77382 the world bank economic review, vol. 16, no. 1 81–108 Trade in International Maritime Services: How Much Does Policy Matter? Carsten Fink, Aaditya Mattoo, and Ileana Cristina Neagu Maritime transport costs significantly impede international trade. This article examines why these costs are so high in some countries and quantifies the importance of two ex- planations: restrictive trade policies and private anticompetitive practices. It finds that both matter, but the latter have a greater impact. Trade liberalization and the breakup of private carrier agreements would lead to an average of one-third lower liner transport prices and to cost savings of up to US$3 billion on goods carried to the United States alone. The policy implications are clear: there is a need not only for further liberalization of government policy but also for strengthened international disciplines on restrictive business practices. The authors propose an approach to developing such disciplines in the current round of services negotiations at the World Trade Organization. Maritime transport costs have a profound influence on international trade. In many cases, their trade-inhibiting effect dwarfs that of customs duties.1 For in- stance, the average incidence of transport cost exceeds that of tariffs on imports from the majority of U.S. trading partners (figure 1). More generally, economic research highlights the role of transport costs in determining geographical pat- terns of trade, production, industrial structure, and income (Venables and Limao 1999). Interesting new work even suggests that transport costs (as an element of trade costs) help explain a variety of puzzles in the field of international macro- economics, such as the well-known home biases in consumption and investment and the excessive volatility of exchange rates (Obstfeld and Rogoff 2000). These observations are interesting from a policy point of view, however, only if some- thing can be done about these costs. Are transport costs exogenously determined by technological developments or can they be influenced by policy? Carsten Fink, Aaditya Mattoo, and Ileana Cristina Neagu are with the Development Research Group at the World Bank. Their e-mail addresses are cfink@worldbank.org, amattoo@worldbank.org, and ineagu@worldbank.org, respectively. This article is part of the World Bank’s research program on trade in services, which is supported in part by the U.K. Department of International Development. The authors thank Marc Juhel and Alexander Yeats for stimulating discussions and Simon Evenett, Bernard Hoekman, Pierre Latrille, Marcelo Olarreaga, Isidro Soloaga, David Tarr, seminar participants at the World Bank, and two anonymous referees for helpful comments. 1. This has been demonstrated in several studies. See Waters (1970), Finger and Yeats (1976), Sampson and Yeats (1977), Conlon (1982), and Amjadi and Yeats (1995). © 2002 The International Bank for Reconstruction and Development / THE WORLD BANK 81 82 the world bank economic review, vol. 16, no. 1 Figure 1. The Relative Importance of Transport Costs and Tariffs for U.S. Imports, 1998 20 45Ëš 18 16 14 Tariff incidence 12 10 8 6 4 2 0 0 2 4 6 8 10 12 14 16 18 20 Transport cost incidence Note: Each dot represents a country. The tariff incidence is calculated as the ratio of actual duties paid over import values. Similarly, the transport cost incidence represents the share of transport charges in import values. Five countries (Benin, Guinea, the Solomon Islands, Togo, and Samoa) exhibit a trans- port cost incidence greater than 20 percent and are not shown. Source: U.S. Bureau of the Census. Some researchers argue that restrictive trade policies keep maritime transport costs high, notably the cargo reservation schemes and monopoly rights granted to providers of port and auxiliary services (Bennathan 1989, Amjadi and Yeats 1995, Francois and others 1996; Hummels 1999). Some also argue that private anticompetitive practices—primarily but not exclusively of the maritime confer- ences—are responsible for keeping costs high (Francois and Wooton 1999, Hummels 1999). However, most observers also argue that both public and pri- vate trade-restrictive policies are becoming less important (White 1988, Franck and Bunel 1991, World Trade Organization [wto] 1998). Yet the available evi- dence suggests that transport costs, especially for liner trade, are not falling— despite dramatic improvements in technology, especially in the form of contain- erization (Hummels 1999). This article seeks to assess the relative importance of public and private trade- restrictive actions in explaining the price of maritime transport services. To Fink, Mattoo, and Neagu 83 measure these prices, we use newly published data on U.S. waterborne trans- port from the U.S. Department of Transportation. A major advantage with these data is that they are broken down by type of service—liner, bulk, and tanker. It is more difficult to put together a comprehensive data set on public policies and private practices, a problem that has inhibited meaningful empirical research in this area. The few attempts to measure the restrictive impact of government policies have only limited coverage (McGuire, Schuele, and Smith 2000), and there has not been, as far as we know, an attempt to use existing information on carrier agreements.2 This article draws on a database, created as part of the World Bank’s services project, which contains information on both policy and private rate-fixing arrangements affecting maritime trade with the United States. These data made it possible to carry out the econometric analysis presented here. Our estimates confirm, first of all, the importance of all the standard deter- minants of transport prices, ranging from distance to technology. More inter- esting, we find that both public policy and private practices continue to exercise a significant influence on maritime transport prices. Somewhat surprisingly, private anticompetitive practices seem to have a stronger influence on prices than public restrictions. What are the implications for policy? The negotiations on maritime trans- port were the only post–Uruguay Round services negotiations that completely failed. This failure implied an unfortunate loss of political momentum for reform of domestic policies and, less obviously, a lost opportunity to develop procompetitive rules. To some extent, an effort was made to develop rules that would ensure nondiscriminatory access to port services.3 But these rules, con- cerned primarily with ensuring market access, did little to protect consumers from the anticompetitive practices of international cartels. An international ini- tiative is needed because these practices cannot be adequately addressed only through national competition policy, given the weak enforcement capacity of small states. A further reason for developing a first-best international response to these practices is to prevent recourse to an inferior national response: recall that the cargo-sharing schemes imposed by many developing countries were primarily a response to the perceived power of conferences. A possible way forward is to strengthen the provision of the General Agreement on Trade in Services (gats ), dealing with anticompetitive business practices to ensure that collusive pricing does not erode the gains from liberalization. 2. Kang (2000) uses the policy indicators developed by McGuire, Schuele, and Smith (2000) to es- timate the impact of restrictive maritime policies on bilateral shipping margins, defined as the ratio of cost-insurance-freight import values to free-on-board export values. This approach suffers from well- known data problems (import and export values are not reported by the same statistical entities) and by the undesirable property that shipping margins vary with unit values of shipped goods. The empiri- cal approach adopted here addresses both of these problems. 3. In some respects, the approach to port services, which can be seen as essential facilities often controlled by major or monopoly suppliers, was analogous to the approach to basic telecommunica- tions networks established in the procompetitive regulatory principles. 84 the world bank economic review, vol. 16, no. 1 I. An Overview of International Maritime Transport Maritime transport services consist of three types of activities: international maritime transport (freight and passengers), that is, the actual transportation service performed once the commodity is on board a ship in a country until the moment when the vessel reaches the destination port of a different state;4 mari- time auxiliary services, that is, any activities related to cargo manipulation in ports and on ships;5 and port services, that is, activities related solely to ship management in ports.6 This article uses data pertaining to restrictions affecting each segment of the market. Due to differences in commodity types as well as to technological improve- ments in the shipping industry (most importantly, containerization), international maritime freight transport has developed specialized branches. Thus, liner ship- ping—meaning maritime transport of commodities by regular lines that publish in advance their calls in different harbors—is distinct from tramp shipping, re- ferring to transport performed irregularly, depending on momentary demand. Typically, liner carriers transport commodities with a higher degree of indus- trial processing using containers, whereas noncontainerized raw materials (crude and refined oil, iron ore, grain, coal, and bauxite), generically known as bulk, tend to be carried in tramp carriers.7 Tramp shipping is generally believed to be a fairly competitive market, mostly free from restrictions (wto 1998). By contrast, liner shipping has traditionally been subject both to private cartel-like arrangements and government restric- tions. This article concentrates on the liner segment of the market. Cargo Reservation Schemes Over time, the most important category of barriers applied to international maritime transport has been various cargo reservation schemes. These schemes require that part of the cargo carried in trade with other states must be trans- ported only by ships carrying the national flag or interpreted as national by other criteria. These policies have typically been justified by either security (self- sufficiency in times of war) or economic (infant industry) concerns. 