Closing the Technology Adoption/Use divide: The...

45
Closing the Technology Adoption/Use divide: The role of Contiguous User Bandwagon Gianvito Lanzolla Cass Business School 106 Bunhill Road, Room 4065 London EC1Y 8TZ, UK Tel. 44 (0)20 7040 5243 email: [email protected] Fernando F. Suarez Boston University School of Management 595 Commonwealth Avenue, Room 546-F Boston MA 02215, USA Tel. (617) 358-3572 email: [email protected] 27 November 2009

Transcript of Closing the Technology Adoption/Use divide: The...

Closing the Technology Adoption/Use divide:

The role of Contiguous User Bandwagon

Gianvito Lanzolla

Cass Business School

106 Bunhill Road, Room 4065

London EC1Y 8TZ, UK

Tel. 44 (0)20 7040 5243

email: [email protected]

Fernando F. Suarez

Boston University School of Management

595 Commonwealth Avenue, Room 546-F

Boston MA 02215, USA

Tel. (617) 358-3572

email: [email protected]

27 November 2009

2

Closing the Technology Adoption/Use Divide:

The Role of Contiguous User Bandwagon

Abstract

A firm may readily subscribe to a new technology, but then fail to use it. This paper

advances existing technology diffusion theory by bringing in a new construct that can

explain the likelihood of technology use after adoption. We define contiguous user

bandwagon and show how this information diffusion mechanism can help in explaining

the time to technology use. We test our hypotheses using data on the adoption and use

of e-procurement technology (n=3158) in the early phase of its diffusion. We find

support for the hypothesis that contiguous user bandwagon is a strong antecedent of

time to technology use.

Key words: technology adoption / technology use divide, contiguous user bandwagon,

information technology adoption and use

3

Closing the Technology Adoption/Use Divide:

The Role of Contiguous User Bandwagon

1. Introduction

New technologies have the potential to trigger many changes: they can change

existing firms’ business models; change the level and characteristics of demand; and

they can greatly affect the competitive positions of different industry players. Indeed,

the introduction of new technologies can sometimes mark the beginning of the end for

established companies that have been successful for decades (Christensen, 1992) and

trigger the emergence of whole new industries with a corresponding wave of new

entrants (Anderson & Tushman, 1990). Existing literature has explained the diffusion of

technology by emphasizing the role of information diffusion (Rogers, 2003; Davis,

1989; Ajzen, 1991; Venkatesh, Speier & Morris, 2002) – i.e. the process by which

information about an innovation is transmitted (Rogers, 2003). According to technology

diffusion theory, mass media communication and “adopter bandwagons” (e.g. Bass,

1969; Abrahamson & Rosenkopf, 1993, 1997) generate self-reinforcing stimuli for the

diffusion of a technology in the market. Mass media communications refers to new

technology appearing in different media while adopter bandwagons refers to previous

adopters of a particular technology playing a role in the information conveyed and

therefore in subsequent adoption patterns.

Despite vast research to date - Rogers (2003) counts more than 5,200 articles on the

topic - a number of issues need to be investigated more carefully if we are to get a better

understanding of technology diffusion. These issues have to do with important

assumptions that are incorporated, explicitly or implicitly, in the existing literature on

4

technology diffusion. We have identified two such areas of concern that we address

more formally in this article.

First, existing technology diffusion literature has, for the most part, equated

technology adoption (i.e. the “purchase” of a technology) to technology use. For

instance, Geroski (2000) states that “[…] early adopting individuals (or firms) have

evidently chosen to use the technology […].” A closer examination of companies

suggests that technology use does not necessarily follow from technology adoption. In

many industries, new technologies are sometimes adopted and then used very little or

not at all. Indeed, this practice is so common in software that the business press has

coined the phrase “shelfware” – software that, once purchased, is put on a shelf and

never used (Economist, 2003). In other words, use after adoption might even be the

exception rather than the rule, at least in some industries and for some technologies. The

scant attention to the differences between technology adoption and technology use

implies that existing technology diffusion literature has also failed to identify and

distinguish the type of organizational actors that are involved in the separate decisions

to adopt and to use a new technology. As we elaborate below, the organizational actors

behind these decisions tend to be different and have different responses to information

diffusion mechanisms.

Second, despite significant advancements in the management literature on the

dynamics of technology, existing technology diffusion theory implicitly assumes that a

technology being launched and adopted in the market emerges at once in its “final

form”. This treatment of technology as an exogenous variable that “appears” in the

market at one point and does not change with time should come as no surprise if we

consider that the theoretical roots of technology diffusion theory are in economics.

5

Traditional economics theory has long been criticized for providing such exogenous

treatment of technology (e.g. Nelson & Winter, 1982), and many of the pioneer authors

in the study of technology diffusion used economic tools and economic perspectives to

build their frameworks. However, modern economics and management theory has

emphasized that technology does evolve over time (Abernathy & Utterback, 1978;

Langlois, 1992; Dasgupta & Stiglitz, 1980; Christensen, 1992; Kahl, 2007). Even after

a new technology is introduced, R&D investment continues and firms keep improving

on the existing technological designs (Anderson & Tushman, 1990; Utterback &

Suarez, 1993). Technologies evolve through a “design hierarchy”, where key decisions

set an improvement trajectory for other less-critical decisions to come (Clark, 1985).

Indeed, the early expressions of a technology in the market are often “rudimentary” and

can differ substantially from the technology form that ends up being adopted by a larger

market in a later stage (Foster, 1986). We argue that by failing to incorporate the notion

that technology evolves over time, existing technology diffusion literature misses an

important fact: the value that organizational decision-makers attach to a specific piece

of information changes over time.

In this paper, we tackle these two areas of concern by bringing in a new construct,

contiguous user bandwagon, that we argue can help explain the time to technology use

after adoption has taken place. Going beyond the traditional treatment of bandwagons in

management literature (e.g. Abrahamson and Rosenkopf, 1993; Fiol & O’Connor,

2003) we define contiguous user bandwagon as the number of new users of the

`technology at the time of adoption by a firm. This construct is an important addition to

the literature for at least four reasons. The first reason stems from the observation that it

formally incorporates into technology diffusion literature the notion that different

6

organizational actors are involved in the decisions to adopt and use a technology –

senior management and the organization’s technical layer, respectively. We expand

upon existing technology diffusion theory (e.g. Roger, 2003) by arguing that the

traditional adopter bandwagon mechanisms (introduced to explain technology adoption)

should be complemented with those related to user bandwagon to explain technology

use. User bandwagons represent information conveyed from actual to potential users of

a technology, and are likely to activate rational-efficiency or fad-based mechanisms

(Abrahamson & Rosenkopf, 1993) that act as powerful enabling factors for an

organization’s decision to use a technology after having adopted it (e.g.: Attewell, 1992;

Geroski, 2000; Burt, 1982; Roger, 2003). We posit that user bandwagons (as opposed to

adopter bandwagons) influence the time elapsed between the adoption of a technology

and its actual use.

The second reason stems from our observation that the value of a given piece of

information decreases with time. This observation implies a non-trivial change because

most of the existing diffusion theory is built upon notions of “cumulative” bandwagon

effects, in other words, effects that start to accumulate from the first day the technology

is in the market up to the time of an agent’s decision. For instance, the notion of

network effects and the resulting “excess inertia” advantages (insurmountable

advantages to those players that have amassed a large installed base of users – see

Farrell & Saloner, 1986) rest precisely in this “timeless” notion of cumulative

bandwagons. In this traditional notion, each adopter, no matter when or where they

adopt, has the same informational value for today’s decision-making agent. However,

recent literature has started to improve upon this notion. For instance, Suarez (2007),

studying the process of mobile telecommunication standards adoption, suggests that not

7

all adopters weigh the same in the decision of an agent; in other words, not all adopters

have the same informational value. Information obsolescence is not exclusively

produced by technological change but is particularly marked in fast-moving industries.

