Informational Cascades in IT Adoption - Semantic Scholar

2 downloads 21977 Views 991KB Size Report
Sep 15, 2003 - Director, MIS Research Center. Carlson School of ... learning and managerial incentives in IT adoption herding. By constructing a synthesis ...... different conditions call out for additional research by IS researchers. A number of ...
PAYOFF EXTERNALITIES, INFORMATIONAL CASCADES AND MANAGERIAL INCENTIVES: A THEORETICAL FRAMEWORK FOR IT ADOPTION HERDING Robert J. Kauffman Professor and Chair, Information and Decision Sciences Director, MIS Research Center Carlson School of Management University of Minnesota Minneapolis, MN 55455 Email: [email protected] Xiaotong Li Assistant Professor of Management Information Systems Department of Accounting and MIS University of Alabama, Huntsville Huntsville, AL 35899 Email: [email protected] Last revised: September 15, 2003 _____________________________________________________________________________ ABSTRACT We have recently observed herd behavior in many instances of information technology (IT) adoption. This study examines the basis for IT adoption herding generated by corporate decisionmakers’ investment decisions. We propose rational herding theory as a new perspective from which some of the dynamics of IT adoption can be systematically analyzed and understood. We also investigate the roles of payoff externalities, asymmetric information, conversational learning and managerial incentives in IT adoption herding. By constructing a synthesis of the critical drivers influencing managers’ IT investment decisions, this study will help business researchers and practitioners to critically address the issues of information asymmetries and incentive-compatibility in firm- and market-level IT adoption. Keywords: Agency problem, asymmetric information, cheap talk, herd behavior, incentives, informational cascades, IT adoption, network externalities, reputations, signaling games. ______________________________________________________________________________ Acknowledgements: The authors wish to acknowledge Yoris Au, John Conlon, Rajiv Dewan, David Hirshleifer, John Gallaugher, Angsana Techatassanasoontorn and Al Wilhite for helpful discussions on related work. We also would like to thank two anonymous referees from INFORMS CIST for their helpful suggestions and comments. ______________________________________________________________________________

Note: Published by New Yorker magazine 1972; reproduced from Bikhchandani, Hirshleifer and Welch (1996).

______________________________________________________________________________ INTRODUCTION In the recent years, there have been many instances of information technology (IT) adoption in which we have observed “herd behavior,” as many investment decisionmakers lost touch with their own cautious value-maximizing approaches to investment decisionmaking, and decided to follow the decision of many “smart cookies.” Bikchandani and Sharma (2001) define herd behavior in terms of three related aspects: (1) the actions and assessments of investors who decide early will be critical to the way the majority decides; (2) investors may herd on the wrong decision; and, (3) if they do make the wrong decision, then experience or new information may cause them to reverse their decisions, and a herd may be created in the opposite direction. Herd behavior has long been studied in other academic fields, including Finance and Economics, where the literature is well developed. (See Bikhchandani, Hirshleifer and Welch, 1996). In some cases, such as stock market bubbles or the DotCom mania, herding is driven—in the words of Federal Reserve Bank Chairman, Alan Greenspan—by people’s “irrational exuberance.” However, recent theoretical and empirical studies suggest that in many other cases herd behavior is rather counterintuitively caused by the decisions of perfectly rational people.

1

Such rational decisions at the individual level sometimes result in significant information and welfare losses in the marketplace and the economy. In IT adoption, rational herding has the potential to generate several problems. First, valuable information about new technologies is often lost (or at least poorly aggregated) when IT managers blindly follow others’ adoption decisions. Second, payoff externalities-driven herding makes early adopters’ decisions disproportionately important, and it gives other adopters little chance to compare and experience different technologies. Third, managers may intentionally imitate others’ adoption decisions because of their career concerns, and those reputationmotivated decisions usually fail to maximize expected IT investment payoffs. The widespread mimicry in IT adoption and the resultant inefficiencies motivate us to investigate the basis for technology adoption herding generated by corporate managers’ decisions. A common and well-studied justification for IT adoption herding is positive payoff externalities like network externalities. Recent studies have indicated that many technology markets are subject to positive network feedback that makes the leading technology grow more dominant (Brynjolfsson and Kemerer, 1996; Gallaugher and Wang, 2002; Kauffman, McAndrews and Wang, 2000). Because positive network feedback makes a company’s IT adoption return rise as more companies adopt the same technology, it usually gives managers strong incentives to adopt the technology with the larger installed base of users. In addition to the studies of positive payoff externalities, recent research in information economics demonstrates how rational herd behavior may arise because of “informational cascades” (Banerjee, 1992; Bikhchandani, Hirshleifer and Welch, 1992 and 1998) or managers’ career concerns (Scharfstein and Stein, 1990; Zwiebel, 1995).

2

Informational cascades occur when individuals ignore their own private information and instead mimic the actions of previous decisionmakers. Those mimetic strategies are rational when private information is swamped by publicly observable information accumulated over time. This is why informational cascading is sometimes referred to as “statistical herding” (Ottaviani and Sorensen, 2000). Like informational cascade models, career concerns models have information economics and Bayesian games as their theoretical foundations, but they distinguish themselves by examining rational investment herding through the lens of agency theory. The primary implication of those models is that managers concerned about their reputations may imitate others’ investment decisions to positively influence others’ inferences of their professional capabilities. Although reputational herding decisions are rational for individual managers, they are usually not in the best interests of those companies who hire their managers to maximize investment payoffs. Empirical evidence of herd behavior and imitative strategies has been recently documented in stock analysts’ equity recommendations, emerging technology adoption and television programming selection (Hong, Kubik and Solomon, 2000; Kennedy, 2002; Walden and Browne, 2002; Welch, 2000). There is also extensive experimental evidence of rational herding and informational cascades in the economics literature (Anderson and Holt, 1997; Hung and Plott, 2001). Another recent experimental study of behavioral conformity is Tingling and Parent (2002) in which senior IT and business decisionmakers instead of college students are used as subjects. Despite the fast-growing rational herding literature and the pervasiveness of imitative behavior in IT adoption, systematic studies of IT adoption herding are still rare in the IS literature. By synthesizing previous rational herding models, this paper proposes an integrated

3

theoretical framework within which the dynamics of IT adoption herding can be better analyzed and understood. The next three sections discuss the underlying theories in greater detail. We investigate the relationship between payoff externalities and IT adoption herding in Section 2. We next demonstrate in Section 3 how information asymmetries and observational learning can lead to informational cascades in IT investment. The problem of managerial incentives in IT investment decisionmaking is the focus of Section 4. We discuss why agency problems predispose the market to reputational herding in IT adoption. Managerial compensation schemes designed to address those incentive issues are also discussed. Section 5 provides a synthesis of critical theoretical drivers of IT adoption herding and brings the ideas together into a single integrative framework. Section 6 concludes the paper with the contributions of this work to ongoing research in IS, and ideas for further research. PAYOFF EXTERNALITIES: DOES ADOPTION HERDING PAY OR HURT? One type of positive payoff externalities that is commonly observed in the IT market is “network externalities” (Economides, 1996; Katz and Shapiro, 1994). Network externalities are sometimes referred to as demand-side economies of scale. (For additional constructs related this area of theory, see Table 1). They stem from the presence of significant technology switching costs and the benefits associated with a large installed base of compatible technologies. In the context of IT adoption, network externalities tend to reward herding decisions by increasing the payoffs to IT adopters who associate themselves with the majority. Herding in the presence of network externalities also decreases IT adopters’ risks of being stranded in technologies with too small installed user bases.

