Dr. Chris Anderson
University of Rhode Island-Env. Nat'l Res. Economics
Kingston-Coastal Institute Bldg.
Kingston, RI  02881
401-874-4587
cma@uri.edu
PI


Prof. Thomas Grigalunas
University of Rhode Island-Env. & Nat'l Res.
Kingston Coastal Institute Bldg.
Kingston, RI  02881
(401)-874-4572


Dr. Yong-Tae Chang
URI-Korea America Joint Marine Policy Reseach Center
Kingston Coastal Institute
Kingston, RI  02881
401-874-2471


Dr. Meifeng Luo
University of Rhode Island-Env. & Nat Res. Sci.
Kingston Coastal Institute
Kingston, RI  02881
401-874-2471

Our broad objective is to assist planners in determining whether they can capture and profitably defend shipping market share as a hub, traditional land-sea cargo port, or regional feeder port. While this is a long-term research program, we propose to, within the allowed two-year project, develop and estimate empirical and theoretical models that will lay the foundation for future work, and an eventual tool for use by planners. We propose to:

1. Improve our understanding of shippers’ port choices by estimating how demand shifts among ports as a function of port and hinterland features
   
2. Identify how incumbent ports can defend their market power by developing a game-theoretic model of infrastructure investment and local monopoly, tested with a controlled economic experiment
   
3. Extend our existing port demand simulation model to reflect shipper preferences
   
4. Evaluate specific port development scenarios of current policy interest by applying the simulation model to determine payoffs for key ports under different strategic development responses in the game theoretic model.
   While the resulting model of market entry and demand for port services will aid planners in determining the proper investment scale for new ports in a competitive environment, it will also help them determine whether there are additional gains from cooperative development, and in assessing the effects of service interruptions among existing ports, due to labor dispute, accident, natural disaster or other event.

Intermodal

As national economies globalize, demand for intercontinental container shipping services is growing rapidly, providing a potential economic boon for the countries and communities that provide port services. On the promise of profits, many governments are investing heavily in port infrastructure, leading to a possible glut in port capacity, driving down prices for port services and eliminating profits as ports compete for business. Further, existing ports are making strategic investments to protect their market share, increasing the chance new ports will be overcapitalized and unprofitable. Our project will enable governments and port researchers to understand how local competition in their region will affect demand for port services at their location, and thus better assess the profitability of a prospective port.

We propose to extend our existing simulation model of global container traffic to incorporate demand-side shipper preferences and supply-side strategic responses by incumbent ports to changes in the global port network, including building new ports, scaling up existing ports, and unexpected port closures. We will estimate shipper preferences over routes, port attributes and port services based on US and international shipping data, and redesign the simulation model to maximize the shipper’s revealed preference functions rather than simply minimize costs. As demand shifts, competing ports will adjust their pricing (short term) and infrastructure (long term) to remain competitive or defend market share, a reaction we will capture with a game theoretic model of local monopoly that will predict changes in port characteristics. The model’s hypotheses will be tested in a controlled laboratory experiment tailored to local port competition in Asia, which will also serve to demonstrate the subtle game theoretic concepts of imperfect competition to a policy and industry audience. We will apply the simulation model to analyze changes in global container traffic in three scenarios: addition of a new large port in the US, extended closure of an existing large port in the US, and cooperative and competitive port infrastructure development among Korean partner countries in Asia.

We propose to approach our objectives through four phases. The first two phases focus on developing theoretical and empirical inputs into a global port demand simulation model: phase I will conduct a statistical analysis of port choice data to enhance our understanding of factors which determine port selection, and phase II will develop a strategic model of port investment to capture and protect market share. In phase III, we will extend our existing port demand simulation model to reflect the results of phases I and II. The fourth phase will apply the simulation model to evaluate demand under key scenarios, including cooperative investment in Asia, and expansion or closure of key ports in the US.

