Economics and Electronic Commerce: Survey and Directions for Research

7/22/2019 Economics and Electronic Commerce: Survey and Directions for Research

1/112

INTERNATIONAL JOURNAL OF ELECTRONIC COMMERCE 5

International Journal of Electronic Commerce/ Summer 2001, Vol. 5, No. 4, pp. 5116.Copyright 2001 M.E. Sharpe, Inc. All rights reserved.

1086-4415/2001 $9.50 + 0.00.

Economics and Electronic Commerce:Survey and Directions for Research

Robert J. Kauffman and Eric A. Walden

ABSTRACT: This article reviews the growing body of research on electronic commercefrom the perspective of economic analysis. It begins by constructing a new framework forunderstanding electronic commerce research, then identifies the range of applicable theoryand current research in the context of the new conceptual model. It goes on to assess thestate-of-the-art of knowledge about electronic commerce phenomena in terms of the lev-els of analysis here proposed. And finally, it charts the directions along which useful work inthis area might be developed. This survey and framework are intended to induce research-ers in the field of information systems, the authors reference discipline, and other areas in

schools of business and management to recognize that research on electronic commerce isbusiness-school research, broadly defined. As such, developments in this research area inthe next several years will occur across multiple business-school disciplines, and there willbe a growing impetus for greater interdisciplinary communication and interaction.

KEY WORDS AND PHRASES: Analytical models, economics, economic theory, electronicmarkets, electronic payments, empirical models, information goods, Internet economy,market structure, software agents, technology investments.

Electronic commerce is a revolution that many industry and academic ob-servers believe will transform the conduct and structure of business as we

know it [107, 181, 210, 254, 356, 357, 380, 395].1

The range of interesting devel-opments and innovations that are occurring in information goods, firm busi-ness processes, electronic marketplaces, and emerging industries in theeconomy further increases the impetus to broaden efforts in research on elec-tronic commerce. First, however,researchers should recognize the necessityof obtaining a firm grasp on the fundamental issues that present themselvesin this new and changing technological context for business. This basic under-standing will give researchers a platform from which to develop useful mana-gerial tools specific to the e-commerce settings where the most interestingdevelopments are occurring. These include e-procurement and supply-chainmanagement, the provision of trust in Internet trade contexts, design guidanceand modes of operation for electronic auctions and electronic financial markets,and search capabilities available to users on the World Wide Web. It is the au-thors goal to contribute a framework for understanding electronic commerceand to provide up-to-date coverage of this remarkable new area of research.

Why Examine Electronic Commerce from the Perspectiveof Economics?

Many of the developments associated with electronic commerce on the Internet

seem new to most managers. But economic analysis provides a long-standingand well-developed theoretical vantage point from which to observe and in-


2/112

6 ROBERT J. KAUFFMAN AND ERIC A. WALDEN

terpret these developments in the context of continuing technological innova-tion in the business economy [40]. The strength of economic analysis comesfrom its maturity, rigor, and analytical techniques, attributes that are highlydesirable for the study of electronic commerce on the Internet.

The formal academic study of economics dates back to the eighteenth cen-tury, with Adam Smiths Wealth of Nations.Since then, the field has matured insuch a way that it is possible to study a range of issues using highly devel-oped theory. These issues include firm-to-firm competition, market structureand performance, auction market mechanics, the dynamics of product pric-ing and design, and industry regulation. Many of these issues are central toan understanding of phenomena related to electronic commerce. The explana-tory power of economics comes from its formal mathematical and analyticalmodeling methodologies, which make it possible to carefully design artificialenvironments in which the model variables affect observed economic out-

comes. The use of mathematics as the linguistic and analytical formalism ofthe key modeling relationships enables the researcher to distill the real worldto its core elements. Working with results derived from these core elements,the analyst is able to make sense of relevant parts of a highly complex world.The analyst is thus able to apply and leverage theory directly, with greaterability to explain economic phenomena. This is also highly attractive in thecontext of research on electronic commerce. There are many instances wherethe theory predicts what will emerge in the marketplace, and many otherinstances where an understanding of the complexities of the e-commercemarketplace is best pursued by understanding the basis for observed changes.

Econometric methods of data analysis are especially well suited for the

study of electronic commerce. The business processes of e-commerce produceplentiful data, and computer technology makes it easier than ever before tocollect them. To make sense of the data, and the informational structure thatunderlies the modeling circumstances that motivate data collection, high-endanalytical tools are necessary, especially ones that work well with very largedata samples. Much of the information available for collection related toe-commerce on the Internet tends to be economic data. Electronic commercebusiness processes generate data on prices, quantities, consumer willingness-to-pay, interfirm competitive dynamics and market performance, and all ofthese play to the strengths of econometrics.

Finally, economic analysis offers researchers an opportunity to sort out ex-planations for changing market structure, identify critical factors that resultin firm success and failure, interpret the speed and breadth of adoption of thenew technologies of the Internet, and much more. Thus, it is appropriate totake stock of the contributions to research on electronic commerce that aremotivated by economic analysis.

Targeting the Emerging Electronic Commerce ResearchLiterature

The literature on electronic commerce research is developing just as rapidlyas business on the Internet, as researchers evaluate new models for digitalbusiness, changing sources of firm profitability, the properties of informa-


3/112


tion goods and the marketplaces in which they transact, the emerging elec-tronic markets of the Internet [336, 340], and even new research methodolo-gies [49, 50, 75, 217]. This article reviews the emerging literature on electroniccommerce from the perspective of economic analysis. In so doing it demon-

strates that economics-motivated research on e-commerce exercises a pow-erful influence on the way people define problems, model competitivecontexts, and frame their understanding of issues related to Internet com-merce. However, to do justice to the literature, it was appropriate to span anumber of fields in which economic analysis and theory have affected tech-nology-related research. These include information systems (IS), computer sci-ence (CS), economics, finance, marketing, and several other relevant areaswhere new and interesting work on information technology (IT) and electroniccommerce is taking place.2

The intent of this review is fourfold. First, a new framework is proposed

within which research in electronic commerce can be understood and as-sessed. The framework supports an evaluative perspective for ongoing andfuture work in the field that ought to be robust to underlying changes in themarketplace. The essence of the framework, as will be illustrated below, isthe set of analysis levels that enable one to examine more and less aggre-gated economic activities, with various actors that play important roles inthe Internet economy. Second, the review identifies the range of applicabletheory from a rather large area of scholarship in economics, and examinesthe insights and predictions that have been offered to date. Taking a broadview of developments in economics makes it possible to gauge the extent ofthe interest in specific theories, and the relevance of the new ways in which

researchers are thinking about the many issues that have arisen in relationto the Internet economy. Third, the review conducts a broad-based assess-ment of the state-of-the-art of electronic commerce research at a number ofdifferent conceptual levels of analysis, based on the authors proposed mul-tilevel framework for analysis. In the process, it becomes apparent that someareas are better developed than others, and that many hold out exceptionalopportunities for new applications of economic analysis. Fourth, the reviewcharts the directions along which some of the most interesting work is likelyto develop, based on the research done to date. The research that is yet tocome will be highly interdisciplinary. If it is to solve the important problems

in e-commerce, it will benefit from a variety of collaborations that are notyet widespread in IS research.

A Framework for Electronic Commerce Research from theEconomics Perspective

The discussion that follows begins by laying out the basis for a conceptualframework that facilitates a clearer picture of the developments in electroniccommerce research that will be reviewed. The framework itself will then bepresented, together with justification for the levels of analysis it includes, basedon prior research in IS and elsewhere. The authors believe that the frameworkitself is an important contribution, since it will help readers to conceptualize


4/112


the kinds of research that have been undertaken and are likely to be empha-sized in the future.