4. International transport as defined by gats excludes cabotage, which refers to transportation of commodities between ports of the same country. 5. In the gats classification, maritime auxiliary services include maritime cargo handling, storage and warehousing, customs clearance, container station and depot, maritime agency, and maritime freight forwarding. 6. In the gats classification, port services include pilotage, towing and tug assistance, provisioning, fueling and watering, garbage collecting and disposal, port captain’s services, navigation aids, shore- based operational services, and emergency repair facilities. 7. Bulk traffic is typically divided into two categories: tanker (including crude oil and oil-related products) and dry bulk (including iron ore, grain, coal, bauxite, and phosphates). Note that the distinc- tion between liner and bulk is not watertight. There exists a gray area that includes break-bulk (that is, loose, noncontainerized cargo transported using liners), general cargo (nonbulk commodities transported on liners without using containers), or containerized goods transported by tramp carriers. Fink, Mattoo, and Neagu 85 Cargo reservation takes various forms. It can be imposed unilaterally, so that ships flying national flags are given the exclusive right to transport a specified share of the cargo passing through the country’s ports. An alternative and more common form involves cargo sharing with trade-partner countries on the basis of bilateral or multilateral agreements. A specific form of such a scheme is the U.N. Conference on Trade and Development (unctad) Liner Code of Conduct or the 40-40-20 rule. This legal instrument, which was adopted in 1974 and entered into force in 1983 by its ratification by more than 70 countries, was meant to counteract the anticompetitive actions of liner conferences, which are cartel- like arrangements. In many cases, access of outside shipping companies to a liner conference used to be restricted,8 so governments applying the Liner Code re- quired these cartels to divide the cargo transported according to the following rule: 40 percent for ships belonging to the exporting country, 40 percent for ships belonging to the importing country, and 20 percent for ships belonging to other countries. These restrictions were intended to encourage the development of the shipping industry of developing countries. Cargo reservation schemes have declined in significance, as more countries have phased them out formally or chosen not to implement them. For example, in Asia, Indonesia adopted an “open seaâ€? policy in the late 1980s, Thailand abolished all cargo reservations in 1993, and Korea’s commitment to phase out its “designated cargo system,â€? made in 1995 on becoming a member of the Organisation for Economic Co-operation and Development (oecd), was fully implemented by 1998. In Africa, Côte d’Ivoire and Senegal are among the coun- tries that have eliminated cargo-sharing schemes. In Latin America, Chile pio- neered liberalization in 1979 and Peru phased out most restrictions in the early 1990s. A further indication of the reduced importance of cargo sharing is the spread of “open registriesâ€? in many countries and the intensification of the “deflaggingâ€? process, that is, the transfer of ships to open registries to enable the ship owners to benefit from more efficient cost conditions (wto 1998). The unctad Liner Code, which was never applied on a large scale, is even less vis- ible today, being applied mostly on routes between West Africa and Europe.9 Nevertheless, the evidence we have obtained on policy suggests that 11 of the countries in our sample, ranging from Benin to India, still have in place reserva- tion policies that at least nominally restrict the scope for trade. Most of these countries subscribed to the unctad Liner Code, whereas others (for example, Brazil, China, and Nigeria) implemented schemes that were similar in spirit. 8. The United States banned closed conferences. Cargo sharing and shipping conferences interacted over time, and in many cases authorities tailored their policies by taking into account the presence of carrier agreements. For example, Chile’s cargo reservation mechanism, before the liberalization of the past decades, was designed so as to favor access of Chilean shipping companies into conferences and to restrain conference pressures on nonaffiliated carriers (Bennathan 1989). 9. In many countries, national shipping companies that had access to the reserved share, but did not possess sufficient technical means for its transportation, used to sell their preferential right to cargo, a practice that resulted eventually in a higher transport cost. 86 the world bank economic review, vol. 16, no. 1 Nicaragua offers an example of a country that imposes a reciprocity condition, that is, access for foreign ships depends on whether their home countries grant Nicaraguan ships access. There are few empirical studies of the impact of cargo reservation. The most frequently cited is Bennathan’s (1989) analysis of the determinants of freight charges in Chile’s export trade with the United States before and after elimina- tion of cargo reservation. The results show that the indicators of cost-based pric- ing have greater explanatory power after liberalization than the indicators of demand-based pricing. This is seen as evidence of increased competition in the liner shipping industry. In another study, Pálsson (1997) suggests that in the South American market, the abolition of cargo sharing led to a decline in shipping rates to and from Europe by 20–50 percent, and to and from the United States by 25– 35 percent. This study also indicates that cargo rates from Europe to Abidjan and Dakar declined by 10–20 percent a year after liberalization in 1995 (Pálsson 1997).10 None of these studies provide any details on how these estimates were obtained, and to our knowledge there is not any cross-country analysis of the impact of cargo-reservation schemes. Price Fixing and Other Cooperative Agreements Maritime carriers enter various types of agreements, which help them enjoy ad- vantages that arise from cooperation on technical or commercial matters. Far from being a recent phenomenon, carriers’ collusive habits are deeply rooted in the his- tory of maritime transportation, and the first shipping conferences, covering the routes between the United Kingdom and Calcutta, date back to 1875. By joining carrier agreements, shipping companies retain their juridical independence but consent to common practices with the other members regarding pricing, traffic distribution, and/or vessel capacity utilization. Examples of carrier agreements that were recognized in U.S. regulation by the end of 1998 were conference agreements, cooperative working agreements, joint services agreements, pooling agreements, space charter agreements, and trans-shipment agreements. Conference agreements are made between two or more ocean common carri- ers, and provide for the fixing of and adherence to uniform tariff rates and con- ditions of service.11 Conferences are the most widespread type of rate-binding agreement. In the United States, conferences are required by law not to restrict the entry and exit of any shipping company. Therefore, shipping conferences in U.S. foreign trade are “open,â€? whereas those covering other routes may be closed 10. There are also some studies of the impact of the U.S. Jones Act, which prohibits foreign shipping firms from transporting goods or people from one U.S. location to another. Estimates of the price- increasing effect range from 100 percent (usitc 1991) of the average world price to a high of 300 percent (White 1988). Francois and others (1996) estimate that the welfare costs of this protection (assuming a conservative 100 percent price difference) are at least $3 billion a year. 11. Because conferences are a characteristic of liner shipping, they are also referred to as liner conferences. Fink, Mattoo, and Neagu 87 to outside carriers.12 The U.S. Shipping Act of 1984 defines cooperative work- ing agreements as agreements that establish exclusive, preferential, or coopera- tive working relationships, but that do not fall precisely within the arrangements of any specifically defined agreement. Only some of the carrier agreements have a rate-binding clause, that is, they declare that they engage in unique price set- ting for transport services provided by all members. The high incidence of conferences and other types of carrier agreements in maritime transport is due to the fact that the United States, the European Union, and many other countries exempt shipping conferences from antitrust regula- tion on the ground that they provide price stability and limit uncertainty regarding available tonnage.13 The exemption from antitrust law is com- pounded by the Federal Maritime Commission’s role in helping police price- fixing arrangements. The 1984 U.S. Shipping Act required all ocean carriers to file their rates with the Federal Maritime Commission and publish their rate and schedule information. Secret discounting on filed rates was considered il- legal. Through the imposition of fines, the commission was authorized to en- sure that the filed rates were actually charged.14 However, conferences were required to allow for independent action, meaning that members could post a rate different from the conference rate, provided they notified the conference in advance. Although this provision created some flexibility, there was prob- ably limited incentive to make public preannounced price cuts that were likely to be matched by rivals. In recent years, the power of conferences has eroded for two reasons. The first is the entrance in the market of strong and efficient outside shipping companies. Containerization and other forms of technological progress have made it pos- sible for outsiders to supply the same services as the conferences at lower costs to consumers. A second development is the change in regulations affecting in- ternational shipping, notably the U.S. Ocean Shipping Reform Act of 1998, which amended the Shipping Act of 1984 and went into effect in May 1999. While preserving the antitrust immunity of the rate-setting conference system, the Ocean Shipping Reform Act allows for the confidentiality of key terms (prices are in- cluded in this category) in contracts between shippers and carriers. This amend- ment is bound to create greater scope for price competition. 