4

Table 1. Key Constructs in the Payoff Externalities Theory Relative to IT Adoption CONSTRUCT

DEFINITION

COMMENTS

Rational IT adoption herding

IT adoption herd behavior that can be justified by an individual decisionmaker’s rationality.

Three types of rational justifications exist for IT adoption herd behavior.

Network externalities

A type of positive payoff externalities. The value of a technology increases as the number of users increases. Sometimes referred to as network effects.

They are common in IT markets, and create payoff incentives for decisionmakers to herd in IT adoption.

Negative payoff externalities

It refers to the situation where a company’s return from adopting a technology decreases as more companies adopt the same technology.

They usually punish IT adoption herding and make informational cascades less likely to occur.

Technology switching costs

It refers to the costs incurred for a user to switch from one technology to another. Users may face technology lock-in if switching costs are significant.

Adopters face significant switching costs have more incentives to join an IT adoption herd.

Tippy markets

Markets subject to strong network externalities make technology competition highly unstable and one technology can quickly emerge as the dominant one.

Tippy markets coexist with massive herd behavior in IT adoption, and can create “winner-take-all” situations.

In technology markets subject to network externalities, IT diffusion processes are often characterized by path dependencies. They represent the situational specifics of irreversible managerial decisions and their impacts on the decisions of others. Many managers believe that network externalities and technology switching costs work in tandem to justify imitative technology adoption. In some cases, strong network effects create a “tippy” technology market, in which one technology very quickly emerges as the dominant product because of massive adoption imitation. In such a winner-take-all tippy market, most managers have strong incentives to associate themselves with the winner by quickly jumping on the bandwagon. Thanks to the adoption irreversibility caused by significant technology switching costs, this herding strategy is usually not a risky one. In many cases, the extrinsic network effects (sometimes also called indirect network externalities) created by complementary products and services can further increase the benefits of herding (Basu, Mazumdar, and Rai, 2003; Gallaugher and Wang, 2003). Although herd behavior driven by network feedback can be easily justified by individual rationality, it usually leads to obvious information and welfare losses. Companies make their IT

5

adoption decisions mainly based on the installed user bases of competing technologies. Consequently, managers do not concentrate on the intrinsic merits and suitableness of competing technologies, and under many circumstances they do not even have enough time to compare all available technology choices because the technology competition could end very quickly in favor of a technology. So does network externality-driven IT adoption herding pay off? Or does it hurt those firms that adopt this way? At the individual level, each decisionmaker gains by joining the herd and taking advantage of the network externalities. However, most decisionmakers lose a chance to deliberate the associated opportunities. Very often, as some have claimed for the VHS video format winning out over the Sony Beta format, the market may end up adopting an inferior technology, which will hurt all adopters in the long run (note that network externality is only one of the factors that made VHS prevail in the competition). Payoff externalities, as a stand-alone justification for rational IT adoption herding, has its limitations. Strong network externalities may not be so pervasive in the technology market as many IT and business strategists expected (Liebowitz, 2002; Porter 2001). As a result, imitative technology adoption strategies driven by those illusive network effects are not even individually rational. Moreover, technology managers sometimes choose to adopt emerging technologies with superior performance instead of imitating others by using the dominant technology. Clearly, there is a tradeoff between the future potential of superior new technologies and the network benefits of current technologies. Adoption herding may not persist if some firms find the benefits of exploring new ITs outweigh those of exploiting the dominant IT with network benefits (Lee, Lee and Lee, 2003). It is also worth noting that payoff externalities can be either positive or negative.

6

Unsurprisingly, negative payoff externalities play an important role in mitigating a technology market’s propensity to adoption herding. They are commonly seen in many competitive business environments where downward-sloping demand curves make a company’s IT adoption payoff decrease as more companies adopt the same technology. Therefore, companies imitate others’ IT investment decisions may be punished by intense ex post competition in the downstream market (Li, 2003). Both the fiber cable network glut and the e-commerce gold rush exemplify how severely IT adoption herding may have been penalized by negative payoff externalities. Interestingly, herd behavior still happens in situations where negative payoff externalities are evidently present (Kennedy, 2002). Because of the limitations of payoff externalities as a justification for rational adoption herding, we need to investigate other theoretical explanations of firm-level IT adoption herding. INFORMATIONAL CASCADES: TOO MUCH OR TOO LITTLE INFORMATION? The theory of payoff externalities-driven adoption herding does not sufficiently emphasize two important features that are present in IT diffusion. The first feature is that information asymmetries and information incompleteness are pervasive in emerging technology markets. Different decisionmakers have their own judgments about the value of a new technology based upon their own private information, and generally no one possesses perfect information in adoption decisionmaking. These information structure problems lead to the second feature: to improve decision quality, decisionmakers try to learn valuable information by observing others’ IT adoption decisions. For those who make adoption decisions earlier, their actions may reveal private information to others, which generates information externalities (Zhang, 1997). (See Table 2 for key constructs in the information cascade theory.)

7

Table 2. Key Constructs in the Informational Cascade Theory Relative to IT Adoption CONSTRUCT

DEFINITION

COMMENTS

Information asymmetry

It occurs when some parties know more relevant information than other parties in business transactions.

It causes potential market inefficiencies, e.g., IT adoption herding.

Information incompleteness

It refers to situations in which decisionmakers do not have complete information to make business decisions.

Pervasive in IT adoption; negatively impacts adoption decisions quality.

Informational cascade

Decisionmakers ignore their own private information that is overwhelmed by publicly observable information, and instead mimic others’ actions

A primary mechanism that causes IT adoption herding.

Observational learning

Social learning process in which decisionmakers acquire new information by watching others’ actions.

Used by IT adopters to obtain new information; it may cause informational cascade.

Word-of-mouth learning

Decisionmakers acquire new information through interactions with others; also referred to as conversational learning.

A social learning mechanism in IT diffusion. Credibility is an issue.

