In this work, we examined the boundary of the applicability of network effects theory. We theorized that when adoption is cheap, the cognitive demands of estimating network effects outweigh the benefit of making optimal adoption decisions. Thus, even in contexts where network effects do exist, we predict that adopters will use simple heuristics to make adoption decisions, if adoption is cheap. We propose that adopters simply do what they observe others doing. Using the context of peer-to-peer file sharing, we conducted three studies comparing network effects against simply following the behavior of others, and found in all three cases that there was no marginal effect of network size on willingness to adopt. However, when subjects were told classmates’ adoption choices, there was a strong marginal effect on willingness to adopt. Put simply, if people are offered the option of downloading a free peer-to-peer software, then changing the network size from 1,000 to 1,000,000 has no effect on their willingness to adopt, but having two classmates express a choice not to download the software, had a large negative impact on a subject’s willingness to download. Thus, our subjects, when faced with the option of a free download, did not behave in accordance with network effects theory, suggesting that there is a boundary on the applicability of the theory to internet adoption behavior.
Network effects theory is one of the bedrock theories of information systems (IS). We use the concept of network effects from our first undergraduate classes to our PhD seminars, from our executive education to our top research journals. However, the breadth of its applicability has been challenged by both academics (Liebowitz and Margolis 1994) and journalists (Useem 2001). The criticism can basically be stated as, “[c]ontrary to common misconceptions, network effects do not come about just because business is conducted on a network. (Liebowitz 2002 p. 4).” Thus, researchers have begun to ask, “in what sorts of environments do network effects occur?”
In this work, we try to address the boundary conditions of network effects from a cognitive perspective. We conceive of network effects as a special case of adoption decisions, and thus, focus on the cognitive processes of the decision maker. For network effects theory to be applicable, decision makers must incorporate the network effects into their decision processes. We argue that it does not always make sense to believe that decision makers do, or should, consider network effects when making an adoption decision, even if network effects are present. Specifically, we argue that if the costs of adoption are low—for example, free downloads, which characterizes a great deal of electronic commerce adoption, particularly P2P file-sharing applications—then it is not really worth the effort of the decision maker to consider all the facts. Thus, we ask the question, “Do individuals consider network effects when adoption costs are close to zero?”
We propose that decision makers use computationally simple heuristics to solve low cost adoption decisions. Specifically, we propose they use observational learning or imitation. If people around them choose to adopt a technology, then they choose to adopt the technology. Thus, our second research question is, “If decision makers do not consider the network effects, what should they consider?”
This approach fills an important niche in our understanding of network effects and its relationship to adoption. We follow the spirit of the Social-Economic-Psychological (SEP) model by considering that “…complex IT-driven phenomena … may involve a web of interdependencies related to constructs either purely of a social, economic, and psychological nature, or to constructs that lie at the interfaces between these areas. (Konana and Balasubramanian 2005 p. 508).” We do so by tying the information in the social and economic environment to the adopters’ cognitive costs of processing that information. Thus, it addresses the decision maker’s role in network effects, rather than just looking at the role of the environment. This is particularly salient for electronic commerce researchers, because so much of the adoption that occurs online is low-cost adoption, and network effect theory is applied to these adoption decisions (Kauffman and Walden 2001). If, as we propose, decision makers choose not to process the network effect information, then researchers need to be more cautious in applying network effect theory. Also, if we are correct, then we may be able to offer some explanation of adoption in the nascent stages of a network’s growth. Finally, our study suggests that wide-scale adoption may be an emergent result of local phenomena, and that small scale adoption patterns may be quite different than those observed at the global level and predicted by network effect theory.
We perform a series of four experiments aimed at understanding how network effects and observational learning affect P2P adoption decisions. We find that observational learning has a very strong impact on people’s willingness to adopt a P2P technology. However, network effects have little effect. We attribute this to individuals’ difficulty in determining how many users a P2P technology needs to make it worth the effort of adopting.