Abstract. Key Words: Trade, ECOWAS, WAEMU, WAMZ, Gravity model, Panel Data. JEL Classification: F11, F15, C23

Transcription

1 Determinants of a successful regional trade agreement in WEST AFRICA By Sam Olofin, Afees Salisu, 1 Idris Ademuyiwa and Joel Owuru Centre for Econometrics and Allied Research Department of Economics University of Ibadan Ibadan, Nigeria Abstract In this paper, we evaluate the determinants of effectiveness of the Economic Community of West African States (ECOWAS) at promoting regional trade in West Africa between 1995 and We employ the modified gravity model (GM) that allows for the inclusion of country specific and country-pair characteristics in addition to the traditional GM variables (income and distance). Our findings reveal that economic size, distance, geographical factors such as common border, landmass, landlockedness of countries and socioeconomic variables like common language, political stability and availability of infrastructure significantly influence intra-regional trade within the ECOWAS region. We also find that the francophones dominated region (WAEMU) is export trade creating while the anglophones dominated region (WAMZ) is trade diverting. Therefore, for ECOWAS to be successful in terms of facilitating intra-regional trade, current efforts at forming a synergy between WAEMU and ECOWAS should take cognizance of promoting trade between members, irrespective of colonial origin. Key Words: Trade, ECOWAS, WAEMU, WAMZ, Gravity model, Panel Data JEL Classification: F11, F15, C23 1 Corresponding author, idris_ademuyiwa@yahoo.com; Mobile:

2 1.0 Introduction The last two decades have witnessed tremendous growth in regional trade agreements (RTAs) in different forms (ranging from free trade agreements to economic and monetary unions and economic partnership agreements) and scope (one off treaty based arrangement and management arrangements). In fact, as at 15 th of January 2012, about 511 notifications of RTAs had been received by the World Trade Organisation/General Agreement on Tariff and Trade (WTO/GATT). Of these, almost 90% were free trade agreements and partial scope agreements while custom unions accounted for the remaining 10%. This represents a significant increase when compared to about 400 agreements reported by Whalley (2006). The table below shows the number of RTAs reported by the WTO for 10 year period averages starting from Table 1: Trends in RTAs notification to GATT/WTO Periods of notification of GATT/WTA Period average of RTAs within the period January Source: Compiled and computed by authors from GATT/WTO website 2 Africa has not been spared of the proliferation of RTAs around the world. On average, each country in the continent belongs to at least four RTAs and the continent now has over 30 RTAs most of which are free trade agreements and economic integration agreements. Like other RTAs in other regions of the world, the objectives of RTAs in Africa include attainment of economic objectives like promotion of regional integration and trade; improvement of regional competitiveness; attraction of foreign direct investment (FDI) and non-economic objectives like conflict prevention and resolution and increasing the region s bargaining power in the multilateral front. The quest to meet the aforementioned objectives has also resulted in the formation of RTAs in West Africa. Precisely, the surge to achieve a successful regional trade agreement has continued 2 The RTAs reported includes free trade agreements, economic integration agreements and custom unions. Some agreements were revised on a future date after initial notifications but were not captured above to prevent a case of double counting.

3 to dominate the agenda of the established economic communities in the region (Salisu et al., 2012). RTAs in the region presently include the Economic Community of West African States (ECOWAS) which consists of 15 out of the 17 West African countries and is the only free trade agreement in the region and a prospective monetary and economic union. Others are the West African Economic and Monetary Union (WAEMU) which is a monetary and currency union with eight francophone countries as members and the CFA Franc as official currency, the West African Monetary Zone (WAMZ) which is also an aspiring currency union with six Anglophone countries as members and the Mano River Union (MRU) which is a cooperative agreement. Notwithstanding their forms, they are all explicitly concerned with the promotion of intraregional trade flows (Ogunkola, 1998) and are not immune from the challenges typical to RTAs in Africa. Africa has potentials to expand its trade (both intra-regional and extra regional trade) and increase its competitiveness so as to grow and attract the much needed foreign direct investment (FDI) through mutually beneficial RTAs. According to Oyejide (2003), trade-led growth can reduce rural poverty when it expands employment in small-holder agriculture and can lower urban poverty when it is associated with increased output and export of labour-intensive manufactures. In other words, the fact that RTAs should inherently be beneficial to Africa is not in doubt as there exists a consensus as to the fact that regional integration efforts and schemes act as avenues for battling the different challenges facing the African continent as a whole. Also, increased linkages among African countries, through an expansion of intra-regional trade, can be a crucial device for creating the necessary growth spillovers and fostering the regional takeoff (Longo and Sekkat, 2004). Despite these facts, the performance of RTAs in Africa especially in the attainment of the economic objectives has been relatively disappointing. 3 For instance, as reflected in the table below, there has been no significant improvement in Africa s contribution to world export in the past four decades. 3 Unlike Africa, the contribution of other regions like Asia has been on the increase during the same period. This is partly due to the increasing competitiveness of the region and its transition to production of value added goods via improved technology.

