Comparison of Deep Regional Integration in the Melitz, Krugman and Armington Models: The Case of the Philippines in Rcep

This paper estimates the impacts on The Philippines of deep integration in a modern mega-preferential trade agreement, the Regional Comprehensive Economic Partnership. The paper assesses how the results differ with three versions of market structure: (i) perfect competition, Armington style; (ii) monopolistic competition Krugman style; and heterogeneous firms, Melitz style. The paper develops a new numerical model of foreign direct investment with heterogeneous firms where firms produce in the host country and demand corresponds to the ?proximity burden,? and is the first to apply a heterogeneous-firms model of foreign direct investment to preferential trade analysis. It also develops an extension of the Krugman model that allows small countries to impact the number of varieties. Both of these model extensions, as well as market structure, are crucial to the results. The trade and foreign direct investment responses are held constant across the three market structures. Lowering trade costs is examined from: (i) the reducing non-tariff barriers in goods; (ii) lowering barriers against foreign services providers, from foreign direct investment and cross-border; and (ii) facilitating trade. The results show that in all three market structures, there are substantial gains from deep integration, but virtually no gains from preferential tariff reduction. Both Krugman and Melitz style models produce significantly larger welfare gains than the Armington structure, especially in the impacts of foreign direct investment or with wider spillover effects on non- Regional Comprehensive Economic Partnership regions. The relationship between the welfare gains in the Krugman versus Melitz models is complex.


Policy Research Working Paper 8587
This paper estimates the impacts on The Philippines of deep integration in a modern mega-preferential trade agreement, the Regional Comprehensive Economic Partnership. The paper assesses how the results differ with three versions of market structure: (i) perfect competition, Armington style; (ii) monopolistic competition Krugman style; and heterogeneous firms, Melitz style. The paper develops a new numerical model of foreign direct investment with heterogeneous firms where firms produce in the host country and demand corresponds to the "proximity burden," and is the first to apply a heterogeneous-firms model of foreign direct investment to preferential trade analysis. It also develops an extension of the Krugman model that allows small countries to impact the number of varieties. Both of these model extensions, as well as market structure, are crucial to the results. The trade and foreign direct investment responses are held constant across the three market structures. Lowering trade costs is examined from: (i) the reducing non-tariff barriers in goods; (ii) lowering barriers against foreign services providers, from foreign direct investment and cross-border; and (ii) facilitating trade. The results show that in all three market structures, there are substantial gains from deep integration, but virtually no gains from preferential tariff reduction. Both Krugman and Melitz style models produce significantly larger welfare gains than the Armington structure, especially in the impacts of foreign direct investment or with wider spillover effects on non-Regional Comprehensive Economic Partnership regions. The relationship between the welfare gains in the Krugman versus Melitz models is complex. This paper is a product of the Macroeconomics, Trade and Investment Global Practice. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/research. The authors may be contacted at dgtarr@worldbank.org.

Introduction
The impact of the new trade theories on the welfare effects of modern preferential trade agreements (PTAs) has not been examined. Notwithstanding the initial results of Arkolakis, Costinot and Rodriguez-Clare (2012) on the welfare equivalence of the Armington, Krugman and Melitz models, further research by Costinot and Rodriguez-Clare (2014), Melitz and Redding (2015) and  has shown that substantial differences appear in the welfare results of these market structures when there are departures from the one-sector stylized model of Arkolakis et al. (2012). Our paper is the first to compare the welfare impacts of the Armington (1969), Krugman (1980) and Melitz (2003) trade theories within the context of PTAs. As argued by Arkolakis, Costinot and Rodriguez-Clare (2012) and Costinot and Rodriguez-Clare (2014), to obtain a fair comparison of the impact of market structure, we adjust elasticities in all three models, such that the trade response is equal in all three of our models.
Our estimated welfare gains in the monopolistic competition models are substantially larger than the estimated gains in the Armington model.
Further, most modern PTAs involve deep integration. Tariffs alone cannot explain the impact of preferential trade agreements (PTAs) on trade. Limão (2016, pp. 307, 312) notes that "observed tariffs can only explain a fraction of the PTA effect and that to fully explain the PTA effect requires an elasticity considerably higher than standard estimates, hence the elasticity puzzle….This justifies the widespread use of dummies in the gravity approach to capture other 2 channels through which PTAs can increase bilateral trade. But it also begs the question of what those channels are." This partly motivates the conclusion of Schiff and Winters (2003) that the real gains from regional trade initiatives come from deep integration.
Despite the evidence that tariffs miss the majority of the impacts of PTAs, the modern general equilibrium simulation analyses of regional trade agreements have focused on tariffs, and have produced rather small estimated welfare gains from preferential trade agreements (PTAs). 2 In this paper, in addition to tariffs, we consider deep integration via three broader trade costs channels through which PTAs may reduce trade costs: time in trade costs; non-tariff barriers on goods; and barriers to services. 3 We find that, across all three market structures, deep integration contributes substantially to welfare; but preferential tariff reduction contributes virtually nothing.
To capture the important commitments that countries make in modern PTAs in business services, we develop the first multi-sector numerical model of foreign direct investment (FDI) with heterogeneous firms that captures the key stylized facts regarding FDI, including the defining characteristics of the knowledge capital model (Markusen, 2002). 4 We apply our FDI model to business services, as these are the sectors in which FDI is most actively negotiated in international trade agreements. Our FDI firms produce in our host country (The Philippines); and, reflecting the literature on the proximity burden in services, the demand structure in the host country distinguishes between services provided with a local presence and foreign produced services. That is, the "proximity burden" literature notes that foreign services provided through FDI are better substitutes for host country domestically provided services than services provided cross-border (Francois and Hoekman, 2010). In our Krugman and Melitz versions, our FDI firms interact under monopolistic competition with endogenous productivity effects in the use of their outputs through the Dixit-Stiglitz mechanism from additional varieties of business services.
As such, our model incorporates the economic theory and the substantial and growing empirical literature based on firm level data showing that foreign direct investment and the wide 3 availability of business services results in total factor productivity gains to the manufacturing sector and the economy of the host country more broadly. 5 We extend to FDI the logic of Arkolakis et al. (2012) Francois and Hoekman (2010) for a survey of the theory and more than a dozen empirical studies that support this finding. In addition, in recent years, several studies that use firm level data support the finding that FDI and the wide availability of business services results in total factor productivity gains to the manufacturing sector and the economy broadly. These include Arnold et al. (2011) for the Czech Republic, Fernandes and Paunov (2012) for Chile, Arnold et al. (2016) for India, Shepotylo and Vakhitov (2015) for Ukraine and Duggan et al. (2013) for Indonesia. Two of the important theoretical papers are Markusen (1989;1995 provide an aggregate estimate of the welfare impact of RCEP, plus eight additional simulations that isolate the various components of RCEP. These latter simulations allow us to assess the relative importance of the deep integration components, as well as preferential tariff reductions.
Regarding market structure, our estimated gains in the Krugman model are about double the estimated gains in the Armington model. We find that the biggest difference in the results between the Armington model and the monopolistic competition models is the difference in the estimated gains from FDI reform in services; both monopolistic competition models produce estimated gains more than twice that of the Armington model for FDI reform in services.
Although not a general result, our overall estimated welfare gains in our Melitz model fall between the Krugman and Armington models.
We take advantage of the calibrated share form of CES technologies and preferences suggested by Rutherford (2002). We reduce information requirements by choosing units such that initial prices are unity. Analogous to the "exact hat" approach of authors such as Dekle et al. (2008) and Costinot and Rodriguez-Clare (2014), 8 we recover the proportional changes in prices and quantities as the equilibrium solution to the numerical problem.
The paper is organized as follows. We provide a review of the related literature in section 2 and an overview of the model in section 3. In section 4 we explain the data that we have developed or used in constructing this model. Results are presented in section 5. We conduct sensitivity analysis in section 6 to (i) spillovers in the policy scenario; (ii) the trade and FDI responses; and (iii) parameters of the model. We conclude in section 7.