In particular, as technology and the way technology is used change over time (Kahl,

2007), information from periods further back in time may become less relevant for

today’s decisions. It follows that our contiguous user bandwagon construct can speak to

the fact that prospective users are more likely to base their use decision on information

coming from recent users of the technology.

The third reason follows from the observation that by implicitly equating adoption

and use, technology diffusion theory overlooks the fact that the competitive

implications of technology adoption and technology use can be quite different. As noted

above, technology adoption will not improve a firm’s competitiveness unless the

adopted technology ends up being used. Moreover, if technology adoption requires a

large investment that does not finally result in technology use, the effect can even be

negative. For instance, FoxMeyer Corporation is reported to have gone bankrupt due

mainly to a failed and costly ERP implementation (Mabert, Soni, and Venkataramanan,

2001). Therefore, in addition to extending the existing diffusion theory, our focus here

on technology use bandwagons has important implications for organizational

performance. Senior managers often decide on the adoption of new technologies and

can have some degree of influence over technology use decisions via changes in

management practices such as incentives and training. By improving their

understanding of the external information diffusion mechanisms that influence the use

of a technology within their organizations, senior managers can develop even more

effective strategies to foster usage and avoid wasteful spending.

8

Fourth, by explicitly bringing technology use in technology diffusion literature, we

shed additional light on technology diffusion patterns. Only technologies that come to

be used can have a long lasting diffusion. Our theoretical extension here prompts

researchers to jointly consider adoption and usage dynamics if we are to improve on the

overall predictive power of diffusion theory. For instance, our theory can help explain

why a technology, despite enjoying phenomenal initial adoption, may then fail to

achieve wide diffusion if it fails to attract users.

We hypothesize below that the stronger the contiguous user bandwagon, the shorter

the time to technology use, after adoption has occurred. We test our hypotheses in the

context of early e-procurement diffusion, a technology introduced in the mid to late

1990s. E-procurement is the term that describes an information technology that enables

the use of electronic marketplaces in different stages of the purchasing process; from

identification of requirements to payment and contract management. In this paper, e-

procurement technology refers to a pre-packaged standard software product used to

facilitate an organization’s interaction with such electronic marketplaces. We have data

on 3158 firms that adopted e-procurement technology from October 1999 to November

2000 (59 weeks) and we observed them until May 2002. Therefore, the overall

observation period spans over 136 weeks (October 1999 to May 2002). In this time

period, e-procurement technology was still in its infancy and rapidly evolving

(Hoffman, Keedy & Roberts, 2002). For the purpose of our analyses, we say that there

is technology use when any of the functionality embedded in the adopted e-procurement

software package has been used for the first time. We say that there is technology

adoption when a firm purchases the e-procurement software. Data about firms’ adoption

9

and use were obtained from one of the largest European e-procurement providers that

agreed to collaborate with our study. We find that only about 13.5% of the sampled

early e-procurement adopters end up being also early users. This is consistent with the

“shelfware” story above and lends additional support to the main concern of this paper.

We find strong empirical support that contiguous user bandwagon is a powerful

antecedent of time to technology use. We draw managerial and policy implications from

our theoretical propositions and empirical findings.

2. Theory and hypotheses

2.1. Technology diffusion literature

The spread of a new technology in a market or user community is commonly

referred to as diffusion (Cooper & Zmud, 1990; Loch & Huberman, 1999). Technology

diffusion literature has pointed to information diffusion as a key force that enables the

spread of new technologies. In particular, technology diffusion – often measured as

adoption rates or adoption time - has been defined as a function of mass media

communication (Fourt & Woodlock, 1960) and information diffusion (Bass,1969;

Mansfield, 1961). Information diffusion processes can take different forms (Geroski,

2000) including: broadcasting and information provision; epidemics and “word of

mouth” processes; and information cascades.

Building upon research findings in economics, sociology, and cognitive and

behavioral theories, Abrahamson and Rosenkopf (1993) and Fiol and O-Connor (2003)

have introduced the construct of “bandwagons” into technology diffusion theory.

Bandwagons refers to a positive feedback loop in which increases in the number of

adopters create a stronger bandwagon, and a stronger bandwagon, in turn, causes further

10

increases in the number of adopters (Abrahamson & Rosenkopf, 1997). Bandwagon

influenced behaviors have been described as ranging from rational behavior based on

positive externalities (e.g. Katz & Shapiro, 1985) to behavior to conform with the sheer

number of organizations that have already adopted the technology (e.g. Abrahamson &

Rosenkopf, 1993; Tolbert & Zucker, 1983; Strang & Macy, 2001; Meyer & Rowan,

1977). Many researchers have used bandwagon theories to explain technology diffusion

for different technologies, industries and geographical zones (Gurbaxani, 1990; Kumar

& Kumar, 1992).

Existing diffusion literature has also pointed to two other clusters of variables to

explain technology diffusion patterns. First, research has shown that the diffusion of a

specific technology depends on some features of the technology itself, such as its

relative advantage vis a vis existing technologies (Loch & Huberman, 1999), the

technology’s compatibility with existing products (Farrell & Saloner, 1985; Katz &

Shapiro, 1994) and complementary technological infrastructures (Katz & Shapiro, 1994;

Shy, 2001); the technology’s complexity and its ability to be trialed and observed

(Rogers, 2003), and on whether the new technology is a product or a process technology

(Bass, 1969; Cabral & Leiblein, 2001). Second, existing literature has pointed to firm

resources and capabilities to explain technology diffusion, but their precise role remains

a subject of controversy. Some studies have postulated a positive relationship with

technology adoption timing (i.e. the higher the level of a firm’s resources, the later a

firm adopts a new technology), while others have proposed a negative relationship. The

argument for a positive relationship suggests that firms with high level of resources

emphasize formal roles and control systems and tend to become more rigid.

Bureaucracy research (Blau, 1970) and organizational ecology studies (Hannan &

11

Freeman, 1989) concur, indicating that the level of firm resources, often operationalized

as firm’s size, is related to higher organizational inertia, higher formalization and

standardization, and structural rigidity. Christensen, Anthony and Roth (2004) suggest

that, as they grow, organizations increasingly rely on processes that over time become

embedded in hard-to-change organizational routines and values. These conditions

prevent large, well-endowed firms from being early adopters of technology. The

argument for a negative relationship suggests that resource-rich organizations are more

likely to be early adopters of technology because of slack resources (Nohria & Gulati,

1996), formal innovation management practices (Van der Ven, 1988), or because their

resources translate in higher absorptive capacity (Cohen & Levinthal, 1990).

2.2. Extending the theory of technology diffusion

One major area of concern in existing technology diffusion literature is that it has

tended to equate technology use with technology adoption when in fact these are two

distinct phenomena that respond to different dynamics. We argue that extending

technology diffusion theory to account for the distinction between adoption and use is

not straightforward as each decision involves different organizational actors. These

differences may be large enough to grant a more careful theoretical and empirical

treatment. Indeed, it is through using a new technology that a firm can trigger changes

in its existing business models, value chain, and inter-firm relationships. We therefore

argue that technology diffusion theory can increase its value and predictive power if we

widen its scope to encompass the antecedents of technology use. Technology diffusion

and technology use literatures have evolved by and large independently and, although

12

both can be said to rely on information diffusion, the specific mechanisms by which

each of them occur are quite different.

