Keywords: Free Trade Agreement, Customs Union, European Partnership Agreement, GTAP, COMESA

A GTAP model analysis of Ethiopia-COMESA Free Trade Agreement, COMESA Customs Union, and European Partnership Agreement. Habtamu Shiferaw Amogne 1 Abstract Common Market for Eastern and Southern Africa (COMESA) is one of the eight building blocks regional economic communities recognized by the African Union (AU) for the establishment of Continental Free Trade Area (CFTA). This study focuses on analyzing the implication of Ethiopia- COMESA Free Trade Agreement (Et-COMESA FTA), COMESA Customs Union (CU), and European Partnership Agreement (EPA) on the economies of COMESA member countries. The study uses standard Global Trade Analysis Project (GTAP) model version 9 database. The result indicates that with full FTA among all countries new FTA member countries reported a large expansion in their GDP while Ethiopia experiences loss in GDP. Nevertheless, compared to full FTA, customs union expands the economies of most COMESA countries. By contrast, EPA shrinks most COMESA countries GDP with the largest loss goes to REA, Zimbabwe, and Kenya. This higher loss in GDP comes from a drop in domestic consumption and investment, although trade balance is positive for most countries. The trade effect on the economy of most COMESA country is mixed, but some countries reported a growth in their export and import. Et- COMESA FTA has positive trade effect for most member countries compared to a customs union. Furthermore, the export and import of most COMESA countries grow with EPA, but with Brexit, the trade effect is lesser. Finally, COMESA countries reported aggregate welfare gain with FTA while CU and EPA result in net welfare loss. EPA reduces the well-being of most COMESA countries, but world welfare improves due to substantial welfare gain from EU-27 and UK. This study concludes that full FTA among COMESA countries and EPA has significant trade effect for most COMESA countries. However, regarding welfare, FTA and CU is preferred to EPA. JEL Classifications: F11, F13, F15 Keywords: Free Trade Agreement, Customs Union, European Partnership Agreement, GTAP, COMESA 1 Corresponding Author: Habtamu Shiferaw Amogne; Economics Department, Kobe University, Address: Apartment 401, 2-22 Ichinosancho, Nada-Ku, Kobe, P.O.Box 657-0012, JAPAN, Tel. 090-8366 - 4964, FAX: +81(0)78-803 - 6877 E-mail: habtamushiferaw@stu.kobe-u.ac.jp. Acknowledgments: The author acknowledges Kobe University for every cooperation and buying GTAP database. My research advisor, Professor Taiji. Hagiwara for his extensive comments and suggestion. I thank Prof. Dr. Jong-Hwan Ko for very useful comments and advice during my presentation at PAPAIOS Conference. Finally, the author acknowledges Reynaldo Senra and Jane Uzunovski for their valuable comments, and help on the editing of the paper. I

1. Introduction Common Market for Eastern and Southern Africa (COMESA) was formed to promote intra-regional trade among member states with the ultimate objective of attaining sustainable growth & development of the countries, and realization of the objective of African Economic Integration (COMESA, 2009). It is one of the eight "building blocks" regional economic communities recognized by the African Union for the establishment of Continental Free Trade Area (CFTA) 2. In 2000, COMESA Free Trade Agreement (COMESA FTA) was launched, and currently, 16 of the 19 COMESA member countries are the signatory to FTA while the rest three non-fta members namely Eritrea, Ethiopia, and Swaziland are negotiating to sign FTA. 3 This FTA provides duty-free, quota-free market access to member States on COMESA originating products. Furthermore, in 2009 COMESA member countries launched custom union and agreed to levy Common External Tariff (CET) against non COMESA member countries. However, the implementation is still under discussion, and many countries did not adjust their national tariffs according to CET rate. There are ongoing debates on the macroeconomic and welfare impacts of regional trade agreements. According to Baier and Bergstrand (2007), FTA increases member countries bilateral trade. However, on welfare ground, a customs union is always Pareto-superior to FTA (Krueger, 1997). On the other hand, a study by Marvel and Karingi (2012), Mureverwi (2016), Yang and Gupta (2007) and Khandelwal (2004) states that trade agreements in Africa are not sufficient to increase intra-africa trade, and more liberalization is needed in Africa focusing on trade facilitation, non-tariff barriers, and strengthening domestic revenue base. The negotiations of non-fta member countries mainly Ethiopia is important. First, Ethiopia is the fourth largest economy in Sub-Saharan Africa (SSA), and the country has an abundant cheap labor force, a market of over 90 million people, and an enormous underexplored hydrocarbon potential. However, Ethiopian economy is highly protected compare to other member countries, and tariff duties range from 0% to 35%, with an average rate of 17%. Second, from geopolitics perspective, Ethiopia has a significant place in the Horn of Africa (Mesfin, 2012). Furthermore, more integration among COMESA member through reduction of protection is vital for the ongoing negotiation undertaken among Africa countries to form Continental Free Trade Agreement (CFTA) and boosting intra-africa trade. Therefore, any trade liberalization policy undertaken by COMESA countries through reduction of protection has large implication for the success of CFTA. The recent trade negotiation between the European Union (EU) and ACP countries, European Partnership Agreement (EPA) was initiated to ensure that the ACP-EU trade relation was compatible with WTO rules (Article XXIV). EPA provides a duty-free and quota-free market access to the EU with 2 These eight are namely: AMU, CEN-SAD, COMESA, EAC, ECCAS, ECOWAS, IGAD, and SADC. 3 The COMESA FTA member countries are Burundi, Comoros, Djibouti, D.R. Congo, Egypt, Kenya, Libya, Madagascar, Malawi, Mauritius, Rwanda, Seychelles, Sudan, Uganda, Zambia, and Zimbabwe. 2

