In this article, we analyze patterns and recent trends of the logistics industry with specific emphasis on the impact of business-to-business (B-to-B) electronic commerce on its industrial organization. From this conceptualization, we develop an optimization-based logistics model comprising two

interlinked networks representing commodity flow and information flow, respectively. The model is operationalized using Monte Carlo simulation to highlight possible scenarios of logistics industry changes in response to the use of B-to-B e-commerce. Through this endeavor, we examine how new technologies affect the cost of information and transportation transactions in the logistics industry.

**********

Introduction

Researchers in the newly emerging sub-field of Internet Geography recently began generating various studies that focus on the impacts of information technologies (ITs) on infrastructural, social (e.g., "digital divide" and consumption patterns), economic, and urban structures (Graham and Marvin 1996, 2001; Wheeler, Aoyama, and Warf 2000; Aoyama 2001a, b; Brunn and Leinbach 2001; Dodge 2001; Zook 2001; Murphy 2002; Sheppard 2002; Aoyama and Sheppard 2003). In particular, studies of Internet Typology have drawn attention to the fact that even in cyberspace, information flows have to be routed along physical infrastructure, through its geographic backbones and switching nodes. The consequences of the spatiality of the Internet suggest variations and inequities in Internet access, speed, and cost, which can have direct implications for regional space economies (Clay and Monk 1997; Odlyzko 2000; Grubesic 2002; O'Kelly and Grubesic 2002). Most research in Internet Geography to date, however, addresses business-to-consumer (B-to-C) electronic commerce, and little theoretical and empirical research has been conducted on business-to-business (B-to-B) electronic commerce.

More broadly, economic geographers have been concerned with various forms of industrial organization and their impacts on innovation, efficiency, and competitiveness with a particular emphasis on lean production, flexible specialization, and just-in-time delivery (as represented in the work by Piore and Sabel 1984; Lipietz 1986; Gertler 1988; Best 1990; Malecki and Oinas 1999). Yet, as discussed by Greis and Kasarda (1997), the issues of logistics and transportation have seldom been granted the central focus of research in these discussions, and have largely been understood simply as an auxiliary sector of secondary importance providing labor-intensive and simple transportation service. The reconfiguration of the division of labor among logistics firms, demonstrated for instance by the emergence of UPS as an integrated logistics service provider, is a new phenomenon and is directly attributed to the use of e-commerce as a medium of information exchange (Haddad and Ewing 2001).

Instead, the movements of goods and their spatial implications have been extensively studied by regional scientists and transportation geographers (Rogerson and Plane 1984; Knudsen 1985). However, the initial assumptions that B-to-B e-commerce has little impact on the locations of facilities, combined with lack of data, have so far hindered progress in research on the role of B-to-B e-commerce. The sheer volume of B-to-B e-commerce today can potentially reconfigure existing distribution networks and shift the relative weight of logistics activities across space.

The aim of this article is to fill these gaps in research by conceptualizing industrial transformations of the logistics industry and modeling its information and commodity flow networks. We first present the recent evolution of the logistics industry in the United States with B-to-B e-commerce adoption, and illustrate the complexities and the contradictory forces that are at work simultaneously. Using these discussions as a foundation we next develop a hypothetical mixed-integer optimization model that allows an analysis of various emerging configurations of the logistics industry.

The logistics industry in the information age

Although the logistics industry is emerging as a knowledge-intensive sector of strategic importance, no consensus has been reached on the geographical implications of the industry. For example, Barnes-Vieyra and Claycomb (2001, p. 17) emphasize the virtual world of information flows, and consider B-to-B e-commerce to possess the revolutionary potential for "location elimination" by overcoming spatial constraints, language barriers, and time zones. In contrast, Li, Whalley, and Williams (2001) argue for the importance of a geographical dimension, arguing that the movement of goods is an intrinsically geographical activity and spatial differences continue to exist in the information infrastructure.

In the following section, we discuss three distinctive evolving processes: disintermediation and reintermediation, disintegration and reintegration, and the migration of logistics functions from real to virtual space. These three processes provide a conceptual foundation for the prescriptive model developed later on.

Disintermediation, reintermediation, and B-to-B e-commerce

Significant cost savings from the disintermediation of supplier relationships and distribution, that is, the elimination of middlemen, is the most widespread assumption about the impact of IT (OECD 2000, p. 205, footnote 27; Jallat and Capek 2001; E-commerce in the Enterprise 2002). Instantaneous transmission of information via e-commerce allows direct relationships between firms and their suppliers, eliminating wholesalers, traders, and other traditional logistics intermediaries (Barnes-Vieyra and Claycomb 2001). Simultaneously, however, a process of reintermediation is taking place in interfirm trade, along with the rise of online logistics intermediaries that provide a wide array of services; they include e-commerce brokers, Internet auctions, industry- or function-specific marketplaces, as well as fully integrated Internet logistics services (Kaplan and Sawhney 2000; Barnes-Vieyra and Claycomb 2001; Lucking-Reiley and Spulber 2001). Some of these online intermediaries emerged as new firms specializing in online transactions, while others emerged out of traditional logistics firms successfully incorporating ITs.

Disintegration and reintegration of logistics functions

In the United States, the overall costs of logistics decreased from 16.1% of GDP in 1980 to 9.5% in 2001. Inventory carrying costs also declined from 7.9% of GDP in 1980 to 3.2% in 2001 (Wilson and Delaney 2001, 2002, Fig. 11). These figures confirm the proliferation of just-in-time delivery, accomplished through internal corporate restructuring and the use of IT, which allow improved visibility of demand as well as inventory in various sectors of the economy. They also reflect the restructuring of the logistics industry itself through simultaneous processes of disintegration (through outsourcing) and reintegration (most frequently through merger and acquisition).