Although observational learning can facilitate information conveyance, it sometimes results in an informational cascade in which most people adopt the same technology independent of their own private information. Similar to the decisionmaking process that appears to be operative in the judicial body depicted in the cartoon at the beginning of this article, the reason why informational cascades occur with IT adoption is because the information revealed through others’ adoption actions may have accumulated enough to overwhelm a decisionmaker’s imprecise private information. The opinions of Supreme Court justices, just like the opinions that expressed in a marketplace in which managers make IT adoption decisions, carry substantial weight with others. In this kind of situation, the action of a decisionmaker (even like a shrewd Supreme Court justice) may not depend on his private information. So mimetic decisions actually become incrementally uninformative to later decisionmakers who may also rationally disregard their own information and imitate the prevailing decision. The outcome is that the valuable private information will be lost, which reduces market efficiency because of poor information aggregation.

8

A related intriguing question is whether informational cascades result from too little or too much information. Corporate decisionmakers frequently struggle with too little information to make sound IT investment decisions. That is why they want to gather valuable information from observing other’s adoption actions. Paradoxically, once they engage in observational learning, they may obtain too much accumulated information, to the extent that it may be strong enough to swamp their private information. As a consequence, an informational cascade will occur and most people will imitate early adopters’ decisions that are rather unfortunately based upon limited information. As two mechanisms that cause IT adoption herding, informational cascades and network externalities are not mutually exclusive; in fact, they can be mutually reinforcing. Informational cascades are sometimes fragile though: they can be stopped or reversed by enough newly-arrived information. For example, many companies will be observed to adopt Technology A over Technology B in a herd when their private information is dominated in an informational cascade, even though everyone knows that the valuable information contained in such a cascade is limited. If some credible information is revealed to support Technology B, the adoption cascade can be quickly stopped or reversed. However, informational cascades are more resilient in the presence of network externalities. Once formed, they will be reinforced by later IT adopters who intentionally agglomerate to reap the benefits of positive network feedback (Li, 2003). The strength of informational cascade theory as an explanation for rational IT adoption herding is its emphasis on social learning under information asymmetries. However, social learning can sometimes mitigate a market’s propensity to be influenced by informational cascades. Most informational cascade models assume that decisionmakers can only infer information from observing others’ actions. This exacerbates the information aggregation

9

problem of rational herding. In a simple world where every decisionmaker truthfully tells public his private information, no valuable private information will be lost and the information aggregation problem disappears. In fact, prior innovation diffusion studies have recognized the significant role played by word-of-mouth learning in affecting technology diffusion (Rogers, 1995). Nevertheless, the effectiveness of conversational information sharing in preventing informational cascades should not be overestimated. The major obstacle for effective word-ofmouth learning under many IT adoption scenarios is that each individual decisionmaker’s incentive for truthful information revelation. Potential adopters can benefit from talking with early adopters if what they are told is credible, but who can guarantee the truthfulness of the socalled “cheap talk”? In competitive environments where most IT adoptions occur, individuals may have strong incentives to misinform others through strategic lying or signal jamming (e.g. Crawford, 2003). At the market level, informational cascades occur when the incentive problems associated with information revelation block credible conversations. At the firm level, decisionmakers’ incentives sometimes cause agency problems that provide another explanation for rational IT adoption herding. MANAGERIAL INCENTIVES IN ADOPTION: PROFITS OR REPUTATION? Since a herd involves a group of decisionmakers, it is natural for researchers to concentrate on understanding the market-level interactive dynamics, such as payoff and information externalities. However, significant developments in agency and incentive theory (Laffont and Martimort, 2002) over the last three decades have nourished a stream of rational herding research that explores the role of managerial incentives in fostering investment herding. (See Table 3 for definitions of reputational herding and incentive incompatibility).

10

Table 3. Key Constructs in the Reputational Herding Theory Relative to IT Adoption CONSTRUCT

DEFINITION

COMMENTS

Reputational herding

Managers concerned about their careers intentionally engage in IT adoption herding to enhance their professional reputations and to increase their own human capital returns.

Driven by implicit incentives of careerconcerned managers and informational asymmetries in IT adoption.

Incentive incompatibility

Economic situations where interests of different parties are not perfectly aligned because of the differences in their objectives and motivations.

Without effective incentive contracts to ensure incentive compatibility, managers may engage in inefficient adoption herding.

Agency problems

Refers to those economic inefficiencies caused by conflicting incentives and information asymmetries between principles and their agents.

Agency problems in IT adoption sometimes cause managers’ sub-optimal decisions, including imitative adoptions.

Traditional capital budgeting theory suggests that profit-maximizing companies look at expected investment returns when they make their IT adoption decisions. However, corporate managers hired by a company’s owners or shareholders may have incentives to deviate from the company’s goals and to pursue their own interests when they make their IT adoption decisions. The conflicts of interest, coupled with information incompleteness, can lead to inefficient outcomes known as lemons problems, adverse selection or moral hazard. In a seminal paper in agency theory, Holmström (1999) shows that reputation-concerned managers are very likely to make inefficient investment decisions in the absence of effective mechanisms to align their own interests with those of their companies. In some cases where managerial incentive problems are present, managers may herd solely for reputational purposes in investment decisionmaking. In their influential reputational herding model, Scharfstein and Stein (1990) make the case that managers tend to intentionally imitate others’ investment decisions. Intentional herding comes from the belief that managers do not wish to run the risk to be associated with those who are not identified as being in the highly-talented group. In a more specific context, reputation and career concerns are found to be important in promoting security analysts’ herd behavior commonly seen in their stock recommendations or earning forecasts (Graham 1999; Hong, Kubik and Solomon, 2000).

11

We believe that reputational herding theory has its distinctive advantages in helping us to understand IT adoption herding. Like other important corporate investment decisions, IT adoption decisions are usually made by senior managers. Because their adoption decisions are not immune to agency and incentive problems, they will imitate others’ decisions to enhance their professional reputations if the situation warrants. But unlike most other investment decisions, IT adoption decisions—especially strategic IT adoption decisions—are more susceptible to reputational herding. This is not because a good decisionmaking reputation is more valuable to IT managers like CIOs than to other mangers. Instead, it is because the informational problems are usually more severe. Under many IT adoption scenarios, managers have to make their IT adoption decisions quickly with very limited information. Because IT adoptions are highly specialized tasks that involve a lot of technical details, there are also significant information asymmetries between the decisionsmakers (IT managers) and their supervisors (the firms’s owners or board). Furthermore, the economic payoffs of many IT investments are notoriously difficult to observe or measure in the short run, which gives managers more room to enhance their reputations at the expense of their companies. Most reputational herding studies use signaling (or signal jamming) games in which managers try to positively influence their supervisors’ and the labor market’s posterior beliefs on their capabilities through their investment decisions. Because firms’ owners or the labor market usually lack concrete evidence to indicate whether an individual IT project will be successful or unsuccessful in the short run, IT managers’ professional reputations will heavily depend on the market consensus reached by peer managers or the short-term stock market reactions. As a result, IT managers that are concerned with their reputations are more likely to exhibit herd behavior in their IT adoption decisions than those who are not.

