WPS6106 Policy Research Working Paper 6106 CGE Modeling of Market Access in Services Elisabeth Christen Joseph Francois Bernard Hoekman The World Bank Poverty Reduction and Economic Management Network International Trade Department June 2012 Policy Research Working Paper 6106 Abstract This paper examines how the applied multi-sector purposes. The third, linked to the data problem, is to computable general equilibrium (CGE) literature has quantify the barriers to be examined. Significant progress moved into quantication of the impacts of greater has been made in modeling foreign direct investment market access for services. This includes discussion of and linking this to productivity, which turns out to multi-sector linkages to the service sector, as well both be important. The paper also provides an example of measuring barriers to trade and investment (generally modeling productivity linkages to openness and domestic with a mix of firm surveys, price comparisons, and regulation, with an applied CGE model of Italy. This econometrics), and how changes in these barriers, illustrates cross-sector linkages and the integration of however measured, have been implemented in the CGE economic data and policy measures to define service literature. Three challenges are highlighted. The first is sector experiments. Priorities for future research include identification of how trade in services takes place and better modeling of market structure, the linkages between how market access is therefore affected by policy. The sectors and the complementarities between different second is to find data sufficiently robust for modeling modes of supplying services. This paper is a product of the Poverty Reduction and Economic Management Network. 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://econ.worldbank.org. The authors may be contacted at E.M.Christen@gmail.com; Joseph.Francois@gmail.com; or Bhoekman@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team ∗ CGE Modeling of Market Access in Services Elisabeth Christen a Universit¨t Innsbruck, and a Johannes Kepler Universit¨t Linz Joseph Francois a Johannes Kepler Universit¨t Linz, and CEPR Bernard Hoekman World Bank, and CEPR June 2012 Keywords: Trade in services, nontariff measures, nontariff barriers, regulation, FDI, productivity, liberalization, CGE. JEL codes: F10 ∗ We are grateful to Peter Dixon for tremendous patience, and to both him and David Tarr for very constructive and helpful comments on earlier versions of this paper. All views expressed are strictly personal and should not be attributed to the World Bank. This is a â€?working draftâ€? version of a handbook chapter. CGE Modeling of Market Access in Services Elisabeth Christen1,2 Joseph Francois1,3 Bernard Hoekman3,4 1 Introduction Services play an increasingly important role in world trade. According to the WTO, the global value of cross-border services exports in 2007 was $3 trillion, or about 20 percent of world trade in goods and services. Because services are less tradable than goods, this number signiï¬?cantly understates the importance of services in world trade. In practice, much services trade is indirect, in that services are embodied in traded goods, or are sold to foreign consumers through direct investment in affiliates. If account is taken of sales of services by foreign affiliates of multinational ï¬?rms, then the value of trade in services rises substantially. Data for 15 OECD countries puts the value of such sales at some $1.5 trillion in 2007 (WTO, 2009). If account is taken of the direct and indirect value added content – that is, if trade is measured in terms of the value that is added by processing of imported components into ï¬?nal products for export as opposed to measuring trade flows on the basis of the gross value of goods crossing the border, the share of services in world trade rises to almost 50 percent (Escaith, 2008; Francois, Manchin, and Tomberger 2012). While the expanding economic importance of services has not gone unnoticed, services have been downplayed in the economic growth and development litera- ture, and have only recently been highlighted in the multi-sector computable general equilibrium (CGE) literature. The trigger for greater attention was the emergence of services on the international policy agenda in the 1980s, in particu- lar the effort that was launched in the Uruguay Round to negotiate multilateral rules on policies affecting trade and investment in services. One result was to mobilize the ï¬?rst analytical contributions to the trade literature (e.g., Deardorff, 1985). This initial effort suggested that in contrast to the presumption that ser- vices were mostly of export interest to high-income countries, many economies had a potential interest in liberalizing trade in services. Thus, for example, many of the poorest developing countries have a revealed comparative advantage in ser- 1 Johannes Kepler Universit¨t, 2 Universit¨t Innsbruck, 3 CEPR, 4 World Bank. We are grate- a a ful to Peter Dixon for tremendous patience, and to both him and David Tarr for very construc- tive and helpful comments on earlier versions of this paper. All views expressed in this paper are strictly those of the authors, and do not reflect the official views or positions of any insti- tution with which they may be affiliated, or of anyone at those institutions. This is a â€?working draftâ€? version of a handbook chapter. This version date: June 2012. 1 vices (Hoekman, 1990).2 This realization helped to overcome some of the early resistance by developing countries to launching negotiations on trade in services and the creation of the WTO General Agreement on Trade in Services (GATS) in 1994. Building on the early analysis inspired by the Uruguay Round, there is now a growing CGE literature on market access in services.3 For services, market access can be difficult to deï¬?ne. In practice it is often hard to judge just how open a particular nation is to foreign participation because explicit (discriminary) barriers will be complemented by domestic regulation, standards and business practices. For this reason, improved market access - as market opening is known in WTO jargon - has tended to be qualitative as opposed to quantitative (e.g., an X percent reduction in a tax-equivalent). In the case of trade in goods the focus of negotiators is on tariffs and non-tariff barriers (quotas, etc.). For analysts, these are relatively straightforward to model, as tariffs can be incorporated directly into models, and non-tariff barriers can be converted into tariff equivalents. In the case of services there is much less scope for negotiators to target numeric measures of policy per se – e.g., the analogue of a 10 percent reduction in tariffs. The focus of policymakers has instead been on qualitative commitments regarding speciï¬?c policies that affect the cross-border movement of consumers and providers of services, or the establishment of foreign providers in a market, and to agree to disciplines on the use of measures that explicitly discriminate against foreign suppliers of services. There is an inconsistency between the basic or balance of payments deï¬?nition of trade, and the legal deï¬?nition of trade enshrined in the WTO’s General Agree- ment on Trade in Services. Under the GATS, four â€?modes of supplyâ€? are deï¬?ned through which trade can occur. These include foreign direct investment (FDI) and temporary cross-border movement of suppliers. As a result there is a tension between how trade in services has come to be deï¬?ned in policy circles and how trade is deï¬?ned in the national accounts and balance of payments, which does not consider movement of factors of production as international trade per se.4 In this paper, we examine how the recent multi-country CGE literature has moved from goods trade into quantiï¬?cation of market access in services. This in- cludes both a discussion of measuring barriers to trade and investment (generally with a mix of ï¬?rm surveys, price comparisons, and econometrics), and a discus- 2 The revealed comparative index is a measure of relative specialization, deï¬?ned as the share in total exports of a country of a given product, divided by the same ratio for the world as a whole. If this measure is greater than one a country is said to have a revealed comparative advantage in the product. The concept was ï¬?rst deï¬?ned and used in Balassa (1965). 3 See WTO (1998) for a survey of the early literature in this area. 4 See Stern and Hoekman (1987), Francois (1990c) and Francois and Hoekman (2010) for further discussion. In practice market access negotiations in services involve a mix of rules on FDI, movement of persons, and cross-border commerce, and thus analysis and models must consider these various modes of supply. 2 sion of how changes in these barriers, however measured, have been implemented in the CGE literature. Until the Uruguay Round, and reflecting the focus of the policy community, the CGE literature mostly focused on goods trade, with services treated as an explicit non-tradable sector. As models have been modiï¬?ed since the mid-1990s to reflect services trade as well as goods trade, three challenges have emerged. The ï¬?rst is identiï¬?cation, in an analytical sense, of how trade in services takes place, and how market access is therefore affected by policy. A second challenge has been to ï¬?nd data on services trade and FDI that are sufficiently robust for modeling purposes. A third, linked to the data problem, has been quantifying the barriers to be examined. We provide an overview of each of these in turn, before turning to the applied policy literature itself. In Section 2 we cover data issues, including challenges linked to quantifying basic trade and investment rela- tionships. In Section 3 we focus on how services ï¬?t into economy wide patterns of production and trade, and highlight features we believe deserve serious attention when modeling market access in a CGE framework. In Section 4 we focus on im- plementation strategies. Building on the discussion in Sections 2 and 3, we stress both policy benchmarking, and methods used to specify benchmark economic structure and policy impact in service sectors. We then turn to the applied CGE literature more generally with respect to services. Section 5 surveys the early contributions as well as recent advances in incorporating FDI and productivity effects of liberalization. In Section 6 we discuss issues raised by the heterogeneity of services sectors. In Section 7 we provide an example of how to model produc- tivity gains linked to domestic regulatory and market access reforms in Italy. We discuss problems and priorities for future research in Section 8. This is followed by closing remarks in Section 9. 2 Deï¬?nitional and Data Issues As already mentioned, in the WTO trade in services is deï¬?ned broadly to include transactions that involve the movement of factors of production. Speciï¬?cally, four so-called modes of supply are deï¬?ned in the GATS. The ï¬?rst of these, mode 1 in GATS-speak is cross-border supply, and applies when service suppliers resident in one country provide services in another country without either the supplier or the buyer/consumer moving to the physical location of the other. An example would be call center services from India to a consumer located in Canada. Mode 2, consumption abroad, refers to a consumer resident in one country moving to the location of the supplier(s) to consume a service. Examples include international tourism or a patient going to another country to obtain medical services. Mode 3, commercial presence, refers to legal persons (ï¬?rms) moving to the location of consumers on a long-term basis to sell services locally through the establishment of a foreign affiliate or branch. Note that what is measured as mode 3 trade are 3 the sales of the services by the foreign-owned entities, generally termed foreign affliate trade in services (FATS), and not the value of the associated foreign direct investment flow or stock. The fourth mode of supply, mode 4 or presence of natural persons, refers to a process through which individual service suppliers temporarily move to the country of the consumer or client to provide the service. An example would be management consulting services provided by an employee of a ï¬?rm based in the UK to a client in Russia. Although technology is making mode 1 increasingly feasible, for many services the need for provider and client or consumer to be in the same place (the proximity burden) remains so strong that delivery must be local, so that foreign ownership (establishment) or temporary movement of provider or consumer is required (modes 2, 3 and 4). The reason why trade in services is deï¬?ned to span these four modes reflects the characteristics of services. Most services are intangible and they are often difficult to store. As a result they are inherently more difficult to trade than tangible goods. Many services can be exchanged only if a provider and consumer or client are in the same place at the same time – which will imply that for trade to occur use will need to be made of modes 2, 3, and 4. An implication for modeling and analysis is that ideally account needs to be taken of all four modes, the factors that affect the costs of using alternative modes and the degree to which the modes are complements or substitutes. Services trade data are available from a number of sources. The OECD, Eurostat and the UN all provide data in some form on bilateral services trade flows (both imports and exports) by partner for up to 24 sectors and subsectors. The most comprehensive coverage of reporting countries among the three sources is the UN, which at present count provides data on roughly 190 countries. Eurostat and OECD provide data for a more limited number of reporters, though with more sector detail than the UN. Eurostat covers 27 EU members plus Croatia, Iceland, Japan, Norway, Turkey, Switzerland, and USA, while the OECD covers 28 countries (all the OECD members apart from Chile, Iceland, Israel, Slovenia, and Switzerland). Time coverage is the deepest with Eurostat, which reports data starting from 1995. IMF data cover almost as many countries as the UN, and for a longer time span, but only for trade with world.5 The construction of datasets for both econometrics and CGE work has re- quired reconciliation and merging of these data in some form. For econometrics, the main issue is reconciling multiple sources, and mirroring is often used to deepen coverage. This means that where country x reports trade with country y, this can be used as a replacement where country y does not itself report its trade with country x. Reconciled bilateral datacubes, expanded to include countries that do not report services trade, are available on this basis.6 5 See Francois and Pindyuk (2010). 6 Again see Francois and Pindyuk (2010). 4 While mirroring reduces the extent of missing data in the bilateral datacube, for CGE modeling this is not enough. This is because, to balance trade data, we need to somehow allocate missing destinations if we are to reconcile bilateral data with total values of trade with world (for example based on data from the IMF). On top of this, official statistics can be inconsistent, in that services data can vary widely depending on the reporting countries. In the case of international shipping and transport, reported data may be inconsistent with flows implied by the cif − f ob margins for goods. The most comprehensive attempt to confront these challenges in the context of balancing global and bilateral service flows has been through the GTAP project. The efforts to reconcile services data within the project are well documented, and reflect rather creative approaches, based on entropy methods, to flesh out the pattern of trade enough to allow construction of a global dataset for modeling cross-border trade in services (see MacDougall and Hagemejer, 2005; as well as Van Leeuwen and Lejour, 2006). Indeed, for the transport and shipping logistics sector, reconciliation of trade data in services is explicitly linked to the parallel reconciliation of data on trade in goods, as services play a critical role in bridging the gap between reported export flows for goods in one country’s national accounts data, and reported import flows in another country’s data (Gelhar, 1996). While data on cross-border services trade have a reputation for being crude, they are easy to work with and much more complete in comparison with data on FDI stocks and flows in services. FDI data are plagued by limited coverage, as well as by conï¬?dentiality constraints that imply that even if available they can- not legally be reported. As discussed below, FDI data compilation efforts have sometimes been linked to speciï¬?c modeling projects, like FTAP (Hanslow et al., 2000), the Michigan model (Brown and Stern, 2001) and the Worldscan model (Lejour, Rojas-Romagosa and Verweij, 2008). As an alternative, the French re- search institute CEPII has endeavored to follow the entropy and reverse gravity methods pioneered for merchandise trade flows to map out a picture of the geo- graphic pattern of FDI reï¬?ned enough for CGE modeling (Boumellassa, Gouel, and Laborde, 2007). However, such estimates reflect potential FDI (based on what we observe, we guess at what we do not know) and so do not necessarily reflect actual stocks and flows. On top of these concerns with FDI stock and flow data, to actually model establishment trade in services (mode 3), ideally we need foreign affiliate trade statistics (FATS) – data on the value of sales of services by foreign affiliates – as well as value added and employment. As things stand, data on stocks of FDI are often used as a proxy for this more comprehensive, but largely unavailable set of FATS data requirements. 