Disintegration not only increased the potential for organizational rearrangement but also, arguably, accelerated the organizational dissociation of the physical movement of goods and the processing of information (Evans and Wurster 1997; Jallet and Capek 2001). This dissociation facilitates outsourcing within the logistics industry that in turn provides the largest cost savings in the e-commerce supply chain, represented by the increasing prominence of third-party logistics providers (3PLs). Today, it is estimated that over three-quarters of Fortune 500 firms use a 3PL (http://www.manufacturing.net). The 3PLs vary from those providing specialized service for just one or two segments of transportation services--which include order assembly, packaging, carrier selection, rate negotiations, fleet management, and freight payment processing--to those providing complete supply chain management functions including shipment planning, inventory management, customer services, and information systems administration, among others (Berglund et al. 1999).

Reintegration resulted from the 3PL category, expanding and including specialists on the implementation and management of the new technologies. In fact, those intermediaries that offer comprehensive supply chain management functions are increasingly being labeled as fourth-party logistics providers (4PLs) or integrated Internet logistics operators (ILOs). These firms are at the forefront of reintegration in the logistics industry, and combine the services of comprehensive 3PL solution providers with B-to-B e-commerce intermediaries (Elli 2001). The trend toward reintegration, however, is still in its initial stages. While most agree that reintegration in the logistics industry is taking place as a result of IT adoption (Jallat and Capek 2001; Lucking-Reiley and Spulber 2001), few firms to date have managed a complete system integration to conduct all aspects of transactions online. While most already operate company-wide accounting software, some are barely growing out of electronic data interchange.

E-logistics: migrating logistics to virtual space

The combined trend of reintermediation and disintegration led to the emergence of new e-logistics intermediaries. They are classified into two broad categories: those that provide logistics services exclusively in virtual space, and those that provide services in both virtual and geographic spaces. While the former outsource transportation, warehousing, and related functions, the latter also provide services in commodity flows ("asset-based" services) in addition to e-commerce operations.

Several types of e-logistics operations can be identified by the origins of firms. The asset-based category mainly comprises traditional 3PLs that have entered the e-commerce market (Lieb and Schwarz 2001), typically through forging partnerships with online B-to-C or B-to-B marketplaces; they provide transportation, warehousing, or other asset-based services for those sites. Some 3PLs developed their own e-commerce platforms to offer one-stop logistics shopping that integrates virtual orders with their physical fulfillment. Other 3PLs are making the transition to a 4PL such as UPS, and offer complete supply chain management services to cater to corporate clients to manage their entire logistical needs (Haddad and Ewing 2001).

Table 1 summarizes the current trends in the logistics industry. Manufacturing firms are increasingly outsourcing their supply chain management functions to specialized firms to leverage their expertise and economies of scale. Their traditional supply chain intermediaries, 3PLs and transportation providers, are raising information technology intensity and are conducting necessary reconfiguration and partial disintermediation in a manufacturer's supply chain to avoid being eliminated. The emerging e-commerce intermediaries, such as online auctions or exchanges, develop new marketplaces and auxiliary services for interfirm transactions in the realm of information flows. The integrated logistics operators (ILOs) that combine asset-based services have the largest growth potential in the restructured logistics industry, through reintegrating various logistics functions within their organization to maximize scale economies, and exploiting new opportunities by using e-commerce.

Assessments of the future impact of the Internet on the logistics industry vary; however, some are predicting that the impact of B-to-B e-commerce will soon overshadow the changes that result from the introduction of just-in-time manufacturing (Golob and Regan 2001), while others argue that "physical factors overpower the virtual" and that close long-term relationships between logistics firms and their customers will endure (Wilson and Delaney 2001, p. 15). Most agree, however, on the intricate nature of the relationships between the two domains that the logistics industry straddles; information flows in virtual space and commodity flows in geographic space.

The broad restructuring processes occurring in the logistics industry that we have identified are the cumulative outcomes of the decisions of many companies, made in light of a competitive environment heavily impacted by B-to-B e-commerce. Due to the relative scarcity of publicly available and coherent empirical data, the attributes and magnitude of the effects of IT on interfirm relations and their spatial manifestations within the logistics industry remain ambiguous. The prescriptive modeling framework, developed in the following section, is intended to serve as an initial step toward systematically assessing the effects of B-to-B e-commerce on the logistics network. The results of this analysis illuminate the types of information most useful to obtain through future in-depth empirical studies. We view this as an iterative process, model results suggesting avenues of inquiry, the results of which may necessitate modifications and reformulations to model structure.

A prescriptive logistics model for B-to-B e-commerce

The prescriptive logistics network model comprises two separate yet linked domains: one for information flows and one for commodity flows.

The logistics network model

The model developed in this section builds upon earlier work on logistics networks incorporating just-in-time manufacturing and interperiod network storage (Ratick et al. 1987; Osleeb et al. 1989; Osleeb and Ratick 1990; Kuby, Ratick, and Osleeb 1991), hub and spoke networks (Bryan 1998; O'Kelly 1998a, b; Horner and O'Kelly 2001), and location-routing and logistics models (see Berman, Jaillet, and Simchi-Levi 1995; Andersson, Jansen, and Waidringer 1997; Chan, Carter, and Burnes 2001; Melkote and Daskin 2001); our model represents a supply chain for a single interchangeable product for a single manufacturer. While the commodity is transported over a geographical transportation sub-network, the information required to process the order and shipment is transmitted over a separate virtual sub-network. The nodes on the two sub-networks represent firms that are potentially available to the manufacturer and fall into one of four classes: supplier nodes linked to both sub-networks; asset-based intermediary nodes linked to both networks that include firms that directly handle movement of goods (ILOs and asset-based 3PLs are also included in this category); information-only intermediary nodes solely operating in virtual space, including e-commerce intermediaries; and the manufacturer linked to both sub-networks.

The mathematical formulation of the model is given below.

