Computer Mediated Markets: An Introduction and Preliminary Test of Market Structure Impacts
Address: Michigan State University, Department of Telecommunication, 409 Communication Arts and Sciences Building, East Lansing, MI 48824-1212, USA
Address: 309 Kennedy Hall, Cornell University, Ithaca, NY 14853, USA.
Address: Human Computer Interaction Institute, Carnegie Mellon University, 5000 Forbes Ave., Pittsburgh, PA 15213.
Electronic commerce may influence the way in which goods are traded between businesses. Many believe that Internet-based business-to-business e-commerce will reduce the extent to which firms buying goods and services are “locked in” to a single supplier. Using a secondary analysis of data collected in late 1996 on firms' use of electronic networks for transactions, we empirically test the effects of Internet use on buyer lock-in. Results are weak, but suggest that using the Internet rather than proprietary computer networks in connecting with external trading partners appears to lessen a buying firm's dependence on its primary supplier. The Internet seems to be especially valuable in allowing small firms to connect to external constituents.
Many researchers have written about the potential of Internet-based electronic commerce to alter the structure of markets in virtually all industries (e.g. Bailey and Bakos, 1997; Bakos, 1997; 1998; Brynjolfsson and Smith, 1999; Sarkar, Butler and Steinfield, 1995; Smith, Bailey and Brynjolfsson, 1999; Steinfield, Kraut and Plummer, 1995, Wigand and Benjamin, 1995). They have concentrated on two areas: the relationship between buyers and sellers, and the role of intermediaries in the production and distribution chain. These discussions have grown out of earlier debates about the role of information technology and electronic communication networks on inter-firm relations and the cost of transactions between firms (e.g. Bakos, 1991; Bakos and Treacy, 1986; Malone, Yates and Benjamin, 1987; Porter and Millar, 1985; Steinfield and Caby, 1993; Steinfield, Caby and Vialle, 1993; Streeter, Kraut, Lucas and Caby, 1996). The argument is that using inter-organizational computer networks results in lower transaction costs, which makes it easier for firms to find new suppliers and to access new markets. According to this reasoning, the Internet is continuing a trend in which interorganizational networks are stimulating the rise of new electronic marketplaces. These new marketplaces are characterized by strong price competition and greater choice for buyers (Bakos, 1997; 1998; Smith, Bailey and Brynjolfsson, 1999; The Economist, February 26, 2000; The New York Times, January 18, 2000). At the same time, the Internet enables the producers of goods and services to develop more direct relationships with their buyers, bypassing most former intermediaries (Wigand and Benjamin, 1995, Choi, Stahl and Whinston, 1997).
Many of the articles in this special issue, however, illustrate that the effects of the Internet on market structures are complex and varied, and not yet well understood. Most of the authors emphasize the increasing, rather than decreasing, role for traditional and new intermediaries on the Internet, providing the downstream services that firms require to make their goods and services more available to consumers (Adelaar, this issue; Klein and Selz, this issue; Palmer, Bailey and Smith, this issue; Schmitz, this issue; Scott, this issue). It now seems safe to conclude that electronic commerce results in significant new opportunities for providers of intermediary services, and that producers of goods who prematurely seek to bypass existing and new digital intermediaries are engaging in risky business (The Economist, June 26, 1999; The Economist, February 26, 2000, The OECD, 1999). For this reason, we do not dwell on the topic of intermediaries in electronic markets in this article.
Less clear, however, is how the Internet affects upstream business-to-business relationships, primarily in the context of producer-supplier linkages. Hence, in this article, our primary question is whether electronic commerce over the Internet is associated with more or less “lock-in” in relationships between buyer and seller firms. We present the rationale and briefly review early evidence about how interorganizational computer networks influence the extent to which buyers are locked into relationships with suppliers. We then provide an empirical test using a secondary analysis of data on electronic exchanges in four industries, concentrating on how the Internet differs from other types of interorganizational networks in commerce. Early research suggested that electronic commerce encouraged the growth of more tightly coupled electronic linkages between business buyers and sellers (Bakos and Brynjollfson, 1997; Kraut, Steinfield, Chan, Butler and Hoag, 1998), and promotes what some are calling “sticky” transactions that keep buyers locked-in (Shapiro and Varian, 1998; Smith et al., 1999). However, because the Internet provides a standards-based technology for electronic commerce, it may reduce the need for firms to invest narrowly in doing business with a specific trading partner. Hence, the Internet may “unstick” these closed trading relationships. In addition, given the Internet's lower costs relative to earlier proprietary networks, these effects may extend to smaller firms who previously could not afford to engage in network-based transactions unless supplied by a controlling trading partner (Steinfield et al., 1995). Consistent with this reasoning, we've seen a number of new business-to-business exchanges in the past several years, in several different industries, such as steel (Metalsite.com, E-Steel.com) and chemical production (Chemdex.com and E-Chemicals.com) (The New York Times, January 18, 2000;). New Web-based exchanges, referred to as business-to-business portals, are in development in the automotive sector, aviation and many other industries as well (The New York Times, March 16, 2000). A crucial feature of these exchanges is that they are open to many buyers and sellers in an industry.
