a Disaggregate Analysis of Port Selection (2004)

Transportation Research Part E 40 (2004) 317–337www.elsevier.com/locate/tre A disaggregate analysis of port selection Matthew B. Malchow *, Adib Kanafani 1 Department of Civil Engineering, Institute of Transportation Studies, University of California, 109 McLaughlin Hall, Berkeley, CA 94720, USA Received 15 January 2003; received in revised form 3 March 2003; accepted 16 May 2003 Abstract With this article we use an alternative form of the discrete choice model to analyze the distribution of maritime shipments among US ports. We model the distribution as a function of the characteristics that describe each shipment and each port. We assume that vessel schedules are fixed in the short-term and examine the assignment to ports for exports of various commodity-types as a function of geographic location, port characteristics, and characteristics of vessel schedules. We find that the most significant characteristic of a port is its location. We show also how the market share predicted for a port can be expected to vary with each commodity-type and each carrier, and we show how the choice process varies for discretionary cargo.
2003 Elsevier Ltd. All rights reserved. Keywords: Port choice; Shipper behavior; Carrier behavior; Shipment routing 1. Introduction Competition between ports has intensified. As a result of containerization, which standardized the transfer process for shipments between ocean and surface transportation, and deregulation, which allowed maritime carriers to set contracts with rail services and establish rates independent of location, the area considered a portÕs hinterland disappeared. Apart from their various marketing efforts, ports compete primarily through their investment program. Ports are improving intermodal facilities to minimize the dwell time of shipments, and they are increasing the storage * Corresponding author. Tel.: +1-510-231-9460; fax: +1-510-231-9565. E-mail addresses: [email protected] (M.B. Malchow); [email protected] (A. Kanafani). 1 Tel.: +1-510-642-3585; fax: +1-510-642-1246. 1366-5545/$ - see front matter
2003 Elsevier Ltd. All rights reserved. doi:10.1016/j.tre.2003.05.001 318 M.B. Malchow, A. Kanafani / Transportation Research Part E 40 (2004) 317–337 space available to terminal operators to allow carriers to concentrate operations. Ports are also dredging their waters so that carriers may deploy larger vessels. By investing to attract business, ports recognize that carriers make two primary decisions, the long-term decision being the deployment of vessels to routes and ports, and the short-term decision being the assignment of shipments to vessels. Thus with the assignment of a shipment to a vessel comes the assignment of a shipment to a port. In this work we apply a choice model to model the assignment of each shipment to a vessel/port and hence to evaluate the competition between ports. We answer three questions: • What factors influence a carrierÕs selection of a port for a shipment? • In what manner and across what domain do ports compete? • What strategies might a port follow to increase its market share? Though the choice-model approach used in this research has not previously been applied to port selection, our approach has been influenced by earlier work. Qualitative analysis of port competition has been done by, among others, Bardi (1973), Foster (1978a,b, 1979), Slack (1985), Hanelt and Smith (1987), and DÕEste and Meyrick (1992a,b). The results have not always been identical, but the authors often suggest that service-related factors were more important than price factors, and that factors within the control of port authorities were often less important than those beyond port control. Brooks (1984, 1985) noted the difference between a characteristicÕs importance and salience. Quantitative analysis can differ between the two, as the ranking of characteristics comes not from word-of-mouth but the results of actions. The scheduling of carriersÕ vessels has also been the subject of much research. Kenyon (1970) and Al-Kazily (1979) explored the development of a carrierÕs maritime network. Hayuth (1981) suggested the formation of a load center by carriers, while Foggin and Dicer (1985) and Slack (1996) evaluated the effects of load centers. Helmick (1994) sought quantitative evidence of the formation of load centers but suggested that other factors, e.g. the presence of tramp lines in routes abandoned by major carriers, prevented confirmation of carrier rescheduling. Lago et al. (2001) found that the rescheduling of vessels by carriers was not drastic but did differ between corridors. They showed how the level at which scale economies were exploited in oceanic transit differed between corridors. Economic models of carrier behavior have come from three directions. First, linear programming to optimize the assignment of vessels has been advanced by Benford (1981), Perakis (1985), Lane et al. (1987), and Perakis and Jaramillo (1991). These models show how the distribution of vessels might be affected by the distribution of traffic but deal with simplified scenarios or are difficult to apply. Second, cost models have been estimated by Jansson and Shneerson (1978), Talley (1990), and Lim (1998) for the general cargo or container shipping industry. Garrod and Miklius (1985) and Jansson and Shneerson (1985) emphasized the importance of shippersÕ inventory costs. Griffiths (1976a,b) measured the optimal size of ports for a given level of traffic. de Neufville and Tsunokawa (1981) confirmed that scale economies existed at ports. Bendall and Stent (1987, 1988), Talley (1988a,b, 1994), and Tongzon (1995) all suggested, or measured, characteristics of the cost function faced by terminal operators. Finally, Allen (1977), Daughety and Inaba (1978) and Daughety (1979) advanced the economic modeling of the decisions made by carriers, in contexts other than the assignment of shipments to ports. Winston (1981a,b) and Nam 2 The Harmonized System is an international six-digit commodity classification developed under the auspices of the Customs Cooperation Council. or origin. We restrict shipments to those for which the carrier of record had a schedule listed with the Journal of Commerce. 54). Kanafani / Transportation Research Part E 40 (2004) 317–337 319 (1997) applied discrete choice models to freight transportation decisions. California. The system classifies goods by what they are. the third pair a subheading. it is the attributes of the service as offered by the carrier that influence that choice process. New York. and the United Kingdom. of shipments within our sample set. Malchow and Kanafani (2001) made an initial application of the choice model to the assignment of shipments to individual ports. Individual countries have extended it to ten digits for customs purposes. and to eight digits for export purposes.M. 3 The data set was reduced to include only shipments that were moved from one of the 48 contiguous United States through one of the eight ports by one of the carriers whose schedules are available from the Journal of Commerce. by choosing a carrier. 