The Directions for Technological Change: Alternative Economic Majorities and Opportunity Costs

Transcription

1 1 The Directions for Technological Change: Alternative Economic Majorities and Opportunity Costs John T. Scott* Department of Economics Dartmouth College Hanover, NH USA Telephone: Fax:

2 2 ABSTRACT: Governments redirect R&D resources. The government supports R&D in technologies and in parts of the country that the private sector is unlikely to support, and the government provides funds for inexperienced businesses and for minority-owned businesses when the private sector will not. Because the distribution of income and wealth is so unequal, our conventional methods of evaluating the social usefulness of the redirection of the resources may not work well. Redirection of R&D resources to meet Schumpeterian and Jeffersonian objectives may actually be efficient despite generating less economic surplus given the current distribution of income and wealth. KEY WORDS: R&D, small business, technological change, technology policy

3 3 I. Government and the Allocation of Research Resources. After two decades of writing about technological change, I have had the opportunity to work during the last four years with the Organization for Economic Cooperation and Development (OECD), the National Institute of Standards and Technology (NIST), the National Academy of Sciences (NAS), and the National Science Foundation (NSF) on technology policy questions. 1 My work with the institutions that address technology policy has convinced me that there is a gap in our knowledge; we do not understand well the opportunity costs of research resources. That is, we do not know the alternative directions for technological change. Discussing the knowledge gap will bring together and provide new understanding about three industrial organization literatures those addressing (1) public versus private R&D spending, (2) the Schumpeterian issue of large, diversified firms with market power versus small, focused firms without substantial market power, (3) the antitrust issues about economic efficiency versus the Jeffersonian ideal of a dispersion of power. This essay considers the social optimality of a government s redirection of technological change. I present new evidence that government can redirect research and development (R&D) effort toward different types of technologies and can also redirect the performance of R&D toward different sorts of companies. I then argue that, potentially, many different directions for technological change are socially optimal, and, consequently, economists need to develop knowledge about opportunity costs that will be useful in the determination of the appropriate choice of the direction of technological change. There is an industrial organization literature that addresses the possibility that private R&D spending is displaced when the government finances the private sector s performance of

4 4 R&D. Some studies have found that sort of substitution for examples, Irwin and Klenow (1996) and Wallsten (1997) with the private sector reducing private R&D expenditures and using the public s funds instead. Others have found a complementary relation examples here include Mansfield (1984) and Scott (1984) with the private sector not only continuing its privately financed R&D effort but increasing it somewhat in addition to performing the publicly financed R&D investments. 2 The results are not inconsistent with each other for the examples chosen because the samples are quite different. 3 Irwin and Klenow observe the Sematech Consortium, while Wallsten studies Small Business Innovation Research (SBIR) grants. Mansfield uses survey methodology to study a sample of major firms in the chemical, oil, electric equipment, and primary metals industries, while Scott uses a cross-section statistical study of over 400 large firms with over 3000 observations on their activities in various industries. Most importantly, both sets of results are consistent with government redirecting the path of technological change in socially optimal ways. Sections II and III describe government policy that redirects technological change and redirects R&D activity toward firms that would not play a role in technological change without government policy. Although certainly not explicitly designed to do so, the public policy I discuss actually addresses the concerns of the Schumpeterian literature. 4 Namely, the technology policy that I discuss recognizes that in some circumstances a purely private sector solution to R&D investment might necessitate large, diversified firms with market power to achieve good economic performance for technological change. Yet, even in those circumstances, public policy can make small, focused firms with no market power viable R&D competitors, while at the same time preserving the Jeffersonian ideal of desirable dispersion of economic, social, and political power. 5 Section IV discusses the conceptual issue of defining the

5 5 opportunity costs of research resources. Section V concludes that evaluations of government technology policies that develop the opportunity costs of research resources will inform both the Schumpeterian questions about market structure and technological change and the antitrust questions about non-economic values such as the Jeffersonian dispersion of power for its own sake. II. Survey Data to Illustrate Redirection of R&D Resources. I shall use new evidence about the Small Business Innovation Research (SBIR) Program to illustrate the redirection of R&D activity by means of government financing of privately performed R&D. Early in 1999, the National Academy of Sciences (NAS) directed a study of the SBIR projects sponsored by the Department of Defense (DoD). 6 As a part of that study, a large sample of SBIR projects were surveyed to gather information about the sponsored projects and the small businesses performing the projects. 7 Among other things, the information identifies the projects for which the small businesses performing the research were able, early in their research projects, to secure substantial amount of outside private financing from either venture capitalists or other companies. 8 I begin here with the 101 DoD SBIR projects begun in 1996 that were continued into the substantial second stage of funding by DoD and not dropped by the performing firms and for which survey data were available. 9 The statistics below are for the subset of 76 projects for which all of the variables used in the analysis were available. In 1996, DoD began its Fast Track Program that encouraged firms to find substantial outside financing early in their research projects. Obtaining such funds to combine with the government funding qualified the project for the fast track of a high priority for the more substantial second stage SBIR funding that followed the initial low-budget research in the first