Transaction Costs, Electronic Markets, and The Internet: A Brief Review
In the latter half of the 1980s, researchers interested in the implications of interorganizational computer networks focused their attention on how such networks reduced the costs of coordination among firms. An influential and early articulation of the emerging theory in this area came from Malone, Yates and Benjamin (1987). Using notions from transaction cost economics (Williamson, 1975), Malone et al (1987) emphasized how electronic linkages between firms could have quite different effects on firms' make versus buy decisions. According to the transaction cost literature, under certain market conditions firms become vulnerable to opportunistic behavior, and as a result choose to make rather than buy needed assets. That is, they choose hierarchical governance rather than market mechanisms to coordinate production. They are more likely to make rather than buy when assets are highly specific, so that there are few suppliers, or when uncertainties and information asymmetries make it difficult for buying firms to develop adequate contracts (Williamson, 1975). Data networks that extend beyond the firm could help, Malone and colleagues argued, in two distinct, but nearly opposite ways. They contended that, in some cases, data networks would be used to integrate tightly the production systems of firms together, binding firms into what they termed electronic hierarchies. Such interfirm relationships permit efficient outsourcing of important activities, without sacrificing control. In other cases, as industries develop more standardized ways of describing products and supporting electronic transactions, networks would have quite a different effect. They could support the creation of electronic markets that perform brokerage functions, thereby efficiently linking buyers and sellers together by reducing their search costs. Examples of the day included the rapid emergence of airline reservation systems and real estate multiple listing services.
In those years, most interorganizational networks were still based upon proprietary systems running over expensive leased telecommunications circuits (Steinfield et al, 1995). As a result, highly specific investments were required to join a network. The hardware and software needed to connect to one supplier or customer would not necessarily work with another. Such technology costs created barriers to exit, and raised users' costs of switching to new suppliers once a network was established. Indeed, information systems theorists pointed out the strategic nature of these systems, particularly for supplier-provided networks that could be used to lock-in buyers (Bakos and Treacy, 1986). Malone and colleagues (1987) felt that pressures towards standardization would lower networking costs, ultimately forcing single supplier networks to open up to other suppliers. Market-based transactions would then dominate over more hierarchical network-based transactions, since the marketplace would encourage more efficient suppliers. Economic analyses further suggested that by reducing buyer search costs, emerging interorganizational networks would empower buyers to find lower cost suppliers, reducing supplier power and ultimately lowering prices (Bakos, 1991).
Most of the early research on market structure effects tended to focus on a few widely cited cases, rather than on broad-based empirical data (Steinfield et al., 1995). The few empirical studies tended to find that interorganizational data networks promoted more tightly coupled relations among firms (see Kraut et al., 1998 for a review). Some found a “move to the middle” occurring, with firms opting for neither market nor hierarchy, but something in between (Clemons, Reddi and Row, 1993). They were reducing their number of suppliers despite the lower costs of coordinating with the market. Bakos and Brynjolfsson (1997) explained this by emphasizing the need to account for non-contractible investments, such as improving quality, information sharing and innovation, that firms must make when conducting business electronically. Kraut et al. (1998) suggest that pre-existing social relations between buyers and sellers may lead firms to develop first the capability for electronic transactions with trusted and established suppliers. Indeed, their data suggest that electronic networks are more likely to be associated with increased producer-supplier integration, and reduced outsourcing.
A critical difference in today's e-commerce environment is the transition from proprietary to open networks (Steinfield et al., 1995). In an open network, costs to join a network are reduced, as are buyer search costs. At a minimum, the lower costs to join an electronic commerce network should enable smaller firms with less resources to benefit. Some evidence from an earlier open data network - the Minitel in France - suggests that lower costs do enable smaller firms to gain from electronic transactions with their suppliers and customers (Streeter et al., 1996). However, Streeter et al. (1996) also found that in the general population of firms in both the U.S and France, there were size effects on network use, regardless of the availability of a low cost public data network. Despite the low cost of hardware and software, there is ample evidence that resources are related to firms' abilities to innovate with technology (Tornatsky and Klein, 1982).
