Preferential Trade Agreements and Trade Liberalization Efforts in the OIC Member States
With Special Emphasis on the TPS-OIC
37
gravity model, in contrast to a general equilibrium model, results are typically presented
aggregated (although they might be obtained from very disaggregated flows). This means that
is possible to obtain the inferred average effect of a tariff in the flow between two partners; but
it is more difficult to determine the effect of a tariff in a particular product. It is possible to
estimate a gravity model for particular products; some complications related to the
identification of the demand variable arise.
An additional distinction applies between gravity models and general equilibrium models (and
also partial equilibrium models as we will see). Whilst CGE models use existing or past data to
make forecast with respect to the value of some variables of interest as a result of a change in a
control variable (a tariff, for example); gravity models are mostly used to evaluate ex post the
effects that policies already have had on the variables of interest. This means that gravity
models are a tool for ex-post analysis rather than ex-ante. By controlling the influence of other
factors, it is possible to identify the effect that an FTA had on the trade between two partners.
A comprehensive review on the application of gravity models for international trade flows and
trade agreements can be found in (Kepaptsoglou, Karlaftis and Tsamboulas, 2010), and also in
Head and Mayer (2013) In their review of empirical findings across 55 different studies using
gravity models between 1999 and 2009, the results are particularly mixed with some studies
claiming no evidence at all of trade diversion or trade creation whilst other claiming important
trade effects. For example, (Endoh, 1999) finds evidence of trade diversion in the context of
the Latin American Free Trade Association, suggesting that being member of this agreement
(controlling for the other variables) reduced imports from the Rest of the World by 22% in
1994 and (Soloaga and Winters, 2001) some evidence of trade diversion in the EU and the
EFTA. On the other hand, (Peridy, 2005) finds important evidence of trade creation (in the
order of 20%-27% of actual exports) in the context of the Euro-Mediterranean FTA.
Interestingly, (Abedini and Peridy, 2008) report important trade effects in the trade flows of
the regions belonging to the GAFTA agreement.
(Cernat, 2003) also uses gravity model to capture the effect of many South-South type of RTAs.
After controlling for the size of the pair of countries, the income per capita, distance, common
language and common border; he uses sets of dummy variables to identify exclusion and
inclusion of different RTAs. The author found that over the period 1994-1998 there was
substantial trade creation in CARICOM, COMESA, ECOWAS and SADC that offset the trade
diversion effects; a mixed result for the ANDEAN COMMUNITY (with imports from excluding
countries falling between 23% and 40%) and a clear domination of trade diversion in the case
of MERCOSUR (a result already found by Chang and Winters, 1999).
Head and Mayer (2013) also review a wide range of studies (over 150) and conduct a ’meta-
analysis’ from which they derive estimates of the average size of the standard coefficients
typically used in a gravity model. Table 1 below is taken from their paper and summarises the
key results. What can be seen from this table is that all the coefficients have the expected sign -
hence the higher is the GDP of either the exporting or the importing country, the higher is
trade; the greater is the distance between countries the lower is trade, and so on. From the