10). 2 The four commodity-types are bulk materials (HS 25. Germany. Brazil. the next pair a heading. The fundamental assumption is that a shipper. while a shipper may be deliberately choosing a particular port for his or her shipments. The data represent exports to eight foreign countries: Australia. and manufactured goods (HS 85). A. The commodity classifications provide variations in the values (and related characteristics) of the shipments. by incorporating additional variables and applying a form of the choice model that accounts for the correlation within the multiple decisions made by each decision maker. Methodology In this research. We want to capture the impact of such factors. implicitly chooses a port for a particular shipment. We select the eight ports in our choice set for two primary characteristics: (i) the volume of trade moved through the port. factors other than location may influence the choice between them. Oakland. South Africa. To estimate the choice model. use. since we use a carrierÕs schedule to measure particular variables. If two ports were geographically close. California. . Washington. Long Beach. not according to their stage of fabrication. The present article extends the work of Malchow et al. Georgia. New York. and Tacoma. we apply a choice model to the assignment of shipments to vessels/ports in order to evaluate the competition between ports.B. We classify shipments into four commodity-types using the first two digits of the harmonized commodity code (HS). Los Angeles. Malchow. the choice set consists of eight ports: Charleston. Japan. Washington. Table 2 shows the distribution. and (ii) the proximity of the port to other significant ports. Table 1 shows the distribution of shipments by country and commodity-type. South Carolina. 3 For each carrier. and the destination countries provide geographic distribution. Seattle. 08. In other words. California. The first pair of digits represents a chapter. Therefore the assumption is that the shipperÕs preference for a port is wholly subsumed within the preference for and choice of a carrier that offers a service through that port. Saudi Arabia. But in the competitive market of shipping it is the attributes of the door-to-door service that a carrier offers that influences the shipper choice. among ports. 2. fabrics (HS 52. foods (HS 07. Egypt. 26). Savannah. we use data that describe shipments exported from the United States in December 1999. by destination and commodity-type HS code Australia 07 08 10 25 26 52 54 85 127 160 7 63 1 33 25 164 All 580 Brazil Egypt Germany Japan Saudi Arabia 2 8 1 18 4 25 17 61 1 33 0 7 0 0 2 15 13 169 7 85 2 26 35 79 722 836 70 352 13 190 12 254 6 64 2 6 0 18 7 69 136 58 416 2449 172 South Africa United Kingdom Total 2 6 0 23 2 21 18 36 57 141 12 56 7 52 28 162 930 1417 99 610 29 365 144 840 108 515 4434 Table 2 The distribution of shipments among ports Rank Port Shipments 1 2 3 4 5 6 7 8 9 10 Oakland. Remember that when modeling the short-term decision. VA Tacoma.5. Because we analyze exports rather than imports. Malchow. We also assume that sufficient space exists for each shipment on vessels scheduled along each route. Vnj of each port as a linear function of five variables: 4 Vnj ¼ aj þ b1
Onj þ b2
Inj þ b3
Hinj þ b4
Cinj þ b5
Pinj . WA Savannah. let us define the scenario in which a carrier selects a port for a shipment. Empty space suggests that each shipment is exported through the optimal port. First. the carrier selects the port and vessel for each shipment simultaneously. CA Charleston. WA 1314 1010 675 650 618 515 462 346 290 254 Many factors affect the choice process being modeled. Before discussing these factors. Kanafani / Transportation Research Part E 40 (2004) 317–337 Table 1 The distribution of shipments. and carriers often transport empty containers to fill slots on outbound vessels. GA Houston. 4 ð1Þ We modeled the decisions with other variables as well. For example. this situation should hold. Thus. During the 1990s. . TX Norfolk. the ratio of imports to exports fluctuated around 1. we can represent each port by the vessel distribution rather than the characteristics of a particular vessel. the explanatory power of the model decreased. CA Los Angeles. A. we assume that the long-term fleet assignment has already been established. We model the systematic utility. In each case.320 M.B. SC Long Beach. we used the average number of sailing days in place of oceanic distance or the frequency of voyages in place of the headway between voyages. CA New York/New Jersey Seattle. the inland distance from the origin of shipment n to port j (km. 5 For the variable Onj . four factors influence the transit time associated with each port: ii(i) i(ii) (iii) (iv) the distance from the origin to the port. The variables Onj and Inj are of course independent of the carrier. Kanafani / Transportation Research Part E 40 (2004) 317–337 321 where. along with Houston. the time needed to transfer the shipment from the ground to the vessel. we use the schedule of all voyages from one of eight United States ports to any port near the destination. Because data for a vessel is maintained only until the vesselÕs voyage has been completed. shipments destined for Germany would also be affected by the nearby ports of Rotterdam and Antwerp. data for December 1999 were no longer available. We measure the capacity of each vessel scheduled along a corridor. we use the shortest sailing distance from port j to the destination of shipment n. data for March 2000 were used to represent the variables. a. This efficiency is ultimately dependent on transit time and cost. We measure the variables Hinj .M. 7 We measure the variables Hinj and Pinj for each carrier directly. the average headway between voyages by carrier i from port j to the destination of shipment n (days). Cinj . in TEUs. For example. through the website MaritimeData.95 existed between the schedules. This is an approximation of the actual sailing distance. separated by three months) and a correlation coefficient of 0. A complete list is given in Appendix A. the time incurred as the vessel calls at other ports in transit. Miami. Malchow. which should approximate the inland rail distance as well. 1000s). the average size of vessels sailed by carrier i from port j to the destination of shipment n (TEUs. 5 In fact. Hinj . and Portland. the probability that port j would be the last port visited by a vessel sailed by carrier i to the destination of shipment n. Inj . implying that carriersÕ schedules did not change much over three months.joc. For each shipment. Likewise. Onj is the oceanic distance to the destination of shipment n from port j (km. and the oceanic distance from the port to the shipmentÕs destination. Cinj . Norfolk. and Pinj for each carrier through an Internet database maintained by the Journal of Commerce. we use the inland road distance. with the remaining variables being measures of the carrier as well as the port.B. The variables in this choice function are selected to represent the common objective of the shipper and the carrier: to get each shipment from its origin to its destination as efficiently as possible.com/scheds/index. and Pinj . 