4 Table 2: Trend in Africa s share in world export Years World Export (in US Million Dollars) Africa Export (in US Million Dollars Source: Compiled and computed by the Authors from UnctadStat (2011) Percentage of African Export in World Export Further, studies have revealed that the impact of RTAs on intra-african trade have been very small (see Yang and Gupta, 2005) especially in comparison with its extra-regional trade (see Cassim, 2001). Further, econometric analyses have also confirmed that regional integration in Africa (through various regional schemes) has been a failure (see Elbadawi, 1997). Similarly, Ogunkola (1998) found intra-ecowas trade to be too low while according to Bundu (1995), after so many years of establishment, no significant impact has been felt on development in West Africa even though countries have adopted the different programs of ECOWAS. 4 A peculiar problem perceived to be hindering the success of RTAs in West Africa in particular is the presence of different agreements and a possible overlap between them. This overlap may result in duplication of responsibilities, potentially conflicting commitments and the waste of already scarce resources (Robert, 2004). Stemming from these problems, some salient questions come to mind. They include; what factors can facilitate the effectiveness of a regional trade agreement in West Africa? Currently, how has the overlap aforementioned affected intraregional trade? While the closest effort at examining the consequent effect of this overlap on trade is by Musila (2005) who focused on the intensity of trade creation and diversion in COMESA, ECCAS and ECOWAS as a whole, we are of the opinion that an inward looking analysis for some RTAs within West Africa in particular will be more insightful. In the present study, we therefore seek to add to the existing literatures in two different ways. First, we extend the gravity model (GM) to capture the determinants of an effective RTA in ECOWAS especially in order to domesticate and update the findings by Longo and Sekkat (2004) on the role of political stability, infrastructure and economic policy in intra-african trade. Second, since two sub-regions have remained prominent in ECOWAS namely WAEMU and 4 Bundu Abass is a former Executive Secretary of ECOWAS.

5 WAMZ, we further examine the possibility of trade creation or diversion between these subregions within ECOWAS. Therefore, apart from the conventional specification that highlights the main determinants of bilateral trade, this study adopts a Vinerian-type gravity model specification with three dummies per FTA with a view to capture the trade creation and trade diversion effect of each RTA. We expect that the presence of trade diversion within the region will reveal a lack of synergy among the RTAs in the region and a possible factor responsible for the setbacks earlier mentioned in West Africa while the presence of a trade creation will prove otherwise. Foreshadowing our main results, we find that economic size, distance, geographical factors such as common border, landmass, landlockedness of countries and socioeconomic variables like common language, political stability and availability of infrastructure significantly influence intra-regional trade within the ECOWAS region. We also find that the francophones dominated region (WAEMU) is export trade creating while the Anglophones dominated region (WAMZ) is trade diverting. Nonetheless, the current effort to synergize WAEMU and WAMZ sub-regions is expected to enhance bilateral trade in ECOWAS. This paper is organized as follows. Section 2 presents some background analysis and stylized facts about the trade relation of ECOWAS member countries. Relevant empirical and theoretical studies are reviewed in section 3. Section 4 describes the methodology employed while section 5 presents the result of estimations carried out. Section 6 concludes the paper. 2.0 Stylized fact about Trade in ECOWAS ECOWAS is a common market and therefore it trades with other regions in the continent, the other RTAs in the world and individual countries (both developed and developing). In this section, we analyze the trends in ECOWAS trade across different periods i.e , and Thus, trends in intra-ecowas trade, ECOWAS trade with other regions in Africa, ECOWAS trade with developing and developed countries and the major RTAs in the world between the periods aforementioned are discussed. Further, it also compares the share of intra-ecowas and extra-ecowas trade in total trade by the region.

6 2.1 Trends in ECOWAS Intra-regional Trade 5 Intra-regional trade in ECOWAS in terms of export, import and consequently total trade can be described as being biased towards a number of countries. Of the total export within the region, Nigeria has the highest percentage between the periods 1995 to 2010 (see table 3 in the appendix). In fact, the country contributed about 36 percent of the total export in the period 1995 to 2000 and this increased to about 40 percent in the period 2006 to The reason for Nigeria s dominance cannot be unrelated to her position as a major crude oil exporter in the region. Closely followed is Cote-d'Ivoire, which accounted for about 35 percent of the total intra- ECOWAS export in the period 1995 to 2000 though it decreased to about 28 percent in the 2006 to 2010 period. Senegal also contributed a relatively high proportion of intra-ecowas export while other countries in the region accounted for less than 5 percent with Gambia and Guinea contributing the least. Cote-d'Ivoire dominated intra-ecowas import by receiving about 23 percent of the regions import between the period 1995 and Also significant is the proportion of import by Ghana, Mali, Nigeria and Burkina Faso with each of them accounting for about 19 percent, 12 percent, 10 percent and 9 percent respectively. For all these countries except Burkina Faso, their intra- ECOWAS import share for 2006 to 2010 reflects an increase relative to their 1995 to 2000 values. That a country like Nigeria does not contribute as much to total import in the region as it does to export confirms the earlier position on the effect of its leadership in crude oil production and reflects its retrogression in terms of commodity trade. On the aggregate, Cote-d'Ivoire accounts for the highest percentage of intra-ecowas trade with almost 29 percent in the periods 1995 to 2000 though this contribution fell to about 26 percent in the 2006 to 2010 period. Expectedly, Nigeria by virtue of its role in intra-ecowas export, seconded with a share of about 24 percent of total trade between 1995 and 2000 which increased to about 26 percent in the 2006 to 2010 period. Other drivers of intra-regional trade in ECOWAS are Ghana and Senegal which contributed about 13 percent and 8 percent between 2006 and 2010 respectively. 5 This section on intra-regional trade excludes four countries namely Liberia, Guinea Bissau, Sierra Leone and Cape Verde because of lack of sufficient data.