6 6 stage of the project. Firms that had previous SBIR awards were required to find more outside financing (to qualify their projects for Fast Track status) than required of firms new to the program. The survey information combined with other information in the SBIR project files of DoD strongly suggests that the government R&D supports projects and firms that are different from those the private sector chooses to support. 10 The government provided support for all of the projects in the sample; the private sector provided outside, third-party support for just a subset of the projects. The private sector was much less likely to provide outside finance early in the research projects even with the government picking up a large portion of the R&D bill for certain types of technologies and certain types of firms. A probit model is used to estimate the probability of obtaining substantial early outside finance, with substantial meaning sufficient to qualify for Fast Track status. The model controls for whether or not the company had previously won a Phase II SBIR award and hence would require more outside support to achieve Fast Track status. The government s matching requirements were more generous to the firms that were new to the SBIR program. For the new firms, less of the total R&D investment was required to come from the third-party investor. A lower probability of outside finance for projects at firms with previous awards is expected, other things being the same, because with less of the investment bill picked up by the government, the expected rate of return to the private investor is less. Other things being the same, a project at a firm with a previous award would require more private investment funds to reach the total amount invested. Yet it would have the same stream of future returns as a project at a firm that was new to the SBIR program and therefore needed less private investment dollars because the government would provide more of the total R&D investment. A

7 7 lower probability of outside finance would also be expected if the firms with previous awards were engaged in research with less commercial potential than the new firms attracted to the SBIR Program by the Fast Track initiative. Apart from the effect associated with previous Phase II awards, the probit model shows that among the set of projects receiving public funding, some technologies and some geographic areas and some types of firms are much less likely to receive early outside third-party financial support for their SBIR projects. Table 1 provides both a parsimonious specification and a more inclusive model to illustrate the stability of the results; a variety of specifications yield similar conclusions. In specification (2) of Table 1, each project was assigned to as many as six different technology areas. Thus, each project could have multiple technology area dummy variables set equal to 1. If instead one replaces specification (2) with the alternative specification using only the primary technology area for each project and including all of the other controls again as well, the effects of the three main variables prior SBIR awards, founders with business experience, and minority ownership are more significant with effects that are slightly larger. 11 Although those results (shown in the preceding footnote) are somewhat larger and more significant, the results shown in Table 1 s second specification use more information and are better estimates in my opinion. For all of the specifications, apart from the effect of previous awards and the area effects of technology and geography, firms whose founders include individuals with a business background and firms that are not minority-owned are more likely to secure outside financing early in the research and development process. Table 2 provides descriptive statistics for the sample. Because the variables are 0-1 qualitative variables, the averages in Table 2 show the proportion of the sample for which a variable takes the value of 1.

8 8 TABLE 1 GOES ABOUT HERE. TABLE 2 GOES ABOUT HERE. Just as for the similar result in Blanchflower, Levine, and Zimmerman (1998), the minority-ownership effect is large, statistically significant, and stable. The minority-ownership effect shown in Tables 1 and 2 is similar across many different specifications. Apparently, the probability of securing outside, third-party finance for these risky SBIR R&D projects is about 0.3 to 0.4 less for companies with minority ownership. 12 In the light of the work of Blanchflower et al., it appears that as with the business loans they study, in the circumstances of the SBIR program, third-party financiers consider the minority-ownership, other observable characteristics of the firm and its R&D project being the same. 13 The foregoing survey data probit results are consistent with, among other things, the hypothesis that the private sector declines to provide financing for risky R&D projects of small businesses, even given fairly high expected rates of return on the R&D costs, because of the costs in addition to the R&D costs that an outside investor would be expected to incur. Those additional costs for the outside financier include the costs of gathering information about the small business performing the R&D, the costs of monitoring the R&D investments, and the costs of mentoring the small business regarding business decisions. The effects of such costs on the probability of obtaining outside finance, as well as differences in appropriability and risk conditions, would be reflected in the probit model s estimated effects for technological and geographic areas and for the various other characteristics of the businesses and the R&D projects.

