GCR 99-780 - Estimating Social and Private Returns from Innovations Based on the Advanced Technology Program: Problems and Opportunities
2. INDUSTRY'S ESTIMATION OF PRIVATE RETURNS FROM NEW TECHNOLOGY
To begin with, let's consider the estimation of private returns from new technology. While this is not the primary interest of the ATP program, it is the area where the most attention (by far) has been devoted. For decades, leading high-tech firms have been faced with the problem of estimating the returns from their R&D (and related) investments. While some firms seem to have felt that retrospective studies of the returns from such investments have been useful, the available studies indicate that few firms, if any, are confident of their forecasts of the profitability of particular R&D projects.
At General Electric, for example, a simple scoring system has been used to estimate the relative profitability of various R&D projects.2 Two values have been calculated, the first being
M x S x G x T,
where M is the estimated size of market (for a new product), S is probable GE share, G is estimated rate of growth of market, and T is a measure of its sensitivity to technological advance. Whereas the first value pertains to the impact of the program if it succeeds, the second value pertains to the probability of success. Specifically, this second value is
D x B x F x O,
where D is a measure of the difficulty of the technical problem, B is a measure of the competitive status of the lab effort, F is a measure of the fit with the laboratory's resources, and 0 is a measure of the ease of transition to operations. These two values have been used to try to determine which R&D projects should be carried out (Steele, 1988).
Given the crudeness of such scoring methods, it is not surprising that firms have experimented with more sophisticated techniques like linear programming. To illustrate how linear programming can be used, suppose that a firm has a list of n possible R&D projects that it might carry out and that to undertake project I would cost C, dollars. Moreover, suppose that project I has a probability of success of P, and that, if successful, it will result in a discounted profit (net of commercialization costs but gross of R&D costs) of it 1. Then, if the firm can spend no more than C dollars on R&D, and if it wants to maximize the expected value of profit, its problem can be represented as follows:
Maximize subject to the constraint that
and X1 = 0,1.
In other words, the firm's problem is to choose the X1 -where X1 = 1 if project I is accepted and 0 if it is rejected - in such a way that it maximizes the expected value of profit, subject to the constraint that the total amount spent on R&D does not exceed C.
Using programming techniques, this problem can be solved. However, this does not mean that techniques of this sort are widely used. One reason why programming techniques have not found more extensive use is that data concerning Pi i and Ci are often very rough. Much more will be said on this score in Section 4. For now, the important point is that surveys indicate that relatively sophisticated quantitative techniques generally have not been used by firms to select R&D projects. Instead, relatively simple (and crude) techniques generally are applied (Steele, 1988; Mansfield et al, 1971).
Besides using quantitative techniques to try to select R&D projects, firms sometimes look at the prospective returns from a sample of the potential innovations emerging from their R&D to provide a rough indication of the potential value of their research. For example, a leading oil firm has carried out a detailed series of decision analyses for this purpose.3 Each year, a number of potential innovations was chosen by the managers of the firm's central research laboratory. These projects had to be at a sufficiently advanced stage so that reasonable estimates concerning prospective costs and benefits could be made. Basically, the present value of the cash flow from the innovation was estimated, given each of a set of possible circumstances. Probabilities were attached to each of these circumstances, and the expected value of the discounted cash flow was calculated for the innovation. The computation of discounted cash flow began with the present year and included R&D expenditures through the pilot-plant stage. (Benefits are discussed below.) Deflated dollars were used throughout each analysis. A discount rate of 5 percent was employed because, according to the firm's personnel, this "corresponds to [the firm's] expected rate of return on its investments, when measured in the context of constant (i.e., deflated) dollars." 4
According to an executive of this firm who was intimately involved with these decision analyses, many projects displayed a similar logical structure: "The initial step in the analysis is the identification of the qualitative benefits to be provided by the innovation. Usually this is a straightforward issue; for example, the project may involve a process improvement, the production of a more valuable product, or the establishment of a completely new business area. Concomitant with this step is a clarification of the 'next best' existing alternative to the new technology. By definition, this alternative is the one that [the firm] would practice in the absence of the innovation. While this initial step may seem obvious, it is critical to stress the importance of clearly identifying the alternate case so that a suitable reference case can be established and that the appropriate incremental benefits can be taken into account."
More specifically, according to this executive, there generally have been several models that have been part of the analysis of each innovation. "At the beginning is the 'Technical Model' in which the key scientific and technological issues are enumerated. Included are topics which need to be quantified before the project can proceed to a pilot-plant or pre-commercial stage. Often these issues are phrased as knockouts, that is, critical-path items that must be resolved favorably for the project to remain viable. Other issues may be characterized as 'figures of merit' and are associated with parameters whose numerical values determine the actual quantitative benefit derived from the innovation."
"The 'Economic Model' follows and includes a detailed computation of the net savings provided by the new technology. This model requires the parameters from the preceding technical model. Often an algorithm is created that calculates the incremental benefits relative to a reference case based on existing technology. The computational guidelines for cost estimation are consistent with [this firm's] standard engineering practice. In fact, direct involvement by engineering personnel in this phase of the analysis is usually required. The output from the model consists of generalized expressions for the incremental capital and operative savings per unit application of the innovation."
"The 'Market Model' is undoubtedly the most difficult model to construct. This deals with how the innovation will affect [this firm's] future business activities over the period of the technology's lifetime. Typically, the development of the model begins with categorizing the types of application areas and identifying the corresponding receiving organizations-(e.g.; affiliates). An effort is then made to obtain broad inputs ... to construct a scenario representative of an implementation schedule. The model may ultimately predict how many refineries would utilize a new process technology as a function of time. Alternatively, the model could predict a market penetration scenario for a new business example using cost projections and assumptions regarding future demand and competitive technologies."
Date created: June 15, 2006
NIST is an agency of the U.S. Commerce Department