Networks, network effects and technology adoption.

摘要

Networks are often composed of heterogeneous participants. This thesis shows that this heterogeneity can be exploited in order to establish causal relationships between the installed base of users of a network technology, and network technology adoption. I look at two main examples.; First, I consider two-sided network effects in electronic payments. Many new technologies have two groups of users, whose ultimate benefit stems from their interaction through a common network. Lack of coordination could lead to an equilibrium where cost savings are not realized, since no one adopts. In the ACH payments network, there are Originators of payments, and Receivers of payments. Since the presence of one group is complementary to the other, one group of users' decisions to join the network has a positive effect on the other group. I use shocks to the installed base on one side of the network, to identify positive network effects on the other side. Likelihood bounds methods that reflect the potential for multiple equilibria in the initial setting further support this result.; Second, I look at direct network effects in communication technologies. I use extensive data on all potential adopters of a firm's internal video-messaging system and their video-messaging patterns. The technology can also be used to watch TV. Exogenous individual shocks to benefits of watching TV are used to identify the causal (network) effect of changes in the installed base on installation decisions. There are two major findings. The first finding is that network effects are "interaction-based". Employees only react to the adoption decisions of those they ultimately communicate with. The second finding is that network effects are heterogeneous: Installation by managers, and workers in "information broker" positions, has a large impact on the installation decisions of employees who wish to communicate with them. Installation by ordinary workers has a far smaller impact. Furthermore these "influential" installers do not internalize the positive effect their installation has on the future installation decisions of others. Optimal technology policy should therefore focus adoption incentives on influential users of a network product, rather than subsidizing everyone's adoption equally.