12. Recently, the European Commission claimed that steps taken by the Trans Atlantic Conference Agreement (taca ) to comply with the “openâ€? conference obligations of U.S. law had constituted an abuse of their dominant position. It was alleged that taca offered inducements to certain shipping lines to enter trans-Atlantic trade as parties to the conference rather than as independents (Levitt 2000). 13. Francois and Wooton (1999). See also Davies (1986), who states that “generally these cartels have been exempted from domestic legislation on competition, primarily because of jurisdictional prob- lems stemming from the international character of the industry, but also because governments have judged them useful for promoting the health of both international trade and national merchant ma- rinesâ€? (p. 300). 14. The rationale for these measures was ostensibly to protect small shippers from being disadvan- taged by their inability to extract discounts from shipping companies. 88 the world bank economic review, vol. 16, no. 1 In response to these developments, two types of arrangements have begun to emerge. First, shipping lines now sometimes enter discussion agreements. These allow conference and nonconference carriers serving a particular trade lane to discuss and share information about rates, costs, capacity, and service. The mem- bers may adopt voluntary rate, capacity, and service guidelines. Second, shipping companies and conferences tend to enter more wide-ranging organizations, such as consortia, alliances, and global alliances. There are two interesting questions, only the first of which we address in this article: How much do the traditional conferences still matter? Although these new arrangements are different from con- ferences from a juridical point of view, how different are they in actual behavior? Some recent events provide implicit evidence of the continued influence of collusive practices. Although price fixing by conferences is exempted from the scope of competition law, the abuse of a conference’s dominant position and the extension of collusion to other areas have provoked the wrath of European competition authorities. In 1992, the European Commission imposed fines against the members of the French/West African Ship Owners Conference.15 The com- mission found that the conference had deterred the entry of other operators by a combination of loyalty agreements with shippers and predatory pricing against nonconference lines. Furthermore, competition between lines belonging to dif- ferent conferences had been prevented by a partitioning of shipping routes: members of one conference were prohibited from operating in the ports served by another conference unless they first obtained membership, through a long, uncertain procedure, of that second conference. In 1998, the European Commission fined the Trans Atlantic Conference Agree- ment (taca) a sum of $314 million. The European Commission concluded that the conference, which controlled more than 60 percent of the traffic crossing the Atlantic at the time, set prices not only for the ocean leg but also for inland transportation by truck or train as well.16 In May 2000, the European Commis- sion imposed a penalty on 15 liner shipping companies that were members of the Far East Trade Tariff Charges and Surcharges Agreement (fettcsa )—an agreement abandoned in 1994 following action by the European Commission. Altogether, the companies controlled 80 percent of the traffic between northern Europe and the Far East. Again, the target of action was not price fixing per se, but the fettcsa members’ collective strategy of not offering discounts from pub- lished fares.17 Finally, reports in the maritime press also suggest that despite an 15. See p. 13 of the Annex of the Communication from the European Community and its Member States to the wto Working Group on the Interaction between Trade and Competition Policy, wt/wgtcp/ w/140, June 8, 2000. 16. As reported by cnn, the event marked a new record for fines imposed by the European Commis- sion on a cartel. This was the first time that any eu authority had assessed the compatibility of liner conference practices with eu competition law. 17. See “fettcsa : Commission fines shipping lines for an illegal price agreement on the Europe / Far East tradeâ€? (DN: IP/00/486), available from the European Commission’s Web page. Fink, Mattoo, and Neagu 89 increase in entry, the limited reductions in transport costs are attributable to the legal privileges granted to shipping company agreements.18 Notwithstanding this evidence, the issue of whether liner conferences were in a position to exercise market power has provoked some debate, primarily over the question of whether liner shipping markets are contestable. One view is that liner shipping markets satisfy a list of a priori conditions of contestability (Davies 1986, Zerby 1988). The entrant and the incumbent lines have access to the same technology and, provided that the market is not affected by any other distor- tions (such as cargo reservation), all shipping lines are equally placed with re- spect to access to cargo. Ship mobility and an active secondhand market imply that no significant sunk costs arise in the industry. Furthermore, the incumbent shipping lines are likely to provide a slow price response to the new entrants, especially if the former are organized in conferences, where price decisions re- quire consensus among members. The frequent entry and exit on certain mari- time routes has been cited as evidence of the contestability of these markets (Davies 1986). An alternative view questions each of these assertions (Pearson 1987, Jankowski 1989). First, it argues that building up goodwill represents a substantial sunk cost, and lines cannot enter and exit markets with complete disregard of the effect this has on their reputation. Advertising and agency costs expended to establish regular liner services are also examples of intangible sunk costs.19 Second, there is evidence that conferences have developed quick price-response mechanisms to respond to entry, including the use of action committees vested with the power to match the rates offered by an outsider. These arguments are in line with the findings of the European Commission cited previously. Finally, the frequency of entry and exit is clearly not convincing evidence of contestability: after all, at least in equilibrium, a contestable market would witness no entry at all. As far as we know, there has been only one attempt to examine econometrically the influence of conferences. Clyde and Reitzes (1995) find no statistically sig- nificant relationship between freight rates and the market share of conferences serving a route. However, they find that the level of freight rates is significantly lower on routes where conference members are free to negotiate service contracts directly with shippers. On this basis, they conclude that the evidence on whether liner conferences are effective cartels is at best mixed. They suggest that there is an alternative interpretation of their results: the industry’s antitrust immunity and tariff filing and enforcement requirements potentially facilitate the ability 18. See, for example, “Obstacles Lie Ahead,â€? 1999 Year-end Economic Review, Bangkok Post, 1999. 19. Franck and Bunel (1991) suggest that it may be appropriate to distinguish between two market segments in liner shipping by the type of entry criterion. In the first, entry is easy and competition from occasional outsiders is strong because they adopt a hit-and-run strategy and are not concerned with staying in business. In the second, an outsider is interested in competing on a long-term basis with the conferences and providing as high-quality services as them, so it is more difficult to comply with the entry conditions. 