Estimates of Goods Market Preferential Liberalization: Beyond Tariffs or Armington
The formation of the Canada-US free trade agreement led to the pioneering work on imperfect competition in applied trade policy analysis by Harris (1984). He showed that if the agreement leads to a more competitive pricing strategy by imperfectly competitive Canadian firms, 8 The term "exact-hat" refers to the characterization of equilibrium impacts in proportional changes. That is, if v is the change in a variable denoted in the benchmark and counterfactual as v and v′, respectively, then v can be summarized as ˆ' there would be substantial welfare gains. The creation of the single market in the European Union led to innovative analysis that required the use of multi-region models with imperfect competition to capture the competition aspects of the single market (Harrison, Rutherford and Tarr, 1996;Smith and Venables, 1988). 9 These studies considered trade facilitation benefits and linked the competitive aspects of the single market to the standards and product regulation issues. Tariff changes were ignored since there already was free trade within the European Union prior to the single market reforms. That is, the single market reforms were exclusively about deep integration.
The North American Free Trade Agreement (NAFTA) led to many numerical studies summarized in the Francois and Shiells (1994) volume. One of the more interesting was by Levy and van Wijnbergen (1995). They use their dynamic model to argue that dynamic incentive problems in adjustment policies for Mexican agriculture imply that adjustment policies should focus on increasing the value of the assets of poor farmers, not their incomes. Preferential arrangements of the European Union with its Mediterranean and Eastern neighbors led to several assessments of deep aspects of these agreements in Armington models, including Harrison, Rutherford and Tarr (1997a) for Turkey and Rutherford, Rutstrom and Tarr (1997; for Morocco and Tunisia. Maliszewska et al. (2009) employed a multi-region model with imperfect competition to examine bilateral deep integration between the EU and five countries in the former Soviet Union: the Russian Federation, Armenia, Azerbaijan, Georgia and Ukraine. They estimated substantial welfare gains to the partner countries of the EU from deep integration, but acknowledge their estimates are upward biased since they are based on a comparative steady-state model. Harrison, Rutherford and Tarr (2002) estimated that Chile would lose from individual preferential arrangements with Southern neighbors unless it lowered its 11 percent uniform tariff in place at the time. However, due to substantial estimated terms-of-trade gains to Chile in partner markets, the collective impact of its "additive regionalism" or "competitive regionalism" strategy produced estimated welfare gains many multiples of unilateral tariff liberalization. 10 Rutherford and Tarr (2003) showed that simply making the Chilean model dynamic will not increase the estimated gains from these agreements if there are no endogenous productivity effects. Balistreri, Tarr and Yonezawa (2015) examined deep integration in East and Southern Africa, while  extended the analysis to a poverty and income distribution 6 application. These studies found significant gains to the member countries from deep integration, but tariffs had only a negligible impact.
In a numerical application of the Eaton and Kortum model, Caliendo and Parro (2015) focus on tariffs and find that NAFTA increased the welfare of Mexico by 1.31 percent, the U.S. by 0.08 percent and the welfare of Canada declined by 0.06 percent. In a model with heterogeneous firms, Caliendo, Feenstra, Romalis and Taylor (2017, p. 3) also focus on tariffs and conclude that "PTAs contributed virtually nothing to total world trade and welfare."