1000s). 7 These records were downloaded from the web and analyzed with a spreadsheet. b. Comparison was made with the schedule for June 2000 (likewise. . For Inj .com. a result quite different from the model resulting from carrier-specific values. i(ii) the charges assessed by the port. We would intuitively expect the carrier-specific values to have more explanatory power as well. The twelve US ports consisted of the eight within the choice set. Instead.shtml. 6 The site at the time of writing was at: http://www. 1000s). We calculate the variable Cinj to represent the average capacity of vessels sailing along the corridor for carrier i. four factors influence the operating cost associated with each port: ii(i) the inland distance from the origin to the port. A. the headway between (or frequency of) voyages was found to be insignificant when included as the average across all carriers. 6 For each destination. which could not be measured due to the complexity and variability of carrierÕs schedules. coefficients estimated in the model. The foreign ports were not constrained to the country that was the destination of the shipment. Malchow. we must mention those variables that are not included in the model but may be significant on a disaggregate level. we would have to collect data for each terminal operated by each carrier at each port within the choice set. Rates also varied little among East Coast ports. Ocean freight rates are no longer publicly disclosed. who analyzed the selection of mode for shipments. Increased port charges could make a port less attractive to a decision maker. we can estimate the importance of the factors that describe each alternative. He found that the rate charged was in most cases insignificant. Second. These estimates influence the likelihood of each decision by the carrier. suggested that port charges are relatively insignificant. 2000). data are not available for intermodal transfer time. which lacks the variability that would be necessary for inclusion into a choice model. Kanafani / Transportation Research Part E 40 (2004) 317–337 (iii) the oceanic distance from the port to the destination of the shipment. two factors would complicate their inclusion into the model.322 M. so long as the shipment arrives at the destination at the expected time. By observing the decisions for multiple shipments. and perhaps more importantly. For example. Thus. However. For each port. First. with a slight difference between the rates for ports on different coasts. There is also some empirical evidence from Nam (1997). the intermodal transfer process could influence each carrierÕs selection of a port for each . The carrier is observed as selecting one port from among the alternatives. 3. and we estimate a factor (referred to as the port-specific constant) that represents the average utility of all unobserved factors. Because of complexities with port tariffs and the prevalence of service contracts. with the most prevalent port charge being wharfage. suggesting that either service characteristics were more important or rates did not vary by alternative. with the shipments moved by each carrier representing a separate group. however. representing economies-of-scale and density.B. or approximately 100 terminals. Second. rates might not be significant in port selection. and we estimate each factorÕs contribution to maximize the likelihood of the observations. These include port charges. and the intermodal transfer process at each port. even if data were available. a carrier could know a shipmentÕs intermodal transfer time prior to a decision only as an expected value. and (iv) the average vessel size. First. and we assume that the carrier has selected the port that provided the greatest utility in the context of that shipment. we estimate the contribution of each factor to its utility. as of the Ocean Shipping Reform Act (OSRA) of May 1999 (Lewis and Vellenga. fortunately. the cost of the transportation services. An alternate formulation of the choice model Under the traditional choice model. port charges are difficult to measure accurately on a disaggregate level. Industry representatives have. the port should not affect the rate that a shipper is willing to pay for transportation services. For two reasons. A. The data that we model represent panel data. inspection of the tariffs and service contracts available through the Federal Maritime Commission showed that the freight rates prior to OSRA varied little among West Coast ports. a shipper cares little about the intermediary points through which a shipment is moved. the utility of a port for a shipment is a linear function of the variables describing that port. Finally. Correlation likely exists among the decisions of a given carrier. Before continuing. e. would require the estimation of 288 constants (36 carriers. Cinj . 9 In mathematical notation. • winj ¼ 1 if carrier i actually sends shipment n through port j.M. The assignment of each shipment to a port would define each distribution. summed over all feasible distributions D. then the unobserved error term would be correlated for each carrierÕs shipments and not distributed with a mean of zero. and thus the constants should vary by carrier. rather than modeling the selection of a port for each shipment. and 0 for all others. A. Onj . • sij ¼ the number of shipments moved by carrier i through port j (Rn winj ). as required for the model estimation. Malchow. The two constraints specify that • 8in. the number of shipments predicted by the distribution to move through each port must equal the number actually observed as moving through that port. This factor could not be included directly since it does not vary across shipments. Kanafani / Transportation Research Part E 40 (2004) 317–337 323 shipment. 