Commissioner

The Effectiveness of Under Antitrust ImmunityCollusionThe Case of Liner Shipping ConferencesPaul S. ClydeJames D. ReitzesFEDERAL TRADE COMMISSIONDecember 1995The Effectiveness of Under Antitrust ImmunityCollusionThe Case of Liner Silipping ConferencesPaul S. ClydeJames D. ReitzesBureau of Economics Staff ReportFederal Trade CommissionWashington, D.C. 20580December 1995FEDERAL TRADE COMl\\fiSSIONROBERT PITOFSKYChairmanMARY L. AZCUENAGACommissionerJANET D. STEIGERCOImnissionerROSCOE B. STAREK, IIICommissionerCHRISTINE A. VARNEYCommissionerBUREAU OF ECONOMICSJONATHAN B. BAKERDirectorRONALD S. BONDAssociate Director for Operations andPolicyPAUL A. PAUTLERAssociate Director for ConsumerProtection and ResearchGARY L. ROBERTSAssociate Director for AntitJ\"1JstDENIS A. BREENAssistant Director for Economic PolicyAnalysisROBERT D. BROGANAssistant Director for AntitrustGERARD R. BUTIERSAssistant Director for ConsumerProtectionTIMOTHY P. DANIELAssistant Director for AntitrustTIMOTHY A. DEY AI(Assistant Director for AntitrustThe authors are economists with the Antitrust DivisionDepartment of Justice and Ernst and Young, respectively. This studyexpands on work performed while the authors were staff economistswith the Advisory Commission on Conferences in Ocean Shipping(ACCOS). ACCOS released its report in Apri11992. Dr. Reitzesalso was an author of \"An Analysis of the Maritime Industry and theEffects of the 1984 Shipping Act,Report of the Federal TradeCommission to Congress and ACCOS (November 19).The views expressed in this study are those of the authors anddo not necessarily reflect the views of the Department of Justice, theFederal Trade Commission, or any individual Commissioner.AcknowledgementsWe wish to thank Tim Daniel for his unflagging support of thisproject and numerous editorial and conceptual contributions. Also, wethank Paul Pautler for both support and comments, as well as RonBond, Curt Jernigan, and John Peterman. This study benefittedgreatly from the suggestions of Dave Butz, Mike Metzger, GregWerden, John Hilke, Mike Ward, and Bob Rogers. We appreciate thestatistical guidance supplied by Luke Froeb and the useful insights intoliner conferences and maritime regulation provided by Allen JacksonRichard Speigel, and Sandy Kusumoto.SummaryCollusion, in the form of tacit or explicit price coordination, isa subject that has fascinated both policymakers and economists formany years. While collusion can certainly stem from formal price-fixing or market-sharing agreements, economists also have noted thattacit collusion can be achieved in a noncooperative setting throughrepeated contact. Chamberlin (1929), Stigler (19), Orr andMacA voy (1965), and others have identified several factors thatdetermine whether collusion can be sustained; chief among thesefactors are the ability of the cartel to identify and punish defectionfrom the collusive outcome, and the ability of the cartel to prevententry. 1 This study analyzes data from the ocean shipping industry toexamine empirically whether the provision of antitrust immunity forprice-fixing agreements and the establishment of a costless mechanismfor policing these agreements are sufficient to foster effective collusionwhen there are no regulatory restrictions on entry.This study analyzes whether liner conferences in oceanshipping act as effective cartels by testing empirically whether the ratestructure in ocean shipping is consistent with cost-based factors, theexercise of market power by conferencespower by firms in a manner , or the exercise of marketunrelated to the conference system. find no statistically significant relationship between freight rates andthe market share of the conference serving the route, which indicatesthat conferences do not act as perfect cartels maximizing the jointprofits of their members. Nonetheless, we do find that the level offreight rates is significantly lower on routes where conferencemembers are free to negotiate service contracts directly with shippers.In the routes included in this study, such freedom existed temporarilyon two routes, both of which had conferences with market shares over80 % . This latter finding provides some support for the conclusionthat some aspects of the conference system may contribute to highershipping rates, particularly when the conference has a sizable marketshare. Finally, we find that increases in market concentration are1 See also Green and Porter (1984), Abreu, Pearce, and Stachetti(1985), Rotemberg and Saloner (1986).associated with statistically significant, but economically smallincreases in freight rates.IntroductionPolicy concerns regarding possible collusion have beenreflected in several legislative and administrative actions, beginningwith the Sherman Antitrust Act's prohibition of any attempt tocombine or conspire ... to monopolize any part of the trade orcommerce among the several States.... \"2 These concerns are stillevident today, as a glance at the recent Department of Justice andFederal Trade Commission Horizontal Merger Guidelines (1992) willconfirm. Economists, however, have provided relativelylittleempirical support for the existence of anticompetitive collusivebehavior in particular industries. The development of such supporthas been :~,ymied by the paucity of detailed firm and industry cost anddemand data needed to assess empirically firm-specific or industry-specific conduct. 3Determining the effects of collusion would be greatlysimplified if firms merely announced to the general public when theywere colluding. This does not happen in most markets becausecollusive pricing behavior typically violates the antitrust laws. Thereare exceptions, though. The prime example is in certaintransportation industries where firms are expressly exempt from theantitrust laws. These industries provide an excellent opportunity forstudying the effectiveness of collusion in an environment where it isconsidered legal.2 Section 2 of the Sherman Antitrust Act (10).3 Ideally, cost and demand data. can be used to estimate eitherfirm conjectures or a parameter that measures the collusiveness ofindustry conduct. For a detailed discussion, see Bresnahans (19)chapter in The Handbook of Industrial Organization.One industry in particularindustry, provides a fertile area for research into the , the international ocean shippingcollusive behavior. In most of the worldeffectiveness ofprice-fiXing agreements that are exeJnpt from , ocean carriers can enter intolitigation. This exemption lowers the potential costs of entering intoantitrust scrutiny andcollusive arrangements. Hence, price-fixing agreements among oceancarriers, blown as conference agreementsare widespread. 4On international routes involving the United Statesconferences enjoy another significant advantage. S. government, the Federal Maritime Commission (hereafterAn agency of theFMC), polices conference pricing agreements at no cost to the cartel.The participants in a conference agreement must collectively file theirfreight rates with the FMCinspection. Any , and those rates are open to publicsecretive discounting on those rates is consideredrebating, by the FMC. The \" and a carrier involved in rebating is subject to a stiff fineFMC devotes a significant portion of its resourcesto investigating alleged rebating activity, and fining .parties.any guiltyMembers of conferences serving U. S. trade routes arehowever, allowed to deviate publicly from the conference rate bytaking independent actionthat differs from the conference rate. (hereafter, IA) and offering a freight rateprovide notification to the conferenceConference members must, howeveran IA rate. Moreover, since an IA rate is publicly available, of their intent to offerbe offered to all qualified , it mustshippers (although the qualified\" may, in fact, be quite small). Given groupillegal and inhibited by the actions of an outside agency, and thatthat secret rate discounting isIn international shipping lanes, these agreements have existedsince the 1870's (see Herman, 1983).5 In the period from 1985 through the middle of 19assessed over $12 million in penalties for rebating and other violations, the FMCof tariff-filing provisions.public rate deviations are immediately recognizablethat it would be difficult to cheat profitably on a conference pricing, one might suspectagreement under the current regulatory structure.Nevertheless, effective collusion in U.S. international oceanshipping may still be hindered by the ability of other carriers to enterthe market6 the heterogeneity of carriers in the market7 and thepresence of large customers. 8 Ocean carriers tend to beheterogeneous for a variety of reasons. In many trades, the majorcarriers have registered their vessels under different national flags.This, by itself, may lead to a divergence in cost conditions since thetreatment of ocean carriers differs widely across nations with regard tosubsidization and tax policy, crewing requirements, and other factors.These differences may supplement competitive differences stemming6 The existing regulatory systenl for ocean shipping, unlike thoseof domestic aviation and trucking prior to their deregulation, allowsany carrier to enter the market.7 When cartel members vary widely in their efficiency, a cartelmay encounter difficulty in maintaining a collusive price. The more-efficient members receive relatively larger gains from cheating on agiven cartel price, since their lower cost structure allows them to profitably expand output by a larger amount. This incentive mayconstrain the sustainable price level for the cartel, Le., the maximumprice level where firms no longer have an incentive to deviate fromthe cartel agreement. For further discussion, see Radner (1977),Osborne and Pitchik (1983), Choi, Menezes, and Tressler (1985),Schmalensee (1987), and Harrington (1991).8 When customers are heterogenous, cartel members may betempted to discount selectively to those customers making relativelylarge purchases. . This type of cheating may be more difficult to detectand punish than a situation where finns can significantly expand salesonly by offering widespread discounts. Thus, the presence of bigbuyers may hold down the cartel's price.from variations across firms in technology and entrepreneurial ability.Customers differ based on the size of their shipments and the qualityof service desired. Purchasers of ocean transportation services (Le.shippers) range from small manufacturers and importers to Fortune500 corporations.TUs study investigates whether the granting of antitrustimmunity and the use of the FMC to enforce pricing sufficient for liner conferences to exercise market power in the U.agreements areinternational ocean shipping market. Alternatively, this analysis couldbe considered a case study of whether the possibility of entry and theexistence of heterogeneity