The literature on determinants of multilateral tradeflows, accounting for the impact of formation of WTO use gravity model. Apart from extended gravity variables like distance and per capita GDP of trade partners, colonial history, linguistic similar-ities, contiguity, etc., country specific fixed effects are accounted for following the suggestions of Anderson and van Wincoop (2003). Following Helpman et al.
(2008), the study will consider all possible bilateral pairs in the econometric exercise. Further, to examine the effects of WTO and other policy variables, some mutually exclusive and exhaustive dummies are taken into account to separate out the WTO membership effect. This follows Dutt et al. (2013). This study adds to the literature by taking into account variables measuring trade costs such as infras-tructure, tariff and non-tariff restrictions along with the variables generally con-sidered in the extended gravity model framework.
It is established in the existing literature that infrastructure is critical for eco-nomic growth and open economies can provide better infrastructure than its closed counter parts (WTO 2004; De 2004; Chakravorty and Mazumdar 2008). As an important determinant of transport cost, infrastructure is considered to be directly related to firm productivity, diversification in production and expansion in trade, increase in output and economic growth (Munnell1992; Easterly and Rebelo1993;
Limao and Venables 2001; De 2004, 2006). However, the effective rate of pro-tection in international trade flow provided by transport cost, in many of the countries, is considerably higher than that by tariff.3 Further, the lack of quality infrastructure creates obstacles to the supply chain of production and trading offinal goods as well.4For Asian countries, De (2004)finds the significant role of trans-action cost and infrastructure in explaining variation in trade flows in a gravity
3WTO (2004).
4Poor infrastructure has been identified as one of the responsible factor that restricts participation of the developing countries in the global value chain (WTO 2014).
model framework. Chakravorty and Mazumdar (2008) develop a model of trade in which provision of infrastructure makes the domestic firm more productive and enables to capture market share from a foreignfirm in both the domestic and export markets. Infrastructure is thus treated as a complementary factor to the trade policies for multilateral trade expansion and, hence, is included in the model.
4.1.1 Measure of Infrastructure
For the measurement of trade facilitating infrastructure (TFI) variable, principal component analysis (PCA) is used in this study. The variables used for measuring TFI across countries during 1995–2010 include volume of air freight measured in metric tonnes times kilometre travelled, flow of port containers from land to sea transport modes and vice versa in a standard size container (20-foot equivalent units), road density measured by the ratio of the length of country’s total road network to the country’s land area, volume of goods transported by railways measured in metric tonnes times kilometre travelled, internet users per 100 people, and telephone lines per 100 people. Data on these variables are collected from World Bank’s World Development Indicator (WDI) database at the country level.
For the sake of brevity, the weighting diagram of the variables included in the infrastructure index is not provided herewith.
4.1.2 The Econometric Model
In order to analyze the impact of formulation of WTO on exportflows, following Rose (2004a), Subramanian and Wei (2007), Tomz et al. (2007), Dutt et al. (2013), De (2004) among others, the following gravity equation is used
Xij¼ b0 Dbij1 Yib2 Yjb3 kbij4 cbij5 TFIb6 ð3Þ where i and j are trade partner countries. The gravity variables are defined as follows:
Xij is theflow of export from ith country to jth country;
Dij is the distance between source and destination country;
Yi is the per capita GDP of the source country;
Yj is the per capita GDP of the destination country;
kij represents the set of time invariant trade influencing geographical, cultural, and historical factors like linguistic similarities, colonial relationship, sharing common land border, etc., among trading countries. Linguistic similarity (Common language) is a binary variable which is unity if both the trading partners have common official language. Again colony (Colonial link) is a binary variable which is unity if i ever colonized j or vice versa, and contiguity (Border) is a binary variable which is unity if i and j share a land border;
cit represents the set of country specific issues like WTO membership, membership of a Regional Trade Agreement (RTA), beneficiary of a Preferential Trade Arrangements (PTA) including offering of GSP, etc., and
TFI is the measure of trade facilitating infrastructure variable for the trading nations
In log-linear form, the Eq. (3) can be expressed as
ln Xijt¼ b0þX Here t represents year, Ei, Mj and Tt represent the country and year specific dummies to have country a year specific fixed effects, and l is the residual dis-turbance term assumed to be well behaved.bi’s are the respective coefficients of the variables used in the regression.