improved rules of origin for ACP countries that have signed WTO-compatible agreements. Recently, nine COMESA member countries have signed interim EPA agreement while the rest are still negotiating. 4 Recently, there are an increasing number of studies analyzing the effects of EPA between ACP and EU. Many African policy-makers, business representatives, and NGOs argue that the EPA agenda is too broad and intrusive for African countries. Besides, with EPA bigger EU companies could flood the continent with cheaper products, destroying emerging local industries. Also, cutting tariffs will lower government revenues that Africa needs to invest in areas including agriculture, health, and education. According to Vollmer et al. (2009), the impact of EPA differs from country to country depending on trade relation between states and initial protection levied on import from EU. Also, full reciprocity with EU is very costly for Africa (S. Karingi et al., 2006). Furthermore, EPA results in budgetary difficulties as a consequence of the loss of trade tax revenue (Bilal and Roza, 2007). Therefore, it is very important to analyze the economic and welfare impact of EPA on the economies of EPA signatory COMESA member countries and EU. However, the United Kingdom (UK) decided to leave EU, Brexit. Therefore, considering this situation the EPA negotiation when UK is a member of EU and the case when there is Brexit will have different economic and welfare impact. As a result, this study analyzes the impact of EPA on cases when UK is a member of EU and when there is Brexit. This paper has three main objectives. First, to analyze the impact of Et-COMESA FTA with both partial liberalization and full liberalization among all COMESA countries. Second, to estimate the effect of implementing CU in an operational FTA on the economies of COMESA countries. Lastly, assessing the impact of potential FTA between COMESA countries and EU under EPA scheme. The last objective considers the situation where UK is a member of the EU and the situation after Brexit. The study mainly focuses on macroeconomics, trade, welfare and output effects of the liberalization policy. To achieve this objective the study uses (GTAP) model (Hertel and Thomas.W., 1997) (Version 9). Also, the baseline GTAP database is adjusted by including tariff changes made among COMESA countries after 2011. The rest of the paper is organized as follows: Section 2 presents an overview of COMESA economies. Section 3 presents the relevant empirical literature. Section 4 explains the model database and model simulations. Section 5 analyses the simulation results under different scenarios. Section 6 provides the systematic sensitivity analysis of the model result. Finally, Section 7 concludes. 2.1. Economic character of COMESA 2. COMESA Economy: A descriptive exposure Demographic changes across countries influence the level and composition of trade both through their impact on comparative advantage and on patterns of demand. As can be seen in Table 1, Ethiopia, Egypt, and D.R.Congo are most populated countries while Seychelles, Djibouti, and Comoros have a subtle 4 Out of 19 COMESA members, 17 are eligible for EPA agreement. Out of 17 countries, eight countries namely Comoros, Djibouti, Eritrea, Ethiopia, Malawi, Sudan, Zimbabwe, and D.R. Congo are negotiating to sign EPA, while the rest countries, Burundi, Kenya, Mauritius, Madagascar, Rwanda, Seychelles, Uganda, Zimbabwe, and Swaziland already signed interim EPA. 3

number of population. Besides, Comoros, Seychelles, Djibouti and Mauritius have small arable land while Sudan and Ethiopia have largest arable land hectares among COMESA member countries. Further, as shown in the last column of Table 1, most countries with abundant arable land have significant agriculture sector value added compare to industry and service sector. Table 1 further reports the relative size of economies of COMESA member countries using GDP. PPPbased GDP data in Table 1 shows that Egypt, Sudan, Ethiopia, and Kenya are the four largest economies among COMESA member countries. In addition, The GDP per capital of COMESA member countries varies widely and ranges from US$711.52 in D.R.Congo to US$ 25172.44 in Seychelles in 2014. The largest GDP per capital for Seychelles, Mauritius, and Libya signals the growth in the economy of these countries and tend to reflect an increase in productivity. Furthermore, Table 1 reports the trade-to-gdpratio measured by the sum of exports and imports divided by GDP. This indicator measures a country's 'openness' or 'integration' in the world economy. Some COMESA countries reported the significant trade to GDP ratio in 2014 (indicating an increasing openness). Trade constitutes 181.29 % of Seychellois's GDP, 114.57% of Mauritius's GDP, and 147.58 % of Libya's GDP. In contrast, Sudan, Egypt, Ethiopia, Burundi, Rwanda, and Uganda have a relatively small trade to GDP ratio of below 50% suggesting plenty of rooms increase openness. Country Arable Land (in Thousand Hectares) Table 1. Economic character of COMESA Population (in millions,2014) GDP (in millions US$, 2014) 4 Per Capital GDP (in US$,2014) Trade, (as % of GDP) (2014) Average Value added ( as % of GDP) (2010-2014) Agri. Ind. Svces Burundi 1200 10.8 7944.82 734.48 41.31 40.09 17.31 42.60 Comoros 65 0.8 1049.93 1363.56 79.92 38.24 11.57 50.19 D.R.Congo 7100 74.9 53238.84 711.52 80.06 22.76 33.71 43.53 Djibouti 2 0.9 2733.70 3120.04 n.a n.a n.a n.a Egypt 2738 89.6 900147.80 10045.78 37.41 12.34 38.37 49.29 Eritrea 690 5.1 n.a n.a n.a n.a n.a n.a Ethiopia 15119 97 138728.89 1430.8 40.74 44.84 11.50 43.66 Kenya 5800 44.9 126449.16 2818.26 51.12 29.18 20.36 50.46 Libya 1720 6.3 93133.61 14879.99 147.58 n.a n.a n.a Madagascar 3500 23.6 32308.91 1373.19 69.38 27.24 16.24 56.52 Malawi 3800 16.7 18611.30 783.83 73.40 31.07 16.20 52.74 Mauritius 75 1.3 22365.09 17730.90 114.57 3.40 25.11 71.49 Rwanda 1182.5 11.3 17975.00 1584.21 46.17 32.96 14.18 52.83 Seychelles 0.08 0.1 2303.93 25172.44 181.29 2.24 13.23 68.59 Sudan 17220 39.4 152767.42 3882.25 19.12 27.44 23.90 48.66 Swaziland 175 1.3 10039.74 7910.84 n.a 6.75 45.49 47.76 Uganda 6900 37.8 63831.94 1689.44 46.83 26.28 19.80 53.92 Zambia 3700 15.7 56946.17 3724.53 n.a 9.96 35.28 54.77 Zimbabwe 4000 15.2 26057.36 1709.14 79.56 13.38 31.12 55.50 (Note) n.a. = not available; GDP per capital is PPP in 2014 (constant 2011, international $). (Source) World Development Indicators (latest update, November 17, 2016)

