group with 89 countries begins in 1965.

Inspection of Figure A5 clarifies some aspects of Figure 3 in our paper and supports the idea that average levels of openness of a subgroup of countries can be used as a reasonable proxy for both the level of world trade globalization and for periods of rise and fall in that level. Figure A5 shows that, except for the U.S., the other groups display generally similar levels of average trade openness and these levels go up and down rather synchronously.

Additional support for our overall measure is the table of correlation coefficients between the group scores and the values estimated using all the cases. These are shown in Table A1.
 
Country

Groups

USA US/GB

France

Seven Fourteen Twenty-four Fifty Eighty-nine
All 148 Countries .51 .84 .85 .73 .92 .94 .74

Table A1: Pearson’s r correlation coefficients of subgroup scores with our estimate of world trade globalization that uses all data.

While the subgroups vary as to how well they are correlated with our overall measure, they are all fairly well correlated with it, except for the United States -- a deviant case.

The erratic fluctuations before 1830 show the fledgling United States of America fighting its way through a world war in which it was allied with the losing side. The Continental blockade, sometimes breached, and poor import statistics probably account for the wildness of the measure of U.S. trade openness in this period. We include these data mainly to display the slim reed that is our window on world trade globalization before 1830. Indeed, the rest of the U.S. performance until about 1960 shows that the United States by itself continued to be a poor reflector of world trade globalization. It was not until the 1960s that the United States experienced increased openness of its trade to the world division of labor.

The trajectory of U.S. trade openness may not be a good proxy for the world economy as a whole, but it is an important window on what happened within the U.S. and its relationship with the larger world-system. While the rest of the world was going through at least one, and possibly two waves of trade globalization between 1830 and 1929, the United States enjoyed a low, and mainly declining, level of trade dependence. This reflects both tariff protectionism and the relatively fast rate of growth of U.S. GDP during this period of territorial expansion, as well as rapid population growth, industrialization and upward mobility into the core of the world-system. U.S. imports did grow mightily, but the domestic economy grew even faster.

This amazing performance was the outcome of internal and international struggles among classes, different sectors within the same classes, and national states. Indeed, it has been argued elsewhere that the U.S. Civil War was mainly a struggle over how the U.S. would be inserted into the larger core/periphery hierarchy (Chase-Dunn 1980). The struggle over tariff policy between 1816 and the Civil War showed how the southern exporters of peripheral agricultural products had political and economic interests that were quite divergent from northern manufacturers. Was the U.S. to continue as an exporter of agricultural raw materials, as the other new states in Latin America did, or was it to use the power of the Federal state to move up the value-added hierarchy into the core of the world-system? The victory of the north in the Civil War meant a consistent policy of trade protectionism to promote import substitution industrialization, a policy that lasted until after World War II. Thus the U.S. success is a poor example for those who want to argue that free trade is a central pillar of economic development.

But our focus here is not the trade history of the individual countries, and neither do we need to assume that all the countries have the same trajectories of trade openness. Rather we are studying the whole system, which is composed of diverse parts with different, but not unrelated, histories.

Another approach for evaluating our Average Openness measure is to examine systematic differences among countries that may be affecting our estimates for earlier years. We have already mentioned that core and peripheral countries often differ in terms of their levels of openness or trade dependence. Peripheral countries tend to be smaller and more dependent on imports and exports, although there are also small core countries with high levels of openness (e.g. Switzerland, the Netherlands). [Openness is not itself a good measure of dependency. What matters in the hierarchical division of labor in the world economy is whether the national exports are high or low in the value-added hierarchy. Little Switzerland, classically exporting fine watches, has a very different sort of openness than Honduras, which exports bananas. "Trade composition" is the concept that captures the nature of exports and imports and this notion has been studied for the whole world-system using network analysis by Smith and White (1993).]

One big problem with our measure of trade globalization is that we have data on few non-core countries early on, so an important piece of the world-system is missing, and this could be biasing our estimation of the level of world trade globalization downward. This is cause for concern because one of the questions we want to answer is whether or not there is a real upward trend, in addition to the obvious cycles. It is possible that the high level indicated for recent decades might be due to the addition of more data on peripheral and semiperipheral countries rather than a real increase in the level of world-wide trade globalization.

Figure A6: Average Unweighted Openness Scores for Core , Periphery and Semiperiphery Groups, 1908-1995

We have divided our list of countries into core, periphery and semiperiphery groups for the period since 1900 following the multivariate approach formulated by Terlouw (1993) (see lists of countries in Table A2 below). Figures A6 plots the unweighted values of these groups. We do not have data for all the countries over the whole time period. Indeed some countries appear for a year and then are missing again for more than a decade. This accounts for some of the spikes in Figure A6. It is obvious that the three groups all experience the same waves of trade globalization, but there are also interesting differences. The peripheral group consistently has higher openness scores than the core group. A t-test rejects the hypothesis of no difference between the mean scores of these groups at greater than the .001 level, and this holds true even when we exclude the years after 1950 (See Tables A3 and A4 below). The semiperipheral countries were consistently lower than the core in average openness before 1970, after which time they became rather higher. The semiperipheral group has always been rather a mixed bag of countries that are pursuing different kinds of development strategies.

After 1976, the core countries reached a plateau that they cycled around until the end of the series, whereas the peripheral countries rose to a height greater than ever before and then fell back in the mid ‘80s and then rose again to their highest height in the early 90s when our series ends. The semiperipheral countries continued to rise and then fell precipitously to the level of the core in 1990.

