Method

In the first approach, we aim at dividing the world according to the opposition between integrated core vs. disarticulated periphery. To achieve this objective, we have used graph methods which enable us to identify the integrated core, whose countries are well connected in the World system with relative symmetric relations with other countries. In contrast, peripheral countries are defined as having strong asymmetric relations with core countries and few linkages with other peripheral countries.

In the first step, we start from a matrix of flows 'country vs. country' (A') that we dichotomize to obtain a Boolean (or binary) matrix (B2). In fact, graph methods like 'K core' requires boolean matrix, in which 1 means the presence of a link and 0 its absence. Hence, it is very important to find the suitable threshold above which the link between two countries will be considered to be effective. Therefore, “we propose a solution based on the geometric mean of four different probabilities of relations between two countries i and j. Geometric mean is justified to avoid the effect of exceptional values and to give an advantage to couple of countries with symmetrical values of interaction, that are more likely to reveal strong links. The geometric mean of the four probabilities creates a symmetric matrix of intensity of relation with value comprised between 0 and 100%. The geometrical mean will introduce a selection in favor of linkage between core countries (which are supposed to be symmetrical) and reduce the effect of linkage between core and periphery (which theory supposes not to be symmetrical). After different tests on threshold, the 5% one appears finally as an interesting compromise as it keeps the most important connection between the different parts of the global core, but it reveals also the different subpart of this core and their associated periphery” (Grasland, 2010).
In the second step, we apply a graph method called 'K-cores' to the Boolean matrix (B1) which precisely separates countries with strong internal relations (core countries) from the other ones (periphery) (C'). This method allows finding cohesive but not complete subgraphs and provides a clear division between central and peripheral nodes (Wasserman & Faust, 1994). Since In massive graphs like those we are studying, it's often difficult to find some cliques but only very cohesive subgraphs, “K-core” method is the most appropriate graph method here.

On the graph of Foreign Direct Investments (FDI), we can notice that, at the 5% threshold we opted for, very few links and nodes are present in comparison with the graph on trade. It is due to the extreme concentration of foreign direct investments into a limited number of countries. Indeed, the regionalization of foreign direct investment with the K core method shows a 9 members core which include USA and West European countries. It highlights the importance of intra-European and transatlantic flows of investment. In contrast with the spatial pattern of trade, Asian countries have a much more marginal position than in the graph of trade. But we must insist on the existence of links between tax havens situated in the Caribbean Islands, Mauritius or Cyprus toward many countries, especially Asian ones. Hence, it might be that some Asian countries are not included in the graph for FDI because some of their investments inwards go through financial intermediaries.

On figure above, we have synthesized these analyses. We have taken into account the average of a standardized “coreness value” between trade and FDI K core' regionalizations. In red, we find the most central countries (USA and West European countries) well interconnected for trade and investments; in dark orange, we recognize an important aggregate of countries which are part of the core or close to it according to trade but not for FDI. This class groups together regional economic powers (Russia, India), cohesive regional area (South American countries, Eastern Asia) and countries well connected to one of the two parts of the maximal sub cliques (Central European countries, some of the Southern East Asian countries and Canada). The two yellow classes are the most marginal according to economic flows. However the darkest one put together countries which form a relatively cohesive area like Middle East around Saudi Arabia and UAE, Southern and Eastern Africa, Balkanic countries or Central America. The light yellow' countries are the less cohesive and the most marginal, they have indeed very few links with countries more integrated than they are (eg. North Korea with China, Bangladesh with India, West African countries with European countries).

Regionalisation of World Trade according to a Core-Periphery Approach (2004-2006)

We will now divide the World countries according to the countries’ position in the division of labour. In order to achieve this objective, we have first launched a principal component analysis on the coefficient of asymmetry of all countries (or blocks) of the CHELEM database for each CITI product in the period 2004-2006. The first factor explains about 1/3 of the total variance (32.8%) and clearly highlights the opposition between countries having a strong positive asymmetry in manufacturing goods and those which have a strong positive asymmetry in raw materials. The second explains 15.1% of the variance and distinguishes the countries with a positive asymmetry of light manufacturing industry especially textile industry (in blue) from those with a positive asymmetry on high technology goods like chemical products or machine tools (in green). These two factors have been completed by a third one in order to exceed the threshold of 50% of explained variance. It opposes countries which have a strong positive asymmetry on the food industry, either raw or transformed, from to the others. In total, the three factors account for 55% of the total variance.