Phase 1: Obtain estimates of demand for port services throughout the global network
In this phase, we will use data on shipment routing to determine the factors that enter in shippers’ decisions about which ports to use, and thus understand how their choices would change in the face of changes in the global port network. We will conduct the analysis on a continental basis, using original separate analyses for the US and Asia; we will adopt Veldman and Buckmann (2003) for European results. For each region, we will obtain data on the routing for a wide range of import and export shipments. For each shipment, we will use the origin (foreign port for imports, domestic city for exports), destination (domestic city for imports, foreign port for exports), routing, timing, value and other manifesto information for every container processed at major ports in the coverage area. We will combine this shipment-specific data with data on port attributes previous studies have suggested are important to port selection (Malchow and Kanafani 2004; Murphy et al. 1992; Murphy and Daley 1994). These include physical characteristics (e.g., berth size, channel depth); productivity measures (e.g., crane speed, number of cranes); size of the commodity demand and supply structure in the port’s hinterland; the transportation infrastructure of the port’s hinterland; and commodity-specific variables. We will consult with industry collaborators to identify other relevant variables. While port service pricing data is unavailable, interviews at US ports suggest per TEU handling charges vary little (see Talley 1994), and that other charges vary with distance of piloting, shipper volume and seasonal variables that can implicitly capture price differences among ports. Supplementing these previous studies, investigator Chang is hosting, through the Global U8 University Consortium, a Seminar on Port Competition in November, at Inha Unviersity, where the investigators will meet with industry representatives, government officials, and academics to improve our understanding of demand and strategic competition.

The relationship between choice of port and these independent variables will be assessed using a conditional logit model, which has been widely used to model choices of route and mode in transportation demand (e.g., McFadden 1974; Malchow and Kanafani 2001, 2004; Veldmann and Bruckmann 2003). Conditional logit (see Sidebar 1) models the probability of choosing each port as a function of the utility (profit) the shipper would receive from using that port, which is function of shipment attributes, port attributes and other factors which are not observed. In making their choice, the shippers trade off among the attributes in unknown ways, which can be measured from the data. The estimated attribute coefficients are weights that indicate the relative importance of each attribute. We will test the hypothesis that each of the factors included in the model using a t-test of the coefficient on each factor to determine whether its effect is statistically significant.

While this analysis is of interest in itself, as knowing more about preferences of shippers will help port managers identify investments that will most affect their appeal to customers, our goal is to integrate these preferences into a global demand model. Hence, we will conduct separate analyses to identify the preferences of US and key foreign trade partners. For the US, we will directly transfer results from a current related project of Anderson and Grigalunas, which uses precisely this method to analyze Port Import Export Reporting Service (PIERS) and Customs data to assess demand for US ports. Asian data will be obtained through efforts of Korean collaborators and through the Korean PORTMIS (Port Management Information System), which maintains similarly thorough database of manifest data, and will be analyzed as part of this project. While the US and Asian groups will approach this task for their respective data, they will be importantly linked by the lead investigators and the need to supply inputs to the demand simulation model in Phase III.

The conditional logit modeling will yield information on the preferences of shippers for port attributes, allowing us to predict how they will react to new ports, closed ports, or changes in attributes of existing ports. This reaction can be captured by changing the attributes for a port (or shipment) and using the estimated coefficients to predict choice probabilities. The demand for each port will be the expected number of containers, given the port attributes resulting from each investment scenario of interest.

Phase II: Develop and test a model of spatially specific port development competition
In this phase, we will develop and test a theoretical model of port authorities’ strategic responses to changes in demand caused by investment of competing ports, or loss of capacity at competing ports. Ports compete with one-another by investing in infrastructure that makes them more appealing to shippers seeking service to their hinterland. Unlike in a standard perfectly competitive market, the investment and pricing decisions of individual ports can affect the market for service, and the extent of that effect depends on the actions of the port of interest, and of the other ports in the region. Such environments where the payoffs of economic agents depend on their own decisions and on the decisions of others are the domain of game theory. As with any modeling exercise, we will simplify the situation to capture the key incentives associated with competition and development; economics is a science of reducing behavior to simple models, and we continue to use simple models because they provide useful insights.