The breakdown of average value added by activity has changed considerably across COMESA member countries over the period 2010-2014. Agriculture is a dominant sector with agriculture value added constituting more than 40% for Ethiopia and Burundi. On the other hand, industry sectors account for more than a quarter of GDP for Swaziland, Egypt, D.R. Congo, Zambia, Zimbabwe, and Mauritius. The value added of the service sector is greater than agriculture and industry sector for most COMESA member countries except Ethiopia. In general, Table 1 shows that the economic characteristics of COMESA member countries are diverse and more trade liberalization in these countries would have a varied effect. 2.2. Trade and protection pattern Table 2 below reports the import share of Ethiopia, COMESA, EU-27, and UK for GTAP aggregated sectors to reflect the existing trade relations among negotiating regions. On the other hand, Table 3 reports estimated average bilateral import tariffs levied by Ethiopia, COMESA, EU-27 and UK on one another's export. The primary source of data for both import and average bilateral tariff is GTAP 9 database, the base year 2011. Therefore, trade liberalizations among COMESA countries mainly FTA members after 2011 is not included in version nine database, and these tariff cuts are included as part of the baseline scenarios explained in Section 4.3. As can be seen from Table 2 below, Ethiopia's import of Petroleum & Chemical, food manufacturing, and service from COMESA member countries is high, but Petroleum & Chemical face a low average tariff. On the other hand, Beverage & Tobacco, Leather, Other Manufacturing, Forestry & Fishery, and Vegetable & fruit import by Ethiopia from COMESA countries face high average tariff but constitutes small import share. Therefore, a substantial expansion of imports by Ethiopia from COMESA member countries are expected on highly protected sectors following Et-COMESA FTA. Furthermore, COMESA countries have significant import share of vegetable & fruit, livestock, and other crops from Ethiopia, but face very low average tariff from Ethiopia. However, COMESA countries have low import share of Motor vehicle part, Forestry and Fishery, Beverage & Tobacco, Other Manufacturing, Wood Paper, and Petroleum Chemical from Ethiopia, but face a high tariff. Therefore, significant improvement in import by COMESA countries is expected in these sectors. The overall Ethiopia-COMESA trade share shows that COMESA countries mainly import agricultural products from Ethiopia while Ethiopia imports mainly petroleum and chemical products. Besides, the average import tariff levied by COMESA member countries from Ethiopia is small compared to Ethiopia's average import tariff on goods originated from COMESA member countries. Table 2 and three further reports the import share between COMESA and EU-27 region. Accordingly, Fabric Metal Equipment, service and petroleum & chemical sector constitute more than 60% of COMESA's import from EU-27. However, these sectors are moderately protected by COMESA member countries. On the other hand, Beverage & Tobacco, Food manufacture and other manufacturing, which have small import share from EU-27 face very high average tariff. Therefore, a massive expansion of imports from EU-27 to COMESA countries on Beverage & Tobacco, Food manufacture, and other manufacturing sectors is expected due to EPA. As can be seen from Table 2, Fabric Metal Equipment, service and petroleum & chemical industries constitute more than 70% of COMESA's imports from the 5

UK while the share of other sectors is subtle, and face a reasonable average tariff. On the other hand, beverage & tobacco, Food manufacturing, leather and other manufacturing sectors constitute a small share of import by COMESA countries from the UK but face a high average tariff. Therefore, with Brexit, small expansion of imports from the UK is expected on beverage & tobacco, food manufacturing, leather and other manufacturing sectors compare to the case when UK is a member of EU. Table 2. Composition of imports by source (Percentage Share of total import, 2011) Commodity Ethiopia import share from COMESA COMESA import share from Ethiopia COMESA import share from EU-27 EU-27 import share from COMESA COMESA import share from UK UK import share from COMESA Grains 0.00 0.87 5.27 0.07 0.20 0.19 Vegetable & Fruit 0.02 51.75 0.64 1.02 1.62 4.95 Oilseed 0.00 2.83 0.01 0.07 0.00 0.07 Other crops 0.21 10.29 0.33 2.86 0.01 7.08 Livestock 0.03 25.52 0.39 0.09 0.10 0.12 Forestry & Fishery 0.02 0.06 0.05 0.23 0.04 0.24 Coal, Oil, and Gas 0.53 0.00 0.50 56.94 0.29 21.87 Food manufacturing 4.18 1.22 6.30 2.69 2.04 5.63 Beverage & Tobacco 0.17 0.04 1.33 0.20 0.67 0.20 Textile & Apparel 0.84 0.09 1.20 2.85 0.72 9.27 Leather 0.11 0.30 0.18 0.44 0.12 0.30 Wood paper 1.80 0.09 4.69 0.35 2.04 1.83 Petroleum & Chemicals 81.65 1.60 18.94 12.39 10.75 10.99 Basic metals 2.70 0.01 6.62 2.39 3.25 0.34 Fabric metal equipment 3.19 1.56 25.44 1.13 30.28 3.70 Motor vehicle part 0.25 0.24 5.57 0.23 4.96 0.29 Other manufacturing 0.86 0.14 0.62 0.16 1.00 0.50 Services 3.41 3.35 21.93 15.87 41.92 32.42 Total import (US$ million) 441.59 542.56 78702.74 80961.01 8929.8 7398.27 (Source) GTAP database version 9 6

Table 3. Bilateral Tariff (Average Ad valorem tariff, 2011) Commodity Ethiopia average tariff on import from COMESA COMESA average tariff on import from Ethiopia COMESA average tariff on import from EU-27 COMESA average tariff on import from the UK Grains 2.25 0.30 4.38 8.04 Vegetable & Fruit 23.44 1.03 11.27 1.56 Oilseed 0.00 2.63 6.45 1.60 Other crops 10.32 1.13 6.14 8.99 Livestock 3.28 0.15 2.27 4.21 Forestry & Fishery 25.93 12.04 3.07 5.80 Coal, Oil, and Gas 5.05 1.86 2.26 2.14 Food manufacturing 15.24 6.94 20.01 27.68 Beverage & Tobacco 31.03 18.45 83.38 235.25 Textile & Apparel 20.39 5.68 11.67 13.63 Leather 30.21 8.29 15.00 25.10 Wood paper 13.73 11.37 6.39 8.62 Petroleum & Chemicals 6.61 10.10 11.42 7.32 Basic metals 8.75 5.86 2.58 4.43 Fabric metal equipment 16.49 8.71 7.78 5.25 Motor vehicle part 15.38 19.45 11.83 14.75 Other manufacturing 26.76 14.34 14.50 23.21 Services 0.00 0.00 0.00 0.00 (Source) GTAP database version 9 3. Literature Review The proliferation of regional trade blocks in Africa have appealed interest among academics and policy makers in Africa. Many studies have been done to analyze the effect of trade liberalization in COMESA and other regional trade agreements in Africa. However, the policy scenarios of trade liberalization measures, the period of assessment and the structures of the model employed vary among these studies. Some studies use a gravity model to analyze the trade liberalization effect while others use partial or general equilibrium models. Musila (2005) using gravity model examine the intensity of trade creation and trade diversion in COMESA, ECOWAS, and ECCAS. The result indicates that the strength of trade creation is higher for ECOWAS followed by COMESA while the trade diversion effect is weak in all regional trade agreements. Furthermore, the result re-enforces the idea that size factors (level of GNP and population), and resistance factors (distance and language) play an important role in the determination of flow of international trade. Similarly, Conroy (2013) analyze the impact of trade creation and trade diversion on COMESA FTA and MERCOSUR FTA using gravity equation. The study points out that both trade agreements have 7