Estimates of Foreign Direct Investment Liberalization in Services
The theory and empirical work on FDI (Markusen, 1989(Markusen, , 1995(Markusen, , 2002Francois and Hoekman, 2010;Dunning, 1985) argues that firms that engage in FDI are intensive in the use of knowledge capital and have created firm specific assets (like blueprints, patents, special formulae or reputation and managerial expertise) that their subsidiaries may use at low cost to their foreign affiliates; but arms-length transactions in these assets are very difficult or impossible. The decision to locate a production facility abroad is based on multiple considerations: (i) in services, which is our focus, the "proximity burden" implies that only local production can effectively compete with host country services; (ii) high transport costs or tariffs may make sales from abroad too costly; and (iii) low production costs in the host country may make FDI advantageous. In addition, we have cited above numerous studies that show that in services, additional providers of services provide productivity gains to local firms. Petri (1997), using a perfectly competitive Armington style model, was the first to capture many of these key features in a CGE framework. His model included separate production 7 structures of the parent and subsidiary and the implications for the demand structure.
Reforms in his model impact the global reallocation of capital. But there is no zero-profit constraint for firms and no endogenous productivity impacts from the actions of imperfectly competitive firms. Building on Petri, Dee et al. (2003) and Brown and Stern (2001)  Markusen, Rutherford and Tarr (2005) developed the first numerical model of FDI in an imperfectly competitive framework that incorporated the above stylized facts with an entry and exit decision by the foreign and host country firms based on zero profit constraint for the firm types. They show that the endogenous productivity effects from the Dixit-Stiglitz externality in their model results in important differences from the implications of a Heckscher-Ohlin model.
Their model was applied to datasets and policy issues of real economies in small open economy models by Jensen, Rutherford and Tarr (2006;2007;2010), Rutherford and Tarr (2008;2010) and Balistreri, Rutherford and Tarr (2009) in Russia, Kenya and Tanzania. But these models could not assess regional preferences in services. Konan and Maskus (2006) Balistreri, Jensen and Tarr (2015) and especially Jensen and Tarr (2012). Balistreri, Jensen and Tarr (2015) have shown that there is an analogy to trade diversion in goods whereby preferential 8 commitments to foreign investors in services could be immiserizing. Jensen and Tarr (2012) extended the analysis to include the impact of improved trade facilitation and the reduction of nontariff barriers in Armenia. Neither Balistreri, Jensen and Tarr (2015) nor Jensen and Tarr (2012) held the trade and FDI responses constant between the Krugman and Armington structures; nor did they consider heterogeneous firms.

Modern Assessment of the Welfare Gains from Trade Liberalization
The early CGE literature was based on constant returns to scale models, where the gains were based on comparative advantage and calculated from "Harberger triangles." The estimated gains from trade liberalization were sometimes characterized by the "Harberger constant," i.e., the gains were generally less than one percent of GDP from trade liberalization. Among others, de Melo and Tarr (1990) and Jensen and Tarr (2003) showed that, even in a perfect competition constant returns to scale model, if there were rents involved, the gains could be many multiples of the gains from the "Harberger triangles." Harrison, Rutherford and Tarr (1997) found that rationalization gains in imperfectly competitive quantity adjusting models did not produce significant gains above perfectly competitive models. Rutherford and Tarr (2002) showed that in a fully dynamic model based on Paul Romer style endogenous growth with gains from variety, the gains from trade liberalization would be many multiples of the gains in a constant returns to scale model. Markusen, Rutherford and Tarr (2005) and Rutherford and Tarr (2008) showed that introducing foreign direct investment in services in a monopolistic competition model would substantially increase the welfare gains. Francois, Manchin and Martin (2013) have summarized many approaches to modeling market structure in CGE models and suggested ways that the alternate model structures could be tested.
Regarding heterogeneous firms, the first effort at a CGE model was by Zhai (2008 This explains why Petri, Plummer and Zhai obtain such large increases in the variety externality 9 in their model. The first numerical model of real economies that has been developed that is consistent with the model of Melitz (2003) is Balistreri, Hillberry and Rutherford (2011). In their multi-region model they find that the welfare gains from tariff reductions are several times larger than with a standard CRTS trade model. Jafari and Britz (2017) (3) seven services sectors in which there is monopolistic competition and foreign direct investment. The cost, production and pricing structures in the three categories differ, but regardless of sector, all firms minimize the cost of production.
Primary factors are skilled labor, unskilled labor and capital (including land). 12 Regarding capital, there is mobile capital and sector-specific capital in monopolistically competitive goods sectors and services sectors with FDI; and 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 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 up. 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 and productivity gain from trade for the importing country depends on the research and development stock of the exporting country. 13 In each of our experiments we hold real government consumption fixed (in both its overall size and in its commodity composition) via an endogenous transfer between the representative household and the public sector. In the Public Finance literature this is referred to as a 'differential' analysis (see, for example, Ballard, 1990). If we assume that the benefits from public expenditures accruing to household is separable from private consumption of goods and leisure, our approach allows us to evaluate welfare impacts without making inferences about the value households place on public expenditures. A reduction in tariff revenue is compensated through a direct transfer so the benefits associated with public expenditure are unchanged, and we can measure welfare using equivalent variation in private consumption. 14

Perfectly competitive goods and services sectors
In these sectors, we employ the "Armington" structure, with goods and services differentiated by the country of origin. Exports are also differentiated from products produced for the home market. For exports and domestic goods, we use a constant elasticity of transformation production function with elasticity of transformation equal to four for all perfectly competitive sectors. Prices in foreign markets to exporters from The Philippines are perfectly elastic in these sectors, but they may change in response to policy variables.

