Buscar

KIM, S and SKVORETZ, J

Prévia do material em texto

http://cos.sagepub.com/
Sociology
International Journal of Comparative
 http://cos.sagepub.com/content/54/2/124
The online version of this article can be found at:
 
DOI: 10.1177/0020715212469513
December 2012
 2013 54: 124 originally published online 14International Journal of Comparative Sociology
Sangmoon Kim and John Skvoretz
Structural embeddedness, uncertainty, and international trade
 
 
Published by:
 http://www.sagepublications.com
 can be found at:International Journal of Comparative SociologyAdditional services and information for 
 
 
 
 
 http://cos.sagepub.com/cgi/alertsEmail Alerts: 
 
 http://cos.sagepub.com/subscriptionsSubscriptions: 
 http://www.sagepub.com/journalsReprints.navReprints: 
 
 http://www.sagepub.com/journalsPermissions.navPermissions: 
 
 http://cos.sagepub.com/content/54/2/124.refs.htmlCitations: 
 
 What is This?
 
- Dec 14, 2012OnlineFirst Version of Record 
 
- Jul 11, 2013Version of Record >> 
 at FACULDADE IMED on April 8, 2014cos.sagepub.comDownloaded from at FACULDADE IMED on April 8, 2014cos.sagepub.comDownloaded from 
http://cos.sagepub.com/
http://cos.sagepub.com/
http://cos.sagepub.com/content/54/2/124
http://cos.sagepub.com/content/54/2/124
http://www.sagepublications.com
http://www.sagepublications.com
http://cos.sagepub.com/cgi/alerts
http://cos.sagepub.com/cgi/alerts
http://cos.sagepub.com/subscriptions
http://cos.sagepub.com/subscriptions
http://www.sagepub.com/journalsReprints.nav
http://www.sagepub.com/journalsReprints.nav
http://www.sagepub.com/journalsPermissions.nav
http://www.sagepub.com/journalsPermissions.nav
http://cos.sagepub.com/content/54/2/124.refs.html
http://cos.sagepub.com/content/54/2/124.refs.html
http://cos.sagepub.com/content/54/2/124.full.pdf
http://cos.sagepub.com/content/54/2/124.full.pdf
http://cos.sagepub.com/content/early/2012/12/12/0020715212469513.full.pdf
http://cos.sagepub.com/content/early/2012/12/12/0020715212469513.full.pdf
http://online.sagepub.com/site/sphelp/vorhelp.xhtml
http://online.sagepub.com/site/sphelp/vorhelp.xhtml
http://cos.sagepub.com/
http://cos.sagepub.com/
http://cos.sagepub.com/
http://cos.sagepub.com/
International Journal of 
Comparative Sociology
54(2) 124 –143
© The Author(s) 2012 
Reprints and permissions: 
sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/0020715212469513
cos.sagepub.com
IJ CS
Structural embeddedness, 
uncertainty, and international trade
Sangmoon Kim
University of North Carolina Wilmington, USA
John Skvoretz
University of South Florida, USA
Abstract
We examine the impact of structural embeddedness, operationalized as a third-party effect, on bilateral 
trade of two goods (apparel and grain) with different levels of product differentiation and transactional 
uncertainty. Specifically, we test two competing hypotheses for how trade ties to third parties affect trade 
in a dyad: the balance hypothesis – common third-party contact(s) mediate information flows between 
two otherwise disconnected actors, thereby encouraging direct dyadic interactions between them and the 
structural hole hypothesis – such contact inhibits dyadic trade. Our longitudinal analyses of international 
trade data show that the balance hypothesis is supported in apparel trade (a differentiated good), whereas 
the structural hole hypothesis tends to be supported in grain trade (a homogeneous good). Implications of 
the findings are discussed.
Keywords
Embeddedness, international trade, product differentiation, social network analysis, triadic closure
International trade is a research topic that displays clear disciplinary differences between econom-
ics and sociology. Economists and sociologists have traditionally had different aims and taken 
different methodological approaches. Economists study trade in the framework of economic actors 
interested in utility maximization, and their models typically use nodal/dyadic and economic 
attributes of actors (countries) as independent variables. Criticizing economic and dyadic reduc-
tionism, sociologists examine trade to show social and structural influences on it. For example, 
world-system researchers analyze trade to show dominance and stratification in the global econ-
omy (e.g. Mahutga, 2006; Smith and White, 1992; Snyder and Kick, 1979). They believe that the 
flows and contents of international trade reflect the underlying hierarchical and exploitative struc-
ture of world economy. World-system models, typically utilizing network analytic methods, focus 
Corresponding author:
Sangmoon Kim, Department of Sociology & Criminology, University of North Carolina Wilmington, Wilmington, NC 
28403, USA. 
Email: kims@uncw.edu
469513 COS54210.1177/0020715212469513International Journal of Comparative SociologyKim and Skvoretz
2012
Article
 at FACULDADE IMED on April 8, 2014cos.sagepub.comDownloaded from 
http://cos.sagepub.com/
http://cos.sagepub.com/
Kim and Skvoretz 125
on non-dyadic effects such as hierarchy and structure which refer to tie patterns that transcend the 
dyadic level. The empirical research in this tradition shows that bilateral trade is influenced by the 
positions that trading dyads (countries) take in world-systems: dense trade among the core and 
sparse trade among the periphery.
Researchers also investigate social influences on trade (Gould, 1994; Ingram et al., 2005; 
Rauch, 1999; Zhou, 2011). They often examine social influences by juxtaposing economic trade 
networks with other social networks, such as immigration, membership in intergovernmental 
organizations (IGOs), and other non-economic characteristics of transacting dyads. Then the cor-
relations (i.e. multiplexity) between the two types of networks are considered as indicators of 
social influences. The empirical results generally support the hypothesis that bilateral trade is also 
socially embedded: non-economic similarities within a dyad are positively associated with eco-
nomic trade.
This article investigates a network structural effect on international trade networks of two goods 
(apparel and grain) over an eight-year period (1992–2000). Expanding the recent efforts in this line 
of research (Kim and Skvoretz, 2010; Zhou and Park, forthcoming), we attempt to identify the 
conditions under which triadic structural embeddedness influences bilateral trade. Compared to the 
studies examining social influences (dyadic multiplexity) on trade, our approach is more structural, 
focusing on triadic influences. Compared to world-system research our approach shares the idea 
that the position that a country has within a structure influences its trade with others. But we are 
interested in a more micro-view, namely, the country’s position within triads and in the contingent 
nature of the positional effect.
We situate our research within the concept of structural embeddedness, and build on recent stud-
ies by Kim and Skvoretz (2010) and Zhou and Park (forthcoming). The focus is on how trade 
within a dyad is influenced by the extent that the dyad shares common third-party contacts. The 
previous studies show that trade within a dyad is positively associated to the extent that the two 
countries in the dyad share common contacts. A suggested mechanism behind this triadic closure 
includes, among others, the information diffusion through common contacts. That is, the connec-
tion to common contacts provides actors with valuable information about each other’s market and 
potential business partners, which reduces transaction costs and thereby promotes direct trade. The 
focus of our research is on this mechanism and the extent to which its effects are contingent on 
market type and product traded, since these characteristics affect information uncertainty.
As many researchers point out, a lack of information about trading opportunities is still an 
important barrier to trade despite the development of information and communication technologies 
(Chaney, 2011; Petropoulou, 2008; Rauch and Watson, 2004). For this reason, overseas contacts 
(includingoverseas branches, subsidiaries, representatives, affiliates, and partners) are important 
sources of trade information (Howard and Herremans, 1988; Keegan, 1974; McAuley, 1993). 
McAuley (1993) finds that personal contacts and agents overseas are the most frequent and useful 
sources of export information among the successful exporters in the UK. In an attempt to explain 
why distance plays the same role today in international trade as it did a century ago, Chaney (2011) 
argues that when distance captures information barriers to trade, which can be circumvented only 
by direct interaction between firms, then the persistent negative impact of distance is explained.
Information barriers, however, do not apply to the same extent across different markets. Network 
structural effects are expected to be more conspicuous in the presence of information uncertainty 
where the ‘information needed to pursue maximization and efficiency is not available, so that 
actors do not know in advance which option will produce the highest profits or the lowest costs’ 
(Leifer and White, 1987: 85–86). Prior research shows that immigration and resulting ethnic ties 
promote bilateral trade, especially in differentiated products, by mitigating the cost of the search 
 at FACULDADE IMED on April 8, 2014cos.sagepub.comDownloaded from 
http://cos.sagepub.com/
http://cos.sagepub.com/
126 International Journal of Comparative Sociology 54(2)
process (Gould, 1994; Rauch, 1996). Portes and Rey (2005) find similar results in the comparison 
of international transactions in corporate bonds with those in government bonds. The market infor-
mation for corporate bonds is more complicated and less accessible than that for government 
bonds. Thus, social and structural embeddedness is more valuable in global corporate bond mar-
kets than in government bond markets.
In this article we contextualize the effect of structural embeddedness by analyzing two goods 
(apparel and grain) with different levels of product differentiation and information uncertainty. If 
the tendency toward triadic closure is due to information diffusion through common contacts, then 
such a tendency should be observed only when information plays a central role in transaction (i.e. 
high information uncertainty). The importance of uncertainty as a factor in economic transactions 
has often been noted (Beckert, 1996) and this article extends its reach to how it influences structur-
ally embedded economic exchanges. The theoretically motivated comparison of grain and apparel 
trade also helps to clarify the generalizability of the embeddedness effect, which much prior 
research fails to address by only examining one industry/sector.
Structural embeddedness
Embeddedness has been a key concept in economic sociology since the 1980s. Scholars use it to 
critique standard economic models which detach economic action from noneconomic, social 