In order to examine the effect of tariff and non-tariff barriers on multilateral trade flows, a measure of tariff and non-tariff barriers is considered for the estimation in a stepwise manner. First, to analyze the sole effect of these trade restricting barriers a set of estimations is carried out by dropping the WTO membership and related dummy variables from the Eq. (4). In that case, Eq. (4) can be represented as
ln Xijt¼ b0þX wheredij represent time variant trade restricting policy instruments like tariff and non-tariff barriers. The second step involves the estimation of the combined effect of tariff and non-tariff barriers along with WTO membership and related policy variables.
Separating Out the Effect of WTO
Following Subramanian and Wei (2007), Eicher and Henn (2011) and the decomposition of trade preferences as carried out by Dutt et al. (2013), the paper also accounts for WTO membership by the trading nations, RTA and PTA using a set of mutually exclusive and exhaustive dummies to estimate the pure WTO effect.
The dummies are as follows
1. Dummy for pure WTO effect (D1): The dummy will take the value 1, if both source and destination countries are members of WTO but they do not belong to a RTA and the destination does not offer any preferential arrangement to the exporter, and zero otherwise.
2. Dummy for pure RTA effect (D2): The dummy will take the value 1, if both the trading partners are members of a common RTA, but at least one of them is not member of WTO and the destination does not offer any preferential arrangement to the exporter, and zero otherwise.
3. Dummy for pure PTA effect (D3): The dummy will take the value 1, if the importer extends preferential arrangements to the exporter but at least one of them is not member of WTO and they do not belong to any RTA, and zero otherwise.
4. Dummy for effect without PTA (D4): The dummy will take the value 1, if both source and destination countries are members of WTO and the trading partners are members of a common RTA, but the destination does not offer any pref-erential arrangement to the exporter, and zero otherwise.
5. Dummy for effect without RTA (D5): The dummy will take the value 1, if both origin and source countries are member of WTO and importer extends prefer-ential arrangement to the exporter but they do not belong to any RTA, and zero otherwise.
6. Dummy for effect without WTO (D6): The dummy will take the value 1, if importer extends preferential arrangement to the exporter and both the countries belong to same RTA but at least one of them is not member of WTO, and zero otherwise.
7. Dummy for having all three effects: The dummy will take the value 1, if both origin and source countries are members of WTO, the trading partners are member of common RTA and importer extends preferential arrangements to the exporter, and zero otherwise.
The estimated coefficient of D1 will thus give the pure effect of WTO on export flows. The asymmetric impact of WTO, if any, can also be examined by separate set of estimations for developed, developing and least developed economies, which is however outside the scope of this paper.
4.2 Data Coverage and Sources
The econometric estimation considers a large dataset of about 5.2 lakh data points with 200 source countries and 234 partner countries for the period 1995–2010. The study includes all countries, developed, developing and least developed. For esti-mation of the model, data on merchandise export value (in US dollar term) are collected for all countries from World Integrated Trade System (WITS) database published by the World Bank. The export data are based on HS (1989–92) clas-sification. Data on per capita gross domestic product (PCGDP) at constant 2000 US
$ prices are collected form World Development Indicator (WDI) database. Data on
distance between the countries (in kilometre) are availed from the CEPII database.5 Other gravity variables including linguistic similarities, colony, contiguity, etc., are also taken from the CEPII database. WTO membership is used as determining factor for exportflows. Information on WTO membership, the Preferential Trading Arrangements (PTA) along with Generalised Scheme of Preferences (GSP), and Regional Trade Agreement (RTA) information are collected from WTO database (www.wto.org). Data on ad-valorem tariff and non-tariff costs of bilateral trading pairs are collected from ESCAP database on comprehensive trade cost.6