7 2.2 Trends in ECOWAS Trade with Developed and Developing Countries ECOWAS trade with developed and developing countries, though biased towards the developed countries, revealed an interesting finding. Of the region s total export, the developed countries accounted for about 68 percent in the 1995 to 2000 period only to reduce to 63 percent in the 2006 to 2010 period. Consequently, developing countries share of ECOWAS export increased from about 32 percent in 1995 to 2000 period to 37 percent from 2006 to For imports, during the periods 1995 to 2000, the developed countries contributed almost 63 percent to ECOWAS import while the developing countries accounted for about 37 percent. However, the trend changed in the period 2006 to 2010 when the share of developing countries increased to about 52 percent and those of the developed countries reduced drastically to about 48 percent. In other words, the developing countries overtook the developed countries during the period 2006 to On the aggregate, while developed countries accounted for about 66 percent of total trade with ECOWAS in the 1995 to 2000 period, the developing countries accounted for 34 percent. This margin closed up in the period 2006 to 2010 as the percentage changed to 57 percent and 43 percent for developed and developing countries respectively. The interesting revelation here is that developing countries are perhaps making efforts to promote trade within their regions. Table 4 (in appendix) and Figure 1 below depict the analysis above. Figure 1: Percentage Share of Developing and Developed Countries in ECOWAS' Total Trade ( ) 120,00 100,00 80,00 60,00 40,00 20,00 0,00 Developing economies Developed economies Source: Graphed by the authors Export Import Total

8 2.3 Trends in ECOWAS trade with African regions Apart from intra-regional trade and trade with developed and developing countries, another insightful categorization of ECOWAS trade relation is those that relate to other regions in Africa. Middle Africa accounted for the larger share of exports from ECOWAS for the period 1995 to 2000 with about 46 percent of the total. Other regions like the Northern and Southern Africa also received a high percentage of ECOWAS export of about 26 percent and about 24 percent respectively. In the period 2006 to 2010 however, this observed trend changed as Southern Africa led other regions with about 58 percent of the region s total export while Middle Africa which had hitherto gotten the highest dropped its share to get about 34 percent. For ECOWAS import, Southern Africa s share increased from about 48 percent in the period 1995 to 2000 to about 54 percent in 2006 to 2010 period while that of Northern Africa dropped from about 28 percent to about 14 percent in the same periods. While most of ECOWAS total trade is with the Southern and Middle Africa, this analysis revealed that ECOWAS trades only on a very small scale or magnitude with the Eastern region of the continent. The sudden rise in the trade relation between ECOWAS and Southern Africa requires specific emphasis especially starting from the turn of the decade. This is perhaps the result of aggressive initiatives by the Southern Africa continent to synergize with other regions of the continent through initiatives like NEPAD (The New Partnership for African Development). Table 5 (in appendix) and Figure 2 below show the results discussed above. Figure 2: Percentage Share of African Regions in ECOWAS' Trade ( ) 120,00 100,00 80,00 60,00 40,00 20,00 0, Southern Africa Northern Africa Middle Africa Eastern Africa Source: Graphed by the authors Export Import Total

9 2.4 Trends in ECOWAS trade with other major RTAs in the world ECOWAS trade relation with 16 of the major RTAs in the world again reflects a lopsided pattern in favor of those in Europe, America and Asia as against those situated even in Africa. In the period 1995 to 2000, the European Union (EU) and the Asia-Pacific Economic Cooperation (APEC) accounted for about 23 percent each of ECOWAS total export to RTAs while the Free Trade Area of the Americas (FTAA) and the North American Free Trade Agreement (NAFTA) received about 20 percent and about 17 percent respectively. Other RTAs contributed below 5 percent. The same story cannot however be told when their respective contributions in the period 2006 to 2010 are examined. Precisely, APEC and FTAA dominated with about 22 percent each and NAFTA s share of ECOWAS export also increased to about 19 percent while the share of EU dropped drastically to just 15 percent. Perhaps this is a reflection of how the global financial crisis has constrained import demand for African goods by the EU countries especially in addition to their domestic crisis. A similar trend is noticeable in the pattern of import by ECOWAS from these RTAs. For instance, the proportion of import from EU dropped from about 38 percent in the 1995 to 2000 period to about 26 percent in the 2006 to 2010 period while those of APEC, FTAA and Asia- Pacific Trade Agreement (APTA) rose from about 23 percent, 10 percent and 7 percent to about 25 percent, 11 percent and 15 percent respectively. On aggregate, in terms of ECOWAS total trade with the major RTAs, one noticeable trend is the continuous lost of contribution by the EU and the increase in the share of APEC and FTAA. Perhaps this implies that the European market is losing its trade linkages with West Africa despite colonial affiliations while trade ties between ECOWAS and the duo of the Asian and American markets is waxing stronger in recent times. These deductions can however be subjected to empirical validation. Table 6 (in appendix) and Figures 3 to 5 reiterates the analysis done here.