across fIm1s thwart effective collusion. If fIrms are not able to form an and customers is sufficient tocartel under the relatively favorable conditions described aboveeffectivemight expect that it would be difficult to form an effective cartel in, onemany markets.Controlling for other factorsthe market share of a liner conference, we test whether an increase inadditional firms join that conference, which presumably occurs when, is associated with a change in the.structure of freight rates. To our knowledge, previous empiricalstudies of liner shipping have not attempted this type of test.9 Manyof these studies argue that cartel market power exists based on theirfinding that freight rates increase as the value of the cargo increases.Given that ocean shipping is now widely containerizedcommodities are typically shipped in uniform boxes that are handled in, so that diversea uniform manner, these value-based. rate differences are considered9 Empirical research on the structure of freight rates Heaver (1973), Bryan (1974), Jansson (1974), Devanney, Livanosincludesand Stewart (1975), Zerby and Conlon (1983), and Jansson andShneerson (1986).evidence of price discrimination (Le.value of service II pricing)resulting from conference market power. Others note that value-based rate differences may still reflectcost differences related to the value of the commodity, such asdifferences in damage liability and service quality.11 Thusare positively correlated with the value of the commodity do not, rates thatnecessarily indicate that conferences have market power. Instead, thistype of rate structure may represent an efficient outcome generated bya competitive market.Another explanation is also possible. Value-based ratedifferences may result from market power that arises independently ofthe conference system. In particular, these rate differences may stemfrom the type of pricing behavior that emerges from firm interactionin a concentrated market. 12 The presence of scale and network10 ThIS claim is based on a Hicksian derived-demand argumentwhere transportation services represent an input in the production of adelivered\" product. When passed through to consumers in entirety,a given increase in the freight rate will lead to a progreSsively smallerpercentage increase in delivered prke as the value of the transportedproduct increases. Thus, if the market elasticity of demand is similarfor a group of products, one would expect that the derived elasticity ofdemand for transportation services would be lower for relatively high-valued products. Recognizing this n~lationship, a conference withmarket power would generally charge higher freight rates for higher-valued commodities (unless, contr~! to most empirical evidencehigher-valued commodities typically face a more elastic marketdemand than lower-valued commodities).11 See Sjostrom (1992) and Butz (1993).12 The exercise of firm-based market power, however, does notimply that firms necessarily earn supranormal profits. The presence(continued.. .(.. .economies may limit the number of sustainable fIrmS on any routethus allowing individual fums to exercise some power over price. Itis also possible that the regulatory structure in ocean shipping,particularly mandatory tariff ruing and enforcement for all carriersfacilitates anticompetitive interaction between carriers inside andoutside of liner conferences. In fact, the distinction betweenconferencr members and outsidecarriers may be limited, since allcarriers must make their freight rates publicly available andconference carriers either can set their own freight rates (through theindependent actionadvanced notice. 13 Freight rates may thus depend more on overallprovision) or exit the conference with littlemarket concentration than on conference market share.We try to contribute to the debate surrounding market powerin liner shipping. If, ceteris paribus, a conferences freight rates on agiven route generally increase as the conference covering that routeincreases its market share, then this outcome is consistent with thenotion that liner conferences exercisl~ market power by themselves.This result could be explained on an efficiency basis only if some costvariable can be identified that is positively correlated with conferencecontinued)of sunk costs could limit the number of firms in equilibrium, witheach of these firms sufficiently large to realize some degree of marketpower under certain types of strategjc behavior. Price discriminationmay arise consequently, but the abiHty. of firms to enter the marketdrives profits to a normal level. In this case, the equilibriumconfiguration would represent a \"quasi\" Ramsey-pricing outcome.For discussion of price discrimination under imperfect competition. see Katz (1984), Borenstein (1985), and Holmes (19).13 Butz (1993) makes a similar point.market share. At best, this is an unlikely proposition.14 In addition,if an increase in a conferences market share is associated with anincrease in the dispersion of the conferences freight rates based oncommodity values, then this result is consistent with the hypothesis thatlarger liner conferences can better discriminate in their pricing on thebasis of commodity value.Our statistical specification measures the effect of a change a given conferences market share on the level of freight rates and thedispersion of those rates on the route covered by the conference. this fashion, we estimate the degree of market power possessed byliner conferences. Additionally, we explore the relationship betweenthe freight rates on a given route and an index of market -concentrationfor that route. This analysis sheds 1 ight on the importance of firm14 An exception might occur if the conferences, themselvesoffered transportation services to the customer. An increase inconference market share might then be associated with higher quality,and possibly more costly, transportation services. for example, if conferences with higher market shares offered moreThis might occurfrequent sailings. Due to the higher-quality service, consumers wouldbe charged higher freight rates as conference size increased.Nonetheless, consumers might still benefit on balance if there weresubstantial improvement in service quality.Howeverwith the exception. of contract servicesno shippingservices are offered by the conference. Instead, individual memberssell their own shipping services to customers. Customers wouldgenerally be expected to judge quality on the basis of the serviceoffered by the carrier instead of the conference. Because our samplespecifically excludes freight rates for contract services, we believe thatdata are free of any methodological problems that would have resultedfrom the inclusion of conference-based shipping services.'-....size, as opposed to conference size, in explaining the rate structure inocean shipping. This study is organized as follows. Section II describes thehistory and current status of U.S. regulation of international linershipping. Section III provides a brief description of our theoreticalmodel, and how we implement it empirically. Section IV contains thetheoretical model that underlies our statistical specification, which isIS Finally, we should mention that one strand of the economicliterature on ocean shipping contends that cooperation among oceancarriers may be necessary to alleviate the problems associated with anempty core(see Sjostrom, 19, and Pirrong, 1992). Under theempty-core hypothesis (see Telser, 1978), the presence of avoitklblecosts and lumpy production preclude the existence of a noncooperativeequilibrium. As applied to ocean shipping, this theory wouldpotentially justify the formation of liner conferences as a means ofimposing stability in the market. Our analysis does not explicitly testthe empty core - hypothesis; insteadbehavior of liner conferences. Since core theory , we focus on the rate-settingwhich industry structure is likely to emergedoes not predict, and offers little insightconcerning the rate behavior under that structurebe used to either affirm or refute the empty-core , our results cannothypothesis.Core theory does predict that unless a coalition of firms (orfirms and customers) forms to impose a sustainable outcomedestructive competition - will emerge with'large swings in rates.Thus, if conferences form to alleviate an empty-core problemmight observe more stable rate behavior than in the absence of, oneconferences. However, it may also hold that in the face of demanduncertainty, a cartel. acting solely for anticompetitive gain will exhibitmore stable rate behavior than a competitive market. Notwithstandingthis problem of interpretation, conferences are a longstandinginstitution, which prevents the assembling of a data set that containsperiods where conferences are absent. Such data would be potentiallyuseful in assessing the relationship between conferences and the stability of freight rates.described in Section V. Section VI describes the data used in thisstudy. Section VII examines the statistical results, and Section VIIIoffers concluding remarks.ll.History and Current Status of U.International Liner ShippingS. Regulation ofConferences in ocean shipping have existed since the late1800's on U.S. international routes. Since then, they have beengranted varying degreesof antitrust immunity under the auspices ofthe Shipping Act of 1916, the 1961 Amendments, and the ShippingAct of 1984.The 1916 Shipping Act allowed ocean carriers to enter intoprice-setting agreements that were immune from antitrust action, butthose agreements had to be filed and approved by the U.S. ShippingBoard. The 1961 Amendments weakened this antitrust immunity byincluding a \"public intereststandard in the approval process forconference agreements conducted by the U.S. Shipping Board'successor, the Federal Maritime Commission. The FMC wasauthorized to disapprove rates that were so \"unreasonably high or lowas to be detrimental to the commerce of the United States. This stricter approval criteria for conference agreements wasgiven practical meaning by the U.S. Supreme Court'Svenskadecision in 196817 which upheld the FMC's ruling that the publicinterest\" standard created the presumption that any conferencerestraint was invalid if it interfered with the policies of U.S. antitrust16 This description of the regulatory structure and marketcharacteristics of U.S. international liner shipping is derived from theReport of the Advisory Commission on Conferences in Ocean Shipping(April 1992), Chapters 2 and 3.17 See Federal Maritime Commission v. Aktiebolaget Svenska Amerika Linien390 U.S. 238 (1968).laws.18 Carriers claimed that this standard impeded the formation ofratemaking, joint venture, service rationalization, and other types ofagreements.By eliminating the public interest\" standard, the Shipping Actof 1984 reversed