Poverty and Shared Prosperity Implications of Deep Integration in Eastern and Southern Africa

Evidence indicates that trade costs are a much more substantial barrier to trade than tariffs are, especially in Sub-Saharan Africa. This paper decomposes trade costs into: (i) trade facilitation, (ii) non-tariff barriers, and (iii) the costs of business services. The paper assesses the poverty and shared prosperity impacts of deep integration to reduce these three types of trade costs in: (i) the East African Customs Union?Common Market of East and Southern Africa?South African Development Community "Tripartite" Free Trade Area; (ii) within the East African Customs Union; and (iii) unilaterally by the East African Customs Union. The analysis employs an innovative, multi-region computable general equilibrium model to estimate the changes in the macroeconomic variables that impact poverty and shared prosperity. The model estimates are used in the Global Income Distribution Dynamics microsimulation model to obtain assessments of the changes in the poverty headcount and shared prosperity for each of the simulations for the six African regions or countries. The paper finds that these reforms are pro-poor. There are significant reductions in the poverty headcount and the percentage of the population living in poverty for all six of the African regions from deep integration in the Tripartite Free Trade Area or comparable unilateral reforms by the East African Customs Union. Further, the incomes of the bottom 40 percent of the populations noticeably increase in all countries or regions that are engaged in the trade reforms. The reason for the poor share in prosperity is the fact that the reforms increase unskilled wages faster than the rewards of other factors of production, as the reforms tend to favor agriculture. Despite the uniform increases in income for the poorest 40 percent, there are some cases where the share of income captured by the poorest 40 percent of the population decreases. The estimated gains vary considerably across countries and reforms. Thus, countries would have an interest in negotiating for different reforms in different agreements.

4 level, we extend the model of BTY to also derive estimated impacts for 2030. We use estimates from the CGE model as inputs in the Global Income Distribution Dynamics (GIDD) microsimulation model to obtain assessments of the changes in the poverty headcount and shared prosperity for each of our simulations. The GIDD is the first global macro-micro simulation tool, which combines a consistent set of price and volume changes from a global CGE model with household surveys at the global level (see Bussolo et al. 2010).
Conceptual innovation is that this paper is the first global trade model to numerically assess the poverty and shared prosperity effects of regional liberalization. It is also the first to examine the poverty and shared prosperity impacts of time in trade costs differentiated by product as well as the impact of liberalization of barriers against foreign direct investors in services.
The essential data problem to assess services commitments has been the lack of estimates of the ad valorem equivalents of the barriers to foreign suppliers of services based on assessments of the regulatory regimes in place. We employ a new database of the ad valorem equivalents of barriers in eleven business services sectors in 103 countries (see Jafari and Tarr, forthcoming), which was aggregated to the sectors and regions of this model (Jafari, 2014c). The estimates of the ad valorem equivalents (AVEs) were possible due to the newly released World Bank survey information on the discriminatory regulatory barriers against foreign suppliers of services on these eleven sectors in 103 countries (see Borchert et al., 2014). Many results in the paper depend crucially on this database of AVEs.
In addition, this paper builds on or adapts the following three databases: (i) trade facilitation-the paper employs the database on the time in trade costs of Hummels et al. (2007) and Minor (2013). We aggregate the database to the sectors and regions of our model. Although a central finding of the studies by Hummels, Minor and their co-authors is that the AVE of time in trade varies across products, most computable general equilibrium modeling of trade facilitation issues have used a single AVE across all products. We show that this more accurate database impacts the results; (ii) foreign affiliate sales-we use the "Global Database of Foreign Affiliate Sales" developed by Fukui and Lakatos (2012). In the Tripartite region, we augmented the database with independent work; and (iii) estimates of the ad valorem equivalents of non-tariff measures developed by Kee, Nicita and Olarreaga (2009).
At the aggregate level, we find that there are substantial gains for all six of our African regions from deep integration in the Tripartite FTA or comparable unilateral reforms to all 5 countries by the EACU to reduce trade costs; but our decomposition analysis reveals that the estimated gains and the magnitudes vary considerably across countries and depend on the reform. Thus, the regions and countries have very different stakes in the various reforms and would have an interest in negotiating for different reforms in different agreements. One striking finding is that in our Near Term model, we estimate that Kenya gains less from comparable unilateral liberalization by the EACU than from the Tripartite FTA, due in part to an umbrella of protection in services markets in the Tripartite region. For goods markets, Wonnacott and Wonnacott (1981) and Harrison, Rutherford and Tarr (2002) have shown that due to market access, there is the possibility of larger gains in preferential agreements than from unilateral liberalization. This extends their result to services markets. Karingi and Fekadu (2009), Jensen and Sandry (2011) and Willenbocket (2013) have executed general equilibrium assessments of the impacts of the Tripartite FTA. They focus either exclusively or primarily on preferential tariff reductions. 8 They find small welfare changes from preferential tariff reduction in the Tripartite FTA, with many countries losing and net gains of only about 0.1 to 0.2 percent of GDP. Our estimates of the impact of tariff changes are consistent with these earlier studies; but, depending on the country or region, our estimates of the gains from reductions of trade costs within the Tripartite area are about 10 to 30 times larger than the estimated gains of preferential tariff reduction--suggesting very different stakes.
The paper is organized as follows. In section 2 we provide an overview of the model. In section 3 we explain the data that we have developed or used in constructing this model. The CGE model results are presented in section 4 and the microsimulation results for poverty and shared prosperity are presented in section 5. Sensitivity analysis is presented in section 6. In section 7, we 8 Jensen and Sandry add a 2 percent uniform reduction in non-tariff barriers on goods and cross-border services to preferential tariff reduction. Willenbockel also executes a scenario with a 5 percent reduction in border crossing costs for all goods based on unpublished TradeMark South Africa estimates of border crossing costs; then the estimated gains increase to 0.4 percent of GDP for the Tripartite region in aggregate.
Although they do not focus on Eastern or Southern Africa, two other interesting general equilibrium assessments of trade policy changes in Africa are the following. Anderson, Martin and van der Mensbrugghe (2006) find that global free merchandise trade would boost real incomes in Sub-Saharan Africa more than proportionately than in other developing countries; but partial liberalization proposals would capture only a small share of the gains. Mevel and Karangi (2012) analyze the removal of all tariffs on goods within the African continent as a whole. They find this would increase intra-African trade by 52 percent, but if trade facilitation measures are also implemented that reduce the time costs of trade by 50 percent, intra-African trade would more than double. conclude with a summary of the key results and the stakes of the regions of our model based on the reform.