10 Figure 3: Percentage Share of the major RTAs in ECOWAS' Export ( ) 80,00 70,00 60,00 50,00 40,00 30,00 20,00 10,00 0,00 Export Export Export Source: Graphed by the authors Figure 4: Percentage Share of the major RTAs in ECOWAS' Import ( ) 120,00 100,00 80,00 60,00 40,00 20,00 0,00 Import Import Import Source: Graphed by the authors Figure 5: Percentage Share of the major RTAs in ECOWAS' Total Trade ( ) 80,00 70,00 60,00 50,00 40,00 30,00 20,00 10,00 0,00 Total Total Total Source: Graphed by the authors

11 2.5 Comparison of Intra-ECOWAS and Extra-ECOWAS trade 6 Expectedly, the share of extra-ecowas trade in total trade is higher than that of intra- ECOWAS trade since ECOWAS is a small region relative to the rest of the world (which are invariably its trading partners). However, the main aim of this sub-section is to see if the share of intra-ecowas trade has increased significantly within the period of study. The result as depicted in table 7 (in the appendix) and figure 6 below shows that there has not been a significant increase in the share of intra-ecowas trade since In other words, one may infer that ECOWAS trade has been more outward looking than inward. Figure 6: Percentage Share of Intra-ECOWAS and Extra-ECOWAS trade in Total Trade ( ) 120,00 100,00 80,00 60,00 40,00 20,00 0,00 Share of Extra-ECOWAS Share of Intra-ECOWAS Export Import Total Source: Graphed by the authors In conclusion, five major stylized facts can be drawn from the analysis done so far on ECOWAS trade relations. First, intra-regional trade in ECOWAS has been biased towards a number of countries namely Cote-d'Ivoire, Nigeria and Ghana while other countries in the region contribute relatively little to trade within the region. Second, in terms of ECOWAS trade with developed and developing countries, the region trades more with the developed world although the gap appears to be closing up in recent times (i.e. from 2006 to2010). Third, the Southern African region increasingly dominates trade with ECOWAS in terms of intra-african trade and it is seconded by Middle Africa which has been losing its contribution overtime. Fourth, out of the 16 6 It is important to note that the total value of intra-ecowas trade (both export and import) used here excludes those of Guinea Bissau, Liberia, Sierra Leone and Cape Verde because of lack of sufficient data. However, the results obtained are not expected to differ significantly from those in this section since this countries exert little influence on the total (precisely less than 10% of the total)

12 major RTAs constituting ECOWAS major trading partners, EU, APEC and FTAA play a more significant role. However, while the share of EU has been declining, those of APEC and FTAA have been increasing. Finally, there has been no significant increase in the ratio of intra- ECOWAS trade to its total trade (i.e. both in terms of intra and extra-ecowas). 3.0 Literature Review 3.1 Empirical Review Empirical research done in areas related to those of this study is multi-facet with different objectives. While some studies have examined trade flows, trade potentials and prospects within other regions of the world (see Fillipino and Molini, 2003; Papazoglou, 2007; Bhattacharya and Bhattacharya, 2007; Athukorala, 2012 among others as noted by 7 Kepaptsoglou et al., 2010) and in Africa (see Ogunkola, 1998; Cassim, 2001; Musila, 2005 and Salisu et al, 2012), some have attempted to understand the determinants and predict the formation of regional trade agreements (RTAs) (see Baier and Bergstrand, 2004; Egger and Larch, 2006; Jayathilaka and Keembiyahetti, 2009 and Chen and Joshi, 2010). Also, apart from studies that have attempted to advance the methodology used in modeling international bilateral trade like Baier and Bergstrand (2004), Baier and Bergstrand (2007), Carerre (2006) Martinez-Zarzoso et al. (2009) to mention but a few, in the past decade, a relatively larger number of studies have focused on investigating the impact of RTAs on regional trade and welfare especially in terms of their tendency to divert or create trade (see Ghosh and Yamarik, 2004; Martinez-Zarzoso, 2004; Musila, 2005; Carrere, 2006; Baier and Bergstrand, 2007; Jugurnath et al., 2007; Magee 2008; Abott et al., 2008; Egger et al., 2008; Martinez- Zarzoso et al., 2009; Hur et al., 2010;Egger and Larch, 2011 and Foster et al., 2011) while some have examined the success factors of existing RTAs (with Vicard, 2009 playing a significant role in this respect). Owing to their policy implications and relevance to the present study, the results from some recent studies on the effect of RTAs on regional trade among member countries worth 7 For the sake of intellectual transparency, it is salient to note that this section gains extensively from the brilliant work done by Kepaptsoglou et al., 2010 especially in getting access to literatures and opinions that were hitherto beyond the reach of the author.

13 examining 8. Nobel Laureate Jan Timbergen (1962) was the first to publish an empirical study in this regard. He found that the effect of RTAs on trade within the British Common Wealth members (Benelux Free Trade Agreement, FTA) to be insignificant accounting for about only a 5 percent increase. Afterwards, results have been mixed, at best (Baier and Bergstrand, 2007). Having noted that FTA variables are endogenous and countries opt into RTAs perhaps for reasons unobservable to econometricians and possibly correlated with trade, Baier and Bergstrand (2007) applied advancements in econometric analysis to estimate the average treatment effect of FTAs on bilateral trade flows between the periods 1960 to 2000 for 96 countries. However with the impression that there is lack of a suitable instrumental variable to replace the RTA dummies, they used a theoretically motivated panel gravity model to carry out their analysis. The result reveals that, contrary to previous studies, the effect of the FTAs on trade flows is quintupled and on average an FTA approximately doubles two members bilateral trade after 10 years. The negation of their results to previous ones cannot be unrelated to their treatment of RTA variables as endogenous. In fact, according to Ghosh and Yamarik (2004), treating RTAs as exogenous variables result into over or under-statement of their estimates. Carrere (2006) also adopted a gravity model specification to assess the ex-post regional trade agreement effect of seven RTAs in 130 countries for a period of 1962 to Again, like Baier and Bergstrand (2007), he made provision for endogeneity of the right hand side variables and unobservable characteristics of each pairs of partner countries. He applied a more generalized panel specification of gravity model derived by Baier and Bergstrand (2002). The result showed that the RTAs have generated a significant increase in trade though often at the expense of the rest of the world. This reflects a clear case of trade diversion. In a related study, Jugurnath et al., (2007) used pooled data for a five year period average between 1980 and 2000 to ascertain if five different RTAs within the Asian-Pacific Region have been trade creating or trade diverting. The analysis which involved 26 countries gave varying result across the RTAs. The result within the Association of South East Asian Nations (ASEAN) and the Australia-New Zealand Closer Economic Relation (CER) showed that there exists trade 8 The authors deliberately selected the reviewed studies bearing in mind the fact that they adopted different methodologies. So we attempt to compare results when different methods were used.