This phase will be accomplished in two tasks, model development and model testing. The model will be tested with a controlled laboratory experiment, in which human subjects play the role of port managers.

• Task 1: Model development
  Port competition includes three elements that are key to understanding competitive investment. First, it is characterized by imperfect competition, where individual ports may cooperate or compete to provide service for a given area. Whether competitive or cooperative outcomes are sustained is of particular interest to port developers in Asia. Second, port competition is spatially dependent, in that investments will generally affect competitors whose hinterlands overlap to a greater extent than more distant competitors. Third, the game is dynamic, in that it is played through time with the option for ports to make investments in different years.
  While model development is the research task, we anticipate integrating elements from three basic game models to develop a single model that generates a range of predictions consistent with the incentives associated with international container port development and competition. The fundamental model of competition will be Bertrand price competition (see Sidebar 2), in which competing ports’ demand curves are substitutes to some shipping customers (captured by parameter b). In this model, the imperfect competitors each set their price anticipating the prices that others will set. In Nash equilibrium, each port is maximizing its profit, given the price levels set by the other players.
  
  The basic Bertrand model will be modified in two ways to incorporate incentives associated with port investment. First, the potential for competition among ports is determined (or limited) by their locations and the extent to which they are substitutes for local and hinterland markets. This can be represented in the model by setting different elasticity of substitution parameters (b) for different ports: those with no local or hinterland competition will have inelastic demands, and those serving similar markets will have demands which considerably affect each other. The equilibrium result will be that ports with little local competition will be able to charge higher prices (or offer fewer services), and that ports competing for the service to the same hinterland will have to lower prices or improve service to maintain market share.
  The second modification to the standard Bertrand model is to integrate a competitive investment environment. For this, we anticipate drawing on two other standard games. Competitive investment in imperfect competition has been widely studied modeled with the monopoly entry game (Cooper, Garvin and Kagel 1997). In the entry game, a number of potential competitors simultaneously decide whether to pay an entry cost to compete with other entrants in a market with a downward-sloping demand curve. If few firms enter, they receive a high price, enough to recover their entry cost and make a profit; if many firms enter, the price is low and entrants may not recover their entry cost. In equilibrium, enough firms enter that each is just able to recover their entry cost, but an additional entrant would not.
  In the port competition game, the entry cost will correspond to the amount paid to develop port infrastructure, which will change the substitution elasticities (port-pair specific bijs) among ports. However, when infrastructure is being developed, it can be expanded (or not maintained) in continuous time. This dynamic element changes the game, adding incentives of a continuous all-pay dollar auction (Shubik 1971), a game where all players bid on a dollar bill, the highest bidder wins the dollar, but all players must pay their bids when the auction ends. Because at any current high bid, every player is better off raising the bid by 1 cent (because then she pays a couple cents more, but wins the dollar), in equilibrium, bidding continues until all but the winning player exhausts their budget. Similarly, ports will continue investing infrastructure to capture and defend market share over which they are competing; because prior investments are sunk costs, the additional investment required to expand or defend market share is always justified. However, because the “prize” of market share is continuous and of decreasing marginal value and constrained by increasing costs of technology and operation, the equilibrium prediction in the port game will not be as extreme as in the dollar auction.
  