9 9 III. Case Studies to Illustrate Social Benefits. Case studies also support the view that risky R&D projects of small businesses often cannot obtain outside private financing despite having fairly high expected rates of return. Additionally, the case studies imply that such projects that would not be financed by the private sector have high social rates of returns because of the spillover effects benefiting firms other than the small business performing the R&D. The cases reported here are the early stages of seven small New England businesses attempting to establish new products with R&D investments that were in substantial part financed with public funds. These are selected results from case studies (Scott, forthcoming) of 14 recent DoD SBIR projects. The 14 projects, all in the New England states of Massachusetts, Connecticut, and Rhode Island, researched a variety of technologies in a variety of business settings. The results here are for the seven projects at seven companies that were in the early years of their corporate life and each trying with R&D projects to establish a major commercial product. The remaining seven projects studied were for larger companies or companies that had been established with earlier products. Of the seven small, young companies examined in the present paper, four are based in Massachusetts, two in Rhode Island, and one in Connecticut. At the time that the research projects studied in this paper began, three of the companies had three employees each, one had five employees, one had seven employees, and two had eight employees. Table 3 provides some characteristics of the seven R&D projects; they were essentially contemporaneous, each beginning in 1995 or TABLE 3 GOES ABOUT HERE.

10 10 The technologies researched by the seven young companies were a tele-robot to be used for surgery, low-temperature batteries for use in space, a high-quality laser for use in communications markets, a sensor made with advanced active materials and used to eliminate harmful vibrations in structures, diagnostic test equipment for discovering defects in aging aircraft wiring, a compact reduced-weight system of gears used for high-precision positioning of equipment, and "quantum dot" electronic phosphors to be used in large screen projection systems. Based on interviews with the seven young companies, the principal finding about the early stages during which they are attempting to use R&D to establish a major new product is that at the outset of the R&D projects, for each project the expected private rate of return in the absence of public funding was less than the hurdle rate for the project required by outside investors, and that rate in turn was less than the expected social rate of return on the project. 14 For the seven projects, the average of the estimates for the private rates of return to the R&D projects without public funding was 28.4%. 15 The average of the seven estimates of the lowerbound social rates of return for the R&D projects was 53.4%. 16 The social rates of return for the projects exceed the private rates of return to the small companies doing the R&D because of anticipated spillovers of the research outcomes yielding commercial benefits for other companies companies other than the small businesses doing the R&D. Based on the interviews with all of the New England respondents, the estimate for the hurdle rate on the R&D projects themselves that would be required by outside private investors before they would make venture funds available for R&D projects of the type examined here was 38.2%. 17

11 11 Using the averages to summarize the story, we have the private rate of return to R&D absent any public funding equal to 28.4%. That expected rate of return falls short of the outside investors' required expected rate of return equal to 38.2%. Thus, in the absence of public funding for the small business research, the R&D would not have been funded and, according to the seven young companies, it would not have been undertaken. Yet, because of the commercially valuable spillovers anticipated, the R&D projects would be socially valuable. The expected lower bound for the social rate of return averaged 53.4% for the seven projects, exceeding even the hurdle rate required by private outside investors. The seven case studies here suggest that when a new business is to be created with R&D performed by a young, small company, market failure may imply private underinvestment in socially valuable R&D. 18 Although the expected private returns for the seven projects were substantial, outside investors required even higher rates of return on the R&D investments themselves, presumably to cover not only the costs of the R&D projects themselves, but the costs of acquiring information about the small businesses and their projects, the costs of monitoring the investments, and the costs of providing marketing and other business guidance to the fledgling companies performing the R&D. Partial public financing of the privately performed R&D of course can be used to raise the private rate of return above the hurdle rate for the R&D project. 19 When spillovers cause the social rates of return to be sufficiently high, such partial public funding allows socially valuable research that would otherwise not be performed. As explained in Scott (forthcoming) and Link and Scott (in NAS, forthcoming), that is the role that public financing has played not only in the seven cases described here, but for DoD SBIR projects more generally. 20

12 12 One interpretation of the case studies is that the government appears to be making grants of R&D funds to small businesses that cannot get outside finance from the private sector. An alternative interpretation would be that these small businesses simply prefer not to share the future profits with outside investors and take the path of least resistance, taking public funds to leverage to the maximum extent their own equity in the R&D projects. The correct interpretation is an open question. The bidding mechanism described in Scott (1998) and Martin and Scott (forthcoming) would provide one way to solve the potential problem of firms using public funds when private funds could have been obtained. Venture capital funds bidding against one another for public support would ideally ensure the minimum expenditure of public funds consistent with the private investors earning a required expected (hurdle) rate of return. IV. Identifying Opportunity Costs. To this point, then, we have evidence that the government is supporting socially useful R&D projects that would not have been funded by the private sector. But I believe there is a problem here that has not been faced squarely. Namely, the opportunity costs of the R&D resources devoted to the projects are not well understood. As a practical matter, evaluations begin by valuing research costs at the present value of the actual research expenditures, rather than attempting to ask directly what the opportunity cost the social value of the best foregone alternative use of the resources might be. Then, there are typically two ways of deciding that in fact such public investments are socially useful. One is to observe that the social rate of return on the projects exceeds some stated social discount rate, or required rate, of return. For example, in evaluations of publicly performed and financed infrastructure research, NIST uses the real social discount rate of seven percent as established by the Office of Management and Budget (OMB). 21 The other is simply to observe that the expected social rates of return are very high