14 creation within the RTAs and with the rest of the world. However, the results for the Asian Pacific Economic Cooperation (APEC), the Southern Cone Common Market (MERCOSUR) and the North American Free Trade Association (NAFTA) indicate that they tend to be trade diverting at the expense of the rest of the world. Using the LINKAGE dynamic Computable General Equilibrium (CGE) model developed by the World Bank for evaluating the determinants of trade flows and policy, Lee et al., (2009) explores the consequences of FTAs among the ASEAN +3 (China, Japan and Korea) and ASEAN +6 (China, Japan, Korea, India, Australia and New Zealand) in terms of economic welfare, trade flows and sectorial outputs. After simulations for the period 2008 to 2015, the results revealed that Singapore, China and other ASEAN countries would realize a relatively large welfare gain while the impact in the European Union (EU) and North America are negligible. Egger and Larch (2011) used a structural analysis of bilateral trade flows informed by the Anderson and Van Wincoop s (2003) framework to examine the effect of the European Agreements enacted in the 1990s on the 15 EU incumbent economies and 10 potential entrants located in Central and Eastern Europe. In a panel data analysis of 167 countries for the periods 1990 to 2001, they found that the agreement resulted into a positive effect on trade in goods. Although the EU s GDP was not responsive to the agreements, the effect on welfare was moderate and the overall effect was better for the potential entrants in Central and Eastern Europe. Finally, it is important to note that other studies including, but not limited to Soloaga and Winter (2001); Musila (2005); Lee and Park (2007) have found RTAs to have positive effect on trade (see kepaptsoglou et al., 2010 for more reviews on this). As rightly noted by Kepaptsoglou et al., (2010) and Baier and Bergstrand (2007), recent evidence do not provide a clear cut evidence as to whether RTAs have been trade diverting or trade creating. Notwithstanding, recent findings have showed that making provision for the endogeneity of the RTA variables in gravity models improves the predictive power of the model and prevent the probable problem of under and overstatement of estimates. Perhaps the negligence of this by earlier studies account for the results they derived which showed that RTAs have negative effect on trade and welfare of countries and regions. It is important to note however that the issues on RTAs and bilateral trade relationship have remained inconclusive (Salisu et al., 2012).

15 On the determinants of a successful regional trade agreement, the results from econometric analyses (usually with the gravity model) have shown that different factors can be linked to the effectiveness of RTAs. Studies have found factors such as the trading partners Gross Domestic Product (GDP), transportation cost (usually proxied with bilateral distance), sharing of a common border, common language, GDP per capita, landmass or area, population of partners among others to be significant to bilateral trade. The table 8 below brings to focus some of these studies; their primary objectives and the major factors found to be significant (except the traditional variables which are normally significant in almost all studies). Table 8: Results of Selected Empirical Studies (in terms of significant determinants) Year Authors Objective Significant Explanatory Variables 2001 Cassim Examine the potentials for trade and the main determinants of trade in the Southern Africa. Common border, Common language, Area, RTA dummy( for SADC) 2004 Longo and Sekkat Examine the impact of infrastructure availability, economic policy and internal political tensions on intra-african trade. Border, Infrastructure, Economic policy, Internal political tension, Colonial ties Carrere Uses a gravity model to assess ex-poste regional trade agreements covering seven RTAs Jugurnath et al. Determine if five RTAs (namely ASEAN, CER, APEC, MERCOSUR and NAFTA) have been trade creating or trade diverting Magee Estimate the effect of regional trade agreements on trade flows controlling for country pair, importer-year and exporteryear fixed effect Vicard Investigate the determinants of the effectiveness of RTAs in enhancing bilateral trade within GATT/WTO Martinez et al. Evaluate the effect of preferential agreements on trade group members and non-members using a static and dynamic model Lohmann Uses the newly constructed language barrier index to test the significance of language barriers to bilateral trade Source: Compiled by the authors Border, Landlockedness, GDP per capita, Infrastructure, Exchange rate, Remoteness, RTA dummy Population, Area, Exchange rate, Tax, Common Language, 4 RTA dummies Population, Adjacency, common language, Area, Landlocked, Colony Contiguity, Common language, Common Colonizer Border, Common language, Island Border, Common language, Landlocked, Area, Island

