24.07.2013 Views

understanding b2b e-market alliance strategies - MISRC - University ...

understanding b2b e-market alliance strategies - MISRC - University ...

understanding b2b e-market alliance strategies - MISRC - University ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

UNDERSTANDING B2B E-MARKET ALLIANCE STRATEGIES<br />

Qizhi Dai<br />

Doctoral Candidate<br />

Information and Decision Sciences<br />

Carlson School of Management<br />

<strong>University</strong> of Minnesota<br />

Minneapolis, MN 55455<br />

Phone: (612) 626-3668<br />

Fax: (612) 626-1316<br />

qdai@csom.umn.edu<br />

Robert J. Kauffman<br />

Professor and Chair<br />

Co-Director, MIS Research Center<br />

Information and Decision Sciences<br />

Carlson School of Management<br />

<strong>University</strong> of Minnesota<br />

Minneapolis, MN 55455<br />

Phone: (612) 624-8562<br />

Fax: (612) 626-1316<br />

rkauffman@csom.umn.edu


ABSTRACT<br />

In the recent rapidly changing environment of the Digital Economy, business-to-business (B2B)<br />

electronic <strong>market</strong>s are adopting cooperative <strong>strategies</strong> in lieu of competitive <strong>strategies</strong> in order to obtain<br />

resources so that they can succeed in the <strong>market</strong>. This paper aims to develop formal theory-based<br />

<strong>understanding</strong> of a range of observed cooperative <strong>strategies</strong> by conducting an empirical study of B2B e<strong>market</strong><br />

strategic <strong>alliance</strong>s. We draw upon research on strategic <strong>alliance</strong>s, intermediation and <strong>market</strong> structure<br />

to explore the factors that motivate firms to enter into interorganizational <strong>alliance</strong>s, and make choices about<br />

how to structure and provide governance mechanisms for them. We employ data from various secondary<br />

sources, and address questions about the motivation, structure and impact of <strong>alliance</strong>s. We investigate the<br />

frequency of <strong>alliance</strong>s that B2B e-<strong>market</strong>s form by testing a Poisson count model. We explain how B2B<br />

firms choose <strong>alliance</strong> structures and whether the cooperative strategy affects the success of B2B e-<strong>market</strong>s by<br />

using two different binomial logistic regression models. Our results show that leading B2B e-<strong>market</strong>s tend to<br />

set up cooperative relations more frequently. They also show that strategic <strong>alliance</strong>s are more likely to<br />

involve high levels of interdependence in governance (e.g., joint equity ownership), if partners are buyers or<br />

suppliers in the online <strong>market</strong>place. However, our results indicate that the survival of B2B e-<strong>market</strong>s is not<br />

significantly related to the number of <strong>alliance</strong>s they form.<br />

KEYWORDS. Alliances, e-commerce, electronic <strong>market</strong>s, e-procurement, B2B e-commerce, count data<br />

analysis.


INTRODUCTION<br />

Business-to-business (B2B) electronic <strong>market</strong>s are an innovative form of interorganizational information<br />

systems (IOS), utilizing the Internet and Web technologies to provide shared infrastructure and a means for<br />

commercial exchange. We define a B2B e-<strong>market</strong> as a firm or a subsidiary of a firm that hosts and operates<br />

Internet and Web-based information systems by which other firms can purchase and/or sell products. They<br />

typically offer electronic product catalogs, price discovering mechanisms, and other <strong>market</strong>-making<br />

functions. A recent study published in the McKinsey Quarterly reported that as B2B e-<strong>market</strong>s experience<br />

growth and <strong>market</strong> change, they have come to find it essential to leverage strategic <strong>alliance</strong>s to gain effective<br />

access to products, customers and new business opportunities (Ernst, Halevy, Monier and Sarrazin, 2001).<br />

This is consistent with literature on strategic <strong>alliance</strong>s which argues that firms in rapidly expanding industries<br />

are more likely to form <strong>alliance</strong>s (Mody, 1993; Teece, 1992). Chan, Kensinger and Keown (1997) also find<br />

that a large portion of strategic <strong>alliance</strong>s that are observed in American industry are formed by information<br />

technology (IT) firms.<br />

For startups, a primary determinant for success is the ability to rapidly develop and <strong>market</strong> products and<br />

services to secure positive cash flow and expand <strong>market</strong> share. By forging cooperative relations with other<br />

firms, new firms are able to develop technological and social resources fast. This sets them up in a position<br />

where it is possible, if other aspects of their business are in order, to outperform their competitors. Baum,<br />

Calabrese and Silverman (2000) found that new biotechnology firms that formed more <strong>alliance</strong>s and were<br />

involved in efficient relationships out-performed other firms in the <strong>market</strong> for initial public offerings (IPOs)<br />

of stock.<br />

Such first-mover advantages are especially critical for firms that seek to compete in environments with<br />

strong network effects, which are characterized by “winner takes all” outcomes (Shapiro and Varian, 1999).<br />

B2B e-<strong>market</strong>s, the focus of the present study, are new digital intermediaries that build upon existing<br />

networks of buyers and suppliers. To achieve success, these firms must be early to develop effective new<br />

service offerings, and bringing them to <strong>market</strong> in good is crucial. But how can they accomplish this task<br />

effectively, given the significant resources that must be required? What kind of “functionality” and services<br />

will it take to win over buyers and suppliers in a new <strong>market</strong>place? And what will it take to be successful in<br />

the longer-term, in spite of the down-<strong>market</strong> in e-commerce services? The answer, we believe, lies in the<br />

formation of strategic <strong>alliance</strong>s by B2B e-<strong>market</strong>s with other organizations. We define <strong>alliance</strong>s as formal<br />

cooperative relationships in which firms pool or exchange resources to engage in a joint endeavor, sharing<br />

costs and returns.<br />

Although we have observed the wide adoption of cooperative <strong>strategies</strong> among B2B e-<strong>market</strong>s, further<br />

investigation is needed to provide additional insights and managerial knowledge about how firms can make


more effective B2B e-<strong>market</strong> <strong>alliance</strong>s. In this paper, we will address the following related research<br />

questions:<br />

What are the contents and intended purposes of strategic <strong>alliance</strong>s that are formed by B2B e<strong>market</strong>s?<br />

Under what circumstances will B2B e-<strong>market</strong>s actually be observed to form <strong>alliance</strong>s? What drives<br />

them?<br />

How will the firms that are involved choose the governance structures of the <strong>alliance</strong>s that they<br />

form? How much interdependence will be observed among firms?<br />

And, finally, how will the <strong>alliance</strong>s affect the success of B2B e-<strong>market</strong>s?<br />

To answer these questions, we will draw upon research on strategic <strong>alliance</strong>s, intermediation and <strong>market</strong><br />

structure to explore the factors that motivate firms to enter into interorganizational <strong>alliance</strong>s, and make<br />

choices about how to structure and provide governance mechanisms for them. We collect data from<br />

secondary sources, and identify four main types of B2B e-<strong>market</strong> <strong>alliance</strong>s: <strong>market</strong>ing <strong>alliance</strong>s, participation<br />

<strong>alliance</strong>s, functionality <strong>alliance</strong>s and connection <strong>alliance</strong>s, involving aligned efforts for the enhancement of<br />

<strong>market</strong> service functionality, buyers and suppliers, product and service distribution, and connections with<br />

potential customers, respectively. We investigate the frequency of the observation of <strong>alliance</strong>s by testing a<br />

Poisson count model, and the <strong>alliance</strong> structure using a binomial logistic regression model. We also test a<br />

separate logit model for the effect of <strong>alliance</strong>s on the survival of B2B e-<strong>market</strong>s.<br />

The paper is organized as follows. Section 2 discusses the background literature and briefly reviews<br />

some of the issues that we face that will be handled by the empirical modeling choices that we make in this<br />

work. Section 3 provides definitions for the kinds of B2B e-<strong>market</strong> <strong>alliance</strong> types that we have observed<br />

during the last five years, and presents the research hypotheses that we will test. Section 4 discusses preempirical<br />

data collection and measurement issues that permit us to translate the theoretical model into an<br />

empirical model. Section 5 presents the details of the empirical models that are tested, and provides<br />

background on the variety of modeling considerations that we made to support effective testing of our<br />

study’s hypotheses. Section 6 presents the analysis results. We report on a baseline estimation using a<br />

Poisson regression count data model of observed frequency of occurrence of B2B e-<strong>market</strong> <strong>alliance</strong>s<br />

explained by our theory. We also present results from a binary logistic regression model that identifies what<br />

affect management’s decisions on developing partnerships that exhibit high levels of interorganizational<br />

interdependence.<br />

LITERATURE AND THEORETICAL BACKGROUND<br />

An <strong>alliance</strong> is a formal cooperative relationship between firms that pool or exchange their resources and<br />

share returns from the pooled investment (Teece, 1992). Along with showcasing the efficacy of cooperative


<strong>strategies</strong> among firms that search for partners to improve their competitiveness, the academic literature has<br />

offered a variety of perspectives that address the issues in <strong>alliance</strong>s (Faulkner and De Rond, 2000; Lorange<br />

and Roos, 1992). Firms are motivated to enter into cooperative relationships by the need for obtaining<br />

complementary resources in a speedy, cost-efficient and flexible fashion (Teece, 1992). Moreover, firms can<br />

strengthen their <strong>market</strong> position and deter entry through partnerships (Tirole, 1997; Bamberger, Carlton and<br />

Neumann, 2001). The value of <strong>alliance</strong>s to firms is reflected by excess stock returns that have been observed<br />

upon the announcement of the formation of strategic <strong>alliance</strong>s, and the subsequent better financial<br />

performance that results in comparison with other firms (Chan, Kensinger and Keown, 1997). An <strong>alliance</strong> is<br />

also viewed as a hybrid organizational form, and in this sense, <strong>alliance</strong>s are set up to minimize transaction<br />

costs and to allocate returns according to property rights (Pisano, 1989; Hennart, 1991).<br />

Competitive Advantage from Strategic Alliances<br />

A strategy of cooperation may enable <strong>alliance</strong> partners to achieve a stronger <strong>market</strong> position together<br />

than they would in isolation. For example, through an arrangement called “code sharing,” airlines cooperate<br />

with each other on connecting flight routes, and thus increase their traffic on the shared routes by gaining<br />

<strong>market</strong> share from other airlines (Bamberger, Carlton and Neumann, 2001). Competing airlines also share<br />

airport facilities. The result is that the smaller partner is prevented from entering the incumbent’s <strong>market</strong> on<br />

a large-scale basis, which secures the latter’s <strong>market</strong> position (Chen and Ross, 2000). In addition to<br />

achieving a strong <strong>market</strong> position, another potential advantage of a strategic <strong>alliance</strong> is to permit a firm to<br />

obtain access to new <strong>market</strong> revenue opportunities or new skills through their partners. This is common in<br />

the biotechnology industry, for example, where small biotechnology firms partner with established<br />

pharmaceutical firms so that the former obtains access to <strong>market</strong> while the latter obtains knowledge in<br />

developing the new drugs (Lerner and Merges, 1997). It is also worthwhile to point out that the<br />

organizational form of strategic <strong>alliance</strong>s gives firms the flexibility of forming and disbanding linkages with<br />

partners swiftly in response to changes in demand or other aspects of their business environment (Mody,<br />

1993; Chan, Kensinger and Keown, 1997). Mody (1993) pointed out that such flexibility enables firms to<br />

explore new technologies and skills without risking financial over-commitment and the potential for financial<br />

distress, and is the most valuable for rapidly growing firms or firms that compete in environments<br />

characterized by rapid changes in product and process technologies.<br />

The advantages of <strong>alliance</strong>s that we cite are realized by firms through the access they obtain to<br />

complementary resources at a lower cost than if they were to develop the capabilities internally (Teece,<br />

1992). The <strong>alliance</strong> literature recognizes three kinds of critical resources in this context: technical,<br />

commercial and social resources (Ahuja, 2000). Technical resources are the skills and capabilities for<br />

developing and offering new products. Commercial resources include firm <strong>market</strong>ing and distribution skills<br />

that can bring products to customers. Social resources reflect the linkages that firms have already formed


and can be leveraged to obtain other resources. Ahuja (2000) showed that firms with more resources are<br />

attractive potential partners but are less inclined to partner with other organizations with few resources.<br />

Strategic <strong>alliance</strong>s represent an even more important strategy for new firms that bring a new form of<br />

business to their customers in the <strong>market</strong>. It is typical that their capabilities (in product development,<br />

support and extension, for example) are doubted. They require assistance to gain legitimacy in the<br />

<strong>market</strong>place in addition to their eager demand for resources. For example, new biotechnology firms signal<br />

their research capabilities by partnering with incumbent pharmaceutical firms (Nicholson, Danzon and<br />

McCollough, 2002). In addition, Baum, Calabrese and Silverman (2000) showed that startup firms that<br />

formed more <strong>alliance</strong>s outperformed their competitors on IPOs. They also provided evidence to suggest that<br />

it is important to achieve relationship efficiencies in strategic <strong>alliance</strong>s. Overall, the research that we cite<br />

indicates that the formation of strategic <strong>alliance</strong>s represents a success factor for startups that promote access<br />

to necessary resources and signaling of their capabilities through the leverage that their partners’ resources<br />

create.<br />

Transaction Costs and Property Rights in Strategic Alliances<br />

If we view strategic <strong>alliance</strong>s as a hybrid organizational form, transaction cost theory gives us the ability<br />

to obtain insights into the governance forms that <strong>alliance</strong>s adopt related to the circumstances under which<br />

they are formed (Williamson, 1989 and 2000). When partnering firms pool their resources or share resources<br />

with each other, they make mutual commitments to relationships which typically are characterized by a<br />

higher level of bilateral dependency than is the case when they use other alternatives in the broader <strong>market</strong>.<br />

