WPS4151 Market structure and market access * Joseph Francois Johannes Kepler Universit¨at Linz & CEPR (London) Ian Wooton University of Strathclyde (Glasgow) & CEPR (London) We examine an issue at the nexus of domestic competition policy and international trade, the interaction between goods trade and market power in domestic trade and distribution sectors. Theory suggests a set of linkages between service-sector competition and goods trade supported by econometrics involving imports of 22 OECD countries vis-´a-vis 69 exporters. Competition in services affects the volume of goods trade. Additionally, because of interaction between tariffs and competition, the market structure of the domestic service sector becomes increasingly important as tariffs are reduced. Empirically service competition apparently matters most for exporters in smaller, poorer countries. Our results also suggest that while negotiated agreements leading to cross-border services liberalization may boost goods trade as well, they may also lead to a fall in goods trade when such liberalization involves FDI leading to increased service sector concentration. Keywords: distribution sector competition, market access, services, trade liberalization, GATS JEL: L16, L8, F12, F13 World Bank Policy Research Working Paper 4151, March 2007 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 view of the World Bank, its Executive Directors, or the countries they represent. Policy Research Working Papers are available online at http://econ.worldbank.org. *Thanks are due to Henrik Horn, Peter Neary, Thierry Verdier, and participants in CEPR's European Workshop on International Trade. Address for correspondence: Joseph Francois, Johannes Kepler Universit¨at Linz, Department of Economics, Altenbergerstrae 69, A-4040 LINZ, AUSTRIA. Email: joseph.francois@jku.at. www.intereconomics.com/francois 1 Introduction In this paper we examine a set of issues at the nexus of domestic competition policy and international trade, the interaction between international goods trade and domestic market structure in trade and distribution sectors. Rules and regulations governing inter- national trade and investment in services are an increasingly important aspect of regional and multilateral trade agreements (Mattoo and Sauve 2003). International negotiations have focused on regulatory restrictions and barriers to cross-border trade and FDI, while research has emphasized quantifying barriers and exploring the role of traded services as inputs to the manufacturing sector. This includes Arnold, Mattoo, and Smarzynska (2006), Francois (1990), Ito and Krueger (2003), and Markusen (1989). We depart from this litera- ture, highlighting the impact of domestic market power in margin services on goods trade. These margin activities include domestic shipping and logistic services, of course, as well as the wholesale and retail sectors and other links in the distribution chain that carries imported goods to the industrial or household consumer. In a very real sense these services make possible any interaction between producers and exporters in one country and final consumers in another. Focusing on market structure in the domestic service sector and its impact on trade in goods, we are able to identify linkages between service sector competition and the value of negotiated market access concessions in goods. An important finding is that ignoring the structure of the domestic service sector may lead to serious overestimation of the market access benefits of actual negotiated tariff reductions. We also find that the market struc- ture of margin sectors matters more for poor and small exporters than for others. Finally, our results suggest that while negotiated agreements leading to cross-border services liber- alization may boost goods trade as well, they may also lead to a fall in goods trade when such liberalization involves FDI leading to increased service sector concentration. 1 In exploring these issues, we are highlighting an important though somewhat ignored aspect of the trading system. In the European Union, for example, internal trade in motor vehicles has been hampered by an antitrust exemption for the distribution and servicing of automobiles. (See both Flam and Nordstr¨om 1995, and Lutz 2004.) Access to the distribution system was also at the heart of a dispute between the United States and Japan involving Kodak and Fuji film (Nanto 1998). These issues also lurk behind the impact on trade of the retail distribution systems both in Switzerland and Japan, as well as the German experience with retailing cartels and the threat of foreign retail entry to established domestic players. With the elimination of trade barriers for textiles and clothing under the WTO's Agreement on Textiles and Clothing in 2005, the market power of such huge buyers as Wal-Mart may also be an important factor in the transmission of price and quantity changes across global textile and clothing markets. Finally, evidence is emerging that the benefits of