significant trade creating effects while MERCOSUR FTA will have a modest level of trade diversion. Also, trade created in the COMESA FTA occurred largely in sectors in which countries have different comparative advantages, indicating that new links are efficient. Many studies like Balistreri et.al (2015), Karingi and Fekadu (2009), Willenbockel (2013), and Makochekanwa (2014) use CGE modeling to investigated the impact of Tripartite Free Trade Area (TFTA) on the economies of member states. The study by Balistreri et.al (2015) analyzes trade cost as the primary trade barrier in (TFTA) 5. The study found that deep integration among the three regions would produce substantial gain, but the estimated gain vary across countries and depends on the trade reform. On the other hand, Karingi and Fekadu (2009) and Willenbockel (2013) using GTAP model analyzes the economic and welfare impact of forming TFTA. The research found an overall benefit from establishing TFTA, but the regional level impact is unbalanced due to the difference in initial protection structure. Similarly, Makochekanwa (2014) using WITS-SMART model analyzes the welfare implication of TFTA and found that there is potential net trade gain, but the implementation of TFTA will lead to loss of tariff revenues, which contribute to a significant proportion of fiscal resources for most countries. Karingi et al.(2002) analyze the impact of implementing COMESA FTA and then forming the Customs Union on the economies of member countries using GTAP model. The study found that COMESA is better off with free trade. However, there is unbalanced benefit across member countries. Besides, FTA gives good outcomes but the customs union must be preferred, and the member countries benefit regarding real incomes and reduction of poverty from the customs union. Also, Dimaranan and Mevel (2008) using similar methodology examine the likely impacts of COMESA customs union and found that custom union results in expansion of trade but most COMESA countries report negative real income. These differences across countries are due to the heterogeneity of the COMESA economies regarding their economic structure, trade and protection patterns. Similarly, Sawkut and Boopen (2010), found that the global welfare increases with COMESA customs union and COMESA countries benefited from the increase, although not to the same degree from forming the Customs Union. In contrast, Nzuma et al.(2009) found that the proposed COMESA customs union will not be beneficial to a majority of the member countries. Besides, to benefit more from the Customs Union more liberalization is needed in the area of harmonization of customs procedures, non-tariff barriers, infrastructural improvements, diversification of production, and measures to include more cross-border transactions with recorded (formal) trade among others. There are currently few studies on the potential macroeconomic and welfare impact of EPA between COMESA countries and EU. A study by S. Karingi et al.(2006) examines the impact of EPA, between COMESA countries and EU on multilateral trade development using both general and partial equilibrium models. The general equilibrium result indicates that with full reciprocity trade relations and general welfare would register positive trends. However, these benefits can only be realized, at the cost of significant and extensive macroeconomic adjustments. Furthermore, the partial equilibrium result shows 5 TFTA is a free trade agreement among COMESA, East African Community (EAC) and Southern African Development Community (SADC). 8

that COMESA countries undergo a customs revenue loss on EU import and a certain level of trade diversion from their trade partners and other COMESA member countries. Besides, Vollmer et al.(2009) analyze the impact of EU-ACP European Partnership Agreement for Sub-Saharan Africa(SSA). The research found that some SSA countries like Botswana, Cameroon, Mozambique, and Namibia would significantly benefit from the interim EPA agreements, while the trade effects for Côte d'ivoire, Ghana, Kenya, Tanzania, and Uganda are close to zero. 4. Methodology The study employed a multi-country, multi-sector general equilibrium modeling approach. (WTO, 2012) States that a general equilibrium analysis explicitly accounts for all the links between the sectors of an economy households, firms, governments, and countries. It imposes a set of constraints on these sectors so that expenditures do not exceed income, and income, in turn, is determined by what the factors of production earn. These constraints establish a direct link between what the factors of production earn and what households can spend. The WTO document further states that the purpose of Computable General Equilibrium (CGE) simulations is to determine the effects of a change in trade policy on the endogenous variables of the model prices, production, consumption, exports, imports, and welfare. The simulation represents what the economy would look like if the policy change or shock had occurred. The difference in the values of the endogenous variables in the baseline and the simulation represents the effect of the policy change. Therefore, the model should be able to predict the effect on macroeconomic, trade, welfare and production patterns if the trade policy was changed. Furthermore, based on the change in welfare, the policy-maker would be able to judge whether the country benefited from the change in policy or not. The study uses the global economy-wide model known as GTAP model (Hertel, Thomas.W., 1997). The study uses the static GTAP model with standard macroeconomic closure to analyze the potential impact of regional integration on COMESA regions. The standard features of the GTAP model are perfect competition, Constant return to scale, Armington assumption in Trade flows, disaggregated import usage by activity, non-homothetic consumer demands and explicit modeling of international trade and investment. GTAP model has the advantage of overcoming the effects of policy changes, at national, bilateral or multilateral levels, on production levels, input factors, volumes of trade and other induced influences on welfare. Furthermore, GTAP model is centered on the reallocation of resources between the sectors of the economy; it is an appropriate instrument for identifying the sectors and countries, which gain or which lose with the change of policy induced by trade liberalization policy. The data used in this study is the version 9 of the GTAP database (Aguiar, Narayanan, and McDougall, 2016). The reference year for the database is 2011. 9