motives (Granovetter, 1992; Mizruchi et al., 2006; Uzzi, 1996; Uzzi and Lancaster, 2004; Zhou, 
2011) and from structural influence (Kim and Skvoretz, 2010). Theoretically, the embeddedness 
concept is closely associated with structural opportunity theory and claims that social relations 
depend on structurally determined opportunities for contact (Blau, 1977; Granovetter, 1973).1 The 
affinity lies in the selection biases hypothesized to structure economic interactions over and beyond 
the random choice assumed in standard economic models. Combining the theories of Blau and 
Granovetter, Fararo and Skvoretz (1987) identify two types of selection biases: homophily (node) 
biases and triad (tie) biases. The former, as Blau (1977) argues, refers to a nonrandom tendency for 
persons to choose to associate with people from the same social background (i.e. ingroup prefer-
ence). The latter, on the other hand, is inspired by Granovetter and ‘refers to nonrandom effects on 
net structure associated with patterns of ties among nodes’ (Fararo and Skvoretz, 1987: 1189); that 
is, a non-random chance for an X–Y relation to form when there is some third party Z linked to 
both X and Y.
Homophily biases relate to ‘dyadic social’ embeddedness. Previous studies of dyadic social 
embeddedness found that economic actors rely on previous (and mostly social) contacts in their 
choice of exchange partners rather than simple utility considerations. Economic transactions are 
socially embedded in this sense, producing long-term benefits such as trustworthiness and willing-
ness to forego opportunism (Galaskiewicz and Wasserman, 1989), while helping to avoid problems 
associated with market uncertainty (Podolny, 1994). Put differently, an economic actor tends to 
choose a partner from an ingroup, within which ties are strong, long lasting, and overlapping, even 
at the cost of short-term efficiency (see Zhou, 2011, for an application of this idea to the subject of 
international trade). Thus homophily, which creates ingroup ties, is one important theoretical ante-
cedent of dyadic social embeddedness.
‘Triadic structural’ embeddedness, on which this paper focuses, is based on the Simmelian 
tradition and suggests that interaction within a dyad is influenced by opportunity or constraint 
created by common third parties. The literature proposes two competing impacts on the dyad of 
common ties to a third party (Obstfeld, 2005). First, the third party may mediate between the two 
actors to whom it is tied and thereby encourage dyadic interaction (the Tertius Iungens: the third 
 at FACULDADE IMED on April 8, 2014cos.sagepub.comDownloaded from 
http://cos.sagepub.com/
http://cos.sagepub.com/
Kim and Skvoretz 127
who connects). Or, the common contact may intervene between the two actors and inhibit their 
direct interaction in order to benefit from brokerage advantages (the Tertius Gaudens: the third 
who enjoys).
Modes of the embeddedness effect
We define structural embeddedness as the structurally determined opportunity for (or constraint 
on) contact between two actors mediated by their common ties to third parties. We operationalize 
it as a third-party effect: the effect on a dyadic relationship between two parties having common 
ties to a third party. There are two competing hypotheses regarding the third-party effects, the bal-
ance hypothesis and the structural hole hypothesis. The balance hypothesis predicts triadic closure. 
That is, common ties to a third party will enhance interaction in the dyad. The structural hole 
hypothesis, on the other hand, mostly focuses on the brokerage opportunity granted to a common 
contact and implies that such a common contact will work to inhibit interaction in the dyad and 
thereby preserve its advantage (Burt, 1992).
Balance hypothesis
Many sociologists and psychologists have observed and reported triadic closure, the tendency for 
X–Y relations to form when they share a common third party, Z (Fararo and Skvoretz, 1987; Feld, 
1997; Friedkin, 1980, 1982; Granovetter, 1973; Heider, 1946; Kossinets and Watts, 2006). While 
this tendency is given different names, such as the ‘forbidden triad’ (Granovetter, 1973), ‘transitiv-
ity’ (Holland and Leinhardt, 1970) and ‘triad bias’ (Fararo and Skvoretz, 1987), the underlying 
foundation is balance theory (Heider, 1946). Balance theory originally applied to the cognitions of 
actors vis-à-vis some third entity to which the actors had either positive or negative attitudes. The 
core of the argument was that such triads would be in ‘balance’. For example, two persons who like 
each other would have the same attitude (either positive or negative) towards the third entity. 
Having different attitudes would create tension in the relation because the triad was ‘unbalanced’. 
When applied to social network data which often measured only positive affiliative ties between 
actors, positive ties were taken to be mutually reciprocated ties and negative ties were taken to be 
the absence of any connection. Hence a triad with mutual ties in two of the dyads and notie in the 
third dyad would be ‘unbalanced’, unstable, and lead to change, namely, the ‘closure’ of the triad 
by the formation of a mutual tie in the third dyad.
There are other theoretical arguments for triadic closure. Galaskiewicz and Wasserman (1989) 
argue that it might occur because common third contacts facilitate information flows between two 
actors. Friedkin (1980, 1982) substantiates this effect in studies of information diffusion via com-
mon third parties in a large sample of faculty members in various disciplines. His findings suggest 
that the probability that B’s awareness of C’s current research (other than through direct contact, 
reading about the research, or hearing a presentation of it), rises steadily as a function of the number 
of B’s contacts (A) who are informed about C’s activities. Furthermore, the stronger the tie is 
between B and the contact(s), the more likely B will receive the information about C’s current work. 
This, argues Friedkin, supports the hypothesis that the diffusion of information is influenced by 
network structure. As Baron and Hannan (1994: 1133, emphasis in original) observe, ‘sociological 
accounts. . . see the availability, nature, and value of information as products, often unintended, of 
social relations’. Based on the above argument, we formulate the balance hypothesis as follows:
Balance hypothesis: Growing structural embeddedness encourages dyadic interaction.
 at FACULDADE IMED on April 8, 2014cos.sagepub.comDownloaded from 
http://cos.sagepub.com/
http://cos.sagepub.com/
128 International Journal of Comparative Sociology 54(2)
Structural hole hypothesis
There are arguments that imply the exact opposite of the balance hypothesis: open triads – two 
actors are disconnected, although they are both connected to a common third party – can be more 
beneficial than closed ones, and thus may be more common. The structural hole hypothesis, most 
closely associated with the work of Burt (1992), is a good example. Burt identifies two different 
ways by which the absence of tie between two actors that share common contacts can benefit the 
actors in a triad. First, the common contact benefits from the absence of tie, which gives it an 
opportunity to broker interactions between its disconnected contacts. Second, it may be that the 
two actors (rather than their common contact) benefit from not choosing each other: already acces-
sible through their common contact, a direct tie between them is redundant and thus reduces their 
network efficiency. Put differently, a direct tie between two actors with common contact(s) is likely 
to elicit redundant information, already accessible through the common contact(s) (Burt, 1992; 
Granovetter, 1973).
The second effect is especially interesting because of its compatibility with the Heckscher-
Ohlin model in economics. Strong structural embeddedness in a dyad (a large overlap in ego net-
works) implies strong homogeneity between two actors with regard to preference and personal 
backgrounds. Rephrased in the context of international trade, strong structural embeddedness 
implies that two countries have similar supply and demand (or factor) endowments. According to 
the Heckscher-Ohlin model, heterogeneity in factor endowments encourages trade. Countries with 
similar factor endowments may produce domestically, rather than trading with each other, and so 
strong structural embeddedness in our terms would discourage bilateral trade according to this 
model. We formulate the structural hole hypothesis as follows:
Structural hole hypothesis: Growing structural embeddedness discourages dyadic interaction.
Uncertainty in trade: Apparel versus grain
In order to examine these competing hypotheses about structural embeddedness, we analyze two 
types of commodities, apparel and grain, whose markets differ in important ways. In this section, 
we compare the two goods in our focal terms – product differentiation and information uncertainty 
in transaction. Further differences between the goods will be discussed in later sections. Close 
consideration of the product differentiation difference suggests a refinement of the structural hole 
hypothesis while similar consideration of the information uncertainty difference suggests a refine-
ment of the balance hypothesis.
The two goods differ in terms of product differentiation. Grain is a homogeneous good in which 
the level of differentiation is relatively low. It is partly for this reason that grain is often traded in 
future markets where buyers, based on short specifications provided, know the quality of the com-
modity they will receive in the future. Also, an international reference price exists – a price that is 
quoted without knowledge of the name of the brand or manufacturer (Rauch, 1999), and price 
reflects the differences in quality reasonably well. Market information for grain is rather simple 
and easily obtained, and economic transactions tend to rely heavily on price.
Apparel, on the other hand, is a differentiated good for which there is no such a reference price. 
Due to relatively low barriers to entry in apparel manufacturing, this industry is highly competitive 
among a large number of small to medium-size firms throughout the world but especially in devel-
oping countries. Apparel manufacturing firms possess various levels of competence and produce 
goods that vary drastically in price, quality, and design. For this reason, apparel industry has been 
 at FACULDADE IMED on April 8, 2014cos.sagepub.comDownloaded from 
http://cos.sagepub.com/
http://cos.sagepub.com/
Kim and Skvoretz 129
characterized as a buyer-driven commodity chain (Gereffi, 1994, 1999), in which the production 
activities are globally sliced up among multiple independent firms, although distributors and mar-
keters in developed countries play central roles.