  Our analysis will look at two types of outcomes, competitive and cooperative. In the competitive Nash equilibrium, ports will invest heavily to compete with one-another. When a large number of ports are serving the same markets (as could happen in Asian export manufacturing regions, or in the US with the inland Midwestern market which can be served well by many ports), the level of investment could be so high that it cannot be recovered (without significant growth in global port demand). In a cooperative outcome, the multiple ports will coordinate on investment levels in which they achieve equal levels of profitability across the global market. The distinction between these two outcomes is of particular policy relevance, because there is concern among port experts that many governments are investing in ports in anticipation of capturing market share, but without adequate consideration for the capacity being developed elsewhere. If the parameters determining returns on investment are such that in the competitive equilibrium can lead to losses, then port development is not a responsible use of public monies; welfare improvements require either cooperative investment levels, or not entering the port services market in the first place. However, outcomes can be improved with cooperation.
  
  
• Task 2: Model testing and demonstration
  The theoretical model will yield a prediction of a competitive equilibria and cooperative outcome of the port investment game. In our experience doing modeling for policy, nonacademics often have a hard time utilizing such purely theoretical results. This is especially true when the model yields counterintuitive results, such as when all or almost all investors lose money. People often wonder if all people are losing money, why can’t they realize this, and reach a better outcome? However, such socially inferior outcomes happen all the time: fishers overharvest fisheries, threatening the sustainability of their own livelihood and suburbanites move outside cities, contributing to congestion on their own commutes. Thus, we plan to test the hypotheses of the model, and demonstrate to policymakers the range of outcomes that may occur in the port investment game.
  
  Our demonstration will be a controlled laboratory experiment, in which human subjects will play the role of port authorities (or governments responsible for port development), choosing levels of infrastructure investment. We will conduct two treatments, each calibrated to a different region of the world to capture different levels of competitiveness. The high competitiveness treatment will be benchmarked to the Far-eastern port market, where China, Korea, Taiwan and Japan compete to ship Asian goods to western markets. The low competitiveness treatment will be benchmarked to the US West Coast market, where container trade is dominated by LA/LB and Seattle/Tacoma, with different local and hinterland markets.
  
  We anticipate our experimental sessions will involve 6-12 subjects, each playing the role of a significant port (or potential port) in the region, within the URI Policy Simulation Laboratory, a 26-workstation facility designed for running economic experiments. Subjects will interact anonymously through computer software developed especially for this experiment. Each subject will be given capital and investment opportunities corresponding to the level of infrastructure at her port. Through a sequence of 15-30 periods (years), each subject will select a level of investment for her port, and will receive payoffs based on the appeal of her port to shippers, given her investments and those of other subjects, as determined by the simulation model (see Phase IV); we may opt to use fewer periods and repeat the game within a session to understand the effects of learning. At the end of the experiment, subjects’ earnings will be totaled and converted to US dollars, and they will be paid in cash for participating.
  
  The experiments will generate data on investment levels and payoffs for each port, and levels of efficiency achieved within the game. These can be compared with the predicted Nash equilibrium and socially optimal (cooperative) outcomes, using a nonparametric Wilcoxon test for the distribution of outcomes. Of particular interest will be the frequency of a non-equilibrium, non-cooperative outcome commonly observed in games with intertemporal, delayed investment (e.g., Moxnes 1998) in which subjects overinvest in early periods, corresponding to a glut of capital and low prices for port services, at which ports cannot recover their investments. Demonstrating that such extreme outcomes are possible is a key contribution of the project.
  
  Economic experiments, the technique awarded the 2002 Nobel Prize in Economics (Vernon Smith), are effective because subjects are paid their experimental earnings in cash, so just as groups who make better investment decisions in the field make more profits, subject groups who make better investment decisions in the laboratory earn more money for participating (Smith 1976; Davis and Holt 1993). Experiments have been used to test theoretical models, demonstrate ranges of possible outcomes (e.g., Anderson 2004) and compare institutions for particular applications. High profile, high value applications of experiments include designing the auction the FCC has used raise more than nine billion dollars selling licenses to bandwidth used by cellular telephones (Banks et al. 2003; Salant 2000; Plott 1997), water allowance trading (e.g., Murphy et al. 2000; Murphy et al. 2003; Cummings, Holt and Laury 2004), the rules under which the Environmental Protection Agency sells pollution permits for sulfur dioxide under the Clean Air Act (Cason 1995; Cason and Plott 1996). In the latter case, experiments demonstrated a perverse incentive that was leading to inefficiencies, and the rules were changed as a result of the experimental results. Investing in laboratory research to fully understand the incentives of competitive actors can improve policy outcomes.
  