12

Furthermore, when IT managers are concerned about their career prospects, imitating the IT adoption decisions of other will be fully rational if doing so will result in a better reputation. The potential inefficiency and welfare losses stem from the conflicting interests among different parties. Therefore, the key to preventing inefficient reputational herding is to address the issue of incentive compatibility. By offering managers appropriate compensation contracts, firms can provide them with explicit incentives to maximize investment returns. Two difficulties arise in the context of IT adoption, however. First, it is usually hard to quantify IT investment payoffs, at least in the short run. Incentives from ambiguously designed performance-based contracts are easily subjugated by the implicit incentives from managers’ career concerns. For example, compensation contracts based on short-term stock price could actually exacerbate the efficiency loss caused by rational investment herding (Brandenburger and Polak, 1996). Second, long-term performance-based compensation, like a bonus of stock options, is thought to be more effective in solving agency problems. However, many IT investments play an instrumental role in strengthening companies’ long-term competitiveness. An IT manager’s decision, unlike a chief executive officer’s decision, generally will have less impact on a firm’s overall performance. So long-term performance-based compensation must be significant enough to dominate a manager’s gains from reputation building. Unfortunately, this usually makes these compensation schemes very expensive to implement. IS researchers are encouraged to study how to overcome this problem in designing optimal incentive contracts for managers who make IT adoption decisions. A SYNTHESIS OF CRITICAL DRIVERS IN ADOPTION HERDING An in-depth understanding of the rational herding theory that we have discussed is the prerequisite for effective analyses of observed herd behavior in IT adoption. In this section, we

13

provide a framework that permits us to systematically evaluate the applicability and the quality of the insights that the theory offers. Our framework cuts the alternative theoretical perspectives on adoption herding with levels of analysis that represent the possible stakeholders in different IT adoption settings. Framework Preliminaries We address two issues in developing our theoretical framework for IT adoption herding. First, almost all the rational herding models in the literature study investment herding in generic investment settings. Although studying herd behavior in generic investments increases the generality of the results, it sometimes fails to capture the distinctive features of certain types of investments like IT adoption. Table 4 shows how certain features of IT investments can increase the likelihood for rational investment herding to occur. Table 4. Rational Herding in Generic Investments and IT Investments RATIONAL HERDING

GENERIC INVESTMENTS

IT INVESTMENTS

Payoff externalitiesdriven herding

Investment herding may arise when there are strategic complementarities among investors who make similar decisions. Negative externalities may curb investment herding in many competitive environments.

Many IT markets are subject to network externalities, a type of positive payoff externality. Network effects and significant IT switching costs are the driving forces behind many instances of IT adoption herding.

Informational cascades (statistical herding)

Decisionmakers imitate others’ investment decisions when their private information is overwhelmed by the information acquired through observational learning.

Information incompleteness and information asymmetries in the IT market make observational learning important for decisionmakers, and thus increase the possibility of an informational cascade.

Instead of helping their companies to maximize investment returns, managers imitate others’ decisions to build their reputations and to increase their own human capital returns.

The financial returns of most IT investments are hard to measure in the short run, which creates incentives for managers to engage in reputational herding when situations warrant it.

Reputational herding

Second, our integrated framework recognizes the differences and relationships among the three explanations for rational IT adoption herding. In order to study IT adoption herding from appropriate theoretical perspectives, we need to understand the strengths and weaknesses of

14

different IT adoption models. A comparison of three types of adoption herding models is given in Table 5. Table 5. A Comparison of IT Adoption Herding Models COMPARISON DIMENSIONS

PAYOFF EXTERNALITYDRIVEN MODELS

INFORMATIONAL CASCADE MODELS

REPUTATIONAL HERDING MODELS

Theoretical foundations

Payoff interdependency and strategic complementarities.

Information economics and Bayesian learning.

Information economics, contracting and agency theory.

Aspects of IT adoption emphasized

Switching costs, network effects and incompatibilities.

Information externality and social learning in IT adoption.

Managers’ incentives and career concerns in IT adoption.

Strengths of the theory

Many IT markets are subject to positive network feedback. Herding models are generally more intuitive and robust.

Models demonstrate how herding arises because of information asymmetries and the problems associated with observational learning.

Shows herding may be caused by managerial incentive problems and thus build a bridge between agency theory and rational herding theory.

Weaknesses of the theory

Hard to explain herding in situations where network externalities are weak or negative payoff externalities are strong.

Simplified assumptions of market information structure and learning processes make these models unrealistic in some settings.

The conditions that lead to herding are more complex and less obvious. Most models only deal with very simple investment settings.

Our framework considers four decisionmaking levels from three different theoretical perspectives. The first level is the business process level at which managers recommend appropriate technologies for their companies to adopt. The second level is the firm level at which companies’ senior management or boards approve or disapprove managers’ adoption recommendations. The third level is the industry level at which companies collaborate to improve the impacts of their IT adoption decisions. An example of this industry-level collaboration is the formation of IT standard groups intended to promote technology compatibility and interoperability. The fourth level is the economy level at which regulators make policies to improve social welfare associated with industry and society-wide IT adoption. Our motivation for proposing a multiple-level framework is not only that an adoption decision often involves many stakeholders at different decisionmaking levels, but also that different herding-related theories can generate distinctive insights to these different stakeholders. For example, managerial incentives are particularly useful in explaining managers’ imitative 15

adoption decisions at the business process level. Similarly, network externalities are usually the driving force behind industry-level IT standard formation. An Evaluative Framework for Assessing Herding Theories of IT Adoption Based upon the four decisionmaking levels discussed above, we present an evaluative framework in Table 6. Table 6. An Evaluative Framework for the Alternative IT Adoption Herding Theories THEORETICAL PERSPECTIVE

BUSINESS PROCESS LEVEL / IT ADOPTERS, IT MANAGERS

FIRM LEVEL / BOARD OF DIRECTORS, SENIOR MGMT

INDUSTRY LEVEL / STANDARD GROUPS,

ECONOMY LEVEL / REGULATORS

Payoff externalities

IT managers often adopt ITs with larger installed bases because of network externalities and significant technology switching costs.

Senior mgmt needs to know that network effects give managers incentives to herd. This type of herding is intended to maximize investment payoffs.

Firms may coordinate their IT adoptions to improve technology compatibility, but they should know that payoff externalities might be negative.

This type of herding is payoff-driven and rational at the firm level, but the resultant social welfare loss may justify regulators’ interventions.

Informational cascades

IT managers sometimes blindly follow others’ adoption decisions because their private information is overwhelmed by publicly observable information.

Senior mgmt needs to know that managers’ decisions are independent of their private information and are thus not very informative in an adoption cascade.

Firms may voluntarily collaborate to share information to reduce the possibility for a cascade to occur, but they must address the issue of credibility in information transmission.

Regulators can facilitate information transmission to avoid potential welfare and information losses. Their actions are particularly effective when cascades are fragile.

Managerial incentives

IT managers concerned with their careers and reputations have implicit incentives to imitate others’ adoption decisions if doing so benefits their careers.

Senior mgmt needs know that managers may engage in inefficient reputational adoption herding without appropriate managerial compensation contracts.