the trend toward greater antitrust scrutiny. The 1984Act shifted the burden of proof so that conference agreements were nolonger subject to an approval process, but instead could be contestedby the F~C. Agreements automatically became effective after 45days urnes3 the FMC sought an injunction on the basis that theagreement was \"likely, by a reduction in competition, to produce anunreasonable reduction in transportation service or an unreasonableincrease in transportation cost. \"19 To date, the FMC haS neverattempted to enjoin a conference agreement, although it occasionallyhas recommended\" changes in the language of an agreement.The 1984 Shipping Act continued the tariff filing and enforcement provisions established by the 1961 Amendments. Theseprovisions require all ocean carriers and conferences to file their rateswith the FMC and publish their rate and schedule information. TheFMC was authorized to enforce that the filed rates were actuallycharged; any secretive discounting on a published rate was consideredillegal and subject to punitive action (Le.fmes) by the FMC. Thetariff filing and enforcement provisions were intended to maintaina nondiscriminatory regulatory process for the common carriage goods by water.... \"20 Apparently, Congress was concerned that18 An exception occurred if the conference could establish a primafade case that the restraint was required by \"a serious transportationneed, necessary to secure important public benefits, or in furtheranceof a valid regulatory purpose of the Shipping Act. 19 See Section 6(g) of the Shipping Act of 1984.20 Section 2 of the Shipping Act of 1984.large shippers may be successful in obtaining lower rates than smallshippers due to a superior bargaining position with carriers.Given that the tariff structure in liner shipping permitsdifferences in rates based on the type of cargo and the volume ofshipment, the tariff filing requirement does not prevent effective pricediscrimination. However, the FMC's role in enforcing filed ratesimplies that conference agreements are policed by an outside agency atno cost to the conference. Thus, the tariff filing and enforcementrequirements potentially facilitate the exercise of conference marketpower.The Shipping Act of 1984 required that conferences be open;any carrier can join or exit a conference agreement with limitedadvance notice and without penalty. There were also no statutoryrestrictions on the ability of carriers to enter any given route (either asan independent carrier or as a member of a conference). Moreoverthe 1984 Act mandated that any conference pricing agreement mustallow meMbers the right of independent action.This right allowsany conference member to offer a rate that differs from the conferencerate, but it must notify the conference in advance (usually ten daysprior) of its intention to do so. In the post-1984 period, independent action has been usedfrequently in specific trade lanes. Data compiled by the FMC alsoindicate that the filed independent-action rates are frequently matchedby other conference members. 22 While the use of independent action21 Before !984, a conference nlember could take independentaction only if the conference agreenlent permitted it.22 For instance, in the 1985-88 period, members of theTranspacific Westbound Rate Agreement (outbound from the U. S .Pacific Coast to Japan) used independent action to initiate filings. 775 rateDuring the same period, there were 45401 additional filings(continued. . .may imply that there have beendefectionsfrom conference rateagreements, the evidence of matching behavior may indicate that someof these defections have been punished. The 1984 Shipping Act also specifically authorized the use ofservice contracts\" Le., contractual arrangements whereby a shippercommits to providing a minimum quantity of cargo. or freight revenueover a fixed time period, and the ocean common carrier or conferencecommits to a certain rate or rate schedule as well as a defined servicelevel (sud.~ as assured space, transit time, or port rotation).23 Priorto 1984, the issuance of a service contract could be challenged asunjustly discriminatory.Although the 1984 Act liberalized thecontracting environment in ocean shipping, certain constraints remain.Service contracts must be fIled with the FMC, and theiressentialtermsare publicly available. Moreover, other shippers that canadhere to the terms of the contract (known as similarly situated\"shippers) are entitled to the identical arrangement from the carrier orconference issuing the contract.Between late 1984 and early 1986, some conferences in theAsian trades allowed their members to enter autonomously into servicecontracts. All other conferences prohibited their members from doingso, and would only issue a conference-wide service contract. In 1986those Asian conferences allowing their members to enterindependently into service contracts revoked that privilege. Casualempiricism reveals that this change in the contracting environment hadcontinued)where conference members matched independent-action rates initiatedby other Inembers. By contrast, members of the Japan, Atlantic, andGulf conference (inbound from Japan to the U.S. Atlantic and GulfCoasts) initiated only 266 independent-action rate filings, and therewere 167 additional filings of matching rates.23 This description of a service contract is based on that providedin Section 3(21) of the Shipping Act of 1984.a significant effect on market behavior; consequently, the entirestructure of freight rates may have been affected.24 Our empiricalanalysis below examines this question in detail.Ill.Brief Description of the Theoretical and EmpiricalModelsOur objective is to determine whether liner conferences caneffectively collude, that is, set shipping rates as if the conferencemembers were a unified, profit-maximizing fIrm. Before presentingrigorously the theoretical model and how we implement it empirically,we briefly describe our conceptual approachassumptions underlying it. , focusing on theAs mentioned above, we wish to test the hypothesis thatmembers of a conference perfectly collude. Our theoretical modeltherefore, assumes that conference nlembers can do so; this permitsour empirical work to test directly this hypothesis. We also assumeconsistent with observed practice, that conferences transport a varietyof commodities of varying values, and that they can set differentfreight rates for different commodities. Based on this paradigm ofperfect collusion, our model predicts that as a conference on a givenroute becomes larger (due to outside firms joining the conference), theconference will be better able to exercise market power. Thisincreased market power will result in an increased ability byconferences to discriminate in their pricing on the basis of commodityvalue. Consequently, as conferences increase in size relative to theshare of their nonconference rivals, freight rates are expected toincrease in general and the differential between the freight rates of two24 For instance, when members of the Transpacific WestboundRate Agreement were allowed to independently enter into servicecontracts in 1985 and 1986, 40% of their cargo was carried undercontract. In 1988, after independent contracting was revoked in favorof a conference-wide service contract, only 1 % of the cargo waS carried under contract.commodities of different value is expected to increase (Le.dispersion of freight rates across , thecommodity values is expected tobecome greater.In setting freight rates across the various commoditiesprofit-maximizing conference will attempt to equate the marginal, aprofitability from shipping the last container of each commodity.Basic economic principles establish that there is an inverse relationshipbetween the profit-maximizing price for a particular service (in thiscase, transporting a particular commodity on a particular route) andthe elasticity of demand for the service. Our theoretical modelcombines i:his relationship with three additional assumptions: (1) thatincreases ill shipping rates are entirely passed through toconsumers; (2) that the market demand elasticity for any final good is- finalconstant with respect to price and invariant over time; and (3) thatfringe II firms do not alter their outputs in response to changes in theconferences output. 2S The result is an equation26 that relates theconferences profit-maximizing price for shipping a particularcommodity on a particular route to five factors: commodity, (2) the conference(1) the value of thes market share with respect totransporting the commodity on that routefor the commodity, (4) the marginal costs incurred in shipping the, (3) the elasticity of demandcommodity on that route, and (5) the \"shadow value\" of conferencecapacity on that route. The \"shadow value\" of capacity is a measureof the value to the conference of adding more capacity; this value is2S While invoking these assunlptions simplifies significantly thederivation of the theoretical modelnecessary to support our ultimate empirical , they actually are stronger thanspecification. Forexample, the same empirical specification emerges if the degree ofpass-through is independent of commodity value and/or if the supplyvarious transported commodities. elasticity of the nonconference \"fringe\" firms is similar across the26 This is equation (3) in Section IV below.positive when the conference is capacity-constrainedzero when the conference has excess capacity., and it equalsTwo testable hypotheses emerge from this analysis. Firsttheory predicts that, holding all other factors constant, thea given route are, freight rates on, in general, positively related to the conferencemarket share on that route. Second, the theory predicts that, holdingall other factors constant, the dispersion of rates across commodityvalues becomes greater as the conferences market share on that routeIncreases.This theoretical construct need only be modified slightly to testthe additional hypothesis that increases in overall market -(as opposed to increases in the conferenceconcentrations market share) contributeto increased freight rates. The only difference is that an index ofmarket concentration would replace the conferencean independent variable explaining the level of freight rates. s market share as27 Asbefore, the two testable implications that emerge are as follows: freight rates generally increase with overall market concentration on a(1)given route, and (2) freight rates will become more dispersed acrosscommodity values as market concentration increases.Transforming the theoretical model described above into aspecificathn that can be estimated empirically requires some finalmodification. First, we assume that the conferences market share oftotal capacity on a particular route is a very good proxy for its marketshare in