90 the world bank economic review, vol. 16, no. 1 of all carriers to collude, not just those carriers that are conference members. Furthermore, in seeking to determine the influence of the market share of con- ferences, they consider only the routes on which conferences exist. They do not test a more basic hypothesis that the critical effect is of the existence of a confer- ence on a particular route rather than its precise share of the market. Restrictions on Port and Auxiliary Services Both port and auxiliary services, particularly cargo handling, have tended to be monopolized. Liberalization of these services has two aspects. One is to ensure that foreign ships serving the domestic market obtain nondiscriminatory access to such services. The second is to allow competition, domestic and foreign, in the supply of the service itself. Seaports are typically coordinated by public or, in fewer cases, private organi- zations called port authorities. Depending on the role assumed by these institu- tions, seaports can be classified into different categories. With landlord ports, the port authority owns and manages port infrastructure and private firms provide the rest of port and maritime auxiliary services; private firms are able to own super- structure and operate assets pertaining to infrastructure by concession or licensing. With tool ports, the port authority owns both infrastructure and superstructure, but private firms provide services by renting port assets through concessions or licenses (for example, Antwerp, Belgium). Finally, with service ports, the port authority owns assets and supplies services by directly hiring employees. Trujillo and Nombela (1999) argue that the landlord port is the most desir- able category from the efficiency point of view, since it allows private enterprise and market forces to play a role in the supply of services, while preventing mo- nopolization of essential assets by private firms. For instance, in the case of Puerto Nuevo in Buenos Aires, six terminals were competitively commissioned to the private sector, with substantial foreign participation in the case of three.20 The government also established free entry in the sector by allowing any operator to build, manage and operate a port for public or private use. These reforms trans- formed Argentine ports from the most expensive in Latin America to among the cheapest. Average charges per container declined from $450 to $120 and con- tainer time at port declined from 2.5 to 1.3 days (Trujillo and Estache 2001). Chile also witnessed a significant improvement in port performance after the competitive allocation of the right to operate ports, with five major world op- erators participating in the bidding consortia (Foxley and Mardones 2000).21 20. Terminals 1 and 2 were originally awarded to an international consortia headed by P&O Aus- tralia in partnership with Fasce sa , a local stevedore company (Trujillo and Estache 2001). Terminal 5 was awarded to an international consortium headed by the Manila-based international operator Inter- national Container Terminal Services, Inc. 21. Hutchinson, P&O, Stevedoring Services of America, hhla and ictsi . In fact, World Bank (2000) reports that the top nine international terminal operators account for 40 percent of the world’s con- tainer liftings. Fink, Mattoo, and Neagu 91 This anecdotal evidence indicates that international participation in the provi- sion of terminal services is now a reality and that the introduction of competition can make a substantial difference in performance. With this broad benchmark in mind, we seek to capture some of the restrictions in place on port and auxiliary services. II. The Model In this section, we develop an econometric model of liner transport prices for U.S. imports. The analysis focuses on the ocean leg of the journey because the data available do not directly capture the price of maritime auxiliary services and port services.22 Nevertheless, the analysis includes policy restrictions affect- ing the latter type of services. This is because the restrictions are likely to have an adverse effect on the efficiency with which these services are supplied to lin- ers and hence push up the costs of liner services—for example, because of longer waiting or unloading times.23 We do not formally derive our estimation equation from a fully specified struc- tural model of competition or collusion among liner companies, but our approach can be best understood in terms of a simple constant-elasticity pricing formula. This pricing rule relates the U.S. dollar price of shipping product k from foreign port i (which is located in country I) to U.S. port j (which is located in U.S. cus- toms district J), Pijk, to the marginal cost for this service, MC(i, j, k), and a markup term, Φ(I, J, k): (1) Pijk = Φ(I, J, k) MC(i, j, k). The markup term is a function of the elasticity of demand perceived by liner companies serving the routes between country I and customs district J for prod- uct k. The pricing formula in equation (1) could, for example, be easily derived from a model of Cournot competition. Taking natural logs of equation (1) yields (2) pijk = φ(I, J, k) mc(i, j, k), where lowercase letters refer to natural logs of the respective variables. Unfortunately, we do not have any direct information on the costs of mari- time transport operations. We therefore decompose the marginal cost term, mc(i, j, k), as follows: (3) mcijk = αJ + λk + γTijk + δdiJ + ηqiJ + Ï?CRI + Ï•1PS1I + Ï•2PS2I. 22. More precisely, the data reflect transport charges incurred in bringing the merchandise from along- side the carrier at the port of export and placing it alongside the carrier at the first U.S. port of entry. 23. The possibility of measurement error provides a more mundane reason for considering the impact of restrictions on the port and auxiliary services. Although in principle the liner transport prices do not include the prices of these services, in practice such a clean truncation may not have been possible. 92 the world bank economic review, vol. 16, no. 1 The first term, αJ, reflects an effect specific to each U.S. customs district. It cap- tures differences across customs districts in port services and other auxiliary services, such as cargo handling, and has been included for the reasons noted above. The second variable, λk, is a product-specific effect that captures differ- ences in the physical properties of shipped goods, such as weight or size. The third effect is a technological effect represented by the share of goods shipped in containers, Tijk. Since containerization is likely to reduce the marginal cost of liner services, we expect the coefficient γ to have a negative sign.24 The fourth cost variable is (the natural log of) the shipping distance between foreign port i and the main port in customs district J, diJ. There is some evidence that the effect of shipping distance on transport cost becomes less important for longer distances (Hummels 1999), and so we expect 0 < δ < 1. Fifth, we include an econo- mies-of-scale effect represented by (the natural logarithm of) the total value of U.S. imports carried by liners (including nontextile goods) between foreign port i and district J, qiJ. If there are economies of scale with regard to traffic originat- ing from the same port, we expect the coefficient η to be negative. Finally, we add three policy indicators that capture restrictions maintained by I’s government affecting the supply of maritime services by foreigners. These restrictions are expected to lead to inefficiencies and the employment of outdated technology. Specifically, CRI is a dummy that indicates whether exporting coun- tries maintain any form of cargo reservation policy for the domestic shipping fleet affecting trade with the United States. PS1I is an index that captures the existence of barriers to the foreign supply of cargo-handling services, considered to be one of the most important auxiliary services. PS2I is an index that mea- sures the number of port services (for example, pilotage, towing, navigation aids, and waste disposal) that are mandatory for incoming ships. In the absence of more direct data on the openness of the port services regime, the extent to which the use of such services is mandatory is used as an indicator of the restrictive- ness of the port services regime. As noted above, the costs of auxiliary and port services are not directly captured by the maritime price data, but restrictions in both are relevant because they could push up the costs of liner services. The markup term, φ(I, J, k), is assumed to depend on the following four vari- ables: (4) φ(I, J, k) = µk + Ï„CRI + ψ1 A1IJ + ψ2 A2IJ. The first term, µk, reflects a product-specific effect that captures differences in transport demand elasticities across sectors. Note that the transport demand elasticities are derived from the final demand for product k in the United States. The second variable is again the variable that captures the existence of cargo reservation policies, which directly limit the extent of competition from foreign liners and thus may push up markups. The third and fourth effects, A1IJ and A2IJ, 24. Over 80 percent of U.S. imports are containerized. Noncontainerized shipments are because certain foreign ports, mostly in the developing world, are not yet equipped with container terminals. Fink, Mattoo, and Neagu 93 are due to the existence of collusive agreements among liner companies on routes between country I and customs district J. We distinguish between two kinds of collusive agreements: price-fixing agreements (which include most conferences) and cooperative working agreements that do not have a binding rate-setting authority. A single agreement typically covers routes between the ports of a for- eign country and one or more U.S. coastal districts that each consists of several customs districts. Because collusion between liner companies is likely to push up markups, we expect both coefficients ψ1 and ψ2 to show a positive sign. But conference and other price-fixing agreements are likely to be more powerful and to have a greater impact on transport prices than cooperative working agree- ments, that is, we expect ψ1 > ψ2. Substituting equations (4) and (3) into equation (2) and inserting an error term, εijk, we obtain (5) pijk = αJ + βk + γTijk + δdiJ + ηqiJ + ψ1 A1IJ + ψ2 A2IJ + ω CRI + Ï•1 PS1 I + Ï• PS I + εijk, 2 2 where βk ≡ (λk + µk), ω ≡ (Ï? + Ï„), and we expect the coefficients on the three policy indicators, ω, Ï•1, and Ï•2, to have a positive sign.25 We calculate the transport price, pijk, as the share of liner transport charges in import values for good k (at the six-digit HS aggregation) multiplied by the unit value of imports. The U.S. Department of Transportation defines transport charges as all freight, insurance, and other charges (excluding import duties) incurred in bringing the merchandise from alongside the carrier at the port of export and placing it alongside the carrier at the first U.S. port of entry.26 However, actual data reported may include charges for port services and inland transportation.27 To reduce the potential bias resulting from differences in inland transportation costs, we exclude observations for which the origin of the import is different from the country of the port of shipment (for example, landlocked countries) as well as all in-transit shipments.28 The appendix provides additional information on the construction and sources of all variables. Table 1 presents an overview of our estimation data set. It covers all U.S. imports carried by liners from the 59 countries for which we could find infor- mation on maritime policies. Data refer to 1998. Liner imports account for around 65 percent of the total value of maritime imports, the remaining 35 per- 25. Because we estimate both product fixed effects and customs district–specific effects, we need to drop one dummy variable (for one customs district) to avoid perfect colinearity among the explanatory variables. 