non-reciprocal tariff preference schemes may be captured by high-income country importing firms, rather than the low-income country exporter firms for which the programs are intended. (See, for example, Olarreaga and Ozden 2005.1) We proceed in this paper as follows. In Section 2 we develop a basic analytical model, involving a domestic distribution sector with market power. It sources both internationally and domestically. We work with this model to examine the impact of imperfect competi- tion in services for the pattern of trade in goods. In Sections 3 and 4, we then examine the impact on gains from trade for both importers and exports. In Section 5, we work with data on competition in distribution and sales in several OECD countries, examining econometrically the issues highlighted in Sections 3 and 4. This involves modeling the interaction between import protection, competition, and the pattern of trade in the con- 1There is also a nascent literature focused on international aspects of competition policy, including Head and Ries (1997), Francois and Horn (2006), and Horn and Levinsohn (2001). However, this literature is not concerned with the interaction between market access and antitrust policy so much as with open economy aspects of merger policy and the beggar-thy-neighbor potential of antitrust. 2 text of a gravity model of trade. We find that imperfect competition in the trade and distribution sectors matters most in the context of free trade areas and customs unions, like the European Union. In the EU, we find intra-EU trade barriers linked to market structure variation in these sectors higher than the average external EU tariff. We also find that market power translates into reduced trade performance when the size and devel- opment (i.e. the bargaining power) of trading partners is unbalanced. We offer concluding comments in Section 6. 2 The Basic Model We focus on the market for imports of a good q that competes directly with a domestic industry. Our primary interest is in the domestic sale and distribution network which we assume to be less-than-perfectly competitive. It exercises market power in sourcing from both domestic and foreign suppliers, and in sales to final consumers. Imports are supplied by competitive, overseas producers. Export and domestic supply are imperfectly elastic. Consequently, due to increasing marginal cost of production, the importing country has some degree of monopoly power in trade. It subjects trade in these goods to an import tax at rate t. This creates a wedge between the cif price pcif and the landed (that is, after duties are paid) import price pm. Export supply qm is represented by the inverse supply function (1). pcif = am + bmqm (1) where pcif is the export price at the border while landed prices inclusive of tariffs are pm = pcif where (1 + t) (2) 3 Similarly, domestic supply qd is an increasing function of domestic price, as reflected in the inverse domestic supply schedule (3). pd = ad + bdqd (3) Consumer demand for the imported good is defined by the inverse demand curve (4). p = x - y(qd + qm) (4) where x and y are constants defining our demand curve. Interaction between suppliers and final consumers takes place through the services of a domestic service sector that facilitates both the movement of imported goods inland and wholesale and retail distribution, mar- keting, and any ancillary services required to sell the goods. These services are supplied by a domestic service sector ­ modeled as a Cournot oligopoly ­ at constant marginal cost.2 The total revenue of a representative firm i in the service sector is: Ri = p(qmi + qdi) (5) where qmi and qdi are the quantity of imports sold by a representative intermediary firm i. We further assume that there are n identical firms in the service market, each having a share s = 1/n of sales. It proves useful to define the index 1 + s as an index of market competitiveness that ranges from a value of 1 to 2. A value of = 1 implies perfect competition (n = ) while = 2 maps to a single firm monopolizing distribution (n = 1). In equilibrium, we may also have = 2 where the service sector acts as a monopolist through perfect collusion in a cartel. Assuming a constant marginal cost c, profits of 2The Cournot approach followed here allows us a direct way to manipulate market structure ­ through induced entry. 4 service firm i are: si = pqi - (pm + c)qmi - (pd + c)qdi (6) From the first-order conditions for profit maximization, quantities will be qd = [bmG + y (G - H)]/A (7) qm = [bdH + y (H - G)]/A (8) q = [bmG + bdH]/A (9) where A (ybd + bm (y + bd)) > 0, G x - c - ad, and H x - c - am The split between imported and domestically sourced q will depend on relative import and domestic supply conditions and the tariff rate .3 3 Markups, Tariffs, and Importer Welfare It is evident that service-sector firms have power on both sides of the market. Their profits are a function of manipulating double margins. On the input side, the price they pay for imports and domestic goods depends on the total quantity bought and the sensitivity of supply to quantity. Similarly, on the demand side, the price at which they sell to consumers is a function of total quantity brought to market. By restricting their trading, the firms are able to both drive down costs in both supply markets and drive up prices, widening 3We are working here with distributors who are willing to source both domestically, internationally, or both. While beyond the focus of this paper, it would also be interesting to explore exclusive distribution networks in the context of open economies. 