4.1. Regional and sectoral Aggregation The GTAP-9 database features 140 countries/regions and 57 tradeable commodities. In this study, the 140 countries/regions are mapped into 17 regions, and the 57 sectors are mapped into 18 sectors (Appendix I & II respectively). The GTAP 9 database identifies only 10 of the 19 COMESA member countries as a separate region while the other nine COMESA countries are aggregated into four GTAP composite regions. As these four GTAP composite regions are almost exclusively composed of COMESA countries, the regional aggregation structure of the database supports almost perfect analytical separation of COMESA and Non-COMESA regions 6. In addition, the aggregation allows a quite detailed analysis of changes in intra - COMESA trade flows, which takes explicit account of the bilateral trade flows among COMESA regions and their trade with the rest of the world. Furthermore, the regional aggregation includes three Non-COMESA regions, EU-27, UK, and ROW. In this study, UK is included as a separate region from EU to analyze the impact of Brexit on EPA negotiation. 4.2. Common External Tariff and Sensitive Product The COMESA customs union was established in 2009, and the member countries agreed to impose CET on non-member countries. The agreed-upon CET rates have four categories of commodities provided by product line in the Common Tariff Nomenclature (CTN). Accordingly, the CTN adopts a four-band classification where scheduled CET rates are 0% for raw material and Capital goods, 10% for intermediate goods and 25% for finished products. In the creation of a customs union, some of the initial tariff rates are higher than the recommended CET rate while in other cases they had to be raised to bring them to the CET rate. As a result, a customs union may reduce or increase protection. Therefore, the design of tariff changes from original levels to the CET rates for customs union scenario is calculated as follow. First, COMESA CTN, specified at the HS6 2007 classification is mapped to MAcMap-HS6 v.3 database. Second, I compute the net tax saving for each aggregated sector at HS code classification. 7 Third, the weighted average tax rate is calculated for the corresponding GTAP sectors. Finally, the weighted average tax rate is used as a shock value for custom union experiment. Accordingly, a negative weighted average shows an increase in protection while a positive value reflects a decrease in tariff protection. (Appendixes III) Under most trade agreements, member countries specify a list of sensitive products that are excluded from the sectors that will be liberalized. Countries often argue for eliminating key products from liberalization for reasons of national interests such as tariff revenue considerations, infant industries 6. The GTAP composition of REA includes Somalia and Mayotte besides to COMESA member countries; RSAC includes Lesotho besides to Swaziland; RSCA include Angola besides to D.R. Congo, and RNA includes Algeria and Western Sahara besides to Libya. 7 Net tax saving is the difference between actual MFN tax rate of COMESA countries and COMESA CET bound rate at HS6 2007 product classification. 10

argument, health issues, and a political and cultural importance of the sector. The framework of the COMESA customs union also allows for exclusions of sensitive products. However, at the time of writing this paper, only eleven countries submit a list of sensitive products. In addition, three of the eleven countries namely, Burundi, Rwanda, and Uganda, agreed to use a similar list of sensitive products with Kenya for the customs union. 8 Therefore, for the above COMESA countries, this study excludes the submitted list of sensitive products from CET calculation. However, for Ethiopia, Egypt, Zambia, Zimbabwe, REA, RNA and RSCA, I select sensitive products using import revenue criteria and exclude top 5% of goods from CET calculation. 4.3. Experiment Design This study begins with GTAP 9 database with the base year 2011, aggregated to the set of regions and sectors specified in section 4.1 above. A baseline scenario is created in this paper by updating the tariff component of the database. The new benchmark contains information on the policy changes, which includes mainly the reductions of duty among COMESA countries mainly from Uganda and D.R.Congo. Therefore, the baseline tariff is adjusted before simulation, but the results are not interpreted in this paper. The primary purpose of including this policy change is to develop realistic and actual policy scenarios for the free trade experiment. Furthermore, the customs union and EPA scenarios consider sensitive products provided by COMESA countries and are thus exempted from the CET. This paper has five different scenarios regarding tariff reduction between the various regions. Under each scenario, tariff among members of regional integration (FTA, CU or EPA) is removed but maintained for other regions. Scenario 1: Et-COMESA FTA (FTA-17). All bilateral ad valorem import tariffs between Ethiopia and COMESA FTA countries are removed. In this scenario, Eritrea and Swaziland are not the members of COMESA FTA. The main purpose of this scenario is to analyze the separate impact of Ethiopia s accession to COMESA FTA on the economies of COMESA member countries. Scenario 2: Full COMESA FTA (FTA-19). A complete removal of Ad valorem import tariff among all COMESA member countries. As Eretria and Swaziland are negotiating to join COMESA FTA, this scenario provides an estimate of the consequence of extending FTA to all COMESA countries. Scenario 3: COMESA Customs Union (COMESA CU). There is full FTA among all COMESA countries whereas CET rate is levied against all non-comesa countries. This scenario provides an estimate of the consequence of implementing the Customs Union by all member countries. 8 Refer 2011 Gazette, Volume 16 Annex 1: list of sensitive product for Kenya, Madagascar, Malawi, Mauritius, and Swaziland. http://www.comesa.int/wp-content/uploads/2016/06/2011gazette-vol.-16-annex-ii-ist-of-sensitive-products.pdf 11

Scenario 4: COMESA FTA with EU (EPA). There is a complete removal of ad Valorem import tariff among all COMESA countries and between EU-28 and COMESA countries. However, CET rate is maintained on imports from ROW. 9 In this scenario, UK is a member of EU. Scenario 5: COMESA FTA with EU after Brexit (EPA+Brexit). This scenario is similar to scenario four, but UK is not a member of EU. Therefore, CET rate is maintained on imports from UK and ROW whereas FTA is maintained among all COMESA and between COMESA and EU-27. Table 4 Experiment Design Scenarios FTA-17 FTA-19 COMESA CU EPA EPA+Brexit Integration COMESA-16 FTA + Ethiopia FTA-17 + Eritrea (REA) + Swaziland (RSAC) FTA-19 + CU FTA-19 + CU + EPA (EU-28) FTA-19 + CU + EPA (EU-27) Source: Authors scenario design. 5. Result and Discussion All scenario s results are designed as a variation of the baseline scenario. The analyses are comparative static; hence, they do not address potentially critical questions relating to the sequencing of reforms and potential dynamic benefits from trade liberalization. When bilateral tariffs are eliminated, relative prices change, and in response, trade flows between countries change, which eventually affect the resource allocations in the economy. It is expected that different sectors in the economy adjust their outputs according to relative price shifts. (Narayanan and Sharma, 2016) States that when an importer reduces tariffs on its partners, the degree of increase or decrease of imports from each of them would depend on two opposite effects. First, trade creation effect enabled by overall expansion in demand for cheaper imports. Second, trade diversion effect created by the expansion of exports by partners facing higher tariff reduction at the cost of others accomplished in terms of a response to price differentials. The following section explains the macroeconomic, welfare, and industry output impact of the different scenarios described above. 9 In this scenario, Egypt and Libya (RNA) are excluded since they are not a beneficiary of EPA. 12