26. If insurance costs are not closely correlated with transport charges, there is the possibility that our transport price variable is distorted. However, this should at least partially be remedied by the inclusion of product fixed effects, because differences in insurance costs are likely to be greatest across products. 27. According to e-mail communication with an official from the U.S. Department of Transportation. 28. Note that we do not exclude the trade originating in third countries and in-transit traffic when calculating total import values qiJ. 94 the world bank economic review, vol. 16, no. 1 Table 1. Overview of U.S. Imports Carried by Liners in 1998 Share of Share of liner liner maritime imports in total Liner transport Liner import imports (%) imports (%) Number of charges charges Countries countries (million $) (million $) Total Non-oila Total Non-oila Developing 37 3,940 82,400 64.88 74.82 48.76 53.10 Industrial 22 3,080 104,500 64.20 66.24 50.73 51.23 Total 59 7,020 186,900 64.71 70.04 50.13 52.38 aExcluding HS category 27. Source: U.S. Department of Transportation and U.S. Bureau of Census. cent being carried by tramp services.29 About half of all U.S. imports (including all modes of transport—maritime, air and road) from the 59 countries consid- ered are carried by liners. III. The Estimates We begin with ordinary least squares estimation of equation (5) over the entire dataset. The error term εijk is assumed to be independently distributed across exporting countries, but we allow for interdependence among observations within each country.30 The results are presented in table 2. Although the coefficients mostly accord with our expectations, this empiri- cal approach has a weakness: it ignores competition from alternative modes of transportation, expressly tramp maritime services (bulk and tanker), air trans- port, and road transport (in the case of Canada and Mexico). For a number of product categories, it is likely that shippers face an explicit tradeoff between the quality and cost of shipping a good by these alternative modes of trans- port. One approach to remedying this problem is to exclude all products for which competition from tramp maritime and air services is important. Since it is difficult to make a clean separation based on product characteristics alone, we adopt a method relying on the revealed importance of the alternative modes. Specifically, we exclude all observations where either the share of air trans- port as a percentage of total imports for shipping product k from country I to customs district J is positive, or the share of tramp services for a particular product k on all routes between country I and district J exceeds 15 percent. 29. However, if we exclude U.S. oil imports (HS category 27), this share rises to 70 percent and liner transport becomes relatively more important for developing countries. 30. Instead of using a fixed-effect specification as in equation (5), we also estimated a model with random product effects and maintaining the customs district fixed effects. This model yielded very similar estimation results. Moreover, the Hausman test rejected the null hypothesis that the individual effects are uncorrelated with our regressors in the model, supporting the use of fixed instead of random prod- uct effects. Fink, Mattoo, and Neagu 95 Table 2. Full-Sample Fixed-Effects Model Variable Estimate Distance 0.298** (4.97) Containerization –0.071** (–2.80) Total liner imports –0.017* (–2.07) Price-fixing agreements 0.488** (5.52) Cooperative agreements 0.050 (1.21) Cargo reservation –0.067 (–0.77) Cargo-handling services –0.203* (–2.44) Mandatory port services 0.357** (2.52) Number of products 4,356 Number of observations 250,237 F-statistic 65.11** Adjusted R2 0.775 *Significant at the 5 percent level. **Significant at the 1 percent level. Note: The dependent variable (liner transport prices), distance, and total liner imports are expressed in natural logs; all other variables are expressed in actual levels; fixed effects are product-specific and U.S. customs district–spe- cific (see text). The regression assumes an independently distributed error term across exporting countries, but al- lows for interdependence among observations within each country. t-statistics are in parentheses. The F-statistic tests the joint significance of all independent variables (except the fixed effects). Source: Authors’ calculations. This reduces our sample size from 250,237 to 98,997 observations. The esti- mation results with the reduced sample are presented in the first column of table 3. Although these results are in line with our expectations, it is possible that the exclusion of observations introduces a sample bias in our estimation. We there- fore adopt a sample selection model, where we estimate the likelihood of a ship- ment having no competition from air and tramp services (as defined above) in two separate probit equations. The explanatory variables in these probit equa- tions are (the natural logs of) the unit value and the unit weight of shipments and, in the case of air transport, a dummy variable that captures the existence of an open-skies agreement between country I and the United States.31 We estimate 31. Because the unit weight is unavailable for selected shipments, the number of observations in the probit regression is somewhat smaller than in the full sample. In the case of tramp services, we also included country fixed effects, except for Benin, for which the share of tramp services was below 15 percent for all observations. As in the liner pricing regression, we assumed that the error term in each probit equation is independently distributed across exporting countries, but allowed for interdepen- dence among observations within each country. 96 the world bank economic review, vol. 16, no. 1 this model using the Heckman two-step estimation procedure, assuming that the error terms in the two probit regressions are uncorrelated.32 The results of the sample selection model are presented in the second to fourth columns of table 3. In the “airâ€? probit equation, the estimated coefficient on unit value is significantly negative and the coefficient on unit weight is signifi- cantly positive, suggesting that valuable and light products are more likely to be sent by air. By contrast, in the “trampâ€? probit equation, the coefficient on unit value is significantly positive and the coefficient on unit weight is significantly negative, indicating that tramp services are primarily used for heavy commodi- ties with low unit values.33 In the final regression, we exclude Mexican and Canadian imports from our (already reduced) sample. For these two countries, road transport is an alterna- tive mode of transport that may compete with maritime and air services. Table 4 presents the estimation results with both the simple reduced sample and the sample selection approach, which are similar to those presented in table 3. Estimates of the Model Coefficients The results from the different estimating methods reveal a reassuring consistency. The estimated coefficient on distance lies between 0.2 and 0.3 and is always sig- nificantly different from both zero and one. This confirms that transport cost increases with distance but less than proportionately. As we expected, contain- erization, as measured by Tijk, works to reduce liner prices, the estimated coeffi- cient being statistically significant. The coefficient on the total value of U.S. imports carried by liners, qiJ, takes a small and significant negative value. This suggests that there are economies of scale with regard to traffic originating from the same port, and that small countries or economies with small trading volumes may be relatively disadvantaged. Consider now the impact of restrictions on trade in maritime services. The most striking finding is the strong positive impact on liner prices of the existence of rate-binding conference and other price-fixing agreements. The existence of cooperative working agreements has a weaker impact that is not always statis- tically significant. These results confirm our expectation that price-fixing agree- ments matter and are more important than cooperative working agreements.34 32. See Maddala (1983, p. 282) for a description of this model. The assumption that the error terms in the two probit regressions are uncorrelated seems reasonable: decisions on whether to ship goods by air or by vessel are likely to be independent of decisions on the mode of maritime transport. 33. Interestingly, the explanatory power is higher in the air probit regression than the tramp probit regression. 