5 3 Conceptual Issues What is different about trade in services compared with trade in goods?7 Services are very heterogeneous, and span a wide range of economic activities. Concep- tually, this diversity masks a fundamental economic function that many services perform: they are inputs into the transaction side of production. One dimen- sion of this ’input function’ is that services facilitate transactions through space (transport, telecommunications) or time (ï¬?nancial services) (Melvin, 1989). This transaction aspect is additional to the role of services as inputs in production. Services are frequently direct inputs into economic activities that determine the productivity of the ’fundamental’ factors of production - labor and capital - and that generate knowledge, goods and other services. Education, R&D and health services are examples of inputs into the production of human capital. In this section, we focus conceptually on channels or mechanisms by which services may be important in an economy wide context, and on the mechanics of international services trade. The following section builds on this discussion when mapping from the issues spelled out here to discuss practical implementation in numerical models. 3.1 Downstream linkages and direct cost channels For applied general equilibrium modeling, the relevance of a focus on services largely hinges on the strength of economy-wide linkages. To illustrate the strength of such linkages, in this sub-section we discuss recent evidence, focusing on how value added structures, and services linkages to trade have evolved over time. Thanks to repeated updating of the GTAP dataset, we have data for both cross border linkages in recent years, and how these have changed since the early 1990s. Table 1 provides a broad overview of services exports (excluding transport and distribution) on a gross and value added basis for Brazil, China, Korea, and the United States (Francois, Manchin, and Tomberger 2012). There are three sets of values in the table. The ï¬?rst, â€?Gross Valuesâ€? refers to exports by sector from standard balance of payments statistics. The second, â€?Direct VAâ€? refers to the direct value added contained in exports. These values reflect an adjustment to gross export values to include only the value of primary factors employed directly within a given sector in production for exports (or domestic absorption). The last set of values â€?Total VAâ€? reflect adjustments to gross export values to include both direct value added, and also value added in a particular sector exported indirectly through downstream linkages. This includes, for example, business services value added embodied in the production of steel for export. 7 What follows draws in part on Francois and Hoekman (2010) and Francois, Manchin and Tomberger (2012). 6 Table 1: Commercial services, % share of total exports (excluding trade and transport) 1992 1995 1997 2001 2004 2007 Gross Brazil 0.9 1.2 7.6 9.2 6.6 4.9 China 12.8 3.9 1.8 2.1 3.1 1.8 Korea 3.7 6.0 7.3 6.0 4.6 3.7 USA 12.9 14.8 15.1 17.8 18.4 15.9 Direct VA Brazil 2.1 2.5 15.1 18.3 13.7 10.1 China 24.5 7.6 2.7 3.3 6.5 3.9 Korea 6.5 11.0 13.5 12.3 9.6 8.0 USA 18.7 19.0 21.7 24.9 27.0 23.9 Total VA Brazil 21.0 20.2 26.2 30.8 29.8 25.8 China 19.8 14.9 8.2 9.7 15.2 13.3 Korea 17.1 21.9 25.0 26.9 24.5 22.4 USA 26.1 24.3 28.5 30.1 31.4 29.1 source: Francois, Manchin, and Tomberger (2012) To generate these values, we begin by denoting a representation of intermedi- ate and ï¬?nal demands as follows: Y = Z − AZ (1) In equation (1), the term Y denotes a ï¬?nal ï¬?nal demand vector, Z denotes a gross output vector, and A denotes a matrix of intermediate use coefficients. Equation (1) therefore deï¬?nes ï¬?nal output with respect to intermediate input requirements. With some manipulation we arrive at the Leontief inverse matrix, also known as the multiplier matrix M. Z = (I − A)−1 Y = MY (2) The multiplier matrix M measures the inputs contained in a unit of ï¬?nal output. In particular, if we assign the sector indexes i, j to the A and M matrices, then a representative element of the M matrix Mij gives the direct and indirect inputs (and thus the sector i receipts) linked to each unit (for example each dollar) of sector j receipts in the data.8 This implies real production activities measured by value of output. For our purposes, it provides a means to trace, through these income flows, the flow of gross activity and value added from intermediate to ï¬?nal goods and services, ostensibly across borders as well as sectors. Because linkages will vary by industry, each industry will be characterized by different multipliers. To focus on value added, we note ï¬?rst that in terms of gross output values Z, ˆ some share of this involves value added within each sector. We deï¬?ne B as the 8 In multiplier analysis with ï¬?xed input coefficients, these values also represent ï¬?xed unit input requirements in value terms, though of course in CGE models one can allow for these coefficients to be endogenous. 7 diagonal matrix indexed over i, j with diagonal elements equal to the value added shares of output Z. We then use M to provide a breakdown of the flow of value added across activities in the form of the matrix V. ˆ V = BM (3) Similar to the Leontief inverse matrix itself, the V matrix identiï¬?es the inputs of value added in each sector related to a unit of ï¬?nal demand. If we multiply V by ˆ the diagonal matrix Y whose non-zero elements are the vector of ï¬?nal outputs, the matrix yields a breakdown of economy-wide value added (the primary component of Gross National Product on a source basis). Similarly, if we multiply V by the ˆ diagonal matrix X whose non-zero elements are the national export vector, we can recover the value added content of exports X (both direct and indirect). ˆ G = VY (4) ˆ H = VX (5) The G matrix and the H matrix give us the set of linkages, both direct and indirect, between value added across sectors. We provide an example for Italy in Tables 2 and 3 below. These are based on GTAP data, aggregated to 10 sectors as indicated in the tables. Starting with Table 2, we can see that 41.76 percent of value added was located in various commercial or market service activities (including for example ï¬?nancial services and ICT services). This follows from reading the bottom row of column S9, which provides the value added in this sector, broken down into the various sectors where these activities fed, ultimately, into ï¬?nal demand. At the same time, much of this actually served as intermediate inputs, so that in terms of ï¬?nal demand for commercial services, this accounted for 33.2 percent of value added (the last column in Table 2). Basically, on a cost basis, almost 10 percent of value added of the economy as a whole involves commercial services used as intermediate inputs by other sectors (the difference between the column and row totals for S9 in the table). If we focus on other machinery, in terms of sales for ï¬?nal consumption (again the last column in the Table) roughly 10 percent of total value added goes into output in this sector. However, only 6.56 percent of economy wide value added (or about two-thirds of the total) involves value added in the sector itself. The rest involved inputs from other sectors (again including market or commercial services, which account for roughly 26.5 percent of total value added costs in the other machinery sector). We next turn to the export structure of Italy on a value added basis. From Table 3, market services account for 37.2 percent of Italy’s exports, in terms of the activity content of trade (last row of S9). What we mean by this is that for every e100 of value added contained in exports in 2007, e37.2 originated in the market services sector. However, if we focus on the value added contained in exports on a sector basis (for example how much value added was embodied in steel exports), we see that manufacturing is where Italy’s exports are concentrated. For example, 8 Table 2: Value Added Content of Final Demand G , Italy % share in 2007 supply sectors S1 S2 S3 S4 S5 S6 S7 S8 S9 S10 Total demand proc text chem other other util mkt pub sectors prim foods cloth petro autos mach manuf consl servs servs S1 primary 0.80 0.03 0.00 0.02 0.00 0.01 0.03 0.03 0.10 0.01 1.02 production S2 processed 0.62 2.04 0.01 0.11 0.00 0.06 0.24 0.16 1.10 0.06 4.41 foods S3 textiles 0.11 0.04 1.88 0.10 0.00 0.05 0.16 0.12 0.87 0.04 3.39 clothing S4 chemicals 0.07 0.03 0.02 1.13 0.00 0.05 0.16 0.16 0.52 0.03 2.17 petrochems S5 autos 0.03 0.02 0.02 0.09 0.65 0.15 0.43 0.09 0.64 0.03 2.16 S6 other 0.12 0.08 0.04 0.27 0.02 5.05 1.39 0.34 2.69 0.13 10.13 machinery S7 other 0.17 0.10 0.06 0.22 0.01 0.15 4.28 0.36 1.85 0.10 7.29 manufacturing S8 utilities 0.16 0.07 0.03 0.29 0.01 0.28 1.28 7.13 2.59 0.19 12.02 construction S9 market 0.46 0.51 0.12 0.73 0.04 0.49 1.17 0.92 28.34 0.41 33.21 services S10 public 0.10 0.11 0.04 0.23 0.01 0.28 0.42 0.42 3.05 19.54 24.20 services Total 2.66 3.03 2.23 3.20 0.74 6.56 9.56 9.72 41.76 20.55 100.00 source: own calculations, using GTAP8 data. motor vehicles and machinery account for 28.3 percent (last columns of rows S5 and S6). Like the results for total ï¬?nal demand in Table 2, in terms of exports, 25.5 percent of total value added costs for other machinery exports follow from inputs of market services (column S9 and column total for row S6). In terms of market access, there is a dual message here. Italy exports a substantial amount of services through the manufacturing sector. As such, the machinery sector has an interest in increased competitiveness and efficiency in the upstream services sectors. At the same time, the service sector clearly has an interest in improved market access for machinery, as this is an indirect channel for export. We return to these issues in Section 7. Returning to Table 1, note that values are influenced by net export positions. For example, the United States runs a net trade surplus in services. As such, while the share of services in total exports on a value added basis is roughly 30 percent, most of this is because of direct exports rather than linkages to (and exports contained in) manufacturing. Thus, 23.9 percent of value added in exports is in services and follows from direct services exports, while just 5.2 percent is linked to indirect exports and downstream linkages. In contrast, for both Korea and Brazil, though commercial services are a relatively low share of exports on a gross value or direct value added basis, they are clearly a major factor in the total cost structure of industry. In both cases, services account for roughly one-quarter of the activity content of exports on a value-added basis. 9 Table 3: Value Added Content of Exports H , Italy % share in 2007 supply sectors S1 S2 S3 S4 S5 S6 S7 S8 S9 S10 Total demand proc text chem other other util mkt pub sectors prim foods cloth petro autos mach manuf consl servs servs S1 primary 10.56 0.18 0.02 0.23 0.01 0.12 0.23 0.28 1.79 0.09 13.51 production S2 processed 0.91 1.94 0.01 0.08 0.00 0.03 0.16 0.12 1.03 0.06 4.35 foods S3 textiles, 0.26 0.02 1.92 0.12 0.00 0.03 0.12 0.12 0.86 0.05 3.50 clothing S4 chemicals, 1.79 0.09 0.04 5.35 0.01 0.11 0.40 0.56 2.68 0.19 11.23 petrochems S5 autos 0.15 0.03 0.03 0.20 2.92 0.29 0.67 0.18 1.82 0.13 6.41 S6 other 0.51 0.09 0.06 0.50 0.05 12.04 2.07 0.59 5.59 0.42 21.92 machinery S7 other 0.90 0.09 0.07 0.39 0.02 0.20 8.48 0.68 3.32 0.19 14.33 manufacturing S8 utilities, 0.05 0.00 0.00 0.02 0.00 0.02 0.11 0.71 0.24 0.01 1.18 construction S9 market 0.36 0.18 0.03 0.40 0.04 0.17 0.44 0.51 19.61 0.29 22.03 services S10 public 0.02 0.01 0.00 0.01 0.00 0.01 0.02 0.04 0.20 1.23 1.54 services Total 15.52 2.62 2.17 7.30 3.07 13.04 12.69 3.80 37.13 2.66 100.00 source: own calculations, using GTAP8 data. For Korea, commercial services accounted for only 3.7 percent of Korean exports on a gross value basis, and 8 percent on a direct value added basis, but for 22.4 percent of all value added contained in exports when embodied exports are included. China proves a notable example as well. While China has opened up and become a dominant exporter of manufactured goods (especially machinery), the contribution of services to total export value added has actually decreased. Indeed, there is an interesting dynamic to China’s service intensity. As China experienced a dramatic increase in exports of manufactured goods, the initial stages used for production saw a drop in total domestic value added in exports, with a rising share of manufactured goods. However, as China has progressed to higher stages of processing, with machinery replacing textiles and clothing as a major export category, there has been a recovery in the total service intensity of China’s exports, from a low of 8.2 percent in 1997 to 13.3 percent in 2007.9 In this section, we have highlighted the importance of services as inputs to manufacturing, and how this relates to the activity content of exports. In terms 9 China has seen a dramatic drop in share terms for services, but this is not something linked to the level of services exports per se, but rather to the very dramatic increase in manufacturing exports since 1992. In 1992, almost half of exports were textiles and clothing, while services were a large share of a small base. Services exports have grown little since – China’s exports are now dominated by machinery exports, on both a value added and gross value basis. See Francois, Manchin and Tomberger (2012) for further discussion. 10 of the political economy of market access, service ï¬?rms have an interest in im- proved market access for downstream manufacturing sectors, while those same downstream ï¬?rms have an interest in greater competition (more open markets) for upstream service sectors. Such linkage effects will be captured in CGE mod- els, but missed in partial equilibrium models. We focus on these linkages further in our example application in Section 7. 