Credible information sharing among firms could reduce the possibility of herding because incentive problems are often exacerbated by information problems.

Regulators concerned about this type of inefficient herding may design policies to encourage effective corporate governance and to promote optimal incentive contracts.

COMPANY CONSORTIA

The table shows the major strategic insights for different IT adoption stakeholders from three theoretical perspectives. Faced by potential adoption herding, these stakeholders usually ask themselves some of the following questions: What can a decisionmaker gain or lose by engaging in adoption herding? How will my strategy affect other stakeholders and how will their strategies affect me? What potential loss can adoption herding cause? Are there any ways to

16

deal with inefficient herding? Their answers to these questions will decide which adoption herding theoretical perspective is the most appropriate and relevant one under their circumstances. For examples, if a company’s board suspects that a manager may gain a lot by imitation, it will scrutinize his decisions for potential incentive problems. If a group of companies are concerned about their insufficient information of a new technology, they may cooperate to share their information to reduce the likelihood for an informational cascade to occur. To conduct an IT adoption case analysis by using our framework, we first need to identify important stakeholders in adoption decisionmaking. Then we examine what their answers will be to some of these herding-related questions. By doing that, we place ourselves in the particular situations that these stakeholders face, which allows us to understand their strategic concerns in IT adoption decisionmaking. These concerns usually determine which theoretical perspective is the most appropriate one. After obtaining the relevant information, our evaluative framework can serve as a roadmap from which appropriate managerial insights can be found. In a more complex setting, multiple stakeholders are crucial to IT adoption decisionmaking and their different concerns suggest that more than one theoretical perspective come into play. So our framework may provide multiple directions in analyzing such a real world case. When this is the case, more effort should be made to find consistent insights from multiple theoretical perspectives to yield a more thorough understanding of the underlying drivers of IT adoption herding. Three Applications of the Evaluative Framework for IT Adoption Herding We now discuss the appropriateness of each theoretical perspective in terms of its general explanatory fit for real world settings in which some form of IT adoption herd behavior has been

17

observed. Payoff Externalities-Driven IT Adoption: The Case of Wi-Fi Adoption. The recent dramatic upswing in market interest for fixed location wireless services, especially Wi-Fi (or Wireless Fidelity), is a good example of IT adoption herding that is occurring primarily on the basis of perceived network externalities. An industry standards organization, the Wi-Fi Alliance (www.wi-fi.org), is involved in the development and propagation of technical standards for further innovations. Recently, the term “Wi-Fi” has been expanded from IEEE 802.11b to apply to IEEE 802.11a, and there is a growing interest to further extend it to the emerging IEEE 802.11g standard. The compatibility externalities arise as vendors design and build new wireless products that are certified to interoperate with one another across manufacturers, and as consumer purchase those devices. The major wireless telecommunication firms have recently been making major commitments to creating public fixed wireless service locations in restaurants, bookstores, malls, airports, and other public spaces. One among this group especially stands out, T-Mobile USA (www.tmobile.com), a mobile telecommunication subsidiary of Deutsche Telekom, which has built almost 2,400 “hotspots” for their Wi-Fi Internet connection services in 24 states. Many Wi-Fi hotspots have become available for free public access to e-mail and the Internet. However, corporate participants have profit-generating business models behind their own efforts. T-Mobile reportedly offers fee-based Wi-Fi connectivity whose attractiveness to consumers depends on the density and extent of the available locations, a network externality for which the wireless-using public has already shown a willingness to pay. Current fees are $29.99 per month for unlimited access in a region, with 500 MBs of data download and transfer capacity per month. Non-metro usage is charged at $0.15 a minute, indicating that the primary targets are

18

corporate users. Figure 1 maps the location density of the Wi-Fi hotspots in the New York, Atlanta and San Francisco Bay Area metropolitan regions. The diffusion levels are varied. Figure 1. T-Mobile USA’s Penetration for Wi-Fi Services in Selected U.S. Urban Regions

New York City Metro: This is a rapid growth area for T-Mobile USA’s Wi-Fi service, creating an installed base of locations and users that is among the most well developed in U.S. urban regions. The number of participating locations reached approximately 232 by mid-May, 2003. Source: T-Mobile USA, Regional Plans, accounts.hotspot. t-mobile.com/ signup plans/ signup_plans_local_ nyc.jsp, May 12, 2003.

Atlanta Metro: This area is earlier in its adoption of T-Mobile USA’s Wi-Fi hotspots. Altogether, the installed base of Wi-Fi hotspots totaled only 57 by mid-May 2003. The externalities associated with the network are still relatively weak, given the size of the metro area, but growing. Source: T-Mobile USA, Regional Plans, accounts.hotspot. t-mobile. com/signup plans/ signup_plans_ local_atlanta.jsp, May 12, 2003. San Francisco Bay Area: This is another area where the positive externalities associated with the Starbuck’s Coffee network of TMobile Hotspots are highly valuable for the firm and consumers who use the Wi-Fi services. The installed base reached 253 hotspots as of midMay 2003. Source: T-Mobile USA, Regional Plans, accounts. hotspot. t-mobile. com/signup plans/ signup_plans_ local_ atlanta.jsp, May 12, 2003.

19

Two important stakeholders are the adopting corporations and the consumers who use Wi-Fi hotspots by setting up fee-based accounts. Recently BusinessWeek magazine reports on a significant upsurge in adoption of these services to as many as 5,000 free hotspots and 18 million global users (Green et al., 2003). Clearly, the market is taking a “rush-is-on” attitude towards finding profitable locations to operate fixed wireless services, and companies are likely to take advantage of the beneficial externalities that they can offer to potential customers, and then monetize through fee-based services, similar to what we saw with electronic banking and shared ATM networks in the 1980s into the 1990s (Clemons, 1990). At the firm level, senior managers and board members should recognize the extent to which the positive network externalities will create incentives for firms to herd, in order to maximize the investment payoffs. At the industry and market level, we expect to see coordination of firm efforts to maximize network externalities, similar again to what we saw with shared electronic banking network adoption, for example, and more recently with business- to-business eprocurement market services adoption. The next few years in the marketplace should see more Wi-Fi adoption driven by the fast-growing installed user base. Although nobody yet has any “hard” evidence about the profits and returns associated with setting up Wi-Fi hotspots, we expect that the actual returns will be more broadly shared, as in the case of the introduction of credit card to gasoline service station operations. The evidence shows credit card-adopting gas stations for a number of years had benefited from sale revenue increase. It is unclear at this time what analogous collateral benefits for Wi-Fi hotspots companies can monetize, although Starbuck’s Coffee has already begun touting the productivity and connectivity boosts that are available to traveling business people who want to use Starbuck’s stores like an “office away from the office,” thus leading to increased food and coffee purchases