However, in contrast to hierarchies, in which one set of owners and managers typically can exercise<br />

unilateral authority, partners share rights to control and monitor activities. They also have the opportunity to<br />

determine how to share returns on the investment. Hence, in such relationships, there are potential<br />

opportunistic behaviors that may diminish the gains from the cooperation (Williamson and Masten, 1995).<br />

To overcome this problem, firms can resort to <strong>strategies</strong> that hold one another “mutually hostage,” for<br />

example, by investing in technology assets that are specific to the interorganizational collaboration, or<br />

obtaining legally enforcable contractual assurances of mutual commitment and non-competition. When they<br />

have a large stake in the relationships or a high degree of uncertainty is involved in the cooperative activities,<br />

firms are more likely to use mechanisms such as equity-based joint ventures as a means to foil opportunism<br />

(Hennart, 1991; Oxley, 1997). Moreover, Allen and Phillips (2000) also showed in an empirical study that<br />

firms obtain maximal benefits when they form <strong>alliance</strong>s while making equity investments.<br />

Partnering firms most often share costs and returns according to the terms specified in formal contracts<br />

they set up at the beginning of their cooperation. However, due to uncertainties in the business environment,<br />

firms typically are not able to include solutions to all possible contingencies that may arise relative to their<br />

contracts, and so they leave numerous issues open for renegotiation, including the sharing or division of<br />

returns. The theory of incomplete contracts (Grossman and Hart, 1986; Hart, 1988) points out that the


argaining power of a firm in an economic exchange is determined by the assets it owns in the relationship.<br />

From this point of view, ex ante property rights ought to determine the ex post allocation of returns when<br />

unexpected situations occur. Therefore, in a strategic <strong>alliance</strong> between firms, the firms ought to be willing to<br />

make a considerable investment in the shared assets if they expect high returns to flow from the cooperative<br />

relationship.<br />

Such practices are noted in an empirical study on biotechnology firms that made equity investments<br />

while initiating joint research and development projects (Pisano, 1989). This gives them significant rights<br />

over the new technologies and products coming out of the joint research and development efforts. Although<br />

equity position is a frequently used mechanism in controlling opportunistic behavior, previous firm<br />

experience with <strong>alliance</strong> will mitigate the need for partial ownership (Robinson and Stuart, 2000). Another<br />

argument based on the theory of incomplete contracts is that in an interfirm coalition, the parties that are<br />

indispensable to the relationship should own the assets that are important to them so that they have the<br />

incentive to make investments in the coalition to maximize the total value of the coalition (Hart and Moore,<br />

1990). In the context of electronic networks, Bakos and Nault (1997) have shown that optimal investment<br />

levels can be achieved when firms that are indispensable to the network participate in the ownership of the<br />

network asset.<br />

In addition to partial ownership, exclusive contracts also turn out to be an effective means for protecting<br />

relation-specific investment considering the incompleteness of interfirm contracts (Segal and Whinston,<br />

2000). These arguments all shed some light on the structure of B2B e-<strong>market</strong> <strong>alliance</strong>s.<br />

B2B Electronic Markets and Digital Intermediation<br />

The basic task of B2B e-<strong>market</strong>s is to enable firms to find desired products, suppliers and customers, or<br />

to create <strong>market</strong>s on the Web (Dai and Kauffman, 2002). They either move the conventional <strong>market</strong>s to the<br />

Web or open new <strong>market</strong>places online which do not have offline counterparts. They also act as <strong>market</strong>making<br />

electronic intermediaries whose value lies in reducing search costs, increasing <strong>market</strong> liquidity,<br />

offering transaction facilitation mechanisms and procurement expertise (Bailey and Bakos, 1997; Bakos,<br />

1997; Chircu and Kauffman, 2001). In addition to <strong>market</strong>-making, B2B e-<strong>market</strong>s also perform another two<br />

roles. The first role is to offer services and products for buyers and suppliers to manage interorganizational<br />

processes and relationships, and the second one is to offer technology adaptation functionalities that<br />

promote interoperability, systems integration and cost-effective connectivity between trading networks (Dai<br />

and Kauffman, 2002). To fulfill these roles, B2B e-<strong>market</strong>s are building up resources and capabilities<br />

through both organic growth and partnerships with other organizations.<br />

B2B e-<strong>market</strong>s that operate on the Internet are relatively new organizational forms in business, and are<br />

often perceived as startups. The implication of this perception is that their business models and capabilities<br />

still need to be recognized and accepted in the <strong>market</strong>place. This is especially true in the recent post-<br />

DotCom boom environment; today most businesses that are based on the Internet are perceived to be very


isky and the <strong>market</strong>place still is going through a shakeout. One way for B2B e-<strong>market</strong>s to win recognition<br />

of their capabilities and value propositions as digital intermediaries is to build their reputation and signal<br />

their quality through partnering with firms that have a strong reputation in the <strong>market</strong>place.<br />

The emergence of B2B e-<strong>market</strong>s also has brought about new opportunities and challenges for the<br />

industry groups that they are serving. For example, they provide new procurement and distribution channels<br />

for the firms that manufacture or consume the products that are transacted over their online <strong>market</strong>places.<br />

And, they represent as a potential threat to traditional distributors serving the same industry groups, since<br />

they enable firms to bypass the traditional intermediaries to transact directly online. In other words, they<br />

bring new competition to the <strong>market</strong> contexts in which traditional intermediaries have competed.<br />

B2B E-MARKET STRATEGIC ALLIANCE FORMATION, STRUCTURE AND OUTCOME<br />

As new organizational forms in the rapidly changing environment of Internet commerce, B2B e-<strong>market</strong>s<br />

need to accumulate resources and build their reputations so that they can “perfect“ their value-added services<br />

and gain recognition in the <strong>market</strong>place. In addition to internal growth, cooperation with other organizations<br />

offers an alternative way to achieve these capabilities. To understand how B2B e-<strong>market</strong>s employ <strong>alliance</strong><br />

<strong>strategies</strong>, we develop a conceptual model by applying the perspectives about <strong>alliance</strong>s in general to B2B e<strong>market</strong>s,<br />

and propose related research hypotheses.<br />

B2B E-Market Alliance Types<br />

In the strategy literature (e.g., Pisano, 1989; Chan, Kensinger and Keown, 1997), strategic <strong>alliance</strong>s are<br />

often categorized according to the tasks that they accomplish. Some of these categories include: <strong>market</strong>ing<br />

and distribution <strong>alliance</strong>s, <strong>alliance</strong>s for joint development of technology, technology transfer <strong>alliance</strong>s, and<br />

manufacturing <strong>alliance</strong>s. Following this rationale, we propose four different kinds of B2B e-<strong>market</strong> <strong>alliance</strong><br />

types.<br />

Marketing <strong>alliance</strong>s permit B2B e-<strong>market</strong> firms to promote and distribute their services.<br />

Participation <strong>alliance</strong>s support the creation of cooperative relationships by B2B e-<strong>market</strong> firms with<br />

other firms that buy and sell on their exchanges. We refer to them as such because the goal is to<br />

ensure the participation of buyers and suppliers in the <strong>market</strong>place.<br />

Functionality <strong>alliance</strong> allow B2B e-<strong>market</strong>s to cooperate with other firms to enhance the set of<br />

functionalities that they offer to facilitate online transactions (Dai and Kauffman, 2002).<br />

Connection <strong>alliance</strong>s are those in which a B2B firm sets up linkages with partners so that partners’<br />

clients can have integrated or preferred access to the electronic <strong>market</strong>places that the B2B firm is<br />

operating.<br />

Table 1 provides an illustration of each of these, through announcements in the press. (See Table 1.)


Table 1. Examples of the Proposed B2B E-Market Alliance Types<br />

ALLIANCE<br />

TYPES<br />

Marketing<br />

Alliance<br />

Participation<br />

Alliance<br />

Functionality<br />

Alliance<br />

Connection<br />

Alliance<br />

EXAMPLES OF ANNOUNCEMENTS<br />

ProNetLink.com (www.ProNetLink.com), the Global Trade Internetwork (OTC: PNLK), today<br />

announced that the Company has finalized a strategic <strong>alliance</strong> with the NetlinQ Group of the<br />

Netherlands for the <strong>market</strong>ing and promotion of ProNetLink.com to businesses and professional<br />

associations throughout Holland. The <strong>alliance</strong> comes as a result of ProNetLink.com's four-week<br />

European <strong>market</strong>ing mission, arranged by the United States Department of Commerce, in July<br />

1999. Under the terms of the agreement, NetlinQ's NetPlus division will promote ProNetLink.com<br />

through a series of advertising and <strong>market</strong>ing initiatives including trade show development,<br />

national advertising campaigns and introductions to NetlinQ's current roster of clients. (Just-<br />

Style.com, 1999)<br />

May 11. DuPont (Wilmington, DE) says it will use the specialized e-<strong>market</strong>place AssetTRADE<br />

(King of Prussia, PA) to buy and sell used equipment–both internally and externally–and will also<br />

take a minority stake in the company. (e-Chemmerce.com, 2001)<br />

Byers Engineering Company and bandwidth.com today announced a strategic <strong>alliance</strong> to offer the<br />

telecommunications industry a unique matchmaking service aimed at reducing the cost of<br />

constructing new fiber routes, wireless networks, and central offices throughout the US and<br />

internationally. Under the terms of the agreement, Byers Engineering Company and<br />

bandwidth.com will co-manage the new service jointly designed by both companies. The co-build<br />

matchmaking service is specifically targeted to provide the telecommunications industry a new and<br />

innovative tool to identify potential partners interested in the co-construction of a variety of<br />

network facilities and infrastructure. (PRWeb, 2000)<br />

ChemCross.com, Asia's largest e-<strong>market</strong>place and portal for chemical companies, and<br />

CheMatch.com a leading Internet-based <strong>market</strong>place and information resource for buying and<br />

selling bulk commodity chemicals, plastics, feedstocks and fuel products announced today that they<br />

have formed a strategic <strong>alliance</strong>. ChemCross and CheMatch have entered into a mutual agreement,<br />

which will allow ChemCross access to CheMatch's Global Trading Network and information<br />

resources. This access will allow ChemCross to <strong>market</strong> CheMatch postings to their rapidly<br />

increasing membership of more than 2000 corporate members. Likewise CheMatch will have<br />

access to ChemCross' platform and information resources to <strong>market</strong> relevant petrochemical<br />

postings to CheMatch's more than 700 member companies (eChemPeople, 2001).<br />

Discriminating this way among the different kinds of strategic <strong>alliance</strong>s, prompts us to consider the kinds<br />

of resources they can bring to B2B e-<strong>market</strong>s. Through <strong>market</strong>ing <strong>alliance</strong>s and connection <strong>alliance</strong>s, a B2B<br />

e-<strong>market</strong> will try to extend its reach to potential customers in other <strong>market</strong> segments, and obtain commercial<br />

resources that are deployed for effective delivery of products and services to customers. Functionality<br />

<strong>alliance</strong>s enable B2B e-<strong>market</strong>s to develop and enhance the services they offer to buyers and suppliers. They<br />

are mainly meant to enable a firm to obtain technological resources that it does not have. Participation<br />

<strong>alliance</strong>s are not set up to acquire access to technological or commercial resources. The rationale is that<br />

participation <strong>alliance</strong>s strengthen the relationship between the B2B e-<strong>market</strong>s and their clients who use the<br />

online <strong>market</strong>places for interfirm transactions. Thus, they enable B2B e-<strong>market</strong>s to gain access to another<br />

critical resource: relational resources.


Research Hypotheses<br />

As our first effort to develop a formal <strong>understanding</strong> about the B2B e-<strong>market</strong> <strong>alliance</strong> <strong>strategies</strong>, we<br />

look into <strong>alliance</strong> formation, <strong>alliance</strong> structure, and <strong>alliance</strong> outcomes, formulate research hypotheses.<br />

Alliance Formation. In general, firms adopt cooperative <strong>strategies</strong> to achieve advantageous <strong>market</strong><br />

positions by obtaining complementary resources. Two sets of factors encourage a firm to develop external<br />

linkages for growth. On the one hand, a firm will have the incentive to cooperate with another organization<br />

when it can gain significant value from this joint effort; on the other hand, it is only able to find desirable<br />

partners when it has shown that it too possesses the necessary resources that will attract partners (Ahuja,<br />

2000). This general rule also may explain the differences among B2B e-<strong>market</strong>s in terms of how often they<br />

are observed to form <strong>alliance</strong>s.<br />

In addition, we expect that B2B e-<strong>market</strong>s are more likely to try to leverage cooperation with other<br />

organizations when such cooperation is perceived to be more beneficial than going it alone. As such, the<br />

perceived value of strategic <strong>alliance</strong>s should be a key driver of their formation. As new organizational forms,<br />

B2B e-<strong>market</strong>s face the critical task of gaining acceptance of their business models and recognition for their<br />

core competencies and capabilities. Since partnering with established firms is an effective means to enhance<br />

reputation and signal product quality to potential customers (Rao and Ruekert, 1994), the need for <strong>market</strong><br />

recognition or legitimacy is likely to motivate B2B e-<strong>market</strong>s to enter into cooperative relationships. This is<br />

especially true for B2B e-<strong>market</strong>s that were founded at the very early days of B2B e-commerce. Why?<br />

Because they faced the challenges of opening up new <strong>market</strong>s for their innovative approaches to doing<br />

business online. Therefore, we propose Hypothesis #1.<br />

H1: The Pioneer B2B E-Market Hypothesis. B2B e-<strong>market</strong>s that were founded earlier will<br />

be observed to form more <strong>alliance</strong>s than later entrants.<br />

Another characteristic of B2B e-<strong>market</strong>s related to their nature as startups is that firms purchasing on the<br />

online <strong>market</strong>places will tend to perceive high procurement risks associated with electronic <strong>market</strong>places<br />

compared to the conventional procurement channels (Chircu and Kauffman, 2001). This perception, in turn,<br />

will affect the perceived effectiveness of B2B e-<strong>market</strong>s in facilitating <strong>market</strong>s for different procurement<br />

needs. Specifically, in the presence of high channel uncertainty, firms will be more willing to use B2B e<strong>market</strong>s<br />

for purchasing indirect products which have low strategic significance (Kauffman and Mohtadi,<br />

2002).<br />

Concerns about data transparency in electronic <strong>market</strong>s may also make suppliers cautious about joining<br />

(Zhu, 2002). They would like to avoid the price competition that might be engendered by electronic<br />

<strong>market</strong>s. Both buyers and suppliers are likely to view electronic <strong>market</strong>places as a riskier channel for<br />

transacting strategic products or exchanging complex and strategic information (Dai and Kauffman, 2000).