Minimize:

[summation over (j)][summation over (i[member of][N.sub.j]] [C.sub.ij][X.sub.ij] + [summation over (j)][summation over (i[member of][N.sub.j]] [E.sub.ij][Y.sub.ij] + [summation over (i)][F.sub.i][W.sub.i] (1)

subject to:

[summation over (j[member of][M.sub.i]] [X.sub.ij] [less than or equal to] [S.sub.i] [for all]i [member of] K (2)

[summation over (i[member of][N.sub.j]] [X.sub.ij] = [summation over (k[member of][M.sub.j]][X.sub.jk] [for all]j (3)

[summation over (i[member of][N.sub.m]] [X.sub.im] = [D.sub.m] (4)

A [summation over (p[member of][P.sub.ij]] [Z.sub.ij.sup.p] [greater than or equal to] [X.sub.ij] [for all]i; [for all]j [member of] [M.sub.i] (5)

[Z.sub.ij.sup.p] [less than or equal to] [Y.sub.kl] [for all](k, l) [member of] [L.sub.p]; [for all]p [member of] [P.sub.ij]; [for all]i; [for all]j [member of] [M.sub.i] (6)

[summation over (j[member of][M.sub.i]] [X.sub.ij] [less than or equal to] A[W.sub.i] [for all]i (7)

[summation over (j[member of][H.sub.i]] [Y.sub.ij] [less than or equal to] A[W.sub.i] [for all]i (8)

[X.sub.ij] = 0 [for all]i; [for all]j [member of] Q (9)

[X.sub.ij] [greater than or equal to] 0; [W.sub.i] [member of] {0; 1}; [Z.sub.ij.sup.p] [member of] {0; 1} (10)

where the decision variables and parameters and sets are as follows:

Decision variables

[X.sub.ij]: amount of goods transported between nodes i and j

[Y.sub.ij]: 1, if information is transmitted between i and j

0, otherwise

[W.sub.i]: 1, if manufacturer m establishes a business relationship with node i

0, otherwise

[Z.sub.ij.sup.p]: 1, if information is transmitted between i and j over a particular path p, composed of one or more links (i, j)

0, otherwise

Parameters and sets

A: a very large number

[C.sub.ij]: cost of transporting goods between i and j

[E.sub.ij]: transaction cost of information transmission between i and j

[F.sub.i]: cost of establishing a relationship with node i (one-time, fixed transaction cost)

[S.sub.i]: supply at node i

[D.sub.m]: final demand at manufacturing node m

[H.sub.i]: set of directly linked information interaction nodes for node i

K: set of supply nodes

[L.sub.p]: set of all links located on a particular path p

[M.sub.i]: set of all destination nodes for goods from origin i

[N.sub.j]: set of all origin nodes for goods for destination j

[P.sub.ij]: set of all paths (composed of one or more links) between i and j

p: a particular path (composed of one or more links)

Q: set of information-only intermediary nodes

The objective function (1) minimizes the total cost of the overall logistics system and comprises transportation costs, information transaction costs, and fixed transaction cost (fixed charge) terms.

Information flow accounts for information-related transaction costs in two ways. First we assume that transaction costs between two firms ([E.sub.ij]) are independent of any "amount" of information exchanged; an information link is therefore either on ([Y.sub.ij] = 1) or o ([Y.sub.ij] = 0). In this respect, the information transaction cost parameters ([E.sub.ij]) are analogous to a fixed cost term, although they are attributes of individual links rather than nodes. These link transaction cost parameters can be used to represent a variety of transaction cost items including the specific type of information technologies employed by both firms, the compatibility of their information technology, the necessity for ongoing quality monitoring, the presence of trust, the degree of collaboration, the synchronization of both firms' business processes, and other measures of the closeness of their relationship.

Second, the fixed transaction cost parameter, [F.sub.i], represents the one-time costs incurred by the manufacturer in setting up a business relationship with a supplier or intermediary, and may include costs for market research, identifying and selecting suppliers, transportation carriers and e-commerce partners, requesting bids, evaluating proposals, and negotiating contracts.

The supply constraints (2) ensure that the amount of goods shipped from each supplier to all possible destinations does not exceed its available supply. Constraints (3) are trans-shipment constraints where the inflows of goods into an intermediary node equal the outflows. Constraint (4) ensures that the manufacturer's entire demand is fulfilled.

Constraints (5) and (6) link the information flow and goods transportation sub-networks. For goods to be transported, information must also flow between the origin and destination nodes for the goods. The routing of this information transmission need not be the same as on the goods transportation sub-network and may pass through a number of intermediate nodes, for example, when supply chain management functions have been outsourced to a 3PL intermediary or when an e-commerce intermediary is involved in partially reintermediating the supply chain. The path set [P.sub.ij] designates possible information routings between two nodes, and each path has an "on/o" status represented by the binary variables [Z.sub.ij.sup.p]. Therefore, as specified by constraints (5), for goods to be transported between i and j, at least one of these information paths must be "on" ([Z.sub.ij.sup.p] = 1), that is, all of its component information flow links [L.sub.p] are "on" ([Y.sub.kl] = 1 for all links (k, l) that are part of path p). Constraints (6) provide for this connection between information flow path status and the status of its component links: path [Z.sub.ij.sup.p] can only be 1 (on) if all of its component links [Y.sub.kl] are 1 (on).

Constraints (7) and (8) ensure that the one-time fixed transaction cost [F.sub.i] of establishing a new business relationship is borne by the manufacturer before any goods movement (7) or information transmission (8) can take place. Constraints (9) define the separate goods-flow and information-flow sub-networks by prohibiting any commodity transportation over links that involve nodes designated as information-only intermediaries. Note that a parallel constraint for the goods flow network is not necessary, since there are no goods-only nodes that are disconnected from the information sub-network.

Finally, equations (10) are the integer and non-negativity requirements for the model variables.