13 13 and much higher than the private required rates of return for the investors involved with the R&D project. The problem is that neither approach directly asks what would be society s best alternative use of the specialized R&D resources. The R&D resources devoted to a particular SBIR project, for example, might have an expected social rate of return of say 80 percent. Yet, there may well be an alternative investment yielding a much higher social rate of return, despite the obvious fact that 80 percent exceeds both the OMB s seven percent and the private required rate of return for the investors involved in the project. If public funds are to be used to redirect the path of technological change, effort must be made to understand the opportunity costs of the R&D resources to be used in government financed projects. Identifying the opportunity costs of the R&D resources may seem daunting enough at this point. But the problem is more difficult than it might seem. Following Griliches (1958) and Mansfield et al. (1977), the social and private rates of return that we have described are based on the concept of total economic surplus and its components, producer and consumer surplus. The problem of determining opportunity costs for the R&D resources would then appear to be the problem of identifying economic surplus for different R&D projects. However, our surplus measures and associated rates of return are based on the current distribution of income and wealth and the choices that consumers make given that distribution. Our surplus measures and choices among projects then reflect what could be called the current economic majority that would cast the dollar votes for various products and services. There are other hidden economic majorities economic majorities associated with other income and wealth distributions that a government should take into account. Indeed, the political process is one way that a hidden economic majority could use the government process to choose a direction for technological change that would not be the choice of the current economic majority. For example, even if the

14 14 current economic majority would prefer to devote R&D resources to an R&D project in the entertainment industry yielding a social return of 80% as compared to the best alternative use of let us say an R&D project to develop low-income housing and medical care that has a social rate of return considerably less than 80%, for some hidden economic majority the inequality might well be reversed. Further, there are non-economic values that may be valued for their own sake despite the fact that they are associated with less total economic surplus than would be possible to obtain. The Jeffersonian ideal of a decentralization of power, whether economic or political or social, for its own sake provides an example. Thus, there are two sensible reasons that governments might direct the use of economic resources away from uses with high social rates of return. First, there may be R&D projects with even higher rates of return conventionally defined that is, defined using actual research expenditures and the present value of total economic surplus given the current distribution of income and wealth. Especially in the redirection of R&D resources that can shape the path of technological change for better or for worse, government needs to address such alternatives explicitly rather than setting a rather arbitrary hurdle rate. Second, governments can rationally choose to deploy R&D resources using an evaluation metric other than the total economic surplus of the current economic majority. 22 V. Concluding Discussion. Using the evidence above about the SBIR program, we see that the U.S. government appears to be funding R&D projects that the private sector does not want to fund. The government supports R&D in technologies and in parts of the country that the private sector is unlikely to support, and the government provides funds for inexperienced businesses and for minority-owned businesses when the private sector declines to provide such support. Are these

15 15 investments improving the present value of the total economic surplus generated by our R&D resources for the current economic majority or a hidden majority? Are they increasing noneconomic measures of value such as achieving the Jeffersonian ideal? Government s redirection of technological change can occur without government financing of privately performed R&D. Government may actually perform the R&D it finances. Typically, that occurs when the government provides infrastructure R&D such as that done by the National Institute of Standards and Technology. For example, infrastructure R&D in information technology can not only redirect private R&D into an area because it is rendered more productive (Scott, 1999), but it can provide an environment where a wider range of types of firms can successfully innovate and compete. 23 Further, government can of course redirect R&D resources without government R&D expenditures. For example, when Title III of the Clean Air Act Amendments of 1990 required that regulations be developed to reduce problems caused by toxic air emissions of specified chemicals, industry responded with new R&D efforts to study and reduce such emissions (Scott, 1996 and 1997). Even when R&D tax credits would be ineffective, an increase in taxes themselves can redirect and stimulate R&D efforts (Scott, 1995). Or, governments can redirect R&D investments by encouraging research joint ventures, although such ventures may not always improve performance of technological change. R&D spillovers are often cited as a reason to support antitrust policy, such as the U.S. National Cooperative Research Act of 1984, that encourages research joint ventures. There is strong evidence of R&D spillovers across industry categories (Scott and Pascoe, 1987). 24 However, the cooperative R&D encouraged by such laws may not promote R&D using shared knowledge because the cooperative R&D may bring together diverse R&D activities that were or could be brought together by competing diversified firms investing independently. The cooperation may