Framework Preliminaries

New Technologies

The technological innovations associated with electronic commerce have ef-fects at several levels of analysis. Economic analysis offers considerable powerfor obtaining insights at all of these levels. For example, consider the on-linebookseller, Amazon.com. Most managers would agree that Amazon is a forcethat has transformed the market structure for bookselling. But the driver ofvalue for consumers may not be the prices and wide selection of books thatAmazon offers in the market. Instead, it may be Amazons technology-

assisted ability to quickly search a large database for specific kinds of books,and to provide recommendations on books similar to those consumers havealready purchased. Such recommendation technologies in e-commerce, aswell as a broad spectrum of other technological innovations, are based oncomputer software and systems, and on telecommunication and hardwaretechnologies. Thus, things become possible in e-commerce that one does notassociate with the bricks-and-mortar world of traditional business. How canintelligent-agent technology be used to automate complex, error-prone busi-ness activities? How do technological standards affect consumers and firms?How can firms implement pricing that optimizes the use of their networks?

New Product Characteristics

Another aspect of commerce affected by the Internet is the pricing ofproducts.

As many authors remind us, the Internet allows for the possibility of manyprices for the same good or service, as well as the opportunity for firms to postand repost prices with simple updates to their databases (in lieu of having tomanually mark and remark prices in a physical department store). In addition,the advent of the Internet provides what Priceline.com (www.priceline.com)

founder Jay Walker has referred to as a virtual death sentence for excess ca-pacity of inventory [268]. Consumers now have access to digital catalogs on-line, and can design and test customized products on their own (e.g., at LandsEnd, www.landsend.com, the catalog clothing emporium that now offers on-line dress up models; and at Dell Computer, www.dell.com, where the con-sumer can construct a personal computer system, component by component).These new capabilities transform the nature of the shopping experience, just asthey transform the process through which sales are made.

Even more interesting, there are some goods that can only be created in thepresence of technology. Digital goods or information goods, such as on-linenewspapers, MP3 music, and Internet-downloadable software products, areenabled only by technological innovations. Information goods differ from tra-ditional goods, and this prompts new questions about their design, their pric-


5/112


ing, and their support in the marketplace [334]. What are the true marginalcosts of production for information goods? What pricing mechanisms are ap-propriate to information goods in order to maximize firm profits? AreInternet products different from their brick-and-mortar alternatives?

The Transformation of Business Processes Within the Firm

Another especially interesting aspect of electronic commerce is the ability thatfirms now have, thanks to the new technologies, to process large amounts ofinformation very quickly, and to store it with nearly perfect accuracy. They canmake these capabilities available in the marketplace in the form of assistancefor consumer purchases of goods and, as well, as information services. This isone of the fundamental differences between e-commerce and traditional com-merce. High-end computing power allows for real-time pricing of a variety of

products (e.g., news and information access, excess inventoryand not juststock prices), electronic auctions (e.g., FirstAuction, www.firstauction.com;MobShop, www.mobshop.com; and Priceline.com), comparison engines (e.g.,MySimon, www.mysimon. com), one-to-one marketing based on collaborativefiltering (e.g., Net Perceptions, www.netperceptions.com), and entirely newbusiness models like Yahoo (www.yahoo.com) and eBay (www.ebay.com).

When used appropriately, the new e-commerce technologies allow firms tostreamline their business processes to achieve lower operating costs and in-crease sales revenue, as well as to improve channel coordination. The newtechnologies can also have a beneficial impact on the overall costs associated

with doing business, such as creating a presence in the marketplace, replacingthe physical infrastructure of a selling organization with a virtual infrastruc-ture, and improving a firms immediacy and responsiveness while broaden-ing its coverage in the marketplace. But how much business value is createdby the new technologies that transform business processes? What new mod-eling and theoretical insights make it possible to better understand how tomanage such processes? How will electronic commerce technologies changethe boundaries of firms?

The Firm in Market Context

The new business models for the firm and the transformed business pro-cesses associated with them influence, and, in turn, are influenced by, themarket context in which they exist [34]. Rapid technological changes cre-ated opportunities for disintermediation, as in the case of the new electronicintermediaries in the travel industry, such as Microsofts Expedia.com(www.expedia.com), Internet Travel Network/GetThere.com (www.getthere.com), and Preview Travel (www.previewtravel.com, now merged withTravelocity.com). But meanwhile, the airlines became increasingly interestedin providing producer-direct electronic booking solutions, even to the pointof cooperating with one another to attract booking traffic with an industry-centric solution [143, 367]. Both developments have led to serious challengesfor traditional travel agency intermediaries in the United States, and point


6/112


up the extent to which firms are affected by their market context.In contrast, the Internet creates an even greater need for intermediaries in

other market contexts [100]. Some technology-enabled auction markets, suchas eBay (www.ebay.com) for mass-market collectibles, would not exist with-

out the Internet. eBay transformed its marketplace by conferring liquidity onthe market for collectibles that had not been possible before the new techno-logical innovations. In a wide range of contexts, including corporate procure-ment (e.g., General Electrics Trade Processing Network Post, GE-TPN,www.geis.com; see also [190]) and the sale of computer and electronics storegoods (e.g., First Auction, www.firstauction.com, and Onsale.com,www.onsale.com), one observes interesting new ways for potential buyers tolocate and interact with vendors, up to the point of transacting with them.What are the determinants of various forms of intermediation on the Internet?How should the new electronic markets of the Internet be designed and oper-

ated, and what effect do they have on the traditional marketplace? How canfirms demonstrate their trustworthiness to consumers?

Aggregating Impacts at the Macroeconomic Level

Most observers recognize that electronic commerce has both microeconomicand macroeconomic effects. With the rapid adoption and diffusion of e-com-merce technologies and the Internet, along with the large amounts of newbusiness capital available for e-commerce firms, the macro-level effects arelikely to be substantial, even if they are difficult to track and accurately mea-

sure with todays tools and approaches. Although there is much debate [166],several studies by leading corporate economists, such as Morgan StanleysStephen Roach, indicate that the prosperity of the 1990s is directly attribut-able to information technology. The huge surge in investment has producedan effect on the economy similar to the one brought about by increased manu-facturing outputs during World War II. Further, the use of electronic com-merce technology is transforming the search characteristics of the labor market,allowing employers with job openings that require commodity-like skills thatare now standard in the marketplace (e.g., programming software, databaseadministration, and networking specialists) to be reached via the Internet.But many questions remain to be explored. In the long run, to what extent will

the capital now forming around the Internet and electronic commerce substi-tute for human labor and physical capital in the economy? How rapidly willthe Internet be adopted in other parts of the world, and what will its impactbe in other economies? What aspects of Internet commerce are most likely tobe regulated first, and what forms should regulation take?

Analytic Framework

Taken together, the questions cited above present many challenges for research-ers interested in studying issues related to electronic commerce. These chal-

lenges apply at every level, from the technology, through products andprocesses, up to firms and markets and the macroeconomy as a whole. To


7/112


organize the discussion of how to address these questions, this article em-ploys a hierarchical analytic framework that recognizes the roles of firms andconsumers as, respectively, value-makers and value-takers. The framework issimilar to the ones proposed by Kalakota and Whinston [210] and Zwass [411,

412], but the analysis levels are parsed somewhat differently, and it includesan economy-wide macro-level tier. This gives the five levels of analysis shownin Figure 1: technology, business process, product, market, and macroeconomy.Figure 1 and the accompanying interpretive text lay out the perspective.

A quasi-hierarchical model is appropriate for a number of reasons, in addi-tion to its exhaustive coverage of the business phenomena observed in theInternet economy. First, examining issues related to electronic commerce re-search from this perspective is useful because economic analysis is tradition-ally segmented based on different levels of aggregation, from the disaggregatedlevel of individuals in the market, all the way to the economy as a whole.

Thus, for example, economics recognizes that individuals behaviors andmotivations are different from the emergent behaviors of systems of individu-als. Similarly, firms formulate strategies with a keen awareness of their marketcompetition and react to potential competitive responses. With a framework ofthis kind, one can apply different types of economic analysis at different levelsof market organization. In the cases referred to here, these include utility theoryand market demand, the theory of contestable markets, and competitive-entry analysis and the dynamics of price-setting in a competitive market.

Second, the analytic levels of the framework are durable over time. Thenew technologies will change the issues that need to be examined at eachlevel, but will not affect future efforts to classify an understanding of the is-

sues in light of the relevant theory. New definitions of business process mayemerge but will still be readily identifiable as business processes. Firm bound-aries may shift (e.g., between buyers and suppliers, and within digital com-munities and virtual marketplaces), just as many new virtual firms of todaydo not have the bricks-and-mortar of traditional firms. The introduction ofnew technologies and new business processes will not change the relevanceof the level of analysis.