A sample problem

A sample problem is developed with five firms available to supply the needed goods to the manufacturer, and five asset-based intermediaries handling transportation and related goods-handling functions. Commodities can only be transported on a direct route composed of two links, from supplier to intermediary to manufacturer. The information flow sub-network consists of all 11 firms represented in the transportation sub-network (Fig. 1), plus two e-commerce intermediaries that are active only in virtual space (Fig. 2).

[FIGURE 1 OMITTED]

The model is implemented with representative (in a relative sense) but arbitrary data for the parameters inferred from the previous sections. Model parameters were obtained using Monte Carlo simulations. We solved the model for 1,000 realizations (parameter sets) obtained by assuming that all parameters are normally distributed, with standard deviations set to 10% of the mean ([sigma] = 0.1 [mu]). (Fig. 1 shows the means [mu] for the unit transportation cost parameters [C.sub.ij] above each link, while Fig. 2 indicates the means for information transaction costs [E.sub.ij] above links and for fixed transaction costs [F.sub.i] above nodes.)

Nodes T4 and T5 are conceptualized as conventional carriers, and, along with suppliers S2 and S3, represent the customary supply chain partners of the manufacturer before the use of electronic commerce alternatives. The fixed transaction costs for those intermediaries and suppliers are set to 0 (Fig. 2), indicating established business relationships based on trust and experience that require no new investments. The mean transportation costs for the links between these established supply chain partners are positioned on the low end of the relative cost spectrum with [mu]([C.sub.S2T4]) = [mu]([C.sub.S3T5]) = [mu]([C.sub.T4M]) = [mu]([C.sub.T5M]) = 45; all other transportation cost parameters involving these traditional carriers are set higher at [mu]([C.sub.SiT4]) = [mu]([C.sub.SiT5]) = 55 (Fig. 1). The traditional intermediaries T4 and T5 are less competitive in their information transaction costs as compared to the e-commerce alternatives. Since they are based on legacy information technologies, the mean information handling and transmission costs for these conventional relationships are comparatively high, with [mu]([E.sub.S2T4]) = [mu]([E.sub.S3T5]) = [mu]([E.sub.T4M]) = [mu]([E.sub.T5M]) = 150 (Fig. 2).

[FIGURE 2 OMITTED]

Second, e-commerce intermediary E1 represents an Internet auction, exchange, or broker that specializes in bringing together suppliers and manufacturers, and hence has links with most suppliers, including an exclusive relationship with S5 (Fig. 2). Such a supplier-broker is representative of non-asset-based e-logistics companies that emerge as a result of the selective reintermediation of supply chains and the outsourcing of certain supply chain functions. In our sample problem, E1 effectively increases the number of potential suppliers by providing information about S5 to the manufacturer. The information transaction costs for E1's relationships with suppliers should consequently be comparatively low; hence, we set [mu]([E.sub.SiE1]) = 50. The mean cost of communicating with the manufacturer is set to an intermediate level of [mu]([E.sub.E1M]) = 100. E1 is not specialized in transportation and other goods-handling services, however, and thus only contracts with transportation intermediary T2 at a relatively high information transaction cost mean of [mu]([E.sub.T2E1]) = 150. Based on their exclusive relationship with E1, T2 can offer competitive transportation costs for supplies from S5 ([mu]([C.sub.S5T2]) = [mu]([C.sub.T2M]) = 45). Conversely, T2's unit cost means for transportation involving other suppliers are much higher at [mu]([C.sub.S1T2]) = [mu]([C.sub.S2T2]) = [mu]([C.sub.S3T2]) = 60 (Fig. 1).

Third, e-commerce intermediary E2 specializes in B-to-B transactions involving transportation and other goods-handling services. This transportation-broker represents a second functional facet of the disintegration and outsourcing of supply chain functions. In the same way that E1 increases M's knowledge about and access to different suppliers, E2 introduces low-cost transportation provider T3 into the market (Fig. 2). This carrier benefits from: its extensive relationship with the e-commerce intermediary, which results in a larger and more geographically dispersed customer base, and being able to operate a more efficient transportation network through the reduction of empty trips. Thus, T3's unit transportation costs are universally positioned on the low end of the spectrum with [mu]([C.sub.SiT3]) = [mu]([C.sub.T3M]) = 45. In addition, E2 can also offer the services of established carriers T4 and T5. In its area of specialization, transportation services, E2's information transaction costs are low and set to means of [mu]([E.sub.TiE2]) = 50. Comparable to E1's cost structure, the mean cost of communicating with the manufacturer is again set to an intermediate level of [mu]([E.sub.E2M]) = 100, while the information links with suppliers--outside of E2's area of specialization--incur higher mean transaction costs of [mu]([E.sub.SiE2]) = 150.

Fourth, firm T1 represents an Internet (e-)logistics operator (ILO) that handles comprehensive supply chain management functions, including services related to e-commerce and information transmission (supplier selection, ordering, billing), as well as goods handling indicative of the latest trend toward the reintegration of multiple logistics functions. Both the mean unit costs for all of T1's transportation links and its mean information transaction costs are assumed to be at an intermediate level. Hence, its mean transportation costs are positioned above those of low-cost intermediary T3, but competitive with those of the other transportation companies in the model at [mu]([C.sub.SiT1]) = [mu]([C.sub.T1M]) = 50 (Fig. 1). Similarly, all of T1's information links have a mean transaction cost of [mu]([E.sub.SiT1]) = [mu]([E.sub.T1M]) = 100 (Fig. 2).

Finally, for companies other than the baseline supply chain partners S2, S3, T4, and T5, we assumed the means for the fixed transaction cost parameters [F.sub.i] to be 250 for the suppliers S1, S4, and S5, and lower for transportation intermediaries T1, T2, and T3 at 100. Business relationships with the e-commerce intermediaries E1 and E2 can be established for fixed transaction costs [F.sub.i] of 50.