3.2 Intermediation and productivity channels The growth of intermediation services is an important determinant of overall economic growth and development because such services support specialization. As ï¬?rm size increases and labor specializes, a greater amount of effort must be devoted to coordinating and organizing the core businesses of companies. This additional activity is partly outsourced to external service providers. The ‘pro- ducer services’ that are demanded and supplied as part of this process are not just differentiated inputs into production. They play an important distinct role in coordinating the production processes needed to generate ever more differenti- ated goods and to realize scale economies. The associated organizational innova- tions and expansion of ‘logistics’ (network) services yields productivity gains that in turn should affect economy-wide growth performance (Francois, 1988,1990a; Deardorff, 2001). A range of service sectors help facilitate increased coordination in more com- plex economies. For example, low cost and high quality telecommunications may generate economy-wide beneï¬?ts, as the communications network is a transport mechanism for information services and other products that can be digitized. Telecommunications are crucial to the dissemination and diffusion of knowledge - the spread of the Internet and the dynamism that that has lent to economies around the world is telling testimony to the importance of telecommunications services. Similarly, transport services affect the cost of shipping goods and move- ment of workers within and between countries. Business services such as account- ing, engineering, consulting and legal services reduce transaction costs associated with the operation of ï¬?nancial markets and the enforcement of contracts, and are a channel through which business-process innovations are transmitted across ï¬?rms in an industry or across industries. Retail and wholesale distribution ser- vices are a vital link between producers and consumers, with the margins that apply in the provision of such services influencing the competitiveness of ï¬?rms on both the local and international markets. Health and education services are key inputs into - and determinants of - the stock and growth of human capital (Hoekman and Mattoo, 2009). While there are serious difficulties that afflict productivity measurement for many service industries (because it is often a major challenge to deï¬?ne the real 11 output of a service sector),10 empirical analysis has found that many services sectors have registered signiï¬?cant productivity growth in recent decades. Dif- ferences in aggregate productivity levels and growth rates across countries are to a large extent driven by differences in the productivity levels/growth rates of producer and business services sectors. These differences varied over time, and by the intensity of ICT investments (Triplett and Bosworth, 2004; Inklaar et al., 2008; Harper et al., 2010). Such variations may in turn be a reflection of policy variables such as regulation, limits on entry into or scaling up of business ser- vices, investment restrictions, etc. as well as factors such as the quality of the educational system in a country. An implication is that incorporation of policies affecting services sectors is important in understanding how such policies affect productivity performance.11 In addition to facilitating gains to specialization in goods, services may them- selves beneï¬?t from increased specialization. Where this is the case, greater variety and quality of services may lead to reductions in (real) prices associated with spe- cialization in services (outsourcing), and so to changes in productivity (welfare) of ï¬?rms (households) that buy services. The productivity-enhancing role of services as inputs contrasts with a long-standing concern in the literature that a steadily expanding service sector may (must) be associated with a declining growth rate of the economy. These concerns have been driven by a presumption that limited potential for productivity improvements in services implies that over time the real costs of – and employment in – services must rise relative to other sectors, reducing the growth potential of the economy to that of the ‘stagnant’ services sectors (the famous Baumol cost-disease – see Baumol, 1967). But given that many services are inputs, an expansion of the service sector can increase growth even if there is limited scope for productivity growth in the services concerned, be- cause greater outsourcing of services entails a reallocation of factors that increase overall output and aggregate productivity. While service sector productivity maps to economy wide efficiency, it is rea- sonable to ask if such productivity in turn hinges on market access. There is an emerging body of evidence that this is indeed the case. This includes, for ex- ample, ECORYS (2012), Baron and Cingano (2011), Francois and Woerz (2008), and Schivardi and Viviano (2010). Much of this work relies on OECD regula- tion indicators, combined with a mix of input-output analysis (as in Section 3.1 above), econometrics, and CGE modeling. The OECD’s product market regula- 10 In contrast to goods it is often difficult to measure prices and thus the output of services industries. However, signiï¬?cant improvements have been made in recent years in measurement of both inputs used and output generated by service industries. See, e.g., Triplett and Bosworth, 2004. 11 See Francois and Hoekman (2010) for further discussion and references to some of the rel- evant literature, including the rapidly expanding ï¬?rm- and industry-level econometric analysis of the spillover effects of reforms affecting the performance of upstream services sectors on the productivity of downstream sectors that use these services as inputs. 12 tion (PMR) index project has produced indexes of regulatory regimes, in terms of their impact of efficiency and competition (OECD 2006). These include indexes on speciï¬?c issues like market access for foreign ï¬?rms, and regulation of pricing and entry conditions, as well as composite indexes. The literature cited above links these measures of regulation to productivity, export performance, and the sector composition of employment. A stylized representation of a basic lesson to be taken from this literature is presented in Figure 1. This plots two different OECD PMR indexes of reg- ulatory restrictions in services for several OECD countries against an index of absolute productivity in services. The two PMR indicators include an index of policy discrimination against foreign ï¬?rms, and an index of domestic regulation of prices. These indexes range from 0 to 7, where a low number implies a an open, â€?good practiceâ€? regime while a high number implies an over-regulated and relatively closed regime. On the horizontal axis, we have plotted estimates of ab- solute productivity levels for ï¬?rms in commercial services, relative to the United States.12 What can be seen in the ï¬?gure is that openness in services appears to be linked to observed patterns of productivity. Greater openness and better regu- lation (meaning lower index values) is associated with higher productivity. Given downstream linkages discussed in Section 3.1, such productivity effects should in turn matter for downstream, sectors as well. Francois and Hoekman (2010) survey a number of empirical papers along these lines, investigating the produc- tivity effects of services trade and FDI liberalization. This work provides more evidence that greater availability of services has a signiï¬?cant positive impact on productivity of the sectors that use the services.13 3.3 Delivery modes and market access channels Delivery of services often has an element of â€?jointness in productionâ€? in the sense that complementary inputs – including other services – are needed to allow ef- fective exchange of (trade in) a service to occur. While this is often also true for goods – various intermediate inputs are needed to produce a unit of ï¬?nal output – jointness in production for services goes beyond this in the sense that often the buyer or consumer is part of the production process. Think for example of management consulting or a doctor’s visit: the quality or value of the transaction will be determined in part by the information that is provided by the client or patient. These complementary inputs can be related to addressing the location or proximity requirements that must be satisï¬?ed for many service transactions to be feasible. Where there is a need for proximity in exchange, factors like dis- tance place a cost burden on certain forms of services delivery. Christen and 12 The ï¬?gure is taken from Francois, Manchin, and Norberg, 2012. Productivity calcula- tions are based on EU-KLEMS data. In Section 7 we use these data in an illustrative policy experiment. This includes regression analysis of the data in the ï¬?gure. 13 Noteworthy contributions include Arnold et al. (2011) and Fernandes and Paunov (2012). 13 OECD  Product  Market  Regula5on  Sub-­â€?Indexes  and  Produc5vity   3.00   2.50   discriminatory  procedures   (against  foreign  ï¬?rms)   domes5c  price  controls   2.00   PMR  indexes  0  to  7   1.50   1.00   0.50   0.00   50.00   70.00   90.00   110.00   130.00   150.00   Absolute  Produc5vity  US  1995=100   Figure 1: OECD regulatory indexes and market service productivity Francois (2009) have labeled this the â€?proximity burden.â€? The proximity burden is recognized implicitly in the policy community, where the cross-border and lo- cal presence (or commercial establishment) components of international service transactions are referred to as modes of supply. Indeed, an important paper by Sampson and Snape (1985) that developed a typology of four modes of exchanging services that includes the establishment aspect of services trade was influential in leading policymakers to deï¬?ne trade in services in the GATS as spanning the four modes of supply. The GATS modes translate into logical channels for modeling services trade and the impact of policy. Modes 1 and 2 will both manifest themselves as more or less standard cross-border trade. What is less obvious is that mode 3 (FDI and FATS) can also involve cross-border trade, because FDI can facilitate trade where imported services are ultimately delivered through foreign affiliates. As such, restrictions on FDI can manifest themselves, econometrically, as effective determinants of cross-border trade (Francois and Hoekman, 2010). In our view, the apparent complementarity between FDI policy and cross-border flows implies that the CGE literature should move toward more explicit linkages between affil- iate sales and cross border sales. Even in models where FDI-based delivery is not modeled explicitly (meaning the bulk of the literature), a focus on cross-border trade implies an implicit role for FDI policy, and for joint modeling of modes 1, 2, and 3 trade. As discussed below, the approach in the CGE literature has so far involved rather crude modeling of nontariff barriers to trade in services as a mix 14 of trade costs and rents, though increasingly with a better basis in ï¬?rm surveys and econometrics. 4 Implementation Issues 4.1 Quantifying policies The need to consider trade, FDI, and cross-border movement of people (suppliers, consumers) has meant that efforts in the research community to characterize and quantify barriers to market access in services (a basic requirement if we want to gauge liberalization in a CGE model) have been problematic. The initial approach (see Hoekman, 1996) was to establish inventory measures of policies, distinguishing between different modes of supply. This involves simply counting the instances in which certain types of measures and policies are observed for various services sectors and modes of supply – cross-boder trade, FDI, etc.14 As an alternative approach, there has also been a branch of research based on price- and cost comparisons by industries across countries (see Kalirajan et al., 2000; Nguyen-Hong, 2000; OECD, 2009). More recently, there is an increasing emphasis on integrating econometric estimates of trade costs from gravity models with ï¬?rm- level, survey-based information on access to and quality of services inputs (such as energy, ï¬?nance, etc.) that can be used to determine what share of observed price margins can be attributed to the effects of policy, which in turn can be used when deï¬?ning services policy reform experiments in CGE models. Such estimates are admittedly crude, but may be the best immediate way forward for the purposes of simulating the effects of liberalization. A range of gravity and price- and cost-based estimates of trade costs in ser- vices, much of which is surveyed in Francois and Hoekman (2010), concludes that the extent of barriers to trade in services may be very substantial, and potentially larger than the barriers presented by conventional trade measures such as tariffs and subsidies (see, e.g., Deardorff and Stern, 2008; Dee, Hanslow and Phamduc, 2003; Dee, 2005; Department of Foreign Affairs and Trade, 1999; Francois, 2000; Francois and Hoekman, 1999; Fontagne et al., 2010). Most often, these ï¬?ndings have been incorporated in CGE analyses in the form of tax or tariff equivalents. Information on the extent to which policies affect prices and costs is impor- tant for accurate modeling of policy reforms, including whether policies create â€?rentsâ€? as opposed to being resource-using (generating â€?wasteâ€?), and the identity (ownership) of the entities and groups to whom any rents accrue. This is a well- 14 See Hoekman and Primo-Braga (1997) for a survey of the early literature on barriers to trade in services. 15 known issue that is not speciï¬?c to services regulation,15 but given the prevalence of regulation of services markets this can have a major bearing on the magnitude of the welfare impacts of policies and policy reforms. For example, if a policy generates rents for domestic groups and liberalization results in a share of these rents accruing to foreign entrants, the result may be lower national welfare.16 Figures 2 and 3 illustrate these concepts. We focus on distinctions between rent and cost generating barriers, and between discriminatory and non-discriminatory market access barriers. In Figure 2, the left panel illustrates domestic demand (the line D) and home supply (the line HS) for services. The gap between these at a given price deï¬?nes the import demand function M that is mapped in the right panel. The foreign supply line (FS) closes the system. In the ï¬?gure, we have imposed a trade barrier that raises the price for imported services, repre- sented as a shift in the FS line to FS’. The result is higher prices. The welfare impact depends on the nature of the barrier. If it is a barrier that generates rents, then we have a consumer (or downstream industry) loss equal to area (1-8-7-2), and revenues or rents equal to area (1-8-4-6), of which terms of trade gains are (3-4-5-6). The net welfare effect is the difference between areas (1-2-3)+(8-5-7) on the one hand, and the terms of trade effects (3-4-5-6) on the other. What is critical is the incidence of the rents. If they accrue to foreigners, then the full area (1-4-6-8) is lost, and the entire exercise generates a welfare loss. Similarly, if the barriers actually raise the cost of foreign delivery, then there are no terms of trade gains, and the full rent area (1-4-6-8) is foregone. In this case, the national welfare/income effect is approximated by area (1-8-7-2), which represents a loss. In Figure 3 we instead introduce a set of regulatory policies that raise cost of operation for both domestic and foreign ï¬?rms. In this second example, we have structured the cost impacts so that we have the same volume of imports as in Figure 2, and no change in domestic supply relative to the benchmark (pre- barrier) regime. Here, the impact of the regulations is modeled as something akin to an uncollected trade tax. In this second example, there is no beneï¬?t to domestic ï¬?rms, imports fall, and there is a net welfare/income loss for the economy of area (9-10-11-12). This area exceeds the discriminatory trade barrier loss (1-8-7-2) in Figure 2. This is because there are even higher prices at home, as domestic ï¬?rms are also affected, and thus are limited in their ability to offset higher foreign supply prices. The outcome involves higher prices, overall income losses, and potentially losses (depending on the domestic cost impact) for domestic ï¬?rms. Along the lines illustrated in Figures 2 and 3, the characterization of services barriers as a mix of deadweight deadweight costs and rents has been important 15 For example, it featured prominently in estimates of the net gains from the EU Single Market program for goods (see e.g., Venables and Smith, 1986), and is a matter that has been the subject of much analysis in the public choice literature. 16 This is of course not a normative argument against liberalization. Instead it points to the importance of ensuring that liberalization results in a more competitive marketplace and thus lower prices for consumers. 