transporting any given commodity on the route. Second, weassume that the elasticity of demand for any particular commodity isindependent of the commodity's value (or, that there is not a strongpositive relationship between a commodity's value and its elasticity ofdemand.) Third, we need to estimate the \"shadow\" value of capacity.Since the \"shadow value\" depends on the total capacity level on theroute, the conferences market share on the route, and cost and27 The market concentration index treats each firm individually whether the firm is a member of the conference or not.demand factors, we include these variables in our empiricalspecification.Lastly, the specification requires an estimate of the marginalcost of shipping a specific commodity on a specific route. We resolvethis by noting that the cost factors specific to transporting a givencommodity along a given route (such as, shipping audio equipmentfrom Japan to the United States) are not likely to vary over the timeperiod covered by this analysis. Thus, we can control for differencesin transportation costs specific to a given commodity on a given routeby including dummy variables for each commodity on each route.Incorporating the four modifications above, we obtain the specificationthat is em.I:Jirically estimated.In sum, our empirical specification stems from the hypothesisthat liner conferences set prices as if the member carriers colludedperfectly. We can test statistically whether the rate behavior of linerconferences is consistent with the behavior predicted by thishypothesis. As mentioned previously, the results would be consistentwith this hypothesis if we observed the following: (1) freight ratesgenerally increasing as conferences increase in size relative to theirnonconference rivals, and (2) the dispersion of freight rates acrosscommodity values increasing as conferences become relatively larger.Our empirical approach also can test whether market concentration, asopposed to conference market share, affects freight rates.IV.The Theoretical ModelW ebase our statistical specification on a model where a linerconference on a given route acts as a residual monopolist, facingcompetition from a fringe of carriers that have refrained from joining28 This is represented by equation (4) in Section V below.:;:; the conference. 29 In this sectionrates that is consistent with joint profit maximization by a cartel, we derive an expression for freightcompeting with outside (Le., fringe) firms. As following sectionshown in the, we can econometrically estimate this expression totest whether the rate behavior of liner conferences conforms with thatpredicted by our model. Our statistical specification also considers thepossibility that market power stems from market concentration, andnot from the conference system.A profit-maximizing cartel operating on a given route wouldattempt to equate the marginal contribution to its profits that itreceives from transporting an additional container of any givenproduct. Hence, for a cartel transporting good on route j, it holdsthat: 30ij ij ij where MP ij (MR(1)ij,marginal contribution to profits (marginal revenue, marginal cost) from transportingan additional container of good i on route 29 The existence of fringe carriers competing with the cartel maybe attributable to a variety of factors. As more carriers join thecartel, the benefits increase from remaining outside the cartel and free-riding off of the relatively high cartel prices. Among othersDonsimoni (1985) and Deneckere and Davidson (1985) find thatn stablecartels are typically incomplete. Firms outside the cartelhave no incentive to enter, and cartel members have no incentive toleave.30 The subscripts denoting time have been dropped forexpositional convenience.== j, In equation (1), MCthat the cartel maximizes profits subject to its capacity limitation;ij refers to short-run marginal cost. It is assumedhence, the term, kj, represents the shadow value of capacity. If thecartel's capacity equalled that level needed to maximize long-runprofits, then the shadow value of capacity, would equal themarginal cost of capacity. N0te that ij (l l/nij ), where rij is the freight rate forcommodity on route j, and nij C is the elasticity of demand that thecartel perceives it faces in transporting commodity on route j.Substituting into equation (1), we obtain:MP.. IJ r..MC.. (2)IJ IJ k..Under appropriate assumptions, we can express nij C as function of the marketelasticity of demand for good and themarket share of the cartel on route j. First, it is assumed that anyincrease in transportation costs is entirely passed through toconsumers. In other wordsdpufdrij which implies that dqufdrl(aqij/ap~(dpij/dr)J (aqufap~, where qij is the total quantity ofgood (from the origin country) demanded in the destination countryon route j, and Pij is the delivered price of good in that country.Second, it is assumed that the markets elasticity of demand for agiven good is constant with respect to price and invariant over time.Henceaqufapij = -nij-(aqufap~(pijlq~ is themarket'(qij/pJ, where ns elasticity of demand for good in the destination country onroute j. Together, these two assumptions imply that dqij/dr (q;/puJ. Under the further assumption that there is no supplyij = response by fringe firms, we then obtain nij C 55 -(dqijldr ~(rijq~ 31 Hence, our model is sufficiently flexible to allow for eithershort-run or long-run profit maximization.j. ij M~ijsj, where sij is the cartel's market share in transporting goodon route Incorporating this last result into equation (2), we obtain:Too IJ I) s..IJ n.. IJ MC.. I) (3)r.. IJ k. MC.. IJ 1Js.1nooi/ Consider equation (3). Within a group of commodities facinga similar market elasticity of demand (Le., 1/\"ij = 7IM Y I), a profit-maximizing cartel would generally set higher freight rates for thehigher-valued commodities (since drufdpij 0).32 This occursbecause although the market elasticity of demand is the same for thisgroup of commodities, the derived elasticity of demand fortransportation services decreases (in absolute terms) as the value of thecommodity increases. When passed through to consumers, a givenincrease in freight rates raises product price by a smaller percentagefor those products that are relatively high-valued. This implies thatconsumers of high-valued commodities are potentially less sensitive tochanges in freight rates. Recognizing this relationship, a profit-maximizing cartel sets freight rates that are generally increasing withrespect to commodity value.Based on equation (3), as the cartel's market share expandsdue to outsidefmns joining the cartel, the rate differential shouldincrease between two commodities of different value (since J2r/dpij0). This behavior would not occur if the rate differentials acrosscommodities were based purely on cost factors.32 More generally, freight rates would increase with respect tocommodity value whenever Pi/1/\" ij is positively correlated with Pij.This behavior necessarily arises unless Pij and rf ij are positivelycorrelated (Le., unless high-valued goods face relatively elasticdemand). There is no empirical support that such correlation exists.. )The above specification can also be obtained using moregeneral assumptions. For instance, instead of assuming full pass-through\" of transportation costs to the product's delivered price, wecan merely assume that the degree of pass-through is independent ofcommodity value. Also, instead of assuming that fringe firms do notchange their supply of transportation services, we can assume thatthese firms respond to changes in the quantity of transportationservices supplied by the cartel, but their response is similar across alltransported commodities.The above model allows for the possibility that linerconferences exercise effective market power. However, anotherindependent source of market power should also be considered. Thepresence of scale and network economies may imply that the oceanshipping industry can sustain a limited number of firms in equilibrium.Additionally, there may be sunk costs involved in serving a givenroute (Le., costs of warehouses, cargo-handling equipment, and otherterminal facilities) that may constrain further the number of firms ableto serve that route. Depending on their strategic behavior, the fewfirms operating on a given route may exercise market power sufficientto permit discriminatory pricing, regardless of whether a conferencesystem exists. If high levels of market concentration allow firms toexercise this type of market power, the associated first-order conditionis identical to equation (3), except that an index of marketconcentration (Le., the Herfindahl index) replaces the cartel's marketshare. 33becomes: 33 In a static oligopoly setting, Cowling and Waterson (1976) andothers have shown that market price is related to the Herfindahl indexof market concentration (~). In these circumstances, equation (3)(continued. . ), (.. . ), T!::e Empirical SpecificationIn equation (3), freight rates are assumed to be a linearfunction of the shadow value of capacity (k), marginal cost (MCand an expression that includes the value of the transported commodity(Pijconference market share and the market elasticity of demandfor the commodity (tfij). The termassuming that the conferences market share is the same forpi!ij/tfij, can be approximated bytransporting any commodity on a given route (Le.the market elasticity of demand is similar across commodities (Le.ij = S), and that34 tfij = rf'-).Incorporating these assumptions into equation (3), weobtain: 35T.. k. MC.. IJ IJ)(4)Under the assumption that market concentration, in addition toconference market share, might affect freight rates, equation (4) wouldbe modified slightly to:continued)T.. IJ ) J)MC.. or r.. k. MC.. IJ IJHln..34 We could instead assume that pij/rf'- ij is positively correlatedwith Pij. See footnote 32.35 Note that Pij represents the import price in the destinationcountry. Since this price is essentially inclusive of transportationcosts, we use the export price in the origin country to avoid spuriouscorrelation.j' ~) T.. IJ k. MC.. IJ III '1n Mn M(5)where is the Herfindahl measure of market concentration on routej. To transform the above equation into a usable statisticalspecification, we must devise a' methodology that: (i) estimates ~, (ii)circumvents the data problems involved in estimating considers possible alternative sources of market power other than linerij, and (iii)conferences .The shadow value of capacity on a given route (Le.depends on the demand for ocean shipping services on , ktt1at routeamount of capacity on that route, the, the conferences market share onthat route, and (possibly) the degree of overall market concentrationon that route. More specifically, with respect to equation (5), weassume that kfunctionCAPf(CAPj, pJj' rj, Swhere