Imperfectly competitive goods sectors (Krugman style model)
Goods in these seven sectors (and all IRTS services) are differentiated at the firm level. import price of foreign goods is simply defined by the import price, and, by the zero profits assumption, in equilibrium the import price must cover fixed and marginal costs of foreign firms.

12
In this model, consistent with firm level product differentiation, we assume that the elasticity of demand in each of the export markets is the Dixit-Stiglitz elasticity of demand. Filipino firms then set marginal revenue equal to marginal costs in each of the seven export markets. If the partner countries of The Philippines reduce their barriers preferentially against exporters from The Philippines, that will induce entry from firms in The Philippines. Introducing downward sloping demand curves into the model, however, means that there are terms-of-trade affects to consider in this model that were not present in the Rutherford and Tarr (2008) model. 15 We present a parameter in each of our scenarios to show the percentage change in the terms-of-trade.
Following Krugman (1980), we assume that imperfectly competitive firms have a fixed cost of production and that marginal costs are constant with respect to output. Then, suppressing subscripts for firms, sectors and regions, total costs are: where TC is total costs, MC is marginal costs, FC is fixed costs, q is output of the firm and p is a vector of factor prices. Following the literature (e.g., Helpman and Krugman, 1985), we assume that the input proportions of fixed and marginal costs are identical, from which it follows that the ratio of fixed to marginal costs is constant. That is, for all firms producing under increasing returns to scale (in both goods and services), we have: Equations (1) and (2), in the Chamberlinian framework, imply that output per firm remains constant, i.e., the model does not produce rationalization gains or losses. The number of varieties affects the productivity of the use of imperfectly competitive goods based on the standard Dixit-Stiglitz formulation. The effective cost function for users of goods produced subject to increasing returns to scale declines in the total number of firms in the industry.

Imperfectly competitive goods sectors (Melitz model)
We follow Melitz (2003)  depend on the productivity draw of firms. Firms do not know their productivity prior to entering the market. They receive a productivity draw from an untruncated Pareto probability density 13 function after entering the market which determines their marginal costs. This gives rise to a heterogeneous structure of firms regarding their productivity, price and output. More productive firms charge a lower price and capture a larger market share. There is a cutoff productivity level, below which, firms do not produce, i.e., they are not active.
Firms face an additional fixed cost of selling in any export market. This gives rise to a second cutoff where only the more productive firms export as well as sell in their domestic market.
Crucially, quasi-rents on exports of the firm must cover the fixed costs of exporting to that market. Filipino exports. This explains our results below where we find that the value of the Dixit-Stiglitz externality is larger in the Krugman model. We discuss these issues in more detail in section 5.3.

Imperfectly competitive service sectors in which foreign direct investment occurs (Krugman style)
In these services sectors, we observe that some services are provided by foreign service providers on a cross border basis analogous to goods supply from abroad. But a large share of business services is provided by service providers with a domestic presence, both multinational and local. 16 Our model allows for both types of provision of foreign services in these sectors. 16 One estimate puts the world-wide cross-border share of trade in services at 41% and the share of trade in services provided by multinational affiliates at 38%. Travel expenditures (20%) and compensation to employees working abroad (1%) make up the difference. See Brown and Stern (2001,  The cost, production, demand and competition structure for firms in this group of industries follows the same structure as the imperfectly competitive goods firms with two differences. The first difference is that we allow multinational service firms to establish a local presence to produce in The Philippines and compete directly with Filipino service firms. Multinational service firms produce a Filipino region-specific variety in The Philippines, which is differentiated from Filipino varieties and the varieties of other multinational services firms in The Philippines. Crucially, all firms (foreign and domestic) incur a fixed cost of operating in The Philippines.
For domestic firms, costs are defined by the costs of local primary factors and intermediate inputs. When multinationals service providers decide to establish a local presence in The Philippines, they will predominantly use Filipino inputs; but they will also import some of the specialized technology or management expertise of the parent firm. That is, foreign direct investment generally entails importing specialized foreign inputs. Thus, the cost structure of multinationals differs from Filipino service providers. Multinationals incur costs related to both imported inputs and local primary factors, in addition to intermediate factor inputs. Provision of services through FDI differs from cross-border provision of services or exports of goods, since services provided through FDI use predominantly local primary inputs, i.e., Filipino inputs in our case.
Our source or parent company produces specialized technology or management techniques and obtains a payment for these goods from its subsidiaries or licensees; but it does not produce the business services that are the sectors of our model. The foreign firms who supply The Philippines through FDI, either sell through FDI or do not produce anything.
For multinational firms, the barriers to foreign direct investment raise their costs of production. The reduction of the barriers lowers these costs, raises the profitability of FDI and induces entry by multinationals until zero expected profit is restored. This leads to a welfare gain from the Dixit-Stiglitz variety externality. In addition, liberalization of FDI barriers frees capital and labor that was used to overcome the barriers for use elsewhere in the economy. In all model variants, including the Armington model, we assume that the reduction in the constraints on foreign direct investment allows the domestic economy to capture rent rectangles. In addition, thirds of services exports were from FDI and about one-third from cross-border sales. See Markusen and Strand (2009,  reducing barriers induces foreign entry until profits are driven to zero, so there are also "Harberger" triangles of efficiency gains.