9 The alternative-specific constant is estimated in the logit model such that the predicted share will be equivalent to the observed share for each alternative. Rn dinj ¼ sij . In this model. A distribution that meets these constraints is considered feasible. Chamberlain (1980) introduced an alternative model that accommodates panel data. Hinj . ð2Þ n. we note that the term Vnj has been replaced by b0 n:j xinj winj and Vnk by b0 n:j xinj dinj . For some carriers the number of shipments would be too small for estimation. and • dinj ¼ 1 for each feasible distribution. and 0 otherwise. let • xinj ¼ the vector of attributes discussed earlier that influence the choice by carrier i of port j for shipment n (i. we model a carrierÕs aggregate distribution of shipments from the set of feasible distributions. Two fundamental properties determine the feasibility of each hypothetical distribution. and • 8ij. Rj winj ¼ 1. and the transfer process at each port likely varies by carrier. however. the port-specific constants 8 If the alternative-specific constants remain constant across carriers. The log-likelihood for all observations is then L¼ X i " X ln exp b0 xinj winj n. 8 ports). rather than maximizing the probability that the carrier selects the chosen port for each shipment.j in which D represents all feasible distributions of the shipments. Second. to the logit P In relating this model P model. This. ChamberlainÕs method. Inj . X d2D X exp b0 xinj dinj !# . It thus affects the constant term associated with each port. and Pinj ). maximizes the probability that the carrier selects the observed distribution from the set of feasible distributions. First. each shipment must be transferred through exactly one port. With this formulation.j !.B. . 8 To account for the individual carriers. we could estimate a set of port-specific constants for each carrier. 00 0.09 0.4 – 0.00 0.38 )0.67 )0. when modeled alone.08 0. the port-specific constant that is estimated for each carrier-port combination. of last (P ) )0. A. The estimate for each of the five coefficients is statistically significant at a level beyond 99%. The Chamberlain model To examine these results further.4 4.00 0.9 0.5 )2.16 )0.01 0.11 0.3 )3.47 )0.10 0.23 0 0.04 )0. our expectation about the impact of vessel capacity is not definite. so we instead use sample .8 )25.04 0. The set of feasible distributions for some carriers would be computationally cumbersome.02 0. no coefficients )6242.324 M. space should always be available for exports. only four of the five estimates have the expected sign. Therefore. Given the trade flows in and out of the US. Table 3 shows the results of the estimation.02 0. We learn later that the impact of vessel capacity. the sum of the constants across the feasible distributions does not vary and thus can not influence the choice. 5.10 0.10 – 0. is actually positive.00 A_Charleston A_Long Beach A_Los Angeles A_New York A_Oakland A_Savannah A_Seattle A_Tacoma 0.01 0.09 )0.24 0.45 0. (More precisely. The negative coefficient of vessel capacity is counterintuitive.00 0.02 – Log-likelihood Log-likelihood. constant across carriers.1 13.07 0.10 0.00 0.5 0. we estimate the Chamberlain model. Choice-model estimation We first estimate a standard multinomial choice model.64 0.8 )1. Kanafani / Transportation Research Part E 40 (2004) 317–337 Table 3 Results of the standard multinomial logit model estimation (ignoring panel effects) Variable Estimate Standard error Z-statistic P -statistic Oceanic distance (O) Inland distance (I) Sailing headway (H ) Vessel capacity (C) Prob. However.13 0. Why might this be? Perhaps there is no immediate advantage in placing a shipment aboard a larger vessel if space is available.00 0.05 0.00 )12.2 disappear from the equation. Malchow.8 5. ink ke ð3Þ with aij .) 4. constants only Log-likelihood.6 )9220.00 0.B.08 0.8 )2.6 )35.2 )8650. in which the probability of choosing port j is given by eVinj Pinj ¼ P V . accounting for panel data characteristics Variable Coefficient estimate Standard error Z-statistic P -statistic Oceanic distance (O) Inland distance (I) Sailing headway (H ) Vessel capacity (C) Prob. 1985).01 0.00 0.70 0.00 0. describes the shipment-decisions better than the traditional model.4 2. no coefficients Table 5 Results of the multinomial logit model estimation. we examine the significance of each variable individually with the Chamberlain model.5 0. as well as the fact that the capacity of a vessel might not affect port selection.8 )7533. we estimate the model as shown in Table 5. Table 4 shows the coefficients estimated with the 1840 observations that were retained. 10 We select five-shipment samples for each of the nineteen carriers that had more than 50 shipments. in which the alternative-specific constants differ between carriers.02 0.5 0.9 )17.00 0.7 )16. The sign of each variableÕs coefficient is consistent with the sign estimated in the multinomial model for each variable except vessel capacity. We also use HausmanÕs test to confirm that the Chamberlain model.05 0.6 sets of the shipments to represent each carrier. of last (P ) )0.00 0.78 )0.02 0.9 )34.4 )6525.003 0. The likelihood-ratio test confirms that vessel capacity is not a significant variable (Ben-Akiva and Lerman. Malchow.00 0.00 0. To understand why. Inland distance provides the greatest explanatory power.001 )15. LIMDEP is the statistical package used to estimate the choice model.1 2.01 )2980. A.5 )6525. .001 )15.01 0.00 0.8 )7533. Due to this inconsistency. accounting for panel data characteristics and ignoring vessel capacity Variable Coefficient estimate Standard error Z-statistic P -statistic Oceanic distance (O) Inland distance (I) Sailing headway (H ) Prob.6 )34. given that space is available. of last (P ) )0. a number too large for us to begin with for consideration. 10 For example.003 0.6 Log-likelihood Log-likelihood from constants Log-likelihood.03 )0. Ignoring vessel capacity.77 )0.12 )0. Kanafani / Transportation Research Part E 40 (2004) 317–337 325 Table 4 Results of the multinomial logit model estimation.01 Log-likelihood Log-likelihood from constants No coefficients )2980. These shipments could be distributed among the eight ports in one of 8527 distributions.2 )0. 11 We discarded sixty observations because the number of feasible distribution was too large for LIMDEP to handle.02 0. Evergreen represented 527 shipments within our dataset.00 0. 