In addition to potentially enabling smaller firms to use network-based transactions, the openness of the Internet implies that alternative suppliers for needed goods and services can be more easily found. Hence, in an environment like the Internet, supplier-led markets are easily circumvented by third-party market makers or buyer-led consortia (Bakos, 1997, 1998). Indeed, as noted above, in many industries, new intermediaries are rapidly organizing such business-to-business market exchanges (New York Times, January 18, 2000). These new exchanges are characterized by strong network externalities which create incentives to include rather than exclude more suppliers and buyers. This does not necessarily spell the end of durable buyer-seller relationships. Streeter et al. (1996) examined buyer-seller relations on Minitel, an earlier open data network used in France. They found that such relationships were longer lasting for Minitel users, and surmised that lock-in might still occur due to the competitive advantages derived from having access to buyer purchase histories. This permitted more targeted offers and added value for buyers, while creating a barrier to switching to a new supplier. Today, the same arguments for the “stickiness” of transactions are now appearing with regard to e-commerce on the web (Shapiro and Varian, 1998; Smith et al, 1999; Brynjolffson and Smith, 1999). Hence, it is important to look anew at the effects of e-commerce over the Internet on market structure. Is it tightening the bonds between existing buyers and sellers, creating opportunities for more lock-in effects, or strengthening the hand of buyers, and encouraging firms to seek out the suppliers who best match required prices and features with each new purchase?
A Preliminary Empirical Test
To shed some further light on this question, we undertook a secondary analysis of data we collected in the fall of 1996 and reported in Kraut et al. (1998). The original study focused on the effect of electronic networks on outsourcing. A national sample of 250 firms in four industries provided information about purchases of selected inputs from suppliers, their relationship with their primary supplier, and their use of data networks and other coordination mechanisms to support transactions (see Kraut et al. 1998 for a full description of the sample and research methods). We first examine the question the of whether the lower cost access to the Internet has indeed “leveled the playing field,” and therefore eliminated the effect of firm resources on the use of electronic transactions between a buying firm and its suppliers. If resources do matter, we would expect to find that:
- ○H1: the greater a firm's resources, the more it uses the Internet for business purposes.
Our second and third hypotheses focus more specifically on the possible influences of Internet use on firms' relationships with their major suppliers. If the electronic markets arguments are correct, we should find that:
- ○H2: greater Internet use is associated with increased connections to various external constituents.
- ○H3: greater Internet use is associated with reduced “lock-in” with suppliers.
The original Kraut et al (1998) study used telephone interviews with managers in charge of acquiring key inputs in each of the sampled industries. The four industries, selected to represent diverse product areas (information products and services as well as manufactured goods) included apparel and pharmaceutical production, magazine publishing, and advertising. Firms were sampled from the Dun and Bradstreet database based on their 4 digit SIC code, with some additional firm demographic data also provided. Firms in the U.S. with at least 20 employees in these industries listed in the April, 1996 Dun and Bradstreet data base were ranked by number of employees, and a stratified sample of large, medium, and small firms (relative to each industry) was selected from each third of the size distribution.
Earlier case studies had helped to identify four key raw materials necessary for production in each of the industries. Managers responded to a series of questions about the specificity of the input, the relationship they had with their primary supplier, and the extent to which they used electronic networks and personal relationships to coordinate their transactions with the supplier. They also indicated the extent to which their firm used the Internet for a variety of purposes. Since the original study focused on inhouse versus outsourced production, suppliers could be either internal divisions of the interviewed firm or external. For this secondary analysis, we only focus on those cases where the suppliers are external firms, since it provides a better test of whether relations are becoming more market-like and less locked in (see Table 1 for the distribution across industries included in this study). We then analyzed what types of firms were using the Internet and for what purposes, and then explored whether Internet use was associated with various measures of external access and “lock-in.”
Sample Size and Industry
Total sales and number of employees were provided by the Dun and Bradstreet database. In the original Kraut et al study, these two measures proved to be highly correlated, and were combined into a standardized and logged scale of firm size (alpha=.91).1 The firm size measure is used here in later analyses Table 2 shows summary statistics for all variables used in the study.
Summary Statistics For Variables Used in Analyses
We included a measure of asset specificity from the Kraut et al (1998) survey in order to examine the extent to which lock-in is caused by this issue. The generic vs. specific object scale measured the extent to which the production input acquired from suppliers was specific to the respondent firm, or was more of a commodity that could be used by any firm in the industry. This scale assessed the extent to which respondents agreed on a 5-point Likert scale with four statements, (e.g., “My firm is the only one that uses the [input],” (reversed), and “The [input] we get from this supplier is fairly standard for the industry”) (alpha=.64).