16 3.2 Methodological Review Different approaches have been adopted in modeling trade flows and effects of RTAs on trade and economic welfare. According to Kepaptsoglou et al., (2010), these approaches can be divided into two. On one hand, we have the simulation models which have to do with replicating trade flows endogenously and examining its impact on bilateral trade, economic growth and welfare. On the other hand, econometrics methodologies have been adopted to make projections and ex-ante forecast based strictly on past and actual data rather than endogenously derived figures. The simulation approach comes in the form of the input output models and the Computable General Equilibrium model (both dynamic and static). The Computable General Equilibrium (CGE) models are advantageous to the extent that they allow for flexibility in modeling (e.g. inculcating the role of transportation cost and tariffs) and extensions beyond some basic assumptions in trade theories. However, the World Bank Economic Prospects (2005) noted that the endogeneity of parameters could be a problem since parameters have no definite statistical property. Some prominent works have adopted this methodology and they include but not limited to, Frankel et al., (1996), Frankel (1997), Baier and Bergstrand (2004), Bond et al., (2004) and Lee et al., (2009). Baier and Bergstrand (2004) developed the first econometric model that predicts FTAs based upon a CGE model of world trade with two factors of production, two monopolistically competitive product markets, and explicit inter-continental and intra-continental transportation costs among multiple countries on multiple continents unlike previous studies that used the CGE models. Lee et al., (2009) on their part adopted the World Bank s LINKAGE CGE model which had been used extensively for comparative analysis of trade integration scenarios and evaluation of the cost and benefit of different RTAs. The model is a recursive dynamic model built around the circular flow of income which has 3 inputs combined using CES technology under a perfect competition assumption. On the demand side, the model assumes the use of differentiated products for both final consumers and the producers on the basis of their origin. Unlike Baier and Bergstrand (2004) who used the CGE model to derive an econometric model, they conducted their study by first carrying out a baseline scenario analysis before carrying out policy experiments to examine the effect of FTAs on the ASEAN +3 and ASEAN +6 countries.

17 Empirical studies for modeling trade and impact of RTAs on trade have been carried out (especially in recent studies) with the use of the gravity model. The reason for the extensive usage and adoption of the gravity model in international trade research in the last few decades cannot be isolated from the robustness and explanatory power of the model. There is no gain saying the number of studies that have adopted this methodology is numerous as it is readily evident in the fact that Kepaptsoglou et al., (2010) reviewed over 75 papers on the gravity model in the last decade (i.e. from year 2000 to 2010) alone 9. However, a number of these recent studies have either used cross-sectional or panel data analysis and the debates inherent in the choice of such methodology worth review. Empirical efforts by Soloaga and Winter (2001), Feenstra et al., (2001), Porojan (2001), Musila (2005), Anderson and Van Wincoop (2003), Baier and Bergstrand (2009) among others as reviewed in Kepaptsoglou et al., (2010) and Salisu et al., (2012) employed the cross-sectional gravity analysis while others like Longo and Sekkat (2004), Martinez-Zarzoso (2004), Carrere (2006), Bhattacharya and Bhattacharyay (2007), Jugurnath et al. (2007), Magee (2008), Martinez-Zarzoso et al. (2009), Vicard (2011) and Athukorala (2012) used the panel analysis. Unlike the panel data framework, the cross-sectional analysis has been widely criticized for its inability to account for time effect, country effect and bilateral trade effect. For instance, Carrere (2006) noted that: discussions about the proper specification of the gravity model have shown that the conventional cross-sectional formulation without the inclusion of country specific effects is misspecified and so introduces a bias in the assessment of the effects of RTAs on bilateral trade. Therefore, he adopted a more general gravity model specification developed by Baier and Bergstrand (2002) to provide for this problem. Further, he showed that the predicted effects of RTAs in terms of trade creation and trade diversion are significantly different when a crosssectional analysis is carried out relative to a random effect panel specification that controls 9 Therefore, a review of studies that used the gravity model will be too voluminous and unnecessary. However, efforts have been made to look at studies on cross-sectional and panel gravity models and the rationale and controversy behind their adoption.

18 unobservable country specific characteristics. In short, he showed that a panel data analysis enables one to isolate the country-pair effects. Although some cross-sectional study like Matyas (1997) and Soloaga and Winter (2001) attempted to provide for these bias by introducing three country specific effects namely exporter, importer and time effects, Carrere sees this as only a restricted version relative his model. Baier and Bergstrand (2007) also noted that a standard problem with cross sectional empirical works is the potential endogeneity of the right hand side variables and the endogeneity bias inherent in them. According to them, the sources of such bias could be omitted variables, simultaneity or measurement error. However, they believe that a more likely cause is the omitted variable bias and that in the use of cross-sectional data, the standard econometric technique to address the problem is estimation via instrumental variables. Ultimately, they posit that three main alternatives techniques for addressing this problem of presence of unobserved time variant heterogeneity in cross-sectional estimation is estimation of a random effect panel, fixed effect panel or differencing the data and using OLS. On this, Egger (2002) is of the opinion that although both the fixed effect and the random effect panel are useful, the random effect should be employed when there is adequate consistency and there is an interest in estimating time invariant effects; otherwise the fixed effect is preferable. Although debates emanate from the choice of using random or fixed effect, Kepaptsoglou et al. (2010) observes that most recent studies prefer fixed effects as their panel correction technique. Other contributors to this debate include, but not limited to Baltagi et al. (2003), Longo and Sekkat (2004), Jugurnath et al. (2007), Magee (2008), Vicard (2011), Athukorala (2012), who made efforts to argue for the inclusion of time effects, country effcets and country pair effects. These effects are easily controlled for in panel data estimations. Perhaps this account for the observation by Kepaptsoglou et al. (2010) that panel data analysis are used in most studies for periods of at least the past 5 years and only few recent studies rely on cross-sectional analysis for drawing their conclusions. Although comparative studies have been undertaken to examine the level of bias imminent in cross-sectional studies, we opine that conducting a similar study in which the problems inherent in the cross-sectional analysis are controlled for and comparing the results with those of an equivalent panel analysis will no doubt be revealing and insightful.

