closing the technology adoption/use divide: the...
TRANSCRIPT
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
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