11 Each coefficient is significant with the exception of vessel capacity. we remove vessel capacity from further models. with 100 samples collected for each of these carriers.03 0.B.M.12 )0. For these shipments.10 0.7 )8.75 )0. expected headway 0.00 0. we estimate a model using only shipments that originated in the central United States.B. inland distance would not vary as much among ports. closer examination of the variables reveals that the variables do play different roles for the discretionary cargo.0 The increase in the probability of being the last port that would be equivalent to a reduction of 1000 km in oceanic transit 258.00 0.4 32. Midwest shipments 15. For discretionary cargo.326 M.16 5. We compare the impact of this to the impact of other variables in Table 7 and find that Table 6 The Chamberlain logit model estimated for discretionary cargo Variable Coefficient estimate Standard error Z-statistic P -statistic Oceanic distance (O) Inland distance (I) Sailing headway (H ) Prob. the results of this model appear similar to those from the model for all shipments.5 equivalent to a reduction of 1000 km in inland transit bI =bP bH =bP bO =bI bO =bH bI =bH Value.01 0.9 . cargo originating in a region that does not contain a port. oceanic transit 3. Kanafani / Transportation Research Part E 40 (2004) 317–337 6.00 0.03 0.0 0.05 0.00 )3820. allowing the impact of other variables to increase.2 Log-likelihood Log-likelihood from constants No coefficients Table 7 The importance of being the last port visited for discretionary cargo Ratio Meaning bO =bP 41.9 2.9 5. Discretionary cargo Here we postulate that the decision made for discretionary cargo.019 0. the probability of being the last port visited is significantly more important.1 91. At first glance.8 0.29 )1.9 )681. Malchow. of last (P ) )0.3 The increase in the probability of being the last port that would be equivalent to a reduction of 1000 km in inland transit 11. Table 6 shows the results.004 )11.0 The increase in the probability of being the last port that would be equivalent to a reduction of one day. differs from that made for cargo originating in a portÕs hinterland.16 The decrease in inland distance (km) that would be equivalent to a reduction of one km.7 The decrease in headway (days) that would be equivalent to a reduction of 1000 km in oceanic transit The decrease in headway (days) that would be 23. A.3 )16. However. all shipments Value.0 )2275. To see how the decision process might differ for discretionary cargo. We find that the average shipment size for a commodity decreases as its average value increases. for discretionary shipments as it is for all shipments. Though we wish to emphasize the magnitude of this variableÕs changing importance.M.7 9. Because the declared value of each shipment is confidential. when compared to other factors. a low-valued commodity being sent from Table 8 Characteristics of the shipments from different commodity groups Commodity # Records Shipment size (metric tons) Average value ($/metric ton) Bulk Fruits and vegetables Fabrics Manufactured 610 2347 509 840 53. truck) that are faster and more expensive than water-based transportation.6 27. 7. Each estimated coefficient has a standard error. Carriers would more likely send lower-valued commodities through nearby ports. A.5 1885 . The shipment size for each shipment corresponds to that filed with the shipmentÕs customs form. This evidence of shifting values supports this idea. relative to the distances. The importance of being the last port visited is three times as large. Commodity-specific models We now consider the proposition that the importance of attributes varies with commodity-type. and we expect the transit time to be less important relative to the operating costs for lower-valued commodities. Malchow. We confirm the statistical difference of these probabilities with the construction of density functions in Malchow (2001b). From the results in Table 7 we also see that a portÕs share of shipments from the Midwest is less affected by the headway between voyages. The negative impacts of distance are the associated transit time and operating costs. A ratio between two estimates has an even larger standard error. The commodity groups examined and their characteristics are shown in Table 8. the Journal of Commerce estimated the value of each shipment according to the trade route and commodity code associated with each shipment.087 All 4434 30.B. For example.9 30. Inland distance is covered by modes (rail. likely to minimize shippersÕ inventory cost. shippers of lower-valued goods would place a lower priority on oceanic distance than on inland distance. we must recognize that uncertainty exists in our estimated values. thus. Many industry analysts suggest that discretionary cargo in particular is sent through the port visited last by a vessel to minimize transit time. and the ratio between two ratios has a still-larger error. Kanafani / Transportation Research Part E 40 (2004) 317–337 327 the relative importance of other variables decreases significantly.8 285 1198 4287 11. We expect the importance of different attributes of each port to vary with the characteristics that describe each shipment. The additional day or so of headway becomes less significant when combined with an additional 48 h of inland transit. 7 0.02 )0. As expected. Again using the likelihood-ratio test. With carrier-specific models. 1996) have used the classification of shipments by commodity-type to show that the transit freight rate.4 )3.01 0.05 )0. a minor exception perhaps due to the perishability of fruits and vegetables.00 0.00 0. Bryan. Kanafani / Transportation Research Part E 40 (2004) 317–337 Table 9 Results of the Chamberlain model as estimated for the different commodity-types Commodity Variable Coefficient estimate Standard error Z-statistic P -statistic Fruits and vegetables (HS 07.12 )0. we find that the model estimated for each commodity is significantly more explanatory than the generic model.05 0. of last (P ) )0.01 0.03 0.00 0.4 3.5 )4.51 )0.00 0.328 M.2 )7.3 )17.47 )0.04 0. Table 9 shows the results of the estimation for each commodity. the marginal rate of substitution between inland and oceanic transit generally increases with the commodity value.00 0. of last (P ) )0.00 0.04 0.7 )11.00 0.01 0.58 )0.12 )0.79 )0. we noted earlier that the rates for commodity-types tend not to vary between individual ports within coastal ranges.4 )20.5 )2.00 )8.43 0.16 0.05 0.04 0.00 0.38 )1.01 0.00 0.01 0.00 0.00 )10.B. 54) (800 simulated shipments) Oceanic distance (O) Inland distance (I) Sailing headway (H ) Prob.00 0.00 Fabrics (HS 52.00 Bulk (HS 25) (1100 simulated shipments) Oceanic distance (O) Inland distance (I) Sailing headway (H ) Vessel capacity (C) Prob.01 0.0 3. .00 0.12 0. on a unit basis. 1974. 1972.00 Manufactured (HS 85) (1300 simulated shipments) Oceanic distance (O) Inland distance (I) Sailing headway (H ) Vessel capacity (C) Prob. whereas a higher-valued commodity might be transshipped via landbridge to a waiting vessel on the East Coast. This condition becomes important when examining the shares for certain commodities from carrier-specific models. of last (P ) )0. of last (P ) )0.5 )6.00 )23.3 5. 12 12 Certain studies (Heaver.03 0.26 )0.7 0.7 0.1 )13.00 0.0 )22. Table 10 shows the marginal rate of substitution between oceanic distance and inland distance for each of the different commodities.00 0.01 California to the United Kingdom would be loaded at a California port and sent on an extended ocean voyage. However. we can examine how a portÕs share is affected by distance and how this impact varies with the value of the commodity. tends to vary with the density of each commoditytype. 