Introduction
In this paper, we obtain results for poverty and shared prosperity in several African countries of deep integration in East and Southern Africa. We do that by first assessing the impacts on the variables that impact poverty and shared prosperity in a computable general equilibrium model. The key variables on which we obtain estimates are the change in the value of real consumption, the change in real wages of skilled and unskilled labor in agriculture and nonagricultural sectors and the change in prices of food and non-food items. We then use those estimates as inputs in the Global Income Distribution Dynamics (GIDD) microsimulation model to obtain assessments of the changes in the poverty headcount and on shared prosperity.
For the near term results, the CGE model employed in this paper is the multi-region trade model developed and explained in Balistreri, Tarr and Yonezawa (2015) and in Balistreri, Tarr and Yonezawa (2014). 9 Here we provide a brief overview. For a detailed description, the interested reader is referred to the Balistreri, Tarr (3) seven services sectors in which there is imperfect competition and foreign direct investment. The cost, production and pricing structures in the three categories differ widely, but regardless of sector, all firms minimize the cost of production. 7 Primary factors are skilled labor, unskilled labor, capital (including land) 10 and natural resources. Regarding capital, there is mobile capital and sector-specific capital in imperfectly competitive goods sectors and services sectors with FDI; and primary inputs imported by multinational service providers, reflecting specialized management expertise or technology of the firm. There is some sector specific capital for each imperfectly competitive firm (and for firms in services sectors with FDI) for each region of the model. In the sectors where there is sector specific capital, there are decreasing returns to scale in the use of the mobile factors and supply curves in these sectors slope upward. We calibrate the elasticity of substitution between sector specific capital and other inputs in each sector so that the elasticity of supply of the firms is consistent with econometric evidence that indicates that the supply response depends on the level of development and the technological complexity of the product. 11 One extension of BTY is that we allow sector specific labor. In BTY, all labor was mobile. Here, in our benchmark equilibrium, we assume that 50 percent of labor is sector specific (both skilled labor and unskilled labor). Value-added is an aggregate of our primary factors with elasticity of substitution σ. Skilled (and unskilled) labor is an aggregate of sector specific and mobile labor with elasticity of substitution 2σ. Thus, the share of sector specific labor may change in a counterfactual scenario, including a comparative steady state scenario.

Comparative Steady State Formulation for the 2030 Solution of the Model
The second important extension is, rather than a primary focus on the comparative static model, we provide equal emphasis on the results from a comparative steady state model; and we make some modeling extensions in the comparative steady state model compared with our earlier applications. The comparative steady state model allows us to provide estimates of impacts of the trade policies in 2030.

