Network Effects in the Adoption of Low Cost Internet Technologies
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1371694
This study examined whether or not the network effects/externality theory could be applied as a behavioral model of lower cost internet technology adoption.
Network externality theory requires that people calculate the value of a network based on the number of members in the network. The paper hypothesizes that whilst network size may be an important determinant of network value, average people do not have any idea how to evaluate and incorporate network size into an adoption decision.
In the study, a simulated website was created for the sharing of course materials between university students. Students were given a scenario in which they needed help text book questions, and told that someone might have posted the answers on the simulated website. A value of $10 was assigned to finding the answers, and a cost of $5 was assigned to joining the network.
Students were given the number of students at their university and the number of students enrolled. They were also told the number of other university students served by the network, and told that 10% of other universities which are served by the network offer the same course with the same book. A participation rate of 100% was assumed, that is, 100% of users made their files available. The network sizes given to the test subjects varied from 7,000 file sharers (giving a 40% probability of a match) to 12,000 file sharers (giving a 60% probability of a match).
Subjects were tested in sequential order. Subject 1 made the decision whether or not he wanted to join the network. The next subject, aware of the previous subject’s decision, was then asked whether or not he wanted to join the network. After the third subject, everyone who indicated that they wanted to join was allowed to simulate their search. They were allowed to keep the $10 if they found the file; otherwise the researcher kept the money.
The results: network size had no significant impact on the subject’s willingness to join. The willingness of subject 1 and subject 2 to join however, did have an effect on subject 3’s willingness to join. The test supports the information cascade hypothesis; the test subjects made decisions based on earlier subjects actions rather than their own private information.
We can think of f(n) as the benefit to each user as a function of network size n. The adopter’s decision however, is based on a calculation of Pr[G|S], the probability that the answers are on the website, given sequence S of independently generated signals consisting of a high signals (previous subjects joining) and b low signals (previous subjects not joining). Subjects choose to join the network when they get more high signals than low, and reject it when they more low signals than high. The reason for this is that the cost of calculating f(n), C(f(n)), is greater than the cost of calculating Pr[G|S], C(Pr[G|S]).
The article acknowledged that the sizes of the network chosen for the study might have been the reason for the results. If there had been greater differences in network size, then perhaps the subjects would have responded differently.