Note that the consumer and the firm are not included in the frameworksanalytic levels. Instead, they are treated as actors who play the role of producers(value-makers) and customers (value-takers) who establish market supply and

demand for new innovations. Unlike users of information systems within thefirm, users of e-commerce systems typically have many choices of whose sys-tem to use, since they are customers and not simply users. Moreover, firms andconsumers are actors at different levels of analysis. Consumers use the technol-ogy, purchase products, and interact with firms in the market through theirbusiness processes. Similarly, firms use technologies associated withe-commerce, then create and sell goods and services via the business processesthey design and deploy.

As electronic commerce matures, firms will increasingly exhibit economi-cally predictable behaviors and abide by well-established economic rules. Thenew economics of the digital economy, which senior executives of incumbentfirms in many markets are today struggling to master, will soon become theold economics, albeit in a marketplace transformed by technology. To be


8/112


Figure 1. A Conceptual Model for Understanding Developments andDirections in Electronic Commerce Research

The technologyassociated with electronic commerce is used by consumers and firms in a variety of ways.

Firms use it to create new kinds of products, especially information goods (e.g., MP3 music recordings,

digital newspapers), and to re-create services, especially information-intensive operational services (e.g., the

use of the Internet for package tracking by UPS and Fedex, as well as the use of Web-based reservation

booking tools by airlines). As a result, firms are actorsin the digital economy that play the role of value

makers. The value they produce can be built up by individual firmsagainst competitors, in alliances offirms,

and within industriesas a whole. By the same token, consumers, who benefit from the innovations and

competitive efforts that firms make, play the role of value takers. Whether as individuals, in homogeneous

groups,or overall in societyat large, they use technology to take advantage of the leveraging benefits

offered (e.g., for improved product search, to acquire more information, to enable infomediation and

enhanced personalization). Consumers willingness-to-pay for innovations by firms has effects at different

levels of analysis. Since firms and consumers play a role in every level of analysis discussed in this article, itis important to make a conceptual distinction between their roles as users of technology.

The discussion in the article identifies five levels of analysisthat exhibit different degrees of aggregation.

These five levels are employed in the synthesis of the literature on economics and electronic commerce. In

the framework, (1) e-commerce technologyenables the creation of (2) new products and (3) new business-

process capabilities by firm and within industries. With these capabilities, firms improve their ability to serve

consumersand more effectively compete in the marketplace. Thanks to the emergence of e-commerce, these

new products and business processes have defined a new marketplace for consumers. In these new (4)

electronic marketsof the Internet, firms and consumers interact with one another in B2C e-commerce, and

among themselves in B2B and C2C e-commerce, thereby transforming the nature of the firm and the

boundaries of market organizations. These changes have the power to reshape industries (e.g., the stock

brokerage industry, the travel industry) and the market structures of the firms within them. Moreover, shifts at

the (5) macroeconomylevel occur in the demographics of employment and firm growth and value, andthere are also changes in the issues that government regulators must track in order to maintain social welfare.

Economics provides a lens of theory with which to explore, understand, and predict these developments.

MACRO

PRODUCTS

TECHNOLOGY

MARKETS

BUSINESS PROCESSES

ENABLES

ENABLES

DEFIN

E DEFINE

USE

USE

ENABLES

AGGREGA

TE

ANALYSIS LEVELS

ACTORS

AGGREGATED

DISAGGREGATED

VALUE MAKERS VALUE TAKERS

CONSUMERS --

Society

Groups

Individuals

FIRMS --

Competitors

Alliances

Industries

MACRO

PRODUCTS

TECHNOLOGY

MARKETS

BUSINESS PROCESSES

ENABLES

ENABLES

DEFIN

E DEFINE

USE

USE

ENABLES

AGGREGA

TE

ANALYSIS LEVELS

ACTORS

AGGREGATED

DISAGGREGATED

VALUE MAKERS VALUE TAKERS

CONSUMERS --

Society

Groups

Individuals

FIRMS --

Competitors

Alliances

Industries


9/112


sure, different facets of prior theory will continue to be emphasized. Theseinclude information goods and product bundles, markets for digital lem-ons [14, 249], the microstructure of Internet-based electronic markets, andzero marginal cost productionjust to name a few. Nevertheless, firms are

organized to make profits, and those that do not are doomed to fail, as re-cently occurred with the fall of many dot-coms.

Bearing these preliminary thoughts in mind, we now turn to the emergingliterature. The discussion begins with a statement of the core business issuesof interest at each given level of analysis in electronic commerce. This is fol-lowed by a presentation of the research and an assessment of its findings andcontributions. Each section of the discussion concludes with recommenda-tions for the directions that electronic commerce research at the given level ofanalysis should take. A range of technologies and relevant economic theoriespertaining to analysis is considered (see Table 1). The recommendations arebased on a sense of the priorities developed through extensive consultationwith leading researchers in the field, participation in conferences and work-shops on electronic commerce, and a range of project research experiences.

Public goods theory, efficient

pricing, game theory, negotia-tion, network externalities and

standards

Table 1. Levels of Analysis, Technology, and Relevant EconomicTheories.

Information Relevant economic

Level of analysis technology theories

Technology

Pricing theory, versioning,

information goods, switching

costs, network externalities and

standards, economies of scale

and scope

Adoption theory, economics of

design, cost-benefit analysis,

allocation of benefits, IT value,

negotiation, economies of scaleand scope, game theory

Auction theory, industrial

organization, transaction costs,

market microstructure,

intermediation, adoption and

diffusion, perfect competition,

returns to scale, optimal market

structure

Agent technology, network

infrastructure, structuralstandards (TCP/IP, HTML, XML,

objects, middleware, encryption,

etc.)

Digital products, information,

information goods, services

Electronic data interchange,

interorganizational systems,

browsers, e-mail, intranets and

extranets

Internet-based businesses,

electronic markets, financial

services and analytics

World Wide Web, Internet-based

corporate recruiting, new labor

markets, electronic payments,fraud and security

Monetary economics, taxation

theory, labor economics,

regulatory economics, publicgoods

Product

Business process

Market

Macroeconomy


10/112


Economic Analysis of Technology-Level Issues in ElectronicCommerce

This section reviews several technology issues with respect to the Internetand electronic commerce. It treats Internet pricing, the application and effectsof intelligent software agents, search costs and technology tools that supportsearch, and finally, issues related to technology standards in e-commerce. Table2 provides an overview of some representative references and the leadingfindings from the research discussed here (see Table 2).

Internet Access Pricing

Description of Issues and Business Problems

The emerging electronic commerce technologies enable operations that wouldnot have been possible in the era before the business economy was so perva-

Internet access pricing

Representative Representative

Issue area references findings

Per-packet pricing makes sense

from the user point of view in thatit minimizes the delay of

important packets. However, a

per-packet scheme may not be a

good solution because it does not

take the providers incentives into

account.

Table 2. Technology Issues, Representative References, and Findings.

Agents provide a viable substitute

for human effort in many market-

transaction applications. The

shortcomings of agents are more

likely to be evident in human-

software interaction than insoftware-soft ware interaction.

While the impact of reduced

search costs has been examined

extensively, the search process

economics have been widely

ignored. More research is

required to understand the

economic benefits of new search

technologies.

The XML standard promises to be

an important issue in e-commerce,but there has been no research to

date on the economic rationale

for this standard.