16 16 actually reduce desirable R&D investment because it reduces the competition among such diversified firms (Scott, 1988; 1993, pp , ). 25 Of course, while cooperation may reduce total R&D investments in the area of the cooperative research, it may increase other R&D because firms who are successful in the cooperative venture may expand the scope of related research. Further, although there is no general presumption that R&D joint ventures encouraged by modification of antitrust laws will improve R&D performance in industry, there is evidence that some types of cooperative R&D lead to R&D investments that would not otherwise occur (Scott, 1996). Additionally, government policy can encourage university participation in research ventures with private firms. The evidence suggests that universities are invited to participate in the ATP research projects with more difficult conceptual problems, but nonetheless their presence reduces the probability that the projects will fail (Hall, Link, and Scott, 1998). 26 The conventional finding of a social rate of return exceeding the private hurdle rate for a project, although important and useful information about social value, is neither necessary nor sufficient as an indicator or the social usefulness of an R&D project. Understanding the need for sufficient indicators of social usefulness places the Schumpeterian and Jeffersonian industrial organization issues in a new light. Public policies to redirect the use of economic resources to stimulate the R&D of small, focused firms without market power or to promote the ideal of a dispersion of power more generally may well be economically efficient by the yardstick of some, perhaps hidden, economic majority. Economists can help to identify other applications of the resources with even higher social returns, and some of those applications may reflect the economic surpluses associated with different economic majorities than the current one. As citizens we can all help decide among the optimal allocations for the different economic majorities and among the solutions to the resource allocation problem using non-

17 17 economic metrics such as the Jeffersonian ideal. Certainly the evaluation task set out here is more difficult than making evaluations based on economic surplus given the current distribution of income and wealth. However, the difficult task is one that I believe government and society should understand and face squarely. Opening evaluations of R&D resource allocations to broader perspectives has a Pandora s Box aspect, but for evaluation of long-run R&D performance, the difficulties may be far less disruptive socially than hiding behind the conventional analysis would be.

18 18 Notes * This essay was prepared as the Presidential Address to the Industrial Organization Society, for the Society s meeting in Boston, January 7, I thank David Audretsch, Bronwyn Hall, Albert N. Link, and Stephen Martin, who have been my co-authors for many of the recent studies cited. Special thanks are due to Albert N. Link who has been my research partner in almost all of the recent projects, to Nancy A. Scott for discussions about alternative economic majorities, and to Danny Blanchflower and Alan L. Gustman for discussions about the effect of minority ownership. I also thank William L. Baldwin and William G. Shepherd for their comments on the first draft of the essay. 1. For OECD, I studied financial leveraging of public funding of private research. For NIST, I studied publicly performed R&D in laboratories of that federal laboratory, partially publicly financed and privately performed R&D of NIST s Advanced Technology Program (ATP), university participation in ATP projects, and the Baldrige National Quality Award of NIST s Office of Quality Programs. For the NAS, I studied the DoD s SBIR program. For NSF, I have studied technology indicators based on licensing agreements. I am grateful to Jean Guinet of OECD, Gregory Tassey of NIST s Program Office, Rosalie Ruegg of NIST s ATP, Charles Wessner of NAS, and John Jankowski of NSF for the opportunities they have provided. 2. Perhaps Lichtenberg s (1987) concern about the link between publicly funded and privately funded R&D and sales to the government versus private sales could explain the findings of Mansfield and Scott and other studies that find a complementary relationship between government and private R&D spending. However, the overarching issue of the social desirability of the government funding remains. To reiterate, even assuming displacement of

19 19 private R&D or assuming that new private R&D is focused on sales to the government rather than to the private sector, the R&D funded by the government and performed privately may well have greater social value, though not greater private value to the R&D performing companies in the absence of any government funding. The facts are all consistent with the government funding having stimulated desirable social investments investments that should be performed from society s standpoint. 3. As is typically the case, one can find studies that reach different conclusions about identical samples. For example, using a different methodology (a case study rather than a time series econometric study), and focusing on the benefits from the consortium rather than on the possible displacement of private spending by government spending, Link, Teece, and Finan (1996) reach a very different conclusion about the Sematech case that Irwin and Klenow (1996) study. 4. Link and Scott (International Journal of Industrial Organization, forthcoming) make this point and elaborate on it in the context of a study of ATP projects. 5. For discussion of the Jeffersonian ideal of dispersion of power and its relationship to the promotion of small business associated with public policies, including antitrust policy, see Baldwin (1987, pp ). 6. The study provides detailed information about the SBIR program of DoD and about the data gathered and is available in National Academy of Sciences (forthcoming). For a seminal discussion and evaluation of the SBIR Program more generally, see Lerner (1996). 7. The survey is described in Cahill (forthcoming). 8. These projects were Fast Track projects, for which the performing companies obtained outside finance early in particular, before Phase II of the SBIR project. In Phase II,

