Recent contributions to the literature showed that an ICT producing sector is not a precondition
to capture the benefits of “Information and Communication Technologies” (ICT).
Timely diffusion of new technology or, from a firm’s point of view, its adoption is at least as
important to promote macroeconomic growth (see e.g. Pilat and Lee, 2001). Understanding
the factors determining technology adoption is thus a highly relevant topic, also from the
policy point of view.
In this paper, we aim at explaining empirically timing and intensity of a firm’s adoption
of ICT as well as of certain elements of this bundle of technologies (Internet, e-selling). To
this end, we estimate a basic as well as an extended version of a model of adoption. We
shall present empirical evidence regarding the most important hypotheses put forward in
the theoretical literature or investigated in previous empirical work; we also take account
of propositions derived from case studies or casual observation. However, the paper does
not strive for further developing the theory of technology adoption.
The analysis refers to ICT in a narrow sense. We do not consider the adoption of other
computer-based technologies such as “Advanced Manufacturing Technologies”. In addition,
we do not investigate the adoption of “old” brands of ICT such as PCs where diffusion
is almost complete. Finally, diffusion of telecommunication devices such as mobile phones
also are excluded from this investigation.
The study, in the first place, is based on a “rank model” of technology diffusion, which,
in explaining inter-firm differences of adoption time and intensity, emphasises differences
among firms with respect to the profitability potential of technology adoption arising from
the heterogeneity of firms. In addition, we take into account information spillovers from
users to non-users, which are the main element of the “epidemic model” of technology
diffusion; see e.g. Karshenas and Stoneman (1995) or Geroski (2000) for comprehensive
surveys of various brands of diffusion models. According to the empirical literature, “rank
effects” and “epidemic effects” are the dominant factors explaining the adoption of new
technology (see e.g. Canepa and Stoneman, 2003).
The data stem from a large survey on the use of ICT we conducted in the Swiss business
sector in autumn 2000. We have at our disposal firm-specific information on, for example,
the time period of adoption of nine technology elements, the proportion of employees using
specific technologies, the range of application of Internet and Intranet, respectively, the
objectives of and obstacles to the adoption of ICT, etc. Moreover, we collected information
referring to various structural characteristics of the firm (size, industry affiliation, etc.).
In addition, we got data pertaining to workplace organisation which also may serve as
determinant of the decision to adopt ICT. The final dataset contains information for more
than 2600 firms.
Our paper adds to previouswork in variousways. Firstly,we can drawon a large (and new)
database in terms of the number of variables and observations as well as the wide coverage of
industries/sectors and size classes. Secondly, by estimating the postulated empirical model
with several types of adoption measures as dependent variables, we are able to identify
differences in the pattern of explanation and to separate robust from shaky relationships; we
expect, for example, that the first use of Internet (a basic element of ICT with a broad range of
application) is driven by somewhat different forces than the introduction of e-selling, whose profitability potential seems to vary substantially across industries (OECD, 2000). Thirdly,
we are able to consider not only inter-firm diffusion but also intra-firm diffusion. To date,
not much research has been devoted to this second type of technology diffusion (see Battisti
and Stoneman, 2003). Fourthly, and perhaps most importantly, our approach adds to the
present understanding of ICT adoption by modelling “rank effects” more comprehensively
than it is the case in most empirical analyses. We specify anticipated profitability of ICT
adoption by taking account of many dimensions of presumed benefits from as well as costs
of adoption. To this end, we use detailed information stemming from our survey which
refers to the relevance of specific objectives of and obstacles to the adoption of ICT as
assessed by the firms themselves.1 In this way, we are able to take into account some factors
such as (technological) uncertainty as well as information and adaptation costs which,
although they seem to be highly important, are usually ignored in the empirical analysis
of adoption decisions (see Karshenas and Stoneman, 1995). Finally, we explore the role of
“New Workplace Organisation” (NWO), which involves practices such as team-working,
flattening of hierarchical structures, decentralising of decision-making, etc., as a factor
determining the adoption of ICT.
The set-up of the paper is as follows: Section 2 is devoted to the conceptual framework
of the empirical analysis. Section 3 provides information on the database as well as a
brief description of the time path of diffusion of ICT in the Swiss business sector. Model
specification and estimation results are shownin Sections 4 (“basic model”) and 5 (“extended
model”), respectively. Finally, we assess the main results and draw some conclusions.
The adoption behaviour of Swiss firms in the field of ICT is characterised by a basic
pattern of explanation which is quite robust across model estimations with different adoption
variables. All categories of explanatory variables we distinguished, though to a different extent, are relevant. Most important are anticipated benefits (primarily those dimensions
reflecting market-orientation and efficiency gains) and costs of adoption (in particular,
investment costs as well as know-how deficiencies and managerial problems), the firm’s
ability to absorb knowledge from other firms and institutions, technological opportunities,
information spillovers from adopters to non-adopters, experience with earlier vintages of a
certain technology, (international) competitive pressure and firm size. In addition to these
firm-specific effects, there is also evidence for industry effects (with a higher probability
of adoption in some high-tech industries, modern services and wholesale trade). These
empirical results, which are based on a large dataset, are in line with the hypotheses put
forward in Section 3, which reflect theoretical thinking, earlier empirical work as well as
some casual observations.