Phase III: Extend simulation model of global port demand to reflect shipper preferences
In this phase, we will extend our existing model, developed over the last several years with URITC funding to understand potential demand for a port at Quonset Point, RI, to reflect shipper preferences in container routing (Luo, 2002; Luo and Grigalunas; 2003; Grigalunas et al, 2001, 2002). The outcomes of the simulations will be used to evaluate port development and security scenarios in Phase IV.
Our existing port demand simulation models movement of all container freight imported to the US from its country of origin to its state of destination, and all container freight exported from each US state to its destination country. International trade data is used to determine the imports and exports of each commodity category for each country, and the volume between each US state-foreign country pair. For each state-country pair, the model selects the route for cargo that minimizes transportation costs, including inland trucking and rail costs, at-sea transport costs, and cargo inventory costs based on the average value of the commodity. Demand at each port is the total number of containers that pass through that port on their cost-minimizing route from their origin to destination.

While this model is the state-of-the-art in modeling port demand, and is useful for characterizing the general effect of medium-scale changes, such as a Quonset container terminal, it has a number of limitations that make it insensitive to small-scale changes and poorly calibrated for large changes. First, the existing model bases preferences for one port over another only on differences in total cost for origin-to-destination transport. In reality, shippers’ choice of ports depends not only on costs, but also many other factors, such as time, frequency, quality of service and risk or uncertainty of delivery. We will modify the existing simulation model to replace the cost minimization objective with the shipper objective function revealed in the conditional logit analysis from Phase I.

The second limitation of our current model is that it estimates potential demand, and thus has no limitations on the capacity of individual ports. In fact, ports have a fixed number of berths, cranes, terminal space and operation methods that constrain the number of containers they can move. Further, as these limitations are approached, service prices must increase as ports must pay overtime for longer operating hours and adopt more space efficient (but costly) yard practices; these price changes will in turn affect pricing and the quality of services. A complete supply side analysis is beyond the scope of this project, however, interviews with port experts indicates ports do not regularly change prices in response to demand, suggesting port responses to demand shifts may be reasonably approximated by imposing capacity constraints at constant prices. In our experience, port managers and terminal operators have a good sense of the current and maximum capacities—both in a short-term pulse of business and sustainable long-term—of their ports. Investigators Anderson and Grigalunas are currently collecting this data for US ports as part of an ongoing project, and a sample of Asian port managers will be contacted by our Asian counterparts for this project. Given Korean Ministry involvement in this project, and interest in the rest of Asia in cooperating with Korea in port development, we expect them to be forthcoming.

The revised model will simulate movement of shipments from their origins to their destinations; where individual shipment data is available, it will be used to identify cargo movement, and where it is not, country-aggregate international trade and consumption data can be used to determine volumes moved from origin to destination. For shipments of each commodity from each origin to each destination, the model will calculate the utility the shipper would receive from each available port, based on the logit model analysis from Phase I. It will then distribute cargo volume among ports in proportion to the predicted choice probabilities given by the logit model. If a port reaches its capacity, the shipments with the lowest opportunity cost will be shifted to their second best ports. This procedure will ensure that the global utility from shipping is maximized subject to the constraints at each individual port.

The model will yield predictions of the amount of cargo that moves through each port, given the port infrastructure and intermodal connections available at each port. From this, we can calculate the payoffs to the port authority from their level of infrastructure; the payoffs from alternative investment levels at competing ports will be used in the game theoretic model to assess Nash equilibrium and possible cooperative outcomes for different port development scenarios in Phase IV.

