The topology of African exports: Emerging patterns on spanning trees

Highlights

• From empirical data, we describe African export linkages through a network approach.

• Limitations of bilateral trade data are avoided by the definition of bipartite networks.

• Two independent networks are defined: a destination share network and a commodity share network.

• Network analysis is complemented with the construction of the corresponding MSTs.

• Emerging topological patterns uncover some stylized facts of African trade.

Abstract

This paper is a contribution to interweaving two lines of research that have progressed in separate ways: network analysis of international trade and the literature on African trade and development. Gathering empirical data on African countries has important limitations and so does the space occupied by African countries in the analysis of trade networks. Here, these limitations are dealt with by the definition of two independent bipartite networks: a destination share network and a commodity share network.

These networks–together with their corresponding minimal spanning trees–allow to uncover some ordering emerging from African exports in the broader context of international trade. The emerging patterns help to understand important characteristics of African exports and its binding relations to other economic, geographic and organizational concerns as the recent literature on African trade, development and growth has shown.

Introduction

A growing literature has presented empirical findings of the persistent impact of trade activities on economic growth and poverty reduction  [1], [2], [3], [4], [5], [6]. Besides discussing on the relation between trade and development, they also report on the growth by destination hypothesis, according to which, the destination of exports can play an important role in determining the trade pattern of a country and its development path.

Simultaneously, there has been a growing interest in applying concepts and tools of network theory to the analysis of international trade  [7], [8], [9], [10], [11], [12], [13], [14], [15]. Trade networks are among the most cited examples of the use of network approaches. The international trade activity is an appealing example of a large-scale system whose underlying structure can be represented by a set of bilateral relations.

This paper is a contribution to interweaving two lines of research that have progressed in separate ways: network analysis of international trade and the literature on African trade and development.

The most intuitive way of defining a trade network is representing each world country by a vertex and the flow of imports/exports between them by a directed link. Such descriptions of bilateral trade relations have been used in the gravity models  [16] where some structural and dynamical aspects of trade have been often accounted for.

While some authors have used network approaches to investigate the international trade activity, studies that apply network models to focus on specific issues of African trade are less prominent. Although African countries are usually considered in international trade network analysis, the space they occupy in these literature is often very narrow.

This must be partly due to the existence of some relevant limitations that empirical data on African countries suffer from, mostly because part of African countries does not report trade data to the United Nations. The usual solution in this case is to use partner country data, an approach referred to as mirror statistics. However, using mirror statistics is not a suitable source for bilateral trade in Africa as an important part of intra-African trade concerns import and exports by non-reporting countries.

A possible solution to overcome the limitations on bilateral trade data is to make use of information that, although concerning two specific trading countries, might be provided indirectly by a third and secondary source. That is what happens when we define a bipartite network and its one-mode projection. In so doing, each bilateral relation between two African countries in the network is defined from the relations each of these countries hold with another entity. It can be achieved in such a way that when they are similar enough in their relation with that other entity, a link is defined between them.

Our approach is applied to a subset of 49 African countries and based on the definition of two independent bipartite networks where trade similarities between each pair of African countries are used to define the existence of a link. In the first bipartite graph, the similarities concern a mutual leading destination of exports by each pair of countries and in the second bipartite graph, countries are linked through the existence of a mutual leading export commodity between them.

Therefore, bilateral trade discrepancies are avoided and we are able to look simultaneously at network structures that emerge from two fundamental characteristics (exporting destinations and exporting commodities) of the international trade. As both networks were defined from empirical data reported for 2014, we call these networks destination share networks (${\mathrm{DSN}}_{14}$) and commodity share networks (${\mathrm{CSN}}_{14}$), respectively.

Its worth noticing that the choice of a given network representation is only one out of several other ways to look at a given system. There may be many ways in which the elementary units and the links between them are conceived and the choices may depend strongly on the available empirical data and on the questions that a network analysis aims to address  [17].

The main question addressed in this paper is whether some relevant characteristics of African trade would emerge from the bipartite networks above described. We hypothesized that specific characteristics could come out and shape the structures of both the ${\mathrm{DSN}}_{14}$ and the ${\mathrm{CSN}}_{14}$. We envision that these networks will allow to uncover some ordering emerging from African exports in the broader context of international trade. If it happens, the emerging patterns may help to understand important characteristics of African exports and its relation to other economic, geographic and organizational concerns.

To this end, the paper is organized as follows: Section  2 presents the empirical data we work with, Section  3 describes the methodology and some preliminary results from its application. In Section  4 we present further results and discuss on their interpretation in the international trade setting. Section  5 concludes and outlines future work.