16 Price' Domes6c'market' Import'market' HS' FS’' 1' 8' FS' 2' 3' 5  7' 4' 3' 6  ''''''' M' D' HS1' HS0' M1' M0' Home'demand'and'domes6c'sales' imports' Figure 2: Discriminatory Market Access Barriers Price' Domes6c'market' Import'market' HS’' HS' FS’' 9' 12' FS' 10' 11' M’' M' D' HS0=HS1' M1' M0' Home'demand'and'domes6c'sales' imports' Figure 3: Non-Discriminatory Regulatory Barriers 17 in CGE-based assessments of potential multilateral liberalization and regional integration schemes.17 An implication for empirical efforts to determine the effects of policy is that it is not enough to focus on modeling price-cost margins as that does not allow one to disentangle the two effects. Better econometrically- based decomposition of the various price and cost margins linked to international service transactions is an important and sorely needed building block in guiding policy formation. In addition, recent focus has also shifted to the potential cost burdens imposed by regulations on domestic and foreign ï¬?rms alike. As Figure 3 illustrates, domestic ï¬?rms may beneï¬?t (or at least not be hurt) from liberalization when it actually involves lifting burdens on both home and foreign ï¬?rms. Another question in quantifying the effects of trade agreements, whether WTO or regional, is the extent to which policies on the ground correspond to commit- ments on paper. This is important if the impacts of new commitments are to be assessed. Both the OECD, with its services trade restrictions database, and the World Bank with a similar exercise, offer evidence on this question. Gootiiz and Mattoo (2009), for example, have created an index measuring the restrictiveness of actual policies in services trade suitable for comparison to the commitments made by countries in the WTO or in regional trade agreements. In the case of the GATS, their work suggests that currently applied policy regimes are much less restrictive than the regimes to which countries are formally committed. That is, the policy bindings negotiated under the GATS allow countries to substantially increase the level of applied trade restrictiveness if they desire to. An implication is not just that there is great scope to lock in the more liberal applied policy regimes, but that modeling the changes in commitments made by countries as a result of a WTO negotiation will overstate the extent of the liberalization achieved in the negotiation. In such cases, new commitments may limit room for policy maneuver, but they imply little change in actual policy. 4.2 Approaches to modeling changes in policy There have been two basic approaches followed in CGE modeling of changes in policies, once prevailing polices have been somehow benchmarked. The ï¬?rst approach involves a focus on cross-border delivery of services. This includes not only modes 1 and 2 under the GATS, but also mode 3, to the extent this involves cross border sales through affiliates. For example, from the U.S. data we know that exports through affiliates are a prominent aspect of cross-border trade. Consistent with this observation, there is also evidence that FDI restrictions in services do lead to less cross border trade as well.18 A common approach in CGE modeling of barriers to cross-border trade in services involves using estimates 17 See e.g. Dee and Hanslow (2001), Hoekman and Konan (2000, 2001) for early CGE ap- plications analyzing the importance of whether services policies generate waste (red tape) or rents. 18 See for example Table 7 in Francois and Hoekman (2010). 18 of dead-weight costs generated by a gravity model regression (or alternatively estimated from price comparisons).19 However, while recent practice has focused on pure trade costs, emerging evidence (e.g. ECORYS, 2009) suggests that the best practice most likely lies in modeling a combination of dead weight costs and rents.20 A problem with much of the early literature is the absence of any empirical estimates of the extent to which policies generated rents and which groups bene- ï¬?tted from such rents. Instead, authors tended to provide a range of results based on different assumptions regarding the extent to which estimated price or cost wedges reflected rents as opposed to waste. In a prospective analysis of services liberalization in Tunisia, for example, Konan and Maskus (2006) assumed that barriers were half rent-creating and half-resource wasting in the absence of em- pirical evidence on this matter. Subsequent work by Dee and Diop (2010) based on better empirical estimates of the welfare effects of barriers concludes that the impacts of services trade reforms will be smaller than those estimated by Konan and Maskus – although in part this also reflects a different baseline and changes in the Tunisian economy during the 2000s. It is obviously important that models incorporate empirical estimates of the magnitude of any rents. There has been extensive recent effort to quantify the effects of nontariff measures (NTMs) af- fecting trade in services and to break down the extent to which they generate rents or costs in the context of regional integration negotiations.21 ECORYS (2009) estimates, based on ï¬?rm surveys, that there is an approximate 60:40 split in the extent to which barriers increase the cost of doing business for all ï¬?rms, and the extent to which generate rents for domestic ï¬?rms. They also provide a CGE-based assessment of reduction in nontariff measures affecting trade between the US and the EU, which suggests that service barriers are at least as important as barriers to trade in goods. A second approach taken in CGE modeling of the effects of changes in policies focuses on both direct trade costs and the ancillary effects of policies affecting FDI. Reflecting our discussion of productivity channels in Section 3, this litera- 19 In multi-region CGE models, modeling services barriers as iceberg or dead-weight trade costs was pioneered with support from the EC to study services trade under the the Millennium Round –now known as the Doha Round (Francois 1999, 2000). This approach has grown from an extension in Francois (1999) to a now standard feature of the GTAP model, following Hertel, Walmsley and Itakura (2001). It has featured in the joint Canadan government and EC (2008) joint study on a EU-Canada Trade Agreement, as well as the ECORYS (2009) study on EU-US non-tariff barriers. 20 This issue ï¬?gured prominently in work by Denise Konan – see e.g., Hoekman and Konan, 2001; Konan and Maskus, 2006; Konan and Kim, 2005; and Konan and Van Assche, 2007. 21 For example the EC has sponsored studies on NTMs in services in North America and Japan. These include ECORYS (2009) and Sunesen et al. (2009), as well as several earlier path-breaking Australian studies, focused on the Asia Paciï¬?c region. Several of the latter are included in Findlay and Warren (2000) and are surveyed in Dee (2005) and Deardorff and Stern (2008). 19 ture places emphasis on direct linkages between productivity and liberalization. A key feature of papers embodying this approach is to allow for imperfect compe- tition and to consider the role of FDI in increasing competition and productivity. This includes for example changes in domestic transport margins, and general productivity and pricing changes in margin sectors like transport and ï¬?nancial intermediation. Examples include Rutherford, Tarr and Shepotylo (2005; 2008), and the FTAP model (Dee and Hanslow, 2001). This approach ï¬?ts well with recent evidence that greater cross border competition in service sectors leads to increased competitive pressure and lower price-cost margins in services.22 A starting point in much of the FDI modeling is Petri’s (1997) assumption that some share of capital is sector and country speciï¬?c (essentially a variation on the Armington assumption), and that FDI allows ï¬?rms in country r to employ some of their country/sector capital in country s. This can follow, for example, from affiliates serving as bridgeheads for import of home country r primary fac- tors, embodied in imports of intermediates, that are then combined with local primary factors and other inputs to provide services locally in destination market o s. Indeed, this is consistent with recent evidence (Fillat-Castej´n et al., 2008; Christen and Francois, 2009) that establishment sales are linked to both FDI levels and cross border sales through affiliates. Whether models emphasize cross-border trade with implicit sales through affiliates, or a mix of direct cross-border sales and explicit modeling of affiliates, it is important to accurately model the extent to which policies raise costs for all ï¬?rms competing in a market as opposed to creating rents for a subset of (local) ï¬?rms, and to accurately model where rents, if any, accrue. Rents under imperfect competition can be highly relevant. Theoretically, it is possible that liberalization of a highly concentrated market leads primarily to a shift of domestic rents to foreign providers, with a resulting loss in domestic welfare (Francois and Wooton, 2001b). An extreme example would be a situation where liberalization entails a shift from a domestic monopoly to foreign monopoly. Less extreme would be a shift from a purely domestic monopoly or oligopoly duopoly to a situation where the market remains highly concentrated but with ownership now shared between domestic and foreign ï¬?rms. Such outcomes can in fact be the objective of a speciï¬?c policy reform if this is conceived and implemented as part of a bilateral or regional preferential trade agreement.23 Of course, this is not a services-speciï¬?c issue, but 22 See for example Molnar and Bottini (2006). Working with ï¬?rm-level data for the OECD, they ï¬?nd large variations over industries and across countries in estimated mark-ups that are associated with differences in the restrictiveness of policies in these sectors. See also Inklaar, Timmer and van Ark, 2008. 23 Jensen and Tarr (2011) and Balistreri and Tarr (2011) ï¬?nd exactly this type of result in their modeling of liberalization in Armenia and Kenya, respectively, where the potential gains for Armenia from preferential liberalization with the EU proves lower if Armenians lose rents from barriers against foreign service-providers that are then shifted abroad. See also Konan and van Assche (2007) for an analysis of telecommunications in Tunisia. 20 is more one that highlights the general importance of ensuring that markets are contestable and entry barriers are removed – which is much more likely to be the case if reforms are implemented on a nondiscriminatory basis. 5 Models of Services Liberalization 5.1 Pioneering work Early CGE models of multilateral liberalization focused on trade in goods, mo- tivated by the Tokyo Round of multilateral trade negotiations (Deardorff and Stern, 1981, 1982; Whalley, 1982). These pioneering efforts treated services as nontradables. One of the ï¬?rst attempts to move beyond trade in goods and consider services was an application of the Deardorff and Stern Michigan model to assess the possible impact of the Uruguay Round as it related to services. Building on early barrier estimates by Hoekman (1996), Brown, Deardorff and Stern (1996) and Brown et al. (1996) revised the Michigan model to allow for services trade. Confronting problems now well known (and outlined above), but less appreciated at the time, extending the model required the Michigan team to estimate services trade flows from the swiss-cheese of aggregated and bilateral trade data (conceptually a data cube, but full of holes). In a procedure similar to the one followed in future extensions of the GTAP database, Brown et al. (1996) used a RAS-type routine to estimate the full bilateral trade matrix. Given Hoekman’s assessment that the Uruguay Round did not actually yield any actual liberalization, the analysis of Brown et al. is more of an estimate of what might be accomplished in future rounds. In this context, they model re- ductions in tariff equivalents of policies restricting trade in services, and conclude that future services liberalization had the potential to yield gains comparable to past accomplishments under the GATT in goods, measured in terms of GDP and trade volumes. More recent estimates suggest that the tariff equivalents grounded in earlier counts of the number of nontariff barriers (NTBs), like those from Hoek- man, overstate actual price impacts of barriers to services trade. At the same time, because the barriers were modeled as tariff-equivalents rather than as NTBs with associated dead weight costs, these early analyses probably also understated the potential welfare gains from liberalization. An important weakness of the early Michigan modeling of trade in services was that FDI was not incorporated into the analysis. In a pioneering contribution to the applied CGE literature, Petri (1997) developed a model that included FDI as well as cross-border trade in services. FDI in the Petri model gives rise to affiliates (foreign-owned plants) that differ from domestic ï¬?rms in the same sector by using inputs â€?importedâ€? from the parent company as well as domestic factors of production. By assuming that consumer demand is differentiated both by place of production (along Armington lines) and nationality of ownership of 21 plants it becomes possible to model the effects of policies that increase the costs of foreign ï¬?rms that are established in a given market (have engaged in FDI). Petri recognized and allowed for the fact that in practice such costs may be discriminatory in nature or affect all ï¬?rms in a sector, and may affect either entry or ongoing operations. 5.2 Post-1990s CGE modeling of services liberalization Subsequent to the early analyses motivated by the Uruguay Round, the CGE literature on multi-country services liberalization has continued to focus on both on the potential of further (future) WTO-based liberalization of trade, and on re- gional trade agreements. For example, Francois (1999,2000), Francois, van Meijl, and van Tongeren (2005) and Kinnman and Lodefalk (2007) model services trade as part of a comprehensive liberalization package under auspices of the WTO. They focus on the multi-country pattern of impacts of improvements in market access on a most-favored-nation basis, with the emphasis on cross-border trade (mode 1). Other multi-country CGE models that incorporated estimates of ser- vices trade barriers and simulated the impacts of liberalization include Hertel (2000), Dee and Hanslow (2001), Brown, Deardoff and Stern (2001), Robinson, Martin and Wang (2002), Decreux and Fontagne (2006), and Francois and Wig- naraja (2008). While the earlier work with the Michigan model included monop- olistic competition, Francois (1999, 2000) explored alternative market structures including both Cournot and monopolistic competition settings. Alternatively, these lead to the primary margin of impact being along either the markup or variety-based productivity channel.24 Major improvements in modelling have centered on incorporating FDI into global models, better representation of dead weight aspects of policy barriers, and incorporating linkages between market structure, prices, and productivity. Building on the initial Petri (1997) paper, working within the ORANI and GTAP family of models,25 Hanslow et al. (2000) and Dee and Hanslow (2001) integrated FDI into the FTAP model. As a result of the FTAP initiative the Australian Productivity Commission pioneered the extension of a standard, static modeling framework to include bilateral FDI in services. Working with the model, Dee and Hanslow (2001) reported estimates that full liberalization of services would yield greater gains than liberalization of remaining barriers to goods trade, driven in 24 The earlier literature calibrated elasticities from estimated markups, while more recent monopolistic competition speciï¬?cations for services are calibrated so that markups ï¬?t assumed trade substitution elasticities. The size of these elasticities is itself a problem. If one is willing to assume that the marginal impact of variation in delivered costs is proportional to the marginal variation in NTB measures themselves (i.e. roughly log-linear with respect to NTB indexes) then gravity-based NTB elasticities can be used for this purpose, as in the case of the NTB elasticities applied in Francois and Hoekman (2010). 25 See Dixon et al. 1982 and Hertel, 1997, respectively. 