The difference in the level of product differentiation between apparel and grain suggests that the 
structural hole effect will have different consequences in the two markets. Specifically, we believe 
that the redundancy effect of common contacts is stronger in homogeneous goods (e.g. grain) than 
it is in differentiated goods (e.g. apparel). The similarity in supply/demand may not necessarily 
affect trade in differentiated goods (e.g. apparel) – two countries exporting women’s dresses may 
still exchange this commodity with each other. However, the same similarity is likely to reduce 
dyadic trade in homogeneous goods (e.g. grain) – two wheat exporters are unlikely to exchange 
that commodity with each other. Thus, the negative effects of structural embeddedness that the 
structural hole hypothesis posits may be limited to grain. We elaborate the hypothesis as follows:
Revised structural hole hypothesis: Growing structural embeddedness reduces dyadic trade of grain 
(because it increases redundancy) but does not affect apparel trade.
The level of product differentiation also affects the extent of information uncertainty that actors 
face in transaction. Apparel with a higher level of product differentiation is likely to have higher 
information uncertainty than grain. Greater information barriers in the apparel market imply that 
search process, involving the identification of profitable trading opportunities and location of suit-
able trading partners, is more lengthy and costly than it is in the grain market (Casella and Rauch, 
2002; Petropoulou, 2008). When formal sources of reliable information are lacking, actors tend to 
rely on personal networks, and therefore, the search for opportunities and partners would be more 
conditioned by networks than by markets (Gereffi, 1999; Rauch, 1999).
The difference in information uncertainty suggests that the balance hypothesis will have differ-
ent consequences in the two markets. Specifically, the trade-enhancing effect of common contacts 
is especially salient in a market where informational barriers and transactional uncertainty are high 
(apparel) and is lessimportant in a homogeneous market (grain), in which reliable information is 
widely available. In this case, dyadic trade depends more on supply and demand conditions in two 
countries than on the information as to who offers what. We elaborate the balance hypothesis as 
follows:
Revised balance hypothesis: Growing structural embeddedness increases dyadic trade of apparel (because 
it decreases information uncertainty) but does not affect grain trade.
The two hypotheses make different predictions for the effect of structural embeddedness on trade 
in grain and trade in apparel. Trade in grain is predicted to be either unaffected by structural embed-
dedness (balance) or negatively affected by embeddedness (structural hole). Trade in apparel is 
predicted to be either positively affected by embeddedness (balance) or unaffected (structural 
hole). Low information uncertainty in grain trade does not discourage dyadic trade, but its low 
product differentiation does. Similarly, while high product differentiation in apparel does not nec-
essarily encourage dyadic trade, the resulting high information uncertainty does.
Measurement of structural embeddedness
In order to measure structural embeddedness, we first obtain similarity index for each actor. This 
index is essentially the rate of common contacts to total contacts. Consider a dyad of X and Y. If X 
has a total of 10 trade partners (excluding Y), of which three are the trade partners of Y as well, the 
 at FACULDADE IMED on April 8, 2014cos.sagepub.comDownloaded from 
http://cos.sagepub.com/
http://cos.sagepub.com/
130 International Journal of Comparative Sociology 54(2)
similarity index for X is 0.3 (3/10). The same index for Y will be 0.2, assuming Y has total 15 trade 
partners (excluding X). Note that the index is asymmetric in the sense that each actor in a dyad has 
a different index. A similar measure can be obtained from valued matrices; the ratio of trade vol-
ume with common contacts to total trade volume.2 Both of the measures vary from zero to one: 
zero indicates no common contacts shared (thus no trade with them), whereas one indicates X’s 
trade partner is always Y’s partner.
As we explain below (see note 4), the dependent variable is symmetric (sum of import and 
export), whereas the similarity indices are asymmetric. For this reason, we symmetrize the indices 
by multiplication (the similarity embeddedness, in short).3 The higher the product, the more similar 
two actors in a dyad are in their trading patterns to the rest of the actors. The maximum (1.0) is 
achieved when both actor-level indices are 1.0. This product measure is very similar to the Jaccard 
coefficient, a popular symmetric measure of dyadic similarity. In fact, the correlations between the 
Jaccard coefficient and our measure are 0.986 in apparel and 0.974 in grain. The Jaccard coeffi-
cient, however, is available for a dichotomized matrix only, and thus assumes that all ties are equal 
in strength. So our product measure is preferable, since it does not have such a limit, nor loses 
information that the Jaccard coefficient delivers.
Our product measure, however, is insensitive to the gap that may exist between two actor simi-
larity indices. In other words, our product measure does not distinguish a dyad with the indices of 
0.6 and 0.4 from another dyad with 0.8 and 0.3. Although the products are same between the two 
dyads (0.24), the latter dyad has a greater gap in actor similarity indices. For this reason, we com-
plement the product measure by taking the absolute difference between the two actor similarity 
scores in a dyad (the dissimilarity embeddedness, in short), and add it as a control variable. The gap 
between actor similarity indices within a dyad has important theoretical implications to the balance 
hypothesis. The balance hypothesis assumes relative equivalence in strength between the two ties 
that connect two actors to their common third party. Consider the forbidden triad, for example. If 
there is a noticeable gap in strength between the two ties, the triad is not a forbidden any more. We 
expect the dissimilarity measure to have a negative effect on apparel trade, as the gap hinders infor-
mation flows through common contact(s).
Data and methods
Data
The data on trade of apparel and grain are obtained from the NBER-UN World Trade Flows 1962–
2000 (Feenstra et al., 2005). This data set was constructed from United Nations trade data by 
Feenstra and Lipsey. Trade data from 1992, 1996, and 2000 organized by the four-digit Standard 
International Trade Classification (revision 2) are used in this study. Of all the commodities traded, 
we extract grain (SITC 0411 – 0482) and apparel (SITC 8411 – 8484) for analyses. All the values 
are reported in American dollars and adjusted for inflation using the American Consumer Price 
Index (CPI). We choose 1992 as the beginning year to maximize the sample. Many countries, par-
ticularly the former Soviet republics, were not established before this year.
Excluding small economies (populations of less than one million) and those with missing data, 
a total of 136 countries are available for analysis (see Appendix for the list). Thus, there are six 
136×136 matrices (trade in grain and apparel in 1992, 1996, and 2000). Each matrix is originally 
asymmetric which distinguishes outgoing (export) and incoming (import) flows. Following the 
common practice in economic literature, the matrices are symmetrized by summing the flows.4 
Thus, each cell represents the sum of imports and exports between two countries in a dyad. This 
results in a total of 9180 (136×135/2) dyads per matrix. A large percent of the cells are zero (no 
 at FACULDADE IMED on April 8, 2014cos.sagepub.comDownloaded from 
http://cos.sagepub.com/
http://cos.sagepub.com/
Kim and Skvoretz 131
trade) ranging from 75.8 percent to 85.7 percent. Previous research sometimes excluded those zero 
cells because they cannot be log-transformed. However, such cases are included in the present 
study because zero-cells are legitimate observations to be explained. In order to do this, we add one 
(unit = $1000) to all cells, so that the zero cells become zero once logged.
It is important to control for confounding factors as suggested by the common ‘gravity’ model 
of trade. Dyadic attributes included as control variables are GDP, population, and formation of free 
trade agreements (FTA). As we discuss below, the fixed-effects model we use for data analysis 
makes unnecessary time-invariant controls (e.g. distance). Information on FTA was obtained from 
the WTO. Data on GDP and population were from the World Development Indicators by the World 
Bank. Because GDP data are not available for three countries in 1992, and for one in 1996, sample 
size declines slightly for those years. Although some researchers use GDP per capita, instead of 
population, the difference is a matter of preference, since the coefficients of the other independent 
variables are not affected. Following standard practice in the research literature, both GDP and 
population are entered in product form (GDPX × GDPY) to account for the increasing return to 
scale, and then the products are logged. Finally, we also note that the GDP product may control for 
the relative world-system status of the two countries: two core nations are likely to have a large 
value for this product while two periphery countries are likely to have a small one with a core–
periphery pair having an intermediate value. In that way, the GDP product when entered into a 
model becomes a way of controlling for world-system position as a determinant of the level to 
dyadic trade.
Methods
For multivariate analyses of data we use conditional logistic regression and fixed-effects model. 
The focus in the present study is on whether or not changes in triadic property lead to changes in 
dyadic property. By introducing country–dyad dummies, a fixed-effects model analyzes the 
changes within each dyadover time, while controlling for observable and unobservable time-
invariant effects such as distance, common language, and adjacency. Specifically, we use the GLM 
procedure in SAS with an ABSORB statement. Like fixed-effects methods, conditional logistic 
regression estimates the effects of the changes in time-varying predictors on the changes in a 
dichotomous dependent variable, while controlling for time-invariant variables. The STRATA 
statement in SAS puts each dyad as a separate stratum, and thus groups together the three observa-
tions over time for each dyad in the process of constructing the likelihood function (Allison, 2005).
Results
Table 1 shows the 10 largest exporters and importers of apparel and grain in 1992 and 2000. 