Thus, we believe that B2B e-<strong>market</strong>s will face more challenges and uncertainties in gaining recognition and<br />

achieving critical mass adoption when they are serving buyers and suppliers who are involved in large-scale<br />

or strategic transactions or products. When this is the case, we argue, e-<strong>market</strong> firms will have greater<br />

incentive to search for external support. This leads us to Hypothesis #2:<br />

H2: The Strategic Product Hypothesis. B2B e-<strong>market</strong>s that deliver strategic products to buyers<br />

will form more <strong>alliance</strong>s than those that are involved in non-strategic products.<br />

The opportunity for partnering is another factor that determines the frequency with which we are<br />

observing the formation of strategic <strong>alliance</strong>s among B2B e-<strong>market</strong>s. Why? Because the purpose of<br />

strategic <strong>alliance</strong>s is to obtain complementary resources, and firms that control or own more resources<br />

typically will be attractive <strong>alliance</strong> targets. Market leaders enjoy higher reputation and are perceived to have<br />

more technological and commercial resources. As a result, we expect that they will be more likely to<br />

develop partnerships. In our context, some B2B e-<strong>market</strong>s are viewed as <strong>market</strong> leaders. They typically<br />

achieve greater visibility, and thus will have more opportunities to form <strong>alliance</strong>s. This leads us to propose<br />

Hypothesis #3:<br />

H3: The Leading B2B E-Market Hypothesis. Leading B2B e-<strong>market</strong>s form more <strong>alliance</strong>s than<br />

others.<br />

Alliance Structure. To protect relationship-specific investments against potential opportunism and<br />

provide incentives to partners to invest in the mutually-beneficial relationship, firms involved in an economic<br />

exchange tend to incorporate partial ownership or exclusivity in their contracts (Segal and Whinston, 2000;<br />

Hennert, 1991). Equity investment increases the controlling and monitoring rights in partners, while<br />

exclusive contracts increase the level of commitment to the relationship. Both mechanisms bind the<br />

partnering firms closer and make them more dependent on each other. In this paper, we will refer to such<br />

relationships as exhibits a high level of interdependence.<br />

B2B e-<strong>market</strong>s are trading networks whose growth creates network externalities. The value of an<br />

electronic <strong>market</strong>place increases with the number of firms that adopt it for procurement transactions. Also,<br />

to the extent that a B2B e-<strong>market</strong> needs to reach a critical mass of participating firms to survive, participating<br />

buyers and suppliers are indispensable.<br />

In addition, to achieve efficient interfirm transactions via the electronic <strong>market</strong>place, both the focal B2B<br />

e-<strong>market</strong> and the firms that are buying and selling on the <strong>market</strong>place ought to make investments in their<br />

information system infrastructures and probably the business processes as well. Bakos and Nault (1997)<br />

utilize the property rights theory of Hart and Moore (1990) to show that that indispensable parties in an<br />

electronic network should own the network assets to ensure that optimal investment levels in IT are achieved.


The electronic linkage between a firm and a B2B e-<strong>market</strong> for integrating their systems and streamlining<br />

business processes is often customized for a particular relationship and cannot be switched for other<br />

applications easily. As a result, the related IT investments are relationship-specific. As a result, we expect<br />

that the relationship between a B2B e-<strong>market</strong> and a participating firm will exhibit a relatively high level of<br />

interdependence. Thus, we propose Hypothesis #4:<br />

H4: The Participation Alliance Hypothesis. Participation <strong>alliance</strong>s are more likely to involve high<br />

levels of interdependence than other kinds of <strong>alliance</strong>s.<br />

Alliances offer opportunities for transferring tacit knowledge between partners, and firms take<br />

measures to protect their core competence from spillover effects, especially when the partners are<br />

competitors or potential competitors (Dutta and Weiss, 1997). The basic argument is that technologically<br />

innovative firms tend to enter into relationships that minimize the chances of tacit knowledge transfer. As<br />

innovators in utilizing the Internet and Web technologies for conducting business online, B2B e-<strong>market</strong>s rely<br />

on their abilities to implement business ideas involving new ITs to succeed in the <strong>market</strong>place. So they<br />

ought to try to limit tacit knowledge transfer when they enter into partnerships. Knowledge spillover is most<br />

likely to occur between competitors due to the large amount of co-specialization, or overlap and<br />

compatibility in technology and skills (Chan, Kensigner and Keown, 1997). As a result, B2B e-<strong>market</strong>s will<br />

try to reduce the chances that their tacit knowledge leaks. This ought to be exhibited in the <strong>alliance</strong><br />

structures that B2B e-<strong>market</strong>s set up with their competitor partners, as we suggest in Hypothesis #5:<br />

H5: The Competitor Alliance Hypothesis. B2B e-<strong>market</strong>s are less likely to form highly<br />

interdependent <strong>alliance</strong>s with competitors and potential competitors.<br />

Alliance Outcomes. Previous research on the value of <strong>alliance</strong> <strong>strategies</strong> shows that they increase<br />

shareholder value and improve the long-term performance of the firm (Chan, Kensinger and Keown, 1997).<br />

In addition, Baum, Calabrese and Silverman (2000) found that the number and efficiency of <strong>alliance</strong>s that<br />

Canadian biotechnology startups formed at their early days is positively related to their IPO performance.<br />

These findings also suggest that <strong>alliance</strong> <strong>strategies</strong> enhance B2B e-<strong>market</strong>s competitive advantage and lead<br />

to better competitive performance. In our context, we will use the survival of the firm as a proxy for its<br />

relative success in the <strong>market</strong>. With these ideas in mind, we propose Hypothesis #6, our last.<br />

H6: B2B E-Market Survival Hypothesis. B2B e-<strong>market</strong>s that have formed more <strong>alliance</strong>s are<br />

more likely to survive.


DATA COLLECTION AND VARIABLES<br />

We next present an overview of data collection, measurement issues and descriptive statistics for the<br />

variables in the study that we will use to test the theory discussed earlier.<br />

Data Collection<br />

To study the research issues, we collected data from Thomson Financial’s (www.tfn.com) Joint<br />

Venture/Strategic Alliances database. This database provides a “one-stop” information source for publicly-<br />

available announcements, including SEC filings, trade publications and international counterparts, and news<br />

wire sources. This database is populated by announcements that Thomson financial collected using such<br />

keywords as “<strong>alliance</strong>”, “manufacturing agreements”, “<strong>market</strong>ing agreements,” “licensing agreements,” and<br />

other related terms.<br />

Data Set and Unit of Observation. We retrieved data entries from January 1995 to February 2002<br />

which at least one participant had an e-commerce business line, or <strong>alliance</strong> activities were reported in the e-<br />

commerce area. This generated 6,241 entries. We filtered these in two steps, retaining <strong>alliance</strong><br />

announcements with at least one participating firm that was operating a B2B e-<strong>market</strong>. Our selection<br />

criteria involved assessments of the database’s business description for <strong>alliance</strong> participants and company<br />

Web sites’ business descriptions. After we completed filtering the data, 426 entries remained. We then<br />

supplemented the Thomson Financial data with Lexis-Nexis (www.lexisnexis.com) information on the same<br />

<strong>alliance</strong> announcements, and retained those data with entries in both databases. We also checked the<br />

business descriptions for the firms to assess whether each could be reasonably claimed to be a B2B e-<strong>market</strong>.<br />

This process resulted in 332 <strong>alliance</strong> entries, involving 200 different B2B e-<strong>market</strong>s.<br />

The unit of observation in our sample is a strategic <strong>alliance</strong> event that is initiated by a business<br />

establishment and is accompanied by an identifiable announcement or news item that describes the contents<br />

of the <strong>alliance</strong>. A business establishment can be a company, a branch or subsidiary of a firm. For example,<br />

Getthere LP (www.getthere.com), a specialist in the area of travel procurement software solutions, operates a<br />

B2B e-<strong>market</strong> and is wholly owned by Sabre Holdings (www.sabre.com). We treat Getthere LP as a<br />

business establishment, and include its announcements in our data set, in spite of its wholly-owned<br />

subsidiary status. In contrast, SciQuest (www.sciquest.com), which offers Internet-based procurement<br />

solutions to pharmaceutical, biotechnology and other research-based organizations in the life sciences<br />

industry, runs its own online <strong>market</strong>place. So it is also included in our data set.<br />

Identification of Market Characteristics. To identify and evaluate relevant characteristics of B2B e-<br />

<strong>market</strong>s and their partnering firms, we compiled data from various sources. For public traded firms, we<br />

collected data from the Mergent FIS online database (www.fisonline.com). For private firms, we used<br />

company Web sites, the Lexis-Nexis database, and the United States Patent and Trademark Office’s<br />

(USPTO) “TESS” Trademark Electronic Search System (tess.uspto.gov). Using these data, we coded the<br />

characteristics of B2B e-<strong>market</strong>s and partnering firms (e.g., year B2B e-<strong>market</strong> was founded, and product


types transacted). The details about variable definition and measurement are described in the following<br />

subsection.<br />

Coding of Variables.<br />

We identified and coded a set of variables in four categories: <strong>alliance</strong> characteristics, partnership<br />

characteristics, B2B e-<strong>market</strong> firm characteristics and product characteristics. The variable names and<br />

definitions are summarized in Table 2. (See Table 2.)<br />

Table 2. Variable Definitions and Measurement<br />

VARIABLES DEFINITIONS<br />

Alliance Characteristics: Continuous Variables<br />

AllianceTiming Log of number of months elapsed from January 1995 to month <strong>alliance</strong> announced<br />

#Alliances Number of <strong>alliance</strong>s<br />

Alliance Characteristics: Binary Variables<br />

MktgAlliance Marketing <strong>alliance</strong>: to promote and distribute B2B e-<strong>market</strong>’s products or services<br />

ParticAlliance Participation <strong>alliance</strong>: Partnership with firms that participate in e-<strong>market</strong>’s online<br />

<strong>market</strong>place as buyers or sellers<br />

FnctAlliance Functionality <strong>alliance</strong>: Enhances functionality<br />

ConnAlliance Connection <strong>alliance</strong>: Obtains access to potential customers<br />

OtherAlliance Alliances for purposes other than the above four types<br />

InterdepLevel Level of interdependence: if <strong>alliance</strong> involves equity position or exclusive agreement<br />

Partner Characteristics: Binary Variables<br />

Intermediary Traditional intermediary<br />

InternetFirm Internet firm<br />

TradeAssoc Trade association<br />

TraditionalFirm Brick-and-mortar firm<br />

B2B E-Market Characteristics: Binary Variables<br />

MktLeader Market leader.<br />

VerticalExch Industry-specific B2B exchange.<br />

ConsortExch Industry consortium-supported B2B exchange<br />

FoundYrij Year when e-<strong>market</strong> firm i was founded (1986-2001), with dummies FoundYr1994 (17<br />

firms founded), FoundYr1995 (15 firms founded), FoundYr1996 (14), FoundYr1997 (11),<br />

FoundYr1998 (24), FoundYr1999 (70), FoundYr2000-2001 (49). 2001 had just 1 firm founded,<br />

so we aggregated it with the firms that were founded in 2000 to form the base case year.<br />

CensoredObs Censored observation: still operating as of August 9, 2002<br />

Exchange Product Characteristics: Binary Variables<br />

DigitalSvcs Business services, digital products transacted<br />

MROSvcs MRO, office supplies transacted<br />

DirectGoods Direct products (e.g., raw materials, parts) transacted<br />

ConsumerGoods Consumer goods transacted<br />

CapitalEquip Capital equipment transacted<br />

OtherGoods Other goods, or all product types transacted<br />

The reader should note the extent of our use of binary variable codings, since many of the variables<br />

indicate the presence or absence of various characteristics. In addition, it is worthwhile to point out that the<br />

binary variable codings, in some cases, do not indicate exclusive categorizations of what a B2B e-<strong>market</strong><br />

does in its business. Instead, it is possible that a firm may have a number of characteristics that are taken


from among a group of variables. This permits us to include binary variables without specifying a “base<br />

case,” as is typical when there are a number of different coding categories.<br />