Phase IV: Simulate global demand for port services under scenarios for change in the port network
We will apply the empirical, theoretical and simulation models developed in Phases I - III to three scenarios of particular policy interest in the US and to our Korean partners in the project:
1. Closure of a major US port (as might be attributed to natural disaster or other incident);
2. Expansion and development of a major US port;
3. Enhanced cooperation among Korea, China, Japan and possibly Taiwan in port infrastructure development.

For each of these scenarios, we will select a port or set of ports of current interest. For the first scenario, we anticipate that the economic significance of LA/Long Beach will makes its closure of particular interest and policy importance: seismic activity in its location makes it prone to natural disaster, and its economic significance makes it a desirable terrorist target. For the second scenario, we anticipate focusing on development or expansion in of an east coast port. Recent interest in a port at Quonset Point makes it an appealing local test case, but if other east coast communities appear closer to development, we will target those options instead, as being of greater current global interest. For the third scenario, local and national Korean governments are currently considering whether and how to cooperate with one-another and other ports in central Asia, including ports in China, Japan and Taiwan to coordinate investment. We will adopt the ports that our Korean partners are mostly considering partnering with, to describe and quantify the benefits of potential cooperation.

Each scenario consists of a set of relevant ports, and investment alternatives for each port. We will apply the simulation model to determine the demand, and thus revenues and profits, for each port under each combination of investments. These payoffs will be used within the game to calculate Nash equilibrium levels of investment, and the potential relative gains from coordination of investment in a cooperative outcome, in the scenarios of interest. These equilibrium outcomes can be used to advise port authorities, local and national governments in selecting a level of port investment that maximizes profits and ensures responsible use of public economic development funds.

Interport Competition Seminar ACL/FA05
Game model development A/W06 A/Sp06
Asian port data collection C/W06
Asian shipping data collection C/W06
Asian data analysis AC/Sp06 AC/Su06
Experiment software develop AL/Sp06 AL/Su06
Experiments A/Fa06,W07,Sp07
Simulation model extension LA/Sp,Su,Fa06,W07
Scenario development ACG/W,Sp,Su07
Scenario evaluation ACLG/Sp,Su07
Manuscript development C/Su06 A/Fa06 ACLG/Su,Fa07
URITC presentation ALG/Sp06,Fa07 ACLG/Fa06
Final Report/Revision ACLG/Su,Fa07
Peer conference ACLG/Fa06,Su07

Letters indicate investigators in charge: Anderson (A); Chang (C); Luo (L); and Grigalunas (G)

Fa -Fall
W - Winter
Sp - Spring
Su - Summer

$535,641.68

This project will support three students, one directly supported graduate student at URI and two at Inha University, funded as match. Both students will be heavily involved and work closely with project leaders Anderson and Chang. At URI we can benefit from the recent expiration of another project making available Simona Trandafir, a 4th year Ph.D. student with interests and experience with both ports and logit analysis of transportation datasets. She will work with Dr. Anderson to design and implement the experimental test and demonstration of the game theoretic model, and to analyze the resulting data. She will also assist Dr. Luo in developing data inputs into the port demand simulation model. We have budgeted for her to attend one conference, such as a U8 or TRB meeting, during the project period. Another graduate student, Tae-Gwon Kim, is a former navigator for Hanjin, a major Korean liner, and his experience will likely lead to hourly involvement in the project.

Two Inha students, namely Seoung-Gon Kim, a PhD student with experience in a transportation and logistics company and an MA degree from the UK,. and either S. Shin or S. Lee, current Masters students, will be involved in the project. They will work with Dr. Chang to collect information on Asian ports and analyze the PORTMIS data to construct a model of import and export demand for Asian ports. They will also work with Dr. Anderson and Ministry officials to specify the cooperative and competitive scenarios to be analyzed in Phase IV, and to interpret and publish the results of the scenario analysis.