Gupta et al. [179]; McKnight

and Bailey [281]; Mendelson[283]

Jennings and Wooldridge [204];

Maes et al. [274]; Vulkan [378];

Wurman et al. [401, 403]

Arunkundram and Bakos [15];

Gupta et al. [173]; Mendelson

and Saharia [284]; Riggins and

Rhee [319]

Glushko et al. [164]

Intelligent agents in e-commerce

Search costs and tools

Definition of e-commerce

standards


11/112


sively networked. However, broad-based access leads to a number of issuesand debates about how to operate the Internet as a technical network, a socialnetwork, and a business value-added network. The practical problems beginwith the mechanics and availability of Internet access, the network conges-

tion brought about by certain pricing schemes, and the different ways busi-nesses and individual users are served within a given pricing regime. A majordebate centers on how to price access to networks in general and to the Internetspecifically [337]. The problem here is that lack of agreement on access pric-ing may lead to crowding on the Internet and inefficient allocation of resources(i.e., bandwidth), thereby delaying important traffic. One camp suggests thataccess should be priced on a per-packet basis; others suggest flat pricingschemes. Both approaches raise important concerns about business incentivesfor the creation of value through Internet access. (For a more detailed presen-tation of this issue, see McKnight and Bailey [280].)

Recent Research

Despite its rapid growth and expansion, the Internet, like other limited-capacity networks, has always had problems associated with congestion. Likethe broadcasting spectrum for radio and television, the Internet is subject to anumber of issues pertaining to the allocation of bandwidth to private firmsand to the public. Unfortunately, the Internet is already viewed as a potentialtragedy of the commons[174], using a term originally proposed by the popu-lation ecologist Garrett Hardin [191]. When Internet users (like the shepherds

in Hardins example of overgrazing in the Spanish Pyrenees Mountains) arenot required to pay for a network resource from which they extract value,they will overuse it. Because any single user only experiences a fraction of theoverall performance degradation to the network that all the new users to-gether cause, one expects the Internet to become so crowded by overuse thatits quality is degraded for all users (as users of cable modembased Internet-access schemes have already learned).

The stream of research that examines this issue area seeks to incorporatenot only the explicit costs of network access, but also the implicit costs result-ing from congestion. MacKie-Mason and Varian provide a useful economic

overview of the interconnection network known as the Internet [269, 270]. Theyexamine its structure and cost, and the network backbone provided by the Na-tional Science Foundation. They propose and justify a per-packet pricing schemethat provides economic incentives to reduce transmissions during peak con-gestion times to free up bandwidth for users with the highest value for con-nectivity. In several articles, Gupta et al. discuss a similar pricing scheme withthe idea of distinguishing the public Internet from the many corporate intranetsthat also take advantage of the network [175179]. They show how overall socialwelfare is greater with a per-packet scheme because delay costs are minimized[180]. In a similar vein, MacKie-Mason et al. show that responsive pricing is avaluable feedback mechanism to regulate traffic on the Internet [271].

More recently, Lin et al. extended this research to virtual private networks(VPNs) [259]. They show that traffic pricing results in greater network wel-


12/112


fare than either a first-in-first-out scheme or a round-robin scheduling scheme.Konana et al. show how this priority scheme can maximize the welfare result-ing from access to real-time databases [242]. In contrast to packet-based pric-ing, Anania and Solomon argue that economic theory teaches that price should

be set to marginal cost, which is essentially zero on a per-packet basis [11].Thus, since a per-packet pricing scheme cannot work in the long term, theysuggest flat-rate pricing. Hallgren and McAdams appeal to theories of publicgoods and externalities to make the case that the Internet does not have thecharacteristics of a private good [185]. Thus, they argue, it should not be regu-lated in the same way that private goods are normally regulated. Becauseefficient allocation of public goods is inconsistent with per-packet pricing,different pricing structures ought to be used. Clark provides a useful over-view of the costs and benefits of flat and per-packet pricing, along with sev-eral hybrid pricing schemes [111].

Mendelson shows that requiring a corporate network to recover its operat-ing costs results in usage below the organization-wide optimum [283]. This isbecause network administrators typically have an incentive to underinvest inbandwidth and charge a higher price for access. Others recognize that themaximization of organization-wide welfare requires the firm to overinvest inbandwidthat least from the point of view of the network administrator.

In contrast to the problem of determining appropriate investment and pric-ing levels for networks within the firm, there is also the issue of how to handlepricing in networks with no central controller. To address this problem, Naultand Tyagi examine how two pricing mechanisms, linear transfer pricing andshared ownership, can be implemented to coordinate the allocation of net-

work resources [296]. They find that both mechanisms lead to underinvestment,but that shared ownership is less prone to do so. Similarly Nault and Dextershow how linear transfer pricing and exclusive territories can mitigateunderinvestment in a network with positive externalities [295]. In a relatedresearch effort, Masuda and Whang examine how a central controller canmaximize system value when demand functions for system services are un-known [278]. Because the price for using the system must be computed dy-namically to maximize the welfare of system users, there will be a delaybetween increased usage and increased price. This delay leads to oscillationsbetween zero usage, when the price is high, and maximum usage, when the

price is near zero. The application of a properly chosen smoothing functionfor price can mitigate oscillations between congestion and zero usage. Theresults suggest a range of up-and-coming approaches that are likely to im-prove resource allocation on the Internet.

Suggested Directions for Future Research

Despite what was said above about some of the advances in how researchersthink about pricing Internet access, much work remains to be done. For ex-ample, Internet access blurs the distinctions typically made between privateand public goods. Many observers recognize the crucial importance of imple-menting a coordination mechanism of some kind on behalf of the private sec-


13/112


tor. This should improve the quality of the infrastructure through appropri-ately formulated investment incentives, and then ratchet up the overall per-formance of the Internet in support of electronic commerce. The reality today,though, is that the Internet is essentially still a public good. In this context,

here are some suggested directions for research that are likely to contributesignificantly to our knowledge of how to handle Internet access pricing.

1. Achieve a better understanding of the importance of bandwidth in pricing Internetaccess.Although complex approaches to resource allocation often make goodeconomic sense in situations where competition is imperfect, they also canlead to undesirable or unexpected outcomes. In the case of the Internet, aspackets are required to contain more data to provide information about thepriority with which they should be handled, the result may be degraded net-work performance. Research at present ignores the cost of additional band-width, treating it as fixed, but extra bandwidth clearly offers the potential to

mitigate the effects of current bottlenecks. Indeed, the time for the addition ofbandwidth on the Internet may be sufficiently small to introduce it as a pa-rameter in future models. Crawford examines several pricing schemes basedon measurable broadcast characteristics, such as access, capacity, usage, andpriority [122]. He considers who pays for the access as an important factor.For example, if buyers pay for access, the set of economic incentives is not thesame as when sellers pay for access. He also considers pricing access for bothproviders of information and users of information, based on the market forcommunication services, which includes both the market for bandwidth andthe market for information. He concludes that viewing Internet access in terms

of a market for communication generates more realistic models than consid-ering bandwidth alone.

2. Model the fundamental elements of interconnections and connection quality rela-tive to Internet-access pricing.Internet service providers (ISPs) have traffic bothwithin their own network and between networks, and the potential impact ofthese interconnections adds a degree of complexity that makes this market-place unique. Understanding how to formulate pricing policies in such settingsis an important direction for research. Afche and Mendelsons model of pricingin this situation is a promising first step [3]. They find that passing data betweennetworks results in increased cost due to a double-marginalization problem.

However, this problem is somewhat mitigated by the ISPs internal traffic. Thus,the within-network-only users subsidize the between-network users. Kim etal. look at the competition among ISPs in a slightly different way [235]. Theyexamine the trade-off between price and connection quality in a duopoly set-ting. Their model shows that firms should respond to one anothers price in-creases with price increases, and that declining technology costs should leadto increased quality and higher ISP pricing. As the new wireless technologiesassociated with mobile commerce become more prominent, the complexity ofthe marketplace will increase, requiring telecommunications providers andISPs to further modify their access-pricing schemes.

3. Begin to reflect next-generation Internet communication schemes in models of ac-cess pricing. New technologies are emerging every year, and much work needsto be done to understand how some of the most recent arrivals on the Internet


14/112


scene affect current approaches to access pricing. For example, Chuang andSirbu identify different types of broadcasting as relevant model parameters[110]. They consider both multicast communication and unicast communica-tion. The costs of multicast communication increase with group size at a con-

stant exponential rate. They propose a per-member charge to eliminatesubsidies accruing to wide-range broadcasters. Noting that computing priceshave declined exponentially over the last 20 years, whereas transmission costshave declined only slightly, they argue that technological innovations in fiberoptics may soon change this relationship. Indeed, this is likely to change theentire economics of network allocation, for node costs are often ignored incost-based models of network resource allocation. Furthermore, it is impor-tant to understand how wireless technologies will affect network pricing. Tra-ditionally, ISPs have charged a flat rate, while wireless phone companies havecharged a per-minute fee. Reconciling these two pricing schemes into a single

wireless Internet scheme is no small problem.