The Alliance Characteristics data are both continuous and binary measures. The continuous variable,<br />

AllianceTiming, represents when the <strong>alliance</strong> was announced publicly, and its value is the log of the number<br />

of months elapsed from January 1995 to the month when the <strong>alliance</strong> was announced. The starting month of<br />

January 1995 was determined based on the overall coverage of our data set. InterdepLevel is the level of<br />

interdependence between the partnering firms, and is one of the key variables that we will use to test our<br />

theory. We coded it with a “1” if the <strong>alliance</strong> involves equity investment or exclusive agreement, and “0”<br />

otherwise. When equity investment or exclusive agreements are present, they indicate a high level of<br />

interdependence. With this approach, the partnering firms are able to assert more control and demonstrate<br />

greater commitment to their mutual relationships. #Alliances is the total of <strong>alliance</strong>s that a B2B e-<strong>market</strong><br />

forms during the time period of our study, from January 1995 up to February 2002.<br />

Our codings for the <strong>alliance</strong> types discussed earlier in this paper are all binary. MktgAlliance codes for<br />

whether the <strong>alliance</strong> aims to promote and distribute the B2B e-<strong>market</strong>’s products and services. The key<br />

phrases that we used for identifying <strong>market</strong>ing <strong>alliance</strong>s were “<strong>market</strong>ing agreement,” “joint <strong>market</strong>ing and<br />

sales,” and “jointly <strong>market</strong> and distribute,” among others. With the search tools that we used, it is also<br />

possible to do the typical searches that search engines support, such as “+joint +<strong>market</strong>ing +<strong>alliance</strong>,” to<br />

require each of the three words to be present in the output to a query. ParticAlliance codes for whether the<br />

B2B e-<strong>market</strong> obtains a participant in its online <strong>market</strong>place through this <strong>alliance</strong>. The key phrases that we<br />

used for coding this variable are roughly as follows: “(Firm A) will use (firm B’s) <strong>market</strong>place to buy (sell),”<br />

“(Firm A) chooses (Firm B) as its online provider,” and so on. Precise queries were not easy, but practice<br />

with the tools permitted us to develop a reasonable level of assurance that we were capturing most of the<br />

necessary strategic <strong>alliance</strong> announcements of this type.<br />

FnctAlliance was coded with a “1” for instances in which the B2B e-<strong>market</strong> announced the enhancement<br />

of service capabilities and functionality in facilitating the <strong>market</strong>, supporting relationship and process<br />

management between buyers and suppliers, and/or improving its technology infrastructures, and “0”<br />

otherwise. The key phrases for the announcement search queries included “jointly develop (service),”<br />

“create new function,” “create new service,” “add new offering,” and other combinations of these general<br />

query terms. Finally, ConnAlliance codes for B2B e-<strong>market</strong>s that tried to extend their reach to potential<br />

users through increased connectivity. Key phrases for search included language similar to the following:<br />

“(<strong>alliance</strong>) gives the customers direct access to (firm B),” “integrate (firm A system) with (firm B system),”<br />

“improve access to customers,” and so on. Although these four types cover most of the <strong>alliance</strong> tasks, there<br />

are other purposes for <strong>alliance</strong>s, and we use OtherAlliance to represent them.<br />

There are four Partner Characteristics variables, all of which are coded 0/1. Intermediary indicates if<br />

the partner of the B2B e-<strong>market</strong> is a conventional intermediary. InternetFirm codes for whether the partner


conducts its major value-added activities on the Internet and Web. This includes Internet commerce firms<br />

that conduct business over the Internet and also Internet service firms that make the Internet itself and the<br />

business conducted through it possible (e.g., Dow Jones Internet Index, www.djindexes.com/jsp/iiFaq.jsp).<br />

TradeAssoc is an indicator to show if the partner is a not-for-profit trade association. TraditionalFirm<br />

indicates if the partner is a brick-and-mortar company. It has a value of “1” if the partner conducts its major<br />

value-added activities offline; otherwise it is “0”. According to this coding, a partner firm must fall into one<br />

and only one type among the following three: InternetFirm, TradeAssoc, and TraditionalFirm. However, a<br />

traditional firm can also be an intermediary.<br />

The next group of binary variables, B2B E-Market Variables, is intended to capture information about<br />

the B2B e-<strong>market</strong>s, but unrelated to their strategic <strong>alliance</strong>s. MktLeader indicates if the B2B e-<strong>market</strong> is a<br />

<strong>market</strong> leader as so designated by Forbes magazine’s “Best-of-the-Web” B2B directories for 2000 and 2001<br />

(available at www.forbes.com/bow/). This is a directory of firms that industry experts perceived as most<br />

promising or active, based on their strategy, execution, and financial status. To qualify for this designation in<br />

our data set, the firm had to be listed in either of the two years. We use the next subset of variables to<br />

characterize the exchange activities of the B2B e-<strong>market</strong> firm. VerticalExch indicates whether the B2B e-<br />

<strong>market</strong> serves a specific industry or a specific business function, which defines it as a vertical exchange.<br />

ConsortExch codes for whether the B2B e-<strong>market</strong> was initiated and backed up by an industry consortium.<br />

We also distinguished among the B2B e-<strong>market</strong> firms in terms of when they were established or founded<br />

through the variable, FoundYr. The fact is that many B2B e-<strong>market</strong> firms failed during the time period that<br />

our data covered, however, some did not and continue to operate even today. To capture this information,<br />

we included another 0/1 variable, CensoredObs, to indicate if the B2B e-<strong>market</strong> was still operating as August<br />

9, 2002, when we last checked its operational status.<br />

The final set of variables that we consider represent the kind of e-procurement activities that a B2B e<strong>market</strong><br />

firm is handling, that is, its Exchange Product Characteristics. DigitalSvcs is coded with a “1” if the<br />

product transacted on the electronic <strong>market</strong>place is business services or information products, and with a “0”<br />

otherwise. MROSvcs indicates that a B2B e-<strong>market</strong> serves a <strong>market</strong>place for purchasing maintenance, repair<br />

and operation (MRO) services and products, or office products. DirectGoods codes for whether there are<br />

buyers on the e-<strong>market</strong>place who purchase their raw materials, parts, and/or components that go into their<br />

own manufacturing and production processes. ConsumerGoods indicates that the B2B e-<strong>market</strong> has buyers<br />

who purchase goods that they resell to consumers. CapitalEquip denotes that firms on the e-<strong>market</strong>place<br />

purchase and/or sell equipment to balance their inventory. Finally, OtherGoods indicates B2B e-<strong>market</strong>s that<br />

do not specify the types of goods that can be transacted on their <strong>market</strong>place, or that allow firms to transact<br />

any types of goods.


Description of the Data Set<br />

In our data set, there are 200 B2B e-<strong>market</strong>s, among which 71 or 36% are <strong>market</strong> leaders that are listed<br />

in Forbes’ “Best of Web” directories. The majority, 68%, of the B2B e-<strong>market</strong>s are vertical exchanges. The<br />

distribution of the founding years of the B2B e-<strong>market</strong>s is shown in Table 3.<br />

Table 3. Distribution of B2B E-<strong>market</strong>s by Year Founded<br />

Year Founded Number of B2B E-<strong>market</strong>s<br />

Before 1995 17<br />

1995 15<br />

1996 14<br />

1997 11<br />

1998 24<br />

1999 70<br />

2000 48<br />

2001 1<br />

Total 200<br />

Many B2B e-<strong>market</strong>s serve more than one product type. Table 3 shows the breakdown of B2B e-<br />

<strong>market</strong>s by the product types that are transacted. (See Table 4.)<br />

Table 4. Distribution of B2B E-Markets by Product Type<br />

Product Type Number of B2B E-<strong>market</strong>s<br />

Business services, digital products (DigitalSvcs) 74<br />

Direct products (DirectGoods) 84<br />

Consumer goods (ConsumerGoods) 21<br />

MRO and office supplies (MROSvcs) 38<br />

Capital equipment (CapitalEquip) 14<br />

In total, we identified 353 strategic <strong>alliance</strong> events in our data set. Among these, 31 <strong>alliance</strong>s had three<br />

partners listed in their announcements. To maintain some explanatory consistency in our modeling, we<br />

chose to eliminate those <strong>alliance</strong>s with three partners; only bilateral <strong>alliance</strong>s are included. This yielded 332<br />

usable strategic <strong>alliance</strong>s in our data set. The distribution of the <strong>alliance</strong>s over the years of the study is<br />

summarized in Table 5. (See Table 5.)<br />

Table 5. Distribution of Alliances by Year<br />

Year Number of Alliance Events<br />

1998 4<br />

1999 30<br />

2000 218<br />

2001 75<br />

2002 5<br />

Total 332


Table 6. Distribution of Alliances by Type<br />

Table 7. Descriptive Statistics<br />

Alliance Type Number of Alliance Events<br />

Marketing 92<br />

Participation 76<br />

Functionality 128<br />

Connection 89<br />

Other 48<br />

VARIABLES Mean<br />

Alliance Characteristics: Continuous Variables<br />

Standard<br />

Deviation Maximum Minimum<br />

AllianceTiming 1.82 0.05 1.93 1.38<br />

#Alliances 1.72 1.43 12 1<br />

Alliance Characteristics: Binary Variables<br />

MktgAlliance 0.28 0.45 1 0<br />

ParticAlliance 0.23 0.42 1 0<br />

FnctAlliance 0.39 0.49 1 0<br />

ConnAlliance 0.27 0.44 1 0<br />

InterdepLevel 0.19 0.40 1 0<br />

Partner Characteristics: Binary Variables<br />

InternetFirm 0.33 0.47 1 0<br />

TradeAssoc 0.02 0.14 1 0<br />

TraditionalFirm 0.43 0.50 1 0<br />

B2B E-Market Characteristics: Binary Variables<br />

MktLeader 0.36 0.48 1 0<br />

VerticalExch 0.69 0.46 1 0<br />

ConsortExch 0.07 0.26 1 0<br />

CensoredObs 0.66 0.48 1 0<br />

Exchange Product Characteristics: Binary Variables<br />

DigitalSvcs 0.37 0.48 1 0<br />

MROSvcs 0.19 0.39 1 0<br />

DirectGoods 0.42 0.50 1 0<br />

ConsumerGoods 0.11 0.31 1 0<br />

CapitalEquip 0.07 0.26 1 0<br />

OtherGoods 0.06 0.24 1 0<br />

Among the 332 <strong>alliance</strong>s, 64 involved equity investments or exclusive agreements. In addition, in 174<br />

cases, B2B e-<strong>market</strong>s had conventional firms as partners; in 151 cases, they formed <strong>alliance</strong>s with Internet<br />

firms; and in the remaining seven instances, they partnered with trade associations. There are 15 cases in<br />

which B2B e-<strong>market</strong> firms partnered with traditional intermediaries, such as distributors. It is important to<br />

note, as we mentioned earlier, that <strong>alliance</strong>s are created to achieve multiple purposes. So it is possible within<br />

our data set for a strategic <strong>alliance</strong> to be coded as being of more than one type. An example is the<br />

cooperation that now-defunct Pricecontainer.com, a B2B trading hub for shippers and carriers, formed with<br />

Nissho Iwai American Corporation, a Japanese trading company, on March 31, 2000. In this <strong>alliance</strong>,


Nissho Iwai American Corporation indicated that it would use Pricecontainer.com for its transaction<br />

logistics, as well as to promote the online <strong>market</strong>place to its own clients. According to our coding scheme,<br />

this partnership is both a participation <strong>alliance</strong> and a <strong>market</strong>ing one. Table 6 shows the number of strategic<br />

<strong>alliance</strong>s for each type. (See Table 6.) In accordance with our theory of “value proposition perfection” of<br />

B2B e-<strong>market</strong> intermediation services presented here and in Dai and Kauffman (2002), we note that the<br />

largest number of strategic <strong>alliance</strong>s emphasize the expansion of <strong>market</strong> service functionality, followed by<br />

<strong>market</strong>ing and connection <strong>alliance</strong>s. Finally, we summarize the descriptive statistics for the data set in Table<br />

7.<br />

EMPIRICAL ANALYSIS APPROACH AND METHODOLOGY<br />

We employ a three-step econometric analysis process to test our hypotheses on strategic <strong>alliance</strong><br />

formation, <strong>alliance</strong> structure and <strong>alliance</strong> outcomes in this research.<br />

Step 1: A Poisson Regression Model for Count Data Analysis of Alliance Formation<br />

To analyze strategic <strong>alliance</strong> formation related to Hypotheses #1, #2 and #3, we examine B2B e-<strong>market</strong>s’<br />

motivation and opportunities to enter into such cooperative relationships. Our unit of analysis is at the B2B<br />

e-<strong>market</strong> firm level. We code #Alliances as the dependent variable. In our B2B e-<strong>market</strong> context, <strong>alliance</strong><br />

announcements are events that occur discretely and infrequently, leading to a limited-dependent variable.<br />

There are numerous models that deal with limited-dependent variables (Maddala, 1993). Among them, the<br />

Poisson model is appropriate in situations where the dependent variable is a count or frequency of<br />

occurrence, and large counts are rare (Cameron and Trivedi, 1986; Winkelmann and Zimmermann, 1995).<br />

Since the total number of <strong>alliance</strong>s that a firm forms indicates the combined effects of its motivation and<br />

opportunities to employ partnering <strong>strategies</strong>, we analyze our data using a Poisson regression model<br />

(Gourieroux and Magnac, 1997; Greene, 2000; Trivedi, 1977, Winkelmann, 1997). Because they also can be<br />

safely assumed to occur independently as well, the Poisson count data regression model is an appropriate test<br />

methodology (Cameron and Trivedi, 1998). Based on this choice, we then will assume that the occurrence of<br />

discrete <strong>alliance</strong> announcement events follow a Poisson distribution:<br />

−λi<br />

i e λi<br />

Pr( Y = yi<br />

) = , (1)<br />

y !<br />

i<br />

y<br />

where yi is the number of <strong>alliance</strong>s (#Alliances) that B2B e-<strong>market</strong> firm i formed during the sample period.<br />

In the above expression, λi generally is a log-linear link function of explanatory variables with<br />

log λi = β’ Xi . In this model, Xi is the vector of explanatory variables for firm i’s <strong>alliance</strong> choices and the β<br />

’s are the parameters to be estimated in the model. In our context, we have selected explanatory variables in<br />

the vector Xi that will proxy for pioneering B2B e-<strong>market</strong>s, strategic products, and <strong>market</strong> leaders. The year<br />

that a B2B e-<strong>market</strong> was founded (coded as FoundYr) indicates if the firm is an early-to-enter B2B e-<strong>market</strong>.<br />

Our reasoning behind this is that the Internet was already becoming commercialized as long ago as 1995.