34. It is possible, in principle, that the formation of collusive carrier agreements is an endogenous variable—that collusion is more likely on more profitable routes. To account for this possibility, we estimated a treatment-effects model that corrects for the possible selectivity bias of the dummy variable on price-fixing carrier agreements (see Greene 1997, pp. 981–82, and Maddala 1983, p. 6). Similar to the sample selection model, we used the Heckman two-step estimation procedure that first estimates a probit model of the selection process and then the regression model with an additional selectivity cor- rection variable. Our explanatory variables in the probit model were exporter gdp, unit weight, unit Fink, Mattoo, and Neagu 97 The evidence on policy restrictions is mixed. The coefficient of the variable capturing the existence of cargo reservation policies is close to zero and not sta- tistically significant in any of the regressions. This result gives credence to the claim that cargo reservation policies no longer exert an important influence on liner trade. The estimated coefficient on the restrictiveness index of cargo-handling services is the only one that has a counterintuitive sign and is statistically signifi- cant in the first set of estimates (table 2), but with other, arguably more reliable methods (tables 3 and 4), it ceases to be significant. Recall that our dependent variable captures the cost of complementary services not explicitly but only to the extent that they feed through into the ocean-leg liner prices. In this respect, the index on the restrictiveness of port policy probably has a stronger claim to significance. Our estimates would seem to confirm this—the coefficient is con- sistently positive and statistically significant. This result also seems in line with current wisdom that the biggest policy hurdles to competitive provision of ship- ping services are to be found at the ports rather than in the ocean leg.35 How- ever, it must be kept in mind that we are only using an indirect measure of port policy restrictiveness. Estimates of the Consequences of Policy Changes The estimated model can be used to calculate hypothetical reductions in trans- port prices due to both the breakup of private carrier agreements and allowing greater competition in the provision of port services. For this purpose, we take the estimated coefficients from the sample selection model in table 3, which we consider to be the most reliable estimates both from an economic and econo- metric standpoint. Table 5 presents the simulated price reductions. The breakup of conference and other price-setting agreements would lead to a more dramatic reduction in transport prices (32 percent) than the breakup of cooperative work- ing agreements (18 percent), whereas the liberalization of port services would cause a 35 percent drop in the price of liner services.36 If we compute the trade-weighted percentage reductions in transport prices across all observations included in the sample selection model, the average total value, and total liner traffic between the exporting country and the importing coast district. Only ex- porter gdp made a positive and significant contribution to the likelihood of observing price-fixing agree- ments. The selectivity correction parameter, however, had a negative sign in the main regression and, accordingly, inflated the coefficient on the price-fixing dummy variable. This result would suggest that rate-binding carrier agreements typically occur on routes with lower prices. This counterintuitive find- ing may be due to the inadequacies of our explanatory variables in the probit equation. Alternatively, the formation of liner agreements may be less an outcome of market forces and more the result of his- torical and institutional forces, which are exogenously determined. 35. Of course, savings from the liberalization of port services are likely to be greater when their full impact on aggregate maritime transport costs is taken into account. 36. The policy simulation makes the simplifying assumption that the mandatory use of certain port services, such as pilotage, cannot be justified by safety or related concerns. This assumption does not seem unreasonable, given that safety standards can also be enforced in more liberal policy environments. Table 3. Reduced-Sample and Sample Selection Models Sample selection model Reduced-sample Variable model Air probit Tramp probit Liner transport prices Distance 0.202** (4.58) 0.228** (5.27) Containerization –0.132** (–3.89) –0.116** (–3.00) Total liner imports –0.018* (–2.39) –0.025** (–3.40) Price-fixing agreements 0.443** (5.78) 0.379** (5.01) Cooperative agreements 0.132* (2.64) 0.202* (2.53) Cargo reservation –1.106 (–1.13) –0.099 (–1.01) Cargo-handling services –0.104 (–1.14) –0.064 (–0.56) Mandatory port services 0.307** (2.26) 0.437** (2.83) Unit value –0.385** (–18.44) 0.130** (13.12) Unit weight 0.448** (17.49) –0.131** (–11.52) Open skies agreement 0.007 (0.07) 98 Sample selection correction (air) 0.399** (3.19) Sample selection correction (tramp) –0.733* (–2.37) Number of products 4,214 4,208 Number of observations 98,997 250,159 250,159 98,815 F-statistic 39.43** 42.51** Adjusted R2 0.779 0.783 Pseudo R2 0.130 0.054 *Significant at the 5 percent level. **Significant at the 1 percent level. Note: The dependent variable (liner transport prices), distance, total liner imports, unit value, and unit weight are expressed in natural logs; all other variables are expressed in actual levels; fixed effects are product-specific and U.S. customs district-specific (see text). All regressions assume an indepen- dently distributed error term across exporting countries but allow for interdependence among observations within each country. The sample selection correction variables are computed following Heckman’s two-step estimation procedure. t-statistics (for liner price regressions) and z-statistics (for probit regressions) are in parentheses. The F-statistic tests the joint significance of all independent variables (except the fixed effects). Source: Authors’ calculations. Table 4. Reduced-Sample and Sample Selection Models without Mexico and Canada Sample selection model Reduced-sample Variable model Air probit Tramp probit Liner transport prices Distance 0.221** (4.12) 0.215** (4.79) Containerization –0.147** (–4.49) –0.142** (–4.13) Total liner imports –0.017* (–2.19) –0.024** (–3.30) Price-fixing agreements 0.464** (5.65) 0.372** (4.60) Cooperative agreements 0.124* (2.58) 0.192* (2.46) Cargo reservation –0.092 (–0.95) –0.089 (–0.88) Cargo-handling services –0.107 (–1.17) –0.068 (–0.59) Mandatory port services 0.298** (2.24) 0.410** (2.69) Unit value –0.402** (–27.28) 0.131** (12.84) Unit weight 0.469** (27.56) –0.132** (–11.30) Open skies agreement 0.021 (0.22) 99 Sample selection correction (air) 0.465** (3.94) Sample selection correction (tramp) –0.694* (–2.26) Number of products 4,190 4,184 Number of observations 97,676 247,673 247,673 97,518 F-statistic 36.74** 50.83** Adjusted R2 0.781 0.784 Pseudo R2 0.136 0.054 *Significant at the 5 percent level. **Significant at the 1 percent level. Note: The dependent variable (liner transport prices), distance, total liner imports, unit value, and unit weight are expressed in natural logs; all other variables are expressed in actual levels; fixed effects are product-specific and U.S. customs district–specific (see text). All regressions assume an indepen- dently distributed error term across exporting countries but allow for interdependence among observations within each country. The sample selection correction variables are computed following Heckman’s two-step estimation procedure. t-statistics (for liner price regressions) and z-statistics (for probit regressions) are in parentheses. The F-statistic tests the joint significance of all independent variables (except the fixed effects). Source: Authors’ calculations. 100 Table 5. Simulated Reductions in Transport Prices Cumulative Breakup of effect of the cooperative Breakup of breakup of working price-fixing private carrier Liberalization Cumulative Simulation agreements agreements agreements of port services total effect 1. Percentage reductions on restricted routes 18.30 31.56 44.09 35.43 63.90 2. Trade-weighted percentage reductions across all observations in our dataset 7.11 18.71 24.25 8.66 30.76 3. Total savings across all observations in our dataset: 100 Absolute value(in million $) 140 371 484 201 637 As a percent of total transport chargesa 5.59 14.80 19.42 7.99 25.59 4. Projected total savings across all exporting countries the world bank economic review, vol. 16, no. 1 and all sectors (in million $)b 575.1 1,522.6 1,997.9 822.0 2,632.7 Note: These calculations are based on the estimated coefficients of the sample selection model in table 3. Given the functional form of the regression equation, the individual effects do not sum to the total effect. aThe share of total savings in total transport charges is equivalent to the unweighted average percentage reductions in transport prices. bThe projected total savings in the last row apply the percentage savings in total transport charges estimated for the reduced sample to total liner transport charges for all U.S. imports. Source: Authors’ calculations. Fink, Mattoo, and Neagu 101 reduction would be 30.8 percent—made up of the cumulative effects of the breakup of carrier agreements (24 percent) and the liberalization of port services (9 percent). Total savings would sum to $637 million of transport charges. To get a sense of the overall magnitudes involved, we can project these savings to total U.S. imports carried by liners across all sectors and all routes. Our simula- tions reveal that the removal of public restrictions to liner trade would lead to savings of up to $822 million and the breakup of private cartels would bring about additional savings of up to $2 billion. There are two important qualifications to these estimates. First, the pattern of restrictions in our limited sample may not