Intelligent Agents in Electronic Commerce


Another issue involves the technologies associated with intelligent agents,which promise new ways for consumers to search and participate in the elec-tronic marketplace of the Internet [273]. These technologies can improve theability of consumers to search, select, negotiate, and transact for products that

fit their preferences [134]. Indeed, software-based intelligent agents offer bothfirms and consumers the possibility of artificial lifebased representation indealing with complex, crowded, fast-paced electronic markets. They also havethe potential to change market structure and market performance. However,research on these technologies is just beginning to be informed by the study ofmarket participants, the theory and structure of the new marketplace, and thedramatic changes in industrial organization. Clearly, the technologies them-selves prompt the application of economics-based thinking.

Recent Research

One of the newest areas in electronic commerce that can be examined throughthe lens of economic theory is the use of intelligent agents and shopbots inelectronic marketplaces. Shopbots are a purely e-commerce phenomenon. Theyprovide a service not easily replicated for consumers in the physical world,and thus may significantly change the rules of competition. Brynjolfsson andSmith were the first researchers to analyze the influence of shopbots on con-sumer behavior [79]. Using a dataset of almost 40,000 book searches conductedvia a shopbot, they conclude that while shopbots substantially weaken themarket positions of branded retailers, brand name and retailer loyalty stillstrongly influence consumer behavior (p. 43). Because more of the purchaseprocess is outsourced in electronic commerce than in physical distribution,consumers seem to rely on brand name as a proxy for noncontractibles, such


15/112


as delivery time, privacy, and overall quality of service. Brynjolfsson and Smithalso compared shopbot and grocery scanner data in terms of their contribu-tion to an understanding of consumer behavior. The true value propositionfor shopbots may be the data they provide about customers rather than the

data about products. It is worthwhile to consider these technologies on theirown, and to better understand their effects from the point of view of eco-nomic analysis.

Jennings and Wooldridge define an intelligent agent as a system that isresponsive to its environment; proactive, in that it exhibits goal-oriented be-havior; and social, in that it can interact with other agents to complete itsgoals [204]. To illustrate this, they provide an overview of agents used in prac-tice. Vulkan examines the same characteristics of agents in the light of eco-nomic theory [378]. He shows how economic theory can be applied to theevaluation and creation of agents. He also offers theoretical directions for re-

search informed by economic analysis, and specifically by game theory.Binmore and Vulkan provide an in-depth analysis of game-theoretic applica-tions to agents, illustrated with their Advanced Decision Environment forProcess Tasks(ADAPT) agents [68]. These are among the most powerful multi-agent applications in existence today, and are capable of handling a variety ofdecision tasks. Binmore and Vulkan also highlight the difficulties with cur-rent theories. For example, intelligent agents behave differently under timepressure and in market games of infinite length. They also discuss ways toinduce users to reveal their true preferences to agents, and show that bargain-ing outcomes are highly sensitive to the number of agents in the system.

Sandholm and Vulkan also use game theory to examine the behavior of

intelligent agents that bargain under time pressure [328]. They find that dead-lines result in a winner-take-all situation in which the agent with the latestdeadline wins. Further research by Sandholm examines the limitations ofVickrey auctions in the context of intelligent agent systems [326]. Several de-ficiencies of the Vickrey auction are known from previous literature, such asbidder collusion, dishonest auctioneers, lower revenue than other types of auc-tions, and the necessity to reveal sensitive information. The use of intelligentagents adds several new difficulties to the process. These include inefficient allo-cation in sequential auctions, untruthful bidding under local uncertainty, andthe need for counter-speculation. In another paper in this stream of research,

Sandholm and Lesser examine boundedly rational intelligent agents that nego-tiate over time [327]. The problem in ongoing negotiations is that accepting anearly offer may mean that a better offer in the future cannot be accepted. To dealwith this issue, the authors propose and justify a variable level-of-commit-ment model in which agents can break contracts by paying a penalty.

In addition to the theoretical implications of agents that negotiate in elec-tronic marketplaces, much research is under way on the practice of buildingintelligent agents. Although this work is primarily design science, the designsappearing in the literature are often very clearly informed and justified byeconomic theory. Research of this kind is taking place at several universityand industry research centers.

Maes et al. present an overview of the projects under way at MIT, togetherwith a framework for electronic commerce based on consumer buying behav-


16/112


ior [274]. The framework consists of six processes: needs identification, prod-uct brokering, merchant brokering, negotiation, payment and delivery, andservice and evaluation. Seven current intelligent agents are classified in thisframework based on the function that each performs. Moukas et al. discuss in

some depth three of the agents developed at the MIT Media Lab: Kasbah,PDA@Shop, and Tte--Tte (commercialized as Frictionless Commerce,www.frictionless.com) [287, 288]. Much of their treatment centers on experi-ence gained in implementing the Kasbah agent [92].

Kasbah is a consumer-to-consumer, multi-agent system that allows indi-viduals to create agents to search prices for them. Users can both buy and sellvia Kasbah by instantiating an agent with their personal goals, such as a highor low price and a negotiating strategy. The buying agents then choose sellingagents and make them an offer. The selling agents can either accept the offerprice and conclude the transaction or reject the offer and move to the next

round of bidding. In the next round, the agents adjust their offer price or ac-ceptance threshold according to the negotiation strategy, and continue bar-gaining. Experiments are now under way with additional features, such as atrust mechanism that allows users to rate the veracity of their counterparts inthe transaction. The interested reader should visit the MIT Media Lab Website for more information on the research programs of the Software AgentsResearch Group (agents.www.media.mit.edu/groups/agents).

The AuctionBot project at the University of Michigan is also concerned withthe design and testing of auction agents [387, 388, 401403]. Wurman et al.apply auction theory to agent behavior in sealed-bid auctions [401]. They showthat different auction structures give buyers and sellers incentives to reveal

their true prices in a single unit auction. Thus, no sealed-bid auction design issimultaneously ideal for both buyers and sellers. They also show that the single-item auction results do not generalize to multi-item auctions. Wurman andWellman address the implications of agents in multi-item auctions [404]. Theyprove the existence of an equilibrium price for bundles of heterogeneous goods.They also point out that intelligent agents are well suited to bundle auc-tions because they express their true valuation more efficiently for an entirebundle than for a single good.

More recently, Wurman et al. develop a standard way to describe auctionrules based on an object-oriented framework [405]. Auctions have three main

classes of activities: bidding rules, information revelation, and clearing policy.These can be parameterized to describe a wide variety of auctions observedin practice. This allows auction designers to have a standard, machine-read-able language for describing the microstructure of auction markets, so thatindependent agents can participate in different auctions. In addition, Fan etal. at the University of Notre Dame, the University of Southern California,and the University of Texas at Austin are conducting research on bundle trad-ing [153, 154]. They have developed a matching program that globally maxi-mizes the surplus of all orders in a market. Some related papers with differentperspectives on auctions and bundling will be reviewed in the section of thispaper that focuses on the design of electronic marketplaces.

Another research group, at the T.J. Watson Research Center of IBM Corpo-ration, is also examining the economics of intelligent agents and shopbots


17/112


[234]. Kephart and Greenwald [229] and Kephart et al. [231], for example,discuss the issues that arise when the information economy moves to widelyadopt such technology. They develop a game-theoretic analytical model thatcharacterizes situations in which shopbots provide price information in a

market. Modeling situations that involve both irrational buyers and nonlin-ear shopbot search costs, they draw conclusions about the qualitative dynam-ics of such marketplaces and the general outcomes for average prices and thenature of their distribution. This research is especially interesting because itforeshadows the emergence of a new prescriptive science for shopbot behav-ior in game-theoretic real-world market interactions (see also [169, 170]).