08) (1500 simulated shipments) Oceanic distance (O) Inland distance (I) Sailing headway (H ) Prob. Malchow.00 0.7 )6.00 )9.01 0.00 0. A.04 0. and the results do agree with our expectations. Brooks and Button.00 0.00 0. This suggests that traffic forecasting models use commodity-specific data to enhance accuracy of predictions.3 0. 9.56 1885 0. The logit model allows infinitesimally small probabilities to exist.15 0.37 0. we see that distance influences port selection most. subject to the constraint that the variable-specific coefficients equal those estimated for the combined data set. if for anything to estimate the implications of rational behavior. Kanafani / Transportation Research Part E 40 (2004) 317–337 329 Table 10 The importance of inland and oceanic transit for each commodity-type Commodity Average value ($/metric ton) The marginal rate of substitution between inland transit and oceanic transit Bulk Fruits and vegetables Fabrics Manufactured 285 1198 4287 11. in other words.087 0. but the predicted share of these alternatives would be zero. we find that the generic model would be rejected for the model estimated for each carrier. we could use the elasticity of choice.M. A. Ports can consider marketing to improve their position in an 13 Recall that to allow the variable to be assigned a finite value. The italicized ports for each carrier are those through which the carrier transported no shipment. However. as described in Train (1986). The probability of being the last port appears to have a very inelastic effect.23 0. American President Lines (APL) and Maersk/SeaLand. Note also that the elasticity with regard to being the last port visited would likely increase were the cargo of the discretionary sort. We estimate the model for two carriers. We first estimate models with port-specific constants for each carrier.B. Carrier-specific models To measure the effect of each variable. The carrier-specific models allow analysis of the share of traffic for each port. 13 With the likelihood-ratio test. If this were true then one might question the worth of models of the type presented here. Malchow. From these estimates. We should be able to model this process to some extent. Port managers could use such models in marketing. ports for which the observed frequency of sailing during March 2000 was zero were assigned an arbitrarily high headway of sixty days. and the effect of the headway between voyages is largest for ports not visited at all. a fundamental belief underlying economic analysis is the rational behavior principle. Carriers will in most cases make rational decisions and it might be that some decisions do not require the same level of analysis. the behavior of each carrier is not consistent. implying that the group deciding often does so without much evaluation. The estimated elasticities are given in Table 11. . Discussion One interview with a carrier suggested that the selection of a port is not entirely predictable.16 All 8. 15 To simplify the analysis.05 0.00 0.90 )1. One hypothetical shipment would be destined for Japan.03 )2.48 )1.29 )1.73 )1. Thus.00 0. .12 0.05 )2. Malchow.40 )1. to be moved by APL. but the predicted advantage is not sufficient to ignore competition.36 )0.24 )1.22 )0.97 )1.18 )1.29 )0.00 0.38 )1.10 )0.35 )1. each port does hold an advantage for shipments within its hinterland.04 0. The predicted market share of the Port of Oakland decreases from 64% within its hinterland to 53% as the origin of the shipment moves inland. and Seattle or Tacoma. Charleston. In addition. we create simpler environment with three of four ports: 14 (1) (2) (3) (4) Los Angeles or Long Beach.00 0.22 )1.16 )0.02 established market or enter a new market.26 )0. 14 The model applies regardless of the available choice set.53 )1.84 )1.330 M.48 )0.67 )1. To show how this model could be used.02 )1.13 )1. of last (P ) APL Cha LB LA NY Oak Sav Sea Tac )1.00 0. 15 The independent variable represents the shipmentÕs origin and moves inland from the Port of Oakland. Oakland.01 )1.10 )1. Kanafani / Transportation Research Part E 40 (2004) 317–337 Table 11 The choice elasticities for the individual ports.17 )1.33 )0. We examine the competition of these ports for a shipment under various scenarios.02 0.23 )0. 1 shows the market share for each port in this scenario.18 )0. Estimates of the effect of certain factors could be used to assess the worth of port investments.00 0. the variable representing the probability of being the last port selected is ignored.17 )0.10 )1.00 Maersk Cha LB LA NY Oak Sav Sea Tac )1.17 )1.48 )1. for two carriers Carrier Port Oceanic distance (O) Inland distance (I) Sailing headway (H ) Prob. because carriers do not operate multiple terminals within a region.08 )2.B.22 )2. Fig.00 0.04 0. and we estimate the distance to each port geometrically. A.18 0.06 )1.62 )1. The market share of competing ports increases as expected to account for this lost share.43 0.52 )1. each port competes with ports in other regions for the assignment of a shipment.02 )1.41 )0.14 )1.43 )1.03 0.11 )1.15 )0.01 0. 17 The generic commodity would be of lower value than the highest-value manufactured goods. we estimate a model for P&O Nedlloyd. Further analysis shows that the potential for competition. Fig. 17 This confirms that lower-valued goods are more likely to be loaded at a neighboring port and transited a longer distance via ocean. For this example. Clearly. is impacted by the value of the good being shipped. In this figure. . Malchow. however. 2. which is located along the opposite coast. Initially. In another hypothetical case.B. the export of a shipment by Maersk to Japan. We suggested earlier that a carrier would transport higher-valued shipments to a distant port to minimize oceanic distance relative to inland distance. a portÕs competition with a landbridge-alternative is greater for higher-valued commodities.M. we replace the Port of Oakland with the Port of Charleston. Kanafani / Transportation Research Part E 40 (2004) 317–337 331 100% 90% 80% 70% Share 60% 50% 40% 30% 20% 10% 0% 0 500 1000 1500 2000 2500 3000 3500 4000 Distance inland from Oakland (km) Oakland Seattle Los Angeles Fig. The impact of inland origin location on portsÕ market shares for a theoretical shipment moved by American President Lines to Japan. 1. and Charleston does not steal significant market share until the origin has shifted even further. while the Port of Tacoma attracts a small market share away from the Port of Long Beach. 16 The Port of Charleston does not begin to steal significant market share away from the West Coast ports until the origin of the shipment has shifted halfway across the country. market share for the Port of Charleston remains insignificant. The impact of inland distance becomes more apparent when comparing market shares of ports along different coasts. time-sensitive goods are more likely to be shipped via landbridge to a port with greater 16 The Port of Long Beach is closer to Oakland than Tacoma. 