GCR 99-780 - Estimating Social and Private Returns from Innovations Based on the Advanced Technology Program: Problems and Opportunities
5. ESTIMATION OF SOCIAL RETURNS FROM NEW TECHNOLOGY
Having looked briefly at the problems firms have encountered in trying to forecast the private returns from their investments in new technology, let's turn to the estimation of the social returns from such investments. Economists have tended to shy away from forecasting in this area; they generally have been content to do retrospective studies. To illustrate, consider the early study that my students and I carried out (Mansfield et al, 1977 a,b). This was the first such study of industrial innovations. Valuable previous work had been done in agriculture (see Evenson, Waggoner, and Ruttan (1979) and particularly Griliches (1958)), but it was necessary to extend this earlier work to handle many basic features of innovation in industry, such as the pricing policies of innovators, and to recognize the effect on displaced products, and the costs of uncommercialized R&D and R&D done outside the innovating organization, as well as environmental effects. Having done so, it was possible for the first time to make comparisons of social and private rates of return.
The innovations in our sample can be divided into three classes: product innovations used by firms, product innovations used by households, and process innovations. Based on an intensive study of each of the innovations in each of these classes, it appeared that the same general kind of model was applicable to all of the innovations in our sample in a particular class. This section describes the model that was used to measure the social benefits in a particular period from product innovations used by firms.
Each of these new products resulted in a potential saving to users. For example, the product innovation in the primary metals industry resulted in a potential saving to makers of household appliances. Thus, each of these innovations could shift downward the supply curve of the industry using the innovation. How far downward this supply curve will shift depends, of course, on the pricing policy of the innovator. If the innovator charges a relatively high price for the new product, the supply curve may shift only slightly. Indeed, if the innovator charges a high enough price, the supply curve will not shift downward at all.
Assume that the innovator decides to set a price for its new product, which yields a profit10 to the innovator equivalent to r dollars per unit of output of the industry using the innovation (for example, r dollars per appliance in the case of the new type of metal). Also, assume that the industry using the innovation is competitive and that its supply curve is horizontal in the relevant range. In particular, assume that, before the advent of the innovation, this supply curve was S1 in Figure 2, and the price charged by the industry using the innovation was P1. After the advent of the innovation, this supply curve is S2, and the price is P2.
Under these circumstances, the social benefits from the innovation can be measured by the sum of the two shaded areas in Figure 2. The top shaded area is the consumer surplus due to the lower price (P2 rather than fl resulting from the use of the innovation. In addition, there is a resource saving, and a corresponding increase in output elsewhere in the economy, due to the fact that the resource costs of producing the good using the innovation are less than P2 Q2. Instead, they are P2 Q2 minus the profits of the innovator from the innovation, the latter being merely a transfer from the producers of the good using the innovation to the innovator. Thus, besides the consumer surplus arising from the price reduction, there is a resource saving amounting to the profits of the innovator.
To illustrate, suppose that the innovator responsible for the new product used by the appliance industry reaps a $100 million annual profit from this innovation. This means that P2 Q2 is an over-estimate of the value of the resources used by the appliance industry. It is too big by $100 million, the amount the appliance industry pays the innovator in profits, because this payment is not in exchange for resources: it is a transfer of profit to the innovator.
Two adjustments must frequently be made in the estimate corresponding to the lower shaded area in Figure 2. First, if the innovation replaces another product, the resource saving cited in the paragraph before last does not equal the profits of the innovator (from the innovation), but these profits less those that would have been made (by the innovator and/or other firms) if the innovation had not occurred and the displaced product had been used instead. This is the correct measure of the resource saving. Second, if other firms imitate the innovator and begin selling the innovation to the industry that uses it, their profits from the sale of the innovation must be added to those of the innovator to get a full measure of the extent of the resource saving due to the innovation.
Using this model, an estimate was made of the social benefit in each period from the investment in each of these innovations. For each innovation, the top shaded area in Figure-2 equals
(P1 - P2 ) Q2 (1 - 1/2 Kn), (1)
where K (P1 - P 2 )1P 2 and n is the price elasticity of demand (in absolute value) of the product of the industry using the innovation. To estimate P1 - P2 , we obtained as much information as we could concerning the size of the unit cost reduction due to the innovation in the industry using the innovation. Based on interviews with executives of the innovating firm, executives of a sample of firms using the innovation, and reports and studies made by these firms for internal purposes, it was possible to obtain reasonably reliable estimates of (P1 - P2 ). And once we had an estimate of (P1 - P2 ), it was simple to compute K. Also Q was generally available from published records. Rough estimates of n were obtained from published studies and from the firms. Since K was generally very small, the results were not sensitive to errors in n. Indeed, the expression in equation (1) could be approximated quite well in most cases by (P1 - P2 )Q2 , which is the total savings to consumers due to the lower price, if they buy Q2 units of the product of the industry using the innovation.
To estimate the additional resource saving from the innovation, which equals the bottom shaded area in Figure 2 (if the adjustments described in the paragraph before the last are unnecessary), the innovator's profit from the new product was obtained from detailed discussions with the firm's executives, as well as study of relevant financial records. For each year, the costs of marketing and producing the innovation, as well as the costs of carrying out the innovation (R&D, plant and equipment, manufacturing start-up, and marketing start-up), were deducted from the innovator's revenues from the innovation. Note that the R&D costs were adjusted to allow for the fact that the innovator invested R&D resources in uncommercialized R&D projects. To make this adjustment, we obtained estimates from each of the innovating firms of the average number of dollars spent on uncommercialized R&D projects per dollar spent on a commercialized R&D project during the relevant period. Then we multiplied the innovator's R&D outlays (in each year) on the innovation by this number in order to get an estimate of the total R&D investment, including a pro-rated allowance for uncommercialized projects. In cases where the adjustments described in the paragraph before the last were necessary, estimates of the foregone profits from displaced products were deducted, and the profits of imitators were added, to the innovator's profits. These estimates were obtained from the relevant firms.
Date created: June 15, 2006
NIST is an agency of the U.S. Commerce Department