22 large part by greater flows of FDI from industrialized, high-income countries to developing nations. A key difference between the Petri (1997) and the Dee and Hanslow (2001) ap- proach is that Petri modelled customers as ï¬?rst choosing among ï¬?rms (suppliers) on the basis of their ownership, and then among the locations that the ï¬?rms were based in. This treatment assumed that in any given sector, Japanese-owned ï¬?rms were closer substitutes for each other than for Australian ï¬?rms in the relevant sector, irrespective of location. By contrast, Dee and Hanslow (2001) assumed that customers choose between phyisical location ï¬?rst (where the services are produced) and then select between suppliers in a given geographic location on the basis of ownership (nationality). As noted by Dee (2003), from an Australian perspective, a US multinational located in Australia is assumed to be a closer substitute for an Australian-owned ï¬?rm than it is for a US ï¬?rm located in the United States. Others have followed the example of the Dee and Hanslow/FTAP effort, in- cluding the Central Planning Bureau in the Hague, which extended its WorldScan model to include FDI in services (Lejour, Rojas-Romagosa and Verweij, 2008), and Brown and Stern (2001) and Brown, Deardorff and Stern (2001), who ex- tended the Michigan model to include FDI along Petri lines. A major limiting factor here is the lack of detailed bilateral FDI data by sector. In addition, an is- sue rarely discussed is the use of FDI data as a proxy for actual (unobserved) sales of services by foreign affiliates when representing the operations of multinational enterprises. While the inclusion of FDI further increases the data challenges confronting modelers (Stone et al., 1999; Hanslow et al., 1999), extending CGE models to include FDI is critical as it is a major channel for the international exchange of services. As discussed subsequently, the extension of CGE models to include FDI has been shown to be very important from a welfare (real income) and growth (productivity) perspective. The results of the simulations of liberalization that emerge from the various CGE models all suggest that there are signiï¬?cant po- tential welfare gains from further liberalization, with both developed and devel- oping countries having an export interest in services liberalization, as well as potentially beneï¬?tting from greater import competition (Francois, van Meijl, and van Tongeren, 2005).More recent work that is anchored on the Markusen (1997) knowledge-capital model of FDI suggests that a major potential source of gain from liberalization comes from reducing barriers to entry into services and the in- crease in the variety of services that is associated with FDI inflows and subsequent â€?downstreamâ€? productivity impacts.26 26 The knowledge-capital model postulates that ï¬?rms can separate the knowledge- (skill-) intensive dimensions of their activities (so-called headquarter services) from production of out- put, which may be dispersed across different locations as a function of differences in factor endowments and minimization of trade and transactions costs. 23 A stream of the literature has centered on country-focused assessments of the impact of services liberalization. A key feature of the models that are used is to allow for Dixit-Stiglitz endogenous productivity effects generated by FDI inflows following a reduction in (liberalization of) policies that restrict entry into these activities – which by their nature often require that foreign providers establish a physical presence in the market (i.e., require FDI). An early example is Konan and Maskus (2006), where a Dixit-Stiglitz structure with FDI leads to productivity gains (and losses) linked to variety (similar to the trade-only speciï¬?cation of Francois, 2000 and Francois et al., 2005). They focus on Tunisia, concluding that the most important component of potential welfare gains from liberalization (which is assumed to be implemented on a non-discriminatory basis by Tunisia) is the removal of barriers against FDI in services sectors. Konan and Maskus argue that increasing international competition on service markets will reduce the ‘cartel effect’ – the markup of price over marginal cost that incumbents are able to charge due to restricted entry, and attenuate what they term the ‘cost inefficiency effect’ – the fact that in an environment with limited competition marginal costs of incumbents are likely to be higher than if entry were allowed. The latter is most important as inefficiency imposes a cost on all sectors and households that consume the services involved. They conclude that removing policies that increase costs can have much greater positive effects on national welfare than the removal of merchandise trade barriers. Instead of the ’standard’ 0.5 to 1 percent increase in real income from goods liberalization, introducing greater competition on services markets that removes cost inefficiencies raises the gains to 6-8 percent. These large potential effects of services liberalization reflect both the importance of services in the economy and the extent to which they tend to be protected through policies that create barriers to entry. In a series of papers, Jensen, Rutherford and Tarr (2006; 2007; 2010); Ruther- ford and Tarr (2008; 2010); Jensen and Tarr (2008); and Balistreri, Rutherford and Tarr (2010) model both direct trade and Petri-type FDI under monopolis- tic competition, combining both approaches found in the literature. Extending the earlier work on services trade with Dixit-Stiglitz variety effects (Konan and Maskus, 2006; Francois, van Meijl, and van Tongeren, 2005; Francois, 2000), a key ï¬?nding of the various Jensen, Rutherford, Tarr, and Balistreri papers is that in an applied CGE setting with FDI, the downstream productivity growth ef- fects of greater competition and choice (variety) in services markets can be very important.27 Rutherford, Tarr and Shepotylo (2005; 2008) assess the impact on 27 Dixit-Stiglitz variety effects are a common feature of the endogenous growth literature; the new economic geography literature; the Krugman style international trade models that incorporate imperfect competition; and models that allow for heterogeneous ï¬?rms. Their im- plications for productivity effects are discussed by Francois (1998) and Francois and Martin (2010). Indeed, it is well understood that variety effects can provide a micro-founded basis for productivity effects of trade, in contrast to ad hoc exogenous productivity shocks that are sometimes assumed in older CGE models. 24 Russia of liberalization of trade in goods and services associated with accession to the WTO. Their estimates are decisively affected by liberalization of FDI in business services sectors and endogenous productivity effects in business services and goods. 5.3 Modeling regional services liberalization The recent CGE literature has also shifted focus to services when examining re- gional integration issues. This is very similar to global or country-based analyses, with the exception of work that is motivated by the European Union and Eu- ropean integration, which centers more on the impacts of differences in national regulation of services. For example, Lejour, Rojas-Romagosa and Verweij (2008) emphasize efforts within the EU to deepen integration of European services mar- kets. The EU is not (yet) a customs union for services (Langhammer, 2005), and deeper integration in this area is politically contested. Lejour et al. esti- mate that while there are potential gains from further integration of EU markets, magnitudes depend on the extent to which intra-EU barriers protect rents, or actually involve dead-weight costs. Hence, they support the notion that how we treat barriers to services trade is important. Political economy factors (gener- ally not found in CGE models) also matter. Kox and Lejour (2006) project that the original 2004 EU Services Directive could have increased intra-EU services trade by 30 to 60 percent and direct investment in services by 18 percent to 36 percent. The revised directive that was adopted in 2006 is unlikely to have such effects given that key aspects of the initial proposal were removed, in particu- lar acceptance of home country regulation. The EU experience illustrates the difficulty for (unwillingness of) polities to converge on common norms and to allow for regulatory arbitrage even in situations where in principle all are agreed that common minimum standards exist. Research on the potential gains from further integration of EU services markets conï¬?rms the ï¬?ndings of the work by Tarr and co-authors discussed above that what matters in welfare terms is to new entry by foreign ï¬?rms (FDI). Lejour, Rojas-Romagosa, and Verweij (2008) conclude that although further integration of EU services markets rsulting from the revised directive will be beneï¬?cial, relative to initial GDP the increases will be limited (less than 1 percent of GDP) because they are unlikey to stimulate a signiï¬?cant net increase in investment as a result of the continued prevalence of country-speciï¬?c regulation (i.e., differences in regulatory regimes will continue to create country-speciï¬?c ï¬?xed costs of entry). Dee (2006) is representative of CGE analysis of shallower forms of regional integration of services markets than the EU. She focuses on the regional trade agreements that have been proliferating in Asia, and investigates whether these have tended to target regulatory restrictions that discriminate explicitly against foreigners, and if so, if these are the restrictions that matter most, in an economic sense. Focusing on banking, distribution, ports, professions, telecommunications, 25 air passenger transport, and electricity generation, she argues that because the agreements covering services are preferential, they liberalize only relatively unim- portant services trade barriers. That is, they do not deal with the policies that result in the greatest markup of prices over costs – which the empirical literature increasingly has found to be associated with nondiscriminatory domestic regu- latory policies that raise ï¬?xed costs and have the effect of restricting entry and reducing competition. Studies of the potential impacts of regional trade agreements often conclude that these can generate signiï¬?cant gains if the effort extends to liberalization of trade in services. In contrast to the literature on modeling of preferential liberalization of trade in goods, services modeling efforts devote less attention to issues of trade diversion and the impacts of discriminatory removal of trade barriers for selected (preferred) trading partners. The assumption frequently is that liberalization – even in the context of a regional agreement – will be ap- plied on a non-discriminatory basis. The extent to which this is indeed the case however is of course an empirical matter and brings us back to the distinction between resource-wasting policies and rent-creating measures, and knowing who gets the rents if there are any. Some CGE analyses of prospective preferential trade agreements model services liberalization as the removal of price wedges (tax equivalents) that accrue to the country that liberalizes. This may be inappropri- ate.28 The papers by Konan and Van Assche (2007), Balistreri and Tarr (2011) and Jensen and Tarr (2011) among others discussed above illustrate clearly that preferential (partial) liberalization can be welfare diminishing if market structures are such as to generate signiï¬?cant rents and the partial reforms result in transfer of such rents to foreign ï¬?rms. 6 Modes of Supply and Sector-speciï¬?city In an overview of the prospective quantitative literature on the potential impacts of services trade liberalization, Whalley (2004) argues that a basic problem with the CGE literature is that the heterogeneity of service activities is typically ne- glected, even though this may have important implications. Liberalization of trade in services touches on aspects of capital movement (FDI) and the (tempo- rary) movement of people, both service providers and services consumers. Indeed, temporary movement of service suppliers (mode 4 in GATS parlance) might offer (arguably) a partial solution to the dilemma of how international migration is best managed given the substantial political resistance that exists against it in 28 This is case for some of the studies that commissioned by the EU to assess ex ante the likely effects of new FTAs with Asian countries – e.g., Boumellassa, Decreux, and Fontagne (2008) and Decreux and Mitaritonna (2007). These assume that the wedges were initially fully captured by the EU by treating them as an export tax, all of which is â€?transferredâ€? to the Asian partner following the conclusion of the FTA. 26 many high-income countries. Building on Walmsley (1999) and working with a multi-region CGE model, Walmsley and Winters (2005) estimate that if OECD countries were to expand temporary access to foreign service-providers by the equivalent of 3 percent of their labor force, the global gains would be greater than those associated with full liberalization of merchandise trade. These gains are driven by higher labor productivity and thus wages of the workers that are permitted to move to locations where labor is scarcer. The gains to those who move outweigh losses incurred by those who stay behind, although these losses are partially offset by income that is expected to be remitted back to source countries (based on observed average rates of remittance transfers). Both developed and developing countries (including the citizens that are permitted to move) would share in these gains, and they would be largest if both high-skilled mobility and low-skilled mobility were permitted. There are of course large political obstacles that must be overcome for such mode 4 trade expansion to be feasible, but move- ment towards liberalization may be possible if designed appropriately. While this is an area where the GATS could play a role, it is more likely that countries will continue to rely on bilateral arrangements to manage such trade, which in turn may be sector-speciï¬?c.29 Rutten and Reed (2009) focus on the impact of health worker movement on the cost structure of the UK health system. They ï¬?nd that while a rise in the National Health Service (NHS) budget yields overall welfare gains, a nominally equivalent policy of enhancing access to the UK for foreign health care providers (yielding the same increase in services) yields even higher overall welfare gains. Cross-border health worker movement, whether temporary of permanent, is likely to become increasingly important given demographic trends in the OECD and the willingness of middle income countries to train such workers (as in the Philippines) for export. Trade in health services can also occur through mode 2 type transactions inso- far as patients move to the location of providers for treatment. While other modes will surely also play a role - e.g., providers may want to establish a commercial presence (engage in FDI) or send health providers abroad on a temporary basis - there is great potential for expanding mode 2 trade. A barrier to such trade is the lack of portability of health insurance in many countries. For example, US fed- eral or state government reimbursement of medical expenses is limited to certiï¬?ed facilities in the United States or in a speciï¬?c U.S. state. This constraint is also signiï¬?cant because it deters elderly persons from retiring abroad. Those who do retire abroad are often forced to return home to obtain affordable medical care. The potential impact of permitting portability could be substantial. Mattoo and Rathindran (2006) ï¬?nd that extending health insurance coverage to overseas care for just ï¬?fteen types of tradable treatments could produce savings for the United 29 See, e.g., Pritchett (2006). See Mattoo and Carzaniga (2003) and Hoekman and Ozden (2010) for discussions of mechanisms that could facilitate agreements to liberalize mode 4 trade. 