3 shows this. the market share that would be captured by each of three ports is shown in Fig. and highervalued. A. as the share predicted for the Port of Charleston for a shipment bound to Japan is shown to differ for a manufactured good relative to the share for a generic shipment. so with the shifting of the origin. Long BeachÕs advantage is lessened. 3. we find that its impact is again greatest when comparing ports from opposite coasts. The market share predicted for an Atlantic port for the shipment of different commodity-types by Maersk to Japan.332 M.B. Ports adjacent along one coast would have near-equivalent oceanic distances. access to the shipmentÕs destination. such that an advantage could be gained only when linking calls for the scheduling of a vessel. Additional models could be evaluated to get more precise estimates. 2. . Kanafani / Transportation Research Part E 40 (2004) 317–337 100% 90% 80% Market share 70% 60% 50% 40% 30% 20% 10% 0% 0 500 1000 1500 2000 2500 3000 3500 4000 Distance inland from Oakland (km) Tacoma Charleston Long Beach Fig. A. 100% 90% Market share 80% 70% 60% 50% 40% 30% 20% 10% 0% 0 500 1000 1500 2000 2500 3000 3500 4000 Distance inland from Oakland (km) All Manufactured Fig. Malchow. In evaluating oceanic distance. The impact of inland origin location on portsÕ market shares for a theoretical shipment moved by Maersk to Japan. additional voyages do little but increase the capacity available for shipments. Fig. or 4+ per month) have been scheduled. 4. Recall that the significance of being visited last is greatest with discretionary cargo. an incremental voyage adds insignificantly to a portÕs market share. The potential impact becomes more apparent when observing the predicted market share as a function of frequency. Though the market share decreases steadily as headway increases. As mentioned earlier. when competing with Los Angeles and Seattle. the Port of Oakland could increase its predicted market share for discretionary cargo from 24% to 85%. for shipments transported by APL from Kansas to Japan. Fig. Fig. 5 shows the impact of sailing frequency on the distribution of shipments between ports. of which it is the decision makerÕs objective to minimize. 6 represents the share predicted for the Port of Oakland. we analyze the importance of being the last port visited by a vessel. so long as a port has voyages scheduled at the frequency of one per week or greater. the actual impact is not as dramatic. To decrease the headway from sixty days to thirty days. The predicted impact of headway between voyages on portsÕ market shares for theoretical shipments moved by APL from Oregon to Japan. the significance of being visited last is much greater for a port with discretionary cargo. but a fifth sailing would reduce the headway by only one day. In a final example. .M. 18 The impact of sailing frequency decreases quickly once a minimum number of voyages (on the magnitude of one per week. Therefore. A. Once a carrier has scheduled a sufficient number of voyages. A second voyage reduces the average headway again from thirty days to fifteen days. a carrier needs to add only one voyage per month.B. Simply by convincing APL to make all of its last calls there. and a third voyage by five days. Malchow. 4 represents the significance of the headway between voyages for a theoretical shipment moved by American President Lines from Oregon to Japan. Kanafani / Transportation Research Part E 40 (2004) 317–337 333 100% 90% Market share 80% 70% 60% 50% 40% 30% 20% 10% 0% 0 10 20 30 40 50 60 Headway between voyages at the Port of Oakland (days) Oakland Los Angeles Seattle Fig. An additional voyage would reduce the expected headway by an insignificant amount. A second voyage (if the voyages were spaced evenly) would reduce the headway by fifteen days. 18 We used headway in the choice model because headway is more linearly related to the time spent by a shipment in transit. 6. with particular regard to the . 10. A. We have found that the variables furthest from the control of port authorities. the oceanic and inland distances. Kanafani / Transportation Research Part E 40 (2004) 317–337 100% 90% 80% Market share 70% 60% 50% 40% 30% 20% 10% 0% 0 5 10 15 20 25 30 Voyages at the Port of Oakland per month Oakland Los Angeles Seattle Fig. We have found other factors to be significant. for theoretical shipments moved by APL from Kansas to Japan. Conclusion These results have in many ways reaffirmed the results of earlier qualitative analysis. Malchow. The impact of being visited last on the market share predicted for the Port of Oakland.B.334 M. but more so in the context of discretionary cargo. 100% 90% Market share 80% 70% 60% 50% 40% 30% 20% 10% 0% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Probability of being last visited (%) All Discretionary Fig. 5. The predicted impact of sailing frequency on portsÕ market shares for theoretical shipments moved by APL from Oregon to Japan. have the greatest impact on carriersÕ distribution of shipments. Port managers could recognize these factors when marketing themselves to carriers. J. 337–343. 97–106.. References Al-Kazily.M.B. H. H. Freight transportation demand revisited: a microeconomic view of multimodal. M. Of interest with future research would be changes that have occurred to the choice process through time as a result of technological change or the development of linked transportation networks..J. An alternative theoretical approach to the evaluation of liner shipping: Part 1... Stent. When forecasting traffic for a port or marketing to carriers.F. A simple approach to fleet deployment. Allen. Brooks. Choice of technology for containerized shipping––a study of interaction between vessels and ports. multicharacteristic service uncertainty and the demand for freight transport. The findings. We must remember also that the distribution of shipments across ports is but one part of the larger picture. A. the exports analyzed here represent only half of port traffic.R. Analysis of covariance with qualitative data.B. Benford... 23–29.F. 1973. in the context of imports. 1987. but it could affect imports if shippers are using terminal space for storage. Estimating service-differentiated transport demand functions.. 1988. analysts should recognize these differences. Choice criteria. 1974. Bardi.B. Transportation Research B 13 (4). Maritime Policy & Management 15 (2).B. in the context of shipment assignment. 35–43. due to imbalances and carrierÕs desire to move shipments quickly. Situational factors. Maritime Policy & Management 8 (4). Inaba.E.. The scheduling of vessels. Daughety. M. An alternative theoretical approach to the evaluation of liner shipping: Part 2.. 281–288. Journal of Transport Economics and Policy 8 (2). Transport Logistics 1 (1).S. K. In another important finding. many of the investments being made might not be necessary. 225–238. MIT Press. 