5 the price-cost margin and boosting profits. The service-sector margins amount to: µd = G( - 1)/ + c (10) µm = H ( - 1)/ + c (11) Equations (10) and (11) lead directly to the following propositions. Proposition 1. The Cournot-Nash mark-up on imports for the domestic trade and distri- bution sectors is a decreasing function of the underlying import tariff. Proposition 2. The Cournot-Nash mark-up on domestic shipments for the domestic trade and distribution sectors is independent of the underlying import tariff. The mark-up over marginal cost for imports declines directly with the tariff. Any attempt on the part of the government to exercise its monopoly power in trade eclipses the ability of the service sector to exercise its market power in the same market. What is the interaction between tariffs, market power, and the volume of trade? Differentiating equation (8) with respect to and yields the following: dqm = (y + bd) [ bdbm (c - x) - ybdam + bmad] < 0 (12) d A2 dqm qm = - < 0 (13) d d2qm d2qm > 0, > 0 d2 dd This allows us to make the following propositions. Proposition 3. Despite the presence of an imperfectly competitive service sector, it re- mains the case that international trade volumes decline with increases in the import tariff. 6 Proposition 4. International trade volumes are inversely related to the degree of concen- tration in the domestic trade and distribution sector, or alternatively the degree of market power exercised in the domestic distribution sector. Proposition 5. The negative impact of a marginal change in market power on trade volumes is greatest in a zero tariff context, and its marginal impact falls with increased levels of import protection or concentration. Hence, the largest impact of imperfect competition in the service sectors will be observed in zero-tariff countries, free-trade areas, customs unions, and under non-reciprocal trade preferences. We focus next on the welfare implications of a range of alternative tariff regimes for the importer, and the role played by service-sector competition across these possibilities. Domestic welfare W is comprised of four elements: service sector profits s, domestic upstream producer profits d, consumer surplus CS, and tariff revenue TR. Thus: W = s + d + CS + TR (14) An explicit expression for service-sector profits is obtained by combining equations (6), (7), and (8). s = ( - 1)[bdH2 + bmG2 + y (G - H)2]/A (15) As both the service-sector profit margin and the volume of trade decline with the tariff, profits of intermediaries decline as the trade tax is increased. The economic profits of the upstream sector can be measured directly by the area between the domestic supply curve and its intersection with the domestic ex-factory price. Combining equations (3) and (7) yields equation (16). d = bd[bmH + y(G - H)]2/2A2 (16) 7 Similarly, consumer surplus CS is simply the familiar triangle under the demand curve (4) and above the final demand price p. This is represented by equation (17). CS = y (bmG + bdH)2 /2A2 (17) Finally, tariff revenue follows directly from equation (8). TR = ( - 1){adA + bm [bdH + y(H - G)]}[bdH + y(H - G)]/A2 (18) Combining equations (15), (16), (17), and (18) with equation (14) yields welfare as a * 15 W 10 s 5 d CS 0 1 2.1 4.1 TR Figure 1: Welfare decomposition with varying rates of 8 function of the basic coefficients of our model. If we then take first-order conditions for welfare maximization, we can solve for the optimal tariff as a function of and the basic demand and supply coefficients of the model. This yields equation (19). = bdy[( - 1)J - amK] - 2bmJK (19) [ambm( - 2)K - bmJ - ambdy]K where J bd(x - c) + ady and K bd + y Figure 1 illustrates domestic welfare and its components for the case of duopoly in the service sector.4 As would be expected, consumer surplus declines monotonically with an increasing tariff, while tariff revenue increases to a maximum and then falls. Consequently, for national welfare, there is an interior solution for the optimal tariff, indicated by in the figure. The loss to the service sector and consumers (d and CS) from an increasing tariff rate is more than offset to the left of the optimal tariff by the combination of rising domestic profits for upstream producers d and tariff revenue TR, while it is only partially offset to the right of the optimal tariff line. The government, in exercising its monopoly power in trade, has the ability to limit the ability of the service sector to extract rents. As has already been established, the profits of the service sector decline with the tariff. Consequently when these rents accrue to domestic agents, the government will wish to moderate its use of the tariff. Indeed, viewed from the perspective of the optimal volume of imports qm the trade-off is complete. This can be seen by substituting equation (19) into equation (8), which yields equation (20). J - amK qm = (20) 2bmK + bdy 4The coefficient values used in Figures 1 and 2 are = 1.5, am = ad = 10, bm = bd = 2, y = 1, x = 20, c = 1. 