Compared to the apparel market, the international grain market is dominated by a small number of 
producers. While grain trade is more centralized in export (production) than in import (consump-
tion), the opposite is true in apparel. Though the top 10 exporting countries account for 85.7 per-
cent (1992) and 84.3 percent (2000) of global grain export, its import is widespread: the top 10 
importers account less than 50 percent of total import. Apparel trade differs from grain trade in this 
respect: the top 10 importers of apparel account for 79.5 percent (1992) and 80.4 percent (2000) of 
global imports, whereas the corresponding shares of top ten exporters are 64.7 percent in 1992 and 
61.1 percent in 2000. Also, in examining the changes in the 10 countries over the eight-year period, 
consumption (importers) of apparel appears more stable than production (exporters), whereas pro-
duction (exporters) is more stable than consumption (importers) for grain.
 at FACULDADE IMED on April 8, 2014cos.sagepub.comDownloaded from 
http://cos.sagepub.com/
http://cos.sagepub.com/
132 International Journal of Comparative Sociology 54(2)
T
ab
le
 1
. 
T
en
 la
rg
es
t 
ex
po
rt
er
s 
an
d 
im
po
rt
er
s 
of
 a
pp
ar
el
 a
nd
 g
ra
in
 in
 1
99
2 
an
d 
20
00
.
A
pp
ar
el
G
ra
in
Ex
po
rt
 1
99
2
Ex
po
rt
 2
00
0
Im
po
rt
 1
99
2
Im
po
rt
 2
00
0
Ex
po
rt
 1
99
2
Ex
po
rt
 2
00
0
Im
po
rt
 1
99
2
Im
po
rt
 2
00
0
C
hi
na
 (
20
.4
%
)
C
hi
na
 (
27
.5
%
)
U
SA
 (
21
.6
%
)
U
SA
 (
31
.1
%
)
U
SA
 (
29
.3
%
)
U
SA
 (
24
.2
%
)
Ja
pa
n 
(1
2.
5%
)
Ja
pa
n 
(1
1.
5%
)
It
al
y 
(9
.4
%
)
It
al
y 
(6
.0
%
)
G
er
m
an
y 
(1
7.
1%
)
G
er
m
an
y 
(1
0.
0%
)
Fr
an
ce
 (
18
.7
%
)
Fr
an
ce
 (
13
.2
%
)
It
al
y 
(6
.1
%
)
M
ex
ic
o 
(5
.1
%
)
H
on
g 
K
on
g 
(8
.3
%
)
H
on
g 
K
on
g 
(5
.9
%
)
Ja
pa
n 
(7
.9
%
)
Ja
pa
n 
(1
0.
0%
)
C
an
ad
a 
(1
0.
1%
)
C
an
ad
a 
(1
0.
9%
)
C
hi
na
 (
4.
7%
)
Br
az
il 
(4
.7
%
)
K
or
ea
 (
5.
3%
)
M
ex
ic
o 
(4
.8
%
)
H
on
g 
K
on
g 
(7
.7
%
)
H
on
g 
K
on
g 
(8
.2
%
)
A
rg
en
tin
a 
(4
.7
%
)
A
rg
en
tin
a 
(8
.1
%
)
Be
lg
iu
m
 (
4.
2%
)
K
or
ea
 (
4.
4%
)
G
er
m
an
y 
(5
.2
%
)
T
ur
ke
y 
(3
.6
%
)
Fr
an
ce
 (
6.
9%
)
U
K
 (
6.
2%
)
G
er
m
an
y 
(4
.6
%
)
A
us
tr
al
ia
 (
6.
9%
)
K
or
ea
 (
4.
1%
)
Ir
an
 (
4.
3%
)
Fr
an
ce
 (
3.
7%
)
In
do
ne
si
a 
(2
.8
%
)
U
K
 (
5.
5%
)
Fr
an
ce
 (
5.
5%
)
C
hi
na
 (
4.
3%
)
G
er
m
an
y 
(5
.9
%
)
N
et
he
rl
an
ds
 (
4.
1%
)
It
al
y 
(3
.9
%
)
T
ur
ke
y 
(3
.3
%
)
G
er
m
an
y 
(2
.7
%
)
N
et
he
rl
an
ds
 (4
.1
%
)
It
al
y 
(2
.9
%
)
U
K
 (
3.
9%
)
C
hi
na
 (
4.
9%
)
G
er
m
an
y 
(4
.0
%
)
U
SA
 (
3.
3%
)
U
SA
 (
3.
3%
)
K
or
ea
 (
2.
6%
)
Be
lg
iu
m
 (
3.
1%
)
Be
lg
iu
m
 (
2.
4%
)
A
us
tr
al
ia
 (
3.
8%
)
T
ha
ila
nd
 (
4.
6%
)
U
K
 (
3.
2%
)
In
do
ne
si
a 
(3
.1
%
)
Po
rt
ug
al
 (
3.
2%
)
U
SA
 (
2.
6%
)
It
al
y 
(3
.0
%
)
N
et
he
rl
an
ds
 (
2.
3%
)
T
ha
ila
nd
 (
3.
6%
)
U
K
 (
3.
9%
)
M
ex
ic
o 
(3
.2
%
)
A
lg
er
ia
 (
3.
1%
)
U
K
 (
2.
6%
)
Ba
ng
la
de
sh
 (
2.
6%
)
Sw
itz
er
la
nd
 (
2.
6%
)
M
ex
ic
o 
(2
.0
%
)
Be
lg
iu
m
 (
2.
6%
)
Be
lg
iu
m
 (
1.
7%
)
Br
az
il 
(2
.9
%
)
N
et
he
rl
an
ds
 (
3.
0%
)
$1
33
 b
ill
io
na
$1
95
 b
ill
io
n
$1
33
 b
ill
io
n
$1
95
 b
ill
io
n
$3
6 
bi
lli
on
$3
5 
bi
lli
on
$3
6 
bi
lli
on
$3
5 
bi
lli
on
a W
or
ld
’s
 t
ot
al
 v
ol
um
e.
 at FACULDADE IMED on April 8, 2014cos.sagepub.comDownloaded from 
http://cos.sagepub.com/
http://cos.sagepub.com/
Kim and Skvoretz 133
The stability and dominance of the major grain exporters fits the characteristics of the producer-
driven commodity chain (Gereffi, 1994). The grain industry has relatively high barriers to entry. 
Large-scale production (and export) of grain requires certain natural endowments, such as area/
quality of arable land and climate. In addition, it also requires large capital investments to expand 
arable land, to improve productivity, and to build infrastructure to store and transport grain to 
major ports. Lacking such resources, agriculture in many developing countries remains at the sub-
sistence level, which results in lower proportions of export to total production, compared to indus-
trialized countries. On the other hand, the relative instability of major apparel exporters implies 
competitive and decentralized production systems, a characteristic of the buyer-driven commodity 
chain (Gereffi, 1994).
Table 2 presents the partial correlation coefficients among the variables with the year effects 
controlled for. There are negative correlations between similarity and dissimilarity measures, rang-
ing from −0.11 to −0.60. That is, as the product of two actor indices increases, their difference 
tends to decline. This is not surprising, since a gap between actor indices tends to lower their prod-
uct. This table also shows that geographically adjacent dyads tend to be more structurally embed-
ded. In other words, adjacent countries have a higher rate of common contacts, although it does not 
necessarily indicate that those common contacts are adjacent to the countries as well. Also, com-
paring the two goods, the negative correlation between structural embeddedness and distance is 
stronger in grain trade than it is in apparel trade.
Table 3 presents some structural characteristics of the trade networks. In all three years, the 
networks of apparel trade are found to be stronger, denser, and less centralized than those of grain 
trade. There was four to five times more trade in apparel than in grain. The average volume of 
apparel trade increased from $14.5 million in 1992 to $18.3 million in 1996 and to $21.2 million 
in 2000.5 On the other hand, the corresponding averages of grain trade are somewhat less consist-
ent: $3.9 million in 1992, $5.5 million in 1996, and $3.8 million in 2000. However, the sudden 
increase in the average volume of grain trade in 1996 reflects a spike in grain prices due primarily 
to the drought in the Midwestern and Western regions of the US during 1995–1996. With the matri-
ces dichotomized (presence or absence of trade), the apparel networks are also denser − 19.7 
Table 2. Pearson’s partial correlation coefficients, controlling for year (N = 26,997).
1 2 3 4 5 6 7 8 9 10
 1. GDP 
 2. Population 0.63 
 3. Distance 0.03 0.07 
Similarity (product) 
 4. Apparel−binary 0.46 0.15 −0.21 
 5. Apparel−volume 0.53 0.21 −0.16 0.78 
 6. Grain−binary 0.07 0.04 −0.36 0.32 0.22 
 7. Grain−volume 0.29 0.20 −0.30 0.28 0.29 0.63 
Dissimilarity (difference) 
 8. Apparel−binary 0.18 0.12 0.04 −0.36 −0.12 −0.20 −0.04 
 9. Apparel−volume −0.17 −0.08 0.11 −0.49 −0.60 −0.19 −0.16 0.58 
10. Grain−binary 0.33 0.24 −0.04 −0.16 0.01* −0.11 0.07 0.58 0.29 
11. Grain−volume 0.09 0.08 0.12 −0.18 −0.09 −0.17 −0.37 0.33 0.25 0.53
*Not significant at the 0.05 level.
 at FACULDADE IMED on April 8, 2014cos.sagepub.comDownloaded from 
http://cos.sagepub.com/
http://cos.sagepub.com/
134International Journal of Comparative Sociology 54(2)
percent in 1992, 23.3 percent in 1996, and 24.2 percent in 2000 – than the grain networks − 14.3 
percent, 16.0 percent, and 15.9 percent, respectively.
Lower volume and density in grain trade may be explained partly by state regulations. Agriculture 
is typically under control and protection of the state rather than left to market mechanisms in order 
to secure a greater degree of self-sufficiency and a higher level of income in the agricultural sector 
(Butler, 1986). Apparel trade has also been under special restrictions such as the Multi Fibre 
Agreement (MFA) that regulated trade in apparel by setting up quotas. Since the early 1960s, the US 
and some other advanced countries limited quantities of apparel that can be imported in order to 
protect domestic manufacturers from cheap foreign competitors. Contrary to the case of grain, how-
ever, those restrictions on apparel trade contributed to trade diversification. Although a vast majority 
of exporting countries did not fill the quota (Fugazza and Conway, 2010), a few major apparel 
exporters (mostly East Asian countries) had to come up with solutions to the quota restriction, which 
include trade diversification (Gereffi, 1999). Small exporters could survive under the quota system, 
in spite of their lower comparative advantage (Fugazza and Conway, 2010).
Network degree centralization index scores obtained by UCINET V (Borgatti et al., 1999) show 
the extent to which the entire network tends to be dominated by only a few countries. As presented 
in the table, the networks of grain trade are more centralized than those of apparel trade, although 
the gap declines over the years. The degree centralization in the apparel networks increases slightly, 
which may be due to the increasing dominance of China (from 20.4% to 27.5% in its share of the 
world’s exports). On the other hand, the network centralization of grain declines, as former com-
munist countries, such as Kazakhstan and Hungary, expand their exports, while other countries like 
Algeria and Iran increase their imports and diversify the sources of those imports.
Tables 4 and 5 present the results from conditional logistic regression and fixed-effects model, 
respectively. Before interpreting the results, it should be noted that Table 4 puts one-sided focus on 
the changes involving zero, that is, emergence or discontinuance of trade tie, while ignoring the 
changes in volume within an already existing tie. In Table 5, on the other hand, equal weight is 
given to every unit change in trade volume, even if the change involves emergence or discontinu-
ation of a tie. In other words, the tie change model (Table 4) does not distinguish the dyads with 
trade ties across all three years from the dyads with no ties at all, since neither type has any change 
in tie formation over the years. The volume change model (Table 5), however, distinguishes the 
two types of dyads, as the former may have different volumes across the years.
The results are generally consistent in both tables. The formation of FTA does not have a signifi-
cant effect on either tie formation (Table 4) or volume (Table 5) in apparel trade, while it does 
impact both for grain trade. Contrary to the common assumption that FTAs promote trade, those 
Table 3. Network structural descriptive statistics of trade in apparel and grain, 1992, 1996, and 2000.
1992 1996 2000
 Apparel Grain Apparel Grain Apparel Grain
Densityb 19.7% 14.3% 23.3% 16.0% 24.2% 15.9%
Average volumea 
(standard deviation)
14.5 (172.4) 3.9 (45.3) 18.3 (237.4) 5.5 (66.5) 21.2 (302.8) 3.8 (45.0)
Network 
centralizationb
61.97% 74.95% 62.84% 72.45% 63.45% 68.82%
aMillion US dollars.
bThe matrices are dichotomized and symmetrized to get these statistics.
 at FACULDADE IMED on April 8, 2014cos.sagepub.comDownloaded from 
http://cos.sagepub.com/
http://cos.sagepub.com/
Kim and Skvoretz 135
T
ab
le
 4
. 
R
es
ul
ts
 (
lo
g-
od
ds
 r
at
io
s)
 fr
om
 c
on
di
tio
na
l l
og
is
tic
 r
eg
re
ss
io
n,
 p
re
di
ct
in
g 
tie
 fo
rm
at
io
n 
w
ith
in
 a
 d
ya
d 
(W
al
d 
C
hi
-s
qu
ar
e)
.
A
pp
ar
el
G
ra
in
 