In a second step, we have launched a hierarchical ascendant classification with the Ward's method on the scores on the three first factors of the PCA in order to regionalize the World countries. Each factor score has been weighted in function of its part of the variance in the PCA. After several essays, we have kept the threshold of 75% of the explained variance. It results in a four group classification for 2006 (Figure 6). The countries in red are the core members (Western Europe without Ireland and Iberic peninsula, Scandinavia without Norway and Iceland, USA, Japan, Korea, Taiwan and Singapore). They have a strong positive factor score on factor 1 and 2. It means that there are mainly exporters of industrials products, especially high technological content. The countries in orange are what we have called the 'semi periphery integrated of the core' or 'semi periphery I' because they also sell industrial goods but in a less massive way than the core countries (Canada and most of the European countries except Balkanic ones). The yellow class aggregate exporters of light manufacturing goods and agricultural products, which we identify as 'semi periphery II'. Most of the East and South Asian countries are part of this class as well as Mexico, Brazil, Colombia, Tunisia and Turkey. The countries in blue are the most peripheral ones because there are mainly raw materials producers and industrial products buyers.

Core Periphery Countries' Classification around 2005

Up to now, we obtain two classifications of countries according to the criteria derived from the theory to distinguish between core and periphery. The first approach has distinguished between countries which form a cohesive and symmetrically interconnected area and the other ones. The second one has allowed classifying core and peripheral countries according to their position in the international division of labour. For each regionalization of the World, we have divided the countries into four classes. By crossing both classifications, we potentially have sixteen different classes. However, as can be seen on Figure 7, only nine cases are filled up (D). This is because, as stated by the theory, there is a relationship between both types of classification even if the discrete classifications we opted for here does not allow testing more in depth this relationship. In simple words, it means that the higher the position in the international division of labour, the more central and integrated the country is in trade or investment flows.

The dark red class is clearly the most central. Core countries (USA, France, Great Britain, Germany, Italy, UEBL, Switzerland and Netherlands) constitute the integrated core. The red class has the same characteristics than the dark red one in the international division of work but is a bit less marginal than the core in terms of directions of economic flows (Spain, Finland, Sweden, Denmark, Japan, Korea and Taiwan). However, this distinction is certainly not essential in qualitative terms. Among the semi periphery I identified in the second approach, some countries are less integrated than other ones but none of them are part of the most cohesive group. In orange, we have mentioned the countries which are a little bit in the margin of the core in the international division of work but directly integrated to it. It is the case of Canada, Norway, Ireland and Central European countries. The yellow countries have a more marginal position within the "semi periphery" (Israel, Portugal, Eastern European countries). There are industrial producers and sellers but are clearly in the margins of the to core countries. The green classes gather countries which are light industrial goods producers and sellers. They might be considered as the factories of the world as long as consumption and low technological products are concerned. Some of these countries are relatively well connected to the core or are part of a relative cohesive group (in dark green), while others are dominated by one economic partner (light green). The dark green class aggregates South-East Asian countries like China, Indonesia or Singapore, South American countries (Brazil, Colombia) and Turkey. Most of them are important economic actors and are in an intermediate position between core countries and their geographical neighbours. In many cases thus, these countries form a cohesive area of exchanges with a relative autonomy from the core countries. Light green' countries only include four countries: Mexico, Tunisia, Vietnam and Sri Lanka. Each of them is strongly related to a big economic power and has a strong asymmetry in term of exchanges with it. Finally, the blue tone classes are the most peripheral ones in the international division of work. They mainly sell primary products and have negative balances for most of manufacturing goods. In dark blue, Russia, Australia, Argentina, Chile, Uruguay, Kazakhstan, Honduras and Guatemala are grouped together. They are primary and food producers but there are relatively well connected either because they are strong economic powers (Russia, Australia) or because they are part of a cohesive regional area (Southern America, CEI). The intermediate blue class presents the same pattern but with smaller economies or in smaller cohesive regional areas (Eastern and Southern Africa, Middle East). In light blue, we find the most isolated and marginal countries (Western and Central African countries and small marginal countries or Caribbean Islands).