We initially solved the problem by treating the means (Figs. 1 and 2) as values for the respective parameters. As expected, this resulted in the goods flow of the baseline, pre-e-commerce supply chain (Fig. 3): goods are transported by carrier T4 from supplier S2 to the manufacturer, and by T5 from S3 to M. The absence of additional fixed transaction costs for contracting with these longstanding supply chain partners, along with the competitive unit transportation costs being offered by carriers T4 and T5, ensures that these intermediaries are part of the cost-optimal solution. The emergence of e-commerce alternatives has changed the way information is processed; no longer are orders and other information passed directly from the manufacturer through the transportation intermediaries to the supplier. Instead, transportation broker E2 can leverage its B-to-B information technology to its advantage by providing more efficient information handling and transmission services. E2 is used to communicate and coordinate with all participants in the supply chain wherever possible (Fig. 4). Only the link between S3 and T5 still relies on conventional, direct information exchange, simply because the network topology of our model assumes that E2 cannot communicate with this supplier. In effect, this scenario shows a partial disintegration and outsourcing of supply chain management functions to e-logistics intermediary E2, and thus the reintermediation of the traditional supply chain. The total supply chain cost in these circumstances is 45,550, comprising transportation costs of 45,000, information costs of 500, and fixed costs of only 50.

Using the above assignments of parameter means, 1,000 scenarios (realizations) of relative cost structures were generated using Monte Carlo simulation. Table 2 shows the aggregate results, focusing on the cumulative outcomes of the decision variables [W.sub.i], which indicate the percentage of solutions that included each supplier and intermediary. The manufacturer's traditional carriers T4 and T5 are part of the solution in 33% and 38% of cases. Among supplier selections, S2 is included in 66% of cases, which is notably more often than other suppliers. S2 has the advantage of being one of the manufacturer's original suppliers with no additional fixed transaction costs to the manufacturer. It is also able to beat the other original supplier S3, which is chosen in 39% of cases, because unlike S3, all transportation intermediaries in our network topology can serve S2 (Fig. 1).

[FIGURE 3 OMITTED]

In 79% of cases, the optimal solution involves transportation broker E2, and in 60% of cases, its associated low-cost transportation provider T3 is chosen. These results confirm that the primary influence on model outcome is transportation costs. The new e-commerce intermediary E2 enables the manufacturer to take advantage of cost-saving transportation options. In addition, as in the scenario based on the parameter means described above, the outsourcing of logistics functions to the e-logistics intermediary E2 provides a secondary source of cost savings in the handling and transmittal of information.

[FIGURE 4 OMITTED]

The other e-logistics intermediary in the model, supplier-broker E1, is included less often, figuring in 27% of solutions. The supply chain partners that E1 introduces into the marketplace, its exclusive carrier T2 and supplier S5, are included in a similar percentage of solutions with values of 27% and 26%, respectively. While E1 and T2 were chosen in 268 cases out of 1000 (Table 2), in only 16 (= 1.6%) cases were no additional intermediaries part of the least-cost solution. Although E1 provides potential access to new supplier S5 for the manufacturer, supply-side factors such as differing costs and quality for products between suppliers have not been included in the present model framework. Hence, the strengths of a supplier-broker Web site like E1 cannot compensate for the deciding factor of transportation costs. This disadvantages E1 and T2, compared with transportation-broker E2 and its associated low-cost carrier T3.

Of particular interest are the cases where the selection of integrated e-logistics company T1 results in a unique reintermediation of the supply chain and a reintegration of management functions within this Internet (e-)logistics operator (ILO). In 7% of cases, T1 is a part of the least-cost solution (Table 2), but only in 15 cases (= 1.5%) does it serve as the sole goods- and information-handling intermediary (Fig. 5). T1 is not chosen more often because it rarely is the transportation cost leader. Using the ILO T1, however, generally results in solutions with the lowest information transaction costs, since the bundling of supply chain management functions at this intermediary avoids further information exchange with other firms, and the costly coordination associated with it. In the 15 cases where T1 provides the optimal solution as the sole supply chain intermediary, then, this information transaction cost advantage outweighs the transportation cost disadvantage. For instance, in one such case where T1 was chosen as the sole intermediary (Fig. 5), the transportation cost was 44,408, but information transaction costs were only 282. Fixed transaction costs of 365 then led to total supply chain costs of 45,055.

[FIGURE 5 OMITTED]

This potential tradeo between transportation costs and information transaction costs is confirmed by a comparison of those solutions that exclusively use the ILO with all other Monte Carlo simulation results (Fig. 6). At 43,330, the 15 cases where the ILO T1 is the sole intermediary have significantly higher average transportation costs than the remaining 985 solutions, with an average of 41,918 (right-hand bars and lower axis in Fig. 6). Comparing information transaction costs, the relationship is opposite: the ILO solutions only require the manufacturer to absorb an average cost of 305, while the average for all other solutions is 503 (left bars and top axis). In addition, fixed transaction costs are somewhat lower for the solutions relying on T1, with an average cost of 358, compared with average costs of 387 for the other cases. Therefore, if Internet logistics operators can be competitive in their transportation cost structure, their inherent advantage in information processing should give these intermediaries a dominant position in the e-logistics marketplace in accordance with the trend toward reintegration of supply chain functions. Otherwise, specialized low-cost transportation providers and associated online marketplaces may be the option of choice, as our model showed by choosing the online marketplace for transportation E2 in the overwhelming majority of cases.

Finally, only a single solution among 1000 simulations makes no use of any of the e-logistics intermediaries T1, E1, or E2. This case represents the baseline, pre-e-commerce scenario of supply chain organization. In this single instance in our simulation, the absence of additional fixed transaction costs is able to outweigh the much higher information transaction costs of a conventional supply chain. In effect, however, the model and Monte Carlo simulations confirm that only under unusual circumstances would it be to the manufacturer's advantage to refrain from implementing e-logistics solutions and associated information technology upgrades. In general, the likelihood of such a "business-as-usual" approach being the model solution depends on the relationship between additional fixed costs as compared with information cost savings. Thus, in a real-world example, this would largely depend on the legacy systems involved and the information technology changes necessary to convert to e-commerce solutions.