is a linearis total capacity on the route, Jd(respectively) the destination countryj and pdj ares income and price level, rj isthe origin country's price level, Sj is the conferences share of totalroute capacity on route j, and is the Herfindahl measure of marketconcentration on route j. As capacity increases on the route, theshadow value of capacity is expected to decrease until it finallyreaches zero (where capacity is no longer binding). An increase in thedestination countrys income or its domestic price level (relative to theforeign price level) is expected to raise import demand, andconsequently, the demand for shipping services on the route. Thuswhen capacity is a binding constraint, we expect kj to be positivelyrelated to Jdj and pdj, and negatively related to rj. Finally, as theconferences market share (or the index of market concentration)increases on the route, the benefit to conference carriers (or carriers inthe market generally) of increased output becomes progressivelysmaller, for the same reason that the marginal revenue is lower for amonopolist than for a perfect competitor. The shadow value ofcapacity is therefore expected to decrease as conference market shareand overall market concentration increases.The effect of these variables on freight rates is identical to thatdescribed above, since freight rates are positively related to theshadow value of capacity (see equation (5)). Of course, if the capacityconstraint is not binding (Le., k0), these variables would notaffect freight rates.The short-run marginal cost of transporting commodity ionroutej (Le.depends on such factors as the distance traveledthe value and weight of the cargo, and the need for special handling(e., refrigeration). Note that these variables are practically invariantover time for a given commodity on a given route36 implying thatdifferences in marginal cost arise across commodities and routes.Thus, we can account for the impact of these variables by using amodel which includes dummy variables for each commodity .37 This approach not only circumvents the data problemson eachroute.involved in accurately estimating marginal costs, it , but as we discusslateralJO eliminates potential biases involved in examining therelationship between freight rates and conference market share (ormarket concentration) from data covering multiple commodities andmultiple routes.Dummy variables for each year (19851988) are alsoincluded in the specification, and an additional dummy variable (lAdenotes those routes for which conference members were allowed toenter independently into service contracts in that specific year. Whileour data excludes rates on service contracts, we hypothesize thatallowing conference members to autonomously enter into servicecontracts could have altered the entire rate structure.36 From year to year, cargo values may change moderately for agiven commodity on a given route. Since annual data for cargo valuewere not available for this study, we assume that relative cargo valuesacross commodities and routes are constant over time.37 For a discussion of such \"fL'(ed-effects\" models, see Judge, etale (1982), pp. 477-502.j + In addition to considering the possibility of joint profit-maximizing cartel behavior, our specification allows thatindependently of the conference systemlead to market power. , market concentration maySince this hypothesis implies that a relationshipexists between freight rates and market concentrationincluded a Herfindahl index of market concentration (hereafter, we haveour specification. This variable is included by itself and interactedwith cargo value.The complete specification is described as follows:roo+ (32SPij + f3~ f3Jlpij (3sCA~ f361'j(6)+ (37(38f3JA+ (3101985 + (3111986+ (3121987 + 13131988.In the above specificationeach commodity on each route by instead expressing each observationwe avoid using dummy variables foron a giver. variable in terms of its deviation from the respective mean variable statistically equivalentfor that commodity on that route. This approach is, and can be interpreted analogously, to theregression where all variables are expressed in terms of their levelsand dummy variables are included for each commodity on each route(see Judge, et al., 1982, pp. 478-481).The above specification is capable of estimating therelationship between conference market share and freight rates. also determine whether the relationship between conference marketIt canshare and freight rates is different for high-valued low-valued commodities. From the above specificationcommodities andfreight rates associated with an increase in conference market share, the change independs on the estimated coefficients from two variables: market share (si) and conference market share multiplied byconferencecommodity value (sj!ij). The first of these estimated coefficients is (31;38 Table 1 defines each of variables used in the analysis...:::: the second is fJ2.39 Note that the sign of this effect may depend oncommodity valuePij. If conference pricing is consistent with the jointprofit-maximizing behavior in our model, then the degree of pricediscrimination should increase with conference market share (implyingthat fJ2 0). In addition, when capacity is not a binding constraintan increase in conference market share should raise freight rates forall commodity values. In that case, we would expect that fJl and40 fJ2 However, when capacity is a binding constraint, anincrease in conference market share should raise rates only for higher-valued commodities. This occurs because, ceteris paribus, an increasein the cartel's market share (due to an outside firm joining the cartel)lowers its perceived marginal revenue. This drop in marginal revenueis relatively larger for those commodities where the derived demandfor transportation services is relatively inelastic. To put its pricestructure hack in equilibrium, a capacity-constrained cartel would raiserates for those commodities with a relatively inelastic derived demandfor transportation services (Le.higher-valued commodities), andlower rates for those commodities with a relatively elastic deriveddemand (Le.lower-valued commodities). Thus, we would expect thatfJl and P2 ~ O. Note that this discussion presupposes thatconferences act as profit-maximizing cartels and are able to pricediscriminate, which may not hold in reality.In fact, conferences may not be able to price discriminate ifnonconference carriers act as cream skimmers.In this case, if aconference attempts to set different freight rates for commodities thatbear the same transportation costnonconference carriers would39 Algebraically, the change in freight rates associated with anincrease in conference market share is: dr;/dsfJl P1Pij.40 As explained above, the coefficient increase in conference market share on the shadow value of capacity.PI measures the effect of anWhen cap.lcity is not a binding constraint, this effect necessarilyequals zero.choose to transport only those commodities with relatively high freightrates. This arbitrageactivity prevents conferences fromdiscriminating across commodities in setting freight rates. Whenconferences act as profit-maximizing cartels under these conditions(and capacity is not a binding constraint), we would expect that anincrease in conference market share would lead to a uniform rateincrease across all commodities (Le., fJt ~ and fJ2 = 0).Now, consider the possibility that market power stems insteadfrom market concentration. According to our model, the degree ofprice discrimination should increase as the market becomes moreconcentrated (Le., fJ4 ~ in equation (6)). When capacity is not binding constraint, freight rates should generally incr~e as marketconcentration increases. 41 We would thus expect fJ3 = and fJ4o. When capacity is a binding constraint, an increase in pricediscrimination should lead to higher freight rates for higher-valuedcommodities and lower freight rates for lower-valued ones. Thisimplies that fJ3 and fJ4 ~ if capacity is binding.Due to limited data, no attempt was made to formulate aninstrument for conference market share or market concentration. 41 That is, the derivative of freight rates with respect to marketconcentration (dr/d~ ) should assume a positive valuefor all commodity values.fJ3 + fJJ1ij42 The reported results also treat total route capacity asexogenous. Industry participants indicated that ocean shipping andshipbuilding are highly subsidized industries, and that capacitydecisions may be heavily influenced by government policies. To seewhether capacity was more heavily influenced by traditional marketforces or national policies, we regressed capacity on a variety of costand demand factors (relative prices, gross domestic product, etc.) aswell as route-specific dummy variables. Consistent with informationgathered in interviews, all of the route dummies were statistically(continued.. .(.. . This econometric approach is therefore subject to some of the samecriticisms that have been aimed at prior empirical research treatingmarket structure as exogenously determined.43 However, our fixed-effects model does avoid potential sources of bias that would arise ifour analysis compared rate behavior across routes instead of withinroutes. Our specification examines the relationship between deviationsover time in the freight rate for a given commodity on a given routeand deviations in other variables pertaining to that routesuch asconference market share and market concentration on the route.Accordingly, we avoid any examination of the relationship acrossroutes between the level of freight rates and the level of market share (or market concentration). Much of the past criticism ofconferencethe exogenous treatment of market shares and market concentrationrelates to its use in the cross-sectional analysis of behavior acrossmarkets, which is avoided in our treatment of the data.As a final point, we note that changes in conference marketshare on a given route are measured holding market concentrationconstant on the route. Our analysis is therefore designed to capturethe movement of outsidefirms into a given liner conference. Ourexamination of the sample data confirmed that most of the variation ina given conferences market share over time was attributable tocarriers entering or leaving the conference.continued)significant, but few cost and demand factors were statisticallysignificant. It might be possible to construct an appropriate instrumentfor route capacity from the subsidy levels for ocean shipping (andshipbuilding) in various countries. Unfortunately, such data wereunavailable.43 For discussion, see Bresnahans and Schmalensees (19)chapters in the Handbook of Industrial Organization. See also Froeband Werden (1991).), VI.DataThe sample consists of port-to-port freight rates charged byliner