9 From equation (20), we can see that from a welfare perspective optimal imports are inde- pendent of the degree of market power in the domestic service sector. The coefficient does not appear in equation (20). In exercising the optimal tariff, the government would seek to target the optimal volume of imports by adjusting the tariff rate to compensate for variations in service sector market power . As a result, the optimal tariff is a strictly decreasing function of the degree of market power in the service sector. This can be shown by differentiating equation (19) with respect to . d (J - amK)(ambdy + bm)(bdy + 2bm) = - (21) d [ambm( - 2)K - bmJ - ambdy]2K The sign of equation (21) is negative whenever qm > 0. These relationships are illustrated in Figure 2, where we plot optimized tariffs, welfare, and quantities for a range of competition index values. The figure is based on the same set of model coefficients as in Figure 1. The key difference is that we are now varying our index of competition and then plotting optimum quantities qm and qd, along with welfare W and the optimum tariff t = ( - 1). As can be seen in the figure, the optimal tariff rate falls with our market power index , as does welfare W and domestic shipments qd, while from equation (20) imports remain fixed. With the additional distortion in the market, in the form of an imperfectly competitive distribution sector, the welfare implications of trade policy become more complicated. It is evident that the optimal tariff declines with increasing concentration in services. Indeed, as illustrated in Figure 2, the optimal tariff when the service sector is a monopoly is a subsidy. In the absence of such an optimal tariff offset by the government, the more concentrated the service sector, the greater its exercise of its market power and, consequently, the lower the trade volume. A tariff further reduces the volume of trade, whereas a subsidy increases the level of imports and hence consumption. Such a subsidy benefits the service sector but, as their profits are part of national welfare, a welfare maximizing government would 10 be prepared to offer it. We summarize the relationship between tariffs, profits, trade, and welfare in the following propositions: Proposition 6. The optimum import tariff is a decreasing function of the degree of market power in the domestic trade and distribution sectors, and with a domestic service monopoly or cartel, the optimum tariff may actually be a subsidy. Proposition 7. There is scope for either the private service sector (through markups) or the government (through tariffs) to exercise market power in international trade, with the optimum tariff implying direct substitution. 24 16 W * 8 t *=100(*-1) qd* qm* 0 1 1.25 1.5 1.75 2 Figure 2: The optimal tariff and welfare when varying 11 4 Market Access and the Exporter Consider the impact of alternative tariff and competition regimes for the exporter. If we are focused on quantity alone, then equations (8), (12) and (13) point to a negative relationship between tariffs and imperfect competition, on the one hand, and export volumes on the other. In addition, taking the cross-derivative from equation (13) we can see that the trade- volume effect of a tariff reduction depends on the underlying trade volume and hence on the degree of competition in the domestic distribution sector. To some extent, tariff reductions may simply lead to a greater exercise of market power by the domestic distribution sector (and vice-versa), nullifying expected direct benefits from tariff reductions in export markets. A second measure of the benefits of improved market access conditions is exporter producer surplus PS. Once again, this is simply the area of a triangle, in this instance the area between the inverse supply curve and the export price: bdH + y (H - G) PS = (22) 2A2 From equation (22) we can calculate the welfare benefit to exporters of improved market access as manifested through increases in export quantities as being simply: dPS K (bmJ + amybd) = - < 0 (23) d 2A2 Further manipulation then confirms that the PS benefit of tariff reductions is a decreasing function of the underlying market power of the service sector. d2PS K (bmJ + amybd) = > 0 (24) dd 2A2 We summarize this section with the following propositions. 