5.1. COMESA FTA 1. Macroeconomic effect Table 5 indicates the macroeconomic impact of COMESA FTA under scenario one and two. When Ethiopia joins COMESA FTA, Kenya experiences the largest growth in GDP (0.27%) whereas Ethiopia s economy contract by 0.23%. However, the impact on other COMESA member countries is insignificant. Table 5 further reports that with full FTA among all COMESA countries (FTA-19), most COMESA member countries experience growth in their GDP compared to FTA-17. RSAC records the largest growth in GDP (3.99%) whereas Zimbabwe and Ethiopia reported contraction of the economy by 0.75% and 0.14% respectively. The main reason for the decline in the growth of GDP for Ethiopia is the reduction in the consumption of domestic commodities and investment, although there is an expansion of the trade balance in scenario one. This decline in domestic consumption is due to an increase in both export and import in both scenario one and two. Thus, there is a potential for change in the production and consumption structure; more of production is exported than in the base case, and more of consumption is imported from Ethiopia. Removal of import tariff among COMESA is expected to result in a significant improvement in the level and direction of trade among member countries by reducing the domestic market price of import. The reduction of import price results in a rise in demand for import by firms for intermediate goods, private households as well government for consumption. Table 5 reveals that most COMESA countries do not benefit from the import and export surge. Kenya and Ethiopia reported the largest increase in export and import with FTA-17 whereas the impact on other COMESA member countries is tiny. However, when the remaining COMESA member countries, Eritrea (REA) and Swaziland (RSAC), join COMESA FTA (FTA -19) the export and import of most COMESA member countries show improvement. As expected, new FTA member countries, Ethiopia, REA and RSAC witness large import growth when they eliminate tariff. The availability of cheap import reduces domestic production cost and increase competitiveness in these countries, resulting in increased export. However, the extent of the rise in both export and import depends on the relative change in price in different sectors driven by tariff reduction. As a result, there is a slight difference in the growth of export and import among COMESA countries. Trade balance of most COMESA countries deteriorates with FTA. Kenya, Egypt reported significant trade balance deficit in scenario one whereas RSAC and Mauritius reported massive deficit with scenario two. In contrast, Ethiopia records large trade balance surplus (US$9.24 million) in scenario one but worsens with scenario two. 13

Table 5. Macroeconomic impact of COMESA FTA scenarios. Region Import value (%) Export value (%) GDP (%) Change in Trade Balance (US$ Million) FTA-17 FTA-19 FTA-17 FTA-19 FTA-17 FTA-19 FTA-17 FTA-19 Ethiopia 0.22 0.5 0.68 0.95-0.23-0.14 9.24-5.95 Egypt 0.04 0.07 0.04 0.06 0.03 0.05-12.11-19.59 Kenya 0.26 0.39 0.15 0.43 0.27 0.21-29.19-25.86 Malawi 0.01 0.28 0 0.06 0.01 0.17-0.24-5.44 Madagascar 0-0.01 0 0 0 0 0 0.1 Mauritius 0 0.57 0 0.21 0 0.54-0.04-28.64 Rwanda 0.01 0.14 0.01 0.13 0.01 0.06-0.03-0.46 Uganda 0.01 0.17 0.01 0.15 0.01 0.1-0.12-1.64 Zambia 0 0.01 0 0 0-0.02 0.02-1.11 Zimbabwe 0 1.31 0.01 3.76 0-0.75 0.09 25.4 REA -0.01 0.73-0.01 0.72-0.01 0.09 0.28-16.53 RNA 0 0 0 0 0 0 0.01-0.22 RSCA 0 0.01 0 0 0 0-0.16-0.25 RSAC 0 7.67 0 3.62 0 3.99 0-46.51 EU-27 0 0 0 0 0 0 6.76 23.08 UK 0 0 0 0 0 0 0.83 0.36 ROW 0 0 0 0 0 0 24.67 103.27 (Source) Model simulation 14

Table 6 Change in GDP Components for New COMESA FTA Countries/regions (in US$ Million) GDP Component Consumption Investment Government Export (-)Import Total Ethiopia FTA-17-57.7-16.42-6.34 28.88-19.54-71.11 FTA-19-37.59 1.46-3.54 39.81-45.87-45.72 REA FTA-17-2.95-0.91-0.59-1.82 2.09-4.17 FTA-19 47.1 30.1 12.11 143.8-160.89 72.22 RSAC FTA-17 0 0 0 0 0 0.01 FTA-19 170.51 97.37 43.05 107.62-154.12 264.44 (Source) Model simulation 2. Welfare effect In GTAP model, the welfare changes are measured by equivalent variation (EV). This change is the amount of money consumer in any region would pay rather than face the changes in prices and quantities resulting from the simulations. The net welfare impact of tariff reduction depends on the relative sizes of trade creation, and trade diversion effects. Trade creation arises when more efficiently produced imported goods replace the relatively inefficiently produced domestic products by increasing import demand. On the other hand, trade diversion occurs when the sources of supply divert from the more efficiently producing non-member countries to the less efficiently producing member countries under the tariff-free access granted to signatory countries. Table 7 reported that Kenya and Egypt experience substantial welfare gain under scenario one. In contrast, Ethiopia suffers considerable welfare loss equivalent to US$ 13 million. The welfare effect for other COMESA member countries from FTA-17 is small, but the overall result shows that aggregate COMESA and world welfare improves by US$39.5 million and US$3.41 million respectively. On the other hand, when Eritrea and Swaziland join the FTA (FTA-19), most COMESA member countries enjoy substantial welfare gain. With full FTA among all COMESA, the largest welfare benefit goes to Zimbabwe, RSAC, REA, Egypt and Mauritius. In contrast, Ethiopia and to a lesser extent Madagascar and Zambia suffers moderate welfare loss. Table 7 further reports that, with full FTA, the aggregate welfare for COMESA and world improve by US$328 million and US$250 million respectively. Considering the non-comesa regions, the study found that all regions experience a loss in welfare and the loss increases with FTA-19 compared to FTA -17. The welfare effect in GTAP model can be decomposed into allocative efficiency, Terms of Trade and Investment-Savings effect. Allocative efficiency is the measured change in the ability to efficiently allocate resource across sectors in the economy. Mathematically, this is just a collection of variations in the tax revenue of a regional household, which represent the government of a country in the real world (Narayanan and Sharma, 2016). Table 7 shows that most COMESA member states reported positive 15