20

(PNM Publications, 2002). Informational Cascades-Influenced IT Adoption: ICQ and Peer-to-Peer Messaging. This interesting example illustrates why informational cascades theory is useful to understand the adoption of IT—even at the level of individual users. Consider the case of Internet-based P2P instant messaging services. On November 15, 1996, the first instant messaging tool, ICQ, the brainchild of four computer enthusiasts in their mid-twenties, was released by Mirabilis, Ltd. (ICQ Inc., 2003). Within six weeks, 30,000 people had signed on to use ICQ, and after six months 1 million. In May 1997, America Online (AOL) released its AOL Instant Messenger, also called AIM. By 1998, after ICQ was acquired by America Online (AOL) for $287 million in June (Brooker, 1999), Red Herring magazine reported that ICQ’s installed base of users had grown to 16 million, one-half of whom were reasonably active users (Needleman, 1998). The rapid growth in installed base continued unabated until the end of 2001, when participation had grown to 117 million users, as shown in Figure 2. Because of ICQ’s first-mover-advantage, it was able to achieve its significant installed base before competing instant messaging tools from Yahoo.com and Microsoft were able to block its growth (Niese and Mook, 2001). Figure 2. Growth in Installed Base of ICQ Users, 1997-2001

Source: Niese and Mook (2001).

21

The observed herd behavior in the early stage of ICQ’s diffusion was largely driven by an informational cascade about the merits of the software’s innovative functionality, by hearsay about the related experience for people who adopted ICQ (especially the immediacy of the replies and the networks of messaging friends), additional conversational interest in Internet chat rooms, and other informal means of information transmission. Another key driver was Mirabilis’ approach to quickly generating an installed user base with the hope of igniting positive network feedback: it employed Netscape’s strategy of subsidizing user adoption by giving away its software, and not really worrying very much about how to nail down an effective revenue and profitability model. Moreover, ICQ relied on the activation of personal networks, and applying different techniques to accelerate the natural contagion that allowed its software to “create a buzz” in the market. However, up to the acquisition of Mirabilis Inc. by AOL and in spite of the expansion of the software’s functionality, ICQ’s position in the market has been subject to strategic vulnerabilities. Since P2P instant messaging tools were not designed to interoperate with one another, other firms with larger pre-existing user bases of complementary products could bring their own products to market. Consumer adoption after the first two or three years has been mostly occurred on the basis of network externalities, since so many others who had adopted ICQ still did not have to pay for it. By the time that Netscape and Microsoft came to recognize the importance of P2P computing and their own market power, ICQ had already gained significant market share. In fact, Download.com, a popular shareware download Web site, reported that there were 100 million ICQ software downloads by 2000 (Cowger, 2000) and 200

22

million by 2002 (Cunningham, 2002). So this case demonstrates how an informational cascade can be sustained by network externalities in IT adoption. A Managerial Incentives-Driven Interpretation: Morgan/Reuters and Riskmetrics. During the early part of the 1990s, innovations in financial risk management were becoming rapidly available. They offered new algorithmic and computational approaches to the assessment of asset price movement correlations and historical volatility in financial market. (Jorion, 2002; RiskMetrics Group, 2003). The new techniques were referred to by the innovating firms in different ways, including risk-adjusted return on capital (RAROC) by Bankers Trust Company, risk management units (RMUs) by Manufacturers Hanover Trust, and value-at-risk (VaR) and RiskMetrics by J. P. Morgan Bank, but the innovations were all rather similar. The new approaches made it possible for financial services firms to gauge the overnight risk of losing specific amounts of money related to portfolio positions in various financial instruments in terms of instrument-to-instrument correlations in movement to exogenous market shocks. JP Morgan, in particular reworked its risk management system, moving from notional market risk limits to VaR-based limits (Contingency Analysis, 2003). The bank’s chairman asked for a report for a daily risk management and treasury operations meeting, which when combined with the profit and loss statements of the different financial market operations of the firm, led to new managerial perspectives that would eventually be sold to the market—only not at first. Morgan made the biggest impact on the marketplace by using its strengths as a money center bank in treasury and risk management advisory services. It created a technique that permitted other financial firms with which it had pre-existing relationships to benchmark the risk of specific portfolios against the risk of market composite portfolio. The benchmarks also made it

23

possible for firms to determine whether their capital coverage was sufficient—neither too great and thus inefficient, nor too small and thus create potential liquidity problems should extraordinary events occur in the market. The market composite portfolio that was constructed by Morgan was made possible through the application of data on similar portfolios that other firms provided, so as to create both absolute and relative assessments of market risk. Morgan first offered RiskMetrics as a free service (Contingency Analysis, 2003). The sharing of data led to the creation of market risk benchmarks for derivative trading, money market and currency trading and credit portfolio evaluation. A significant amount of adoption occurred in rapid-fire fashion during 1993 to 1997 for Bankers Trust’s and Morgan’s respective risk management products, shortly after the “rogue trader” scandal that led to the bankruptcy of Barings Bank Ltd. in 1992. Morgan initially positioned VaR and RiskMetrics for use in existing treasury relationships, and was able to make a strong case relative to the managerial incentives associated with adopting innovations to aid in the measurement and management of financial risk. There was widespread agreement among the major stakeholders—senior management and boards of directors, industry-level financial risk management groups, and banking and finance regulators that greater investments needed to be made to control financial risk. So there was pressure to adopt from these quadrants of the economy. For managers who choose not to adopt the innovation, the damages on their reputations would be devastating if their firms were badly hurt in market shocks as a result of their decisions. So most reputation-motivated and risk averse managers had strong incentives to adopt the technology. Furthermore, many of them might also realize that early adoption of this risk management innovation was a beneficial career move, in terms of opportunities to lead the practice, to develop additional sophisticated knowledge of new techniques, to improve job

24

mobility and centrality in human capital terms. Overall, the short-term clustering of adoptions of RiskMetrics and RAROC were suggestive of reputational herding, in that it was becoming clear that senior financial managers at leading financial firms did not wish to get left behind, even without convincing evidence to indicate that the business payoff of adoption would be high enough to justify the cost. With Reuters’ subsequent acquisition of a major part of J. P. Morgan’s RiskMetrics, and the formation of at-arm’s-length RiskMetrics Group the market is continuing to adopt these innovations in a variety of forms. Discussion The perspectives that we offer on the potential explanations for IT adoption herding under different conditions call out for additional research by IS researchers. A number of possible problems come to mind, in addition to the complications that are likely to develop when mixed theoretical explanations seem more appropriate than any single interpretation. First, perceptions about the variance and correlation of others’ beliefs can change the tendency towards herding and interfere with adoption. Herd behavior has an important precursor: the expectations of the economic agents who decide to adopt a technological innovation must be in alignment. We believe that perceptions about the variance of others’ beliefs provide another important basis for explaining observed IT adoption behavior. When there is significant observed variance of others’ beliefs about the business value of an IT, a potential adopter is unlikely to imitate others’ decisions. As suggested in Lohmann (2000), the complex interactions of the market participants’ expectations will influence the likelihood that herd behavior occurs. Second, influence costs may muddle adoption decisionmaking, but adopters will tend to discount them to an appropriate degree over time. We have pointed out earlier that credible