12 Proposition 8. The market-access benefits of tariff reductions in export markets are in- versely related to the degree of market power exercised by the domestic trade and distribution sector in the export market. Proposition 9. The benefits of market access concessions can be offset by increases in the degree of market power exercised by the domestic trade and distribution sector in the export market. The first of these propositions formalizes the dependence of goods market integration in the European Union (recall the EU autos exemption) on distribution sector competition. The second goes directly to the heart of the Fuij-Kodak dispute. To use a technical GATT/WTO term, nullification and impairment can follow from changes in domestic regulation of the distribution sector. 5 Empirics We turn next to an empirical exercise, focusing on whether the basic effects we have discussed, imperfect competition in distribution affecting market access in goods, matters in an empirical sense.his involves estimating reduced form gravity equations of bilateral trade flows, based on tariffs, distance, and exporter-specific dummies for a cross-section. (See for example Disidier and Head 2003 and Anderson and van Wijncoop 2003.) We include measures of distribution sector competition, as a check on our theoretical results developed above. Our basic data for this exercise are summarized in Table 1. From the OECD (2000), we work with two estimates of the degree of competition in the road freight and retail distribution for some, but not all, OECD members. This includes an index of barriers to entry in the sector, and also what can be interpreted as an overall or composite index of 13 the degree of competition in the sector. These estimates provide a single set of indexes for the late 1990s. For trade, we work with bilateral merchandise trade data extracted from UNCTAD's COMTRADE database and matched to bilateral import protection data taken from GTAP (2005). These data are for 2001. The tariff data offer the advantage of including a bottom-up concordance from detailed tariff data to aggregate bilateral trade flows, including preferential tariff rates. We also have included estimates of the trade-tax equivalent of export barriers as part of the basic trade barrier data (primarily the quota regime on textiles and clothing). In addition, bilateral export data have been adjusted to reflect estimated freight margins. For 69 countries as exporters, we have matched bilateral import data to other country-specific data for the 22 OECD importers covered by our set of OECD indexes on the distribution and freight sectors. We also incorporate data on distance, common language, and common borders from Guillaume, Mayer and Zignago (2004). Finally, as we are unable to use importer dummies for this exercise, we also include data on importer GDP and per-capita income from the World Bank (2005), following the older (pre fixed effects) gravity literature. After matching trade data to our competition data, we have 1,725 bilateral trade flows to work with involving OECD countries as importers in 2001. Our estimating equation is a reduced-form gravity equation, utilizing the data discussed above and augmented to reflect our propositions based on equation (8). Since we are working with a single year, using exporter dummies controls for fob prices, while value flows map to quantities if we normalize these prices to unity. Defining imports by country 14 j from country i as Mi,j, we work with the following estimating equation. qm,i,j = 0 + 1 ln(GDPj) + 2Disti,j + 3 ln(i,j) + 4LANGi,j (25) + 5BORDERi,j + 6 ln(Indexj) + 7 [ln(Indexj) ln(i,j)] + 8 [ln(PCIi) ln(Indexj) ln(i,j)] + 9 Di + 10NAFTAi,j + 11EEAi,j + i,j ,i i The Di terms are dummy variables assigned to each exporter, to reflect the set of exporter- specific variables that remain fixed across importers. The variables NAFTAi,j and EEAi,j are also dummies, capturing joint membership in either the North American or European free trade bloc. The terms Disti,j and Ti,j measure bilateral distance and import barriers (a combination of trade-weighted import tariffs and trade tax equivalents of export restraints) as a share of total import value. We expect the coefficients applied to these variables, 2 and 3 to both be negative. Recall that the Index term is meant to capture the effects related to in the discussion above. From the expressions in (8), we expect 6 to be negative as well. We expect the interaction term to be positive, based on equation (13) and Proposition 5. We have also included the interaction term 8 to allow for possible variations in the impact of tariff and competition-related barriers depending on the level of development of the trading partner. We explore this issue further below with split-sample regressions. Table 2 presents robust regression results for equation (25), based on both versions of our competition index. We have reported robust regression results because the Breusch- Pagan (1979) Chi-squared test statistic (as implemented in STATA) leads us to reject the hypothesis of homoscedasticity at any conceivably reasonable level of significance. Further examination with