27 States of over $1 billion a year even if only one in ten American patients travel abroad. The lower costs of health services abroad offer the opportunity to extend medical beneï¬?ts to people who currently are not insured. This is a mode of trade that has tended to be ignored in CGE analyses. These examples illustrate that CGE modeling needs to consider the interac- tions and complementarities between modes of supplying and obtaining services. Each service sector will have distinct features that affect supply and demand, including regulatory requirements. Thus analysis must take into account the sector-speciï¬?c nature of policies that affect trade by impacting on the incentives (ability) to utilize the various modes of supply that are relevant for a particular sector. To date the CGE literature has not focused on such issues. This is one area where future research could have a high payoff, because in the absence of an accurate characterization of the relationships between modes of supply simu- lations, the effects of policy reforms may be far off the mark. 7 An Example The literature reviewed above covers a broad range of issues. In this section, we focus on an illustrative example. This builds on our earlier discussion of value added linkages between the services sectors and the rest of the economy. We focus on the case of Italy. Table 4 below provides a breakdown of the value added structure of the Italian economy. In Table 4, column A corresponds to values in Table 2, while column C and D values correspond to row and column totals in Table 3.30 Clearly Italy’s economy is highly service intensive. On a value added basis, 72 percent of activities are in the service sector (column A). This is also reflected in the structure of exports. On a gross value basis, 82 percent of exports in terms of gross value are in manufacturing (column B). However, when we focus on the activity content of exports on a sector basis (column C in Table 4), this drops to 53 percent for manufacturing, while the services share of total value added in exports rises from 17 percent to 43 percdent. This reflects forward linkages, by which we mean that while most Italian exports are in goods, the production of those goods requires a great deal of service inputs. As such, in terms of the activities where value added activities take place when producing for export, much of this is in services. Finally, the last column (column D) provides a breakdown of which export sectors actually contain domestic value added. Here, we are looking at the total value added content of exports, broken down by sector, e.g., the share of Italian value added contained in total exports that is embodied in exports of automobiles. Like the gross value basis in column B, manufacturing is the platform for most exports on a value added basis (column D). Basically, 30 An Excel based version of the calculations for Tables 2, 3, and 4 is available on request, as well as GEMPACK code for replicating value added decompositions for countries in the GTAP database, and a RunGTAP version of the application reported in this section. 28 Italy exports a great deal of intermediate services embodied in goods (machinery, vehicles, chemicals and petrochemicals, etc), as reflected in the earlier discussion on Table 3. Table 4: Italy: value added composition of the economy, 2007 value value added value added added gross exports in exports: in exports: total share of forward backward economy total value linkages linkages A B C D primary sectors 0.03 0.01 0.04 0.02 processed foods, beverages 0.03 0.05 0.03 0.05 textiles and clothing 0.02 0.07 0.04 0.07 chemicals, petrochemicals 0.03 0.14 0.08 0.11 motor vehicles 0.01 0.07 0.02 0.06 other machinery 0.07 0.28 0.15 0.28 other manufacturing 0.10 0.21 0.20 0.22 utilities, construction 0.10 0.01 0.05 0.01 commercial services 0.42 0.15 0.36 0.18 public services 0.21 0.01 0.02 0.01 total 1.00 1.00 1.00 1.00 source: own calculations, using GTAP8 data. While Italy is similar to other high income economies, in terms of service intensity, it is an outlier in terms of service sector productivity. We make this claim on the basis of estimates of absolute and relative productivity levels, using data from the EU-KLEMS database and following the methodology developed by Inklaar,Timmer and van Ark (2008).31 On this basis, US multi-factor productiv- ity (MFP) in market services (i.e. non-public) in 2007 was 145% of 1995 levels. At the same time, in 2007 Italy’s productivity was 84.8 percent of the US level for 1995. Italy, along with a number of other Eastern and Southern European states (and Japan) lag relatively far behind Northern Europe and North America in terms of service sector productivity. As a follow-up to our discussed in Section 3 regarding emerging evidence that the cross-country pattern of productivity in services is linked to regulatory regimes, in Table 5 we have organized the data from Figure 1, related to produc- tivity and regulation indexes, transformed for regression analysis. As such, all 31 This involves ï¬?rst constructing detailed output (value-added) and input (labour and capital) price indexes, and then making productivity comparisons on the basis of Tornqvist indexes. The resulting data provide cross-country benchmarking of absolute productivity for selected years, implicitly linked by relative productivity growth rates for the intermediate years. See Francois, Manchin, and Norberg (2012) for further discussion. The data used as the basis for our calculations are the EU KLEMS database (November 2009 release) and the GGDC Productivity Level database (Inklaar et al. 2008). 29 data are in logs, and for the PMR indexes we have re-scaled them to range from 1 to 8. The resulting regression is summarized in Table 6.32 Table 5: Productivity and Regulation in Services, 2007 productivity PMR PMR US 1995=100 market access price controls ln(A) ln(1+index) ln(1+index) Australia 4.741234 0.000000 0.565314 Austria 4.497638 1.178655 0.336472 Belgium 4.882857 0.000000 0.500775 Czech Republic 4.198892 1.299283 0.336472 Denmark 4.904586 0.405465 0.095310 Finland 4.680383 0.000000 0.182322 France 4.839986 0.405465 0.182322 Germany 4.873890 0.000000 0.765468 Hungary 4.299110 0.223144 0.587787 Italy 4.440765 0.510826 0.788457 Japan 4.250411 1.178655 0.875469 Slovenia 4.173714 0.864997 0.988611 Sweden 4.970974 0.510826 0.470004 United Kingdom 4.785076 0.000000 0.182322 United States 4.979338 0.000000 0.405465 basis: OECD and EU-KLEMS. Table 6: Regression analysis – MFP and regulation robust OLS regression discrimination against foreign ï¬?rms -0.3796*** -.3975 *** domestic price regulation -.3822 ** -.3735 * intercept 4.9861*** 4.9952*** summary stats Rsq: 0.61 n:15 n:15 F(2,12)=18.93 F(2,12)=7.99 pr>F=.0002 pr>F=.0006 note: Regressions are based on 2007 MFP indexes. Indexes are rescaled OECD PMR sub-indexes. All variables are in logs. Signiï¬?cance levels: .01=***, .05=**, .1=*. We assume that the coefficients capture, in a reduced form, the impacts of variations in openness and domestic regulation on MFP in the service sectors. On the basis of the coefficients above, we then construct an experiment wherein we move the Italian regulatory regime toward the North European norm, as proxied by the regulatory index values for the average for Northern European Members of 32 The correlation of MFP against various OECD PMR sub-indexes is stronger for some than others. The correlation between MFP and openness (discriminatory procedures against foreign ï¬?rms) is 0.6799. The correlation between MFP and domestic regulation (here represented by the price regulation index) is 0.4989. 30 the EU. From the estimated coefficients, and the indexes themselves, this implies a combined or composite productivity gain in market services in Italy of 66.3 percent. Of this, most follows from domestic regulation (38.0 percent), with the rest following from removing discriminatory barriers, that is, greater openness (20.5 percent). Although closed by West European standards, Italy is actually relatively close to the OECD norm in terms of openness (Austria, Japan, Slovenia, and the Czech Republic all rank more closed). It is an outlier in terms of domestic regulation, being near the upper range for overall domestic regulation issues.33 The CGE experiment involves an application of the basic GTAP model, using GTAP8 data (benchmarked to 2007, the same year as our regulatory and MFP data). The basic aggregation scheme is summarized in Table 7. Table 7: Database aggregation sectors regions primary sectors Italy processed foods, beverages rest of European Union textiles and clothing rest of world chemicals, petrochemicals motor vehicles other machinery other manufacturing utilities, construction commercial services public services In the basic GTAP family of models, changes to MFP can be modeled through efficiency coefficients (Hertel, Ianchovichina, and McDonald, 1997). In particular, national production is modeled as follows. Xi,r = aoi,r [γV A,i,r V Ai,r Ï?i + γIN,i,r INi,r Ï?i ]1/Ï?i (6) where σi = 1/(1 − Ï?i ) V Ai,r = avai,r CES (Φi,r : σi,V A ) (7) INi,r = CES (Qj,i,r : σi,IN ) (8) In equations (6), (7), and (8), the terms X, V A, and IN refer to domestic goods/services, composite value added, and composite intermediate inputs, in- dexed for sector i and production region r. The elasticity of substitution at the upper-nest, the value added nest, and the primary input nests are represented by σi , σi,V A , and σi,IN . Primary factors are represented by Φ. Qj,i,r represents the set of intermediate inputs from sector j used by sector i in region r. Both value added and intermediate inputs are modeled as inputs to a ï¬?nal CES-based pro- duction function (6). The full set of production equations for the system is CES, 33 From Table 5, we are estimating the impact of moving Italy from a log market access index of 0.510826 to 0.312551, and from a log price control index of 0.788457 to 0.339374. 31 characterized for the value added nest, for example, by the substitution elasticity σi,V A . The terms avai,r and aoi,r are productivity parameters applied to the value added and ï¬?nal production nest for the block of equations (6), (7), and (8). For the experiment here we will use our estimates of MFP gaps linked to regulation – total 66.3 percent, domestic regulation 38.0 percent, and openness 20.5 percent – and apply them through the coefficient avai,r . The basic macroeconomic impact of our productivity experiment is summarized in Table 8; the effects on sectoral value added shares is reported in Table 9. Table 8: Macroeconomic Impact of Service Sector Reform openness domestic total to foreign regulatory impact establishment environment GDP, % change 21.86 5.00 16.87 equivalent variation 23.58 5.48 18.09 consumer prices -10.02 -2.45 -7.58 producer input prices for commercial services -21.27 -5.26 -16.01 source: Own estimates from CGE model and database. Table 9: Comparison of Value Added Shares Italy 2007 Italy EU26 pre-experiment post-experiment primary 0.03 0.03 0.03 manufacturing 0.22 0.25 0.27 services 0.76 0.72 0.70 source: Own estimates from CGE model and database. Starting from the value added composition of the Italian economy as summa- rized in Table 4, the large impact on GDP should not be surprising. The market (commercial) services sector accounts for 42 percent of Italian value added. As such, a 66.4 percent productivity shock implies a 27.8 percent gain in value added, which is close to the GDP and equivalent variation changes in the CGE model – between 22 and 24 percent. Even here, the gains are not quite 27.8 percent, reflecting substitution effects in the model (adjustments in input mixes, and trade substitution effects) as well as underlying distortions not reflected in the back of the envelope estimate. As we move away from simple microeconomic compar- isons, the more detailed estimates from the CGE model provide greater insight. For example, consumer prices drop substantially with regulatory reform in ser- vices (Table 8). They also drop with better conditions for foreign establishment. However, the greatest impact follows from domestic rules and regulations. Table 10 focuses on the resulting shift in the Italian structure of production. Here, 32 Table 10: Detailed Value Added Composition value value added value added added gross exports in exports: in exports: total share of forward backward economy total value linkages linkages A B C D primary sectors 0.03 0.01 0.03 0.01 processed foods, beverages 0.03 0.04 0.04 0.04 textiles and clothing 0.02 0.06 0.04 0.06 chemicals, petrochemicals 0.03 0.13 0.08 0.09 motor vehicles 0.01 0.07 0.02 0.05 other machinery 0.07 0.23 0.15 0.23 other manufacturing 0.10 0.18 0.20 0.18 utilities, construction 0.12 0.01 0.05 0.01 commercial services 0.36 0.28 0.37 0.31 public services 0.23 0.01 0.02 0.01 total 1.00 1.00 1.00 1.00 source: Own estimates from CGE model and database. we see the implications of productivity gains in services, combined with the tier- sector linkages summarized by Table 4. Comparing Tables 4 and 10, we see the drop in value added for commercial services drives the shift shown in Table 9. The efficiency gains in the sector free-up resources, which are then shifted to other sectors. Because of the substantial drop in commercial service input prices for manufacturing ï¬?rms (Table 8), this also means there is less value added from commercial services embodied in manufactured goods. Focusing on column D in the two tables, the end result is that commercial services, being more competitive, are exported directly rather than indirectly through goods. The goods sectors also are more price competitive, as falling input prices upstream are effectively equivalent to productivity gains downstream. The end result therefore is more direct exports of services and a shift in value added toward manufacturing as resources are freed up in the now more competitive service sector. There are of course other ways to approach changes in Italian regulatory policy and openness. For example, in models with monopolistic competition in services (e.g. Balistreri, Rutherford and Tarr 2010; Konan and Maskus, 2006; Francois, van Meijl, and van Tongeren, 2005) productivity effects can follow from changes in service-sector variety, again with downstream implications. However, this type of productivity effect is partially â€?implicit,â€? in that national accounts-based MFP measures do not capture variety effects well. They would however reflect ï¬?rm selection effects found in ï¬?rm heterogeneity models. Thus, while variety and selection effects can work as an endogenous mechanism linked to ava and/or ao in equations (7) and (6) above, the determinants are inherently different from basic MFP measures.34 34 See Francois, Manchin, and Martin (2012) for a technical discussion on modeling variety effects. Indeed, one strategy is to explicitly link the coefficient ao in equation (6) functionally 33 8 Setting Future Research Priorities Given of the prevalence of increasing returns and imperfect competition in many services industries, the welfare effects of liberalization will depend on market structure and the contestability of the service industries concerned and the na- tionality of ownership of ï¬?rms. In the services context the prevalence of imperfect competition implies that liberalizing trade in services is likely to encourage greater specialization, thus helping to realize increasing returns where these exist (Fran- cois, 1990b). Even if a country does not happen to have comparative advantage in certain services, liberalization may have a positive effect by encouraging further fragmentation of production activities, fostering exports of merchandise and/or other services. If mode 3 is the main mode of supply, the prices affected by liber- alization are internal prices, so that the associated terms of trade effects can be neutral or even positive (as a result of improving international competitiveness) (Dee and Sidorenko, 2006). One potentially important issue related to market structure is market power in trade and distribution sectors. This is closely related to the macroeconomic literature on price pass-through, which also highlights the structure of retail sec- tors in determining transmission of border prices to consumers and downstream industry. It points to potential impacts on goods trade linked to market struc- ture in both international transport services and domestic distribution services. On the side of econometric and theory-based reasons to take this seriously, Raff and Schmitt (2009) examine the potential for trade liberalization in goods to lead to increased concentration in the retail sectors, while Francois and Wooton (2010) focus on the impact of combined oligopsony/oligopoly pricing in retail and wholesale trade on the gains from trade liberalization. They suggest that, at least in the EU context, market power in these sectors may effectively offset realized liberalization in goods sectors. Bradford (2005) analyzes the impacts of distribution margins in eight OECD countries - deï¬?ned as the ratio between the value of output in producer and con- sumer prices for 124 products. The focus is on deriving an estimate of the spe- ciï¬?c distribution margins (costs from factory gate to consumer, including whole- sale/retail trade and transportation) by explicitly controlling for the impact of trade barriers on producer prices. His estimates range from a low of around 60 (i.e., 60 cents to move a dollar worth of output measured at world prices from the domestic factory gate or from the dock to consumers) for Canada, the Nether- lands, the UK and the US to a high of almost 100 for Japan. He uses a CGE model to assess the welfare impacts of inflated margins, running an experiment in which margins are reduced to the lowest level observed in the sample for each of the 124 products. Bradford’s simulations suggest that inefficiencies in distri- to changes in varieties. The approach is explained in Francois (1998) with respect to the GTAP class of models. 