2.2.1. Contiguous User Bandwagon and the Time to Technology Use

In existing diffusion theory, information “bandwagons” typically refer to the

number of adopters without consideration to how many of those adopters come to

actually use the technology. For instance, Katz and Shapiro (1986) and Shy (2002)

outline that the extent of adoption externalities depends on the expected final network

sizes - a phenomenon that is often also called “adopter bandwagon” (Abrahamson &

Rosenkopf, 1993). However, as noted in the introduction, within an organization, the

decisions to adopt or use a technology are typically made by different stakeholders. For

instance, Leonard-Barton and Deschamps (1988), in their study of the introduction of a

new software package (an “expert system”), make a clear distinctions between the “top

management ‘authority decision’ to adopt the innovation” and the “target end-user’s

adoption decisions” (p. 1253). Similarly, after surveying many cases of ERP

implementation, Mabert et al. (2001) conclude that the “Adoption of ERP was generally

a top-down decision” (p. 75). Senior managers tend to be responsible for the decision to

adopt a new technology because adoption requires the approval of significant capital

expenditures (e.g. ERP, CRM) and sometimes even a strategy change that can only be

endorsed by an organization’s senior level. Thus, their decisions to adopt a new

technology tend to be influenced by adopter bandwagons and are typically based on

rational efficiency (Westphal, Gulati, & Shortell 1997), the “symbolic value” of the new

technology (DiMaggio & Powell, 1983; Meyer & Rowan, 1977), “managerial

13

improvisation” (Orlikowski, 1996), call options (Miller & Folta, 2002), or simply "me

too" behavior (Strang & Macy, 2001).

However, when it comes to using a new technology, the key actors are not senior

managers but the “technical layer” of the organization (Bacharach, Bamberger, &

Sonenstouhl, 1996) – i.e. those who are asked to replace an existing technology and

associated processes with a new technology, processes, and routines. More often that

not, there is a marked disconnect between the senior management’s cognition and the

technical layers’ beliefs as to the real need to adopt the new technology, and the

associated costs and benefits of using it. In his study of how organizations adopt total

quality management, Zbaracki (1998: 613) notes that “after leaders decide to implement

TQM, they pass it to other members in the organization”. He then describes the

frustration of the “other members” when they try to integrate TQM into their daily

routines. Similarly, Leonard-Barton and Deschamps (1988) note that the use of a

technological innovation is a process that involves “numerous individual ‘secondary’

adoption decisions by target users even after successive layers of management have

passed along the ‘authority decision’." In spite of specific incentives or practices that

senior management may put in place to promote the use of a new technology, existing

literature suggests that the technical level in charge of putting the new technology to use

tends to be cautious and conservative and often does not respond as planned to

traditional incentives (e.g. Leonard Barton, 1992; Edmondson, Pisano and Bohmer,

2001; Bresnahan, Brynjolfsson, & Hitt, 2002). There is a powerful reason for the

technical layer’s reluctance: new technologies disrupt existing roles and routines and are

surrounded by high uncertainty, and therefore tend to entail high costs for the technical

layer (Black, Carlile, & Repenning, 2004). It follows from our argument above that the

14

technical layer of the organization tends to be very careful when it comes to making a

decision whether to undertake the significant cognitive and organizational costs

necessary to use a new technology.

Adopter bandwagons can point to the existence of a new technology and the fact

that other companies have purchased it, but they fail to transmit relevant information

regarding the use of the technology (Attewel, 1992). Prospective users in the technical

layer of the organization look for information that can speak of the “expected personal

outcomes of adopting the innovation… ‘What will its advantages and disadvantages be

in my situation?’ ‘How complex will the innovation be for me to use?’” (Agarwal &

Prasad, 1998). We argue that users place great value in information coming from other

users of the same technology when it comes to assessing the uncertain costs and

benefits of using a technology. For instance, prospective users (the technical layer) of

complex technologies such as ERP often require vendors to arrange for visits to existing

ERP deployments in order to talk directly with other users about their implementation

experience. These arguments can also be explained with structural equivalence theory

(Burt, 1982; Hedstrom & Swedberg, 1998) that predicts actors’ behavior based on the

set of “linkages”, not necessarily ties, existing among actors. Burt (1982) argues: “two

people identically positioned in the flow of influential communication will use each

other as a frame of reference for subjective judgments and so make similar judgments

even if they have no direct communication with each other" (1982: p. 1293).

We argue that the decision to use a new technology will be affected by a different

kind of bandwagon which we call user bandwagon – i.e. the number of technology

users.

15

As shown above, a second area of concern in existing diffusion literature is the

treatment of technology itself. Contrary to what existing diffusion theory implies, a new

technology does not remain unchanged once it appears in the market, nor is it used

always in the same way. Management scholars have long studied the evolution of

technology, with particular emphasis on the implications for firm survival and

performance. Abernathy and Utterback (1978) characterized the early phase of

technology as “fluid”, a phase marked by a high rate of innovation in product features

and architecture. Anderson and Tushman (1990) conveyed the same idea when

describing the early phase of technology diffusion as an era of “ferment”. During the era

of ferment, technology evolves constantly, and technological change adds to the

uncertainty that surrounds technology-related decisions by organizations. This

uncertainty makes it especially hard for users to commit to technology-specific learning

(Schmalensee, 1982; Carpenter & Nakamoto, 1989). As time goes by and technology

evolves, new information becomes available which reduces the level of uncertainty

surrounding the new technology – e.g. particular variations of the new technology may

be selected out of the market or new, improved technologies may be introduced. Not

only technology changes over time, but also the way that people use it. Kahl (2007)

argues that use is a learning mechanism that can generate knowledge to reinforce a

particular use as well as generate knowledge about different uses. In this context,

information that reduces uncertainty is valued highly, particularly by the technical layer

that has to commit significant personal resources to use the new technology effectively.

We argue that, during the diffusion of a new technology, potential users tend to

place greater value on new bandwagons, that is, bandwagons formed in the time periods

just prior to the adoption decision. We call these bandwagons contiguous bandwagons.

16

We argue that prospective users will be positively influenced by contiguous

bandwagons because these may signal new, up-to-date information and convey less

uncertainty about the technology’s true potential benefits.

Our focus on contiguous bandwagon expands upon and complements the

conventional focus on the cumulative level of adopter bandwagon that has traditionally

been associated with technology diffusion (e.g. Abrahmson and Rosenkop, 1993). By

considering the cumulative level of adopters from all past periods, existing literature

implicitly assumes that information from different periods has the same importance for

the organizational actor making a decision about a new technology. We argue that this

is not the case.

Combining user bandwagon and contiguous bandwagon, we define the construct of

contiguous user bandwagon as the number of new users of the technology at the time of

a firm’s technology adoption. As elaborated above, contiguous user bandwagon should

act as powerful antecedent of time to technology use because: (a) the information relates

to users, not adopters, of the technology (as we discussed in the previous paragraph) and

it is therefore considered more relevant and reliable by prospective users; (b) the

information is “new”, that is, relates to users realized during the time period of a firm’s

technology adoption decision. Hypothesis 1 follows:

Hypothesis 1. The time between technology adoption and technology use

(time to technology use) will be inversely related to contiguous user

bandwagon.

17

From our arguments above, it follows that a prospective user’s behavior may also be

positively influenced by specific contiguous user bandwagons that relate to users that

share similar “structural” characteristics with the prospective user (e.g. Lazarsfeld &

Merton, 1964; Burt, 1982). Abrahamson and Rosenkopf (1993: p. 493) argue that

economic actors should not be thought as independent agents; rather, they should be

grouped into “collectivities”, groups of agents where information diffuses more easily.

Several empirical studies have provided evidence that structurally equivalent

collectivities can predict behavior (e.g. Burt, 1987; Harkola and Greve, 1995). There

are several ways of grouping organizations into collectivities. One obvious grouping is

by geographical proximity. The studies pioneered by Hagerstand (1952) show that

“spatial interactions” generally have a positive impact on information diffusion. The

basic idea here is that information is more easily transferred and runs a lower risk of

integrity loss when the agents are geographically closer. This is particularly important

for complex information such as that derived from the use of a new technology. We

define contiguous user bandwagon by location as contiguous bandwagons generated by

new users located in the same geography of a prospective user. We posit,

Hypothesis 1.a. The time between technology adoption and technology use

(time to technology use) will be inversely related to contiguous user

bandwagon by location.