allocative efficiency in both scenario 1&2. In contrast, Zimbabwe, REA, and RSCA get welfare loss in scenario 1 and RNA in scenario 2. For most COMESA countries the major gain in welfare comes from positive allocative efficiency (e.g., Kenya and Egypt in scenario 1; Zimbabwe, Kenya, and REA in scenario 2). Regions Table 7. Welfare Decomposition for COMESA FTA (US$ million) Allocative Efficiency Terms of Trade FTA-17 Investment Savings Total (EV) 16 Allocative Efficiency Terms of Trade FTA-19 Investment Savings Total (EV) Ethiopia 1.1-6.15-8.13-13.18 4.45-2.6-5.52-3.7 Egypt 2.6 10.67 5.62 18.9 6.67 16.3 8.03 31 Kenya 7.1 15.33 13.16 35.6 11.43 9.2 10.08 30.7 Malawi 0.05 0.06 0.05 0.16 1.29 1.03 0.74 3.06 Madagascar 0 0 0 0 0.19-0.5 0.01-0.3 Mauritius 0 0 0.01 0.02 1.72 19.3 5.18 26.2 Rwanda 0.06-0.15 0.01-0.1 0.46 0.04 0.06 0.56 Uganda 0.23-0.48 0.04-0.2 1.35 1.04 0.36 2.76 Zambia 0.06 0.09-0.07 0.07 0.23-1.3 0.36-0.7 Zimbabwe -0.13-0.06-0.04-0.2 174.41-29 -27.56 118 REA -0.82-0.48-0.08-1.4 36.33 4.92 1.66 42.9 RNA 0.06-0.07-0.13-0.1-0.34 0.39 0.29 0.34 RSCA -0.08-0.02 0.09-0 0.01 0.48-0.02 0.48 RSAC 0 0 0 0 31.55 69.1-24.09 76.5 EU27-1.17-5.74-1.26-8.2-5.69-26 3.36-28 UK -0.2-1.11-0.13-1.5-0.65-1.9-0.35-2.9 ROW -5.39-11.9-9.16-26 -12.15-62 27.31-47 Total World 3.41 250 (Source) Model Simulation The terms of trade, in theory, are defined using the C.i.f price of imports relative to the f.o.b. prices of exports. The domestic price of imports goes down when the FTA is implemented. Hence, elimination of import tariffs automatically leads to a reduction in the import prices. Terms of trade loss result from lower import price than export price arising from tariff elimination. Table 7 further reports that with FTA-17,

Egypt and Kenya experience substantial gain in their terms of trade while other COMESA member countries reported low or negative terms of trade with the highest loss goes to Ethiopia (US$6.15 million). However, with FTA-19, RSAC, Egypt, Mauritius and REA reported a large gain in their terms of trade. The difference between investment and saving in a country adjusts to equate the real trade balance. Therefore the saving investment effect from COMESA FTA in Table 7 above moves in line with the direction of trade balance. 5.2. COMESA Customs Union 1. Macroeconomic effect Table 8 shows the macroeconomic and trade effect of COMESA customs union. The move towards customs union slightly improves GDP of most COMESA member countries. Egypt, Malawi, Madagascar, Mauritius, Zambia, REA, and RSCA reported modest growth in their GDP but lose regarding export and import growth. On the other hand, Ethiopia, Kenya, Rwanda, and Uganda reported growth in their export and import but loses in terms of GDP from the Customs Union scenario. For non-comesa regions, EU- 27, UK, and ROW, there is a trade surplus, but they lose in terms of GDP. As can be seen from Table 9, the decline in GDP for some COMESA countries is mainly due to a flood of cheap imports and a resulting reduction in consumption of domestic commodity and investment, although there is an improvement in trade balance (e.g., Ethiopia, Kenya, Uganda and Rwanda). However, for other countries, the gain in GDP results from the increase in trade balance except for Zimbabwe, REA, and RNA. For Zimbabwe, REA and RNA the main reason for the growth of GDP is an increase in domestic consumption, investment and government consumption, although trade balance is negative. Therefore, there will be a change in the production and consumption structure for most COMESA countries because of the customs union; for Ethiopia, Kenya, Rwanda, and Uganda more consumption is imported, and production is exported compared to the base year scenario. As COMESA countries reduce import tariff against non-member regions through customs union, their import demand increase. However, the existing tariff of some COMESA member countries is below CET rate, and customs union may, in turn, increase protection for most COMESA countries. As a result, the growth of export and import for most COMESA countries is negative. 17

Table 8. Macroeconomic effects of COMESA Custom Union Region Import Value (%) Export Value (%) GDP (%) Change in Trade Balance (US$ million) Ethiopia 0.06 0.22-0.05 4.02 Egypt -0.18-0.22 0.08 16.87 Kenya 0.06 0.35-0.16 21.74 Malawi -0.14-0.07 0.07 1.92 Madagascar -0.33-0.27 0.08 2.4 Mauritius -0.08-0.09 0.05 1.3 Rwanda 0.06 0.08-0.05 0.24 Uganda 0.09 0.12-0.02 0.74 Zambia -0.02-0.01 0.02 0.75 Zimbabwe 0.02-0.36 0.39-12.26 REA -0.05-0.07 0.16-3.06 RNA 0.05-0.1 0.12-108.65 RSCA -0.12-0.06 0.09 6.85 RSAC -0.02 0.07 0 2.48 EU27 0 0 0 33.28 UK 0 0 0 2.38 ROW 0 0 0 29.01 (Source) Model Simulation 18