be representative of the pattern of restrictions in trade of all products across all routes. Second, competition from other modes of transport for some products may limit the ability of carrier agree- ments to fix prices. But note that our simulation pertains to the savings arising from goods carried to the United States alone. The imports of the United States are only about a fifth of total world merchandise imports. So global gains from the elimination of all forms of restriction are likely to be substantially larger, particularly if we take into account the indirect benefits from reducing impedi- ments to trade. IV. Conclusion Our estimates confirm the general belief that cargo reservation policies, which proliferated in the 1970s and 1980s, are no longer an important barrier to trade. However, it emerged that both public policy, specifically in the form of restric- tions on the provision of port services, and private practices continue to exer- cise a significant influence on maritime transport prices. Interestingly, private anticompetitive practices have a stronger influence on prices than public re- strictions do. These results challenge the notion that collusive carrier arrangements have lost their significance over the past decade. In defense, maritime industry sources fre- quently point to the fact that liner operators hardly break even and, on this basis, argue that there is little scope for price reductions. But it is well known that pro- tection and cartel-like behavior in the presence of fixed costs can lead to ineffi- cient entry and reduced profitability. The benefits of competition typically arise not only from increased allocative efficiency—that is, pricing close to costs—but also from increased internal efficiency—that is, a reduction in costs. There may be scope for increasing this latter type of efficiency in the maritime industry. Our results need to be qualified. First, we focused only on routes leading to the United States. Although there is need for further research on other routes, the paucity of transport data in other countries is a major constraint. Second, the analysis herein has focused solely on the maritime leg of the transport journey and has not examined distortions on the inland section. Evidence suggests that the ocean leg accounts for a little more than a third of total door-to-door ship- ping charges (oecd 1968, Livingston 1986). Unfortunately, there are no com- 102 the world bank economic review, vol. 16, no. 1 prehensive data on such charges. An ambitious future research program would seek to disaggregate the components of door-to-door shipping charges and sub- ject them to an analysis similar to that carried out in here. A critical component of such a program would be to develop better measures of the restrictiveness of port and auxiliary services than have been used here. Notwithstanding these qualifications, this article has certain implications for policy. The elimination of policy restrictions to trade in maritime transport ser- vices is likely to produce substantial gains. Many of these restrictions can be removed unilaterally, and the gats can be used to bind the openness to reduce uncertainty and the possibility of policy reversals. But it is not enough to elimi- nate policy restrictions. There is also a need to deal with the private anticompetitive practices of international maritime cartels. Large states can probably tackle such practices unilaterally through their own competition laws, despite the extra- territoriality problems involved. But small states with limited enforcement capacity are at a disadvantage; the problem is accentuated by the fact that major trading countries have diluted the application of competition disciplines to the maritime sector. One positive development described earlier is the elimination by the United States of some of the provisions in its shipping law that helped police price-fixing arrangements. Whether collusion can be sustained in the ab- sence of such facilitating devices is open to question. But we would argue that there is cause for concern as long as the basic rate-setting conference system continues to enjoy antitrust immunity. An international initiative would seem desirable. One approach would be to deal with the problem by creating sector-specific competition rules, as in the case of basic telecommunications. Or, if such anticompetitive practices also affect other services sectors, there may be a need to strengthen the general gats disciplines. Currently, Article IX of the gats (which deals with private anticompetitive prac- tices) has little substance, providing only for an exchange of information and consultations. The current round of services negotiations offers an opportunity to strengthen this provision. What form could such a strengthening take? We believe that the harmoni- zation of either sector-specific or general competition rules is probably neither feasible nor necessary. Our proposal is much simpler and would involve the cre- ation of two obligations. The first would end the exemption of collusive agree- ments in the maritime sector from national competition law. The second would create the right of foreign consumers to challenge anticompetitive practices by shipping lines in the national courts of countries whose citizens own or control these shipping lines. The second obligation is necessary to deal with a possible failure to enforce and already has a precedent in the wto rules on intellectual property and government procurement.37 Would it be feasible to create such rules? History does not provide cause for optimism. The procompetitive rules in basic telecommunications, in line with 37. See Mattoo and Subramanian (1997) for an elaboration of this argument. Fink, Mattoo, and Neagu 103 most wto rules, were designed to protect the market access rights of foreign suppliers, and conventional political-economy forces supported their creation. To establish rules that enable small countries to protect their consumers from foreign oligopolies will be far more difficult. In fact, the negotiating history of the gats reveals successful opposition to the strengthening of Article IX from some of the countries that exempt maritime conferences from the scope of their antitrust laws. However, the reluctance of many developing countries to make liberalization commitments under the gats did not strengthen their case. One strategy in the current round of services negotiations would be for a coalition of developing coun- tries to put forward an offer of substantial liberalization conditional on the strength- ening of Article IX. By targeting the twin maladies of maritime trade, such a strategy, if successful, would provide substantial global benefits. Appendix: Data Data on liner transport charges, import values, the percentage of containerized cargo, total imports carried by liners and the market share of tramp services are from the Waterborne Trade Database compiled by the U.S. Department of Trans- portation. The containerization variable is measured in terms of the weight of goods shipped. Tramp services are defined as bulk and tanker services. Unit values, unit weights, and the market share of air services are computed from the U.S. Merchandise Imports Database published by the U.S. Department of Com- merce. This source does not publish data separately by foreign and U.S. ports; we therefore have to use these variables at the more aggregate level, that is, U.S. trading partners and U.S. customs districts. Shipping distances were kindly provided from a private service called BP Marine. Some missing ports that are included in the Waterborne Transport Database had to be approximated by the closest neighboring port. Information on private carrier agreements between U.S. coastal districts and individual coun- tries comes from the Federal Maritime Commission (1998). We excluded agree- ments signed before 1970 and also those with an unspecified regional coverage (for example, the Far East), because the de facto coverage of such agreements may only relate to a few particular routes. The potential bias introduced by this exercise is likely to be small because most routes covered by such regional agree- ments are also covered by country-specific agreements. As mentioned in the text, we construct two dummy variables to account for the presence of carrier agree- ments on maritime routes. The first refers to conferences and other price-fixing agreements and the second captures cooperative working agreements that do not have a binding rate authority. Data on the existence of open-skies agreements were taken from the Web site of the U.S. Department of Transportation. The three indicators of trade restrictions are constructed based on informa- tion compiled from the following sources: wto (1994), various wto Trade Policy Reviews, gats schedules of commitments (available online at http://gats- info.eu.int/index.html), apec Individual Action Plan submissions (available 104 the world bank economic review, vol. 16, no. 1 online at http://www.apecsec.org), unpublished oecd documents, eclac (1999), eu Market Access Database (available online at http://mkaccdb.eu.int), and various editions of the National Trade Estimate Report on Foreign Trade Barriers com- piled by the U.S. Trade Representative (available online at http://www.ustr.gov). In some cases, Greg McGuire kindly supplied data from the above sources. The cargo reservation dummy variable is assigned a value of 1 if a country has a bilateral agreement involving cargo sharing with the United States, if it is a signatory of the U.N. Code of Conduct for Liner Conferences and applies Article 2 of the code in its trade with the United States, or if it sustains any kind of unilateral cargo reservation scheme; and 0 otherwise. The cargo-handling ser- vices index measures restrictions or special requirements imposed in a country to potential foreign