Another common way of grouping organizations is by industries or groups of

organizations that produce close substitutes (Porter, 1980). We define contiguous user

bandwagon by industry as contiguous bandwagons generated by new users of the

technology operating in the same industry of a prospective user. We posit,

18

Hypothesis 1.b. The time between technology adoption and technology use

(time to technology use) will be inversely related to contiguous user

bandwagon by industry.

Basic firm descriptive characteristics are another commonly used way to group

organizations based on “structural equivalence” (Abrahamson and Rosenkopf, 1993).

For instance, international banks’ behavior is more likely to be affected by actions taken

by other similar international banks than by local banks' actions. A company’s legal

status is another important descriptive firm characteristic which is often used to

categorize firms. It follows from our arguments that, for instance, prospective users

from public firms will be more likely to be influenced by bandwagons created by actual

users in other public firms. Public firms indeed face different regulatory, shareholder

and tax environments than private firms. We define contiguous user bandwagon by

legal status as contiguous user bandwagons generated by actual technology users that

have the same legal status as a prospective user. We posit,

Hypothesis 1.c. The time between technology adoption and technology use

(time to technology use) will be inversely related to the value of contiguous

user bandwagon by legal status.

3. Methods

3.1. Sample and technology context

We test our hypotheses in the context of e-procurement technology at the early stage

of its diffusion. E-procurement describes the use of electronic marketplaces in every

stage of the purchasing process; from identification of requirements through payment,

19

and contract management. The e-procurement technology considered in this study was

sold in standard packages and did not require major customization. The technology

vendor did not charge an upfront fee but only a fixed fee per completed transaction on

actual use of the technology that did not vary across firms. We can then reasonably

assume that there were no asymmetric disincentives for organizations to experiment

with the technology. In addition, interviews conducted with the technology vendor and

some adopters suggest that technology implementation time was fairly constant and

independent from firm characteristics such as size or industry. We also asked about

channels by which users collected information about other users of the technology. In

addition to traditional channels – e.g. media and interpersonal communication, where

applicable - it also emerged that the e-procurement technology provided users and

potential users with access to an electronic database (marketplace) which published

real-time information on companies adopting and using the technology. Prospective

users had full access to these data and could set up their own search criteria – e.g.

location, industry or legal status.

Data for our analysis were obtained from one of the largest European e-procurement

providers that agreed to share with us its database on adoption and use. We sampled

3158 firms that adopted the e-procurement technology from October 1999 to November

2000 (59 weeks) and we observed them until May 2002. Therefore, the overall

observation period spans over 136 weeks, from October 1999 to May 2002. For each

firm adopting the e-procurement technology, the dataset provides some firm-specific

data – e.g. firm’s location and firm’s SIC - and tracks the timing of firms’ activity with

the technology – e.g. time of purchase and time of use. Our sample exhibits right

20

censoring: in week 136 when we stopped observing our firms, 86.44% of the firms that

had subscribed to this technology had not yet used it.

This dataset has at least three key strengths. First, data are structured according to a

uniform reporting criterion and, therefore, they are easily comparable. Second, there is

no response bias because the data capture hard variables that were automatically

recorded on the technology vendor database. Third, having sampled firms that have

adopted the same e-procurement technology, the sample can be considered self-

controlled for differences across technologies.

3.2. Measures

Dependent variable

Time between Technology Adoption and Technology Use (Time to Technology Use).

Time to technology use is a positive variable calculated as the time elapsed between

adoption (Week of Adoption in Table 1) and first use (Week of First Use in Table 1) or

to the end of the observation period if the firm did not use the technology. We say that

there is technology use when any of the functionality embedded in the adopted e-

procurement software package has been used for the first time. We say that there is

technology adoption when a firm purchases the e-procurement software.

Independent and control variables

Contiguous User Bandwagon. Contiguous user bandwagon is operationalized by

computing the number of firms that use the e-procurement technology for the first time

the week that a prospective user purchases the technology. Building on this definition,

we define Contiguous User Bandwagon by Industry as the number of firms in the same

21

industry of a prospective user that use the e-procurement technology for the first time

the week that the prospective user purchases the technology (industry defined at the first

digit of the users’ SIC). We define Contiguous User Bandwagon by Location as the

number of firms in the same geographical location of the prospective user that use the e-

procurement technology for the first time the week that the prospective user purchases

the technology (geographical location defined at the province level). We define

Contiguous User Bandwagon by Firm’s Legal Status as the number of firms of same

legal status as the prospective user that use the e-procurement technology for the first

time the week that the prospective user purchases the technology (see below for a

definition of firm’s legal status).

Contiguous Mass Media Communication. As reviewed earlier, mass media

communication is a main diffusion mechanism, and is considered a rapid and efficient

means of informing companies about the existence of new technologies (Fourt &

Woodlock, 1960; Mahajan, Muller, & Bass, 1990). We define the variable Contiguous

Mass Media Communication and compute it as follows. For each period considered in

our analysis, we carried out a search in the FACTIVA database counting the number of

occurrences of the keywords “e-procurement” and “e-marketplace”. These keywords

well describe the technology under consideration. The FACTIVA database searches

more than 9,000 sources - including the Wall Street Journal and the Financial Times

and has often been used by other researchers for searches on business-oriented media.

Following our reasoning above, Contiguous Mass Media Communication is defined as

the number of occurrences in the business press of the above keywords for each week

included in the analysis.

22

Contiguous Adopter Bandwagon. Adopter Bandwagon is another diffusion

mechanism considered in technology diffusion literature. We compute Contiguous

Adopter Bandwagon in any given week as the number of new firms that purchase the e-

procurement technology in that week.

Cumulative Adopter (User) Bandwagon. We compute cumulative adopter (user)

bandwagon as the total number of technology adopters (users) from the introduction of

the technology. It is important to note that contiguous user bandwagon is highly

correlated with both cumulative adopter bandwagon and cumulative user bandwagon.

However, as we have elaborated extensively above, contiguous and cumulative

bandwagons are different constructs. In order to include both contiguous and

cumulative variables in our models and avoid biased estimations that arise from multi-

collinearity we used a generated new regressor (Wooldridge, 2002: 115-116), namely,

the predicted residuals of a regression of cumulative adopter bandwagon against

contiguous user bandwagon. This involves formalizing the relationship to model the

overlap between the two measures and using the residuals as predictors. In this way, we

can separate out the cumulative adopter effect from the contiguous user effect. This is a

standard technique when correcting for multi-collinearity (Kennedy, 1992: p. 210-211).

Industry. We created dummy variables to capture differences across industries. The

three main industries are Manufacturing (40.7% of the sample), Wholesale (13.6%) and

Services (27.5%).

Censor. We captured right censoring by using a dummy variable that takes the value

of 1 if a firm has used the new technology, and 0 otherwise.

Legal Status. We created dummy variables representing four types of firms’ legal

status: Limited Liability company, Corporation, Partnership, and Personal Firm.

23

Italy. A significant percentage of the sampled firms are based in Italy and thus a

dummy variable captures this country effect.

Time. We created dummy variables for each time period included in the analyses, in

order to control for time-related effects; including, for instance, cumulative adopter

bandwagons, cumulative user bandwagons, marketing campaigns, and stage of

adoption.

Table 1 reports descriptive statistics for our variables and Table 2 provides a

correlation matrix.

------------------------------------------------

Please insert Table 1 and Table 2 about here

------------------------------------------------

4. Models and results

To test our hypotheses on the role of Contiguous User Bandwagon in the time to

technology use, we fitted survival models (Cleves, Gould, & Gutierrez, 2002; Hoesmer

& Lemeshow, 1999) on our data. To draw preliminary insights on time to technology

use, we first ran a non-parametric estimation. We estimate the hazard function without

covariates, that is, the probability of a firm using the adopted technology within a short

time interval, conditional on not having used the technology up to the starting time of

the interval. The survivor function is therefore:

24

------------------------------------------------

Please insert Figure 1 about here

------------------------------------------------

Figure 1 shows that the hazard is not constant and displays negative duration

dependence, dλ(t)/dt < 0. Given the shape of the non parametric hazard function, we can

estimate our model using the accelerated failure time form1 of logistic distribution.

