Reference: “Innovations and Technological Spillovers” by M. Ishaq Nadiri – pdf at http://ideas.repec.org/p/cvs/starer/93-31.html

Recently in class we began discussing the idea of diffusion in networks. This notion of diffusion can be useful for modeling many things including the spread of information or rumors, the epidemiological problem of disease outbreaks, or (as mentioned in class) the acceptance of certain new technologies. The model developed in class thus far has focused on the idea of a network of individual who play coordination games among themselves to determine their optimal behavior. As usual, this model is a simple abstract to which many complications can and should be added to obtain a richer model of these complex processes. One such complication that remains relatively open to further research is the notion of defining the complex network on which these diffusions occur.

Whether we are concerned with a network of individuals or of larger economic entities, such as companies or even entire industries, it is equally interesting and important to understand the dynamics of how new technological advances (products or processes) move from one node to the next. In order to study and model these dynamics, however, we must have an accurate representation of the structure across which the technologies are being diffused. Several studies cited by M. Ishaq Nadir in a section on “technology flow methodology” in the referenced article show that this problem is difficult to solve in its own right. These studies have tried to define various network structures including those that span several industries and those that connect firms within a single industry. The difficulty in do these analyses is the lack of a large quantity of reliable data about the borrowing of R&D results. These difficulties aside for now, most often the purpose of these research projects was to determine the rate of return due to R&D spillovers in high-tech industries, a direct product of network externalities. However, these network definitions could also be used to further enhance our models of the diffusion process itself.

Whereas the previously mentioned studies intended to determine the returns to private investors as well as to society derived from the diffusion of new technologies, it would be interesting to build these empirically defined industry networks into the diffusion model. This type of analysis would reveal which firms and industries or types of firms and industries fall into dominant positions within the network. Also it would be useful to measure the strength of each link in the network in terms of the ability of on player’s decision to influence that of his neighbors. Incorporating these specific real-life structures into the diffusion model would certainly make the problem more difficult, but the end product could be a very useful analytical tool that could aid in the decision making of not only individuals and firms but also of the federal government who subsidizes a large portion of the total R&D performed each year.

Overall this does not seem like such a unique proposition. In all of the network models that we have encountered this semester, a clearer understanding of the physical, social, or economic network to which we are trying to apply a simple model will certainly lead to more complicated but more rewarding discoveries. The interesting notion here is that we can apply a method first attempted for a slightly different purpose and enhance it by looking at not only industry network structures but also those that connect industry to academia and industry to consumers to achieve some truly revealing results about the diffusion of technological innovations through a network.