We measured two types of network use, as well as the extent to which various constituents had access to the firm via computer networks. Network use with a major supplier (E-link) measures the reported importance of “any type of computer communication that allows you to exchange information” with a major supplier. Respondents indicated how important E-link was in six separate stages in the process of acquiring a key production input, including: 1) searching for and selecting a supplier; 2) developing the specifications of the key input; 3) negotiating the terms of the acquisition such as price, delivery date, and so on; 4) ordering the input; 5) monitoring the quality of the good or service; and 6) fixing problems after the order. The composite scale had a high reliability in the Kraut et al (1998) study (alpha = .94). Note that these connections could be via proprietary connections or public networks, like the Internet. To distinguish the importance of using the Internet over and above the general use of computer networks, we created an index from a series questions that asked whether or not the firm used the Internet for six different business purposes: 1) production of a key product; 2) electronic mail communication; 3) placing and receiving orders; 4) sending and receiving invoices; 5) providing product or service information to customers; and 6) electronic funds transfer (alpha=.79). For a comparison, the original survey also asked how much firms used any sort of network for the same six purposes. Figure 1 depicts this comparison, and highlights that, at least as late as the fall of 1996, the main uses of the Internet were still for email and to supply product information, even though firms were using other data-networks for more transaction-oriented purposes.
Our measure of external constituent access via computer networks was derived from a set of six items that asked “how frequently do the following groups access computers in your company or communicate with your company electronically.” The groups include: 1) employees located at the same sites as the computers; 2) employees at other sites who connect to the computers remotely; 3) customers; 4) suppliers; 5) other companies with whom the firm works; and 6) the general public. The responses range from 1 (never) to 5 (very often). These items were factor analyzed, and two factors emerged with eigenvalues greater than one. The four items representing external constituent access loaded together on the first factor (see Table 3). External constituents include suppliers, other firms, customers and the general public. Employees include local and remote employees. These groupings have high face validity, and for later analyses, we created a scale for external constituent access by taking the average score across the three high loading items (alpha=.75).
Rotated Factor Loadings for Access to Firm By Computer Network Items
Four different indicators of the extent to which firms had a “locked-in” relationship with their supplier were measured in the original Kraut et al (1998) survey. These included the percent of the needed input acquired from the supplier, the length of time doing business with the supplier, the extent of firm investments in specific hardware and software in order to do business with the supplier, and the extent of firm investments in specific knowledge in order to do business with the supplier. The first two are single item measures, while the later two combined two items that asked on 5 point scales whether doing business with the supplier required specialized equipment (alpha=.81) or knowledge (alpha=.67).
We use a series of multiple regression analyses to test the three hypotheses. In each regression, we controlled for variations in Internet use among the four industries by including a set of industry dummy variables. We also included firm size to control for the potential influence of having greater resources available for networking activities. The extent to which required inputs were specific to the buyer was also entered into regressions where the extent of lock-in was the dependent variable. Additionally, the extent to which firms report having network-based interactions with their major supplier was entered into each regression. This enabled us to see if the reported Internet use had effects over and above any prior proprietary network use, as the electronic market hypotheses predict.
H1 posits that firms with greater resources will make more use of the Internet for business purposes. Table 4 shows the regression of firm size, along with dummy variables for industry and our measure of network use with the major supplier (E-link). Contrary to much of the early expectations about the Internet, firm size was a significant predictor of Internet use. This suggests that even though the Internet is open, there are still enough costs associated with its use that firm resources matter. There were some industry differences, with greater Internet use in the information industries (publishing and advertising) than in manufacturing. Interestingly, there was no relationship between the use of electronic networks for transactions with a firm's major supplier and Internet use. This suggests that, at least in late 1996, the Internet was used in very different ways than the proprietary networks that firms relied on for transactions with their suppliers. Our earlier figure showing little use of the Internet for transactions supports this view.
Regression Predicting Internet Use
Our second hypothesis proposed that use of the Internet would be associated with greater access to the firm by a variety of external constituents, since the Internet's openness and shared costs make connections to any outside entity lower in cost and the same for all. This, of course, is the underlying mechanism that permits the firm to access a wider range of suppliers, and ultimately reduce lock-in. Regression results here offer fairly strong support for this hypothesis, but with some very interesting qualifications (Table 5). The more firms relied on network links with their major supplier, the more external constituents such as suppliers, other business partners, customers and the public, accessed the firms by computer networks. This is basically true almost by definition, but, note that as firms used the Internet more, external access was increased over and above the increases due to their specific supplier network links. This result suggests that the Internet did open firms up to greater access by their external constituents.