Using this distribution, the hazard function is given by:

With being the hazard function and the shape parameter. We introduce

covariates by defining as a function of a set of regressors:

This model allows us to estimate the effect of each explanatory variable on

duration. If a coefficient displays a negative sign, it implies that the variable decreases

time to technology use (increases the probability of earlier use). No left censoring is

present in our database given that, for each firm, the exact time of adoption is known.

Models 1 throughout 7 in Table 3 show the results of our estimations.

------------------------------------------------

Please insert Table 3 about here

------------------------------------------------

1 An accelerated failure time form is characterized by its conditional survival function S(t|Z=z) for a duration T, with . is the baseline survivor function, and if it is specified parametrically, we get a parametric model (as in our case). The model used depends on how we define

(we used a log-logistic model here).

25

Model 1 in Table 3 is a baseline model containing our control variables, contiguous

mass media communication and contiguous adopter bandwagon. In Model 2, our key

construct, contiguous user bandwagon, is entered. In Model 3, we add cumulative

adopter bandwagon. Models 4 through 6 include additional explanatory variables in

steps: contiguous user bandwagon by industry, contiguous user bandwagon by location,

and contiguous user bandwagon by legal status. Model 7 contains all variables of

interest.

All Models are highly significant as shown by their log likelihood (p<0.001). Using

Model 1 as a baseline, all additional variables in models 2 through 6 are significant

when taken together, according to the likelihood-ratio tests reported in the Appendix2.

Contiguous adopter bandwagon is significant in all models (p <0.001). In contrast,

contiguous mass media communication is significant only in the first two models. The

coefficient of contiguous adopter bandwagon is positive, suggesting that higher level of

contiguous adopter bandwagon increases the time to technology use. The coefficient of

cumulative adopter bandwagon (residuals) is significant (p <0.001) and negative in all

models. Model 2 shows that contiguous user bandwagon is significant (p<.001) and its

coefficient, as expected, is negative3. It follows that contiguous user bandwagon has a

negative impact on the time to technology use – i.e. it shortens the gap between

technology adoption and use.

2 Likelihood-ratio tests whether the parameters used in a given model are significant by comparing the likelihood of this model with the likelihood of a model without parameters using:

under with Q the number of restrictions.

26

In Model 3, and hereafter, we add the residuals of cumulative adopter bandwagon

against contiguous user bandwagon. Contiguous user bandwagon is still highly

significant (p<.001) and negative. Hypothesis 1 is therefore supported.

We test the effect of contiguous user bandwagons by industry, location, and legal

status in Models 4 to 6, respectively4. As expected, the coefficients of all contiguous

bandwagon variables are negative and significant (p<0.001). Hypotheses 1a, 1b and 1c

are therefore supported. Model 7 is a full model where all variables of interest are

entered5. In this Model, the coefficients of all contiguous bandwagon variables have the

expected negative sign. Contiguous user bandwagon and contiguous user bandwagon by

legal status are significant (p<.05). This suggests that the effect contiguous user

bandwagon remain significant even after the other contiguous bandwagon measures are

entered in the model.

We performed several other analyses to check for the robustness of our results, not

included in this paper. We replicated the analyses above by testing contiguous user

bandwagon against cumulative user bandwagon (using the residual approach noted

above for cumulative adopters) and our key results were confirmed. We then estimated

our models by entering time dummies for all periods. Time dummies provide an

alternative way to control for cumulative adopter bandwagon, cumulative usage

bandwagon and other time-related trends – e.g. the effect of marketing campaigns.

Model estimations show that contiguous user bandwagon is a significant predictor

(p<0.001) of time to technology use even after including time dummies.

4 In each of these models, the cumulative adopter bandwagon variable expresses the residuals of cumulative adopter bandwagon against contiguous user bandwagons by industry, contiguous user bandwagons by location, and contiguous user bandwagons by legal status, respectively. 5 In this Model, cumulative adopter bandwagon expresses the residuals of cumulative adopter bandwagon against contiguous user bandwagon.

27

Finally, we ran models with contiguous user bandwagons that incorporate new users

from periods further back in time – i.e. not only new users at a firm’s time of

technology adoption. In particular, we ran Model 3 with several redefined contiguous

user bandwagon measures that incorporated users from two periods, three periods and

four periods back, respectively. Model estimations show that the coefficients of the

lagged contiguous user bandwagons rapidly decrease in their magnitude as the

information refers to periods further back (the variable still retains significance and its

negative sign). |This test provides some support to an important claim in our paper that

relates to our “contiguous” measures, i.e. that prospective users tend to discount the

value of information coming from periods further back.

5. Discussion and final remarks

Technology use is an important topic to be investigated; after all, a new technology

can only have an impact on firms and industries if it is used and, as stated above,

technology use does not necessarily follow adoption. This paper advances existing

literature on technology diffusion by proposing a new construct, contiguous user

bandwagon, and showing theoretically and empirically how this construct can help

explain the time to technology use6. To address this issue, our proposed new construct

directly addresses two assumptions that existing literature has often made, if only

implicitly: (a) that the antecedents of technology adoption and technology use are the

same; (b) that organizationl actors value all information much in the same way,

irrespective of their sources and “newness” (time period the information comes from) .

6 As noted by an anonymous reviewer, our theoretical contribution can be classified as “invention by extension” (Dubin, 1978).

28

We have shown that by incorporating contiguous user bandwagon in technology

diffusion theory – that is, when used to complement cumulative constructs that have

been the focus of this theory to date -- we can develop a more comprehensive and useful

theory of how technology spreads and influences firms and industries.. One of the main

reasons underpinning the technology adoption / technology use divide is that different

organizational actors are responsible for making the adoption decision and for using and

implementing the new technology -- senior management and the firm’s technical layer,

respectively. These actors not only are different but they respond differently to

information stimuli. We argue that adopter bandwagons (and the buzz and hype that

typically surrounds them) tend to exert a larger influence on senior management than

they do on the technical layer of the organization (Abrahamson & Rosenkopf, 1993).

Drawing from different theoretical perspectives, we argue that the technical layer tends

to have a more conservative approach when it comes to technology given the fact that,

when a new technology is adopted, they have to go through a painful process of change

in routines, processes and cognitive maps. Ultimately, only technologies that are used

can enjoy long lasting diffusion and our theory explicitly urges researchers (and

practitioners) to jointly consider adoption and usage to gain a better understanding of

long term technology diffusion patterns. We have argued that decision-makers may tend

to heavily discount information from earlier periods, a feature that existing literature on

technology diffusion has largely overlooked. This consideration led us to move beyond

the traditional conceptualization of bandwagons. Our contiguous user bandwagon

construct departs from conventional wisdom in diffusion theory and bandwagon studies

that define and operationalize bandwagons as the cumulative level of adopters – that is,

information coming from all time periods since the launching of a technology.

29

The empirical results presented here provide strong support for the hypothesis that

contiguous user bandwagon is an important antecedent of time to technology use.

Furthermore, we find that contiguous user bandwagon by location is also a significant

antecedent of time to technology use. Our models also include other contiguous

bandwagon control variables. Particularly interesting are our results for contiguous

adopter bandwagons. This variable shows a significant effect on time to technology use

yet its net effect is positive, increasing the time to use. This result may apparently look

surprising. Yet, our arguments above made it clear that adopters have different

motivations and costs when compared to users; thus, it should come as no big surprise

that users may actually react with some skepticism to bandwagons triggered by

adopters. For instance, users may perceive management’s decision to adopt as a “fad” or

simply a “me too” behavior (Strang & Macy, 2001), and resist the use of the new

technology.