19 Following Jacob Viner (1950), some studies have also attempted to divide the total effect of belonging to the same RTA or not on trade into a trade creation or trade diversion effect. The focus of such studies which include Solaoga and Winter (2001), Martinez-Zarzoso (2004), Carrere (2006), Jugurnath et al. (2007), Magee (2008), Martinez-Zarzoso et al. (2009) among others, is to compare the magnitude of each of these integration effects so as to ascertain if RTAs are trade creating or trade diverting on the aggregate. Apart from the aforementioned, other modifications to the gravity model unknown to most will not come as a surprise as the model exhibits a high level of flexibility, though Anderson and Van. Wincoop (2003) noted that not all of such re-specifications have theoretical foundations. 10 Despite the success of the gravity model in analyzing international trade flows, it acceptability was marred by its lack of theoretical foundations. This, perhaps account for the neglect of the model from the late 1960s to the late 1980s. As noted by Carrere (2006), the model has notwithstanding acquired a second youth partly due to its recent extensive use to study trade patterns and mainly because justifications and theoretical explanations have been developed for it based on international trade theories like Hecsher-Ohlin theory of trade and increasing return theory among others (see Anderson, 1979; Bergstrand, 1985; Helpman and Krugman, 1980; Deardorff, 1998; Baier and Bergstrand, 2001; Evennet and Keller, 2002; Anderson and Van Wincoop, 2003 among others for a succinct review). The paper by James Anderson in 1979 titled a theoretical foundation for the gravity equation is probably the first notable effort at such. Having reviewed some of the major methodological issues relating to estimating the effect of RTAs on trade across regions of the world and the adoption of the gravity model in particular, this study therefore seeks to make some contributions to the ongoing debate. 4.0 Methodology 4.1 Model Specification and Estimation Technique Although it has its origin in physics and precisely in the postulates of Sir Isaac Newton s Law of Universal Gravitation, the gravity model has no doubt earned itself a near universal acceptance as it has been applied to a range of academic disciplines (including geography and 10 For instance, Bergeijk and Brakman, (2010) shed light on some advancements and applications of the gravity model in recent times.

20 sociology). In international economics, the gravity model has become the main tool for estimating the determinants, patterns and effects of bilateral trade since it was first adopted by Nobel Laureate Timbergen (1962) and Linnerman (1966). Particularly, the model has gained popularity with analyzing the effects of RTAs on trade and welfare in different geographical and economic regions of the world. In this respect, a gravity model involves regressing bilateral trade on a series of explanatory variables, then using dummy variables to ascertain whether this relationship is affected by the existence of RTAs (Jugurnath et al., 2007). These explanatory variables include the traditional variables of the model namely economic sizes (usually proxied with GDP) and transaction cost (usually represented with the distance between trading partners) and other variables that have been incorporated into the model overtime. According to Head (2003), the model in its conventional form can be expressed as below;!!"# =!!"#!!!!!!" (4.1) Given the multiplicative form of equation (4.1), the model can be re-specified in a log-linear form after taking the natural logs as below;!"!!"# =!!!"!!" +!!!"!!" +!!!"!!"# +!!!"!!"# +!!"# (4.2) Where the inclusion of the!!"# makes it estimable with OLS and!!"# is bilateral trade between countries i and j and!! and!! are the GDPs or economic size equivalent of countries i and j respectively.!!" represents bilateral distance between the two countries while!!! denotes remoteness of the trading partners from the rest of the world. Other variables have been incorporated in different studies notwithstanding to augment the afore-stated variables though most of them lack theoretical justification as noted by Anderson and Van Wincoop (2003). Given the strong empirical prowess and theoretical acceptance of the gravity model for the analysis of bilateral trade flows, this study adopts the model. In line with the main thrust of the study, two gravity models are specified and estimated. The first model is estimated with the aim of examining the determinants of an effective RTA within ECOWAS countries while the second incorporates the effects of RTAs in West Africa on trade in a procedure similar to those of Carrere, (2006), Jugurnath et al., (2007) and Martinez-Zarzoso et al., (2010). It is the authors belief that for ECOWAS to be the successful and effective