Regressions Predicting Access to Firm by Computer Network
However, the interaction between the Internet use and firm size is especially noteworthy (Figure 2). Smaller firms were more likely to experience greater external constituent access with increased use of the Internet, while there was actually lower external constituent access for the larger firms who went on the Internet. This suggests that among larger firms, the Internet served different purposes. At least in 1996, the Internet had not replaced the use of other types of interorganizational networking by large firms to connect electronically with external constituents. On the other hand, smaller firms appeared to be using the Internet as a surrogate for proprietary network links, and for them, the Internet increased their accessibility to external constituents.
Our third hypothesis proposed that Internet use would be associated with reduced lock-in with suppliers. We had four measures that tapped different aspects of lock-in: the proportion of the needed input purchased from a firm's primary supplier, the numbers of years the firm had been doing business with their primary supplier, the extent to which the firm had made specific investments in hardware or software to work with their primary supplier, and the extent to which working with their primary supplier required specific knowledge from the firm. Regression analyses offered weak and somewhat mixed support for this hypothesis (Table 6). Firms who had electronic connections with their primary suppliers were more likely to have made specific investments in equipment and knowledge, which suggests that such network use was associated with more lock-in. Using the Internet was unrelated to these measures of lock-in. However, the interaction effect on the proportion of the input acquired from the primary supplier reveals interesting network effects (Figure 3). When firms that relied more on any electronic links to their major suppliers also used the Internet more, they bought less of the needed input from those particular suppliers. This result suggests that increased Internet use reduced dependence on suppliers, but only for firms that had already been engaging in electronic transactions to purchase needed inputs. This result is suggestive of an important market structure effect of the Internet: It appears to reduce lock-in, but only among those firms that already consider electronic transactions with their main supplier to be important.
Regression Predicting Indices of Lock-in
Discussion and Conclusions
Our secondary analysis of the Kraut et al (1998) data has provided mixed support for the expectation derived from transaction cost economics that the Internet will influence market structure by limiting supplier power and reducing lock-in. Internet use was related to increased computer-based access to a firm from external constituents. We also found some evidence for an Internet effect on a direct measure of lock-in, i.e., the proportion of a needed input that is purchased from a firm's major supplier, but only for firms that already relied on network-based transactions with their major supplier. We found no evidence that using the Internet reduces the supplier-specific assets needed to do business with suppliers.
The analyses did reveal, however, that a number of the influences of network use were moderated in some way by the size of the firm. Larger firms used the Internet more. In contrast, smaller firms with greater use of the Internet were more likely than larger firms to experience increased access from external constituents.
One interpretation of these findings is that, like other technological innovations, the Internet requires resources to gain from its use, and therefore larger firms do use it more (Tornatsky and Klein, 1982). But in this study, larger firms did not use the Internet to give greater computer access to external constituents, perhaps because they were already using electronic connections with their major suppliers. It would appear that the primary use of the Internet was for providing product information, not for executing transactions. Presumably, larger firms had access to more secure network connections for transactions, and, in 1996 at least, were not willing to trust the Internet for these purposes. On the contrary, resource-constrained smaller firms might have turned to the Internet because it did offer an affordable means of electronically networking with their external constituents. Such Internet use does appear to have reduced lock-in, —as indicated in the proportion of inputs acquired, with key suppliers, but only for those firms that were already conducting electronic transactions with their suppliers in the first place.
As a final note, our analysis has to be considered tentative, given the state of Internet use at the time data were collected. In late 1996, although there were already clear trends towards business use of the Internet and several important business-to-business initiatives, such as CommerceNet, it may simply have been too early for notable market structure effects to appear. The tendencies we see in our data may be the early signs of influences in the same direction, or they may be simply holdover effects from pre-Internet networking approaches. Clearly, the new growth in business-to-business portals over the past year signifies important market structure impacts. New research that explicitly examines the way in which these new portals operate to structure the relationships between buyers and sellers is now needed. Are portals, for example, only used by firms to identify a relevant trading partner, with whom they then develop a longer term relationship? Or are portals turning the procurement process across many industries largely into a spot market structure? Who is benefiting from such electronic marketplaces? Does firm size still enable either buyers or sellers to extract greater benefit from such electronic market exchanges, or is the power of larger players reduced? Answers to such questions can help shape business practice and information technology policy in the coming years as the Internet matures as a venue for conducting business.
Since we are using the same data as the Kraut et al (1998) study, when we use their scales, we provide the original reliability alphas that they reported.