The vanguard DotCom firms, including Amazon.com, eBay, and Chemdex, emerged around then, and have<br />

been widely perceived as the archetypal pioneers in e-commerce. As a result, B2B e-<strong>market</strong>s that were<br />

founded in the year 1995 or earlier are considered to be pioneers.<br />

The second factor that we hypothesize to affect formation of strategic <strong>alliance</strong>s is the product types that<br />

B2B e-<strong>market</strong>s serve. Among the product types that we identified in Table 2, MROSvcs (maintenance, repair<br />

and operation services and products) and CapitalEquip (capital equipment, usually from excessive inventory)<br />

are non-strategic products to buyers. However, we believe that DirectGoods (raw materials, parts and<br />

components used in production processes), ConsumerGoods (products for reselling to consumers), and<br />

DigitalSvcs (business services, and information products) are strategic products. Why? Because these<br />

products directly affect the product and service quality of the buyers.<br />

The third factor that we examine for <strong>alliance</strong> formation is the <strong>market</strong> position of the B2B e-<strong>market</strong>,<br />

which is indicated by the variable MktLeader. In addition, we also include the variables VertExch and<br />

OtherGoods as control variables. This yields the following equation for explanatory variables in the Poisson<br />

regression model:<br />

log λ = β + β ⋅ MktLeader + β ⋅VerticalExch<br />

+ β ⋅ ConsortExch<br />

+ β ⋅ DigitalSvcs<br />

+<br />

i<br />

β ⋅ DirectGoods<br />

+ β ⋅ ConsumerGoods<br />

+ β ⋅ MROSvcs + β ⋅ CapitalEquip<br />

+<br />

5<br />

β ⋅ OtherGoods +<br />

9<br />

0<br />

1<br />

∑<br />

j<br />

6<br />

γ FoundYr<br />

j<br />

2<br />

ij<br />

3<br />

7<br />

Finally, the FoundYrij variables are dummy variables for founding year j for firm i. (We designated the<br />

founding years, 2000 and 2001, as the base case for testing the FoundYr effects, and so the dummy variable<br />

FoundYr2000-2001 is actually not included in our model.)<br />

In accordance with our first three hypotheses, we expect to observe positive coefficients for the<br />

following explanatory variables: FoundYr1994 and FoundYr1995 in support of H1; DigitalSvc, DirectGoods,<br />

and ConsumerGoods in support of H2; and MktLeader in support of H3.<br />

Step 2: A Binomial Logit Model to Explain B2B E-Market Alliance Structures<br />

In structuring strategic <strong>alliance</strong>s, partnering firms can choose to enter into highly interdependent<br />

relationships by obtaining equity positions or specifying exclusive agreements, or they may stay with a<br />

simple formal contract. The former and the latter cases represent interorganizational governance structures<br />

in strategic <strong>alliance</strong>s that involve relatively high and low levels of interdependence, respectively. In Step 2 in<br />

our econometric analysis, we model and analyze the factors that affect management’s decisions about<br />

developing highly interdependent <strong>alliance</strong>s. Since the choice between high and low level interdependence is<br />

a choice variable for the B2B e-<strong>market</strong> firm, and it can be represented as a 0/1 binary variable, we will use a<br />

limited-dependent variable binomial logistic regression model for our test of the theory (Hosmer and<br />

Lemeshow, 2000; Maddala, 1993). The general form of a logit model is:<br />

8<br />

4<br />

(2)


exp( β ' X )<br />

Pr( Y = 1)<br />

=<br />

, (3)<br />

1 + exp( β ' X )<br />

where Pr() indicates probability, and Y is the binary choice dependent variable, and X is a vector of<br />

explanatory variables. 1<br />

In the B2B e-<strong>market</strong> context, the dependent variable of interest is InterdepLevel, the observed level of<br />

interdependence in the <strong>alliance</strong>. We will test for the statistical significance of two separate effects.<br />

Hypothesis #4 states that participation <strong>alliance</strong>s are more likely to have a high level of interdependence.<br />

Hypothesis #5 posits that <strong>alliance</strong>s are less likely to exhibit a high level of interdependence when competitors<br />

are involved. B2B e-<strong>market</strong>s are digital intermediaries, and compete against conventional intermediaries for<br />

buyers and vendors, as suggested by research in Internet-based intermediation (Chircu and Kauffman, 2001).<br />

As a result, partners who are conventional intermediaries are also competitors from the perspective of B2B e-<br />

<strong>market</strong> firms. Moreover, B2B e-<strong>market</strong>s rely on their competence in Internet and Web technologies to<br />

design and deliver their products and services, and hence their <strong>market</strong>s overlap with those partners who are<br />

also Internet firms. In this sense, a B2B e-<strong>market</strong> is partnering with a competitor if the partner is an Internet<br />

firm.<br />

Therefore, we will use the variables ParticAlliance (participation <strong>alliance</strong>s), Intermediary, and<br />

InternetFirm as explanatory variables in our empirical model. In addition, we will include variables on other<br />

<strong>alliance</strong> types, partner types and B2B e-<strong>market</strong> characteristics as control variables. With these considerations<br />

in mind, the empirical model is as follows:<br />

exp( Z(<br />

X ))<br />

Pr( InterdepLevel<br />

= 1)<br />

=<br />

,<br />

1+<br />

exp( Z(<br />

X ))<br />

with the function of explanatory variables, Z(X) structured as follows:<br />

Z(<br />

X ) = β + β ⋅ ParticAlliance<br />

+ β ⋅ MktgAlliance<br />

+ β ⋅ FnctAlliance<br />

+ β ⋅ ConnAlliance<br />

+<br />

β ⋅ OtherAlliance<br />

+ β ⋅ Intermediary<br />

+ β ⋅ InternetFirm<br />

+ β ⋅TradeAssoc<br />

+<br />

5<br />

β ⋅ MktLeader + β ⋅VerticalExch<br />

+ β ⋅ ConsortExch<br />

+ β ⋅ AllianceTiming<br />

9<br />

0<br />

1<br />

10<br />

6<br />

2<br />

Considering that a partner firm is classified as one and only one type out of the three types<br />

11<br />

7<br />

(TraditionalFirm, InternetFirm, and TradeAssoc), we use TraditionalFirm as the base case, and so it is not<br />

included in the regression. In accordance with Hypotheses #4 and #5, we expect the coefficients for the<br />

variable ParticAlliance to be positive, while those for Intermediary and InternetFirm ought to be negative.<br />

Step 3: Binomial Logit Model on Alliance Outcomes<br />

1 Since binomial logit regression is relatively well known in IS research, we do not provide a lot of details about the<br />

estimation process, the interpretation of the model, the appropriate diagnostics for the statistical significance of the<br />

results, or other issues that relate to functional form. For the interested reader who would like more information, we<br />

recommend the following sources: Agresti (2002), Greene (1999), Harrell (2001) and Hosmer and Lemeshow (2000).<br />

3<br />

12<br />

8<br />

4<br />

(4)<br />

(5)


In Step 3 we examine the outcomes of strategic <strong>alliance</strong>s in the B2B e-<strong>market</strong> firm sector. We use the<br />

operational status of B2B e-<strong>market</strong>s as of August 9, 2002 as the basis of measurement. If a B2B e-<strong>market</strong><br />

was still operating by August 9, 2002, then we consider it to be surviving in <strong>market</strong> competition up to that<br />

time. This is an instance of a very general phenomenon called censoring, which occurs when no event is<br />

observed to happen for a participating firm in the study. The corresponding variable, CensoredObs, is set to<br />

“1” for firms that were still operating as of August 9, 2002, and for those that failed in <strong>market</strong> competition,<br />

we code CensoredObs to “0.” Hypothesis #6 suggests that the number of <strong>alliance</strong>s that a B2B e-<strong>market</strong> has<br />

formed tends to increase its survivability, and so our econometric test is aimed at finding out whether this is<br />

true. Therefore, we use CensoredObs as the dependent variable, and #Alliances as the explanatory variable.<br />

Considering that previous empirical research has shown curvilinear relationship between the number of<br />

<strong>alliance</strong>s and the rate of innovation (Deeds and Hill, 1996), we also include the square of number of<br />

<strong>alliance</strong>s, #Alliances 2 , in our model.<br />

Since the dependent variable has binary values, we again use a logit model in our test. Similar to the<br />

models that we tested in Steps 1 and 2, we also include variables on founding year, product types, and other<br />

B2B e-<strong>market</strong> characteristics as control variables. The general form of the estimation model is:<br />

exp( Q(<br />

X ) + γ ⋅ FoundYr)<br />

Pr( CensoredObs<br />

= 1)<br />

=<br />

1 + exp( Q(<br />

X ) + γ ⋅ FoundYr)<br />

In this model, the function Q(X) includes the details of the test that we employ to substantiate our theory. It<br />

is given by:<br />

Q(<br />

X ) = β + β ⋅#<br />

Alliance + β ⋅ (# Alliance)<br />

+ β ⋅ MktLeader + β ⋅VerticalExch<br />

+<br />

β ⋅ Consortium + β ⋅ DigitalSvcs<br />

+ β ⋅ DirectGoods<br />

+ β ⋅ ConsumerGoods<br />

+<br />

5<br />

β ⋅ MROSvcs + β ⋅ CapitalEquip<br />

+ β ⋅ OtherGoods<br />

9<br />

0<br />

1<br />

10<br />

6<br />

2<br />

2<br />

7<br />

11<br />

In accordance with Hypothesis #6, we expect to observe a positive estimated coefficient on #Alliances, but a<br />

negative one for #Alliances 2 .<br />

ESTIMATION RESULTS AND DISCUSSION<br />

We used LIMDEP 7.0 (www.limdep.com) to estimate the above three empirical models on <strong>alliance</strong><br />

formation, structures and outcomes. Our tests of the study hypotheses proceed according to the three stages<br />

outlined in the preceding section. The first step examines the frequency with which B2B e-<strong>market</strong>s form<br />

<strong>alliance</strong>s by testing a Poisson count model with #Alliance as the dependent variable. The second step aims to<br />

explain the governance structures that B2B e-<strong>market</strong>s set up for their partnerships with other organizations,<br />

and especially to analyze the degree to which they are dependent on each other for controlling rights and the<br />

scope of their interfirm business activities. The third step investigates whether <strong>alliance</strong> <strong>strategies</strong> enhance<br />

B2B e-<strong>market</strong> survivability. The latter two tests involve logit regression models. For each model, we<br />

3<br />

8<br />

4<br />

(6)<br />

(7)


perform diagnostics on pairwise correlation, multicollinearity and other model specification issues. We also<br />

report on and interpret the estimation results.<br />

Step 1: Poisson Regression for Count Data on Alliance Formation<br />

To begin our analysis, we first checked for problems with pairwise correlations between all the<br />

explanatory variables and control variables. The correlation matrix is shown in the Appendices. See Table<br />

A1 at the end of the paper. The highest pairwise correlation is 0.337, which is well below the frequently-<br />

used threshold of 0.6 suggested by Kennedy (1998). In order to detect multicollinearity among the<br />

explanatory variables, we calculated variance inflation factors (VIFs) (Neter, Kutner, Nachstheim and<br />

Wasserman, 1996). Our calculations show that the highest VIF is 1.852—values in excess of 10 would be a<br />

cause for concern—and so we have no evidence for multicollinearity among the explanatory variables.<br />

Due to the source of our data, firms with no <strong>alliance</strong>s are not included in our sample; so the dependent<br />

variable is truncated above 0, where the lower bound of occurrences of <strong>alliance</strong>s occurs. In this case, the<br />

Poisson model stated earlier in Equation 1 is modified as follows, to handle the Y > 0 condition:<br />

−λi<br />

i e λi<br />

yi!<br />

Pr( Y = yi<br />

| Y > 0)<br />

=<br />

Pr( Y > 0)<br />

y<br />

We fit our data using this left-truncated Poisson model with the explanatory variables that are included in<br />