M
od
el
 1
M
od
el
 2
M
od
el
 3
M
od
el
 1
M
od
el
 2
M
od
el
 3
FT
A
0.
10
 (
0.
04
)
0.
12
 (
0.
05
)
0.
03
 (
0.
00
)
0.
61
* 
(4
.2
1)
0.
62
* 
(4
.2
9)
0.
59
#
 (
3.
79
)
G
D
P
0.
30
 (
1.
47
)
0.
28
 (
1.
31
)
0.
10
 (
0.
15
)
– 
0.
88
**
* 
(1
2.
30
)
– 
0.
89
**
* 
(1
5.
00
)
– 
0.
95
**
* 
(1
6.
40
)
Po
pu
la
tio
n
– 
2.
49
**
* 
(1
5.
05
)
– 
2.
58
**
* 
(1
5.
67
)
– 
2.
49
**
* 
(1
4.
72
)
– 
1.
57
**
 (
7.
69
)
– 
1.
73
**
 (
9.
20
)
– 
1.
73
**
 (
9.
15
)
Si
m
ila
ri
ty
 
em
be
dd
ed
ne
ss
 (
Bi
na
ry
)
3.
19
**
* 
(1
7.
03
)
3.
26
**
* 
(1
7.
57
)
– 
36
.7
6*
 (
3.
78
)
– 
0.
50
 (
0.
70
)
– 
0.
35
 (
0.
33
)
– 
19
.2
8 
(2
.1
7)
D
is
si
m
ila
ri
ty
 
em
be
dd
ed
ne
ss
 (
Bi
na
ry
)
0.
33
 (
0.
61
)
0.
43
 (
1.
00
)
1.
14
**
* 
(1
1.
28
)
1.
19
**
* 
(1
2.
05
)
In
te
ra
ct
io
n 
(G
D
P*
Si
m
. 
Em
be
dd
ed
ne
ss
)
0.
82
* 
(4
.4
8)
0.
38
 (
2.
10
)
Y
ea
r 
19
92
– 
1.
31
**
* 
(5
7.
99
)
– 
1.
34
**
* 
(5
7.
71
)
– 
1.
30
**
* 
(5
4.
15
)
– 
1.
00
**
* 
(4
6.
93
)
– 
1.
12
**
* 
(5
4.
74
)
– 
1.
10
**
* 
(5
2.
48
)
Y
ea
r 
19
96
– 
0.
36
**
* 
(1
1.
81
)
– 
0.
37
**
* 
(1
2.
36
)
– 
0.
36
**
* 
(1
1.
35
)
– 
0.
31
**
 (
10
.1
8)
– 
0.
36
**
* 
(1
3.
61
)
– 
0.
36
**
* 
(1
2.
97
)
– 
2 
Lo
g 
L
22
73
.3
22
72
.7
22
68
.1
25
34
.7
25
23
.3
25
21
.2
#
p 
<
 0
.1
0;
 *
 p
 <
 0
.0
5;
 *
* 
p 
<
 0
.0
1;
 *
**
 p
 <
 0
.0
01
.
 at FACULDADE IMED on April 8, 2014cos.sagepub.comDownloaded from 
http://cos.sagepub.com/
http://cos.sagepub.com/
136 International Journal of Comparative Sociology 54(2)
T
ab
le
 5
. 
G
LM
 e
st
im
at
es
 o
f f
ix
ed
 e
ffe
ct
s 
m
od
el
, p
re
di
ct
in
g 
th
e 
ch
an
ge
s 
in
 d
ya
di
c 
tr
ad
e 
vo
lu
m
e 
(s
ta
nd
ar
d 
er
ro
r)
a .
A
pp
ar
el
G
ra
in
 