Imperfectly competitive service sectors in which foreign direct investment occurs (Melitz style).
We develop a new innovative model of FDI with heterogeneous firms. The key departure from the Krugman style model of FDI described above is that the marginal costs of firms depend on the productivity draw of firms. Foreign firms who wish to enter the market of The Philippines do not know their productivity prior to entering the Filipino market. They receive a productivity draw from an untruncated Pareto probability density function after entering the market which determines their marginal costs. This gives rise to a heterogeneous structure of foreign firms regarding their productivity, price and output. There is a cutoff productivity level, below which, foreign firms do not provide the service in The Philippines.
We assume foreign firms who provide services through FDI in The Philippines, produce in The Philippines and use predominantly Filipino primary inputs. We believe this is a crucial property of FDI. For example, the two major mobile telephone companies in The Philippines, which are primarily owned by nationals of Hong Kong SAR, China and Singapore, use almost all Filipino workers. Consistent with the knowledge capital model, our multinationals operating in The Philippines import specialized technology or management techniques from their parent firm; but they do not produce the final product in the source country.
Our model is also consistent with the proximity burden literature in services that argues that a local presence is required for foreign firms to compete effectively with host country services (see Francois and Hoekman, 2010). A local presence in services allows FDI provided services to be better substitutes for host country services than cross-border services. It follows that services provided through FDI and cross-border services are not perfect substitutes. For example, banks with a physical presence in The Philippines are a better substitute for Filipino banks in The Philippines than foreign banks that provide banking services cross-border on an electronic basis. More strikingly, in services such as mobile telephone services or trucking services, it is very difficult for foreign firms to compete with domestic firms without a physical presence in the host country. Our model incorporates the stylized fact that FDI services are better substitutes for host country services than cross-border services. 16 Our source or parent company produces specialized technology or management techniques and obtains a payment for these goods from its subsidiaries or licensees; but it does not produce the business services that are the sectors of our model. The foreign firms that supply The Philippines through FDI, either sell through FDI or do not produce anything. So the productivity cutoff for these firms to be active in the host country market is their sole productivity cutoff.

Labor-Leisure Choice
We assume that there is a labor-leisure choice, where utility may be represented by a CES function that is weakly separable between leisure and all goods and services consumption. We consider an aggregate non-leisure consumption good C, with a dual price P. Utility is: Where 0 C = leisure and C = consumption of the aggregate good/service.
Let E = the total time endowment of the consumer/worker; W = the wage rate; L = labor supply; and P = the price index of goods and services. The demand for leisure is: The uncompensated elasticity of leisure demand with respect to the wage rate is: and the uncompensated elasticity of labor supply with respect to the wage rate is 17 : We evaluate the labor supply elasticity in the neighborhood of the initial equilibrium, where we choose units such that W = P = k = 1. Then we have: The values of L and Y are data in the initial equilibrium. In the special case of a single primary factor, zero trade balance and zero non-labor factor income, the elasticity of labor supply reduces to: Then the elasticity of labor supply is positive if an only if the elasticity of substitution exceeds unity.
We prefer to avoid assuming a value of the time endowment E and the share of the time endowment devoted to leisure μ, since these may be arbitrary and Ballard (2000) has shown that the impact can be substantial. Instead, based on empirical estimates, we assume that the value of the uncompensated elasticity of labor supply is 0.2 and the compensated elasticity of labor supply of 0.7. Given the unshown expression for the compensated elasticity of labor supply, these values allow us to solve for (calibrate) the parameters on the right-hand side of the elasticity of labor supply.

The Multi-Sector Feenstra Ratio
Feenstra derived a measure we call the Feenstra ratio that precisely measures the welfare impact of the variety externality in a one-sector monopolistic competition model. We extend that measure to a multi-sector, multi-region Krugman or Melitz model. Details are in Balistreri and Tarr (2018, appendix A). Here we provide the key results.
Despite the fact that all varieties are consumed everywhere in the Krugman model, Feenstra (1994;2010) has shown that in a one-sector model, the welfare impact of a variety depends on its expenditure share. From Feenstra (2010, Theorem 2), the Feenstra ratio is: where e is the unit expenditure function, and t t p q are the vectors of prices and quantities in period t, t I is the set of goods available in period t at prices t p , I is the set of goods available in both period t and t-1 and (...) SV P is the Sato-Vartia index that shows the ratio of the unit expenditure function in the two periods if the sets of available goods in the two periods are identical. Feenstra's theorem tells us that the proportional change in the unit expenditure function is the product of two terms that are: (i) a measure of the change in the prices charged by firms on goods common to the two periods (the Sato-Vartia index); and (ii) a ratio that is a measure of the value of the variety gain, the Feenstra ratio. The methodology requires distinguishing changes in the price charged by firms on goods available in both periods from the variety externality. Feenstra shows that 1 ( ) t I   is the expenditure on new goods relative to total expenditure in period t, so a larger number of new goods in period t will tend to lower ( ) t I  . A value of 18 1.01 for our Feenstra ratio means that the cost of a unit of utility declined by one percent due to new varieties.
In a multi-sector, multi-region model, we define the Feenstra ratio for region s as: where is  is the economy-wide expenditure share (absorption share) on goods or services of sector i in region s; and is F is the Feenstra ratio for sector i in region s defined as: PDS t is the Dixit-Stiglitz price index in sector i in region s, and SV is P is the Sato-Vartia index for sector i in region s. Given that the Sato-Vartia index remains an "ideal" index of firm level prices at the sub-sector level, and the Dixit-Stiglitz index is the unit cost of goods in sector i taking variety into account, our index is F is precise at the sector level. If we have a one-sector model, our measure reduces to the ratio of Feenstra (2010). Given the existence of intermediates in our model, the aggregation across sectors is an approximation.