Tesauro and Kephart model the behavior of software agents that have theforesight to exhibit strategic behavior when they interact with competing soft-ware agents [359]. The algorithmic approaches by which they ensure that stra-tegic behavior is not marred by interactions among myopic agents provides a

mechanism to prevent the occurrence of undesirable behaviors, such as agentprice wars and long-drawn-out interaction processes. Kephart et al. offer someadditional consideration of the agent-driven price wars that occur in model-ing formulations involving two sellers in the presence of a price/quality modelor an information-filtering model [232, 233].

The issue of how learning occurs in agent economies is a central concern ofthe research program at the IBM T.J. Watson Research Center. Tesauro andKephart report on the extent to which Q-learning, a form of agent-based prob-lem-solving that draws on knowledge of reinforcement-learning procedures,enables the solution of Markov decision problems [201, 360, 381]. Such prob-lems occur in settings where new information has the potential to affect an

agents state; for example, making it willing to transact with other agents whenit previously was not willing to do so. Tesauro and Kephart provide a way toeffect approaches to computational economics that permit meaningful inter-actions among software agents. For additional details on this work, the inter-ested reader should see some of the following papers at the IBM informationeconomics research Web site (www.research.ibm.com/infoecon): [189, 230, 234,346, 358].

Similarly, researchers at the University of Minnesota have been developinga marketplace infrastructure called MAGNET in which agents trade [121].The marketplace provides support for different transactions, controls fraud

by keeping track of participants reputations, arranges for payment, and pro-vides an agent-based language for interaction. Tsvetovatyy et al. discuss anearlier version of this infrastructure called MAGMA, an early attempt to cre-ate the sort of B2B agent-based marketplace that is currently being written upin the popular press [366]. The primary insight is that economic principles ofmarket design could be implemented in a computerized system, which wouldallow for interaction of economically motivated intelligent agents. Collins etal. continue this research by examining how expected utility can be used totransform probabilistic estimated of outcomes into expectations upon whichan agent can act [120]. Using the MAGNET infrastructure, they are able toimplement agents that optimize customers desires over both risk and at-tributes. For a more in-depth look at the MAGNET market, the interested readershould see (www.cs.umn.edu/research/airvl/magnet).


18/112



As the number of intelligent agents entering the marketplace increases, re-search on the commercial applications of intelligent agents will become espe-

cially worthwhile. In the meantime, however, there are many opportunitiesfor laboratory-based artificial-life experiments of agent-based economies andmarket settings. Noted below are several directions that will be of significantinterest in the next several years.

1. Study the impact of the technical transparency of software agents in Internet mar-kets. Vulkan, who has produced some of the most engaging thinking about thekinds of research issues that await study, suggests a number of extensions forthe economic analysis of agents in electronic marketplaces [378]. He addresses,for instance, the idea of agent credibility in Internet markets, an example ofone type of demeanor that agents can take on. A software agents credibility

in making threats, like the credibility of a threat by a human negotiator ortransactor, can fundamentally change the dynamics of market games. If thereare software agents that can make credible threats regarding transactions andtrades in repeated game contexts, this will have an important bearing on howone might approach market design and the setting up of interactions amongbidding players and potential sellers. Vulkan notes that software agents,through a form of technical transparency, can make completely crediblethreats because they can reveal their software code and permit the market toestablish knowledge of their true parameters. Threat-making behavior isjust one aspect of a software agents demeanor that may be revealed by tech-nical transparency. Others are the agents approach to buying and selling in

repeated games, the nature of the agents approach to value maximization,and how the agent processes or filters new information from the market.

2. Conduct research that emphasizes the principal-and-agent nature of software agentsin support of market transaction-making.The idea of user-agent interactions onthe Internet is also important, and can be studied with perspectives from agencytheory in economics. For example, a persons intelligent purchasing agentmay wish to know how to represent the purchasing demeanor of its princi-pal in an Internet market. To do this, it must extract the persons utility forthe trade item that is being sought. A related issue is the extent to which theprincipal will trust the software agent, based on knowledge of the agents

action code. Yet another issue worthy of investigation pertains to the marketsettings in which multiple agents interact. Indeed, recent news in the businesspress suggests that the effect of agents on markets will be a major concern inthe Internet economy.

3. Develop more knowledge about how to build agents that are robust informationprocessors in markets with complex product characteristics. One solution to thisproblem is to build learning agents. A number of promising research effortsare already under way. For example, preliminary findings show the possibil-ity for market failure in an electronic market with nonadaptive agents [228].To answer this type of problem, Zacharia et al. examine how to build an agentmarketplace for services where the quality of the service provided is unknown[407, 408]. They establish a reputation for service providers based on prior


19/112


performance. Since such a market rapidly reaches a stable equilibrium, theyconclude that the reputation mechanism is a viable solution for markets inwhich quality is ex anteunknown. Similarly, Oliver examines the question ofwhether agents can learn to find viable bidding strategies [302, 303]. The re-

sults suggest that agents are able to find such strategies, but like human be-ings, they may anchor on successful strategies, and thus are vulnerable toexploitation. Overall, Oliver finds that agents perform on par with humanbeings in a variety of market situations. Goh et al. corroborate this result [165].Research on variations of these themes will be especially useful in the comingyears, as the Internet provides more and more capabilities for the support ofcomplex products

4. Continue to explore and develop new theories regarding the impact of new tech-nologies in Internet markets. Economic theory offers a panoply of insights onhow to design exchange markets, and many opportunities will soon arise toapply these ideas to the design of agent protocols in automated negotiationsettings. At present it is widely believed that agents reduce transaction costsin markets, including Internet markets. Whether this is true in Internet mar-kets, however, is an empirical question. Varian argues that agents and inter-mediaries may increase prices in certain markets if selling firms make use ofthem [375]. The argument goes like this: If a firms price changes are quicklymet by other firms, then in equilibrium there will be no incentive for a firm toreduce prices, as this will only lower profits. However, if firms can imitate oneanothers prices instantaneously, then there is an incentive to raise prices. Thiseffect is observed in the airline industry, for example, where information tech-

nology is used to keep prices high through revenue-yield management andmarket price signaling practices. What remains to be seen is whether suchpractices migrate over time to the electronic markets of the Internet (e.g., CDs,books, consumer electronics). Research that investigates some of these issuesof robotic collusion in greater depth should be worthwhile [125, 223]. Fi-nally, according to Barua et al., the Internet is also likely to be a place in whichfirms can pursue opportunities to leverage their capabilities and knowledgethrough electronic collaboration [47, 48]. Their initial work in this area is sug-gestive of the rich potential for discovery that future research holds.

Search Costs and Tools


The advent of powerful tools for Internet search offers great promise in thearea of information access for consumers. However, laying the groundworkfor the best kind of search capabilities and understanding their impact onconsumers and businesses poses many practical problems. The lower boundon the number of indexable Web pages on the Internet was about 800 millionunique pages when Lawrence and Giles published their recent study on thisissue, and the figure will be well into the billions by the time the present ar-

ticle appears in print [247]. According to Lawrence and Giles, no single searchengine indexed more than one-third of the possible pages. In spite of this less-


20/112


than-complete coverage, the Inktomi (www.inktomi.com) search engine (usedby Yahoo until recently, when it announced a switch to Google, www.google.com) has been valued at as much as 130 times its revenues.

With more than 1 billion Web pages now available via the Internet, technol-

ogy is the only answer to the problem of efficient search for users. Improvingsearch efficiency (and the market liquidity of goods for sale) will have a majorinfluence on the value that electronic commerce search technologies can de-liver [7]. Search research applies not only in the consumer arena, but also inthe area of interorganizational systems and business-to-business e-commerce.Vendors can provide additional value to their clients by enabling them to eas-ily locate appropriate information, and electronic marketplaces can providevalue by enabling firms to effectively search across a range of potential part-ner firms. However, maximizing search efficiency is both a problem of tech-nology (in terms of what it is feasible to build) and a problem of economics (in

terms of what can be done efficiently with the technology).