M
od
el
 1
M
od
el
 2
M
od
el
 3
M
od
el
 1
M
od
el
 2
M
od
el
 3
FT
A
– 
0.
15
 (
0.
10
)
– 
0.
14
 (
0.
10
)
– 
0.
15
 (
0.
10
)
0.
59
**
* 
(0
.1
2)
0.
59
**
* 
(0
.1
2)
0.
60
**
* 
(0
.1
2)
G
D
P
0.
32
**
* 
(0
.0
5)
0.
32
**
* 
(0
.0
5)
0.
22
**
* 
(0
.0
5)
– 
0.
14
* 
(0
.0
6)
– 
0.
14
* 
(0
.0
6)
– 
0.
13
* 
(0
.0
6)
Po
pu
la
tio
n
– 
1.
48
**
* 
(0
.1
4)
– 
1.
47
**
* 
(0
.1
4)
– 
1.
40
**
* 
(0
.1
4)
– 
0.
65
**
* 
(0
.1
7)
– 
0.
65
**
* 
(0
.1
7)
– 
0.
63
**
* 
(0
.1
7)
Si
m
ila
ri
ty
 
em
be
dd
ed
ne
ss
 (
V
al
ue
d)
0.
27
**
* 
(0
.0
6)
0.
23
**
 (
0.
07
)
– 
11
.2
0*
**
 (
1.
06
)
– 
0.
16
* 
(0
.0
7)
– 
0.
15
* 
(0
.0
7)
0.
96
 (
1.
08
)
D
is
si
m
ila
ri
ty
 
em
be
dd
ed
ne
ss
 (
V
al
ue
d)
– 
0.
08
 (
0.
06
)
– 
0.
01
 (
0.
06
)
0.
01
 (
0.
06
)
0.
00
 (
0.
06
)
In
te
ra
ct
io
n 
(G
D
P*
si
m
. 
em
be
dd
ed
ne
ss
)
0.
25
**
* 
(0
.0
2)
– 
0.
02
 (
0.
02
)
Y
ea
r 
19
92
– 
0.
54
**
* 
(0
.0
4)
– 
0.
54
**
* 
(0
.0
4)
– 
0.
50
**
* 
(0
.0
4)
– 
0.
25
**
* 
(0
.0
5)
– 
0.
26
**
* 
(0
.0
5)
– 
0.
25
**
* 
(0
.0
5)
Y
ea
r 
19
96
– 
0.
18
**
* 
(0
.0
3)
– 
0.
17
**
* 
(0
.0
3)
– 
0.
16
**
* 
(0
.0
3)
– 
0.
03
 (
0.
03
)
– 
0.
03
 (
0.
03
)
– 
0.
03
 (
0.
03
)
R2
0.
91
0.
91
0.
91
0.
82
0.
82
0.
82
N
27
,0
02
27
,0
03
 