Terms-of-Trade Parameter
We define the terms-of-trade for region r as the weighted average of export prices divided by the weighted average of import prices: In our definition for the terms-of-trade, is p is the composite price of the i-th good in region s. This is the dual price variable that we define as the left-hand side variable in Balistreri and Tarr (2018, appendix B). is p incorporates the technology and the optimization, and its depends on the Armington, Krugman or Melitz market structure. The other parameters are weights, which we define as follows.
Define isr X is the total value of country r's expenditures on good i from country s (excluding tariffs


Where isr tar is the tariff rate on imports from region s shipped to region r. Note that for exports, there is a unique weight associated with any good in a particular region; but the import weights Indonesia has many characteristics similar to The Philippines that makes it a good proxy: their annual per capita incomes are close. According to the IMF in 2016, in thousands of USD on a PPP basis, Indonesia's per capita income was 10.5 and The Philippines is 8.3; they are both island economies with populations of more than 100 million and both are ASEAN members. 20 In the "other manufacturing" sector, of 44 estimates, we excluded two outliers of 33.5 and 35.3; other values were less than 9. In "textiles and apparel," of 23 estimates, we excluded three outliers with values above 33.5. All other estimates were less than 11.4.

FDI Response.
By similar reasoning to holding the trade response constant across the market structures, it seems appropriate for a fair comparison across the market structures to hold the FDI response constant across the market structures as well. For the FDI response, the obvious choice for calibration of the FDI response is the scenario where we reduce barriers against FDI in the Krugman model. We then adjust the elasticities in the Armington and Melitz models that determine the FDI response to match the FDI response of the Krugman model. Having determined the elasticities of the Armington, Krugman and Melitz models, based on the trade and FDI responses for these two scenarios, we hold those elasticities fixed in all scenarios. In section 5, we provide more details on the magnitude of the trade and FDI responses and the adjustments necessary to hold these constant.

Elasticities of Supply.
Beginning with the path-breaking work of Coe and Helpman (1995), a rich literature now exists that has empirically investigated the transmission of knowledge through the purchase of imported intermediate goods and through foreign direct investment. Coe and Helpman found that OECD countries benefit from foreign research and development (R&D), that they benefit more from trading with countries that have a larger stock of research and development, and that the benefits are greater the more open the country is to foreign trade. Moreover, while in large countries the elasticity of total factor productivity (TFP) with respect to domestic R&D capital stocks is larger than that with respect to foreign R&D capital stocks, the opposite holds in small 22 countries; that is, foreign R&D is more important for small countries. Coe, Helpman, and Hoffmaister (1997) extend these results based on a sample of 77 developing countries. They find developing countries that do little R&D on their own, have benefited substantially from industrialized country R&D through trade in intermediate products and capital equipment with industrialized countries. Regarding the impact of FDI in services, we have cited numerous papers above that establish that FDI in services increases TFP in the economy.
In summary, this literature shows that the purchase of intermediate inputs from industrialized countries and FDI in services is an important mechanism for the transmission of R&D and productivity growth in developing countries. For small developing countries, trading with large technologically advanced countries is crucial for TFP growth. In our model, the parameter that reflects the ability of a region to increase total factor productivity through the transmission of modern technologies is the elasticity of firm supply with respect to the price. The greater the elasticity of firm supply in a sector the more varieties will be received in response to a price increase with respect to that country. Based on these considerations, in the increasing returns to scale sectors of our model, we assume that the elasticity of firms' supply with respect to price is 3 for all IRTS sectors in The Philippines, India and the Rest of ASEAN; 4 for China and 5 for all other regions. These values are in a footnote to table 4. We conduct sensitivity analysis on these parameters, to determine the impact of these parameter values on the results. authors. In this manner, they score the regulatory regimes of the 11 sectors in the 103 countries according to the Australian authors' criteria.

Estimates of AVEs of Barriers to Cross-Border Trade in Services. Several authors have used gravity models to estimate AVEs of barriers to cross-border trade in services.
The most comprehensive of these efforts is by Francois, Hoekman and Woerz (2007

Estimates of the AVEs of the Costs of Time in Exporting and Importing.
In order to estimate the impact of improved trade facilitation, in this paper we apply the time cost of trade dataset of Hummels and Schaur (2013)

Estimates of the AVEs for Non-Tariff Measures (NTMs) for the Regions of our Model
Cadot and Gourdon (2014) 26 In the case of agriculture, we apply the AVE of the non-tariff barriers in our model to agriculture and forestry; and to wheat and cereals. In manufacturing we applied the AVE of the non-tariff barriers to beverages and tobacco; chemicals, minerals and metal products; electronic equipment; food products; other manufacturing; petroleum and coal products; textiles and apparel; meat and dairy products; processed rice and sugar.

Tariff and Trade Data
Trade data and tariff rates are taken from the GTAP 9.1 database. For tariff rates, GTAP uses the MAcMap-HS6 database, described in Guimbard, Jean and Mimouni (2011

Social Accounting Matrices
The core structural data of the model comes from the GTAP 9.1 dataset. 27 Exports and imports and by trading partner of The Philippines from this source are available in tables 2.