25

information transmission is challenging in many competitive business environments. Some firms—especially those who have already adopted a new IT and are looking forward to the adoption of other firms to build up the related payoffs from the network effects—may overstate the benefits of an IT to others. This becomes an influence cost to others in the market because of the possibility that other potential adopters make up their minds to adopt on the basis of consistently inflated estimates. Third, incomplete contracts, suboptimal ownership structures and bargaining problems need to be emphasized in some complex IT adoption cases that involve multiple firms. Sometimes adoption herding arise because of contract incompleteness and incentive incompatibility in joint IT investments. The theory of incomplete contracts (Hart and Moore, 1990) provides a basis for understanding many complex issues of joint IT investment. Some of these issues are asset ownership, inter-organizational governance and bargaining. It merits further research to investigate how these issues interact to affect IT adoption herding dynamics. CONCLUSION This study provides a new theoretical framework for understanding some observed forms of IT adoption herding and the related fundamentals of managerial decisionmaking for IT investments. Herding, as a type of human imitative behavior, tends to result rather naturally from people’s imperfect reasoning and cognition. In many cases, inefficient IT adoption decisions may be attributed to the adopters’ bounded rationality (Au and Kauffman, 2003). Thus, it will be helpful to explore some of the potential behavioral justifications of observed investment decision herding under certain circumstances (Schiller, 1995). It is worth noting that our exposition in this article should not be interpreted as a basis for rationalizing the herd behaviors that are commonly observed in IT investment decisionmaking. Nevertheless, we

26

believe that the rational herding theory synthesized in our framework is likely to be very useful in many settings where bounded rationality plays a significant role. In fact, many problems associated with bounded rationality can be studied in fully rational models with imperfect information and vice versa (Conlisk, 1996). By formally analyzing IT adoption herding in a somewhat simplified rational world, we are able to pay attention to the relevant information and incentive problems, without sacrificing the real world applicability of most of the managerial insights that will be generated. Our framework suggests payoff externalities, informational cascades and managers’ career concerns as three interrelated explanations for the kinds of imitative decisionmaking behaviors that are observed in IT adoption. We investigated network externalities, observational learning and managerial incentives as critical drivers influencing managers’ IT investment decisions. By doing so, we demonstrated the relevance and the importance of our proposed framework to academic researchers, business strategists and IT investment managers. Instead of introducing various rational herding models as standalone theories, we emphasized their inherent relationships in an attempt to acquire more distinctive insights for IT adoption herding. For example, we showed why an adoption herd is much more likely to form in an emerging technology market where both information asymmetries and network effects come into play. We also demonstrated why reputational herding is more common in IT adoption processes where agency problems are compounded by severe information problems. We believe that future studies within our framework will not only enhance our understanding of herd behavior in IT investment, but also offer insights and contributions to the rational herding literature in Economics. Most informational cascade models downplay the importance of conversation and information sharing across firms because of the credibility issue. However,

27

word-of-mouth communications are usually very useful in IT diffusion and social learning in general (Ellison and Fudenberg, 1995; Rogers 1995), and the growing literature on “cheap talk” also sheds light on the effectiveness of conversational learning in strategic interactions (Crawford and Sobel, 1982; Farrell and Rabin, 1996). So future studies that explore the role of conversation in IT adoption and diffusion have the potential to generalize the prior informational cascade models. Future studies of IT adoption herding can also provide additional insights related to the theory of reputational herding. In most reputational herding models that we have seen to date, the outcomes of managers’ investment decisions are observable ex post in a mechanistic manner. So the labor market and firm owners can infer managers’ capabilities through Bayesian updating. However, as we pointed out earlier, many IT projects are strategically instrumental to the firms that invest in them, and yet their payoffs are hard to measure in the short run. This complicates the inference process that leads to managerial decisions about IT investments, and thus requires the extension of reputational herding models so that they will yield more useful insights in this managerial context. Finally, we believe that the framework proposed in our study can also make contributions to the IS literature in some contexts other than IT adoption. For example, we frequently observe herd behavior in IS curriculum design, research topic selection, consumer online shopping patterns, IT firms’ advertising campaigns, R&D spending in developing certain kinds of information systems, and so on. Our study suggests that IS researchers ought to pay close attention to a number of related issues when they study a decisionmaker’s behavior in those contexts. The first issue is payoff interdependence. It is important to try to figure out whether a decisionmaker’s decision will hurt or benefit others who make the same decision. The second issue is information gathering through learning. It is useful to understand whether a

28

decisionmaker can learn information from the similar decisions of others, and whether the information gathered is credible. The third issue is the incentives of different stakeholders. Here, we want to know whether there are conflicts of interest among the different stakeholders, and if so, whether a specific decisionmaker will have sufficient incentive to pursue her own interests at the expenses of other stakeholders. Our future research will take up some of these issues. REFERENCES [1] Anderson, L. and C. Holt (1997) “Information Cascades in the Laboratory,” American Economic Review, 87, 5, 847-862. [2] Au, Y., and R. J. Kauffman, (2003) “What do You Know? Rational Expectations in IT Adoption and Investment,” Journal of Management Information Systems, 20, 2, in press. [3] Banerjee, A, (1992), “A Simple Model of Herd Behavior,” Quarterly Journal of Economics 107, 3, 797-818. [4] Basu, A. Mazumdar, T. and S. Rai (2003), “Indirect Network Externality Effects on Product Attributes,” Marketing Science, 22, 2, 209-222. [5] Bikhchandani, S., Hirshleifer, D. and Welch. I. (1992), “A Theory of Fads, Fashion, Custom, and Cultural Change as Informational Cascades,” Journal of Political Economy, 100, 5, 992-1026. [6] Bikhchandani, S., Hirshleifer, D., and Welch, I. (1996). “Informational Cascades and Rational Herding: An Annotated Bibliography,” mimeo, Yale University. [7] Bikhchandani, S., Hirshleifer, D. and Welch, I. (1998), “Learning from the Behavior of Others: Conformity, Fads, and Informational Cascades,” Journal of Economic Perspectives, 12, 3, 151-170. [8] Bikhchandani, S. and S. Sharma (2001), “Herd Behavior in Financial Markets,” IMF Staff Papers, Vol. 47, 3, 279-310. [9] Brandenburger, A., and B. Polak (1996), “When Managers Cover Their Posteriors: Making the Decisions the Market Wants to See,” Rand Journal of Economics, 27, 3, 523-541. [10] Brooker, K. (1999), “No Sales? No Profits? No Problem. That’s What AOL Said When It Bought Net Startup ICQ,” Fortune, February 15, 1999. [11] Brynjolfsson, E., and C. Kemerer (1996), “Network Externalities in Microcomputer Software: An Econometric Analysis of the Spreadsheet Market,” Management Science, 42, 12, 1627-47. [12] Clemons, E. K. (1990), “MAC: Philadelphia National Bank's Strategic Venture in Shared ATM Networks,”Journal of Management Information Systems, 7, 1, summer 1990, 5-26. [13] Conlisk, J. (1996), “Why Bounded Rationality?” Journal of Economic Literature, 34, 669700. 29