a T
he
 9
17
9 
co
un
tr
y-
pa
ir
 d
um
m
ie
s 
ar
e 
no
t 
re
po
rt
ed
.
#
p 
<
 0
.1
0;
 *
 p
 <
 0
.0
5;
 *
* 
p 
<
 0
.0
1;
 *
**
 p
 <
 0
.0
01
.
 at FACULDADE IMED on April 8, 2014cos.sagepub.comDownloaded from 
http://cos.sagepub.com/
http://cos.sagepub.com/
Kim and Skvoretz 137
established between 1992 and 2000 had no trade-creatingeffect for apparel. A possible explanation 
includes that the period under investigation is either too early or too short to capture the effects of 
the FTAs that are important to apparel trade (e.g. NAFTA). A positive effect of FTA on grain may 
be explained by initial level of market protection and degree of product differentiation. According 
to Goto (1997), they are two key parameters upon which the impact of FTA depends. That is, the 
higher the initial level of protection and the lower the product differentiation, the greater the impact 
that an FTA has on bilateral trade. Goto (1997) finds that FTA tends to have a greater impact on 
agricultural trade flows than on manufacturing trade for the same reasons. Grant and Lambert 
(2005) also report a similar finding.
Change in income (GDP) does not have significant effects on tie formation (Table 4), while 
significantly enhancing the volume of apparel trade (Table 5). Apparel has a positive income elas-
ticity, meaning that its demand rises with income and thus, a positive sign on it was expected. 
While the increasing demand may lead to an increase in the number of apparel trade partners, the 
results do not support the expectation (Table 4). The same variable (GDP), on the contrary, has a 
negative effect on grain trade. This indicates a negative income elasticity: an increase in income 
reduces consumption of grain. This is a well-known trend in grain trade, particularly among middle 
and high income countries. The coefficients of population growth consistently show a negative 
effect on both apparel and grain. Controlling for GDP and other factors, growing population 
reduces trade of both goods, but of apparel especially.
The analyses consistently show that the changes in structural embeddedness have positive 
impacts on both the changes in tie formation (Table 4) and the changes in trade volume (Table 5) 
for apparel, which supports the balance hypothesis. This implies that when market information is 
complex, trading firms tend to rely on their networks for transaction. How about grain trade? Due 
to a lower level of product differentiation, we expected negative impacts of structural embedded-
ness. The findings are somewhat less consistent. The changes in structural embeddedness have no 
significant impact on the formation of trade tie for grain (Table 4). However, they have a negative 
and significant (t = –2.28 and –2.03) effect on the changes in trade volume of grain (Model 1 and 
2 in Table 5, respectively). This means, for example, that when the similarity indices of two actors 
in a dyad changes from 0.6 – 0.4 to 0.8 – 0.6 (the same difference but a higher product), trade vol-
ume within the dyad declines. This was expected due to growing redundancy.
Finally, regarding the gap between two actor similarity indices, we expected a significant and 
negative effect on apparel trade. As the tables show, however, it is non-significant, although it 
approaches to the statistical significance (t = –1.48) with the expected sign in the trade volume 
model (Model 2 in Table 5). Interestingly, the same variable is positive and highly significant (χ2 = 
11.28, p = 0.0008) in grain trade (Model 2 of Table 4), meaning that a new tie is likely to form, when 
the gap between actor similarity indices is growing in a dyad. For example, a tie is more likely to 
form, when the similarity indices of two actors in the dyad changes from 0.6 – 0.4 to 0.8 – 0.3 (the 
same product but a higher difference). This finding may have to do with declining redundancy. That 
is, when the gap in actor similarity indices is growing, one becomes less redundant to the other.
Are the effects of structural embeddedness in apparel trade present across all types of dyads? It 
is possible that the triadic effects are contingent upon the combination of two countries in a dyad. 
For example, the triad of the US, Uruguay, and Kenya (i.e., one core and two peripheries) in which 
the US is the common contact, may be still open, even if the latter two would know each other’s 
market better due to their common contact. They may not have sufficient demands on differenti-
ated goods. However, the same scenario is less likely in such a triad, the US–France–Germany 
(three core countries). That is, triadic closure is more likely when both actors in a dyad are core, 
but less likely when both are peripheral. To investigate this possibility, we add an interaction term 
 at FACULDADE IMED on April 8, 2014cos.sagepub.comDownloaded from 
http://cos.sagepub.com/
http://cos.sagepub.com/
138 International Journal of Comparative Sociology 54(2)
between GDP and structural embeddedness (see Model 3 of Tables 4 and 5). The latter, of course, 
represents the triadic effects, and the former proxies the combination effect of dyad. If a dyad con-
sists of two core countries, the product of their GDPs will be greater than that of the dyad with two 
peripheral countries.
The results support the contingency effect. While the interaction term is not significant in 
grain trade, it is positive and highly significant in apparel trade, meaning that the effects of 
structural embeddedness are positively associated with the size of economies in a dyad. 
Specifically, the parameter estimates show that structural embeddedness does not increase the 
odds of tie formation (Table 4) or trade volume (Table 5) in apparel until the logged product of 
the two countries’ GDPs reaches 44.92 and 45.31, respectively (the mean logged product GDP 
= 47.28). In fact, these two values are very close to the logged product (44.14) of two econo-
mies at the 25th percentile (e.g. Nepal and Haiti, whose GDPs are $3.7 billion and $4.0 billion, 
respectively). This means that the positive effects of structural embeddedness in apparel trade 
do not apply to the dyads with small economies (and thus small demands). These results reveal 
another condition, along with the level of information uncertainty, for triadic closure – a dyadic 
attribute of economic size.
Discussion
International trade is shaped not only by the dyadic characteristics of trading countries, but also by 
their relations with other countries. Operationalizing structural embeddedness as the extent of trade 
ties to common third parties, this article examined its effect on bilateral trade. Specifically, we 
tested two hypotheses – the balance and the structural hole hypotheses – in two markets with dif-
ferent levels of product differentiation and transactional uncertainty. The multivariate longitudinal 
analysis revealed that when actors transact differentiated goods, the presence of common contacts 
encourages transaction within a dyad. The resulting triadic closure is likely due to the higher uncer-
tainty that actors face and try to reduce through their common contacts. On the other hand, when 
they transact homogeneous goods, the presence of common contacts discourages transaction within 
a dyad. We believe that this finding reflects the redundancy effect of structural embeddedness in 
markets with relatively homogeneous goods.
In order to further test the arguments, we ran the same GLM models in Table 5 separately for 
the dyads whose distance is below and above the mean distance.6 If the positive effect of structural 
embeddedness on apparel trade reflects information barriers, it must be clearer among the dyads 
with above average distance. Likewise, if growing homogeneity in factor endowments is behind 
the negative effect of structural embeddedness in grain trade, such an effect would be stronger 
among the dyads with above average distance, since there is even less reason to exchange homo-
geneous goods with distant partners. The results (not reported to save space) confirm the expecta-
tions. The coefficient on the similarity embeddedness in apparel trade not only increases but also 
gets more significant among the above average distant dyads. On the other hand, the same variable 
becomes non-significant among the below average distant dyads.The same pattern (with the oppo-
site sign) is found in grain trade: growing structural embeddedness reduces grain trade among 
distant dyads, whereas it is non-significant among relatively adjacent dyads.
Our findings can be generalized to other markets. The balance effects are likely to be found in 
a market where goods are differentiated and barriers to entry are low, both of which increase trans-
actional uncertainty. On the other hand, the structural hole effects are possible in a market where 
goods are rather homogeneous and/or barriers to entry are high.7 Since many finished goods are 
more or less differentiated, the balance effects may be rather common across trade networks of 
 at FACULDADE IMED on April 8, 2014cos.sagepub.comDownloaded from 
http://cos.sagepub.com/
http://cos.sagepub.com/
Kim and Skvoretz 139
finished goods, especially when Third World producers are visible due to lower entry barriers. 
Likewise, since many raw and intermediary goods are less differentiated, and their markets are 
often dominated by a smaller number of producers, their trade networks are more likely to show 
the structural hole effects.
More speculatively, our results suggest a process by which the overarching structure of core and 
periphery of world-system theory occurs and is maintained. In this core–periphery structure, much 
triadic closure occurs among core nations and very little among peripheral nations, while open 
triads often have a core nation at the center of two peripheral nations trading with each of them. 
Such a pattern can occur and be maintained if much of the trade between core nations is in differ-
entiated, high uncertainty commodities whereas the trade between core and peripheral is in homog-
enous and undifferentiated commodities. Even within the trade of differentiated goods where we 
expect the balance effects, a triad with two peripheral countries and a core county at the center may 
remain open, as our further analysis shows, because of insufficient demand on such goods within 
the peripheral countries. On the other hand, when a core country establishes a trade tie with a 
periphery for differentiated goods, the former is constrained by available information which it 
gains from its established foreign partners. In other words, between two peripheral countries (with 
a similar level of wages), trade is more likely with the one that is a partner of one’s partner. And 
this effect is particularly conspicuous between remote countries.
Why is (valuable) trade information relayed through common contacts in apparel trade, instead 
of being held by common contacts, as the structural hole hypothesis posits? While the answers to 
this question require data at the firm level, we note two possibilities. One is that common contacts 
deliberately choose to share information with two disconnected alters. Although network closure 
eliminates potential brokerage opportunities for common contacts, it also creates other advantages 
(Obstfeld, 2005). For example, introduction of trading partners who are otherwise disconnected 
may be reciprocated later, which further enriches the networks of common contacts. The other pos-
sibility is the coordination problems that exist among multiple common contacts. That is, even if 
common contacts have motivations to control information flows and to benefit from brokerage 
opportunities, it becomes hard for a single actor to practically achieve it, when faced with other 
actors in a similar position. This may be particularly true in the markets where competition is high 
(e.g. the apparel market), so that information naturally diffuses along the ties.
Finally, our results cover a relatively short period of time (1992–2000) so may not capture 
longer-term changes in the global supply and demand of apparel and grain. This limitation will 
be especially serious, if the demand and supply of the goods show non-random (e.g. cyclical) 
patterns which are not fully captured by the eight-year observation period (see for example, 
Hopkins and Wallerstein, 1994). Furthermore, the period is practically under the MFA regime, 
which tends to diversify international apparel trade. ‘Triangle manufacturing’ (Gereffi, 1999) is 