Basically, the general pattern of explanation is quite similar to what has been found
in studies related to “Advanced Manufacturing Technologies”; see Canepa and Stoneman
(2003) for a comparison of several empirical studies related to the adoption of these technologies.
Rank and epidemic effects are the most important drivers of the adoption of both
bundles of technologies. An interesting difference pertains to firm size, one of the most
prominent variables in studies of technology adoption. Firm size mostly exerts a strongly
positive influence on adoption. In case of ICT, we get a more differentiated picture. We
find positive size-effects only up to a threshold of about 200 employees; moreover, in the
case of the intra-firm diffusion of the Internet, medium-sized firms seem to have the highest
propensity to adopt the new technology.
Our results strongly confirm the usefulness of modelling anticipated profitability of technology
adoption more comprehensively than it is the case in most empirical models.We take
account of a whole set of revenue and cost components. In particular, we are able to include
factors such as technological uncertainty, information problems and adjustment costs, which
are neglected in most previous studies, although their importance is stressed in the literature.
We find that know-how deficiencies, managerial problems as well as costs and financing of
ICT are the most important obstacles to introduce these technologies, whereas there is hardly
any evidence for a negative impact of (technological) uncertainty and switching costs. Some
of these results, in particular those related to management problems and switching costs, differ
from those we got in earlierwork dealing with “Advanced Manufacturing Technologies”.
The relative importance of the explanatory variables is thus technology-specific. The analysis
also shows that ICT is not only a cost-reducing, efficiency-enhancing technology but
also exhibits a great potential to generate competitive advantages based on new output
characteristics (product innovations, improving customer-orientation, after-sales services,
etc.).
Furthermore, we identified some interesting differences among results pertaining to specific
types of adoption variables, i.e. the timing of the adoption of specific ICT elements
(inter-firm diffusion of Internet and e-selling) and the intensity of use of ICT (intra-firm
diffusion of ICT in general and of Internet in particular). Firstly, it turned out that the adoption
of e-selling is driven to a greater extent by (anticipated) benefits related to customerand
market-orientation as well as by epidemic and learning effects than it is the case for the
introduction of the Internet; for the Internet, efficiency gains through optimising production
processes and relations with suppliers as well as absorptive capacity are more important
determinants of timing decisions. Secondly, we also find some differences in explaining inter- and intra-firm diffusion of ICT. Absorptive capacity, firm size, cost of technology
as well as anticipated benefits from improving internal processes are more important as
determinants of adoption in the case of intra-firm diffusion, whereas quality-oriented and
customer-related advantages are more relevant for timing decisions. The overall effect of
the postulated explanatory variables on adoption is stronger in case of intra-firm diffusion.
This result does not seem implausible, since the firms’ resource commitment often is low
in an early phase of the adoption process, whereas beyond a certain level of (intra-firm)
diffusion intensifying the use of ICT becomes more complex and expensive (e.g. transition
from stand-alone to network technologies). Notwithstanding these differences across various
adoption variables, it has to be stressed that the basic pattern of explanations is not very
different for the adoption variables distinguished in this paper.
Finally,we found evidence for the influential role “NewWorkplace Organisation”(NWO)
plays in decisions related to the adoption of ICT. Team-working, decentralised decisionmaking
and flattening hierarchical structures seem to be the most relevant dimensions of
NWO that favour the adoption of ICT, whereas, not surprisingly, we do not find any impact
of job rotation. However, the results regarding NWO may be affected by potential endogeneity
of this variable. To circumvent this difficulty, we investigated the reverse causality
running from the adoption of ICT to the introduction of NWO, and we found evidence for
this proposition as well. In addition, we introduced time lags, alternatively for the ICT and
the NWO variable; the results provide some indication of a more sluggish adaptation of organisational
structures as compared to technology adoption (“organisation” as a quasi-fixed
factor in the short run). These findings seem to be consistent with those of some recent studies
which found that ICT and NWO, at least in the longer run, are complementary elements
of a strategy to increase efficiency of production and quality of products.
However, the conclusions regarding the role NWO plays in the process of ICT diffusion
are very preliminary. Further research is required to investigate in greater detail the relationship
between the seemingly complementary variables ICT and NWO. As stressed by
Brynjolfsson and Hitt (2000, p. 35), correlations between NWO and ICT are not sufficient
to prove complementarity. However, the same authors also indicate that after an empirically
evaluation of possible alternative explanations, complementarity is often the most plausible
interpretation.10 To get more insight, the use of simultaneous estimation techniques with
ICT andNWO (and perhaps human capital as well) as endogenous variables would be helpful.
Moreover, panel estimation, provided that suitable data are available, could contribute
to uncover the dynamic relationship between ICT and NWO.