9–14. Transportation Research 11 (1). Discrete Choice Analysis: Theory and Application to Travel Demand. 1996.. We could expect the decisions to be similar for imports.. 1979. Chamberlain. Maritime Policy & Management 14 (4). The demand for freight transportation: a micro-approach.F. 145–155. 1985. Daughety. F. H. M.F. Kanafani / Transportation Research Part E 40 (2004) 317–337 335 location their port holds in a vesselÕs schedule. the first port visited.. A. Ben-Akiva. Button. The determinants of shipping rates: a North Atlantic case study. Managers might recognize that. Malchow. A. 161–173. Dissertation submitted to the University of California at Berkeley. 1984. Brooks. 1980. W. Brooks. The capacity available on vessels could be of more importance for imports. 223–228. Transportation Journal 13 (1). On measuring cargo handling productivity. MA.. M. Maritime Policy & Management 12 (2). due to the trade imbalance between imports and exports. 1978. Lerman. Carrier selection from one mode. A.. Transportation Research Record 668. Bendall. Review of Economic Studies 47.. Stent. G. Cambridge. I. S. 1979.. Bryan. 23–30. Further examination could be made of the carriersÕ selection of the last port to visit or. . An equally significant (if not more significant) task faced by the carriers is the assignment of vessels to particular routes. In addition. E. could affect the future predictions for port traffic. 1981. Part 2. Maritime Policy & Management 11 (1). coupled with our analysis of shipment distribution.. except for certain factors.. choice behavior varies significantly across carriers as well as commodities. Bendall. 1977. 1985.. One such factor is the desire of importers to concentrate traffic around a particular distribution center. A. 21–30. Cargo handling productivity. Regression analysis of ocean liner freight rates on some Canadian export routes. Storage space should not affect the decision for exports. Maritime Policy & Management 8 (2). Maritime Policy & Management 18 (3). 127–138. 160–176. and solution approaches.B. Heaver.. R. The ocean shipping reform act of 1998. 1970. 1988b. D. W. Containerization. Carrier selection in a RO/RO ferry trade: Part 2. 293–303. G. Jansson. 1991. Miklius. D. Malchow. 1985. 2001. 27–34. 147–155. 361–373. B.... A... A model of scheduled liner freight services: balancing inventory cost against shipownersÕ costs.M. 3–28.. Uyeno. 121–129. Journal of Transport Economics and Policy 19 (1). A second look at fleet deployment.N. 41–43. 1988a. ChiltonÕs Distribution Worldwide 77 (1). DÕEste. . 32–36. 115–126. Trans-Pacific trade. 1978. Maritime Policy & Management 17 (3). Meyrick. Perakis. I. Elements in inter-port competition in the United States. Unloading facilities at an iron-ore terminal.. Slack.. inter-port competition and port selection. Hanelt. M. 287–293. Maritime Studies and Management 3 (4). Y. University of Miami. Logistics and Transportation Review 8 (1). D. 361–373.336 M.. 2000. A. A. Fleet deployment optimization for liner shipping: Part 1. J. 1985.K. University of California. Foggin. 1972. ChiltonÕs Distribution Worldwide 77 (1). Shneerson. Conceptual framework for the decision process.. Hayuth. Foster.K.. 209–214. W. 1976a. 1998. A disaggregate analysis of factors influencing a portÕs attractiveness. Lewis. liner shipping. The Logistics and Transportation Review 21 (3). B.. Maritime Policy & Management 25 (4). Containerization and the load center concept. Maritime Policy & Management 15 (2). Maritime Policy & Management 19 (2).. An analysis of carriersÕ schedules and the impact on port selection. Talley. T. Concentration and Connectivity in the North Atlantic Liner Port Network.. Jansson. Kanafani. 61–76. Foster.E. 1–24. T. Dissertation submitted to the Department of Civil Engineering. The role of US ocean ports in promoting an efficient ocean transportation system. Kanafani / Transportation Research Part E 40 (2004) 317–337 de Neufville.. Optimal containership size. Kanafani. R.. Maritime Policy & Management 19 (2). Griffiths. Talley.. Transportation Journal 39 (4).D. ChiltonÕs Distribution Worldwide 78 (1). DÕEste..N. 1979. Planning and scheduling for efficiency in liner shipping. T. G.D. 385–391. 83–90. 1992a. Griffiths. Economies of scale in container shipping.I. Disappearing hinterlands: the impact of the logistics concept on port competition.B. S. Lane. J. Malchow. 183–200. Economic Geography 46 (1).. S. Foster. Garrod. Malchow. Transportation Research E: Logistics and Transportation Review 33E (3). 163–167.B. 1970–1990. 109–125. Vellenga.. T.. Maritime Policy & Management 12 (3). Kenyon.D. An analysis of port selection..O. Economic Geography 57 (2). Optimum throughput and performance evaluation of marine terminals.. Maritime Policy & Management 15 (4). Carrier selection in a RO/RO ferry trade: Part 1. 223–231. 1978b. 44–48. and conference rates. 1994. The optimal ship size: a comment. 237–243. K. Background.... 2001. 1976b.. 195–215. Port pricing: a cost-axiomatic approach. Helmick. Smith. A.K.. G. 1994. Maritime Policy & Management 21 (1). Dissertation submitted to the Department of Management and Logistics. Ports: what shippers should look for.K. Talley. Economies of scale of general cargo ships. M. 289–300. Heaver.H.O.. Tsunokawa. Dicer.S.. 1992b. Lim. Maritime Policy & Management 23 (3). D. Productivity and returns to scale of container ports. S. Presentation at 2001 International Association of Maritime Economists Conference.. 327–331. Optimal handling capacity at a berth. Review of Economics and Statistics 60. Perakis.. Talley.. 1985. Decision factors and attitudes. Lago. 1987.S. J... 2001.D. 82–91. 1990. J. 1978a. Meyrick. 1981... Maritime Studies and Management 3 (3). T.. A study of the estimation and aggregation of disaggregate models of mode choice for freight transport. Shneerson. Maritime Policy & Management 14 (2). Slack.. Journal of the Transportation Research Forum 26 (1).N. J. 1987..L. P. J. Maritime Policy & Management 12 (4). D.. 1985. W. 1996. 1997. problem formulation. Journal of the Transportation Research Forum 28 (1). 165–175. Ports: what shippers do look for.. 1985. D.. A. M. J. Malchow.. A. Shipping lines as agents in the port industry. Jaramillo. K. Maritime Policy & Management 28 (3). WhatÕs important in a port. W. W. The dynamics of west coast container port competition. D... 1981. Nam.M. Hong Kong. Train. Econometrica 49 (4). Kanafani / Transportation Research Part E 40 (2004) 317–337 337 Tongzon. Transportation Research A 29A (3).E. A.M. C. Determinants of port performance and efficiency.. Malchow. Cambridge. The MIT Press.. Winston. J.B. 1995.L. 1981a. 1981b. 1986. . Qualitative Choice Analysis. A disaggregate model of the demand for intercity freight transportation. 981– 1006. MA. 245–252. C. A multinomial probit prediction of the demand for domestic ocean container service. K. 243–252. Journal of Transport Economics and Policy 15 (3)... Winston.