34 bution reduce imports and impose substantial welfare costs, the magnitude of which is similar to that caused by border trade barriers. As is stressed in other CGE papers that analyze the impact of services policies, the extent to which inflated margins reflect excess costs or are captured by domestic agents as rents is important. If it is assumed that half of the excess margin generates rents, the simulation generates an increase in real income (equivalent variation) of 1.7 percent across the sample.35 Francois and Wooton (2001a) focus on a related issue, developing a theoretical structure where trade requires transport costs supplied by a shipping sector op- erating as an oligopoly. Working with both their theoretical model and numeric examples in a simple CGE model, they link shipping prices to the combination of supply and demand elasticities. Hummels, Lugovsky, and Skiba (2009) of- fer empirical support for such a framework, ï¬?nding that shipping ï¬?rms charge much higher freight rates when demand is relatively inelastic. In general, the message from this literature is that with intermediate service ï¬?rms exercising market power on two price-cost margins (meaning they act collectively as both an oligopoly vis-a-vis ï¬?nal consumers and an oligopsony vis-a-vis producers), the gains from trade in goods hinge on the degree of competition in service sectors. As such, trade and FDI policy in services may impact directly and substantively on trade in goods. This issue is little explored in a CGE framework, in part due to data constraints.36 In principle, the GTAP database includes estimates of bilateral shipping costs. These data are admittedly crude, and do not include information on market structure, per unit and per weight charges, and the vari- ation in these charges by route. However, with new data sources emerging, this situation should improve.37 International trade in services is dominated by a small number of large ï¬?rms, and much of global trade takes place within these ï¬?rms. For example, trade within multinational companies (affiliate trade) accounted for 25.9 percent of U.S. exports of private services in 2005 and for 22 percent of U.S. imports of private services. Affiliated trade in business, professional and technical services accounted for 50.1 percent of total exports and for 69.6 percent of total imports in 2005. In the United States, the top 8 consumer lenders accounted for 75 percent of receipts, the top 4 international trade ï¬?nancing companies accounted for 70 percent of receipts, the top 8 securities ï¬?rms accounted for 50 percent of receipts, and the top four direct life insurance carriers accounted for 81 percent of receipts. Similar 35 In a parallel paper, Bradford and Gohin (2006) calculate that a 10% reduction in Japan’s ï¬?nal goods distribution margins would beneï¬?t it as much as worldwide free trade would. In a ï¬?nding similar to that of Konan and Maskus (2006), they also ï¬?nd that, compared to trade opening, reducing margins leads to smaller inter-sectoral production shifts and thus may en- gender less political opposition. 36 A related issue is restrictions on trade in cabotage services – see e.g., Arce et al. (1996). 37 For example, the OECD (2011) has recently organized a dataset on precisely these values, and indications are that this will be updated over time. 35 concentration patterns hold in Europe (Christen and Francois, 2009). As such, an oligopoly approach to trade and FDI in services is likely to offer a better mapping than the large group, monopolistic competition approach now followed in the literature. Given the importance of large ï¬?rms, there ought to be substantial payoff from more explicit modeling of services ï¬?rm behavior in CGE models. While the recent body of theory on ï¬?rms and trade focuses on goods producing ï¬?rms, the same basic theoretical framework also promises valuable insights for service ï¬?rms, once we reinterpret physical transport costs as the costs that are associated with overcoming the proximity burden (see Helpman, 2006). The ï¬?rst insight relates to â€?naturalâ€? elements of ï¬?rm costs (those that are independent of policy). Where the most efficient ï¬?rms engage in internal trade and FDI, since in heterogeneous ï¬?rm models market share primarily goes to the more efficient ï¬?rms, we should see most cross-border services trade taking place within multinational ï¬?rms rather than through unaffiliated (arms-length) sales. Going further, such ‘establishment transactions’ should be more important than unaffiliated sales for the same reasons. Another insight relates to the impact of policy. In the theoretical literature, internalization hinges on a mix of costs affecting internal and arms-length delivery, as well as multi-plant versus single plant costs and distance costs. If we map these to the different modes of supplying services, concessions made in trade agreements or unilateral changes in national policies that affect these different costs should affect the choice of modes, the relative importance of establishment sales and arms-length versus internal cross-border sales in predictable ways. This suggests potentially fruitful application of the recent theory on goods-producing multinational enterprises and internalization of transactions costs to formulate questions that map to observed patterns in the services trade and policy data. Yet another area where further research is a priority concerns explicit mod- eling of regulation that affects all ï¬?rms in a sector or industry in addition to analysis of the impacts of discriminatory trade and FDI entry barriers. As noted previously and has been stressed in the policy literature, the measures that af- fect the contestability of services markets are often a mix of discriminatory and nondiscriminatory taxes and regulation. The latter cannot be (are not) captured by tariff-equivalents-cum-estimates of price wedges. More important, a focus on discriminatory policies – while the focus of trade negotiations and thus of interest to policymakers and stakeholders – may neglect the fact that the nondiscrimina- tory regulatory policies that prevail have a greater impact on welfare. One reason is that the latter have the potential of generating greater ‘waste’ than discrimi- natory measures – which by their nature can be expected to generate rents for the domestic industry. There are a number of papers in the literature that focus on the effects of domestic regulation as well as the impacts of discriminatory policies – e.g., Kox and Lejour (2006) and Lejour, Rojas-Romagosa and Verweij (2008), as well as much of the Australia-based research covered above. Doing more to accurately 36 characterize both types of measures and the likely impacts of reform is important for work in this area to be policy-relevant.38 9 Conclusions Because of the dominance of the service sector in most economies, it is clearly appropriate that CGE modeling of market access has begun to move away from an exclusive focus on goods trade. However, this is a relatively recent development, and there is a need for more investment in theory and data before incorporation of services trade and investment can move beyond being rather crude. The credence that should be placed in the numbers generated by CGE assessments of market access in services depends very much on the validity of the modeling assumptions made and the data that are used. While the accuracy of the speciï¬?c numbers generated is certainly open to question, the conclusion that services liberalization can generate much larger welfare effects than goods liberalization is probably robust.39 Yet clearly, to be more informative, CGE analyses need to be able to draw on empirical research that estimates the effects of policies on markups and costs. This need includes investing effort in better representation of the role that ser- vices play in distribution, communications, and investment in modern economies, their role as intermediate inputs and a determinant of aggregate productivity of the economy (through downstream as well as direct productivity impacts), and the interaction between market structure, trade, FDI in services, and industry performance in goods. While the available econometric results are rough and imprecise, they suggest that service-related trade policy reforms that were imple- mented by most countries on a unilateral basis during recent decades may have been an important force for productivity growth. Evidence is also growing that the price effects of protection (trade costs) in the service sectors may be quite large compared to remaining import protection of goods sectors. Ignoring policies affecting trade and investment in services, and more generally regulatory policies 38 See the example in Section 7, as well as Balisteri, Rutherford, and Tarr (2009). In the latter, the authors work with survey data to benchmark barriers to foreign service ï¬?rms as well as regulatory barriers that affect both foreign and domestic business service providers in Kenya. They identify large potential gains from reform (up to 50% of baseline consumption in the long-run). However, this is not from removing prevailing market access restrictions per se, but rather from regulatory barriers that are applied on a non-discriminatory basis. This message is consistent with our own assessment of Italy in this paper. Domestic regulation is likely as important, if not more so, as market access conditions. 39 One of the authors of this paper recalls a presentation by Bob Stern of the Brown, Deardorff, Fox, Stern (1996) results. Stern characterized the modeling process at the time as follows. â€?With modeling, we are lifting policy rocks, looking underneath for big effects. We don’t know much about services trade yet, but given the size of the sector, the rocks are big, and so the effects underneath are likely to be as well.â€? 37 that determine the performance of domestic services sectors, may be an impor- tant reason why CGE studies assessing the effects of trade reforms may miss the mark. Notwithstanding the caveats regarding data quality and the limited speciï¬?city (accuracy) of the characterization of services trade policies and their effects on the different modes of supply, a basic message that emerges from the current literature is that liberalization of services matters, perhaps much more than trade in goods. However, much depends on how well the characteristics and economic functions of different services are captured, the accuracy of estimated or assumed impacts on costs and prices of services, whether policies create rents or simply raise costs, if there are rents, what share accrues to foreign factors, and whether policy reforms are applied on a discriminatory basis or not. These are all areas where future research should focus and is likely to have a high payoff. 38 10 References Arnold, J., Matoo, A. Javorcik, B. (2011), â€?Does Services Liberalization Ben- eï¬?t Manufacturing Firms: Evidence from the Czech Republic,â€? Journal of International Economics, 85(1),136-146. Arce, H., Francois, J., Flynn, J., Reinert, K. (1996), â€?Commercial Policy and the Domestic Carrying Tradeâ€?, Canadian Journal of Economics 29(1), 181-198. Balassa, B. (1965), â€?Trade Liberalisation and Revealed Comparative Advantage,â€? The Manchester School, 33: 99-123. Baldwin, R., Francois, J., Portes, R. (1997), â€?Costs and Beneï¬?ts of EU Enlarge- ment to the Eastâ€?, Economic Policy 12(4), 125-176. Balistreri, E., Tarr, D. (2011), â€?Services Liberalization in Preferential Trade Ar- rangements - The Case of Kenyaâ€?, World Bank Policy Research Paper No. 5552. Balistreri, E., Rutherford,T., Tarr, D. (2009), â€?Modeling services liberalization: The case of Kenya,â€? Economic Modelling 26(3), 668-679. Barone, G, and F. Cingano (2011), â€?Service Regulation and Growth: Evidence from OECD Countries,â€? The Economic Journal, 121 (September), 931-957. Baumol, W. (1967), â€?Macroeconomics of Unbalanced Growth,â€? American Eco- nomic Review 57: 415-26. Boumellassa, H., Decreux, Y. Fontagne, L. (2008), â€?Economic Impact of a Po- tential FTA Between the European Union and Asean,â€? CEPII report 2008-4. Boumellassa, H., Gouel, C., Laborde, D. (2007), â€?A multisector multicountry FDI database for GTAPâ€?, Paris: CEPII working paper. Bradford, S. (2005), â€?The Welfare Effects of Distribution Regulations in OECD Countriesâ€?, Economic Inquiry, 43(4), 795-811. Bradford, S., Gohin, A. (2006), â€?Modeling Distribution Services and Assessing Their Welfare Effects in a General Equilibrium Frameworkâ€?, Review of De- velopment Economics 10(1), 87-102. Brown, D., Stern, R. (2001), â€?Measurement and modeling of the economic ef- fects of trade and investment barriers in servicesâ€?, Review of International Economics, 9(2), 262-286. Brown, D., A. Deardorff and R.M. Stern. (1996), â€?Modelling Multilateral Trade Liberalization in Services,â€? Asia-Paciï¬?c Economic Review 2:21-34. Brown, D., Deardorff, A., Stern, R. (2001), â€?CGE Modeling and Analysis of Multilateral and Regional Negotiating Options,â€? in Stern, R. (ed.), Issues and options for United States-Japan trade policies, Ann Arbor, University of Michigan Press. Brown, D., Deardorff, A., Fox, A., Stern, R. (1996), â€?The liberalization of services trade: potential impacts in the aftermath of the Uruguay Round,â€? in W. Martin and L.A. Winters (eds.), The Uruguay Round and the Developing Countries. Cambridge: Cambridge University Press, 88-124. 39 Christen, E., Francois, J. (2009), â€?Modes of Delivery in Servicesâ€?, CEPR Discus- sion Paper No. 7912. Deardorff, A. (1985), â€?Comparative advantage and international trade in in- vestment services,â€? in R.M. Stern (Ed.), Trade and Investment in Services: Canada/US Perspectives, Toronto, Ontario Economic Council, pp.39-71. Deardorff, A. (2001), â€?International Provision of Trade Services, Trade and Frag- mentationâ€?, Review of International Economics 9(2), 233-248. Deardorff, A., Stern, R. (1981), â€?A Disaggregated Mode1 of World Production and Trade: An Estimate of the Impact of the Tokyo Round,â€? Journal of Policy Modeling, 3(2), 127-152. Deardorff, A., Stern, R. (1982), â€?Economic Effects of the Tokyo Roundâ€?, Southern Economic Journal 49(3), 605-624. Deardorff, A., Stern, R. (2008), â€?Empirical Analysis of Barriers to International Services Transactions and the Consequences of Liberalization,â€? in Mattoo, A., Stern, R., Zanini, G. (eds.), A Handbook of International Trade in Services, Oxford: Oxford University Press. Decreux, Y., C. Mitaritonna (2007), â€?Economic Impact of a Potential FTA Be- tween the European Union and India,â€?CEPII-CIREM report for the European Commission. Decreux, Y., Fontagne, L. (2006), â€?A Quantitative Assessment of the Outcome of the Doha Development Agenda,â€? CEPII Working Paper 2006-10. Dee, P. (2003), â€?Modelling the Policy Issues in Services Trade,â€? Economia Inter- nationale 94: 283-300. Dee, P. (2005), â€?A Compendium of Barriers to Trade in Servicesâ€?, Washington DC: World Bank working paper. Dee, P. (2006), â€?East Asian economic integration and its impact on future growthâ€?, The World Economy 30: 404-23. Dee, P., and N. Diop (2010), â€?The Economy-wide Effects of Further Trade Re- forms in Tunisia’s Services Sectors,â€? World Bank Policy Research Working Paper 5341. Dee, P., Hanslow, K. (2001), â€?Multilateral liberalisation of services trade,â€? in Stern, R. (Ed.), Services in the International Economy, University of Michigan Press, Ann Arbor. Dee, P., Sidorenko. A. (2006), â€?The Rise of Services Trade: Regional Initia- tives and Challenges for the WTOâ€?, in C. Findlay and H. Soesastro (eds.), Reshaping the Asia-Paciï¬?c Economic Order. London: Routledge. Dee, P., Hanslow, K., and Phamduc, T. (2003), â€?Measuring the cost of barriers to trade in servicesâ€?, in Takatoshi Ito and Anne Krueger (eds.), Trade in Services in the Asia-Paciï¬?c Region, Chicago: University of Chicago Press. Dee, P., Francois, J., Manchin, M., Nordas, H., van Tongeren, F. (2011), â€?The Impact of Trade Liberalisation on Jobs and Growth: Technical Noteâ€?, OECD Trade Policy Working Papers, No. 107, OECD Publishing. 