suppliers of cargo-handling services (foreign suppliers means, in this case, locally registered companies with foreign participation in their capi- tal or branches of firms established in other countries). The index values are 0 if there is no restriction, 0.25 if minor restrictions exist, 0.5 if a joint venture con- dition is imposed, 0.75 if a very high national participation in the capital of the company is required, and 1 if foreign companies are not allowed to provide cargo- handling services at all. In selected countries, consultations with industry experts suggested a slightly different assessment of policy restrictiveness in cargo han- dling than the one implied by the sources listed above. But a modification of the rankings did not lead to substantial changes in our empirical findings. The index on mandatory port services assigns a score of 0.125 for the existence of each of the following mandatory services: pilotage, towing, tug assistance, navigation aids, berthing, waste disposal, anchorage, and other mandatory services. Table A-1 lists the countries for which we could find information on the three policy indicators and that are included in our estimation set. The table also shows the assigned values of these policy variables as well as the average value of the dummy capturing the two types of collusive carrier agreements (the latter lying between 0 and 1, if not all U.S. coastal districts are covered by the agreements affecting a particular country). Fink, Mattoo, and Neagu 105 Table A-1. Indicators of Maritime Policy and Carrier Agreements, 59 Countries Price-fixing Cooperative Cargo Cargo-handling Mandatory carrier working Country reservation services port services agreements agreements Argentina 0 0 0.13 0.00 1.00 Australia 0 0 0.13 1.00 1.00 Belgium 0 0 0.06 1.00 0.00 Benin 1 1 0.00 0.00 0.00 Brazil 1 0.5 0.75 0.00 1.00 Brunei 0 0 0.00 0.00 0.00 Canada 0 0 0.13 0.00 0.00 Chile 0 0 0.25 0.43 1.00 China 1 0.5 0.00 0.00 0.00 Colombia 0 0.5 0.13 0.50 1.00 Costa Rica 0 0 0.00 0.00 1.00 Côte d’Ivoire 0 0 0.25 0.00 1.00 Cyprus 0 1 0.31 0.00 0.00 Denmark 0 0 0.06 1.00 0.00 Dominican Republic 0 0.25 0.25 0.50 1.00 Ecuador 0 0 0.00 0.43 1.00 Egypt 1 0.75 0.75 0.00 0.00 El Salvador 0 0 0.00 0.00 1.00 Finland 0 0 0.25 0.00 0.00 France 0 0 0.38 1.00 0.00 Germany 0 0 0.38 1.00 0.00 Ghana 1 1 0.50 0.00 1.00 Greece 0 1 0.19 0.00 0.00 Hong Kong 0 0 0.25 0.00 0.00 Iceland 0 0 0.13 0.00 0.00 India 1 0 0.00 0.00 1.00 Indonesia 0 1 0.06 0.00 0.38 Ireland 0 0 0.13 1.00 0.00 Italy 0 0.25 0.50 0.38 0.00 Jamaica 0 0.5 0.00 0.00 0.60 Japan 0 0.75 0.13 0.89 1.00 Korea, Rep. of 0 0 0.38 0.00 0.00 Malaysia 0 0 0.25 0.00 0.38 Mauritius 0 1 0.38 0.00 0.00 Mexico 0 0.5 0.38 0.00 1.00 Morocco 1 0.5 0.13 0.00 0.00 Netherlands 0 0 0.50 1.00 0.00 New Zealand 0 0 0.38 1.00 1.00 Nicaragua 1 0 0.00 0.00 1.00 Nigeria 1 0 0.50 0.00 1.00 Papua New Guinea 0 0.5 0.00 0.00 0.00 Peru 0 0.5 0.00 0.50 1.00 Philippines 0 0.5 0.00 0.00 0.38 Poland 0 0.25 0.00 0.00 0.00 Portugal 0 0 0.13 1.00 0.00 Romania 0 0 0.63 0.00 0.00 Senegal 0 0 0.00 0.00 1.00 Singapore 0 1 0.38 0.00 0.33 (continued) 106 the world bank economic review, vol. 16, no. 1 Table A-1. (continued) Price-fixing Cooperative Cargo Cargo-handling Mandatory carrier working Country reservation services port services agreements agreements Spain 0 0 0.06 1.00 0.00 Sweden 0 0 0.06 1.00 0.00 Taiwan 0 0.5 0.00 0.00 0.00 Thailand 0 0.5 0.63 0.00 0.38 Togo 1 0 0.00 0.00 0.00 Tunisia 0 0.5 0.13 0.00 0.00 Turkey 0 0 0.00 0.43 0.00 United Kingdom 0 0 0.31 1.00 0.00 Uruguay 0 0 0.00 0.00 1.00 Venezuela 1 0 0.00 1.00 1.00 Vietnam 0 0 0.00 0.00 0.50 Note: The indicators on “price-fixing carrier agreementsâ€? and “cooperative working agreementsâ€? show the average value of the 0-1 dummy variable used in the estimation. This value lies between 0 and 1 if not all U.S. coastal districts are covered by the agreements affecting a particular country. Source: Authors’ calculations. References Amjadi, Azita, and Alexander Yeats. 1995. “Have Transport Costs Contributed to the Relative Decline of Sub-Saharan African Exports? â€? Policy Research Working Paper No. 1559, World Bank, Washington, D.C. Bennathan, Ezra. 1989. “Deregulation of Shipping: What Is to Be Learned from Chile?â€? World Bank Discussion Papers No. 67, World Bank, Washington, D.C. Clyde, Paul S., and James D. Reitzes. 1995. “The Effectiveness of Collusion under Anti- trust Immunity, the Case of Liner Shipping Conferences.â€? Federal Trade Commission, December. Conlon, R. M. 1982. “Transport Cost and Tariff Protection of Australian Manufactur- ing.â€? Economic Record 58(160):73–81. Davies, J. E. 1986. “Competition, Contestability and the Liner Shipping Industry.â€? Journal of Transport Economics and Policy 2(3):299–312. eclac (U.N. Economic Commission for Latin America and the Caribbean). 1999. “In- ventory of Measures Affecting Trade in Services.â€? Manuscript, United Nations, New York. Federal Maritime Commission. 1998. “Carrier Agreements in the U.S. Oceanborne Trades.â€? Bureau of Economics and Agreement Analysis, Washington, D.C. Finger, J. M., and Alexander Yeats. 1976. “Effective Protection by Transportation Costs and Tariffs: A Comparison of Magnitudes.â€? Quarterly Journal of Economics 90(1):169–76. Foxley, Juan, and José Luis Mardones. 2000. “Port Concessions in Chile: Contract De- sign to Promote Competition and Investment.â€? Viewpoint: Public Policy for the Pri- vate Sector, Note No. 223, World Bank, Washington, D.C. Francois, Joseph, and Ian Wooton. 1999. “Trade in International Transport Service: Fink, Mattoo, and Neagu 107 The Role of Competition.â€? Centre for Economic Policy Research Discussion Paper No. 2377, Centre for Economic Policy Research, London. Francois, Joseph, Hugh Arce, Kenneth Reinert, and Joseph Flynn. 1996. “Commercial Policy and the Domestic Carrying Trade: A General Equilibrium Assessment of the Jones Act.â€? Canadian Journal of Economics 29(1):181–98. Franck, Bernard, and Jean-Claude Bunel. 1991. “Contestability, Competition and Regu- lation, the Case of Liner Shipping.â€? International Journal of Industrial Organization 9(1)141–59. Greene, William H. 1997. Econometric Analysis, 3d ed. Upper Saddle River, N.J: Prentice Hall. Jankowski, W. B. 1989. “Competition, Contestability and the Liner Shipping Industry.â€? Journal of Transport Economics and Policy 23(2):199–203. Hummels, David. 1999. “Have International Transportation Costs Declined?â€? Manu- script, University of Chicago, Chicago, Ill. Kang, Jong Soon. 2000. “Price Impact of Restrictions on Maritime Transport Services,â€? in C. Findlay and T. Warren, eds., Impediments to Trade in Services: Measurement and Policy Implications. London: Routledge. Levitt, M. 2000. “The Treatment of Liner Shipping Under eu and us Law: The Transat- lantic Conference Agreement,â€? in S. J. Evenett, A. Lehmann, and B. Steil (eds), Anti- trust Goes Global. Washington, D.C.: Brookings Institution Press. Livingston, Ian. 1986. International Transport Costs and Industrial Development in the Least Developed Countries. unido/IS.616, Vienna: unido. Maddala, G. S. 1983. Limited Dependent and Qualitative Variables in Econometrics. Cambridge, Mass.: Cambridge University Press. Mattoo, A., and A. Subramanian. 1997. “Multilateral Rules on Competition Policy: A Possible Way Forward.â€? Journal of World Trade 31(5):95–115. McGuire, Greg, Michael Schuele, and Tina Smith. 2000. “Restrictiveness of International Trade in Maritime Services,â€? in C. Findlay and T. Warren, eds., Impediments to Trade in Services: Measurement and Policy Implications. London: Routledge. Obstfeld, M., and K. Rogoff. 2000. “The Six Major Puzzles in International Macroeco- nomics: Is There a Common Cause?â€? nber Working Paper No. 7777, National Bu- reau for Economic Research, Boston. oecd. 1968. Ocean Freight Rates as Part of Total Transport Costs. Paris: oecd. Pálsson, Gylfi. 1997. “Containerized Maritime Trade between West-Africa and Europe: Multiple Ports of Call versus Hub-and-Spoke.â€? Mimeo, World Bank, Washington, D.C. Pearson, Roy. 1987. “Some Doubts on the Contestability of Liner Shipping Markets.â€? Maritime Policy Management 1:71–78. Sampson, G. P., and A. J. Yeats. 1977. “Tariff and Transport Barriers Facing Australian Exports.â€? Journal of Transport Economics and Policy 11(2):141–54. Trujillo, L., and G. Nombela. 1999. “Privatization and Regulation in the Seaport Indus- try.â€? Policy Research Working Paper No. 2181, World Bank, Washington, D.C. Trujillo, Lourdes, and Antonio Estache. 2001. “Surfing a Wave of Fine Tuning Reforms in Argentina’s Ports.â€? Mimeo, World Bank, Washington, D.C. U.S. International Trade Commission. 1991. “The Economic Effects of Significant U.S. Import Restraints, Phase III: Services.â€? usitc Publication 2442, Washington, D.C. Venables, A. J., and N. Limao. 1999. “Geographical Disadvantage: A Heckscher-Ohlin- 108 the world bank economic review, vol. 16, no. 1 von Thunen Model of International Specialization.â€? Policy Research Working Paper No. 2256, World Bank, Washington, D.C. Waters, W. G. 1970. “Transport Costs, Tariffs, and the Patterns of Industrial Protec- tion.â€? American Economic Review 60(5):1013–20. White, Lawrence J. 1988. International Trade in Ocean Shipping Services. Cambridge, Mass.: Ballinger Publications. World Bank. 2000. “Module 2: The Evolution of Ports in a Competitive Worlds, Mod- ule 2.â€? From the World Bank Port Reform Tool Kit, World Bank, Washington, D.C. wto . 1994. “Questionnaire on Maritime Transport Services.â€? Negotiating Group on Maritime Transport Services (s/ngmts/w/2), World Trade Organization, Geneva. ———. 1998. “Maritime Transport Services.â€? Background Note by the Secretariat (s/c/ w/62). World Trade Organization, Geneva. Zerby, J. A. 1988. “Clarifying Some Issues Relating to Contestability in Liner Shipping and Perhaps Also Eliminating Some Doubts.â€? Maritime Policy and Management 15(1):5–14.