Equation 2. The Poisson model assumes equidispersion (Cameron and Trivedi, 1998), which means that the<br />

conditional mean given by E[yi | Xi ] = exp (β’Xi) equals the conditional variance, Var [yi | Xi]. This<br />

assumption implies that the expected value of the count yi changes only with the explanatory variables. A<br />

failure of the assumption of equidispersion has similar qualitative consequences to a failure of the<br />

assumption of homeschedasticity in the linear regression model. That is, the standard errors of the estimated<br />

model parameters will be large so that the estimation will be inefficient.<br />

To test if our dataset violated the equidispersion assumption, we conducted the regression-based test on<br />

over-dispersion as discussed by Cameron and Trivedi (1990). The idea behind the test is that the value,<br />

{y - E[y]} 2 – E[y], should have a mean value of zero under equidispersion. The test hypotheses are:<br />

H0: Var[yi] = λi<br />

H1: Var[yi] = λi + α g(λi)<br />

The two suggested formats for g(λi) are λi and λi 2 . Hence, under equidispersion, the coefficient α should be<br />

zero. The overdispersion test results are reported in Table 8. (See Table 8.) In both cases, the coefficient α<br />

is not significantly different from zero, which implies that we cannot reject the equidispersion hypothesis,<br />

and that the parameter estimates will be efficient. Next, we consider the estimation results in more detail.<br />

(8)


Table 8. Test Results for Overdispersion in the Poisson Regression Model<br />

FUNCTION α p-VALUE<br />

g (λI)=λi -0.291 0.194<br />

g (λi)=λi 2 -0.094 0.522<br />

Note: Number of observations in model: 200. The lack of significance of the<br />

estimated values of α, -0.291/-0.094, suggests that α g(λi) =0, which is<br />

consistent with accepting the null hypothesis of equidispersion, Var[yi] = λi<br />

The maximum likelihood estimation results are reported in Table 9. (See Table 9.) The estimated model<br />

has a χ 2 value of 208.93, indicating good model fit. We note that ConsortExch is not significant, however,<br />

which led us to test a second reduced model with the ConsortExch variable removed. The reduced model has<br />

a χ 2 value of 208.92, showing that there is little loss of fit, and so we adopt it for our hypothesis tests.<br />

Table 9. Poisson Regression Results for B2B E-Markets Strategic Alliance Formation (Step 1)<br />

VARIABLE COEFF STD ERROR p-VALUE<br />

Constant -0.593 0.369 0.108<br />

E-Market Characteristics<br />

MktLeader 0.288 0.179 0.107<br />

VerticalExch -0.778*** 0.197 0.000<br />

Product Characteristics<br />

DigitalSvcs 0.413* 0.221 0.061<br />

DirectGoods 0.765*** 0.218 0.001<br />

ConsumerGoods 1.295*** 0.289 0.000<br />

MROSvcs -0.372 0.243 0.125<br />

CapitalEquip -0.551* 0.315 0.080<br />

OtherGoods -0.127 0.289 0.771<br />

Founding Year<br />

FoundYr1994 0.242 0.432 0.575<br />

FoundYr1995 1.664*** 0.274 0.000<br />

FoundYr1996 0.802** 0.342 0.019<br />

FoundYr1997 -0.505 0.731 0.490<br />

FoundYr1998 1.045*** 0.303 0.001<br />

FoundYr1999 0.213 0.282 0.451<br />

Note: Left-truncated Poisson regression; χ 2 = 208.92***; 14 DF, log<br />

likelihood = -205.38; *** = significant at .01 level; ** = 0.05 level; and<br />

* = 0.1 level. Number of observations in model: 200. 2000 and 2001<br />

are aggregated together and form the base case in the model for the<br />

founding year variables.<br />

The results show that the model has a good fit with the data with a χ 2 of 208.92 (p = 0.000), indicating<br />

that at least one of the explanatory variables has significant effect. We note that MktLeader has a marginally<br />

significant positive coefficient (2.88, std. dev. = 1.79, with p < 0.107). This provides weak support for<br />

Hypothesis #3, the Leading B2B E-Market Hypothesis, which states that <strong>market</strong>-leading B2B e-<strong>market</strong>s tend<br />

to form more <strong>alliance</strong>s with other organizations. Based on the results of this, it appears to be the case that


although some B2B e-<strong>market</strong>s may be perceived to perform better in the <strong>market</strong>place, such perceptions do<br />

not seem to confer any extraordinary advantages on the firm over others in obtaining external resources.<br />

We can test Hypothesis #1, the Pioneer B2B E-Market Hypothesis, using FoundYr2000-2001 as the base<br />

case. The effect will be present if founding year predicts the frequency of <strong>alliance</strong> formation. This captures<br />

the idea that first movers in this <strong>market</strong>place may have more motivation to seek partnerships or greater<br />

capabilities to attract other firms to form strategic <strong>alliance</strong>s. Our results show that FoundYr1995, FoundYr1996,<br />

and FoundYr1998 have significant positive effects on the number of <strong>alliance</strong>s; FoundYr1994 did not have a<br />

significant effect. Therefore, Hypothesis #1 is only partially supported, to the extent that B2B e-<strong>market</strong>s<br />

founded in the early years of e-commerce (1995 and 1996) formed more <strong>alliance</strong>s than others that entered<br />

into this <strong>market</strong> later.<br />

B2B firms that were set up before 1995 often were conducting business in related offline <strong>market</strong>s and<br />

had developed certain sources when they began Internet <strong>market</strong>s. As a result, they would not need external<br />

resources as much as firms that started B2B e-<strong>market</strong>s right at the beginning of Web-based e-commerce. For<br />

example, Buyerzone.com Inc. (www.buyerzone.com) was founded in 1992 as a middleman serving small<br />

and medium-sized businesses, and by the time it launched its online <strong>market</strong>place in 1997, it had already<br />

obtained experience with buyers and suppliers. This observation may explain why firms that were set up<br />

before 1995 did not form more <strong>alliance</strong>s than other firms.<br />

The coefficients on DigitalSvcs (0.413. p = 0.061), DirectGoods (0.765, p = 0.001), and ConsumerGoods<br />

(1.295, p = 0.000) are all positive and significant, as suggested by our Hypothesis #2, the Strategic Product<br />

Hypothesis. The results indicate that B2B e-<strong>market</strong>s that operate online <strong>market</strong>places for these strategic<br />

products are more likely to employ <strong>alliance</strong>s. In contrast, we note that MROSvcs and CapitalEquip have<br />

negative effects on the number of <strong>alliance</strong>s that B2B e-<strong>market</strong>s enter into, which supports the hypothesis<br />

from the opposite side of the issue.<br />

In addition to the explanatory variables, we also note that vertical e-<strong>market</strong>s tend to have fewer <strong>alliance</strong>s,<br />

indicated by the negative coefficient of VerticalExch (-0.778, p = 0.000). Our tentative explanation is that<br />

vertical e-<strong>market</strong>s are focused on specific industries, and thus, they have restricted scope for developing<br />

cooperation. Another reason may be that vertical exchanges perform in a more predicable environment than<br />

horizontal exchanges, because they are playing in their <strong>market</strong> niches. To the extent that industry-specific<br />

exchanges accumulate their knowledge about this industry, they reduce <strong>market</strong> uncertainty and thus diminish<br />

the need for external resources.<br />

Step 2: Binomial Logit Model to Explain B2B E-Market Alliance Structures<br />

In the second step of our empirical tests, we examine the <strong>alliance</strong> structures using the model which is laid<br />

out in Equation 4 and 5. We again checked the pairwise correlations and calculated the VIFs for all the<br />

explanatory variables. The correlation matrix is shown in the Appendices. See Table A2. The tests show no<br />

evidence of pairwise correlation, or multicollinearity; the largest multicollinearity VIF is 1.76, which is well


within the acceptable range. We fitted a binomial logit model with our data set and obtained the maximum<br />

likelihood estimates. The estimated model shows a good fit with the data (χ 2 = 64.52, p = 0.000) 2 . Another<br />

means to measure the goodness-of-fit of a binomial logit model is concordant and discordant pairs analysis.<br />

This assesses the accuracy of the model in predicting the dependent variable (Agresti, 2002). A concordant<br />

pair occurs when the fitted value is consistent with the observed value; otherwise it is a discordant pair. The<br />

percentage of concordant pairs in the total observations can be used as an indicator of the model’s predictive<br />

validity. The percentage of concordant pairs for our data is 83.4%. The results and concordant pairs<br />

analysis are shown in Tables 10 and 11. (See Tables 10 and 11.)<br />

Table 10. Binomial Logit Model Results for B2B E-Market Alliance Structures (Step 2)<br />

VARIABLE COEFF STD ERROR p-VALUE ODDS RATIO<br />

Constant 17.338*** 6.162 0.005<br />

AllianceTiming<br />

Alliance Types<br />

-4.679*** 1.462 0.001 0.01<br />

MktgAlliance 0.540 0.363 0.137 1.72<br />

ParticAlliance 1.391*** 0.447 0.002 4.02<br />

FnctAlliance 0.081 0.415 0.845 1.08<br />

ConnAlliance -0.936* 0.566 0.098 0.39<br />

OtherAlliance 0.405 0.526 0.441 1.50<br />

Partner Characteristics<br />

Intermediary -1.583* 0.885 0.074 0.21<br />

InternetFirm -0.642* 0.373 0.085 0.53<br />

TradeAssoc -0.232 1.186 0.845 0.79<br />

E-Market Characteristics<br />

MktLeader 0.476 0.324 0.142 1.61<br />

VerticalExch 0.688* 0.352 0.051 1.99<br />

ConsortExch -2.298** 1.159 0.047 0.10<br />

Model: Binomial logit; χ 2 = 64.52***, 12 DF, log likelihood = -130.49; significant at 0.01<br />

levels ***, 0.05 **, 0.1 *. No founding year dummies are significant, so no<br />

FoundYr coefficient estimates are not included in the testing model. Number of<br />

observations:332.<br />

The results in Table 10 show the coefficients of the variables, and their effects on the observed level of<br />

interdependence which are captured by the odds ratios. The odds ratio is defined as an approximate<br />

measurement for the relative probability of an outcome under different levels of an explanatory variable<br />

2 Binary logit models have error terms whose distribution under the assumption that the fitted model is correct is<br />

unknown (Neter, Kutner, Nachtsheim and Wassermann, 1996) As a result, using R 2 to assess goodness-of-fit for a logit<br />

model is not appropriate. There are a couple of statistical measures that can be applied. A common method is to<br />

examine the difference between the residuals of the model under the constraints that all regression coefficients are zero<br />

and the residuals of the estimated model. This difference can be tested as a χ 2 statistic. In this paper, we use this model<br />

χ 2 as the measure for the goodness-of-fit of logistic models. An alternative is the deviance-χ 2 statistic. It indicates<br />

percent of uncertainty under the null hypothesis that the model has fit well (Hauser, 1978). We are able to reject the<br />

null hypothesis when the deviance-χ 2 is “large.” The statistic can also be used to assess goodness-of-fit for more<br />

parsimonious models.


(Hosmer and Lemshow, 2000). The effect of ParticAlliance on the interdependence level is positive and<br />

significant (1.082, p = .005, odds ratio = 4.02). This is consistent with Hypothesis #4, the Participation<br />

Alliance Hypothesis, which argues that participation <strong>alliance</strong>s are likely to show a higher level of<br />

interdependence. The positive odds ratio indicates that participation <strong>alliance</strong>s are 4.09 times as often as non-<br />

participation <strong>alliance</strong>s to involve a high level of interdependence. The other two explanatory variables,<br />

Intermediary (-1.667, p = 0.055, odds ratio = 0.21) and InternetFirm (-0.884, p = 0.047, odds ratio = 0.53),<br />

show significant negative effects on the interdependence level in <strong>alliance</strong>s. Similarly, from the odds ratios of<br />

Intermediary (0.21) and InternetFirm (0.53), we can tell that it is nearly five times (1 / 0.21) and twice (1 /<br />

0.53) as infrequently as with other firms for a B2B e-<strong>market</strong> to engage in highly interdependent relationships<br />

with an intermediary or an Internet firm, respectively. These results support Hypothesis #5, the Competitor<br />

Alliance Hypothesis.<br />

Table 11. Concordant Pairs Analysis for Dependent Variable, Interdependence Level<br />

PREDICTED<br />

0 1<br />

0 258 (77.7%) 10 (3.0%)<br />

OBSERVED<br />

1 45 (13.6%) 19 (5.7%)<br />

Note: Concordant pairs of predicted and observed values of the dependent variable occur on the<br />

northwest-southeast diagonal; discordant pairs appear on the northeast-southwest diagonal. The<br />

concordant pairs total 277, discordant pair 55, indicating good model fit for the dependent variable.<br />

In addition to the above variables, we also note the significant effects of AllianceTiming, ConnAlliance<br />

(-0.936, p = 0.098, odds ratio = 0.39), VerticalExch (0.688, p = 0.051, odds ratio = 1.99), and ConsortExch<br />

(-2.298, p = 0.047, odds ratio = 0.10). Our interpretation is that B2B e-<strong>market</strong>s are more likely to enter into<br />

closely-interdependent relationships in the earlier years, and vertical exchanges are more likely to do so than<br />

horizontal exchanges. Otherwise, consortium-sponsored B2B exchanges are less likely to form highly-<br />

interdependent relationships. B2B e-<strong>market</strong> partnerships in setting up interfirm linkages or integrating their<br />

systems and activities with other trading systems are less likely to have high levels of interdependence.<br />

Step 3: Binomial Logit Model for Alliance Outcomes<br />

In this final step, we examine the <strong>alliance</strong> outcome by testing a binomial logit model which is laid out in<br />

Equations 6 and 7. As in the first two steps, we check the correlation and multicollinearity for the<br />

explanatory variables. The correlation matrix is shown in Table A3 in the Appendices. The highest<br />

correlation value is between #Alliance and FoundYr1995, which is 0.375. Overall, there is no problematic<br />

pairwise correlation. The highest-valued multicollinearity VIF was 1.94, indicating no problems.<br />

We estimated the model using all the explanatory variables in Equation 7, and found VerticalExch to be<br />

insignificant, so we removed it and re-estimated a reduced model with no reduction in model fit: the model<br />

χ 2 had only a small change from 44.79 to 44.78. (See Table 12.)