conferences serving the United States between 1985 and 1988.. Pursuant to Section 18 of the Shipping Act of 1984, the FMCcollected conference freight rates for the most popular commodities ona given route, accounting for at least 50 percent of the volume on thatroute. The data covered fourteen conferences carrying outbound andinbound freight between the U.S. Atlantic and Pacific coasts andJapan, Germany, Italy, and Australia.44 Freight rates (T) wereexpressed as total charges per twenty-foot-equivalent (Le.TEU - thestandard measure of container volume), including all relevantsurcharges, such as bunker-adjustment, currency-adjustment, terminal-handling, and container-yard fees. Due to the use of independentactionconference carriers occasionally offered more than one ratefor the transportation of a given commodity. In that situation, weused the rate under which the majority of the cargo was transported.Commodity value (Pij) was derived from export and importdata collected by the Bureau of the Census and compiled by theMaritime Administration of the U. S. Department of Transportation.From these data, we used standard conversion factors to translate fromcommodity value per long ton to commodity value per TEU.Capacity (CAPconference market share (S), and marketconcentration (~) were compiled from data furnished by Lloyd'Maritime Information Services. Lloyd's collects capacity data (inTEUs) for each carrier on a given route, and identifies whether thecarrier belongs to the conference covering that route.45 Conference44 Appendix A lists the routes and the commodities used in oursample.45 To determine carrier capacity on a given route, each vessel'capacity is multiplied by the number of voyages it made on that route. (continued...(.. . market share is the proportion of total capacity operated by conferencecarriers. Market concentration is measured by the Herfindahl indexthat is, each carriers share of total capacity is squared and then thesefigures are summed for all carriers serving the route.The destination country's gross domestic product and pricedeflator (:tdj and pdj, respectively) and the origin countrys pricedeflator (P') were obtained from International Financial Statisticspublished by the International Monetary Fund. The price indices wereadjusted for exchange rate movements.The fmal data set, containing 620 observations, is one of thelargest ever used to analyze rate behavior in international' oceanshipping. Since there are four annual observations (from 1985 to1988) for each commodity on each route, our sample contains 155commodity-route combinations. Table 1 contains a description of eachvar~able and its data source, while Table 2 contains summary statisticsfor key variables.continued)during that year. Theseadjusted\" vessel capacities are then summedover all vessels used by a given carrier on that specific route.VII.ResultsTable 3 presents regression results for the entire sample of 620observations.46 The results offer some support for the hypothesisthat increased market concentration leads to higher freight rates. total effect of market concentration on freight rates (Le.The/d~ 026 0000018pijis positive and statistically significant at the10% (5%) level for commodity values exceeding $77268($103,172),representing 24 % (19%) of the total sample. the variable that interacts market concentration with commodity valueSince the coefficient onis positive and statistically significant (at the are also consistent with the hypothesis that increased market10% level), our resultsconcentration leads to greater price discrimination on the basis ofcommodity value. 46 Note that the If- where each observation on a given variable is expressed in terms of itsstatistic in Table 3 is based on the regressiondeviation from that variablethat route. In s respective mean for that commodity onother wordsIf- omits the explanatory power providedby the dummy variables for each commodity on each route.However, the calculation of the degrees of freedom must recognizethat these 155 dummy variables are implicitly included (see Judge, etal., 1982). Hence, the degrees of freedom equal 620 - 452.13 - 155 =47 Since the coefficient on market concentration alone statistically significantis not, and the coefficient on market concentrationinteracted with commodity value is statistically significantare consistent with profit-maximizing behavior in a concentrated, our resultsmarket when capacity is not a binding constraint. statistical tests further support that capacity is not binding; in Additionalparticular, an F-test does not reject the null hypothesis that thecoefficients equal zero on all variables (Le.CAlj, relevant to determining the shadow value of capacity.j, plj' r(continued. . .\\.. .The results in Table 3 do not support the hypothesis that anincrease in conference market share leads to higher rates or greaterprice discrimination. Neither of the variables that include conferencemarket share are statistically significant; moreover, an F-test does notthese variables. reject the null hypothesis that the coefficients equal zero on both Despite the apparent lack of a statistical relationship betweenconference market share and freight rates, our results indicate thatchanges in conference rules, 'Such as those pertaining to servicecontracts, are associated with changes in freight rates. Specifically,the coefficient is negative and statistically significant for the IAvariable, which identifies those routes where conference memberscould enter independently into service contracts within a givenyear. 48 Ceteris paribus, freight rates were approximately $590 lower(on average) on those routes allowing independent actionon servicecontracts during that particular year. This reduction represents about19 % of the mean freight rate in our sample.continued)It should be noted that, contrary to the predictions of ourmodel, the coefficient on the origin countrys price level is positiveand statistically significant at the 5 % level. Our model posits that thisvariable affects the destination country'demand the origin country. It is for the products ofquite possible that this variable insteadcaptures movements in wages and input prices in the origin countrythat are relevant to the cost of transportation services on that particularroute.48 Two outbound routes fit this categorization. In 1985, theconferences on the U. S. East Coast - Japan and the U. S. West Coast -Japan routes permitted their members to enter independently intoservice contracts with shippers. The conferences on these routes hadmarket shares of 82 % and 85 % in 1985.Independent actionindividual conference carrier to enter into an agreement with anon service contracts allowed anindividual shipper without any conference restriction on the terms orservices provided under the agreement. These filed with the FMCcontracts were then, but certain elements of the contract were notpublicly available (Le., the name of the shipper and the carriercertain information relating to service provisions). , andadvent of independent action on service contracts could have increasedConsequently, thethe cost to the conference of monitoring the rate and service offeringsof its members. Further, unlike independ~nt action on regular discountmust be offered to all shippers of thattariffrates where the commodity on that route, contracts could be written in a mannerwhich effectively allowed the conference carrier to offer - a selectivediscount to that particular shipper (or a small group of similarly-situated\" shippers). Due to these circumstances, permittingindepende'lt action on service contracts could have increased theattractiven~ss of cheating and inhibited the ability of the conference todetect and punish cheating. The associated reduction in conferencemarket power might have resulted in lower freight rates for all typesof transactions, including the spot\" (i.e., noncontract) transactionsthat comprise our sample.The above findings are consistent with three primaryconclusions: 1) conference rate behavior is not consistent with thejoint profit-maximizing behavior predicted by our model; (2)conference rules nonetheless may affect the exercise of market powergiven that rates were lower when conference members were allowedto enter independently into service increase in market concentration is associated with increased freightcontracts with shippers; and, 3) anrates, particularly for high-valued commodities. With respect third conclusion, it is important to consider whether the positiveto therelationship between market concentration and freight rates iseconomically significant, as well as statistically significant. Table 4shows the effect on freight rates of a one-standard-deviation (hereafterI-s.) increase in market concentration, evaluated at variouscommodity values.49 At the median commodity value (of $31912),a 1-s.d. increase in market concentration is associated with a rateincrease of $12, which represents only 0.4% of the average freightrate. Calculated at the highest commodity value (of $415493), a 1-d. increase in market concentration is associated with a $277increase in freight rates, or about 8.9% of the average freight rate.Because the substantial U.S. trade deficit between 1985 and1988 led to considerably heavier ocean traffic on inbound routes, wedecided to examine whether the results change qualitatively wheninbound and outbound routes were analyzed separately. Table 5presents summary statistics, and Table 6 presents the regressionresults. soAccording to Table 6, the results from the pooled data stemlargely from behavior observed on inbound routes. For those routesmarket concentration is a statistically significant determinant of freightrates, while conference market share is not. On outbound routes49 Our fixed-effects model essentially measures the relationshipover time between the freight rate for a given commodity on a givenroute and the variables relevant in determining that rate. Since freightrates for a given route are influenced only by changes in marketconcentration on that route, we derived the standard deviation ofmarket concentration used in Tables 4 and 7 from an average of theindividual standard deviations of market concentration calculated foreach route (based on changes in the routes concentration over time).In Table 4, which includes all routes, this average value is 384; inTable 7, which includes outbound routes only, it is 407.so As shown in Table 6, we omitted from these regressionscertain variables that were invariant across outbound routes (i.origin-country price index) or inbound routes (Le., destination-countryprice index and GDP). Since these variables only changed from year-to-year, they were perfectly collinear with the yearly dummyvariables .neither market concentration nor conference market share is asignificant determinant of freight rates. 