Recent Research

Although search is obviously an important economic issue, the specifics ofsearch costs have been largely ignored in the literature on electronic com-merce. In early work on database design, Mendelson and Saharia recognizethat there is a trade-off between database completeness and search costs [284].They show that searching all possible data is often economically inefficient,because the marginal cost of additional search will outweigh the marginal

value of the additional information yielded by the search. Ojala provides a setof rules for maximizing the value of business database searches, and her ideasare readily applicable to Internet searches [301]. Arunkundram and Bakospresent a search model that incorporates market size into search costs, so thata larger market increases search costs even as IT drives them down [15]. Fanet al. built and tested a system that uses genetic algorithms to provide searchcapabilities compatible with specific contexts [155]. They find that such adap-tive search methods lead to more effective search. Two of the most interest-ing technologies that have emerged in the last several years, collaborativefiltering and predictive search, now extend the way we think about search

on the Internet. Previously search was viewed as user-driven. But today,collaborative filtering and predictive search technologies exploit character-istics of the user that can be discovered in on-line interactions with a Website, enabling the presentation of information that is uniquely suited to theuser. Such predictive presentation of information can reduce the user s needto search, and thus change the potential cost-benefit of an individuals use ofthe Internet.

Riggins and Rhee discuss how corporate intranets can facilitate building alearning organization by linking various parts of the organization with part-ner organizations to build knowledge [319]. This raises the question of how tobuild an enterprise-wide intranet to facilitate search and retrieval of informa-

tion by agents [18]. Gupta et al. have developed a methodology that deter-mines users utility of time based on their observed actions [173]. If such


21/112


methods can be incorporated into search technologies, this might set the stagefor individually optimized, on-demand search. The literature on marketinghas already recognized the economic viability of the new search technologies[406]. All of these technologies should be evaluated in terms of their costs for

adopting organizations and how their benefits can be measured.


The literature on electronic commerce lacks a deep understanding of Internetsearch costs, search technologies, and how consumers use them. Since so manymodels are built on the proposition that e-commerce reduces search costs, suchan understanding will become necessary as e-commerce research matures.

1. Expand research efforts to more fully understand the dynamics of technology-

assisted search in Internet markets. The marketing faculties at the University ofPennsylvania and Columbia University are building an understanding of con-sumer search through the use of the Wharton Virtual Test Market(wvtm.wharton.upenn.edu), an experimental market that focuses on the studyof consumer behavior [58]. This research uses detailed demographics frommore than 10,000 people in more than 80 countries. One of its first findingsrevealed that information search is the most important predictor of on-linebuying, and thus that search is an important factor in electronic commerce.

2. Search costs affect not only buyers but sellers. Dewan et al. examine a model ofsupplier narrowcasting in which a supplier may choose to selectively adver-

tise its products [135]. They show that the impact of buyer search costs is moreimportant than the cost of narrowcasting in making the narrowcasting deci-sion. The logic is that if buyer search costs are very high, then the only way fora seller to inform buyers is by narrowcasting, for they will not search on theirown. This perspective offers an interesting contrast to the traditional view ofreduced search costs. Moreover, reduced consumer search costs may generatevalue for sellers by reducing advertising costs.

3. Further explore the impact of search-facilitation technologies in electronic commerce.Another emerging aspect of Internet search is meta-search, the process of search-ing among search engines. Different search engines allow individual users to

search differently, either because they have different algorithms or because theybetter fit the users cognitive talents. The latest generation of search-facilitationsoftware (e.g., PurpleYogi.coms Discovery System, www.purpleyogi.com) pro-vides capabilities for user-profile specification and knowledge-based search. Inaddition, users may prefer (for whatever reason) to use different search en-gines, either in general or for different or specific kinds of tasks. Further, userscan easily use multiple search engines in a single search session. Thus, thedynamics of competition among search engines exhibits some peculiaritiesthat may not be found in other marketplaces. Mukhopadhyay et al. find thatit makes economic sense for search engines to provide direct links to com-petitors and even forge partnerships with them [289]. They also find that tocompete effectively, new entrants into the search engine market do not needto be technologically superior, but only different in terms of what they in-


22/112


dex. This research is especially noteworthy in that it represents one of thefirst in-depth explorations of the economics of search engine selection and use.

Definition of Electronic Commerce Standards


The study of the establishment of technical standards for various aspects ofthe Internet (e.g., music and video file formats, and markup languages) is oneof the most interesting directions that future research can take [313]. Cusumanoand Yoffie see standards as the major factor in the outcome of the browserwars between Netscape and Microsoft [127]. In a world where standards areuncertain and competitors are closely matched, it does not take much to de-stabilize the marketplace and make prior market leaders almost irrelevant.

Recent Research

A related issue is the potential for the Extensible Markup Language (XML) toemerge as a new standard for Web page markup and data representation oncorporate Web sites and in interorganizational-systems data exchange. XMLsegments data representation and page presentation, thereby making it pos-sible for organizations to use their business-process data models to representdata embedded in the pages of their Web sites. Glushko, Tenenbaum, andMeltzer argue that an XML standard will enable electronic commerce to cre-

ate a significantly higher level of value for users and vendors than is possiblewithHypertext Markup Language (HTML), which emphasizes page format-ting [164]. This advantage occurs because standard definition tags can be ap-plied to data (e.g., Discount_Price or List_Price) so that they becomemachine-readable, setting up opportunities for automated purchasing in B2Bsupply-chain management operations. Despite the promise of the new tech-nology, however, the current marketplace is fragmented, and potential spon-sors for this technology are unwilling to commit to a general standard whosedefinition they will not be able to fully control. The situation is reminiscent ofthe early role of Sun Microsystems with respect to Java, which is still propri-etary, but which the marketplace would like to see brought forward as an

unsponsored standard.


The literature on electronic commerce standards lacks depth, thus providinga fertile ground for future research.

1. Develop a better understanding of the role of the XML standard for data sharing ine-business. Some of the major questions that will occupy e-commerce research-ers and practitioners in the coming years are how to define an XML standard,who should own the standard, and what the standards impact on organiza-tional productivity will be. This research will provide a core understanding of


23/112


standards that can be applied again and again as different standards are de-veloped and implemented. The literature on IS should be a good place to de-velop a rich economic understanding of technology standards, as IS, more sothan other fields, is concerned with the development of standards that are

economically motivated rather than legally motivated.

2.Understand how a broader set of technology standards will affect electronic com-merce. Many questions about HTML and XML remain to be answered. Forexample, how do they constrain the possible applications of e-commerce? Andhow will the data model-focused XML change the potential for documentsharing and value-bearing applications for business? Overall, researchers needto consider the effects of standards for these and other e-commerce technolo-gies on incentives for designing new technology.

In addition to XML, a variety of other standards and technologies enableand constrain the activities that comprise electronic commerce. Only visual

and audio information is deliverable via e-commerce technologies, and thiswill limit the types of transactions that are feasible. At the same time, theability to calculate and monitor in real-time enables certain transactions, likeon-line stock trading, that would not otherwise be possible. In their examina-tion of this topic, Lal and Sarvary define the digital characteristics of productsversus their physical characteristics [245]. Physical characteristics, such ascomfort, fit, and feel, cannot be delivered via e-commerce technologies.Subramani and Walden make a similar digital-tangible distinction [349, 351].However, no one has specifically examined the limitations imposed bye-commerce technologies. There is a need for researchers to build a more ex-

act understanding of the fundamental constraints and opportunities imposedby electronic commerce technologies.

Economic Analysis of Product-Level Issues in ElectronicCommerce

In this section, the product level of the framework will be discussed in moredetail. The following issues are prominent at this level: pricing digital prod-ucts; the characteristics of other Internet products; bundling, sharing, and dif-ferentiating products; and adoption issues for products with respect to thenetwork externality benefits they confer upon users. Table 3 provides an over-

view of some representative references and findings to orient the reader tothis area of the literature (see Table 3).

Pricing Digital Products


The type of product most suited to electronic distribution is the digital prod-uct or information good, which can be entirely digitized by the producer, down-loaded from the Internet by consumers upon demand, and freely sent overtelecommunication lines. The term product is used in the strict economicsense to refer to either a good or a service. Thus, both MP3 music and a digital


24/112


service like a search on Yahoo are digital products. On-line magazine content,digital music, software objects and packages, digitized maps, and informa-tion archives all qualify as digital products. Their key distinguishing featureis their near zero marginal cost of production. For a more comprehensive over-view, see Hurley and Varian [202].