Szroeter's (1978) test statistic points to a pervasive problem, involving 15 roughly half of the right hand side variables. Many of these relate to the exporter fixed effect variables, indicating for example greater variance in the data involving some exporting countries than others. This is not surprising, as we have included relatively small aggregate trade flows (all flows over $10,000), usually involving a range of least developing countries. In these cases, bilateral trade flows may be a function of historical/structural variables unique to a given country pairing. Given the pervasiveness of the problem, there is a not an obvious single adjustment to be made to the data. We therefore resort to robust least squares, involving Huber-type (1981) robust regressions as implemented in STATA. These results are what are shown in Tables 2 and 3. Turning first to Table 2, this reports the results for equation (25) with both indexes. Relevant coefficients are significant in the 0.05 to 0.01 range or better, with the sign pre- dicted from our theoretical analysis for the direct effect from competition. (Where we have expectations of sign, the one-tailed significance results in the table are appropriate. This includes both competition indexes.) An F-test for the joint significance of the competition coefficients 6 and 7 rejects the null hypothesis that the coefficients are jointly zero at the .001 level. Country fixed-effect coefficients are not shown, though they are all generally significant at the 0.001 level across all regressions. The pattern of results for competi- tion fits expectations. Basically, these results suggest that tariffs and reduced competition both have a dampening effect on estimated trade flows, consistent with our theory-based propositions in the previous sections of this paper. Table 3 presents a further decomposition of patterns in the data, based on split-sample regressions. Implicit in the analysis above is that competition matters more as importers have more market power. In terms of the previous section, this depends on the rela- tive slopes of the supply and demand schedules, in conjunction with the general level of competition in the service sector itself. In a more general sense, we may expect import- 16 ing/distribution firms to have more market power vis-`a-vis smaller suppliers. At the same time, exporters in lower income countries may be less organized, and less adept, in hold- ing their own against market power exercised by buyers. (Imagine WalMart negotiating supplier contracts in Jamaica, as opposed to in Canada.) In Table 3 we explore this is- sue by making the following splits in the data. The first split involves OECD trade with low-income countries (defined as having a per-capita income below $1000 in 2001 dollars), and all other trade. For the second split, we divide the sample into OECD trade where the importer is large (with a nominal GDP greater than $500billion) and the exporter is small (defined as having a nominal GDP below $100billion), versus all other trade. For the final split, we examine OECD trade where the importer is large and the exporter is both poor and small. In all cases, we find that the correlation in the data between exports to the OECD and competition is greater when there is likely to be greater market power, in the sense that it matters more for smaller and poorer exporters. The structure of the retail and distribution sector in the OECD countries is more of a trade barrier for small and low-income countries than it is for exporters from higher income and larger economies. Finally, Table 4 is our attempt to convey a sense of the magnitudes involved, not so much statistically but rather economically. In the table, we have taken the tariff coefficient from Table 2, combined with sample values for EU competition indexes and a competition coefficient estimated for the intra-EU15 subset of our full sample. We have used these to calculate a trading cost- or tariff-equivalent from changing the degree of competition in the sample of EU countries, for intra-EU (i.e. duty-free) trade. Hence, for example, from the first column of numbers in Table 4, moving France to the average level of competition in distribution across the EU would be comparable to eliminating a 4.2 percent tariff against its EU partners. Moving to the most competitive level in the sample would correspond to the elimination of an 8.4 percent tariff. In the table, these trading cost equivalents range 17 between 0.0 and 8.4 percent of the value of trade, with most between 3.0 and 4.0 percent of the value of trade. The patterns of results in Tables 2, 3, and 4 suggest that variations in the degree of domestic competition matters for trade. Indeed, problems with competition in domestic distribution and trade activities are likely to themselves act as barriers to trade. In a European context, this means that continued competition exemptions for automobiles, for example, should indeed be expected to hinder trade substantially. In the context of