These results support the notion that both core and non-core countries are experiencing changes in trade globalization synchronously, and that up until 1975 these groups had similar levels of openness, with peripheral countries usually a bit higher than core countries. After 1975 we see a divergence. The core countries plateau at level that is higher than the level reached in the earlier waves of globalization, while the peripheral countries continued to rise to a much higher level.

 

Table A2: Country list showing categorization of Core, Peripheral and Semiperipheral Countries
 
Periphery Core Countries Semiperiphery
AFG-Afghanistan KWT-Kuwait USA-United States BRA-Brazil
ALB-Albania KGZ-Kyrgyz Republic GBR Wales CHN-China
DZA-Algeria LAO-Lao PDR FRA-France HKG-Hong Kong
AGO-Angola LVA-Latvia AUS-Australia IRN-Iran, Islamic Rep.
BHR-Bahrain LBN-Lebanon DNK-Denmark ISR-Israel
BGD-Bangladesh LSO-Lesotho ITA-Italy KOR-Korea, Rep.
BRB-Barbados LBR-Liberia SWE-Sweden ZAF-South Africa
BLR-Belarus LBY-Libya ESP-Spain SGP-Singapore
BLZ-Belize LTU-Lithuania JPN-Japan MEX-Mexico
BEN-Benin MAC-Macao NLD-Netherlands ARG-Argentina
BOL-Bolivia MKD-Macedonia, FYR AUT-Austria IND-India
BWA-Botswana MDG-Madagascar CAN-Canada IDN-Indonesia
BGR-Bulgaria MWI-Malawi BEL-Belgium Taiwan
BFA-Burkina Faso MYS-Malaysia CHE-Switzerland
BDI-Burundi MLI-Mali DEU-Germany
KHM-Cambodia MRT-Mauritania NZL-New Zealand
CMR-Cameroon MUS-Mauritius IRL-Ireland
CAF-Central African Republic MDA-Moldova NOR-Norway
TCD-Chad MNG-Mongolia FIN-Finland
CHL-Chile MAR-Morocco PRT-Portugal
COL-Colombia MOZ-Mozambique
COG-Congo MMR-Myanmar
CRI-Costa Rica NAM-Namibia
CIV-Cote d'Ivoire NPL-Nepal
HRV-Croatia NIC-Nicaragua
CUB-Cuba NER-Niger
CYP-Cyprus NGA-Nigeria
CZE-Czech Republic OMN-Oman
DOM-Dominican Republic PAK-Pakistan
ECU-Ecuador PAN-Panama
EGY-Egypt, Arab Rep. PNG-Papua New Guinea
SLV-El Salvador PRY-Paraguay
ERI-Eritrea PER-Peru
EST-Estonia PHL-Philippines
ETH-Ethiopia POL-Poland
RWA-Rwanda PRI-Puerto Rico
SAU-Saudi Arabia ROM-Romania
GAB-Gabon RUS-Russian Federation
GMB-Gambia, The SEN-Senegal
GEO-Georgia SLE-Sierra Leone
GHA-Ghana LKA-Sri Lanka
GRC-Greece SDN-Sudan
GTM-Guatemala SUR-Suriname
GIN-Guinea SYR-Syrian Arab Republic
GNB-Guinea-Bissau TZA-Tanzania
GUY-Guyana THA-Thailand
HTI-Haiti TGO-Togo
HND-Honduras TTO-Trinidad and Tobago
ARE-United Arab Emirates TUN-Tunisia
URY-Uruguay TUR-Turkey
HUN-Hungary UGA-Uganda
IRQ-Iraq VEN-Venezuela
YEM-Yemen, Rep. VNM-Vietnam
YUG-Yugoslavia, FR (Serbia/Montenegro) ZMB-Zambia
ZAR-Zaire ZWE-Zimbabwe
JAM-Jamaica
JOR-Jordan
KAZ-Kazakstan
KEN-Kenya

____________________________________________________________________________________
 
t-Test: Two-Sample Assuming Unequal Variances
core vs. periphery 1908-1995 
Core
Periphery
Mean
0.225598273
0.288585325
Variance
0.004604849
0.009302946
Observations
88
88
Hypothesized Mean Difference
0
df
156
t Stat
-5.010296877
P(T<=t) one-tail
7.28752E-07
t Critical one-tail
2.350489012
P(T<=t) two-tail
1.4575E-06
t Critical two-tail
2.607703209
Table A3: T-test of mean differences between Core and Periphery Average Openness Scores (Unweighted), 1908-1995

 
 
t-Test: Two-Sample Assuming Unequal Variances
core vs. periphery 1908-1949
Core
Periphery
Mean
0.193067921
0.259427199
Variance
0.003638978
0.011485727
Observations
42
42
Hypothesized Mean Difference
0
df
65
t Stat
-3.496897049
P(T<=t) one-tail
0.000427012
t Critical one-tail
2.385095286
P(T<=t) two-tail
0.000854023
t Critical two-tail
2.653614501

Table A4: T-test of mean differences between Core and Periphery Average Openness Scores (Unweighted), 1908-1949

 


 

Figure A7: Imports/GDP for the United States

 

Figure A8: Average Openness for the U.S., Great Britain and France

Figure A9: Average Openness for the Group of 7.

Figure A10: Average Openness for the Group of 14.

Figure A11: Average Openness for the group of 24.

Figure A12: Core Country Shares of World GDP

Source: Maddison (1995:180,227)

 

 

Figure A13: Trade and Investment Globalization