Share of Market Captured by Foreign Direct Investors in Services
In the business services sectors of our model, it was necessary to calculate their market

Results: RCEP and Its Key Components: Comparing Market Structures
We evaluate RCEP and its components in three model classes: Melitz models (so as to not directly impact the variety externality); we achieve the equivalent increase in aggregate exports by adjusting the shape parameter in the Pareto distribution in all IRTS goods and services sectors. We scale down the shape parameter in our application by 0.98 (to a value of 4.49), which is a slightly more heterogeneous distribution than the distribution of Balistreri, Hillberry and Rutherford (2011) Kehoe (2003), that to match the trade response (in this case the FDI response) in the Armington model, elasticities have to be increased to unrealistic levels. 34 Having calibrated these elasticities to match the FDI response, we hold them fixed in all scenarios.

Scenario Definitions
We evaluate the impact of the principle changes under negotiation by the RCEP member and the aggregate of those changes. These changes are: (i) 50 percent reduction of the ad valorem equivalent of the barriers against providers of services through both foreign direct investment and through cross-border delivery of services; (ii) a full reduction in the tariffs by The Philippines on RCEP members and reciprocal action by the RCEP members on Filipino exports (ASEAN tariffs are assumed to be maximum 0-5 percent in the benchmark equilibrium); (iii) a twenty percent reduction in the ad valorem equivalent of non-tariff barriers in goods by The Philippines and reciprocal action by the RCEP members on Filipino exports; and (iv) a 20 percent reduction in the time in trade costs on both imports and exports. In the case of time in trade (or trade facilitation) costs, reforms to facilitate trade with RCEP partners inevitably will, at least partially, reduce these costs for third countries; thus, we include a spillover effect on non-RCEP countries of five percent. In the sensitivity analysis, we consider spillover impacts in NTMs and in services.
In order to assess the relative importance of the reforms, and to identify the reforms in RCEP most important to The Philippines, we decompose the reforms into their components and simulate each separately. In tables 5-7, the heading of the table specifies what we are assuming in our central RCEP scenario and in the scenarios that assess each of the reforms separately.

Effects of RCEP in the Krugman Model
Our aggregate and decomposed results for our Krugman model are presented in table 5.
There are nine scenarios, with the head of the columns labeled 1-9. Welfare gains are 3.1 percent of the value of goods and services consumption in the benchmark equilibrium or 2.5 percent of GDP. 35 Since our model is comparative static, these gains are to be interpreted as annual There are three sources of these gains. First, these deep integration measures result in gains since we assume that the maintenance the barriers imposes real resource costs. That is, liberalization of the barriers frees capital and labor that was used to overcome the barriers for use elsewhere in the economy allowing the domestic economy to capture rent rectangles. These rent recapture gains are also present in our perfectly competitive model. The second source of gains is that the reduction in barriers induces additional supply. In our Krugman model, additional supply derives from entry by new firms into the market (an extensive margin). In the perfectly 35 Note that we do not include the imputed value of leisure in our denominator. 29 competitive model, existing RCEP firms will sell more. So, there are "Harberger" triangles of efficiency gains (an intensive margin). 36 There is a third source of gains that is present in our Krugman model, but does not exist in the perfectly competitive model. This is the gains from the Dixit-Stiglitz externality due to additional varieties in the Krugman model. Take  It is striking that our results indicate that the change in import tariffs within RCEP does not contribute to the increase in welfare in the Philippines. Unlike the measures of deep integration we have just discussed, we apply the conventional assumption that tariffs do not induce the expenditure of capital and labor in the home country (and thus do not induce rents).
Consequently, there are no recaptured rents to be gained from tariff liberalization. Further, preferential tariff reduction may induce trade diversion, i.e., it may induce purchases from suppliers who are less efficient than third country suppliers. It has been well documented in economic theory (beginning with Viner, 1950), that the possible induced switch to less efficient suppliers in the partner regions away from more efficient suppliers on world markets may result in net losses from preferential tariff reduction. As noted above, small or no gains in the welfare impact of preferential tariff reduction is consistent with other modern assessments of the impact of preferential tariff liberalization, as in Caliendo and Parro (2015) and Caliendo et al. (2017).
These are the reasons we have focused our work on deep integration.

Effects of RCEP--Armington Compared with Krugman
Decomposed results for the Armington model are in table 7. Aggregate welfare gains from RCEP as a percent of benchmark consumption are: 1.5 percent in the Armington model as opposed In the Krugman model, when barriers against Filipino exports by RCEP partners are removed, there is an extensive margin response from the entry of new firms that magnifies the gains from trade. That is, the additional profits on exports induce more entry as the quasi-rents on exports help to cover the fixed production costs. Firm entry continues until there are zero profits.
Since all domestic firms have the same cost structure in the Krugman model, any firm that supplies the export market also supplies the domestic market. So, there are more domestic varieties available as a result of increased profits on exports which increases welfare above the Armington model. Crucially for our interpretation below regarding the difference between the Melitz and Krugman mhodel, since there is no reduction of barriers against foreign firms in The Philippines 31 in this scenario, there is no loss of domestic varieties. Our Feenstra ratio shows an estimated gain in the net value of varieties.