an example, where an apparel exporting country (e.g. South Korea) bypasses quota restriction 
by manufacturing in a third country (e.g. Honduras) and shipping it directly to advanced indus-
trial countries (e.g. US). If South Korea traded finished apparel with Honduras before the trian-
gle manufacturing started, then the balance effects in apparel trade might be facilitated by the 
MFA. Thus, future research should cover a longer and more recent period of time (after 2005 
when the MFA was formally phased out) to evaluate the generality of the patterns we found in 
this study.
Acknowledgements
We thank the editor and the anonymous reviewers of International Journal of Comparative Sociology for their 
helpful comments.
 at FACULDADE IMED on April 8, 2014cos.sagepub.comDownloaded from 
http://cos.sagepub.com/
http://cos.sagepub.com/
140 International Journal of Comparative Sociology 54(2)
Funding
This research received no specific grant from any funding agency in the public, commercial, or not-for-profit 
sectors.
Notes
1. The definitions of structure vary, however. For example, Blau (1977) defines it as population distribu-
tions among social positions along various lines. For Granovetter (1973), it refers to network structure, 
created by social ties between individuals.
2. Research shows that both the strength and the number of ties to common contacts are important predic-
tors of dyadic interaction. Granovetter’s forbidden triad emphasizes the strength of the ties that B and 
C have with their common third contact (A), as he argues, ‘If A-B and A-C ties exist, then the amount 
of time C spends with B depends (in part) on the amount A spends with B and C, respectively’ (1973: 
1362). Friedkin (1980) and Kossinets and Watts (2006) find that the number of common contacts is also 
an important predictor of dyadic interaction.
3. The product measure assumes more than proportional decay in information flows at lower levels of actor 
similarity. For this reason, we also use the average measure, which produces identical results.
4. Whether to use asymmetric or symmetric trade as a dependent variable is essentially about whether or 
not nodal attributes are turned into dyadic attributes. When asymmetric trade volume is used as a depend-
ent variable, the model is:
TX-->Y = Exporter(X)’s nodal attributes + Importer(Y)’s nodal attributes + XY’s dyadic attributes
TY-->X = Exporter(Y)’s nodal attributes + Importer(X)’s nodal attributes + XY’s dyadic attributes
 For example, if GDP and Distance are predictor variables, then
TX-->Y = beGDPX + biGDPY + bdDISTANCEXY
TY-->X = beGDPY + biGDPX + bdDISTANCEXY.
 When these two equations are added,
TXY = (be+bi)(GDPX) + (be+bi)(GDPY) + 2bd(DISTANCEXY).
 That is, the model assumes that the volume of trade between X and Y is a function of X and Y’s nodal 
attributes and their dyadic attributes.
 On the other hand, when symmetric trade is used as a dependent variable, the model is:
TXY =bs( GDPX * GDPY) + bdDISTANCEXY.
 That is, nodal attributes (GDPs) are turned into a dyadic attribute by multiplication. Comparing this with 
the earlier model, a difference is whether to put GDP as nodal or dyadic attribute.
5. Although the volume of apparel trade increases over the years, its share of total world trade declines from 
3.83 percent in 1992 to 3.39 percent in 2000. The share of grain trade to world’s total shows a similar, if 
less consistent, trend: 1.04 percent in 1992, 1.05 percent in 1996, and 0.60 percent in 2000.
6. Since distance is a time-invariant variable, it cannot be added to a fixed-effect model.
7. We believe that barriers to entry tend to become higher in homogeneous goods markets, as producers attempt 
to achieve economies of scale in order to lower price, a most importantcompetitive advantage in such markets.
References
Allison PD (2005) Fixed Effects Regression Methods for Longitudinal Data: Using SAS. Cary, NC: SAS 
Press.
Baron J and Hannan MT (1994) The impact of economics on contemporary sociology. Journal of Economic 
Literature 32: 1111–1146.
Beckert J (1996) What is sociological about economic sociology? Uncertainty and the embeddedness of eco-
nomic action. Theory and Society 25(6): 803–840.
 at FACULDADE IMED on April 8, 2014cos.sagepub.comDownloaded from 
http://cos.sagepub.com/
http://cos.sagepub.com/
Kim and Skvoretz 141
Blau PM (1977) Inequality and Heterogeneity: A Primitive Theory of Social Structure. New York: Free Press.
Borgatti S, Everett MG and Freeman LC (1999) UCINET V for Windows. Cambridge, MA: Analytic 
Technologies.
Burt RS (1992) Structural Holes: The Social Structure of Competition. Cambridge, MA: Harvard University 
Press.
Butler N (1986) The International Grain Trade: Problems and Prospects. New York: St Martin’s Press.
Casella A and Rauch J (2002) Anonymous market and group ties in international trade. Journal of International 
Economics 58: 19–47.
Chaney T (2011) The gravity equation in international trade: An explanation. Unpublished and preliminary 
manuscript. Available at: http://home.uchicago.edu/tchaney/research/Distance.pdf (accessed October 2011).
Fararo TJ and Skvoretz J (1987) Unification research programs: Integrating two structural theories. American 
Journal of Sociology 92(5): 1183–1209.
Feld SL (1997) Structural embeddedness and stability of interpersonal relations. Social Networks 19(1): 91–95.
Feenstra RC, Lipsey RE, Deng H, et al. (2005) World trade flows: 1962–2000. NBER Working Paper No. 
11040.
Friedkin N (1980) A test of structural features of Granovetter’s strength of weak ties theory. Social Networks 
2: 411–422.
Friedkin N (1982) Information flow through strong and weak ties in intraorganizational social networks. 
Social Networks 3: 273–285.
Fugazza M and Conway P (2010) The impact of removal of ATC quotas on international trade in textiles and 
apparel. Policy Issues in International Trade and Commodities Study Series No. 45, UNCTAD.
Galaskiewicz J and Wasserman S (1989) Mimetic processes with an interorganizational field: An empirical 
test. Administrative Science Quarterly 34(3): 454–479.
Gereffi G (1994) The organization of buyer-driven global commodity chains: How U.S. retailers shape 
overseas production networks. In: Gereffi G and Korzeniewicz M (eds) Commodity Chains and Global 
Capitalism. Westport, CT: Praeger, 95–122.
Gereffi G (1999) International trade and industrial upgrading in the apparel commodity chain. Journal of 
International Economics 48(1): 37–70.
Goto J (1997) Regional economic integration and agricultural trade. Policy Research Working Papers (1805), 
World Bank.
Gould DM (1994) Immigrant links to the home country: Empirical implications for U.S. bilateral trade flows. 
Review of Economics and Statistics 76(2): 302–316.
Granovetter M (1973) The strength of weak ties. American Journal of Sociology 78(6): 1360–1380.
Granovetter M (1992) The sociological and economic approaches to labor market analysis: A social structural 
view. In: Granovetter M and Swedberg R (eds) The Sociology of Economic Life. Boulder, CO: Westview 
Press, 233–263.
Grant JH and Lambert DM (2005) Regionalism in world agricultural trade: Lessons from gravity model esti-
mation. Paper presented at the American Agricultural Economics Association Annual Meeting.
Heider F (1946) Attitudes and cognitive organization. Journal of Psychology 21: 107–112.
Holland PW and Leinhardt S (1970) A method for detecting structure in sociometric data. American Journal 
of Sociology 76(3): 492–513.
Hopkins TK and Wallerstein I (1994) Commodity chains: Construct and research. In Gereffi G and 
Korzeniewicz M (eds). Commodity Chains and Global Capitalism, Westport, CT: Praeger, 17–20.
Howard DG and Herremans IM (1988) Sources of assistance for small business exporters: Advice from suc-
cessful firms. Journal of Small Business Management 26(3): 48–54.
Ingram P, Robinson J and Busch M (2005) The intergovernmental network of world trade: IGO connections, 
governance, and embeddedness. American Journal of Sociology 111(3): 824–858.
Keegan WJ (1974) Multinational scanning: A study of the information sources utilized by headquarters exec-
utives in multinational companies. Administrative Science Quarterly 19(3): 411–421.
Kim S and Skvoretz J (2010) Embedded trade: A third-party effect. Social Science Quarterly 91(4): 964–983.
Kossinets G and Watts DJ (2006) Empirical analysis of an evolving social network. Science 311: 88–90.
Leifer E and White H (1987) A structural approach to markets. In: Mizruchi M and Schwartz M (eds) 
Intercorporate Relations. Cambridge: Cambridge University Press, pp. 85–108.
 at FACULDADE IMED on April 8, 2014cos.sagepub.comDownloaded from 
http://cos.sagepub.com/
http://cos.sagepub.com/
142 International Journal of Comparative Sociology 54(2)
Mahutga M (2006) The persistency of structural inequality? A network analysis of international trade. Social 
Forces 84(4): 1863–1889.
McAuley A (1993) The perceived usefulness of export information sources. European Journal of Marketing 
27(10): 52–64.
Mizruchi MS, Stearns LB and Marquis C (2006) The conditional nature of embeddedness: A study of borrow-
ing by large U.S. firms, 1973–1994. American Sociological Review 71(2): 310–333.
Obstfeld D (2005) Social networks, the Tertius Iungens orientation, and involvement in innovation. 
Administrative Science Quarterly 50(1): 100–130.
Petropoulou D (2008) Information costs, networks and intermediation in international trade. CEP Discussion 
Paper No. 848.
Podolny JM (1994) Market uncertainty and the social character of economic exchange. Administrative 
Science Quarterly 39(3): 458–483.
Portes R and Rey H (2005) The determinants of cross-border equity flows. Journal of International Economics 
65(2): 269–296.
Rauch JE (1996) Trade and search: Social capital, sogo shosha, and spillovers. NBER Working Paper 5618.
Rauch JE (1999) Networks versus markets in international trade. Journal of International Economics 48(1): 
7–35.
Rauch JE and Watson J (2004) Network intermediaries in international trade. Journal of Economics and 
Management Strategy 13(1): 69–93.
Smith D and White D (1992) Structure and dynamics of the global economy: Network analysis of interna-
tional trade, 1965–1980. Social Forces 70(4): 857–893.
Snyder D and Kick E (1979) Structural position in the world system and economic growth, 1955–1970. 
American Journal of Sociology 84(5): 1096–1126.
Uzzi B (1996) The sources and consequences of embeddedness for the economic performance of organiza-
tions: The network effect. American Sociological Review 61(4): 674–698.
Uzzi B and Lancaster R (2004) Embeddedness and price formation in the corporate law market. American 
Sociological Review 69(3): 319–344.
Zhou M (2011) Intensification of geo-cultural homophily in global trade: Evidence from the gravity model. 
Social Science Research 40(1): 193–209.
Zhou M and Park C (forthcoming) The cohesion effect of structural equivalence on global bilateral trade, 
1948–2000. International Sociology. doi: 10.1177/0268580912443577
Appendix: List of countries included in the analysis. 
Albania Cote d’Ivoire Indonesia Mozambique Sudan
Algeria Croatia Iran Nepal Sweden
Angola Cyprus Ireland Netherlands Switzerland
Argentina Czech Republic Israel New Zealand Syria
Armenia Dem. Rep. Congo Italy Nicaragua Tajikistan
Australia Denmark Jamaica Niger Tanzania
Austria Djibouti Japan Nigeria Thailand
Azerbaijan Dominican Rep. Jordan Norway Togo
Bahrain Ecuador Kazakhstan Oman Trinidad & Tobago
Bangladesh Egypt Kenya Pakistan Tunisia
Belarus El Salvador Korea Rep. Panama Turkey
Belgium-
Luxembourg
Estonia Kuwait Papua New Guinea Turkmenistan
Benin Ethiopia Kyrgyzstan Paraguay Uganda
(Continued)
 at FACULDADE IMED on April 8, 2014cos.sagepub.comDownloadedfrom 
http://cos.sagepub.com/
http://cos.sagepub.com/
Kim and Skvoretz 143
Albania Cote d’Ivoire Indonesia Mozambique Sudan
Bolivia Fiji Lao P. Dem. Rep. Peru UK
Brazil Finland Latvia Philippines Ukraine
Bulgaria France Lebanon Poland United Arab 
Emirates
Burkina Faso Gabon Liberia Portugal Uruguay
Burundi Gambia Libya Rep. Moldova USA
Cambodia Georgia Lithuania Romania Uzbekistan
Cameroon Germany Madagascar Rwanda Venezuela
Canada Ghana Malawi Saudi Arabia Viet Nam
Chad Greece Malaysia Senegal Yemen
Chile Guatemala Mali Sierra Leone Zambia
China Guinea Mauritania Singapore Zimbabwe
China Hong 
Kong SAR
Guinea-Bissau Mauritius Slovenia 
Colombia Haiti Mexico South Africa 
Congo Honduras Mongolia Spain 
Costa Rica Hungary Morocco Sri Lanka 
Appendix: (Continued)
 at FACULDADE IMED on April 8, 2014cos.sagepub.comDownloaded from 
http://cos.sagepub.com/
http://cos.sagepub.com/

Continue navegando