Because digital products have a near zero marginal cost and are easily re-producible by anyone with a computer, how should firms like Netscape/America Online, Inc., and Microsoft price their Internet browser software? Ithad cost millions of dollars to create the first copy of Netscapes Navigatorsoftware, but making subsequent copies is almost free. In a similar vein, howdoes an electronic brokerage or stock market quote vendor price time-sensi-

Digital products are character-

ized by zero marginal cost, sothe traditional economic result of

pricing at marginal cost does not

allow firms to stay in business.

Thus, firms must find clever ways

to price digital products based

on the customers willingness to

pay rather than minimization of

the items production costs.

The e-commerce environment

changes the characteristics of the

products available, either

through bundling with expertsystems or changes in the

shopping process. Because of

this it is necessary to understand

the impact of these changes on

consumer behavior.

As firms seek ways to sell to

heterogeneous customers,

economics presents a number of

ready-made solutions. Their value

depends on the nature of the

heterogeneity.

Value on the Internet is rooted in

the actual or expected installed

bases of users of a given

technology. Researchers are

seeking to understand how to

create, measure, and leverage

beneficial network externalities

to improve the performance of

products, business processes,

firms, and markets.

Representative Representative

Issue area references findings

Pricing of digital products Bhargava and Choudhary [63];

Brynjolfsson and Smith [79];Clemons, Hann and Hit t [115];

Croson and Hann [123]; MacKie-

Mason et al., [272]; Shapiro and

Varian [334]; West [390]

Table 3. Product Issues, Representative References, and Findings.

Lal and Sarvary [245]; Liang and

Huang [256]; Lynch and Ariely

[267]; Shankar et al. [333]

Bakos and Brynjolfsson [36, 37];

Bakos et al. [39]; Brooks et al.

[73]; Chen and Hitt [93]; Shapiro

and Varian [334]

Cusumano and Yoffie [127];

Prasad and Harker [312];

Shapiro and Varian [335]

Characteristics of Internet

products

Bundling, sharing, and

differentiation

Adoption and network

externalities


25/112


tive information, such as stock quotes? During the first few years of wide-spread diffusion of the Internet, 20-minute delayed quotes were available free,but real-time quotes often required a consumer to have an account with aninvestment firm. The characteristics of stock quote information, as a digital

product, are telling. Entrepreneurs will always find a way to deliver 15-minutedelayed quotes for free (and then 10-minute and 5-minute delayed quotes,etc.), fueling consumer demand, changing the markets expectations, and put-ting pressure on the status quo.

This recently occurred with a market leader, E*trade, which now offers freeaccess to real-time quotes. Other firms have followed suit. As a well-knownInternet-sector security analyst, Bill Burnham, notes, Were now on the slip-pery slope towards free real-time quotes everywhere, for everybody. . . . Oncethat becomes fairly widespread, its on to the next trick [236]. Of course, some-where in this mix, someone does pay for the quote vendors information ser-

vice, since its cost of production is not zero. However, e-brokers have begunto recognize that such information goods act as infrastructure for accountholders who are unable to effectively trade via a Web site that does not pro-vide market information [330].

Recent Research

Managers are concerned about how new technologies will change pricing inthe electronic commerce environment. The primary difficulty with pricinginformation goods comes from the economic result that firms in a competitive

market will charge marginal cost for a product. Since information goods tendto have zero marginal cost, in the long run firms should be forced to give themaway. A good starting point for readers who are new to these issues is Informa-tion Rules: A Strategic Guide to the Network Economy[335, esp. chap. 2]. Theauthors give the example of the digital Yellow Pages, which were very costlyto produce, but once they had been encoded in digital form could be repro-duced at essentially no cost. The first firm to produce the digital Yellow Pagesproduct soon went out of business, as other firms were willing to undercut itsprice all the way down to zero. Today there are a variety of digital YellowPages available on-line at no charge. This is precisely the information goods

pricing problem. However, the fact that consumers are connected to computerequipment allows firms to implement complex and even personalized pricingschemes, which would not be possible with the traditional posted-price mecha-nism. Cusumano and Yoffie take an in-depth look at the way that pricing offree Internet browsers has developed, and the relation between pricing strat-egy and market adoption [127]. A similar problem exists in the area of publicavailability of government databases, where new production, pricing, andaccess issues have arisen. West investigates on-line database vendors, includ-ing those involving the provision of data from U.S. government sources, andhow they have responded to technology changes over time by adjusting theirpricing [390].

Differential pricing (also referred to as price discrimination) occurs when afirm can effectively split or segment its overall market, such that it can charge


26/112


the different segments of its market different prices. This enables the firm tocollect additional surplus from the market that it is not able to obtain with aone-price approach (see Figure2).

The producer firm in Figure 2 can produce Q0at price P

0without price-

discriminating, but it is willing to trade the revenue associated with Area B((P

0 P

L) (Q

0 Q

H]) for the greater revenues associated with Area A ((P

H

P0)QH)and Area C (PL(QL Q0]).How does this work in practice? Varian provides a nontechnical introduc-

tion on how differential pricing of information goods is accomplished in themarket, and what firms can expect in return for doing so [373, 374]. Economictheory offers a great deal of insight about the efficacy of differential pricingapproaches. Such approaches are of considerable value to firms that sell in theInternet marketplace. First, one must emphasize how different digital prod-ucts are from traditional products. Often, no physical production is involved,at least in terms of resources other than human labor and intellectual capital[206]. Second, digital products frame the issue of intellectual property rights,

because digital products can be reproduced for free and then distributed byanyone who has acquired an initial copy. Third, digital products are typicallyinterdependent with standards and technological compatibilities (or incom-patibilities) that may pave the way for product success or doom a great prod-uct idea to failure. As a result, network externalities usually play an importantrole in the adoption of digital products in the marketplace, and, as a result, intheir ultimate success.

Shapiro and Varian look at product versioning in the case of informationgoods and digital products [334]. In their view, versioning is a necessity inthese situations, because of the logic of marginal cost-based pricing. Versioningenables higher-willingness-to-pay consumers to pay positive prices for theitem, which generates revenues for the firm and a large margin. But in a com-petitive market, there will be some consumers whose value for a given infor-

Note: The producer sells QHfor P

Hand Q

L-Q

Hfor P

L.

Figure 2. Differential Pricing

C

A

Price

Quantity

DemandFunctionP0

Q0

PH

PL

QH QL

B

Gain

Loss


27/112


mation good is only marginally greater than zero. Nonetheless, with a nearzero marginal cost of production for that information good, some firms willstill have an incentive to serve such consumers.

This sets the stage for major players in the market to provide information

products to these consumers at a price near to zero. However, they cannot dothis without making the features of the near-zero price product somewhatless attractive than those of the full-cost product. This is especially true for thehigher-willingness-to-pay consumers. And thus the market arrives atversioning as a solution, with older and less capable versions of a softwareproduct or information good made available at a very low price. As a result,lower-willingness-to-pay consumers will accept the less capable version ofthe product, while higher-willingness-to-pay consumers will choose the full-featured, higher-price version. Croson and Hann propose that the concept ofswitching costs explains why upgrades for software products are often much

less expensive than the original purchase [123]. Similarly, Chen and Hitt findthat on-line brokers have significant control over their switching costs [95].Using a measure of switching costs, the authors are able to evaluate the mo-nopoly power of on-line brokerages as the product of switching cost andinstalled base [94]. They find support for the prediction that monopoly power,as described above, increases the prices brokerages are able to charge.

Recent research by Bhargava and Choudhary [63] and Bhargava et al. [64]indicates that versioning may not be a good solution to the pricing of informa-tion goods, where the marginal cost of quality does not increase at an increas-ing rate. The authors show that in such circumstances, selling multiple versionsof a good at multiple quality levels decreases profits relative to selling only

the highest-quality good. Although the existing literature on price discrimi-nation focuses on the case where the cost of quality increases at an increasingrate, they argue that this is not true of information goods, where marginal costis often zero. The rea

Documents

Economics and Electronic Commerce: Survey and Directions for Research