[FIGURE 6 OMITTED]

Conclusion

In this article, we reviewed the evolution of B-to-B e-commerce and its effects on the logistics industry. We first identified industry trends, which were then operationalized within a prescriptive logistics modeling framework comprising information and transportation networks in order to test the possible outcomes of various relative transaction cost structures. By modeling commodity and information flows on separate yet linked sub-networked domains, we demonstrated the nature of the inter-relationship between logistics in geographic space and e-logistics in virtual space. Our results suggest that competitive transportation costs can still play an important role in the logistics industry, even after the introduction of B-to-B e-commerce. However, in effectively all scenario solutions, B-to-B e-commerce intermediaries were adopted, which demonstrates the persistence of the advantages that such intermediaries can provide, and auguring well for the continued growth of e-logistics.

The single-objective static optimization model that we presented minimizes the total cost of the supply chain in a single instance in time. Modern manufacturing processes that rely on just-in-time delivery and a lean production strategy generally require two additional objectives. First, minimum response time is vital to demand-driven supply chain management (Hadjiconstantinou 1999; Ralston 1999; Liker and Wu 2000; Arntzen and Shumway 2002). Second, an objective in supply chain management is to minimize the cash-to-cash cycle time, which is equivalent to minimizing the amount of goods in transit or storage in the system at any one time. The model can be extended to incorporate these multiple real-world objectives by explicitly incorporating time and allowing for goods storage at intermediary nodes between one time period and the next (Ratick et al. 1987; Osleeb and Ratick 1990). The objective of minimizing inventory in the system can be expressed in terms of inventory carrying costs, and added to the existing objective function that minimizes total supply chain costs. The dynamic formulation permits response time to be represented as a second objective function.

Such a multi-objective dynamic model extension enables many supply chain situations to be incorporated into our modeling approach in a more realistic manner. For instance, information transmittal and goods shipment often occur asynchronously, as in the case of standing orders where goods are transported continuously, yet information exchange is required only at long intervals. Other examples include an expiration date for fixed investments, allowing the model to account for technological progress. Another extension acknowledges possible restrictions on information exchange in a time period, such as bandwidth limitations. If the capacity of a link were exceeded, the remaining information would either need to be transmitted in the next time period, or a second information link between the nodes in the same time period would have to be opened. In either case, the appropriate additional charges for the extra information will be incurred.

The spatial problems in a logistics/e-commerce system can have a significant impact on firms' ability to engage in logistics activity. The locations of suppliers, intermediaries, and the manufacturer all affect physical as well as information access, and transportation pricing may vary between urban vs. rural contexts. A variety of real geographical factors will also need to be reflected in estimating the information and transportation costs. Thus, the next step for this study includes empirical research to generate better approximations for transaction costs, particularly for information transmission and related information-handling tasks. Further theoretical and empirical work on the effect of electronic commerce on logistics and distribution systems would contribute to this end. Both e-commerce and logistics are likely to increase their importance in our understanding of the future of the space-economy; thus, continued research on this theme would be essential in furthering our assessment on the role of technologies in spatial structures.

The continuing evolution of the logistics industry and its relationship with e-commerce also suggest an augmentation of the optimization model with a dynamic agent-based approach that simulates the reconfiguration of the logistics marketplace and the firms participating in it (Blake and Gini 2002; Collins, Ketter, and Gini 2002; Gijsen, Szirbik, and Wagner 2002; Karageorgos, Thompson, and Mehandjiev 2002; Turowski 2002). This dynamic simulation component can allow for the new entry and exit of intermediaries and suppliers, and permit the modification of parameters of existing supply chain firms. For instance, if an intermediary relying on legacy information technology continuously loses market share in successive cycles, it would either be forced to exit the market or reinvest in a technology upgrade.

The persistent importance of commodity flow transport costs in the logistics industry suggests the presence of limitations for strictly virtual solutions to logistical efficiency. The frictions, constraints, and obstacles that geography brings to commodity flows are by no means overcome or made irrelevant as a result of greater involvement of e-logistics operators and cyber-agents. Simultaneously, the information transaction costs incorporated in this model provide an assessment of the presence of frictions in cyberspace interactions in the form of the quality of information transmission (i.e., not only speed and accuracy but also in terms of transparency and customer service quality), which are influenced by a variety of operational obstacles in terms of consistency and transferability of databases at the interfirm level. The resulting logistics networks therefore represent a juxtaposition of geographic and functional frictions operating simultaneously in the two dimensions of commodity flows and information flows, at times exacerbating and at other times neutralizing each others' effects. Thus, with the introduction of B-to-B e-commerce, the logic of commodity flows and the logic of information flows are intertwined and generate the newly emerging industrial organization of the logistics industry.

Table 1 Principal Processes Reshaping E-Logistics Networks

Participants        Logistics Functions      Trends

Manufacturers       Supply chain management  Outsourcing
Carriers            Transportation           IT intensification
                                             required; else:
                                             disintermediation
3PLs                Warehousing, etc.
Virtual B-to-B      Electronic marketplace   Reintermediation
intermediaries
Internet logistics  All logistics functions  Reintegration and
operators                                    reintermediation

NOTE: IT, information technology; 3PL, third party logistic provider.

Table 2 Results of Model Solutions after 1,000 Parameter Simulations

Supplier/intermediary                  Selected in ... percent of cases

S1                                     34%
S2 (Traditional supplier)              66%
S3 (Traditional supplier)              39%
S4                                     35%
S5 (Exclusive partner of E1)           26%
T1 (Internet logistics operator)        7%
T2 (Exclusive partner of E1)           27%
T3 (Exclusive partner of E2)           60%
T4 (Traditional carrier)               33%
T5 (Traditional carrier)               38%
E1 (E-commerce supplier-broker)        27%
E2 (E-commerce transportation-broker)  79%

Acknowledgements

The authors thank the Editor and three anonymous reviewers for their constructive comments on earlier drafts of this paper, any remaining errors are solely the responsibility of the authors. Support for this project is being provided through a grant from the National Science Foundation, Geography and Regional Science Program (BCS 0350697); and through the Henry J. Leir Fellowship program of the Ridgefield Foundation.