Contiguous mass media communication does not reach significance in most of our

models. It is interesting to consider that, in the time span considered in our analysis,

mass media communication created very high expectations regarding the advantages

and promises of e-procurement and other Internet-based technologies. Yet, this

bandwagon may have affected the behavior of adopters but not that of users. As we

argued earlier, users are less likely to be influenced by business media communication

than adopters 7. Moreover, our result here further suggests that users do not respond to

the same stimuli than adopters. Overall, these results provide further support to one of

the key ideas of the paper – i.e. the need to differentiate between antecedents of

adoption and use.

7 We thank an anonymous reviewer for pointing out this potential explanation.

30

Finally, our analyses confirm that our theory does add predictive power to the extant

bandwagon literature by showing that contiguous user bandwagon does have a

significant (negative) impact on time to use even after controlling for the cumulative

level of adopters (or users). Indeed, our theory can help not only predict time to use

but also improve our predictive power in terms of the overall technology adoption and

diffusion patterns of a given technology. Only technologies that are used can have a

long lasting diffusion; our theory prompts research and practitioners to consider both

aspects (adoption and use) of the overall diffusion process.

5.1 Avenues for future research

There are some limitations that apply to this paper that open interesting avenues for

further research. We have conducted our study in a specific context (e-procurement

technology) which can be considered a process technology. Although we believe that

our propositions are general enough to be applicable to other contexts, this should be

done with the usual caveats. Further studies could focus on the antecedents of

technology use in other phases of the technology diffusion cycle (we have focused on

the early diffusion phase), or replicate our study with other technologies to provide

interesting comparisons. Also, our measure of technology use itself may be improved

upon. In this paper, we have considered use as a discrete event: a firm either uses or

does not use the technology during the time of analysis. We measured this by looking

at the time of first use of the e-procurement system. However, it could be argued that

there are different degrees and forms of use, and a future study could try to provide that

extra granularity in the analysis – e.g. by capturing the persistence or effectiveness of

use by different users. Last, building upon existing technology diffusion literature, we

31

assume that prospective users, like adopters, are influenced by mass media

communication. This aspect should be explored more, both theoretically and

empirically.

5.2 Managerial and policy implications

The results of our study suggest that technology producers and technology vendors

should pay attention to contiguous bandwagons, particularly in what relates to

“managing” the use characteristics of their new technology so as to enable contiguous

user bandwagons. For instance, technology vendors could create strategic action plans

to increase user bandwagons. To some degree, this is already happening in sectors such

as software. Nearly all software product companies have set up user and developer

groups, designed to diffuse technology information to current and future users.

Our results point to contiguous user bandwagon as an important mechanism for the

long-term success of the new technology. A “boom” in early sales of a new technology

might be followed by a sudden drop if the technology is not accepted by its potential

users. Technology producers and vendors should be aware that the timing of introducing

a new technology is a key variable and should critically consider the implications of

rushing a new technology to the market if that can have negative implications for user

acceptability. For senior managers making technology adoption decisions, our research

flags a warning to the risk of “shelfware” that can result from choices that do not take

into account the antecedents of technology use. Decision-making managers should

strive to have a better understanding of users within their own organizations and the

implementation risks and obstacles associated with new technologies, and should

facilitate the communication between prospective uses in their organization with actual

32

users of a technology in other organizations. In this light, Internet-based social networks

– e.g. blogs, Web 2.0, or more recently Twitter - could provide an important platform to

initiate and trigger these user bandwagons8.

Our results also have implications for policy makers. When it comes to new

technologies, policy makers often want to find ways of accelerating diffusion, but

diffusion of a technology is not successful unless the technology is both purchased and

used. Policy makers should increase their awareness of the differences between adopters

and users of new technologies, and correspondingly re-tool their instruments and

policies used to support effective technology diffusion. For instance, policy makers

could place greater emphasis on policies that foster and promote technology use and not

just on technology adoption. As the technical layer of organizations is crucial in

technology use decisions, policies designed to ease the transition of these organizational

actors from the old to the new technology could include training, user workshops and

other policies destined to promote user awareness and exchange. Technology use, and

not technology adoption, should be the final aim of policy makers.

8 We thank an anonymous reviewer for raising this implication.

33

References

Abernathy, W. J., & Utterback, J. M., 1978. Patterns of industrial innovation.

Technology Review, 80, 97–107.

Abrahamson, R., & Rosenkopf, L., 1993. Institutional and competitive bandwagons:

using mathematical modeling as a tool to explore innovation diffusion. Academy of

Management Review, 18: 487-517.

Abrahamson, R., & Rosenkopf, L., 1997. Social Network Effects on the Extent of

Innovation Diffusion: A Computer Simulation. Organization Science, 8: 289-309.

Agarwal, R. & Prasad, J. 1998. The antecedents and consequents of user perceptions in

information technology adoption, Decision Support System, 22(1)

Ajzen I., 1991. The theory of Planned Behavior. Organizational Behavior and Human

Decision Processes, 50: 179-211.

Anderson, P., & Tushman, M.L., 1990. Technological discontinuities and dominant

design: a cyclical model of technological change. Administrative Science Quarterly,

35: 604 – 633.

Attewell, P., 1992. Technology diffusion and organizational learning: the case of

business computing. Organization Science, 3: 1 – 19.

Bacharach, S., & Bamberger, P., Sonenstouhl, W., 1996. The organizational

transformation process: The micropolitics of dissonance reduction and the alignment of

logics of action. Administrative Science Quarterly, 41: 477-506.

Bass, F.M., 1969. A new product growth model for consumer durables. Management

Science, 15: 215-227.

Beaudry A., & Pinsoneault, A., 2005. Understanding user responses to Information

Technology: a copy model for user adaptation. MIS Quarterly, 29: 493 -524.

34

Black, L., Carlile, P., & Repenning, N. Formalizing Theoretical Insights from

Ethnographic Evidence: Revisiting Barley’s Study of CT-Scanning Implementations.

Administrative Science Quarterly, forthcoming..

Blau, P. M., 1970. A Formal Theory of Differentiation in Organizations. American

Sociological Review, 35: 201-218.

Bresnahan, T. E., Brynjolfsson, E., & Hitt, L., 2002. Information Technology,

Workplace Organization and the Demand for Skilled Labor: Firm-level Evidence.

Quarterly Journal of Economics: 117: 339-376.

Burt, R., 1987. Social contagion and innovation: cohesion versus structural equivalence.

American Journal of Sociology, 92: 1287 – 1335.

Cabral, D., & Leiblein, M., 2001. Adoption of a Process Innovation with Learning-by-

Doing: Evidence from the Semiconductor Industry. Journal of Industrial Economics,

49.

Carpenter, G. S., & Nakamoto, K., 1989. Consumer preference formation and

pioneering advantage. Journal of Marketing Research, 26: 285-298.

Christensen, C. M., 1992. Exploring the limits of the Technology S-curve, Part I:

Component Technologies. Product and Operations Management Journal, 1: 334-357.

Christensen, C. M., Roth, E. A. & Antony, S. D. 2004. Seeing what’s next. Boston,

MA: Harvard Business School Press.

Clark, K., 1985. The interaction of design hierarchies and market concepts in

technological evolution. Research Policy, 14: 235-251.

Cleves, M., Gould, W.M., & Gutierrez, R.G., 2002. An Introduction to Survival

Analysis using Stata. College Station, TX: Stata Press.

35

Cohen, W. M., & Levinthal, D.A., 1990. Absorptive capacity: a new perspective on

learning and innovation. Administrative Science Quarterly, 35: 128 – 152.

Cooper, R.B., & Zmud, R.W., 1990. Information Technology Implementation Research:

A Technological Diffusion Approach. Management Science: 36: 123 – 139.

Czepiel, J.A., 1975. Patterns of Interorganizational Communications and the Diffusion

of a Major Technological Innovation in a Competitive Industrial Community. Academy

of Management Journal, 18: 6-24.

Dasgupta, P., & Stiglitz, J., 1980. Uncertainty, Industrial Structure and the Speed of

Rand. Bell Journal of Economics, 11: 1-28.