Basic Theory of the Endogenous Capital Stock in the Comparative Steady State
Model. In the comparative static model, we assume that the capital stock is fixed and the rental rate on capital is endogenously determined. In the comparative steady state model, the logic is reversed: the real return on capital is fixed, but we allow the capital stock to adjust to its steady 8 state equilibrium along with all of the model features we employ in our comparative static model.
The comparative steady state model is based on the assumption that investors demand a given rate of return on capital in order to invest in a given country. We assume that the rate of return demanded by investors for each country or region is initially in long run equilibrium. If a trade policy or other type of shock happens to induce and increase in the rate of return on capital so that it exceeds the initial rate of return, investors will invest and expand the capital stock. Expansion of the capital stock drives down the marginal product of capital, i.e., it drives down the rental rate on capital. A new equilibrium in the comparative steady state model is determined when the capital stock rises sufficiently that the real rate of return on capital falls back to the initial level. 12 To analyze trade policy, this comparative steady state approach has been employed by many authors, including Tarr (1996, 1997b) and Baldwin et al. (1999) and Francois et al. (1996). The approach, however, dates back to the 1970s, when both Hansen and Koopmans (1972) and Dantzig and Manne (1974) developed it.

Endogenous Investment in the Comparative Steady State Model.
We have made an important modification in the modeling approach of the above studies to adjust for an upward bias in the estimated welfare gains. The approach employed in the above studies ignores the foregone consumption necessary to achieve the higher level of investment and, thus, is an upper bound estimate on the long run gains within the framework of the model assumptions. Based on the relationship between the capital stock and the cumulative depreciated value of investment, 13 we follow Francois et al. (2013) and assume that investment increases in proportion to the increase in the capital stock. Since consumers obtain utility only from consumption, if the capital stock 12 The rate of return on investment in our model is the rental rate on capital divided by the cost of a unit of the capital good. We allow both mobile and sector specific capital to be endogenously determined in the comparative steady state model. 13 In an intertemporal model, capital is usually measured as the cumulative undepreciated value of investments over time. That is where K(t) and I(t) are the capital stock and investment in period t, respectively, and d is one minus the depreciation rate. If we assume that investment in each period is constant equal to I, then the capital stock in period t is: It follows that for a given time period and fixed depreciation rate, the percentage change in the constant value of investment in each period is equal to the percentage change in the capital stock, i.e.,: (Hicksian equivalent variation) by the percentage increase in the population. 14 In many of our African countries the percentage increase in the labor force exceeds the percentage increase in the population. A greater share of the population in the labor force should increase welfare, so we reduce the estimated welfare by the percentage increase in the population, not the percentage increase in the labor force.

