How is this model linked to empirical evidence?
crucial importance, because with low differentiation high inter-industry price elasticity a and low e, welfare decreases with the entry of new users. This result is
coherent with the analysis of the impact of entry on the steady state TFP: in Section 4.2, we said that A
d
decreases with entry if the differentiation and the extent of spillover are low. The decreasing welfare is then driven by this negative effect on
A
d
, which decreases consumer surplus and total profits. We obtain the following result:
Simulation result
8
: when differentiation and extent of network effect are low, total welfare decreases with entry.
6. How is this model linked to empirical evidence?
The question of the impact of the network effect on the TFP is closely related to the Solow productivity paradox. Indeed, the major part of the empirical literature
on the Solow productivity paradox measures the effect of productivity investment on TFP at the industry or firm level, while our model analyses TFP at the steady
state i.e. when the TFP is constant. Nevertheless, we think that the relationship we found between the extent of network effect and the productivity level offers some
interesting insights about the impact of information technologies on TFP.
As stated in the introduction of this paper, in recent years, the relationship between information technology and productivity has become a source of debate.
As consequence, studies attempting to measure the impact of IT on total factor productivity are very heterogeneous.
Aggregate level studies are quite disappointing: the overall negative correlation between productivity and the advent of computers underlies many of the arguments
that information technology has not helped the United States productivity, or it has been counterproductive Jorgenson and Stiroh, 1995; Baily, 1996. Indeed, the
relationship between white-collar productivity and information technology is very hard to measure, especially with aggregate data Brinjolfsson and Yang, 1996.
As going down to the industry level and emphasizing cross-sectional effects should help to control many problems that arise from aggregation Siegel and
Griliches, 1991, we have investigated the micro-economic conditions that foster investment in communication technologies. We claim that telecommunications and
information technologies are characterised by the network externality effect, distin- guishing them from a simple substitute for labour or other kinds of capital inputs.
This opens the possibility of a more accurate specification of the impact of IT on total factor productivity via the network effect, as a proxy for the spillovers
obtained by using the communication technologies. Our results show that, ignoring the complex interplay between market conditions, number of users, extent and
productivity of network effect may bias the measurement of the IT contribution to the firms’ performance.
Our simulation results suggest that the impact of IT is non-neutral with respect to the characteristics of the firms analysed. This is confirmed by Brinjolfsson and
Hitt 1993, who find that up to half of the excess returns imputed to IT could be
attributed to firm-specific effects. It is thus worthwhile looking at the empirical results and ranking firms or industries according to the expected extent of
spillovers. This expected extent of network effect could be considered as important as the Yale technology flows matrix, that represents the technological closeness
among sectors, and is used in RD studies as a tool to measure the pure knowledge spillover intensity Van Mejil and Van Zon, 1933.
Following our predictions, we could expect that activities requiring a large communication network, an intensive extent of spillovers, and good technological
opportunities, like for example, retailing and banking services, would experience a positive impact of IT on total factor productivity
11
. In general, at the firm level, the results of IT on productivity seem to be more clearly associated with the kind of
industry to which the firms’ sample belongs, as our results confirm. The results of our model regarding the positive impact of the network effect on
the consumers’ surplus are consistent with Bresnahan 1986, who estimated the benefits to consumers of declining computer prices using the hedonic price index
method. Finally, our results on total profit’s analysis offer a possible answer to the debate
frequently raised by management science in attempting to understand whether the usage of IT leads to more concentrated or less concentrated industries. The
argument that favours the first hypothesis is that network technologies have the advantage of organising the production process with a lower number of subordi-
nates, higher control, and just-in-time scheduling Mintzberg, 1982. The second view refers to the decentralisation effect that is possible when efficient communica-
tion networks are used Hubey, 1990; Brinjolfsson, 1993. The results of our model show that the causality between centralisationdecentralisation and communication
technologies is not driven simply by the usage, as the above mentioned studies claim. We find that the equilibrium number of users is determined by combining
market conditions, productivity of the communication network, and extent of the network externality effect.