Effects of RCEP: Melitz Compared with Krugman
Our This real exchange rate depreciation provides an incentive to sell more exports compared to selling on the domestic market. Then the argument above applies again, but to a lesser degree. That is, the variety effects in the Krugman model will again be slightly larger due to the absence of the selection effect on Filipino firms that reduces domestic varieties. Our calculated Feenstra ratios show larger gains from the value of varieties in the Krugman model in these cases.
We do not believe the ranking of Krugman versus Melitz in this application is a general result. First, it depends on our extension of the Krugman model. Further, using a multi-region model of world trade,

Sensitivity Analysis
We conduct three types of sensitivity analysis. First, we conduct policy sensitivity with respect to an assumption of wider spillovers with respect to the rest of the world. In these policy scenarios, we assume non-RCEP regions obtain a reduction of ten percent of ad valorem equivalents of barriers against FDI in services and five percent in non-tariff measures. We

Sensitivity of Deep Integration to Spillovers or Liberalization to Non-RCEP Regions.
Baldwin (2014) has argued that compared to regional preferences regarding tariffs, the deep integration aspects of 21 st century regional agreements are relatively difficult to limit to partners to the agreement; further, global value chain considerations lead to a "multilateralization" of some of the deep integration aspects of 21 st century regional agreements. 37 Feenstra (1994;2010) has shown that if new varieties are consumed in lower quantities, then their impact on the quality adjusted price index or the expenditure function will be less compared to where the same number of varieties are consumed in equal amounts. Arkolakis et al. (2008) argue that the firms with the highest productivity are already present in the market prior to the trade liberalization. Thus, the Melitz model would imply that the firms that enter after trade liberalization have lower productivity and lower value of sales on average compared with incumbent firms. Calling this the "curvature" effect, Arkolakis et al. (2008) apply the Feenstra methodology to conclude the variety gains from trade in the Melitz model are muted.
We argued above that measures that reduce the time in trade costs for RCEP members would inevitably convey at least a partial benefit to non-members. For that reason, in our central scenario, we allowed non-RCEP regions to realize a five percent reduction in their time in trade costs on exports to and imports from The Philippines, while RCEP members obtained a 20 percent reduction. Regarding preferential liberalization of barriers against foreign investors in services, it is an unsettled question of how feasible it is to exclude third countries from preferential liberalization in services. 38  and "Harberger triangle" type efficiency gains. There is a substantially larger increase in the wider spillover case, however, in the estimated welfare gains in both monopolistic competition models. Compared with our central scenario, they increase by 1.8 percent of consumption in both the Krugman and Melitz models, to 4.9 percent and 3.9 percent, respectively. Our Feenstra 38 If the preferential agreement grants equivalent rights to third country firms located in the partner region, the preferential arrangement becomes somewhat multilateral. The rules of origin would impact how multilateral the preferential liberalization becomes. What rules of origin apply in practice is an unsettled question both in the literature and in practice. Fink and Jansen (2009) note that, typically, FTAs require that enterprises eligible for the agreement's preferences are incorporated under the laws of one of the partner countries. Further, to qualify for preferences, the enterprise must have "substantial business activities" within the region. This indicates that preferences do not extend to enterprises located in third countries if they are not incorporated with substantial business interests in the region. As an example of these principles, Fink and Molinuevo (2007) note that in East Asia non-parties can benefit from the preferences provided in the FTA, as long as they establish a juridical person in one of the FTA member countries and are commercially active in that country. But again, the preferences for non-parties are enterprise specific and do not extend to enterprises without a commercial preference with substantial business interest.
35 ratio parameter shows a larger increase in the Dixit-Stiglitz variety externality in the case of spillovers which primarily explains the widening of the difference in the estimated welfare gains between the monopolistic competition models and the Armington model. These results suggest that substantial gains can be achieved for the members of a regional agreement if the trade costs reduction of deep integration can be extended to third countries.

Sensitivity to the Trade Response and FDI responses
In these scenarios, we test the sensitivity of the welfare results to the trade response and the FDI response. As in our central scenarios, for the trade response sensitivity, we focus on the scenario where we reduce the ad valorem equivalents of the non-tariff measures. We choose this scenario since it is the scenario closest to a gravity model estimate of the trade response impact of a reduction of trade costs. For the FDI response, we choose the scenario where we reduce the barriers against FDI. Our results are in table 9.

Sensitivity to the Trade Response
To an increase in their price, i.e. an adverse terms-of-trade effect. With the larger elasticity of demand for imports, the adverse terms-of-trade effect is larger, which explains the result in the table. We reran the Armington scenario with close to perfectly elastic supply of imports (so there is no adverse terms-of-trade effect) and obtained the result that the welfare gains are higher than in the high trade elasticity case.

37
The Krugman and Melitz models produce very close estimates of the changes in welfare as a percent of consumption. In both the Krugman and Melitz models, the welfare impacts vary inversely with the Dixit-Stiglitz elasticities. The FDI response is not changing substantially between these low and high parameter values, so the differences in the welfare estimates are primarily determined by the valuation of varieties.
There is no difference in the welfare estimates in the Armington cases in the table. This reflects the lack of a change in the FDI response due to the upper bound on the elasticities of substitution of 90.

Piecemeal Parameter Sensitivity of the Krugman Model Welfare Estimates in the RCEP Scenario
In these simulations, we examine how robust are the welfare estimates in our central scenario, which is RCEP in the Krugman model. We examine the sensitivity of the results to ten sets of parameters listed in

Conclusions
We found that The Philippines can experience a welfare gain of 3.  .
They show that the ranking of the relative gains will depend on these model variants as well as scenarios, parameters and data. That is, a Melitz or Krugman style model will not dominate the other regarding welfare impacts in all model variants.

IRTS Goods
Beverages and Tobacco

Welfare is Hicksian Equivalent Variation (EV) as a Percent on Benchmark Consumption
*For a solution to the Melitz model, the Pareto shape parameter is restricted to be greater than the elasticity of substitution minus one: a(i) > [σ(i)-1]; since σ(i) = 5.1 in textiles and apparel, we must restrict a > 4.1. Then our multiplier on "a" may not be below .9 **See