Summary of the Global Income Distribution Dynamics (GIDD) Model
We use the Global Income Distribution Dynamics (GIDD) model, developed by Bussolo, de Hoyos, and Medvedev (2010), to estimate distributional effects. GIDD is a "top-down" micro simulation framework that exploits heterogeneity observed in household surveys to distribute macroeconomic shocks. These shocks are aligned with a macroeconomic model such as the CGE model used in this paper. More specifically, we impose consistency between the GIDD and the CGE models in this paper in various ways. First, both use the same United Nations projections in aggregate population and age and education structures. The GIDD then uses estimates from the CGE model as inputs into the household model. In particular, as inputs into the calculation of changes in per capita household income, the GIDD uses CGE model estimates of differentiated wages for skilled, unskilled, agricultural and non-agricultural labor and changes in the prices of agricultural and non-agricultural goods. 15 . Finally, all household incomes are adjusted proportionally so that the percentage change in the aggregate of household incomes in the GIDD is consistent with the CGE model's estimate of the percentage change in real income.
GIDD was developed by the World Bank's Development Prospects Group and was inspired by previous efforts involving simulation exercises (Bourguignon et al., 2002;Bourguignon et al., 2008;and Davies 2009). Previous examples of application using CGE outputs and GIDD include the effect of agriculture distortions in the global economy (Dessus, et al., 2008, Bussolo et al., 2010, the effect of demographic change on Africa (Ahmed et al., 2014) and external and internal shocks in Africa (Devarajan et al., 2015). 11 The first step in the microsimulation exercise is to implement a set of changes in the household surveys' demographic structure, as explained in section 3.5. The second step is to adjust factor returns by skill and sector in accordance with the results of the CGE model. The GIDD imposes an entirely new vector of earnings on each worker, conditional on that worker being in sector s and having and educational attainment e. The third step adjusts the average income/consumption per capita to guarantee that it changes exactly in line with the CGE results.
Lastly, GIDD constructs a household-specific deflator to adjust for changes in relative prices.
The price deflator is constructed using initial and final prices indexes of food versus non-food from the macro model and household-specific budget consumption shares for food and non-food observed in micro data.
Beginning with a distribution of earnings from labor by sector and skill [ , ] in the macro data, define a set of wage gaps as follows: , , and a similar set of wage gaps for the macroeconomic counterfactual scenario: The micro data will have also have a set of wage premia, which, in general, will differ from the CGE data. Analogous to equations 5 and 6, define: where , are the wage premia based on averages by skill group and sector in the household data; ′ , are the average earnings of labor in sector s and skill group e based on the household data; ′ , are the average earnings of unskilled labor in agriculture based on the household data; and the "hat" with apostrophe symbols such as ′ are the predicted values at the household level as 12 a result of the policy change. All right hand side values of equation 7 are known from the initial household data. It is necessary to calculate the counterfactual wage gaps , . In the GIDD, these will be calculated as: We may calculate the left hand side of equation 9, since the three values on the right hand side are known from equations 5, 6 and 7. Equation 9 implies that even if initial wages differ between the CGE and micro models, the percentage change in the wage gaps will be consistent across the two models. By passing on percent changes in wage premia by type of worker, instead of percent changes in wages, the possibility of wage gaps moving in opposite directions in the macro and in the household data is eliminated. Within each group of workers, distributional changes occur; but, on average, for any group of workers, the relative wages for each type of worker is constrained to be consistent with the corresponding growth rates from the CGE model.
Given the known values in equations 5-9, and defining average wages for unskilled labor in agriculture as numeraire, so that ′ , ′ , , it is possible to calculate the percentage changes in average wage income of households in sector s and skill level e that are consistent with wage gaps expressed in Equation 9: The changes in real per capita incomes brought about by a change in relative prices of food versus non-food can be approximated by the following linear expression:

18
where in Equation 17 is the real per capita income adjusted for changes in relative prices of food versus non-food. is the counterfactual measure of real per capita income of household h for the analysis of poverty and shared prosperity.

Estimates of the Ad Valorem Equivalents of the Trade Costs, Population Projections, and Foreign Direct Investment Shares
Given the primary importance of the ad valorem equivalents of the barriers against foreign suppliers of services, the time in trade costs and the non-tariff barriers, we discuss those estimates here and present the estimates in tables 2a-2f. Since it involves a new data set, we also discuss the estimates of the shares of domestic services markets captured by foreign direct investors. Full documentation of the data set is available in BTY. Finally, given their importance for our 2030 CGE model and the microsimulation work, we also discuss the population and skill mix projections.

Ad Valorem Equivalents (AVEs) of the Barriers Against Foreign Suppliers of Business Services
Our estimates in the services sectors necessitated the development of a new global database of the ad valorem equivalents of discriminatory barriers against foreign providers of services. This was possible only because of the availability of a new World Bank database of survey information on the discriminatory regulatory barriers in 11 services sectors in 103 countries described in Borchert, Gootiiz and Mattoo (2014). 16 Borchert et al. produced "Services Trade Restrictiveness Indices," but did not transform their indices of the regulatory regimes into ad valorem equivalents.
Our methodology uses the World Bank database for an assessment of the regulatory regimes, but builds on a series of studies, supported by the Australian Productivity Commission, to convert assessments of services regulatory regimes into ad valorem equivalents for all 11 sectors in 103 countries. This work is documented in Jafari and Tarr (2015). The aggregation to the sectors and regions of our model is documented in Jafari (2014c). In the cases of Kenya and Tanzania, additional information was available and the estimates are taken from Jafari (2014a) and Jafari (2014b).

Estimates of the Ad Valorem Equivalents of the Costs of Time in Exporting and Importing
In order to estimate the impact of improved trade facilitation, in this paper we apply a new data set of the time cost of trade based on the path-breaking work of Hummels and Schaur (2013) and Hummels et al. (2007). Using the estimates of Hummels and his co-authors, Peter Minor (2013) provided estimates for the regions and products in the GTAP database on a bilateral basis.
We use estimates from Peter Minor, which we aggregate to the sectors and regions of our model, yielding the cost of trade by product and country on a bilateral trade basis. Detailed documentation of the steps we have taken, and a brief explanation of the methodology may be found in appendix C of BTY.