[14] Contingency Analysis, (2003). “RiskMetrics,” in RiskGlossary.com, in association with Contingency Analysis, Boston, MA, available on the Internet at www.riskglossary.com/articles/riskmetrics.htm. [15] Cowger, J. (2000), “CNET Announces 100 Million Downloads for #1 Product on Download.com,” press release, CNET Networks, San Francisco, CA, June 16, 2000. Available on the Internet at www.cnet.com/aboutcnet/0-13613-7-2089600.html. [16] Crawford, V., and J. Sobel (1982), “Strategic Information Transmission,” Econometrica, 50, 6, 1431-1451. [17] Crawford, V. (2003), “Lying for Strategic Advantage: Rational and Boundedly Rational Misrepresentation of Intentions,” American Economic Review, 93, 1, 133-149. [18] Cunningham, W. (2002), “ICQ First: 200,000,000 Downloads on CNET,” CNET Download Newsletter, CNET Network, San Francisco, CA, May 21, 2002. Available on the Internet at www.icq.com/info/press/cnet200.html. [19] Economides, N. (1996), “The Economics of Networks,” International Journal of Industrial Organization, 16, 4, 673-699. [20] Ellison, G., and D. Fudenberg (1995), “Word-of-Mouth Communication and Social Learning,” Quarterly Journal of Economics, 110, 1, 93-125. [21] Farrell, J., and M. Rabin (1996), “Cheap Talk,” Journal of Economic Perspectives, 10, 3, 103-118. [22] Gallaugher, J. and Y. Wang (2002), “Understanding Network Effects in Software Markets: Evidence from Web Server Pricing”, MIS Quarterly, 26, 4, 303-327. [23] Graham, J. (1999), “Herding Among Investment Newsletters: Theory and Evidence”, Journal of Finance, 54, 1, 237-269. [24] Green, H., S. Rosenbush, R. O. Crockett, and S. Holmes. (2003), “Wi-Fi Means Business,” BusinessWeek, 3830, April 28, 2003, 86-92. [25] Hart, O., and Moore, J. (1990), “Property Rights and the Nature of the Firm,” Journal of Political Economy,” 98, 6, 1119-1158. [26] Holmström, B. (1999), “Managerial Incentive Problems: A Dynamic Perspective,” Review of Economic Studies, 66, 1, 169–182. [27] Hong, H., J. Kubik and A. Solomon (2000), “Security Analysts’ Career Concerns and Herding of Earnings Forecasts,” Rand Journal of Economics, 31, 1, 121-44. [28] Hung, A., and C. Plott (2001) “Information Cascades: Replication and an Extension to Majority Rule and Conformity-Rewarding Institutions,” American Economic Review, 91, 5, 1508-20. [29] ICQ, Inc. (2003). “The ICQ Story,” ICQ Inc. Available on the Internet at company.icq.com/info/icqstory.html. [30] Jorion, P. (2000), Value-at-Risk: The New Benchmark for Managing Financial Risk, McGraw-Hill, New York, NY. [31] Katz, M., and C. Shapiro (1994), “System Competition and Network Effects,” Journal of Economic Perspectives, 8, 2, 93-115. [32] Kauffman, R., McAndrews, J. and Y. Wang (2000), “Opening the “Black Box” of Network Externalities in Network Adoption,” Information Systems Research, 11, 1, 61-82.

30

[33] Kennedy, R. (2002), “Strategy Fads and Competitive Convergence: An Empirical Test for Herd Behavior in Primetime TV Programming,” Journal of Industrial Economics, L, 1, 5784. [34] Laffont, J., and D. Martimort (2002), The Theory of Incentives: The Principal-Agent Model, Princeton University Press: Princeton, NJ. [35] Lee, J., Lee, J., and H. Lee. (2003), “Exploration and Exploitation in the Presence of Network Externalities,” Management Science, 49, 4, 553–570. [36] Li, X. (2003), “Informational Cascades in IT Adoption,” Communications of the ACM, forthcoming. [37] Liebowitz, S. (2002), Rethinking the Network Economy: The True Forces That Drive the Digital Economy, Amacom Publishing: Saranac Lake, NY. [38] Lohmann, S. “I Know You Know He or She Knows We Know You Know They Know: Common Knowledge and the Unpredictability of Informational Cascades," in D. Richards (Ed.), Political Complexity: Non-Linear Models of Politics,” Ann Arbor, MI: University of Michigan Press, 2000, 137-173. [39] Needleman, R. (1998), “The Product Line: Do You ICQ?” Red Herring, 60, November 1. Available on the Internet at www.redherring.com/mag/issue60/line.html. [40] Niese, N., and Mook, N. (2001), “ICQ Celebrates Five Years,” BetaNews, November 15. [41] Ottaviani, M., and P. Sorensen (2000), “Herd Behavior and Investment: Comment,” American Economic Review, 90, 3, 695-704. [42] PMN Publications. (2002). “Starbuck’s, T-Mobile Launch Wi-Fi Hotspots,” PMN Publications, Surrey, UK, August 21, 2002. [43] Porter, M. E. (2001), “Strategy and the Internet,” Harvard Business Review, 79, 3, 62-79. [44] RiskMetrics Group (2003). “About RiskMetrics Group.” Available on the Internet at www.riskmetrics.com/aboutus.html. [45] Rogers, E. (1995), Diffusion of Innovations, 4th Edition, New York, NY: Free Press. [46] Scharfstein, D., and J. Stein (1990), “Herd Behavior and Investment,” American Economic Review, 80, 3, 465-479. [47] Schiller, R. (1995), “Rhetoric and Economic Behavior: Conversation, Information and Herd Behavior,” American Economic Review, 85, 2, 181-185. [48] Tingling, P. and M. Parent (2002), “Mimetic Isomorphism and Technology Evaluation: Does Imitation Transcend Judgment?” Journal of the Association for Information Systems, 3, 113-143 [49] Walden, E. and G. Browne (2002), “Information Cascades in the Adoption of New Technology,” L. Applegate, J. DeGross, and R. Galliers (Eds.), Proceedings of the TwentyThird Intl. Conf. on Information Systems, Barcelona, Spain, December 2002, 435-443. [50] Welch, I. (2000), “Herding Among Security Analysts,” Journal of Financial Economics, 58, 3, 369-96. [51] Zhang, J. (1997), “Strategic Delay and the Onset of Investment Cascades,” Rand Journal of Economics, 28, 1, 188-205. [52] Zwiebel, J. (1995), “Corporate Conservatism and Relative Compensation,” Journal of Political Economy, 103, 1, 1-25.

31