40 Dixon, P., B.Parmenter, J. Sutton and D. Vincent (1982), ORANI: A Multisec- toral Model of the Australian Economy, Contributions to Economic Analysis 142, North-Holland Publishing Company, pp. xviii + 372. Department of Foreign Affairs and Trade (1999), Global Trade Reform: Main- taining Momentum. Australian Department of Foreign Affairs and Trade, Canberra, www.thecie.com.au. ECORYS (2009), â€?Non-Tariff Measures in EU-US Trade and Investment - An Eco- nomic Analysisâ€?, report to the European Commission Reference: OJ 2007/S 180-219493, Rotterdam. ECORYS (2012), â€?Spill-overs from Malfunctioning Services Markets and Eco- nomic Performanceâ€?â€?, report to the European Commission, Rotterdam. Escaith, H. (2008), â€?Measuring trade in value added in the new industrial econ- omy: statistical implicationsâ€?, WTO, mimeo. Fernandes, A. (2009), â€?Structure and performance of the services sector in tran- sition economies,â€? Economies in Transition, 17(3), 467-501. Fernandes, A., Paunov, C. (2012), â€?Foreign direct investment in services and manufacturing productivity: evidence for Chile,â€?Journal of Development Eco- nomics, 97(2): 305-21. o Fillat-Castej´n, C., J. Francois, and J. Woerz, (2008), â€?Cross-Border Trade and FDI in Services,â€? CEPR Discussion Paper 7074. Findlay, C., Warren, T. (eds.) (2000), Impediments to Trade in Services: Mea- surement and Policy Implications, Routledge, London and New York. Fontagne, L., Guillin, A., Mitaritonna, C. (2010), â€?Estimations of tariff equiva- lents for the services sectorsâ€?, Paris: CEPII discussion paper. Francois, J. (1988), Trade in Producer Services and the Realization of Returns Due to Specialization, University of Maryland: Reprinted by University of Michigan Press. Francois, J. (1990a), â€?Producer Services, Scale and the Division of Labor,â€? Oxford Economic Papers 42: 715-29. Francois, J. (1990b), â€?Increasing Returns Due to Specialization, Monopolistic Competition, and Trade in Producer Services,â€? Canadian Journal of Eco- nomics 23(1): 109-24. Francois, J. (1990c), â€?Trade in Nontradables: Proximity Requirements and the Pattern of Trade in Servicesâ€?, Journal of Economic Integration, 1990(a): 31- 46. Francois, J.F. (1998), â€?Scale Economies and Imperfect Competition in the GTAP Model,â€? GTAP Technical Papers 317, Center for Global Trade Analysis, De- partment of Agricultural Economics, Purdue University, 1998. Francois, J.F. (1999), â€?Economic Effects of a New WTO Agreement Under the Millennium Round,â€? report to the European Commission Trade Directorate. Francois, J. (2000), The Next WTO Round: North-South stakes in new mar- ket access negotiations, CIES Adelaide and the Tinbergen Institute, CIES: Adelaide. 41 Francois, J., Hoekman, B. (1999), â€?Market access in the service sectorsâ€?, Tin- bergen Institute, manuscript, cited in B. Hoekman (2000), â€?The next round of services negotiations: identifying priorities and options,â€? Federal Reserve Bank of St Louis Review, 82(4), 31-47. Francois, J., Hoekman, B. (2010), â€?Services Trade and Policyâ€?, Journal of Eco- nomic Literature 48(3), 642-692. Francois, J., M. Manchin, and P. Tomberger (2012), â€?Services Linkages and the Value Added Content of Trade,â€? Johannes Kepler University working paper. Francois, J., M. Manchin and H. Norberg (2012), â€?Openness in Services and Economy-wide Productivity Gaps,â€? Johannes Kepler University working pa- per. Francois, J. and W. Martin (2010), â€?Ex Ante Assessment of the Welfare Im- pacts of Trade Reforms with Numerical Models,â€? in John Gilbert (ed.) New Developments in Computable General Equilibrium Analysis for Trade Policy (Frontiers of Economics and Globalization, Volume 7), Emerald Group Pub- lishing Limited, pp.379-434. Francois, J., M. Manchin and W. Martin (2012), â€?Market Structure in General Equilibrium Models of Open Economies,â€? in P.B. Dixon and D.W. Jorgenson eds Handbook of Computable General Equilibrium Modeling Elsevier. Francois, J., van Meijl, H., van Tongeren, F. (2005), â€?The Doha Round and Developing Countriesâ€?, Economic Policy 20, 350-391. Francois, J., Pindyuk, O. (2010), â€?Consolidated Data on International Trade in Servicesâ€?, Vienna: WIIW discussion paper. Francois, J., Shiells, C. (eds.) (1994), Modelling Trade Policy: Applied General Equilibrium Assessments of North American Free Trade. Cambridge: Cam- bridge University Press. Francois, J. and G. Wignaraja (2008), â€?Economic Implications of Asian Integra- tion,â€? Global Economy Journal (BE Press), vol. 8(3), pages 1-48. Francois, J., Wooton, I. (2010), â€?Market Structure and Market Accessâ€?, The World Economy 33(7), 873-893. Francois, J., Wooton, I. (2001a), â€?Trade in International Transport Services: The Role of Competitionâ€?, Review of International Economics, 9(2), 249-261. Francois, J., Wooton, I. (2001b), â€?Market Structure, Trade Liberalization and the GATSâ€?, European Journal of Political Economy 17, 389-402. Gehlhar, M. (1996), â€?Reconciling Bilateral Trade data for Use in GTAPâ€?, GTAP Technical Paper No. 1, Purdue University. Gootiiz, B., Mattoo, A. (2009), â€?Services in Doha, What’s on the Table?â€?, Journal of World Trade 43(5), 1013-1030. Government of Canada and European Commission (2008), â€?Assessing the Costs and Beneï¬?ts of a EU-Canada Economic Partnership,â€? report published jointly by the Government of Canada and the European Commission. 42 Hanslow, K., Phamduc, T., Verikios, G. (2000), â€?The Structure of the FTAP Modelâ€?, Melbourne: Productivity Commission Research Memorandum MC- 58. Harper, M.J., B. Khandrika, R. Kinoshita, and S. Rosenthal (2010), â€?Nonmanu- facturing industry contributions to multifactor productivity, 1987-2006,â€?Monthly Labor Review, June: 16-31. Helpman, E. (2006), â€?Trade, FDI, and the Organization of Firmsâ€?, Journal of Economic Literature 44(3), 589-630. Hertel, T. (2000), â€?Potential gains from reducing trade barriers in manufacturing, services and agriculture,â€? Federal Reserve Bank of St Louis Review, 82(4). Hertel, T. (Ed.) (1997) Global Trade Analysis: Modelling and Applications, Cam- bridge University Press, Cambridge. Hertel, T.W., E. Ianchovichina, and B.J. McDonald (1997). â€?Multi-Region Gen- eral Equilibrium Modeling,â€? Chapter 9 in J.F. Francois and K.A. Reinert, eds, Applied Methods for Trade Policy Analysis: a Handbook, Cambridge Univer- sity Press: Cambridge. Hertel, T. W.; Walmsley, T. L., and Itakura, K., (2001) â€?Dynamic Effects of the ˆ Ëœ ONew AgeO Free Trade Agreement between Japan and Singaporeâ€?. Journal of Economic Integration, 16(4); 446-448. Hoekman, B. (1990), â€?Services-related Production, Employment, Trade, and For- eign Direct Investment: A Global Perspective,â€? in P. Messerlin and K. Sauvant (eds.), The Uruguay Round: Services in the World Economy. Washington D.C: The World Bank. Hoekman, B. (1996), â€?Assessing the General Agreement on Trade in Servicesâ€?, in W. Martin and L.A. Winters (eds.), The Uruguay Round and the Developing Countries. Cambridge: Cambridge University Press, 88-124. Hoekman, B., Konan, D. (2000), â€?Rents, Red Tape and Regionalism: Economic Effects of Deeper Integration,â€? in B. Hoekman and J. Zarrouk (eds.) Catch- ing Up with the Competition: Trade Opportunities and Challenges for Arab Countries. Ann Arbor: University of Michigan Press. Hoekman, B., Konan, D. (2001), â€?Deep Integration, Nondiscrimination and Euro- u Mediterranean Free Trade,â€? in J¨rgen von Hagen and Mika Widgren (eds.), Regionalism in Europe: Geometries and Strategies After 2000, Kluwer Aca- demic Press. Hoekman, B., Ozden, C. (2010), â€?The Euro-Mediterranean Partnership: Trade in Services as an Alternative to Migration?â€? Journal of Common Market Stud- ies,48(4), 835-58. Hoekman, B., Mattoo, A. (2009), â€?Services, Economic Growth and Development,â€? in J. Marchetti and M. Roy (eds.), Opening Markets for Trade in Services. Cambridge University Press. Hoekman, B., Primo-Braga, A. (1997), â€?Protection and Trade in Services: A Survey,â€? Open Economies Review, 8: 285-308. 43 Hummels, D., Lugovskyy, V., Skiba, A. (2009), â€?The Trade Reducing Effects of Market Power in International Shippingâ€?, Journal of Development Economics 89, 84-97. Inklaar, R., M. Timmer and B. van Ark (2008), â€?Market services productivity across Europe and the US,â€? Economic Policy, 23: 139-194. Jensen, J., Tarr, D. (2008), â€?Impact of local content restrictions and barriers against foreign direct investment in services: the case of Kazakhstan accession to the WTO,â€? Eastern European Economics, 46(5), 5-26. Jensen, J., Tarr, D. (2011), â€?Deep Trade Policy Options for Armenia - The Im- portance of Services, Trade Facilitation and Standards Liberalizationâ€?, World Bank Policy Research Paper No. 5662. Jensen, J., Tarr, D. (2010), â€?Regional Trade Policy Options for Tanzania - The Importance of Services Commitmentsâ€?, World Bank Policy Research Paper No. 5481. Jensen, J., Rutherford, T, Tarr, D. (2006),â€?Telecommunications Reform within Russia’s Accession to the WTO,â€? Eastern European Economics 44(1), 25-58. Jensen, J., Rutherford, T., Tarr, D. (2007), â€?The Impact of Liberalizing Barriers to Foreign Direct Investment in Services: The Case of Russian Accession to the World Trade Organization,â€? Review of Development Economics, 11(3), 482-506. Jensen, J., Rutherford, T., Tarr, D. (2010), â€?Modeling Services Liberalization: the Case of Tanzania,â€? Journal of Economic Integration, 25(4), 644-675. Kalirjan, K., McGuire, G., Nguyen-Hong, D., Schuele, M. (2000), â€?The Price Impact of Restrictions on Banking Servicesâ€?, in C. Findlay and T. Warren (eds.), Impediments to Trade in Services: Measurement and Policy Implica- tions. London: Routledge. Konan, D., Van Assche, A. (2007), â€?Regulation, market structure and service trade liberalization,â€? Economic Modelling 24: 895-923. Konan, D. and K. Kim (2004), â€?Beyond Border Barriers: The Liberalization of Services Trade in Tunisia and Egyptâ€?, The World Economy 27(9), 1429-1447. Konan, D., Maskus, K. (2006), â€?Quantifying the Impact of Services Liberalization in a Developing Countryâ€?, Journal of Development Economics 81, 142-162. Kox, H., Lejour, A. (2006), â€?The Effects of the Services Directive on Intra-EU Trade and FDIâ€?, Revue Economique, 57(4), 747-69. Langhammer, R. (2005), â€?The EU Offer of Service Trade Liberalization in the Doha Round: Evidence of a Not-Yet-Perfect Customs Unionâ€?, Journal of Common Market Studies 43(2), 311-325. Van Leeuwen, N., Lejour, A. (2006), â€?Bilateral services Trade Data and the GTAP databaseâ€?, the Hague: CPB Memorandum 160. Lejour, A., Rojas-Romagosa, H., Verweij, G. (2008), â€?Opening services markets within Europe: Modelling foreign establishments in a CGE frameworkâ€?, Eco- nomic Modelling, 25(5), 1022-1039. 44 Markusen, J. (1997), â€?Trade versus Investment Liberalizationâ€?, NBER working paper 6231. Mattoo, A. and R. Rathindran (2006), â€?Does Health Insurance Impede Trade in Health Care Services?â€?, Health Affairs 25(2), 358-368. Mattoo, A., Carzaniga, A. (eds.) (2003), Moving People to Deliver Services, Oxford University Press and World Bank, Washington, D.C. McDougall, R., Hagemejer, J. (2005), â€?Services Trade Data,â€?in B. Dimaranan and R. McDougall (eds.), Global Trade, Asistance, and Production: The GTAP 6 Data Base, Center for Global Trade Analysis, Purdue University. Melvin, J. (1989), â€?Trade in Producer Services: A Heckscher-Ohlin Approach,â€? Journal of Political Economy, 97: 1180-96. Molnar, M., Bottini, N. (2008), â€?How large are competitive pressures in ser- vices markets? Estimation of mark-ups for selected OECD countriesâ€?, paper presented at the OECD Technical Workshop on Trade Barrier Assessment Methodology, December. Nguyen-Hong, D. (2000), â€?Restrictions on Trade in Professional Servicesâ€?, Pro- ductivity Commission Staff Research Paper (Canberra: Ausinfo). OECD (2006), â€?Product Market Regulation in the Non-Manufaturing Sectors of OECD Countries,â€? Economics Department Working Paper 530. OECD: Paris. OECD (2009), â€?Methodology for Deriving the STRIâ€?. OECD:Paris. Download- able at http://www.oecd.org/trade/stri. OECD (2011), â€?Clarifying Trade Costs In Maritime Transportâ€?, Paris: OECD Trade And Agriculture Directorate, TAD/TC/WP(2008)10/FINAL. OECD:Paris. Petri, P. (1997), â€?Foreign Direct investment in a computable general equilib- rium frameworkâ€?, Paper prepared for the Conference, Making APEC work: Economic Challenges and Policy Alternatives, March 13-14, Keio University, Tokyo. Pritchett, L. (2006), â€?Let Their People Come: Breaking the Gridlock on Global Labor Mobilityâ€?, Washington DC: Center for Global Development. Raff, H., Schmitt, N. (2009), â€?Imports, Pass-Through, and the Structure of Retail Marketsâ€?, Kiel Working Paper No. 1556. Robinson, S., Wang, Z., Martin, W. (2002) â€?Capturing the implications of services trade liberalization,â€? Economic Systems Research, 14, 3-33. Rutherford, T., Tarr, D. (2008), â€?Poverty Effects of Russian WTO accession: modeling ’real’ households with endogenous productivity effects,â€? Journal of International Economics, 75(1), 131-150. Rutherford, T., Tarr, D. (2010), â€?Regional Impacts of Liberalization of Barriers against Foreign Direct Investment in Services,â€? the case of Russia’s accession to the WTO,â€? Review of International Economics, 18(1), 30-46. Rutherford, T., Tarr, D., Shepotylo, O. (2005), â€?The Impact on Russia of WTO Accession and The Doha Agenda: the importance of liberalization of barri- ers against FDI in services for growth and poverty reductionâ€?, in T. Hertel 45 and L.A. Winters (eds.), Putting Development Back into the Doha Agenda: Poverty Impacts of a WTO Agreement, New York: Palgrave McMillan. Rutten, M., Reed, G. (2009), â€?A comparative analysis of some policy options to reduce rationing in the UK’s NHS: Lessons from a general equilibrium model incorporating positive health effectsâ€?, Journal of Health Economics 28(1), 221-233. Sampson, G., Snape, R. (1985), â€?Identifying the Issues in Trade in Servicesâ€?, The World Economy, 8, 171-81. Schivardi, F. and E. Viviano (2010), â€?Entry Barriers in Retail Trade,â€? The Eco- nomic Journal, 121 (March), 145-170. Stern, R., Hoekman, B. (1987), â€?Issues and Data Needs for GATT Negotiations on Services,â€? The World Economy, 10, 39-60. Sunesen, E., Francois, J., Thelle, M. (2009), â€?Assessment of Barriers to Trade and Investment between the EU and Japanâ€?, Copenhagen Economics: Report to the European Commission ref no. TRADE/07/A2. Triplett, J.E. and B.P. Bosworth (2004), Productivity in the US Services Sector; New Sources of Economic Growth, Washington DC: Brookings Institution. Venables, A., Smith, A. (1986), â€?Trade and Industrial Policy Under Imperfect Competitionâ€?, Economic Policy 1, 622-72. Walmsley, T. (1999), â€?Incorporating International Capital Ownership into the GTAP Model: Results for Asia-Paciï¬?c Trade Liberalisation,â€? Centre of Policy Studies/IMPACT Centre Working Papers 72, Monash University. Walmsley, T., Winters, L. (2005), â€?Relaxing the Restrictions on the Temporary Movement of Natural Persons: A Simulation Analysis,â€? Journal of Economic Integration 20(4), 688-736. Whalley, J. (1982), â€?An Evaluation of the Recent Tokyo Round Trade Agree- ment Using General Equilibrium Computational Methods,â€? Journal of Policy Modeling, 4(3), 341-61. Whalley, J. (2004), â€?Assessing the Beneï¬?ts to Developing Countries of Liberal- ization in Services Trade,â€? The World Economy, 27(8), 1223-53. WTO (2009), International Trade Statistics. Geneva: WTO. WTO (1998), â€?Economic effects of services liberalization: overview of empirical studies,â€? WTO document S/C/W/26/Add. 1, May 29. 46