Estimates of the Ad Valorem Equivalents (AVEs) for Non-Tariff Measures (NTMs) for the Regions of Our Model
Our estimates of the AVEs of NTMs are based on the estimates of Kee et al. (2009).
Building on Kee et al. (2008), Kee et al. (2009)  Although the benchmark equilibrium incorporates tariff free trade between partners in free trade agreements or customs unions, the report of the East African Community (2012) shows that non-tariff barriers remain a very significant problem. Consequently, we assume the ad valorem equivalents of the non-tariff barriers apply to all countries.

Share of Market Captured by Foreign Direct Investors in Services and by Cross-Border Sales of Services
For each country or region in our model, it was necessary to calculate the market share of foreign direct investors by source region in the business services sectors of our model as well as the share of cross-border services in each of our regions for these seven sectors. For cross-border sales of services, we use the trade data from the GTAP 8.1 data set.
Our primary data source for foreign affiliate sales is the database developed by Fukui and Lakatos (2012). Fukui and Lakatos combine Eurostat data for 41 countries with an econometric model to estimate the missing values and thus produce estimates for all regions and sectors in the GTAP data set. For the share of sales in the sector by the host country, we use the GTAP data set for total sales in the sector and subtract the total of foreign affiliate sales from total sales to obtain the host country share of sales. In the case of insurance services in African regions, we used the Axco database (for a complete list of companies) and publicly available information on ownership shares of companies as our primary data sources. In the case of telecommunications services in our six African regions, we used national communications commission data and other publicly available sources on ownership of companies, taking South Africa as our proxy for SADC. Details are in appendices D and E of BTY. In Kenya and Tanzania, professional associations of lawyers and engineers in these countries provided data on the number of professionals, both total and nonlocal. Details and documentation of the calculations for Kenya and Tanzania are available in Jafari (2014d).

Population and Skill Mix of the Labor Force
The population growth adjustment is particularly important in countries with high fertility rates, such as those in Sub-Saharan Africa. In practical terms, the adjustment for population growth allows the analysis to explicitly take into account changes in the size of the working-age population. We perform population and education projections during the first stage of the microsimulation model and in creating the Business as Usual scenario for the comparative steady state CGE model. For each country, we construct the demographic profile in two steps. First, the age and gender composition is exogenously determined following medium variant estimates from the World Population Prospects (United Nations Department of Economic and Social Affairs, 2015).
In a second step, following Bourguignon and Bussolo (2012), country-specific educational profiles are constructed using initial educational achievement levels observed in the household surveys with some conservative yet simple assumptions about educational progress.
As mentioned earlier, we employ age and gender totals based on data from the United Nations' (2015) medium variant population projections. In terms of education, we assume that as the population ages, the average educational attainment in a country increases through a pure pipeline effect, as younger and more educated cohorts replace older cohorts. For example, if at time t half of the population in the cohort formed by individuals between 25 and 30 years of age have postsecondary education, then after ten years (at t+10), half of the population between 35 and 40 will have post-secondary education. Furthermore, for younger cohorts we imposed the assumption that there is no improvement in enrollment and graduation rates from those observed at time t. In other words, the average educational attainment of these young cohorts in the future is equal to the average educational levels of the 20 to 24 cohort of time t. This is a conservative assumption given that the 20 to 24 cohort observed at time t may not have the maximum educational level attainable. 19

CGE Results: Deep Integration in Eastern and Southern Africa
We present both "Near-Term" results in table 3 table 3. 19 In practical terms, the micro-simulation model recalibrates each household sample weight to match the age, gender, and education projected totals. A new probability distribution can be obtained by solving an optimization problem based on a minimum cross-entropy criterion as in Olivieri et al. (2014 As a point of reference to compare with the Tripartite FTA, we first consider what is at stake for the EACU members from narrower deep integration within the EACU alone and wider unilateral liberalization. As members of a customs union, we assume that the EACU members act collectively on all actions in our scenarios. In our benchmark equilibrium we assume that tariff free trade prevails within each of the three regional groups of the Tripartite area, but the barriers that lead to high trade costs apply to all countries and regions. In tables 2a-f, we show the benchmark ad valorem rates of distortion for all barriers we apply in the six African regions of our model. We elaborate in section 6.3 on an approach to reduce these barriers. Consequently, we take a more modest 20 percent reduction in the ad valorem equivalent of these barriers, with no spillover to countries excluded from the agreement.