Table 12. Binomial Logit Model Results for the Effects of B2B E-Market Alliances (Step 3)<br />

VARIABLE COEFF STD ERROR p-VALUE ODDS RATIO<br />

Constant -0.426 0.697 0.541<br />

B2B E-Market Characteristics<br />

#Alliances 0.093 0.380 0.806 1.10<br />

(#Alliances) 2 -0.019 0.046 0.671 0.98<br />

MktLeader -0.565 0.400 0.158 0.57<br />

ConsortExch 1.659* 0.872 0.057 5.25<br />

Product Characteristics<br />

DigitalSvcs 1.302*** 0.463 0.005 3.68<br />

MROSvcs -0.502 0.469 0.284 0.61<br />

DirectGoods 0.315 0.456 0.490 1.37<br />

ConsumerGoods 1.023 0.645 0.113 2.78<br />

CapitalEquip -0.925 0.689 0.179 0.40<br />

OtherGoods 2.061* 1.165 0.077 7.85<br />

Founding Years<br />

FoundYr1994 1.569** 0.750 0.037 4.80<br />

FoundYr1995 3.205** 1.366 0.019 24.66<br />

FoundYr1996 2.787** 1.142 0.015 16.23<br />

FoundYr1997 -0.363 0.775 0.640 0.70<br />

FoundYr1998 1.599** 0.678 0.018 4.95<br />

FoundYr1999 -0.064 0.434 0.882 0.94<br />

Model: Binomial logit; χ 2 = 44.78***, 16 DF, log likelihood = -105.14; *** = .01 level, ** = 0.05<br />

level, * =0.1 level. Number of observations: 200.<br />

The results in Table 12 show that the model is a good fit for the data (χ 2 = 44.78, p = 0.000). In addition,<br />

we also match the predicted and observed values of the dependent variable to check the goodness-of-fit of<br />

the model. Table 13 shows that the percentage of the concordant pairs is 74.5%.<br />

Table 13. Concordant Pairs Analysis for Dependent Variable, Censored Observation<br />

OBSERVED<br />

PREDICTED<br />

0 1<br />

0 35 (17.5%) 32 (16%)<br />

1 19 (9.5%) 114 (57%)<br />

Note: The concordant pairs total 149, discordant pairs 53, indicating good model fit.<br />

The estimation results show that the coefficients of both #Alliances and (#Alliances) 2 were not<br />

significant. One reason might be that our dataset does not include firms that have not formed any <strong>alliance</strong>s,<br />

and thus, the data will not reflect a full effect of <strong>alliance</strong> <strong>strategies</strong> on firm performance. An improvement<br />

would be to include B2B e-<strong>market</strong>s in our data set that have no <strong>alliance</strong>s (yet). Another possible reason is<br />

that there are other factors, such as the general <strong>market</strong> and relevant industry dynamics, that influence B2B e-<br />

<strong>market</strong>s survival, and should be controlled for in the model. Moreover, it is not only the number of <strong>alliance</strong>s,


ut also the management of <strong>alliance</strong> portfolios that affect how successful B2B e-<strong>market</strong>s can become in the<br />

<strong>market</strong>.<br />

Although the effect of #Alliance is not significant, our results show several other interesting effects.<br />

First, the DigitalSvcs (1.302, p = .005) has a significant positive effect, indicating that B2B e-<strong>market</strong>s<br />

operating online <strong>market</strong>places for business services or information products have an advantage over others<br />

because of the higher product transactability relative to physical goods. Second, the significant positive<br />

effects of three founding year variables, FoundYr1994, (1.569, p = 0.037), FoundYr1995 (3.205, p =0.019), and<br />

FoundYr1996 (2.787, p = 0.015), tell that B2B e-<strong>market</strong>s enjoy first-mover advantages which may be<br />

reinforced by the positive network effects in the growth of electronic trading networks. Third, industry<br />

consortium-sponsored B2B e-<strong>market</strong>s have a better chance to survive in the <strong>market</strong> as indicated by the<br />

positive effect of ConsortExch (1.659, p = 0.057). This effect may be explained by the observation that<br />

consortium-supported exchanges tend to have more financial resources to sustain their operation.<br />

CONCLUSION<br />

Although <strong>alliance</strong>s have been common among B2B e-<strong>market</strong>s (Lenz, Zimmerman and Heitmann, 2002;<br />

Rajgopal, Venkatachalam and Kotha, 2002), there is a need for formal knowledge about their <strong>strategies</strong> in<br />

partnering with other organizations. Our study draws upon theoretical and empirical research in strategic<br />

<strong>alliance</strong>s, transaction costs and property rights analysis, and digital intermediaries, to examine cooperative<br />

<strong>strategies</strong> in the context of B2B e-<strong>market</strong>s. We first identified four main types of B2B e-<strong>market</strong> <strong>alliance</strong>s<br />

based on the nature and tasks of the <strong>alliance</strong>s, including <strong>market</strong>ing, functionality, participation and<br />

connection <strong>alliance</strong>s. This classification helps us look into B2B e-<strong>market</strong> cooperation in a structured manner.<br />

We conduct empirical tests and point out several patterns in B2B e-<strong>market</strong> <strong>alliance</strong> <strong>strategies</strong>, including<br />

<strong>alliance</strong> formation and structuring.<br />

This study reveals that not every B2B e-<strong>market</strong> uses cooperative <strong>strategies</strong> in the same manner because<br />

of differences in both motivation and opportunity. More specifically, B2B e-<strong>market</strong>s that entered into this<br />

<strong>market</strong>place at the early stage of e-commerce are more likely to have allied with other organizations due to a<br />

need for <strong>market</strong> acceptance. For the same reason, B2B e-<strong>market</strong>s offering strategic products to buyers are<br />

also inclined to form more <strong>alliance</strong>s. In addition, leading B2B e-<strong>market</strong>s are more likely to develop<br />

partnerships since they are perceived to have more resources and thus have more partnering opportunities.<br />

B2B e-<strong>market</strong>s also structure their <strong>alliance</strong> contracts differently according to the level of interdependence<br />

between partners. B2B e-<strong>market</strong>s tend to engage in highly interdependent relationships when their partners<br />

are firms that join their online <strong>market</strong>places for transactions. This permits them to obtain optimal level of<br />

investments in their trading networks from these partners. In contrast, they are more likely to have less<br />

interdependent relationships with Internet firms and conventional intermediaries so that they can protect core<br />

competence from spilling over potential competitors. Moreover, we tested the effects of <strong>alliance</strong> <strong>strategies</strong>


on firm survivability. Although the nature of the effect is consistent with our hypothesis, it is not significant.<br />

This means that just forming more <strong>alliance</strong>s is not enough to improve B2B e-<strong>market</strong>s’ ability to survive<br />

competition.<br />

Further study is needed to investigate the effect of <strong>alliance</strong> <strong>strategies</strong> on firm performance. One topic for<br />

future research is whether firm survivability is affected by the presence or absence of <strong>alliance</strong>s. Another<br />

related issue is how the <strong>alliance</strong> portfolio of a B2B e-<strong>market</strong> affects its performance. Analyses about these<br />

issues will enhance our <strong>understanding</strong> about B2B e-<strong>market</strong> cooperative activities and offer more insights to<br />

support success <strong>strategies</strong> in the arena of B2B e-commerce.<br />

In our study, we had difficulty in obtaining data on the financial performance of B2B e-<strong>market</strong>s because<br />

most B2B e-<strong>market</strong>s are private firms and data about firm characteristics and performances, such as annual<br />

revenues or sales, are not available from public sources. As a result, such data are not included in our<br />

empirical tests, and we use the founding year of B2B e-<strong>market</strong>s as the proxy for these omitted firm<br />

characteristics. However, there still exist concerns about effects of omitted variables. Additional data on<br />

such firm characteristics as size and financial status would enhance our study. Moreover, our analysis is<br />

based on firm-level data that is aggregated over the whole period of time of the study. We could create more<br />

insights into <strong>alliance</strong> <strong>strategies</strong> if we were able to disaggregate the data over time and study the path of the<br />

changes in <strong>alliance</strong>s over time.<br />

REFERENCES<br />

Agresti, A. Categorical Data Analysis, Second Edition, New York, NY: John Wiley and Sons, 2002.<br />

Ahuja, G. “The Duality of Collaboration: Inducements and Opportunities in the Formation of Interfirm<br />

Linkages,” Strategic Management Journal (21:3), March 2000, 317-343.<br />

Allen, J. and Phillips, G. M. “Corporate Equity Ownership, Strategic Alliances, and Product Market<br />

Relationships,” Journal of Finance (55), 2000, 2791-2815.<br />

Bakos, J.Y. "Reducing Buyer Search Costs: Implications for Electronic Marketplaces," Management Science<br />

(43:12), December 1997, 1676-1692.<br />

Bailey, J., and Bakos, Y. “An Exploratory Study of the Emerging Role of Electronic Intermediaries,”<br />

International Journal of Electronic Commerce (1:3), Spring 1997, 7-20.<br />

Bakos, J. Y., and Nault, B. "Ownership and Investment in Electronic Networks," Information Systems<br />

Research. (8:4), 1997, 321-341.<br />

Bamberger, G. E., Carlton, D. W., and Neumann, L. R. “An Empirical Investigation of the Competitive<br />

Effects of Domestic Airline Alliances,” Working paper, Graduate School of Business, <strong>University</strong> of Chicago,<br />

Chicago, IL, 2001.<br />

Baum, J. A. C., Calabrese, T., and Silverman, B. S. “Don’t Go It Alone: Alliance Network Composition and<br />

Performance in Canadian Biotechnology,” Strategic Management Journal (21:3), March 2000, 267-294.<br />

Cameron, A. C., and Trivedi, P. K. “Econometric Models Based on Count Data: Comparisons and<br />

Applications of Some Estimators and Tests,” Journal of Applied Econometrics (1), January 1986, pp. 29-54.


Cameron, A. C., and Trivedi, P. K. "Regression Based Tests for Overdispersion in the Poisson Model",<br />

Journal of Econometrics (46), December 1990, 347-364.<br />

Cameron, A. C., and Trivedi, P. K. Regression Analysis of Count Data (Econometric Society Monograph<br />

No.30), Cambridge, UK: Cambridge <strong>University</strong> Press, 1998.<br />

Chan, S., Kensinger, J. and Keown, A. “Do Strategic Alliances Create Value?” Journal of Financial<br />

Economics (46:2), 1997, 199-221.<br />

Chen, Z. G., and Ross, T. "Strategic Alliances, Shared Facilities and Entry Deterrence," RAND Journal of<br />

Economics (31), Summer 2000, 326-344.<br />

Chircu, A. M., and Kauffman, R. J. "Intermediation in Electronic Markets: An Analytical Model,” in<br />

Proceedings of the 2001 Workshop on Digitisation of Commerce: E-Intermediation, International Institute on<br />

Infonomics / Maastricht Economic Research Institute on Innovation and Technology, Maastricht,<br />

Netherlands, November 2001.<br />

Dai, Q. and Kauffman, R. J. “To Be Or Not To B2B? An Evaluative Model for E-Procurement Channel<br />

Adoption”, The 5th INFORMS Conference on Information Systems and Technology, November 2000, San<br />

Antonio, TX, available on CD_ROM.<br />

Dai, Q., and Kauffman, R. J. “Business Models for Internet-Based B2B Electronic Markets,” International<br />

Journal of Electronic Commerce (6: 4), Summer 2002, 41-72.<br />

Deeds, D. L, and Hill, C. “Strategic Alliances and the Rate of New Product Development: An Empirical<br />

Study of Entrepreneurial Biotechnology Firms,” Journal of Business Venturing (11), January 1996, 41-55.<br />

Dutta, S. and Weiss, A. M. "The Relationship Between a Firm’s Level of Technological Innovativeness And<br />

Its Pattern of Partnership Agreements," Management Science 43(3), March 1997, 343-356.<br />

e-Chemmerce.com “DuPont Signs AssetTRADE for Used Equipment,” May 11, 2001. Available on the<br />

Internet at www.e-chemmerce.com/c/0500/p/05004p.html.<br />

eChemPeople. “ChemCross and CheMatch Form Strategic Alliance,” February 20, 2001. Available on the<br />

Internet at www.echempeople.com/News/Feb_news.htm.<br />

Ernst, D., Halevy, T., Monier, J.., and Sarrazin, H. “A Future for E-Alliances,” McKinsey Quarterly (2),<br />

2001, 92-102.<br />

Faulkner, D., and de Rond, M. (Edtors). Cooperative Strategy: Economic, Business and Organizational<br />

Issues, Oxford <strong>University</strong> Press: Cambridge, UK, 2000.<br />

Gourieroux, C., and Magnac, T. (Guest Editors). “Special Issue on Event Count Data Methods,” Journal of<br />

Econometrics (79: 2), August 1979.<br />

Greene, W. Econometric Analysis, Fourth Edition, Englewood Cliffs, NJ: Prentice Hall, 2000.<br />

Grossman S. J., and Hart, O. D. “The Cost and Benefit of Ownership: A Theory of Lateral and Vertical<br />

Integration,” Journal of Political Economy (94), 1986, 691-719.<br />

Harrell, F. E., Jr. Regression Modeling Strategies: With Applications to Linear Models, Logistic Models and<br />

Survival Analysis, New York, NY: Springer-Verlag, 2001.<br />

Hart, O. D. "Incomplete Contracts and the Theory of the Firm," Journal of Law, Economics, and<br />

Organization (4:1), Spring 1988, 119-140. Reprinted in O. E. Williamson and S. G. Winter (Editors), The<br />

Nature of the Firm, Oxford <strong>University</strong> Press: New York, NY, 1991.<br />

Hart, O. D., and Moore, J. “Property Rights and the Nature of the Firm,” Journal of Political Economy (98:<br />

6), 1990, 1119-1158.<br />

Hauser, J. “Testing the Accuracy, Usefulness and Significance of Probabilistic Choice Models: An<br />

Information Theoretic Approach,” Operations Research (26: 3), 1978, 406-21.