Submitted: August 8, 2003. Revised version accepted: August 3, 2004.

References

Andersson, S., K. Jansen, and J. Waidringer. (1997). "Optimisation of a Generic Multi-Mode Transportation and Logistics Network." In Information Systems in Logistics and Transportation, 329-34, edited by T. Bernhard. Oxford: Pergamon Press.

Aoyama, Y. (2001a). "Structural Foundations for Electronic Commerce: A Comparative Organization of Retail Trade in Japan and the United States." Urban Geography 22, 130-53.

Aoyama, Y. (2001b). "The Information Society, Japanese Style: Corner Stores as Hubs of E-Commerce Access." In Worlds in Electronic Commerce, 109-28, edited by S. D. Brunn and T. R. Leinbach. Chichester, UK: Wiley.

Aoyama, Y., and E. Sheppard. (2003). "The Dialectics of Geographic and Virtual Spaces." Environment and Planning A 35, 1151-56.

Arntzen, B. C., and H. M. Shumway. (2002). "Driven by Demand: A Case Study." Supply Chain Management Review January/February, 34-41.

Barnes-Viera, P., and C. Claycomb. (2001). "Business-to-Business E-Commerce: Models and Managerial Decisions." Business Horizons 44(3), 13-20.

Berglund, M., P. van Laarhoven, G. Sharman, and S. Wandel. (1999). "Third-Party Logistics: Is There a Future?" The International Journal of Logistics Management 10(1), 59-70.

Berman, O., P. Jaillet, and D. Simchi-Levi. (1995). "Location-Routing Problems with Uncertainty." In Facility Location: A Survey of Applications and Methods, 427-52, edited by Z. Drezner. New York: Springer-Verlag.

Best, M. (1990). The New Competition: Institutions of Industrial Restructuring. Cambridge: Harvard University Press.

Blake, M. B., and M. Gini. (2002). "Introduction to the Special Section: Agent-Based Approaches to B2B Electronic Commerce." International Journal of Electronic Commerce 7(1), 7-8.

Brunn, S., and T. Leinbach, eds. (2001). Worlds of E-Commerce: Economic, Geographical and Social Dimensions. Chichester, UK: Wiley.

Bryan, D. (1998). "Extensions to the Hub Location Problem: Formulations and Numerical Examples." Geographical Analysis 30, 315-30.

Chan, Y., W. B. Carter, and M. D. Burnes. (2001). "A Multiple-Depot, Multiple-Vehicle, Location-Routing Problem with Stochastically Processed Demands." Computers and Operations Research 28, 803-26.

Clay, K., and T. Monk. (1997). "What's Next for Internet Data Analysis? Status and Challenges Facing the Community." Proceedings of the IEEE 85(10), 1563-71.

Collins, J., W. Ketter, and M. Gini. (2002). "A Multi-Agent Negotiation Testbed for Contracting Tasks with Temporal and Precedence Constraints." International Journal of Electronic Commerce 7(1), 35-58.

Dodge, M. (2001). Mapping Cyberspace. London: Routledge.

E-commerce in the Enterprise. (2002). Great Predictions that Never Happened. February 26, available at http://www.itworld.com/nl/ecom_in_ent/02262002/

Elli, S. A. (2001). "Who's Who? Sorting out the e-Logistics Players." Logistics Management and Distribution Report 40(4), E3-9.

Evans, P. B., and T. S. Wurster. (1997). "Strategy and the New Economics of Information." Harvard Business Review 75(5), 71-82.

Gertler, M. (1988). "The Limits to Flexibility: Comments on the Post-Fordist Vision of Production and its Geography." Transactions: Institute of British Geographers 13, 419-32.

Gijsen, J. W. J., N. B. Szirbik, and G. Wagner. (2002). "Agent Technologies for Virtual Enterprises in the One-of-a-Kind Production Industry." International Journal of Electronic Commerce 7(1), 9-34.

Golob, T. F., and A. C. Regan. (2001). "Impacts of Information Technology on Personal Travel and Commercial Vehicle Operations: Research Challenges and Opportunities." Transportation Research Part C: Emerging Technologies 9(4), 87-121.

Graham, S., and S. Marvin. (1996). Telecommunications and the City: Electronic Spaces, Urban Places. London: Routledge.

Graham, S., and S. Marvin. (2001). Splintering Urbanism. Networked Infrastructures, Technological Mobilities and the Urban Condition. London: Routledge.

Greis, N. P., and J. D. Kasarda. (1997). "Enterprise Logistics in the Information Era." California Management Review 39(4), 55-78.

Grubesic, T. H. (2002). "Spatial Dimensions of Internet Activity." Telecommunications Policy 26(7-8), 363-87.

Haddad, C., and J. Ewing. (2001). "Ground Wars. UPS's Rapid Ascent Leaves FedEx Scrambling." Business Week May, 64-68.

Hadjiconstantinou, E., ed. (1999). Quick Response in the Supply Chain. Berlin: Springer-Verlag.

Horner, M. W., and M. E. O'Kelly. (2001). "Embedding Economies of Scale Concepts for Hub Network Design." Journal of Transport Geography 9, 255-65.

Jallat, F., and M. J. Capek. (2001). "Disintermediation in Question: New Economy, New Networks, New Middlemen." Business Horizons 44(2), 55.

Kaplan, S., and M. Sawhney. (2000). "E-Hubs: The New B2B Marketplaces." The Harvard Business Review 78(3), 97-106.