Deep Preferential Integration within the East African Community ("EAC Central
Barriers on foreign providers of services: We take a 50 percent cut in these barriers.
On July 1, 2010, the East African Community adopted a Common Market protocol that called for 21 the free movement of services within the five member states, along with the free movement of goods, capital and labor. 22

Preferential Reduction of Barriers against EACU Service Providers. In the Near
Term (2030)  EAC, there are production and consumption efficiency gains, which explain the difference between the total welfare gains and the recaptured rents.

EACU Unilateral Liberalization ("EAC Liberal")
The above estimates indicate that there are gains from deep integration within the EAC.
With a combined nominal GDP in 2013 of only about US$121 billion (or US$297 on a purchasing power parity basis), 23 however, the EACU is not a large market, and economic theory indicates that there should be substantially greater gains from integrating into the world trading environment. As a point of reference, in the scenario labeled "EAC Liberal," we assess the extent of these larger gains from unilateral extension by the EACU of the reforms to lower trade costs.

Scenario Definition.
In EACU Liberal, we extend the liberalizations of non-tariff barrier and services barriers implemented in "EACU Central" to all trading partners in the world.
In the case of the time in trade costs, we assume the EACU countries implement equivalent reforms to those in the EACU Deep Integration scenario to reduce the time in trade; but we do not extend these outside of the EACU on the grounds that the improvements that can be made are primarily regional and reciprocal and we already convey a 5 percent cut in these barriers for countries outside of the EAC. In tables 3 and 4, the results and policy changes are listed.
In the Near Term scenario, we see that for Kenya and Uganda, the gains are about twice as large as in EACU Central; for Rwanda the gains increase substantially from 1.4 percent to 4.95 percent of consumption. The biggest increase in welfare is for Tanzania; the welfare gain dramatically increases from 0.95 percent of consumption to 7.11 percent of consumption. We decompose the EACU Liberal scenario to explain the differences across the EACU countries.
In the case of Tanzania, the big increase in welfare is clearly due to the broader liberalization of non-tariff barriers. The wider liberalization of non-tariff barriers results in a welfare gain of more than five percent of consumption, whereas the welfare gains were only 0.17 percent of consumption in the EACU Central case. This large increase is explained by two factors: (i) as shown in respectively. The estimated gains for the four members of the EACU in our model are all substantial, ranging from 6.5 to 11.1 percent of consumption, annually.

Trade Facilitation Only Results
The "only trade facilitation" scenario shows that a significant part of the explanation for creates substantial gains for insurance services suppliers from Kenya in COMESA markets. We verified this explanation by executing a scenario in which we preferentially liberalize services barriers within the Tripartite area, but exclude preferential reduction in insurance services barriers.
In this scenario, the estimated Near Term gains to Kenya from "only services liberalization" within Tripartite fall dramatically from 1.39 to 0.2 percent of consumption. The theory paper of Wonnacott and Wonnacott (1981) emphasized that improved market access in export markets of goods could lead to preferential trade agreements dominating unilateral trade liberalization; and this was demonstrated by Harrison, Rutherford and Tarr (2002) for goods. Our result for Kenya extends those earlier results in goods to market access in services, and shows that preferential agreements can yield larger gains than unilateral liberalization due to market access gains.
The other region that reaps substantial gains from services reform in the Tripartite area is COMESA. This is explained by the high services barriers in COMESA, especially in insurance, yielding large rent capture from liberalization, and efficiency gains from better access to relatively efficient services suppliers.

Adjustment Costs and the Political Economy of Regional Trade Liberalization
25 Due to the product mix differences across countries of the aggregated sectors of our model. 26 The partner country AVE is also relevant in assessing impacts.