51 During the period coveredby this analysis, routes. It , significant excess capacity existed on outbound, butnot inboundseems likely that the presence of this excesscapacity prevented the exercise of market power on outbound routes.The results on the inbound routes are qualitatively similar tothose obtained from the pooled data. One difference is that anincrease in market concentration is associated with a statisticallysignificant increase in freight rates for a larger portion of commoditiesin the inbound sample than in the pooled sample (e., at the10% (5%) level of significance, 45% (40%) of the sample as comparedto 24 % (19%)). A comparison of Tables 4\" and 7 also shows that thisincrease in freight rates is relatively larger in magnitude in theinbound sample. For instance, a 1-s.d. increase in marketconcentration on inbound routes is associated with an $50 increase inthe freight rate at the median commodity value, representing about4% of the average inbound freight rate (see Table 7). At the highestcommodity value, this increase in market concentration is associatedwith a rate increase of $439, or about 12.6% of the average inboundfreight rate.In closing this section, we acknowledge Stiglers (19)hypothesis that a cartel's ability to reach a joint profit-maximizingpricing agreement may be facilitated as the number of colluding firmsdeclines. In other words, his model predicts that prices may be higherwhen the cartel contains firms with large market shares instead of51 An F -test does not reject the hypothesis that the coefficients the inbound and outbound regressions are the same; however theresults on the outbound routes are qualitatively different from thosepreviousl:, mentioned.smallones.52 To test this hypothesis, we constructed a conferenceHerfindahl index, which represented the sum of the squared values ofeach conference members share of conference capacity. We addedthis variable, by itself and multiplied by cargo value, to thespecification in equation (6). This variable was found not to bestatistically significant in either the full sample or the samples inbound 3'ld outbound routes only. Thus, we found no apparentrelationship between the concentration of the conference and thebehavior of freight rates.VDI. Concluding RemarksThe U. S. international ocean shipping industry enjoys antitrustimmunity and a conference system that allows carriers to enter intopricing agreements which are then monitored and enforced by agovernment agency. Previous empirical analysis has concluded thatthis system allows the exercise of cartel-based market powerevidence that freight rates are increasing with respect to cargo value., based onThis finding has been interpreted as consistent with discriminatorypricing by an effective cartel, but the literature has ignored thepossibility that these rate differences stem from cost differences orother sources of market power. In this study, we attempt to addressthese shortcomings by directly examining the relationship betweenfreight rates and both conference market share and marketconcentration. Our statistical approach controls for cost differencesacross commodities and routes.A !though we fmd no significant relationship betweenconference market share and freight rates, our evidence indicates that52 Another effect may work in the opposite direction of thatsuggested by Stigler. Firms with relatively large market shares mayface lower costs (or flatter marginal cost curves). This result suggeststhat relatively large firms may gain more from cheating on a givencartel price. Consequently, as firms become larger, the sustainablecartel price may decline.% . freight rates were significantly lower on those routes where individualconference carriers were allowed to enter into service contracts withindividual shippers. These results perhaps when combined with relatively high conference market sharesuggest that some conference rulesmay allow carriers to maintain rates at levels higher than they wouldotherwise. Market power is undermined when carriers within aconference are allowed to independently contract with shippers.Our results show that an increase in market concentration isassociated with increased rate levels and increased rate dispersion onthe basis of commodity value; this is consistent with the hypothesisthat greater market concentration leads to greater market power.Nonetheless, a one standard deviation increase in concentrationincreases rates only 0.4% of the median freight rate for the entiresample. \"llen we divided the sample between inbound and outboundroutes, the above results continued to hold for the inbound routesonly. Relative to the entire sample, the increase in freight ratesassociated with increased market concentration was larger inmagnitude and statistically significant for a wider commodities in the inbound sample. range ofoutbound routesAt the median freight rate for, a one standard deviation increase in marketconcentration increases freight rates approximately 1.4 outbound routes, where there may have been considerable excess53 Even when firms apparently wield market power (Le.above marginal cost), there may still., set pricebe a cost-the positive relationship between market concentration and price. based explanation forapplied to ocean shipping, it is likely that there are some fixed (butnot sunk) costs involved in serving a given route. If those costs wereto rise, some carriers would exit the route. market concentration and freight rates would increaseConsequently, bothincrease might only be sufficient to allow , but the ratefixed costs. ThusfirIng to again cover their, the exercise of market power by - does not recessarily imply that those firms earn supranormal profits.individual flI1Ilscapacity, the analysis could not identify a significant statisticalrelationship between freight rates and either market concentration orconference market share. Perhaps the presence of substantial excesscapacity eliminated any sources of market power. In general, our results show that a perfectly-collusive outcomeis unlikely even under the favorable conditions that e~ist inoceanshipping. Liner conferences do not, or cannot, maximize joint profitseven when rate information is publicly available and a governmentagency attempts to prevent cheating on collusive pricing agreements.The sustainablecartel pricing structure apparently divergesconsiderably from that which maximizes the joint profits of itsmembers. A cartel's market power may erode in the presence ofexcess capacity, firm heterogeneity, and multidimensional competitionwhich all arise in liner shipping. In particular, this study shows thatcollusive prices may be adversely affected by increased contracting,which allows firms to expand service offerings and engage in selectivediscounting.In light of the liner shipping industrys unique regulatorystructure, there may be another explanation for the above results. Therequired public filing of freight rates and the associated enforcementof those rates, which applies to all carriers and conferences, mayfacilitate anticompetitive interaction between firms inside and outside54 Ocean carriers have been particularly concerned with excesscapacity i:.~ recent years, spawning both formal and informalagreements to reduce capacity. These agreements include theTranspacific Stabilization Agreement, a so-calleddiscussionagreementbetween conference ~arriers and nonconference carriers tovoluntarily limit capacity on U.Far East routes. The Trans-AtlanticAgreement, a ratemaking agreement involving liner transportationbetween the United States and Northern Europe, also containsprovisions to reduce capacity. It is quite possible that theseagreements are intended to inhibit any procompetitive pricing behaviorthat stems from the existence of excess capacity.of liner conferences. At the same time, conferences may be inhibitedin maintaining internal pricing discipline by their members' ability toautonomously set rates (through the independent action provision)and exit the conference at any time with limited notice. Thisreasoning suggests that the importance of conference market share asan indicator of market power in ocean shipping may be less importantthan market concentration. What does our analysis imply about the costs and benefitsfrom the regulation of liner shipping? Our finding that freight rateswere significantly lower when conference carriers were free tonegotiate directly with shippers provides some support for theproposition that the conference system imposes costs on consumersthrough higher freight rates. At the same time, however, it isimportant to point out that this finding is based on routes where theconferences market share is relatively high (over 80%), and thatfreight rates are not directly related to the market share of theconference serving the route. Further, it is also possible thatconferenc~ provide some offsetting benefits, such as increasedefficiency in providing a network of ocean transportation services.Coordinated activities among ocean carriers may be beneficialwhen those activities lead to improved rationalization of resources(e.g., through vessel-sharing and space-chartering arrangements) orinduce efficiency-enhancing investments (e., through consortia thatoffer additional transportation services). It is unclear, however, thatcollective pricing behavior actually facilitates those coordinatedactivities which improve efficiency. Moreover, even if collectivepricing is required to encourage efficiency-enhancing investment, theassociated anticompetitive harm may still outweigh the efficiencybenefits. Certain regulatory aspects of the open conference systemparticularly the ability of carriers to enter or leave the conference withlimited notice, would presumably discourage carriers from usingss For further discussion of this point, see Butz (1993).conferences to engage in efficiency-enhancing joint investment andrationalization activities. It is this tradeoff between the potential for anticompetitiveharm and the potential for increased efficiency that should form thebasis for analyzing coordinated activities in ocean shipping andassessing whether antitrust immunity should be continued. A similarcost-benefit test could be usefully applied to other aspects of theregulatory structure in ocean shipping, such as tariff filing andenforcement and restrictions on contracting.S6 Butz (1993) and Reitzes (1993) consider these issues in detail..