Davis F.D., 1989. Perceived Usefulness, Perceived Ease of Use, and user Acceptance of

Information Technology. MIS Quarterly, 13: 319 -340.

Davis, F.D. Bagozzi, P., & Warshaw, P.R., 1989. User Acceptance of Computer

Technology: A Comparison of Two Models. Management Science, 38: 475 – 487.

Desanctis, G., & Poole, M. S., 1994. Capturing the complexity in advanced technology

use: adaptive structuration theory. Organization Science, 5: 121-147.

DiMaggio, P.J., & Powell, W.W., 1983. The iron cage revisited: institutional

isomorphism and collective rationality in organizational fields. American Sociological

Review, 48: 147 – 160.

Dubin, R. 1978. Theory Building. New Yoork: Oxford University Press.

Duplaga, E. A.; Astani, M. 2003. Implementing ERP in Manufacturing. Information

Systems Management, 20(3).

Economist (The). 2003. At your service. 367: 8323.

36

Edmondson, A. C., Bohmer, R., & Pisano, G. P., 2001. Disrupted Routines: Team

Learning and New Technology Adaptation. Administrative Science Quarterly, 46: 685-

716.

Farrell, J., & Saloner, G., 1985. Standardization, compatibility, and innovation. Rand

Journal of Economics, 16: 70-83.

Foster, R., 1986. Innovation: the attacker’s advantage. New York, NY: Summit

Books.

Fourt, L. A., & Woodlock, J.W., 1960. Early Prediction of Market Success for New

Grocery Products. Journal of Marketing, 25: 31-38.

Geroski, P. A., 2000. Models of technology diffusion. Research Policy, 29: 603-625.

Gurbaxani,V., 1990. Diffusion of computing networks: The case of Bitnet.

Communications of the ACM, 33: 65-75.

Hagerstand, T, 1952. The propagation of Innovation Waves. Lund, Sweden: Lund

Studies in Geography 4.

Hannan, M., & Freeman, J., 1989. Organizational ecology. Cambridge, MA: Harvard

University Press, 1989

Hedstrom, P & Swedberg, R. 1998. Social Mechanisms. Cambridge, UK: Cambridge

University Press.

Hoesmer, D., & Lemeshow, S., 1990. Applied Survival Analysis. New York: Wiley,

1990.

Hoffman, W., Keedy, J., & Roberts, K., 2002. The unexpected return of B2B.

McKinsey Quarterly, 3.

Huber, G.P. 1991. Organizational Learning: The Contributing processes and the

Literature. Organization Science, 2: 88-115.

37

Kahl, S., 2007. Utilizing Use: The Potential Impact of Customer Use Patterns on

Technology and Industry Evolution. AOM Best Paper Proceedings.

Katz, M., & Shapiro, C., 1986. Technology adoption in the presence of network

externalities. Journal of Political Economy, 94: 822-841.

Kenney, P. 1992. A guide to Econometrics. MIT Press, Cambridge, MA.

Kumar, U., & Kumar V., 1992. Technological innovation diffusion: The proliferation of

substitution models and easing the user’s dilemma. IEEE Transactions on Engineering

Management, 39: 158 – 166.

Langlois, R. N., 1992. External Economies and Economic Progress: The Case of the

Microcomputer Industry. Business History Review, 66: 1-50.

Lazarus, R.S., 2000. Towards better research on stress and copy. American

Psychologist 55: 665-673.

Lazarsfeld, P. F. & R. K. Merton 1964. Friendship as Social Process: A Substantive and

Methodological Analysis. In Monroe Berger et al. Freedom and Control in Modern

Society, New York: Octagon.

Leonard Barton, D., 1982. Core Capabilities and Core Rigidities: a Paradox in

Managing New Product Development. Strategic Management Journal, 13: 111-125.

Leonard Barton, D. & Deschamps, I. 1988. Managerial Infleunce in the Implementation

of New Technology, Management Science, 34(10).

Lewis, W., Agarwal, R., & Sambamurthy, V., 2003. Sources of Influence on Beliefs

About Information Technology Use: An Empirical Study of Knowledge Workers. MIS

Quarterly, 27.

Loch, C.H., & Huberman, B.A., 1999. A punctuated equilibrium model of technology

diffusion. Management Science, 45: 160 – 177.

38

Mahajan, V., Muller, E., & Bass, F.M., 1990. New Product Diffusion in Marketing: a

review and Directions for research. Journal of Marketing, 54: 1-26.

Mansfield, E., 1961. Technical Change and the Rate of Imitation. Econometrica, 29:

741-766.

Mabert, V. A., & Soni, A. & Venkataramanan, M.A. 2001. Enterprise Resource

Planning: Common Myths Versus Evolving Reality. Business Horizons, 44(3).

Meyer, J.W., & Rowan, B., 1977. Institutionalized Organizations: Formal Structure as

Myth and Ceremony. American Journal of Sociology, 83: 340-363.

Miller, K.D., & Folta, T.B., 2003. Option value and entry timing. Strategic

Management Journal, 23: 655-665.

Moore, G., & Benbasat, L., 1991. Development of an Instrument to Measure

perceptions of Adopting and Information Technology Innovation. Information System

Research, 38: 217-230.

Nohria, N., & Gulati, R., 1996. Is slack good or bad for innovation?. Academy of

Management Journal, 39: 1245-1264.

Orlikowski, W., 1996. Improvising Organizational Transformation over Time: A

Situated Change Perspective. Information Systems Research, 7: 63-92.

Orlikowski, W. J., 1992. The duality of technology: rethinking the concept of

technology in organizations. Organization Science, 3: 398-427.

Orlikowski, W. J., 2000. Using technology and constituting structures: a practice lens

for studying technology in organizations. Organization Science, 11: 404-428.

Rogers, E.M., 2003. Diffusion of innovation. New York: Free Press.

Schmalensee, R., 1982. Product differentiation advantages of pioneering Brands.

American Economic Review, 72: 349-368.

39

Shy, O., 2001. The economics of network industries. Cambridge University Press, New

York.

Strang, D., & Macy, M. W., 2001. In Search Of Excellence: Fads, Success Stories, and

Adaptive Emulation. American Journal of Sociology, 107.

Suarez, F., 2005. Network Effects Revisited: The Role of Strong Ties on Technology

Selection. Academy of Management Journal, 48: 710-720.

STATA 2005. STATA Manuals, TX(College Station); Stata Press.

Taylor, S., & Todd P., 1995. Understanding Information technology Usage: A test of

Competing Models. Information System Research, 40: 1069 – 1092.

Tolbert, P.S., & Zucker, L.G., 1983. Institutional sources of change in the formal

structure of organizations: The diffusion of civil service reform: 1880-1935.

Administrative Science Quarterly, 28: 22-39

Utterback, J. M., & Suarez, F. F., 1993. Innovation, competition, and industry structure.

Research Policy, 22: 11-21.

Van De Ven, A. H., 1986. Central problem in the management of innovation.

Management Science, 32: 590-607.

Venkatesh, V., Speier C., & Morris, M.G., 2002. User Acceptance Enablers in

Individual Decision Making about Technology: Toward an Integrated Model. Decision

Science, 33: 297-316.

Westphal, J., Gulati, R., & Shortell, S., 1997. Customization or conformity? An

institutional and network perspective on the content and consequences of TQM

adoption. Administrative Science Quarterly, 42: 366-394.

Wooldridge J.M. 2002. Econometric Analysis of Cross Section and Panel. Cambridge,

MA Cambridge, MA MIT Press

40

Zbaracki, M. J. 1998. The rhetoric and reality of Total Quality Management.

Administrative Science Quarterly, 43: 602-638.