20 20 much more substantial R&D investments, both greater in dollar amount and greater in time, are made after the initial, low-cost quick look of Phase I established the potential for the investment. 9. These 101 observations are a subset of the larger sample used, for different purposes, by Cahill (forthcoming) and Audretsch, Link and Scott (forthcoming). 10. I thank my fall term, 1999, seminar students at Dartmouth College who undertook the analysis of the data, with permission from NAS, as their term project. The results that I discuss were originally discovered by my students: Sara J. Becker, Daniel E. Bernstein, Virginia F. DeJesus-Rueff, Meghan M. DeMark, Jason C. Deeken, Christopher G. Ellis Jr., Daniel L. Herlihy, Athena A. Maikish, and Benjamin M. Moor. 11. For the specification analogous to Table 2 s second specification but using just the information about primary technology area, the estimates are prior awards with df/dx = -0.35, z = -1.9, P> z = 0.06, business founders with df/dx = 0.42, z = 2.3, P> z = 0.02, and minority ownership with df/dx = -0.38, z = -2.3, P> z = In this specification just three observations were lost because of perfect prediction, so n = Thus, for example, other things being the same, a minority-owned firm would have a probability of securing early outside finance of 0.1 to 0.2 if for a non-minority-owned firm the probability of early outside finance were However, given the unique circumstances of the SBIR program, it would be possible, although I think unlikely, that firms with minority owners are less likely to want to use outside finance if they can get DoD SBIR financing without the higher priority for funding that Fast Track status would confer. In the work of Blanchflower, Levine, and Zimmerman, the firm had

21 21 to apply for a loan to be denied a loan; here, a minority-owned firm could have simply decided not to seek outside funds and hence not to apply for Fast Track status. 14. Scott (forthcoming) provides details about all of the case studies and about the interview-based methodology for estimating the rates of return. Additional discussion of the methodology is provided in Link and Scott (in NAS forthcoming). 15. The private rate of return without public funding for the seven projects averaged with standard deviation The lower bounds for the social rates of return for the seven projects averaged with standard deviation The social rates of return are lower bounds because only the producer surplus created by the innovative investment was captured in the returns, leaving new consumer surplus unmeasured. 17. Only seven of the respondents in the full sample of 14 projects were confident enough about the answer to offer an estimate. For the seven respondents, the average hurdle rate for these high-risk small business R&D projects was with a standard deviation of Not surprisingly given their limited experience, among the seven young firms that are the focus of this study, an even smaller proportion were confident about the hurdle rate for outside investors. Just two of the seven provided an estimate, with the average being 33%. For purposes of the discussion in the text, I use the estimate based on the full sample, because I suspect it is more accurate. However, the story remains the same if the average estimate for the smaller sample is used instead. 18. Martin and Scott (forthcoming) consider market failure more generally and suggest various forms of public support that are tailored to the particular type of market failure.

22 The minimum amount of public funding needed is the amount necessary to raise the private rate of return to the hurdle rate. Of course, typically much more than that minimum amount of funding is provided. For the seven projects studied here, the private rate of return given the public funding provided was 56%. Scott (1998) proposes a bidding procedure that would minimize the amount of public funding needed to ensure that socially valuable research projects are performed by innovative small companies. 20. DoD SBIR projects may embody circumstances where partial public funding of privately performed research is especially likely to work well. The projects are typically in areas where the government has procurement interests. Even when the projects have obvious commercial, non-military applications, the research is related to the mission of a DoD agency. Nelson (1982, pp ) has observed that for "government procurement-oriented R&D... a government agency spends to further its own reasonably well defined purposes, and its evaluation of the technology emanating from R&D determines whether the new technology will be used or not." Given "a recognized public interest in certain kinds of advances" along with "a government agency [that] stands ready to see that the fruits of R&D are employed" (Nelson, 1982, p. 461) direct government grants to the private sector can work well. 21. Office of Management and Budget (1992) and Link and Scott (1998). 22. The arguments can be made for economic resources in general, rather than for R&D resources in particular. My focus in this essay is of course on the redirection of R&D resources, and it is for policy regarding R&D resources that I have had first-hand experience with primary data and its sources. 23. Government laboratories, such as the National Institute of Standards and Technology in the U. S., perform public R&D to provide infrastructure technology such as measurement

23 23 and test methods, standard reference materials, and scientific and engineering databases to the private economy (Tassey, 1999, pp ). Without public investment in such infrastructure technology, industry left on its own would be expected to underinvest in supporting technology infrastructures because of market failures (Tassey (1999, p. 23, p. 40). 24. Scott (1993, pp ) reviews several other studies that demonstrate effects of R&D spillovers across industry categories. Sharing of knowledge across industries and firms can of course be contractual, and in addition to research joint ventures that have been extensively studied, licensing of patented technology provides another important way to share technology. Link and Scott (1999) document the importance of licensing in pharmaceuticals, industrial chemicals, electronics, and software, and demonstrate that licensing arrangements are determined by complementarities in the patent portfolios of the licensors and licensees. 25. Note that this possibility is distinct from the empirical fact that no general statistical relationship between seller concentration and R&D intensity has been established when other factors affecting R&D are controlled. In the absence of complete controls for firm and industry effects, relationships can be found, but with a complete set of controls any relationship vanishes (Scott, 1984; Scott, 1993, pp , ). Hence, broad antitrust policies aimed at affecting seller concentration cannot be justified based on the facts available about concentration and R&D. Certainly there is ample reason to expect relationships between the number and size distribution of firms in an industry and their R&D investments (Scott, 1993, chapter 8; Baldwin and Scott, 1987). My point is that there is no evidence of a broad statistical relationship once firm and industry effects are controlled.

