Analysis of Agri-Food Trade Structures
To Promote Agri-Food Trade Networks
In the Islamic Countries
8
not have an in-built concordance to SITC Rev.3, which means that an extra step would be
required to obtain the data required to construct the Annex 1 categories.
For these reasons, the analysis uses UN Comtrade as the primary data source in this report. The
data are extracted directly in SITC Rev.3 format, limiting the dataset to the products identified
in Annex 1. Fields are then added to the dataset to capture the aggregate groupings listed in
Annex 1, so that the data accord with the product categories. Data are extracted for the period
1995-2016, the most recent period for which data are widely available, in line with the objective
of this Study; 2017 data are not used as there is often a significant lag with country reporting to
UN Comtrade, so data for that year would likely be a major understatement of the true level of
trade. Taking all products, countries, and years together, the database totals over 8.2 million
observations (where an observation is a report of a bilateral trade flow at the SITC Rev.3 product
level).
International trade databases typically include data on the value of goods traded, but also
quantities traded. The analysis is exclusively in terms of values, as is typical of applied work in
international trade. The reasons are two-fold. First, units of quantity differ across products (e.g.,
kg for solids, liters for liquids) and so summing within the categories within Annex 1 would not
always give consistent results. Second, most trade policy measures applied to agri-food products
at the border are ad valorem tariffs, i.e. they depend on the value of the goods being imported,
not specific tariffs, which depend on quantity. As such, customs administrations have a strong
revenue-raising incentive to ensure that values are recorded correctly. They do not have that
same incentive in relation to quantities. As such, most trade economists regard quantity data as
much less reliable than value data, and almost all applied work in international trade is done in
terms of values. That is the approach followed here.
1.2. Empirical Methodologies
Much of the empirical work in this report is conducted by means of descriptive statistics.
Common approaches include calculating trends and growth rates, as well as breakdowns of total
flows by product categories. The evolution of agri-food trade is considered at the global level,
and then also among OIC regional groups. These methods are standard in the literature and
widely used in policy reports, and do not require a detailed explanation.
Two elements of the methodology require further explanation, as they are less well known
outside the academic community. The first is network analysis. This area is an emerging one in
terms of the literature.6 Network science and applied mathematics have developed a number of
general purpose tools to produce summary statistics that capture the position of nodes
(countries) in complex networks. Two are of particular policy relevance. The first is degree
centrality, which is a measure of the number of countries that each country is directly connected
to. In a matrix of export flows, it is the number of export destinations a country has. The policy
relevance of this measure is that it coincides with the idea of geographical export diversification,
which is of particular importance to developing countries looking to develop new trade flows.7
The second is eigenvector centrality. The rationale is that this measure captures the position of
6 See for example: Shepherd, B. (2017). “International Input-Output Linkages and Exogenous Shock Transmission: A Simple
Model and Evidence from the Global Financial Crisis.” Working Paper DTC-2017-3, Developing Trade Consultants.
and Shepherd, B., and L. Archanskaia. (2014). “Evaluation of
Value Chain Connectedness in the APEC Region.” Report Prepared for the APEC Policy Support Unit. Singapore: APEC.
7 Shepherd, B. (2010). “Geographical Diversification of Developing Country Exports.” World Development, 38(9): 1217-1228.