- APPENDIX AROUTES AND COMMODITIES CONTAINED IN THE SAMPLEAustralia U. S. Atlantic CoastApplesBeef (frozen)BeerCaseinCheeseCanned FruitGlutenNickelPears10.Wine11.Woolustralia . S. Pacic CoastAutomobile PartsBeef (frozen)BeerCaseinGlasswareGlutenSteel SheetsSteel TubesTin Plate10.Wire Rods11.ZincS. Atlantic Coast AustraliaAutomobile PartsBoxboardCandyCigarette TowCitric AcidClayGlass (flat)GlasswareMagazines10.Plastic Sheeting11.Tires12.Tobacco13.WhiskeyS. Pacic Coast -ustraliaAutomobile PartsBoraxCitrusDried FruitLubricating OilMarine EnginesNutsPulses (beans)Roadmaking Equipment10.Salmon11.Vegetables (frozen)German;ys. N. Atlantic CoastApple JuiceAutomobile PartsBeerChemicalsEnginesFurnitureG~1SswareHam (canned)Iron Castings10.Metal Working Machinery11.Moulding Machines12.Offset Presses13.Plastic Foils14.Tires15.Wine16.Wine GumsGerman;yS. S. Atlantic CoastAutomobile PartsBeerEnginesFurnitureHamSteel PipesSteel StripsTextile MachinesTitanium Dioxide10.WineS. N. Atlantic Coast rmanyAutomobile PartsCorn SeedElectrodesEnginesLumberMedical SuppliesPhoto gr ap hie Equ i p men tPork OffalsRoadmaking Equipment10.Rubber (synthetic) 11.Tobacco12.Yarn (synthetic)13.VeneersU. S. S. Atlantic Coast German;yAutomobile PartsClayF emsGrapefruitLumberPeanutsPork OffalsRubber (synthetic)Tobacco10.Wood PulpU. s. tlantic Coasts. tlantic Coast IJfJ1J!.Automobile PartsCopper and Brass ScrapCeramic TilesFootwearFurnitureMacaroniMarble TilesPaper (for printing)TextilesTomatoes (peeled)10.Tractor Parts11.Tractor Tracks12.Vermouth13.WineJaoan S. Atlantic CoastAudio Equipment Aluminum Sheets Auto Parts (of iron and steel) Auto Parts (panel group) Auto Parts (other) Bolts, Nuts, and ScrewsElectrical GoodsMachine ToolsMotorcycles10.Porcelain11.Textiles12.Tires13.Tractors14.Transportation Equipment15.TV Cameras16.Video and TV Receiving SetsCorn SeedHidesLumberPeanutsRagsTobaccoWaste PaperWood Pulps. Atlantic Coast JapanButterfishLubricating OilPeanutsTobaccoWood PulpJaDan s. Pacific Coasts. Pacific Coast JaDan10.11.Auto Parts (other)Auto Parts (panel group)Audio EquipmentElectrical AppliancesMotorcyclesPhotocopying EquipmentPorcelainTiresTV CamerasVideo and TV AccessoriesVideo and TV Receiving SetsBeefCottonFrench FriesMilk Carton StockSalmonWood Pulp, \", \", \", \", \", \", \", \"ReferencesAbreu, D., Pearce, D., and E. Stachettiwith Imperfect Monitoring, II' Journal of Economic Optimal Cartel EquilibriaTheory, Vol. 39(1985), pp. -251-269.Borenstein, S.Price Discrimination in Free Entry Markets,RandJournal of EconomicsVol. 16 (1985), pp. 380-397.R. Bresnahan, T.Empirical Studies of Industries with Market Power\" inSchmalensee and R. Willig, eds.The Handbook of IndustrialOrganizationVol. 2., (Amsterdam: Elsevier Science Publishers, 19),pp. 1011-1057.Bryan, 1.Regression Analysis of Ocean Liner Freight Rates on SomeCanadian Export RoutesJournal of Transport Economics and Policy,Vol. 8 (1974), pp. 161-174.Butz, D.Ocean Shipping Economics: Free Trade and AntitrustImplications,Contemporary Policy IssuesVol. 11 (1993), pp. 69-80.Chamberlin, E.Duopoly: Value Where Sellers Are FewJournal of EconomicsVol. 43 (1929), pp. 63-Quarterly100.Choi, E., Menezes, C., and J. TresslerTheory of Price-FixingRingsQuarterly Journal of EconomicsVol. 100 (1985), pp. 465-478.Cowling, K. and M. WatersonPrice-Cost Margins and MarketStructure, II EconomicaVol. 43 (1976), pp. 267-274.Deneckere, R. and C. DavidsonIncentives to Form Coalitions withBertrand CompetitionRand Journal of EconomicsVol. 16 (1985), pp.473-486.46 , \", \", \", \", \", \", \", \"Devanney, J., Livanos, V., and R. StewartConference Ratemakingand West Coast of South AmericaJournal of Transport Economics andPolicy, Vol. 9 (1975), pp. 154-177.Donsimoni, M-Stable Heterogeneous CartelsInternationalJournalof Industrial OrganizationVol. 3 (1985), pp. 451-466.Froeb, L. and G. WerdenEndogeneity in the Concentration-PriceRelationship: Causes and Consequences\" Department of Justice WorkingPaper No. EAG 91-, 1991.Green, E., and R. PorterNon-cooperative Collusion Under ImperfectPrice InformationEconometricaVol. 52 (1984), pp. 87-'100.Harrington, J. Jr.The Determination of Price and Output Quotas in aHeterogeneous CartelInternational Economic ReviewVol. 32 (1991),pp. 767-792.Heaver, T.The Structure of Liner Conference RatesIndustrial EconomicsJournal ofVol. 21 (1973), pp. 257-265.Herman, A.Shipping Conferences(Deventer: Kluwer, 1983).Holmes, T.The Effects of Third-Degree Price Discrimination inOligopoly,American Economic ReviewVol. 79 (19), pp. 244-250.Jansson, J.Intra-Tariff Cross-Subsidization in Liner Shipping, \" Journalof Transport Economics and Policy, Vol. 8 (1974), pp. 294-311.and D. Shneerson, liThe Effect of Capacity Costs and DemandElasticities on the Structure of Liner Freight Rates, II Logistics andTransportation ReviewVol. 22 (1986), pp. 3-25.Judge, G., Hill, R., 1982). , Griffiths, W., Lutkepohl, H., and T-C. LeeIntroduction to the Theory and Practice of EconometricsJohn Wiley & Sons(New York:, \", \", \", \", \", \", \", \", \"Katz, M.Price Discrimination and Monopolistic CompetitionEconometricaVol. 53 (1984), pp. 1453-1472.Orr, D. and P. MacA voy, \"Price Strategies to Promote Cartel Stability,EconometricaVo1. 33 (1965), pp. 186-197.Osborne, M. and C. PitchikProfit-Sharing in a Collusive Industry,European Economic ReviewVol. 22 (1983), pp. 59-74.Pirrong, S.An Application of Core Theory to the Analysis of OceanShipping MarketsJournal of Law and EconomicsVol. 35 (1992), pp.-131.Radner, R.Excess Capacity and Sharing Profits: A Game-TheoreticApproach\" mimeo, Bell Laboratories, 1977.Reitzes, J. Ocean Shipping Economics: CommentContemporaryPolicy IssuesVol. 11 (1993), pp. 81-85.Rotemberg, J. and G. SalonerA Supergame-Theoretic Model of PriceWars During BoomsAmerican Economic ReviewVol. 76 (1986), pp.390-407.Schmalensee, R. Competitive Advantage and Collusive OptimaInternational Joumal of Industrial OrganizationVol. 5 (1987), pp. 351-367.Inter-Industry Studies of Structure and Performance,\" in R.Schmalensee and R. Willig, eds.The Handbook of IndustrialOrganizationVol. 2., (Amsterdam: Elsevier Science Publishers, 19),pp. 951-1009.Sjostrom, W., \"Price Discrimination by Shipping ConferencesTheLogistics and Transportation ReviewVol. 28 (1992), pp. 207-216., \"Collusion in Ocean Shipping: A Test of Monopoly and EmptyCore ModelsJournal of Political Economy, Vol. 97 (19), pp. 1160-1179.Stigler, G.A Theory of Oligopoly,Journal of Political Economy,Vol. 72 (19), pp. 44-61.Telser, L.Economic Theory and the Core(Chicago: University ofChicago I~Jess, 1978).Zerby, J. and R. ConlonJoint Costs and Intra-Tariff Cross-Subsidies:The Case of Liner Shipping,Journal of Industrial EconomicsVol. 31(1983), pp. 383-396. TABLE 1Definitions of Variables and Data SourcesVARIABLEfreight rate per container for commodity on route conference rate data collected by the Federal Maritime Commission)(fromconference market share on route (from Lloyd's MaritimeInformation Services)Pijvalue per container of commodity on route (from export andimport data collected by Bureau of the Census and compiled by theMaritime Administration of the U.S. Departmenrnf Transportation)Herfindahl index of market concentration on route Maritime Information Services)(from Lloyd'CAP..total capacity on route (measured in TEU'Lloyd's Maritime Information Services)- i., containers) (fromy,dgross domestic product in destination country on route (in dollars)Fund) (from International Financial Statistics of the International Monetaryprice index for destination country on route (converted into dollars)(from International Financial Statistics of the International MonetaryFund)price index for origin country on route (converted into dollars)(from International Financial Statistics of the International MonetaryFund)dummy variable denoting whether conference on route j allows members to independently enter into service contractsits1985198619871988 = dummy variables for years 1985-TABLE 2Summary Statistics For Entire Sample(620 observations)VariablemeanstandardnnmmummaximumdeviationFreight Rate104598722239(Revenu~ perTEU)Commodity814628568-415493ValueConferenceMarket ShareMarket906Concentration990528285(Herfindahl)Capacity5800266933654048461Destination848836160Country GDP435(in billions ofdollars)Destination118Country Price100190IndexOrigin Country124100190Price Index*(*TABLE 3Estimation of the Determinants of Ocean Freight RatesVariableCoefficient(t-value)Conference Market Share 1.23)Conference Market Share x Commodity Value (sjlu000017(0.70)Market Concentration (~)-0.026(-0.31)Market Concentration x Commodity Value (Bpij0000018(1.80)*Capacity (CAP)-0.000032(-0.17)Destination Country GDP (J1d)(1.07)Destination Country Price Index (P'(1.14)Origin Country Price Index (P'(2.93)**Independent Action Dummy (IA590.4709)**1985243.(2.05)**1986-43.(-0.78)198770.1.14)1988129.461.14)number of observations620(degrees of freedom)(452)*) indicates significance at the (5 '0) level in a two-tailed testTABLE 4The Effect on Freight Rates of a One-Standard-DeviationIncrease in Market Concentration(Evaluated at Varioos Commodity Values)Commodity Value5680712345415493(in dollars)percentile * 25%50%75%100%Change+12+41+277dollarsfreightrates:asa %0.3 %+0.1 %+0.4%+ 1.3%+8.of themeanfreightratepercentage of commodity values in the sample that lie below thespecified dollar level (i.e., $568 ($415493) is the minimum(maximum) commodity value)TABLE 5~ary Statistics For Samples of Inbound and Outbound RoutesVariablemeanstandardminimummaximumdeviationInbound (352 obs.Freight Rate48267399114,239(Revenue per TEU)Commodity Value19380,512178415,493Conference MarketShareMarket Concentration956116528285(Herfindahl)Capacity623862725,99750,49698Origin Country140100190Price IndexOutbound (268 obs.Freight Rate607344722087(Revenue per TEU)Commodity Value12353,638568276754Conference MarketShareMarket Concentration841792544127(Herfindahl)Capacity522,4504,83272,687048,461Destination Country835711160849GDP(in billionsof dollars)Destination Country136100190Price IndexTABLE 6Estimation of the Determinants of Ocean Freight Rates on Inbound and Outbound RoutesInboundOutboundCoeff.Coeff.Variable(t-value)(t-value)Conference Market Share 1.811.58)(-0.56)Conference Market Share x000017000047Commodity Value (sRij(0.55)(1.02)Market Concentration (~)039-0.(0.35)1.09)Market Concentration x0 .000002500000o61Commodity Value (HRij(1.85)*(0.40)Capacity (CAlj)-0.000340003529)(1.21)Destination Country GDP (J1d(1.13)Destination CountryPrice Index (pi)(0.79)Origin CountryPrice Index (P'(2.15)**Independent Action Dummy (IA)284.1.06)1985149.1.(0.81)(1.54)1986112.1.52)(-0.15)198725.70.(-0.26)1.00)198811.21109.(-0.07)(-0.93)number of observations352268(degrees of freedom)(254)(1)*) indicates significance at the 10%level in a two-tailed testTABI,E 7The Effect on Inbound Freight Rates of a One-Standard-DeviationIncrease in Market Concentration(Evaluated at Various Commodity Values)Commodity Value178172625765415493(in dollars)percentile *25%50%75%100%Change +24+33+50+97+439dollarsfreightrates:asa %+0.+ 1.0%+ 1.4%+ 2.8 %+ 12.of themeanfreightratepercentage of commodity ,values in the sample that lie below thespecified dollar level (i.e., $8,178 ($415493) is the minimum(maximum) commodity value)