29
We find that, in general, preferential trade liberalization (tariff or NTB reduction) results in substantially muted output changes at the sector level compared with unilateral liberalization. Take Tanzania as an example (which has the highest AVEs of its NTBs in our data set). In the Near Term model, the gains from unilateral reduction of NTB barriers alone are about 25 times greater than the gains from liberalization within EACU alone. But, the maximum output decline at the sector level from NTB liberalization within EACU alone is two percent, but with unilateral liberalization, we estimate output declines of 9.5 percent for textiles and apparel, 11.8 percent for other manufacturing and 13.8 percent for wood and paper products. Thus, although the welfare gains of preferential liberalization are dramatically smaller than unilateral liberalization, the adjustment costs are also smaller.
To illustrate, we quantify the adjustment costs estimate for Tanzania, by adopting the . * * μ / (4).
We calculate equation 4 for Tanzania for our three principal scenarios. Regarding β, in our Near Term model simulations, we estimate the number of workers that must change jobs by sector and skill type. Taking a weighted average across all sectors and skill types of labor for Tanzania, we calculate that: β = 0.0049 for the EACU Central scenario; β = .0117 for the EACU Liberal scenario; and β = .0146 for the Tripartite scenario. That is, in the EACU Central scenario, we 27 For an explanation of the methodology, see Morkre and Tarr (1980, chapter 3) or Matusz and Tarr (2000). 28 This is the sum of compensation for subsistence labor (that receives 29.8 percent of value added) and all other categories of labor (which receive 30.6 percent). See Jensen, Rutherford and Tarr (2010).
30 estimate that about one-half of one percent of labor must change jobs, and between 1 percent and  These results are consistent with the evidence from empirical studies, summarized by Matusz and Tarr (2000), which has shown that the adjustment costs of trade liberalization are dramatically smaller than the welfare gains. 33 However, policy makers often receive strong lobbying from those who suffer adjustment costs from trade liberalization, while those who gain are more diverse or may not realize they will gain from trade liberalization; so the gainers typically do not lobby for liberalization or lobby much less vigorously. Thus, these results explain some of the appeal of regional liberalization to policy-makers, despite the usually larger net gains of broader unilateral or multilateral liberalization. 29 The Tripartite scenario contains preferential tariff reduction, whereas the EACU Liberal scenario does not any tariff changes; this explains the larger adjustment costs for the Tripartite scenario.  Matusz and Tarr (2000) summarize the evidence on the adjustment costs of trade liberalization.

Poverty Headcount Results
Results for the poverty headcount, percentage of the population below the poverty line and shared prosperity are displayed in Table 6. We find that deep integration in the Africa region reduces poverty beyond what would be achieved in the baseline-business as usual scenario for 2030. Deep integration within the Tripartire region has a larger impact on poverty reduction on the EACU member countries than deep integration within the EACU alone. In the Tripartite scenario, the results suggest a net reduction of 4.2 and 3.23 million of people living with less than PPP$1.25/day in EACU and COMESA-SADC, respectively. Unilateral liberalization by the EACU members would lift even more out of poverty--an estimated 5.31 million in the region, but would have virtually zero effect in COMESA-SADC.

Sensitivity Analysis
In this section, we assess the impact of parameter values and the key modeling assumption of rent capture on the results. We focus on the Near Term model. Through our "piecemeal sensitivity analysis," we will determine the most important parameters for the results, and we will assess how important the rent capture assumption is for the results. We examine the three aggregate policy scenarios: EACU Deep Integration, Tripartite FTA with Deep Integration and EACU Unilateral Liberalization. Our results are presented in table 7.

Impact of Rent Capture Assumption
In our central scenarios we assume that: it takes capital and labor to overcome the barriers; the rents from the barriers are "dissipated;" and the rents are recaptured by the domestic economy in the central scenarios. It is possible, however, that some of the barriers are not dissipated, but instead generate rents that are captured by domestic agents in our initial equilibrium. If so, then the rents that are captured initially by domestic agents would not be available as a net welfare gain since they are a loss to domestic agents; and the welfare analysis for rents is analogous to tariff loss. The "triangle" of efficiency gains will remain, but the welfare gains should be smaller when there are initial rents captured by domestic agents.
In This reflects large available rents in the case of services and non-tariff barriers in trade with regions outside of Africa.
On the other hand, the estimated gains fall considerably in the cases of Tanzania and COMESA. Table 7 shows that in the case of Tanzania, the welfare gains fall to about 60 to 70 percent of their original level in the case of EACU Deep Integration or Tripartite FTA scenarios.
In the case of EACU unilateral liberalization, however, the welfare fall is dramatic--from more than 7 percent of consumption to about 1 percent of consumption, deriving mostly from the lack of capture of the large rents that are impacted by unilateral liberalization of NTBs in Tanzania. For COMESA in the Tripartite FTA, the gains fall from 1.4 to 0.5 percent of consumption due to no rent capture in services. Results for Rwanda are between Kenya and Tanzania in terms of percentage reduction of the gains due to initial rent capture, but the pattern of a much stronger drop in the percentage of gains in the unilateral liberalization scenario prevails due to no capture of rents in services liberalization.