21 regional arrangement it aspires to be, there must not be any form of trade diversion within the region. Stemming from equation 4.2 above, the first model is specified as below; ln!"#$%&!"# =!! +!!!"!"#!" +!!!"!"#!" +!!!"!!"# +!!!"!"##$!" +!!!"!"##$!" +!!!"#$%#!"# +!!!"#$%"$&!"# +!!!"#$!%&'($!" +!!!"#$!%&'($!" +!!"!"!"#!!" +!!!!"!"#!!" +!!"!"!"#$%!" +!!"!"!"#$%!" +!!"!"!"#$#%!" +!!"!"!"#$#%!" +!!"!"#$%&'!" +!!"!"#$%&!!" +!!"# (4.3) This specification allows for the inclusion of both country specific characteristics (like landlockedness, area, economic policy, infrastructural development and political stability) and country pair characteristics (like border and language). The dummy for landlocked is operationalized such that it takes up the value of one if a county is landlocked and zero otherwise while the dummies for border and language take the value of one if the trading partners share a common border or common language and zero otherwise. The infrastructure variable is computed as an average of road length per capita and number of telephones per capita while the political stability variables are gotten from different indicators. The economic policy variable used is the flow of FDI into the country as this is believed to reflect, to a large extent, the level of confidence of rational investors on the economy. Apriori expectation dictates that economic size of trading partners is positively related to trade between them so that!! and!! are expected to be positive. The distance variable is a proxy for transportation cost and therefore higher the distance is expected to mean an increase in transportation cost and consequently a reduction in bilateral trade (so!! < 0). Countries with a common border and language are expected to trade more with one another base on this level of affinity, so we expect that!!,!! > 0. Landlocked countries do not readily have access to the sea and consequently transaction cost are relatively greater and trade is often debarred, hence!!,!! should be negative. Availability of infrastructure and a stable economic policy is expected to promote trade and hence!!",!!",!!" and!!" are anticipated to be positive. Also, political instability and absence of violence is expected to foster trade so that!!" and!!" to be positive. GDP per capita is often used to proxy the level of development of trading partners. Therefore, a high level of development in either partner should result into higher trade, so we expect!! and

22 !! to be positive. The sign of the coefficient of area is indeterminate as revealed by Jugurnath et al., (2007). In order to examine the trade creation or diversion impacts of RTAs within ECOWAS, RTA dummies for WAEMU and WAMZ are added to equation 4.3 above. Therefore the resulting gravity equation can be specified as below;!"!"#$%&!"# =!"#!"#$% (4.3) +!!!!!"!!!!"#!"!"#!" +!!"!!!!"#!" +!!"!"#!" (4.4)!!!. 11 According to Marinez-Zarzoso et al. (2009), the gravity model represents a good counterfactual to identify the effects of an RTA, since it suggests a normal level of bilateral trade for a given sample and dummies are used to capture above or below normal levels of trade resulting from an RTA. The!"#!" will take the value of 1 if the exporting country is a member of the trading bloc k and 0 otherwise. Table 9 below provides cursory information about the interpretation of the RTA dummies. Positive coefficient of!"#!" (i.e.!!" > 0) implies that!"#! is trade creating as it shows that members exports to non-members are higher than the level obtainable in the absence of an agreement. In the same vein, when!!" is greater than zero, it implies that members imports from non members are higher than their normal level. These are indications that!"#! is trade creating. The integration dummy!"#!"!"#!" will be one if both source and reporting countries are members of the same RTA and zero otherwise. Therefore, a positive value of!!" will imply that the extent to which members of the!"#! trades with one another is higher than the level obtainable in the absence of an agreement and hence this tells us if intra- RTA trade has increased or decreased relative to those obtainable without the formation of such RTA. Table 9: Interpreting Static Integration Effects Coefficients Extra-bloc Imports(!!" ) Exports(!!" ) Intra-bloc Sign The language dummy in equation 4.3 has been dropped in subsequent estimations involving RTAs to prevent a case of multicollinearity.

23 !!" + Pure TC(M) TC+MD (!!" >!!" )or MD (!!" <!!" ) Pure TC(X) TC+XD (!!" >!!" ) or XD (!!" <!!" )!!" - ME MD + MC XE XD + XC Source: Martinez-Zarzoso et al. (2009) Note: TC denotes trade creation in terms of imports (M) or in terms of exports (X), MD and XD denotes import and export diversion, respectively, ME and XE denotes extra-bloc import and extra-bloc export expansion respectively, and MC and XC denote intra-bloc import and extra-bloc export contraction respectively. Explicitly, equation (4.4) above can be re-specified as below in the context of WAEMU and WAMZ as the existing regional agreement of some forms within ECOWAS.!"!"#$%&!"# =!"#!"#$%!"!" 4.3 +!!"!"#$%!"!"#$%!" +!!"!"#$%!" +!!"!"!"#!" +!!!!"#$!"!"#$!" +!!"!"#$!" +!!"!"#$!" (4.5) This study makes provision for the likely impacts of efforts geared at having a synergy between ECOWAS as a whole and WAEMU. Precisely, these efforts were initiated with the signing of a cooperation agreement between the duo in Ouagadougou (Burkina Faso) in 2003 and later in Abuja (Nigeria) in These agreements were aimed at developing a common plan of action on trade liberalization and macroeconomic policy convergence. In order to take account of these developments, equation 4.5 is estimated in different period variants. The first is for the period 1995 to The second variant covers the period before the agreements were properly internalized being from 1995 to 2004 while the third variant covers the post agreement period which is 2005 to We estimate the two models noted above (i.e. equations 4.3 and 4.5 where equation 4.3 is our model 1 and equation 4.5 is our model 2). We employ the Least Square Dummy Variable (LSDV) approach of fixed effects to estimate models 1 and 2. The LSDV approach is relevant in this case as earlier mentioned as it allows for the inclusion of dummy variables to capture both the country specific and country pair characteristics. Ignoring these specific effects when in fact they exist in the trade model may lead to bias result and misleading inferences (see for example, Baltagi, Egger and Pfaffermayr, 2003 and Carrere, 2006).