This project leverages a current project to Grigalunas and Anderson and a previous URITC project to Grigalunas and Luo. First, Grigalunas and Anderson’s project involves assessing the likely effect of a port closure, or security-related slowdown, on the US economy. For this project, they are estimating a model of US port demand akin to that described in this proposal (the results of which accomplish Phase I for the US), visiting ports to identify excess capacity within the port system (which can be integrated in the simulation model of Phase III), and estimating increases in port and shipping costs which will affect final market prices in the US. While their contract prohibits using confidential of sensitive data, the information, experience, contacts and results acquired through that research add significantly to the value of this project. Second, Grigalunas and Luo developed the cost-minimization based port demand simulation model through several years of URITC funding, directed primarily at assessing demand for a container terminal at Quonset Point, RI. Phase III extends this model to our current application.

This project is also an outgrowth of the research and organizational efforts of the Global U8 Consortium, a group of eight universities in the US, Europe, Asia and the south Pacific with expertise in global engineering, economic and social aspects of transportation and logistics issues. Its objective is identify and bring expertise to transportation research problems of global scale. The need for this research, to help in both port development and security planning, was first identified and discussed at the Ports and Logistics Workshop of the Global U7 (now U8) Consortium in February, 2005. At that meeting, a core set of expertise was identified, and a team led by investigator Anderson (URI), an expert in applied game theory; investigator Chang (Inha U., Korea), a logistics researcher with substantial background in port issues; investigator Luo (URI), an economist and computer scientist who has developed a port demand simulation model; and investigator Grigalunas (URI), an economist specializing in port economics and environmental problems associated with port usage.

The results of this project will be distributed through a number of avenues. We expect to produce at least four publications in peer reviewed journals, such as Transportation Research or TRB publications: one on the game model, one on the experiment, one on the simulation model, and one reporting analysis of our scenarios. The peer review process will assure other audiences of the scientific basis for our conclusions. Chief among those other audiences is port development and regulatory authorities, with whom we have direct relationships. Korean port development authorities are directly involved in this project, offering substantial (contingent) cash match, and Anderson and Grigalunas have established relationships with the US Coast Guard’s port economics group. The Global U8 Consortium hosts regular meetings of academics from around the world (including one logistics conference to be held at URI, Sep. 2006, in connection with the Consortium’s Third Presidential meeting). It invites participants from government and industry; we anticipate making several presentations to this group. It is further likely that U8 will host a follow-up meeting to the Interport Competition Seminar, with government and industry representatives. Finally, Inha has budgeted to host two workshops (one at the beginning and one at the end) of the project with international port managers, industry and government officials. The early workshop will provide us with contacts and background information; the later workshop will convey the results of our project. Such a meeting could include participating in the experimental software, to demonstrate to those in attendance the strategic incentives faced by port developers.

The improved understanding of competition in spatially overlapping markets with cumulative intertemporal investment, and how this relates specifically to the maintenance and development of the global port network, will benefit shippers, liners, port authorities and governments around the world. For shipped goods manufacturers and liners, understanding where competition is likely to result in excess capacity of port services can help them target production locations and future ship deployments for areas with low-cost (or high-tech) port services, or with capacity of future growth. For local and national governments deciding whether and how to allocate public funds to port development, the model provides a specific framework to allow them to think about how other ports in the region will react to their investments, how demand will shift, and therefore whether they will be recover their investment. They may find that only moderate or cooperative investment is preferable to a competitive environment, and is a better use of public resources. For the specific scenarios outlined in Phase IV, alternatives and outcomes will be analyzed explicitly. Finally, the competition and demand model can help governments develop supply chain security, as the model can suggest how demand will shift if a particular port is closed for a short or long time. While recent political attention in the US has focused on terrorist attacks, such information is also valuable in planning for hurricanes, typhoons, tsunamis and earthquakes.

Port competition, investment game, port overcapacity, shipper behavior, port choice, port demand simulation