24 The Advanced Technology Program within the U.S. Department of Commerce's National Institute of Standards and Technology provides partial public funding for generic and technology-enabling research performed by private companies.

25 25 References Audretsch, D. B., Link, A. N., and Scott, J. T., "A Statistical Analysis of the National Academy of Sciences' Survey of Small Business Innovation Research Awardees: Analyzing the Influence of the Fast Track Program, in the conference volume for the National Academy of Sciences Symposium on the Assessment of the SBIR Fast Track Program, May 5, 1999, forthcoming. Baldwin, W. L., Market Power, Competition, and Antitrust Policy (Homewood, Illinois: Irwin, 1987). Baldwin, W. L., and Scott, J. T., Market Structure and Technological Change, in the series Fundamentals of Pure and Applied Economics, vol. 17 (Chur; London; Paris; New York: Harwood Academic Publishers, 1987). Blanchflower, D. G., Levine, P. B., and Zimmerman, D. J., "Discrimination in the Small Business Credit Market," manuscript, Dartmouth College, November 1, Cahill, P., Fast Track: Is it Speeding Commercialization of DoD SBIR?, in the conference volume for the National Academy of Sciences Symposium on the Assessment of the SBIR Fast Track Program, May 5, 1999, forthcoming. Griliches, Z., Research Costs and Social Returns: Hybrid Corn and Related Innovations, Journal of Political Economy, 66 (1958), pp Hall, B. H., Link, A. N., and Scott, J. T., Universities as Partners in Research Joint Ventures, a report to the Advanced Technology Program, National Institute of Standards and Technology, December Irwin, D. A., and Klenow, P. J., High-tech R&D Subsidies Estimating the Effects of Sematech, Journal of International Economics, 40 (1996), pp

26 26 Lerner, J., The Government as Venture Capitalist: the Long-run Impact of the SBIR Program, NBER Working Paper 5753 (NBER, Cambridge, Massachusetts, 1996). Lichtenberg, F. R., The Effect of Government Funding on Private Industrial Research and Development: A Re-Assessment, Journal of Industrial Economics, 36 (September 1987), pp Link, A. N., and Scott, J. T., Public Accountability: Evaluating Technology-Based Public Institutions (Boston, Massachusetts: Kluwer Academic Publishers, 1998). Link, A. N., and Scott, J. T., "Estimates of the Social Returns to SBIR-Supported Projects," in the conference volume for the National Academy of Sciences Symposium on the Assessment of the SBIR Fast Track Program, May 5, 1999, forthcoming. Link, A. N., and Scott, J. T., Development of an Industrial Database on Licensing Patterns, Final Report submitted to the National Science Foundation, Division of Science Resources Studies, Research and Development Statistics Program, SGER Project , July Link, A. N., and Scott, J. T., Public/Private Partnerships: Stimulating Competition in a Dynamic Market, International Journal of Industrial Organization, forthcoming. Link, A. N., Teece, D. J., and Finan, W. F., Estimating the Benefits from Collaboration: The Case of SEMATECH, Review of Industrial Organization, 11(October 1996), pp Mansfield, E., R&D and Innovation: Some Empirical Findings, in R&D, Patents, and Productivity, ed. Z. Griliches, chapter 6, pp (Chicago: University of Chicago Press for the National Bureau of Economic Research, 1984).

27 27 Mansfield, E., Rapoport, J., Romeo, A., Wagner, S., and Beardsley, G. Social and Private Rates of Return from Industrial Innovations, Quarterly Journal of Economics, 91 (1977), pp Martin, S., and Scott, J. T., "The Nature of Innovation Market Failure and the Design of Public Support for Private Innovation," Research Policy, forthcoming. National Academy of Sciences, Board on Science, Technology, and Economic Policy, National Research Council, The Small Business Innovation Research Program: The Fast Track Pilot, a conference volume for the National Academy of Sciences Symposium on the Assessment of the SBIR Fast Track Program, May 5, 1999, forthcoming. Nelson, R. R., Government Stimulus of Technological Progress: Lessons from American History, in Government and Technical Progress, ed. R. R. Nelson, pp (New York: Pergamon Press, 1982). Office of Management and Budget, Circular No. A-94: Guidelines and Discount Rates for Benefit-Cost Analysis of Federal Programs, Washington, D. C., Scott, J. T., "Firm Versus Industry Variability in R&D Intensity," in R&D, Patents, and Productivity, ed. Z. Griliches, chapter 10, pp (Chicago: University of Chicago Press for the National Bureau of Economic Research, 1984). Scott, J. T., "Diversification versus Co-operation in R&D Investment," Managerial and Decision Economics, 9 (September 1988), pp Scott, J. T., Purposive Diversification and Economic Performance (Cambridge; New York: Cambridge University Press, 1993). Scott, J. T., "The Damoclean Tax and Innovation," Journal of Evolutionary Economics, 5 (February 1995), pp