41

Table 1 – Sample: Descriptive Statistics

Observations Mean Std. Dev Min Max

Time of First Use 3158 124.6 29.70373 22 136

Time of Adoption 3158 35.2 15.33683 2 59

Time to technology Use 3158 90.4 34.18454 1 135

Censor 3158 0.13 0.342 0 1

Industry - Agric. And Fish. 3158 0.01 0.098606 0 1

Industry – Constructions 3158 0.002 0.050275 0 1

Industry – Mining 3158 0.03 0.178488 0 1

Industry – Manufacturing 3158 0.41 0.491334 0 1

Industry - Transportation 3158 0.04 0.188133 0 1

Industry – Wholesale 3158 0.14 0.342342 0 1

Industry – Retail 3158 0.05 0.226346 0 1

Industry - Financial Serv. 3158 0.03 0.163703 0 1

Industry - Other Services 3158 0.27 0.446513 0 1

Industry – Public Administration 3158 0.002 0.435753 0 1

Legal Status - Personal company 3158 0.15 0.357531 0 1

Legal Status - Partnership 3158 0.17 0.374335 0 1

Legal Status - Limited Liability 3158 0.41 0.492714 0 1

Legal Status - Incorporated 3158 0.21 0.406654 0 1

Italy 3158 0.98 0.144126 0 1

Table 2 – Correlation Coefficientsa

V1 V2 V3 V4 V5 V6 V7 V8 V9 V1: Time of Adoption 1 V2: Cumulative Adopter Bandwagon 0.9833 1 V3: Cumulative User Bandwagon 0.97 0.9436 1 V4: Contiguous User Bandwagon 0.9015 0.9164 0.8838 1 V5: Contiguous Adopter Bandwagon 0.137 0.1374 0.115 0.2389 1 V6: Contiguous Media Bandwagon 0.7676 0.7894 0.6633 0.7022 -0.0184 1 V7: Contiguous User Bandwagon (SIC) 0.5839 0.5927 0.5589 0.6478 0.1364 0.4998 1 V8: Contiguous User Bandwagon (Location) 0.325 0.3595 0.2967 0.4172 0.1303 0.2632 0.3042 1 V9: Contiguous User Bandwagon (Legal Status) 0.6858 0.6817 0.6639 0.7393 0.1682 0.5602 0.5036 0.3427 1

a Correlations > |0.1| significant at p < .001

Table 3 – Model estimations

Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7

Cont. User Bwg. -0.0818** -0.163*** -0.0911* (0.0260) (0.0351) (0.0456)

Cont. User Bwg. (Legal Status) -0.461*** -0.159* (0.0985) (0.0818)

Cont. User Bwg. (Location) -0.700*** -0.0964 (0.152) (0.0899)

Cont. User Bwg. (Industry) -0.435*** -0.159 (0.0942) (0.0818)

Cumulative Adopter Bwg (Res.) -0.00124*** -0.000933*** -0.000976*** -0.000851** -0.00133*** (0.000372) (0.000255) (0.000232) (0.000259) (0.000374)

Cont. Adopter Bwg 0.00925*** 0.0126*** 0.0117*** 0.0126*** 0.0126*** 0.0127*** 0.0113*** (0.00247) (0.00265) (0.00266) (0.00254) (0.00254) (0.00253) (0.00267)

Cont. Media Mass Comm. Bwg -0.0118*** -0.00657** -0.00312 -0.00273 -0.00307 -0.00262 -0.00243 (0.00168) (0.00231) (0.00252) (0.00254) (0.00252) (0.00253) (0.00257)

Industry – Agric. and Fishing -1.300 -3.797 -3.949 -3.922 -3.967 -3.945 -4.100 (1.442) (2.203) (2.192) (2.192) (2.193) (2.191) (2.188)

Industry – Mining 32.15 27.28 28.23 26.51 28.18 25.60 29.79 (1879) (1209) (1548) (1060) (1547) (908.8) (2182)

Industry – Constructions 0.214 -2.331 -2.315 -2.317 -2.363 -2.384 -2.410 (1.014) (1.955) (1.945) (1.945) (1.947) (1.946) (1.944)

Industry – Manufacturing -1.115 -3.696* -3.776* -3.560* -3.800* -3.843* -3.542* (0.588) (1.775) (1.766) (1.788) (1.768) (1.768) (1.794)

Industry – Transportation -0.652 -3.085 -3.180 -3.158 -3.162 -3.214 -3.315 (0.843) (1.873) (1.863) (1.863) (1.864) (1.864) (1.862)

Industry – Wholesale -0.805 -3.317 -3.379 -3.316 -3.395 -3.437 -3.420 (0.656) (1.799) (1.790) (1.792) (1.792) (1.792) (1.792)

Industry – Retail -2.921 -2.892 -2.894 -2.889 -2.917 -2.948 -3.021 -1.866 (1.857) (1.847) (1.847) (1.848) (1.848) (1.846)

Industry - Financial Services -1.054 -3.537 -3.587 -3.560 -3.572 -3.596 -3.580 (0.907) (1.902) (1.892) (1.892) (1.893) (1.893) (1.890)

Industry – Other Services -0.360 -2.847 -2.866 -2.730 -2.865 -2.925 -2.740 (0.611) (1.782) (1.773) (1.781) (1.774) (1.774) (1.783)

Industry – Public Admin. -4.859* -7.195* -7.371** -7.246** -7.353** -7.389** -7.459** (2.281) (2.821) (2.790) (2.803) (2.793) (2.800) (2.804)

Legal Status - Personal Comp. -0.430 -0.325 -0.223 -0.256 -0.257 -0.214 -0.193 (0.941) (0.942) (0.939) (0.938) (0.939) (0.939) (0.938)

Legal Status - Partnership -1.593 -1.500 -1.460 -1.479 -1.469 -1.366 -1.285 (0.909) (0.910) (0.907) (0.906) (0.907) (0.909) (0.909)

Legal Status - Limited Liab. -1.441 -1.450 -1.385 -1.430 -1.399 -1.097 -0.828 (0.877) (0.879) (0.877) (0.876) (0.877) (0.904) (0.918)

Legal Status - Incorporated -3.220*** -3.258*** -3.214*** -3.256*** -3.212*** -3.030*** -2.823** (0.886) (0.889) (0.887) (0.885) (0.887) (0.896) (0.901)

Italy -1.575 -1.435 -1.435 -1.417 -1.363 -1.412 -1.340

(1.383) (1.384) (1.376) (1.377) (1.379) (1.377) (1.375)

Constant 14.93*** 16.39*** 16.71*** 16.29*** 16.46*** 16.19*** 16.09***

(1.547) (2.250) (2.248) (2.254) (2.235) (2.243) (2.254)

/ln_gamma 0.883*** 0.878*** 0.871*** 0.872*** 0.872*** 0.871*** 0.869***

Log Likelihood -2033.3 -2028.3 -2022.8 -2022.7 -2022.6 -2022.1 -2019.5 Standard errors in parentheses *** p<0.001, ** p<0.01, * p<0.05

44

Figure 1 – Smoothed hazard estimate

45

Appendix 1 – Joint Likelihood Ratio Tests Joint Likelihood Ratio Test Likelihood-ratio test LR chi2(1) = 10.11 (Assumption: model1 nested in model2) Prob > chi2 = 0.0015 Likelihood-ratio test LR chi2(2) = 21.21 (Assumption: model1 nested in model3) Prob > chi2 = 0.0000 Likelihood-ratio test LR chi2(2) = 21.28 (Assumption: model1 nested in model4) Prob > chi2 = 0.0000 Likelihood-ratio test LR chi2(2) = 21.50 (Assumption: model1 nested in model5) Prob > chi2 = 0.0000 Likelihood-ratio test LR chi2(2) = 22.49 (Assumption: model1 nested in model6) Prob > chi2 = 0.0000 Likelihood-ratio test LR chi2(5) = 27.69 (Assumption: model1 nested in model7) Prob > chi2 = 0.0000 Likelihood-ratio test LR chi2(3) = 6.49 (Assumption: model3 nested in model7) Prob > chi2 = 0.0901