Karageorgos, A., S. Thompson, and N. Mehandjiev. (2002). "Agent-Based System Design for B2B Electronic Commerce." International Journal of Electronic Commerce 7(1), 59-90.

Kuby, M., S. J. Ratick, and J. Osleeb. (1991). "Modeling US Coal Export Planning Decisions." Annals of the Association of American Geographers 81, 627-49.

Li, F., J. Whalley, and H. Williams. (2001). "Between Physical and Electronic Spaces: The Implications for Organizations in the Networked Economy." Environment and Planning A 33, 699-716.

Lieb, R., and B. Schwarz. (2001). The Year 2001 Survey: CEO Perspectives on the Current Status and Future Prospects of the Third Party Logistics Industry in the United States. Accenture Consulting and Northeastern University, available at http://web.cba.neu.edu/~rlieb/2001ceopaper.dot

Liker, J. K., and Y.-C. Wu. (2000). "Japanese Automakers, U.S. Suppliers and Supply-Chain Superiority." Sloan Management Review 42(1), 81-93.

Lipietz, A. (1986). "New Tendencies in the International Division of Labor: Regimes of Accumulation and Modes of Regulation." In Production, Work, Territory: The Geographical Anatomy of Industrial Capitalism, edited by A. J. Scott and M. Storper. Boston: Allen & Unwin.

Lucking-Reiley, D., and D. F. Spulber. (2001). "Business-to-Business Electronic Commerce." Journal of Economic Perspectives 15(1), 55-68.

Malecki, E. J., and P. Oinas, eds. (1999). Making Connections: Technological Learning and Regional Economic Change. Aldershot, UK: Ashgate.

Melkote, S., and M. S. Daskin. (2001). "An Integrated Model of Facility Location and Transportation Network Design." Transportation Research Part A: Policy and Practice 35(6), 515-38.

Murphy, A. (2002). "The Emergence of Online Food Retailing: A Stakeholder Perspective." Tijdschrift voor Economische en Sociale Geografie 93(1), 47-61.

Odlyzko, A. (2000). "The Internet and Other Networks: Utilization Rates and Their Implications." Information Economics and Policy 12(4), 341-65.

OECD (2000). OECD Economic Outlook 67 (June). Paris: Organisation for Economic Cooperation and Development.

O'Kelly, M. E. (1998a). "A Geographer's Analysis of Hub-and-Spoke Networks." Journal of Transport Geography 6(3), 171-86.

O'Kelly, M. E. (1998b). "On the Allocation of a Subset of Nodes to a Mini-Hub in a Package Delivery Network." Papers in Regional Science 77(1), 77-99.

O'Kelly, M. E., and T. H. Grubesic. (2002). "Backbone Topology, Access, and the Commercial Internet, 1997-2000." Environment and Planning B: Planning and Design 29(4), 533-52.

Osleeb, J. P., and S. J. Ratick. (1990). "A Dynamic Location-Allocation Model for Evaluating the Spatial Impacts for Just-in-Time Planning." Geographical Analysis 22, 50-69.

Osleeb, J. P., S. J. Ratick, M. J. Kuby, H. E. Olson, L. G. Antle, and A. F. Hawnn. (1989). "Coal Logistics System." Transportation Research Record 1222: Ports, Waterways, and Inland Water Transportation, 6-13.

Piore, M., and C. Sabel. (1984). The Second Industrial Divide. New York: Basic Books.

Ralston, B. A. (1999). "GIS and Logistics: Potential Contributions and the Current State-of-the-Art." Conference paper presented at the International Workshop on Geographic Information Systems for Transportation (GIS-T) and Intelligent Transportation Systems (ITS), Hong Kong, April 26-28, available at http://www.jlgis.cuhk.edu.hk

Ratick, S. J., J. P. Osleeb, M. Kuby, and K. Lee. (1987). "Interperiod Network Storage Location-Allocation (INSLA) Models." In Spatial Analysis and Location-Allocation Models, 269-301, edited by G. Rushton and A. Ghosh. New York: Van Nostrand Reinhold.

Rogerson, P. A., and D. A. Plane. (1984). "Modelling Temporal Change in Flow Matrices." Papers of the Regional Science Association 54, 147-64.

Sheppard, E. (2002). "The Spaces and Times of Globalization: Place, Scale, Networks, and Positionality." Economic Geography 78(3), 307-30.

Turowski, K. (2002). "Agent-Based e-Commerce in Case of Mass Customization." International Journal of Production Economics 75, 69-81.

Wheeler, J. O., Y. Aoyama, and B. Warf. (2000). "Introduction: City Space, Industrial Space, and Cyberspace." In Cities in the Telecommunication Age: Fractured Geographies, 3-17, edited by J. O. Wheeler, Y. Aoyama, and B. Warf. New York: Routledge.

Wilson, R., and R. V. Delaney. (2001). Managing Logistics in a Perfect Storm (12th Annual "State of Logistics Report"). Cass Information Systems and ProLogis, available at http://www.cassinfo.com/2001%20Press%20Conference%20Final.PDF

Wilson, R., and R. V. Delaney. (2002). Understanding Inventory--Stay Curious (13th Annual "State of Logistics Report"). Cass Information Systems and ProLogis, available at http://www.cassinfo.com/2002%20Press%20Conference%20Full.pdf

Zook, M. A. (2001). "Old Hierarchies or New Networks of Centrality? The Global Geography of the Internet Content Market." American Behavioral Scientist 44, 1679-96.

Yuko Aoyama, Samuel J. Ratick, Guido Schwarz

Graduate School of Geography, The George Perkins Marsh Research Institute, Clark University, Main Street, Worcester, MA

Correspondence: Yuko Aoyama, Graduate School of Geography, The George Perkins Marsh Research Institute, Clark University, 950 Main Street, Worcester, MA 01610-1477

e-mail: yaoyama@clarku.edu