multilateral negotiations in the World Trade Organization (WTO), this also means that WTO-based liberalization of these service sectors under the GATS (General Agreement on Trade in Services) may also mean improved market access conditions for affected goods sectors along the lines developed here. More broadly, this supports the notion that the benefits of trade for exportering countries is a function of their market power vis-`a-vis trade and distribution firms in the importing countries. At the same time, increased FDI flows in the service sectors, if they lead to increased concentration and less rather than more competition in distribution and trade services, ironically may lead to an erosion of market access conditions for goods, both in a customs union and bilateral setting. 6 Summary and Conclusions The pattern of trade in goods depends on a number of factors. Recent work has stressed transport costs and its linkages to the geography of production and trade. We take a different slant here. In this paper we examine the interaction between trade in goods and the degree of market power exercised by the domestic trade and distribution sectors ­ the so-called margin sectors. We first develop a theoretical model that allows us to highlight interactions between the degree of competition in domestic service sectors and the pattern of trade in goods. This is followed by an econometric exercise involving the import patterns 18 of 22 OECD countries vis-´a-vis 69 trading partners. Our theoretical results point to an expected linkage between service sector competition and goods trade. At least in theory, the domestic service sector can serve as an effective import barrier. This is also supported by our econometric results. These point to statistically significant linkages between effective market access conditions for goods and the structure of the service sector. From back of the envelope calculations, they also point to economically/qualitatively significant effects. (See Table 4.) What all this means is that, by ignoring the structure of the domestic service sector, we may be seriously overestimating the market access benefits of actual tariff reductions given the existence of imperfect competition in the margin sectors. We also find that the competition of margin sectors matters more for poor and small exporters than for others. Finally, our results suggest that GATS-based services liberalization may boost goods trade as well, if it leads to more competition in the distribution and trade sectors. Where GATS-based liberalization involves FDI and increased concentration, such service sector liberalization may instead have the unintended effect of eroding market access conditions for goods. 19 References Anderson, J. and E. van Wincoop (2003). "Gravity with Gravitas: A Solution to the Border Puzzle." American Economic Review 93(1): 170-192. Arnold, J., Mattoo, A., and B. Smarzynska (2006). "The Productivity Effects of Services Liberalization: Evidence from the Czech Republic. " working paper: World Bank. Breusch, T.S. and A.R. Pagan (1979). "A Simple Test for Heteroscedasticity and Random Coefficient Variation." Econometrica 47: 1287-1294. Disdier, A.C. and K. Head (2003). "Exaggerated Reports on the Death of Distance: Lessons from a Meta-Analysis." working paper, TEAM, Universit´e de Paris I Panth´eon Sorbonne. Flam, H., and H. Nordstrom (1995). "Why Do Pre-Tax Prices Differ so much across ¨ European Countries." CEPR Discussion Paper No. 1181. Francois, J.F. (1990). "Producer Services, Scale, and the Division of Labor." Oxford Economic Papers 42: 715-729. Francois, J.F. and H. Horn (2006). "Antitrust in Open Economies." CEPR discus- sion paper, forthcoming in V. Ghosal and J. Stennek, eds., The Political Economy of Antitrust, Elsevier-North Holland. Guillaume, G., T. Mayer and S. Zignago (2004). "Notes on CEPII's distance mea- sures." CEPII discussion paper, March. Global Trade Analysis Project (2005). The GTAP Database version 6: public-release version. GTAP consortium, Purdue University. 20 Head, K. and J. Ries (1997). "International Mergers and Welfare under Decentralized Competition Policy." Canadian Journal of Economics:30(4), 1104-23. Horn, H. and J. Levinsohn (2001). "Merger Policies and TradeLiberalization." Eco- nomic Journal 111 (April): 244-273. Huber, P. (1981). Robust Statistics. John Wiley & Sons: New York: 153-199. Ito, T. and A.O. Krueger editors (2003). Trade in Services in the Asia-Pacific Region: NBER-East Asia Seminar on Economics, Volume 11. Chicago and London: University of Chicago Press, 2003. Lutz, M. (2004). "Pricing in Segmented Markets, Arbitrage Barriers and the Law of One Price: Evidence from the European Car Market." Review of International Economics 12(3): 456-475. Markusen, J.R. (1989). "Trade in Producer Services and in Other Specialized Interme- diate Inputs. " American Economic Review 79: 85-95. Mattoo, A. and P. Sauve editors (2003) Domestic Regulation and Services Trade Lib- eralisation, Washington and Oxford: Oxford University Press and World Bank. Nanto, N. (1998). "The Kodak-Fuji Film Trade Dispute at the WTO." Congressional Research Service: Washington. OECD (2000). "Regulatory Reform in Road Freight and Retail Distribution." paper ECO/WKP (2000)28, Paris. Olarreaga, M. and C. Ozden 2005. "AGOA and Apparel: Who Captures the Tariff Rent in the Presence of Preferential Market Access?" World Economy 28(1): 63-77. 