Table 9 Change in GDP Components for Customs Union Scenario (US$ million) GDP Components Consumption Investment Government Exports (-) Imports Total Ethiopia -9.93-8.58-1.27 9.42-5.4-15.8 Egypt 165.5-6.71 18.96-130.5 147.4 194.6 Kenya -46.15-19.9-10.89 32.24-10.5-55.2 Malawi 3.59-2.1 0.81-1.33 3.26 4.22 Madagascar 6.31-1.95 0.71-8.53 10.92 7.46 Mauritius 4.9-1.06 0.84-4.36 5.65 5.97 Rwanda -2.44-0.71-0.26 1.14-0.9-3.17 Uganda -3.16-0.78-0.25 5.2-4.46-3.44 Zambia 1.77 0.54 0.59-1.38 2.13 3.65 Zimbabwe 40.34 6.94 7.36-10.73-1.53 42.39 REA 96.61 16.87 13.1-14.96 11.9 123.5 RNA 151.14 166.65 64.79-76.75-31.92 273.9 RSCA 71.36 23.89 19.79-42.69 49.54 121.9 RSAC -0.37-1.97-0.03 2.12 0.36 0.11 EU27-96 -60.75-33.75 4 30-157 UK -11.13-3.75-3.69-1.88 4.25-16.2 ROW -144-88 -38 1 27-242 Total 228.35 18.64 38.81-237.9 237.6 285.5 Source: Model simulation 2. Welfare effect Table 10 shows that COMESA customs union is welfare improving for most COMESA countries; RNA, Zimbabwe, and Ethiopia reported substantial welfare gain while Egypt, Kenya, REA and RSCA experience considerable welfare loss. For Ethiopia, Rwanda, and Uganda the large gain in welfare results from an improvement in the efficiency of resources allocated to the sector. In contrast, for Madagascar, Mauritius, Zambia, Zimbabwe and RNA, the efficiency gain results from positive terms of trade effect. In general, the welfare impact of COMESA customs union differs from country to country depending on the change in the efficiency of resource allocation and relative change in import and export price. COMESA customs union reduce the aggregate welfare of COMESA and world by US$50.59 million and US$80.66 million respectively. Besides, COMESA customs union result in a welfare loss for non- COMESA regions. 19

Table 10. Welfare Decomposition for COMESA Customs Union (US$ million) Region Allocative Efficiency Terms of Trade Investment- Savings Total (EV) Ethiopia 13.09-1.69-6.81 4.59 Egypt -75.63 35.77 12.6-27.26 Kenya 8.64-11.18-9.92-12.46 Malawi -1.27 0.49 0.1-0.68 Madagascar -1.53 1.54 0.07 0.08 Mauritius -0.58 1.64 0.7 1.76 Rwanda 1.67-0.05-0.03 1.59 Uganda 6.85-1.15 0.15 5.86 Zambia -0.19 0.96-0.43 0.34 Zimbabwe -2.06 3.25 12.5 13.69 REA -50.57 6.48 1.35-42.75 RNA -3.78 17.45 6.48 20.15 RSCA -11.47 5.31-9.21-15.37 RSAC 0.34-0.1-0.38-0.13 EU27 30.29-44.56-0.92-15.19 UK 0.7-1.2-1 -1.5 ROW 4.91-12.99-5.28-13.36 World Total -80.66 (Source) Model Simulation 20

5.3. COMESA - EU FTA: European Partnership Agreement 1. Macroeconomic effect As shown in Table 2, COMESA member countries have large import share from EU, particularly in industrial and service sector. Therefore, a complete removal of import tariff from COMESA to EU will have a potential impact on the trade between these regions. This effect comes from two sources. First, EU imports, which were previously taxed, will enter COMESA markets duty-free once an EPA is in place. Therefore, there will be a substitution of domestic goods for cheap imports from EU. Second, trade liberalization under an EPA will make some EU products more affordable (since exempted from customs duties) than products previously imported from other sources still subject to customs duties; this trade diverted from non- EPA origin to the benefit of EU imports. Table 11 shows that all EPA signatory COMESA member countries lose in terms of GDP except RNA; the loss is smaller when UK exit from EU (Brexit). The largest loss in terms of GDP goes to REA, Zimbabwe, Kenya and Ethiopia. In contrast, EU-27 and UK reported a slight increase in GDP growth, but Brexit result in lower GDP growth for the UK compared to the case when UK is a member of EU. Table 11 further reports that EPA results in a growth of import for most COMESA countries except Mauritius and Egypt; with Brexit, the impact is smaller. Similarly, there is an increase in export of most COMESA countries with EPA scenario except Egypt, Mauritius, RNA, and RSAC; the impact is less with Brexit. The largest growth in export for EPA scenario goes to Zimbabwe, REA, Kenya and Ethiopia whereas REA, Madagascar, RSCA, and Ethiopia have large import growth. The impact of EPA differs from country to country depending on the trade relation and initial protection imposed by the country. Furthermore, with EPA, EU-27, and UK experience slight growth in their export and import, whereas when UK exit from EU, UK's import decrease by 0.01%. Table 11 indicates that Ethiopia, Egypt, Kenya, and Zimbabwe reported large trade surplus while RSCA, RNA, Zambia, and Madagascar reported significant trade deficit with EPA. Brexit has similar trade balance effect, but the trade balance deficit and surplus are small with Brexit. On the other hand, EPA agreement worsens the trade balance of EU-27 and UK by US$880.17 million and US$161.13 million respectively. However, UK reported a trade balance surplus of US$57.22 million following Brexit while the deficit to EU-27 increases further with Brexit. As can be seen from Table 12, the main reason for the decline of GDP for Ethiopia, Egypt, Kenya, Malawi, Mauritius, Zimbabwe and, REA are the drop in consumption of domestic commodities, investment, and Government consumption, although trade balance is positive. For Rwanda, Uganda, RSCA, Madagascar, Zambia and, RSAC deterioration of trade balance contributes to the decline in GDP besides to a fall in consumption, investment and government consumption. The main reason for the decrease in consumption of domestically produced commodity for COMESA countries is the increase in both export and import due to the complete removal of tariff by COMESA countries for EU origin product. Furthermore, the increase in import and export also change the production and consumption structure; the 21