28 28 Scott, J. T., "Environmental Research Joint Ventures among Manufacturers," Review of Industrial Organization, 11 (October 1996), pp Scott, J. T., "Schumpeterian Competition and Environmental R&D," Managerial and Decision Economics, 18 (1997), pp Scott, J. T., "Financing and Leveraging Public/Private Partnerships: The Hurdle-Lowering Auction," STI (Science, Technology, Industry) Review, 23 (1998), pp Scott, J. T., "The Service Sector's Acquisition and Development of Information Technology: Infrastructure and Productivity," Journal of Technology Transfer, 24 (April 1999), pp Scott, J. T., An Assessment of the SBIR Program in New England: Fast Track Compared with non-fast Track Projects," in the conference volume for the National Academy of Sciences Symposium on the Assessment of the SBIR Fast Track Program, May 5, 1999, forthcoming. Scott, J. T., and Pascoe, G., "Purposive Diversification of R&D in Manufacturing," The Journal of Industrial Economics, 36 (December 1987), pp Tassey, G., "R&D Trends in the U.S. Economy: Strategies and Policy Implications," Planning Report 99-2, Program Office, Strategic Planning and Economic Analysis Group, National Institute of Standards & Technology, U.S. Department of Commerce, April Wallsten, S., Can Government-Industry R&D Programs Increase Private R&D? The Case of the Small Business Innovation Research Program, manuscript, November 1997, Stanford University.

29 29 Table 1. Determinants of the Probability of Outside Third-Party Finance Early in the R&D a x df/dx (z, prob> z ) prior Phase II SBIR awards (-2.42, 0.016) (-1.67, 0.095) founders with business background (2.64, 0.008) (1.46, 0.143) minority ownership (-1.65, 0.100) (-1.82, 0.069) technology area effects b no yes geographic area effects c no yes agency effects d no yes observed probability predicted probability (at the means of the variables) number of observations e chi-square (degrees of chi-square(3) = 16.2 chi-square(18) = 48.8 freedom) probability > chi-square pseudo R Ln Likelihood a From the probit model of the probability F, with all other variables at their means for these qualitative variables, the table shows df/dx for the discrete change in x from 0 to 1, and shows z statistics for the underlying coefficients in the probit index of the x variables along with the associated probability value.

30 30 b The technology areas are computer, information processing and analysis; electronics, materials, mechanical performance of vehicles, weapons, and facilities; energy conversion and use; environment and natural resources; and life sciences. Each project has been assigned to as many as six different technology areas, and therefore can have as few as one or as many as six of the technology area effects. The results are quite similar whether all of the information about technology areas is used in that way, or if instead, just the primary technology area for each project is used. c The geographic areas are the west, southwest, mid-west, south, southeast, east, and northeast. The west is left in the intercept of the second specification s underlying probit model. d The agencies are the Air Force, Army, Ballistic Missile Defense Office, Defense Advanced Research Project Agency, Defense Special Weapons Agency, and the Navy. The Army is left in the intercept of the second specification s underlying probit model. e Recall that each project was assigned to as many as six different technology areas. All but one of the 76 projects has electronics as one of their technology areas, and all but two had materials as one of their technology areas. There was just one project assigned to the Defense Special Weapons Agency. Of course, then, the dummies for the technology areas of electronics and materials when not equal to 1, and the dummy for the Defense Special Weapons Agency when not equal to zero, predicted failure perfectly. As a result those three dummy variables were dropped and four observations were not used in the second specification.

31 31 Table 2. Descriptive Statistics a Variable Mean Standard Deviation Fast Track Business founders Prior Phase II SBIR Minority ownership Computers Electronics Materials Mechanical Energy Environment Life sciences Air Force BMDO DSWA Navy DARPA Army West Southwest Midwest South Southeast East Northeast a The number of observations for each variable is 76. All variables are 0-1 qualitative variables. Their means, therefore, show the proportion of the sample with the given characteristic. Thus, 39.5 percent of the sample had Fast Track status because significant outside finance was obtained early in the research project. All but one of the projects had one of its up to six different technology areas in electronics, and so forth.