21 Szroeter, J. (1978). "A Class of Parametric Tests for Heteroscedasticity in Linear Econo- metric Models." Econometrica 46: 1311-1327. World Bank (2005). World Development Indicators. Washington DC. 22 Table 1 Database Overview (Value Data Reported in logs) name description mean max min GDP Importer gross domestic product in 12.797 16.126 10.858 billions of dollars in 2001 Source: World Bank (2005). PCI PPP-based per-capita income, dollars, 2001. 9.675 10.517 7.709 Source: World Bank (2002). M Imports, millions of U.S. dollars in 2001. 4.695 12.011 -4.605 Source: UNCTAD COMTRADE and GTAPv6.2 databases. = 1 + t MFN trade-weighted tariff (adjusted 0.028 0.670 -0.123 for trade preferences where available) based on a concordance of WTO, UNCTAD, and MACMAPS tariff data. Source: GTAPv6 database. Dist Distance between national capitals, from 8.332 9.884 2.821 the CEPII database of distance measures. Source: Gaulier, Mayer, and Zignago (2004). Border Sharing a common border. 0.041 1.000 0.000 Source: Gaulier, Mayer, and Zignago (2004). Lang Sharing a common language 0.059 1.000 0.000 Source: Gaulier, Mayer, and Zignago (2004). Index1 Overall index of competition in the 0.735 1.548 -0.223 freight/distribution sectors. Source: OECD (2000). Index2 Index of barriers to entry in the 0.747 1.705 -0.357 freight/distribution sectors. Source: OECD (2000). NAFTA A dummy variable for the case where 0.005 1.000 0.000 importer and exporter are both in the North American Free Trade Area. EEA A dummy variable for the case where 0.221 1.000 0.000 importer and exporter are both in the the European Economic Area. Note: The scale of competition indexes in levels ranges from 0-6, for least to most restrictive regimes. For countries reported as an interval by the OECD, the mid-point has been used. Index data are available for 22 OECD countries. Trade data are grouped by these 22 importers and by 69 exporting countries. Applied tariff data and distance data have been matched to these bilateral trade pairs. 23 Table 2 Robust Regressions: Gravity Equation of Bilateral Trade model 1 model 2 general index index of of competition entry barriers 1: GDPj 0.959 0.956 (62.86)*** (62.33)*** 2: Disti,j -1.057 -1.046 -(28.51)*** -(28.11)*** 3: ln(i,j) -1.836 -1.994 -(3.30)*** -(3.60)*** 4: LANGi,j 0.599 0.595 (7.19)*** (7.14)*** 5: BORDERi,j -0.033 -0.001 -(0.30) -(0.01) 6: ln(Indexj) -0.300 -0.242 -(7.73)*** -(7.80)*** 7: [ln(Indexj)ln(i,j)] 4.527 8.020 (1.00) (2.24)** 8: [ln(PCIi)ln(Indexj)ln(i,j)] -0.778 -1.185 -(1.46) -(2.77)*** 10: NAFTAi,j 0.631 0.684 (1.92)* (2.09)** 11: EEAi,j -0.105 -0.158 -(0.99) -(1.48) Summary statistics for estimates from robust regressions variables 78 78 observations 1701 1633 df 1622 1554 F: H0(0 = 1 = ... = 11 = 0), Pr > F 328.86, 0.0 318.59, 0.0 Summary statistics for OLS regressions R2 0.878 0.877 Note: Robust regressions are estimating using Huber method as implemented in STATA, with default convergence criteria. t-statistics are reported in parentheses , *, **, and *** indicating 0.15, 0.10, 0.05, and 0.01 levels of significance for a two-tailed test, or 0.075, 0.05, 0.025, and 0.005 where a one-tailed test is instead appropriate, as discussed in the text. 24 Table 3, Robust Regression Estimates Competition Coefficients with Split Samples model 1 model 2 general index index of of competition entry barriers Exporter is poor -0.339 -0.328 -(3.72)*** -(4.43)*** Rest of Sample -0.271 -0.193 -(6.46)*** -(5.78)*** A Large importer and a small exporter -0.366 -0.269 -(4.65)*** -(4.48)*** Rest of Sample -0.286 -0.239 -(6.93)*** -(6.77)*** A Large importer and a small, poor exporter -0.327 -0.299 -(2.46)*** -(2.75)*** Rest of Sample -0.279 -0.208 -(7.00)*** -(6.43)*** Note: Robust regressions are estimating using Huber method as implemented in STATA, with default convergence criteria. t-statistics are reported in parentheses , *, **, and *** indicating 0.15, 0.10, 0.05, and 0.01 levels of significance for a two-tailed test, or 0.075, 0.05, 0.025, and 0.005 where a one-tailed test is instead appropriate, as discussed in the text. 25 Table 4 Trade-cost Equivalents for Intra-EU Trade for Changes in Competition in Member States, % move to most move to average competitive EU regime EU regime Austria -3.4 -7.5 Denmark -1.3 -5.3 Finland -1.5 -5.6 France -4.2 -8.4 Germany 3.9 0.0 Great Britain -0.4 -4.4 Greece -0.4 -4.4 Ireland 3.0 -0.9 Italy -1.7 -5.8 Netherlands 3.0 -0.9 Portugal -0.6 -4.7 Spain -0.4 -4.4 Sweden 1.9 -2.1 Note: Based on competition index 1, Table 2 coefficient for tariffs, and a split-sample regression estimate of the competition index for the sub-sample of intra-EU trade. 26