Social Returns from R&D: Findings, Methods and Limitations

Abstract

Economists have made dozens of studies of the social returns from particular new products and processes. In 1977, my students and I made what seems to have been the first study of the social returns from a sample of industrial innovations, but earlier studies had been made for agriculture. To extend our sample and replicate our study, the National Science Foundation commissioned two follow-on investigations, which obtained results quite similar to our own.(1). The model we used to derive the social benefits from an industrial innovation is relatively straightforward. Suppose that the innovation is a new product used by firms, and that it reduces the costs of the industry using the innovation. The social benefits from the innovation can be measured by the profits of the innovator from the innovation plus the benefits to consumers--that is, consumers of the good produced by the industry using the innovation--due to whatever reduction occurs in the price of this good due to the innovation. Because a variety of adjustments have to be made, the calculations are not as simple as this statement seems to imply (and the task of collecting data can be arduous and expensive), but this conveys the spirit of the analysis. Moreover, this type of model can be applied to innovations of many kinds, not just new products used by firms. Using a model of this sort, economists have found that the social rate of return--that is, the rate of return to society as a whole--from investments in industrial innovations tends to be very high. For the 17 innovations in our 1977 study, the median social rate of return was about 50 percent. For the two follow-on studies, each including about 20 innovations, the median social rates of return were even higher. Also, in each of these studies, the social rate of return from an innovation was, on the average, at least double the private rate of return to the innovator from its investment in the innovation. In agriculture, too, there has been considerable evidence that the social rate of return from investments in research and development has been high, often over 40 percent. Also, a recent study of a medical innovation, CT scanners, came to the same conclusion, the estimated social rate of return being over 200 percent (2). R&D Expenditures of Industries and Firms Besides studying individual innovations, economists have used econometric techniques to estimate the rate of return from an industry's or firm's R&D expenditures. Typically, it is assumed that the output of a particular industry or firm is a function of the amount of labor and capital used, as well as the amount of R&D done by the industry or firm itself and by its suppliers. Based on historical data, statistical methods are used to estimate the relationship, holding other factors constant, between output and the amount of R&D. Given this relationship, one can obtain rough estimates of the rate of return from R&D. One advantage of this approach over that described previously is that there is no need to choose samples of individual innovations, which may be biased for one reason or another. (In the case of studies of innovations that are known in advance to have been successful, the estimated rate of return obviously is biased upward.) Also, this approach is relatively inexpensive. Whereas a careful study of an individual innovation can take many person-weeks, econometric studies of hundreds of firms in dozens of industries can be done relatively cheaply because they rely largely on published data (or data that have already been collected for other reasons). But econometric studies of this sort have many problems of their own. Since they basically are built around estimated statistical relationships between a firm's or industry's rate of productivity increase and its R&D expenditures, there are many nettlesome problems that must be faced. Output measures, particularly in many science-based industries, are often treacherous for these purposes. …