Hennart, J.-F. “The Transaction Cost Theory of Joint Ventures: An Empirical Study of Japanese<br />

Subsidiaries in the United States,” Management Science (37: 4), April 1991, 483-97.<br />

Hosmer, D. W., and Lemeshow, S. Applied Logistic Regression, Second Edition, New York, NY: John<br />

Wiley and Sons, 2000.<br />

Just-Style.com. “ProNetLink.com Continues International Expansion with Alliance with NetlinQ Group,”<br />

December 20, 1999. Available on the Internet at www.just-style.com/news_detail.asp?art=1785&c=1.<br />

Kauffman, R. J. and Mohtadi, H. “Information Technology in B2B E-Procurement: Open versus Proprietary<br />

Systems,” Working paper, MIS Research Center, Carlson School of Management, <strong>University</strong> of Minnesota,<br />

Minneapolis, MN, 2001.<br />

Kennedy, P. A. Guide to Econometrics, 4th edition, Cambridge, MA: MIT Press, 1998.<br />

Krzanowski, W.J. Principles of Multivariate Analysis: A Users's Perpective. Oxford, UK: Clarendon Press,<br />

1988.<br />

Lenz, M.; Zimmermann, H-D.; and Heitmann, M. “Strategic Partnerhsips and Competitiveness of Businessto-Business<br />

E-Marketplaces: Preliminary Evidence from Europe,” Electronic Markets (12:2), Spring 2002,<br />

100-111.<br />

Lerner, J., and Merges, R. P. "The Control of Strategic Alliances: An Empirical Analysis of Biotechnology<br />

Collaborations," Working Paper 6014, National Bureau of Economic Research, Cambridge, MA, 1997.<br />

Lorange, P. and Roos, J. Strategic Alliances, Blackwell Publishers: Cambridge, MA, 1992.<br />

Maddala, G. S. Limited-Dependent and Qualitative Variables in Econometrics, Cambridge, UK: Cambridge<br />

<strong>University</strong> Press, 1983.<br />

Mody, A. “Learning from Alliances,” Journal of Economic Behavior and Organization, (20) 1993, 151-170.<br />

Neter, J., Kutner, H. M., Nachtsheim, C., and Wasserman, W. Applied Linear Regression Models, 3rd<br />

Edition, New York, NY: McGraw Hill, 1996.<br />

Nicholson, S., Danzon, P. M., and McCollough, J. “Biotech-Pharmaceutical Alliances As a Signal of Asset<br />

and Firm Quality,” Working Paper #W9007, National Bureau of Economic Research, Cambridge, MA, 2002.<br />

Oxley, J. E., “Appropriability Hazards and Governance in Strategic Alliances: A Transaction Cost<br />

Approach,” Journal of Law, Economics and Organization (13), 1997, 387-409.<br />

Pisano, G., “Using Equity Participation to Support Exchange: Evidence from the Biotechnology Industry,”<br />

Journal of Law, Economics, and Organization (5), 1989, 109-126.<br />

PRWeb. “Byers Engineering Company and BandWidth.com Announce Launch of New Telecommunications<br />

Carrier-to-Carrier,” February 7, 2000. Available on the Internet at<br />

www.prweb.com/releases/2000/2/prweb11831.php.<br />

Rajgopal, S.; Venkatachalam, M.; and Kotha, S. “Managerial Actions, Stock Returns, and Earnings: The<br />

Case of Business-to-Business Internet Firms,” Journal of Accounting Research (40:2), May 2002, pp. 529-<br />

557.<br />

Rao, A. R. and Ruekert, R.W. "Brand Alliances As Signals of Product Quality," Sloan Management Review,<br />

Fall 1994, 87-97.<br />

Robinson, D. T., and Stuart, T. E. “Just How Incomplete Are Incomplete Contracts? Evidence from the<br />

Biotech Strategic Alliances,” Working paper, Graduate School of Business, <strong>University</strong> of Chicago, 2002.<br />

Available on the Internet at home.uchicago.edu/~dtrobins/working/<strong>alliance</strong>s20.pdf<br />

Segal, I. R.; and Whinston, M. D. “Exclusive Contracts and Protection of Investments,” RAND Journal of<br />

Economics (31:4), Winter 2000, 603-633.


Shapiro, C., and Varian, H. R. Information Rules: A Strategic Guide to the Networked Economy. Harvard<br />

Business School Press: Boston, MA, 1999.<br />

Teece, D. J. "Competition, Cooperation, and Innovation," Journal of Economic Behavior and Organization<br />

(18) 1992, 1-25.<br />

Tirole, J. The Theory of Industrial Organization, MIT Press: Cambridge, MA, 1997.<br />

Trivedi, P. K. (Guest Editor). “Special Issue: Econometric Models of Event Counts,” Journal of Applied<br />

Econometrics (12: 2), May-June 1997.<br />

Williamson, O.E. “Transaction Cost Economics,” Chapter 3 in R. Schmalensee and R. Willig<br />

(Editors) The Handbook of Industrial Organization, North-Holland: Amsterdam, Netherlands, 1989.<br />

Williamson, O.E. “The New Institutional Economics: Tacking Stock, Looking Ahead,” Journal of<br />

Economic Literature (38), 2000, 595-613.<br />

Williamson, O.E., and Masten, S.E. (Editors). Transaction Cost Economics. Theory and Concepts,<br />

Edward Elgar Publishing, Aldershot, UK, 1995.<br />

Winkelmann, R. Econometric Analysis of Count Data, Second Edition, Heidelberg, Germany:<br />

Springer-Verlag, 1997.<br />

Winkelmann, R. and Zimmermann, K. F. "Recent Developments in Count Data Modeling: Theory and<br />

Application", Journal of Economic Surveys (9), 1995, 1-24.<br />

Zhu, K. “Information Transparency in Electronic Marketplaces: Why Data Transparency May Hinder the<br />

Adoption of B2B Exchanges,” Electronic Markets (12:2), Spring 2002, 92-99.


Appendix.<br />

Table A1. Correlation Matrix for Poisson Regression Model on Alliance Formation<br />

Mkt Vertical Consort Digital Direct Consumer MRO Capital Other Found Found Found Found Found Found<br />

Leader Exch Exch Svcs Goods Goods Svcs Equip Goods Yr1994 Yr1995 Yr1996 Yr1997 Yr1998 Yr1999<br />

MktLeader 1.000<br />

VertExch -.032 1.000<br />

ConsortExch .165 .175 1.000<br />

DigitalSvcs -.157 .033 -.048 1.000<br />

DirectGds .152 .231 .203 -.337 1.000<br />

ConsumGds .019 .183 .034 -.262 -.225 1.000<br />

MROSvcs -.120 -.140 -.133 .078 .053 -.124 1.000<br />

CapEquip .165 -.257 -.075 -.170 .044 -.030 -.133 1.000<br />

OtherGds .022 -.308 -.063 -.176 -.195 -.079 .006 -.063 1.000<br />

FYr1994 -.039 -.003 -.084 -.085 -.114 .071 .127 .057 -.070 1.000<br />

FYr1995 .066 .015 -.004 .018 .065 -.098 -.041 .071 .022 -.087 1.000<br />

FYr1996 .165 -.170 -.075 -.088 -.075 .034 .117 -.075 .117 -.084 -.078 1.000<br />

FYr1997 .004 -.039 .020 .133 -.161 -.083 -.005 -.066 .045 -.074 -.069 -.066 1.000<br />

FYr1998 .176 -.103 -.041 .004 -.034 -.076 .056 .080 .056 -.113 -.105 -.101 -.089 1.000<br />

FYr1999 -.150 .076 -.160 .024 -.030 .091 -.142 -.037 -.072 -.224 -.209 -.201 -.177 -.271 1.000<br />

Note: Pairwise correlations are based on observations for 200 firms in the data sample. The highest observed pairwise correlation is -.337 between DigitalSvcs and<br />

DigitalGoods.<br />

Table A2. Correlation Matrix for Binomial Logit Model on Alliance Structure<br />

Alliance Mktg<br />

Timing Alliance<br />

AllianceTiming 1.000<br />

MktgAlliance -0.087 1.000<br />

Partic<br />

Alliance<br />

Fnct<br />

Alliance<br />

Conn<br />

Alliance<br />

Other<br />

Alliance<br />

Internet<br />

Firm<br />

Intermediary<br />

Trade<br />

Assoc<br />

Mkt<br />

Leader<br />

ParticAlliance -0.054 -0.097 1.000<br />

FnctAlliance -0.028 -0.089 -0.299 1.000<br />

ConnAlliance 0.084 -0.056 -0.297 -0.158 1.000<br />

OtherAlliance -0.017 -0.116 -0.201 -0.233 -0.187 1.000<br />

InternetFirm 0.010 0.037 -0.261 0.057 0.425 -0.071 1.000<br />

Intermediary -0.145 -0.037 0.192 -0.053 -0.066 -0.005 -0.196 1.000<br />

TradeAssoc 0.070 0.003 -0.030 -0.073 -0.089 0.001 -0.132 -0.032 1.000<br />

MktLeader -0.093 -0.043 0.049 -0.097 0.057 -0.070 -0.014 -0.002 0.007 1.000<br />

VerticalExch 0.051 -0.095 0.126 0.029 -0.012 -0.000 -0.115 0.073 -0.064 -0.020<br />

ConsortExch 0.113 -0.095 0.042 -0.006 0.015 -0.013 -0.018 0.107 0.040 0.150<br />

Note: number of observations = 332<br />

Vertical<br />

Exch<br />

Consort<br />

Exch<br />

1.000<br />

0.210 1.000<br />

34


Table A3. Correlation Matrix for Binomial Logit Model on Alliance Outcome<br />

#Alliance<br />

Mkt<br />

Leader<br />

Vertical<br />

Exch<br />

Consort<br />

Exch<br />

Digital<br />

Svcs<br />

Direct<br />

Goods<br />

Consumer<br />

Goods<br />

#Alliance 1.000<br />

MktLeader 0.163 1.000<br />

VerticalExch -0.014 -0.032 1.000<br />

ConsortExch -0.000 0.165 0.175 1.000<br />

DigitalSvcs -0.014 -0.157 0.033 -0.048 1.000<br />

DirectGoods 0.099 0.152 0.231 0.203 -0.337 1.000<br />

ConsumerGoods 0.068 0.019 0.183 0.034 -0.262 -0.225 1.000<br />

MROSvcs -0.064 -0.120 -0.140 -0.133 0.078 0.053 -0.124 1.000<br />

CaptilEquip 0.055 0.165 -0.257 -0.075 -0.170 0.044 -0.030 -0.133 1.000<br />

OtherGoods -0.002 0.022 -0.308 -0.063 -0.176 -0.195 -0.079 0.006 0.063 1.000<br />

FoundYr1994 -0.065 -0.039 -0.003 -0.084 -0.085 -0.114 0.071 0.127 0.057 -0.070 1.000<br />

FoundYr1995 0.375 0.066 0.015 -0.004 0.018 0.065 -0.098 -0.041 0.071 0.022 -0.087 1.000<br />

FoundYr1996 0.082 0.165 -0.170 -0.075 -0.088 -0.075 0.034 0.117 -0.075 0.117 -0.084 -0.078 1.000<br />

FoundYr1997 -0.090 0.004 -0.039 0.020 0.133 -0.161 -0.083 -0.005 -0.066 0.045 -0.074 -0.069 -0.066 1.000<br />

FoundYr1998 0.117 0.176 -0.103 -0.041 0.004 -0.034 -0.076 0.056 0.080 0.056 -0.113 -0.105 -0.101 -0.089 1.000<br />

FoundYr1999 -0.132 -0.150 0.076 -0.160 0.024 -0.030 0.091 -0.142 -0.037 -0.072 -0.224 -0.209 -0.201 -0.177 -0.271 1.000<br />

Note: number of observations = 200.<br />

MRO<br />

Svcs<br />

Capital<br />

Equip<br />

Other<br />

Goods<br />

Found<br />

Yr1994<br />

Found<br />

Yr1995<br />

Found<br />

Yr1996<br />

Found<br />

Yr1997<br />

Found<br />

Yr1998<br />

35<br />

Found<br />

Yr1999

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!