Piecemeal Sensitivity Analysis (other than rent capture)
We see that the central results are rather robust with respect to most of the parameters. In

Overall Gains by Region
In this paper, we have estimated the impact of reducing non-tariff barriers, improving trade facilitation and reducing services costs in the EAC, the Tripartite FTA and unilaterally by the EAC.
We estimate that deep integration within the EACU alone will produce significant benefits for its members, but if expanded to include COMESA and SADC in a Tripartite deeply integrated FTA, it would substantially increase the benefits for our EACU countries. We find that SADC and COMESA would gain substantially from the Tripartite FTA, but their gains are about one-half the gains of Kenya, Tanzania and Uganda from deep integration within the Tripartite FTA. Tanzania and Rwanda gain dramatically more from unilateral liberalization rather than preferential liberalization, but Kenya gains less from EACU Liberal (Unilateral) than from the Tripartite FTA.

Sources of Gains by Region
Our decomposition analysis reveals the reasons for these differences: (i) the augmented  Schiff and Winters (2003) have found that the largest gains from regional integration come from the deep aspects of the agreements, so our results are consistent with the broader empirical and theoretical literature.
The near term results for the Kenya case are especially interesting since it gains more from the Tripartite FTA than from unilateral liberalization by the EACU in all three dimensions of our trade costs. We explained these results due to better market access under a protected regional umbrella and the logistics of trade facilitation reform. This extends the Wonnacott and Wonnacott (1981) result to services. Our results provide a rationale for the strong support of the Tripartite FTA by the Kenyan authorities.

Poverty and Shared Prosperity Results
Our poverty results suggest that the effects of the trade reforms are pro-poor. For example, our estimates are that, as a result of deep integration within the Tripartite agreement, the number of poor in the whole region would be reduced by 7.43 million; and the incomes of the poorest 40 percent of the population increase from 1.5 percent in the case of SADC to 7.7 percent in the case of Kenya. Focusing on poverty reduction in Kenya, Tanzania, Uganda, and Rwanda, compared 36 with Tripartite integration, these four countries would see poverty reduced by an additional 1.1 million from unilateral liberalization. This indicates that while deep regional integration is an important step in helping to reduce poverty, the region should not ignore the additional gains available from broader trade liberalization.
We find that countries are affected in different ways from the different aspects of the reforms, which could lead to countries lobbying for different reforms in terms of choosing the strategy that maximizes poverty reduction. This is typical in trade negotiations, as a country will often have to make a "concessions" in areas it is not highly motivated to reform, in order to obtain and if they were to do so, they would be expected to achieve significant gains for the poor, comparable to the EACU countries.
Despite increases in the incomes of the poorest 40 percent of the populations in all of our trade policy scenarios (where the country or region was an active participant in the reform), there are two concerns. First, in some cases, inequality increases, i.e., the incomes for the richest 40 percent increase faster than the increases in incomes of the poorest 40 percent. Second, we have estimated that adjustment costs are only a small percentage of the gains from trade liberalization; nonetheless, the poor are often very badly equipped to handle adjustment costs. This highlights the need for effective safety net programs to be in place to assist the poor 37

Policies to Reduce Non-Tariff Barriers and Services Costs and to Improve Trade Facilitation
Cadot and Gourdon (2014)  Crucially, there should be public-private partnerships that involve the active engagement of the business community, economic policy officials and regulators organized around the supply chain as a whole in a sector or area of trade. The process would be overseen by a focal point within government with a mandate to coordinate and oversee all regulation that directly affects supply chain efficiency. To be effective in reducing trade costs requires coherence and coordination across many government agencies and collaboration with industry. The process should generate information on sources of trade costs through regular assessments of regulatory trade barriers and costs, and concrete action agendas and proposals for reforms.

Business Services
Air

Business Services
Air

Business Services
Air

Business Services
Air

Business Services
Air         0.0 -0.1 *Reductions apply to the EACU countries or the Tripartite countries depending on whether the scenario is EACU or Tripartite. ** Trade facilitation within EACU is part of the "EACU liberal" scenario also.  wood and paper products 3.2 6.3 9.5 Key: σ(qi, qj): Elasticity of substitution between firm varieties in imperfectly competitive sectors σ(va, bs): Elasticity of substitution between value-added and business services σ(D, M): Elasticity of substitution between domestic goods and imports in CRTS sectors σ(M, M): Elasticity of substitution between imports from different regions in CRTS sectors σ(L, K): Elasticity of substitution between primary factors of production in value added σ(A1,…An): Elasticity of substitution in intermediate production between composite Armington aggregate goods εROW, εEU, εCHINA, εUSA εEACU, εCOMESA, εSADC: Vectors of elasticities of imperfectly competitive firms' supply in the Rest World, EU, China. USA, EACU, COMESA and SADC with respect to the price of their outputs. θr: Share of rents in services sectors captured by domestic agents θm: Shares of value added in multinational firms due to specialized primary factor imports Source: Authors' estimates.