|Make your work easier and more efficient installing the rrojasdatabank toolbar ( you can customize it ) in your browser.|
Technology Policy and Economic Growth
Michael Borrus and Jay Stowsky
BRIE Working Paper 97
Michael Borrus is co-director of the Berkeley Roundtable on the International Economy (BRIE) and adjunct Professor at the University of California, Berkeley;
Jay Stowsky is Director of Research Policy and Development for the University of California system and served, from 1993-95, as Senior Economist for Science and Technology on the staff of the President Clinton's Council of Economic Advisers.
This was prepared for Investing in Innovation. Lewis Branscomb, (ed.) (Cambridge, MA: MIT Press, 1997)
Table of Contents
Between the idea
Technology policy is obscured in deep shadow: The idea in American practice is to let the market decide industrial fortunes, but a half century of government sponsorship of new technology industries from jet aircraft to electronics and biotechnology suggests a different reality: Unacknowledged U.S. practice often contradicts what is preached, to enormous economic benefit. With the cover of the Cold War gone, it is time to move technology policy into the light -- not to the patchy fluorescence of Bill Clinton's first term, but to the bright spot of center-stage. In what follows, we suggest why and how this should be done.
A nation's standard of living is the most significant indicator of its economic performance. Productivity growth (the rate of growth in output per unit of input, usually output per worker), income distribution, and the unemployment rate are the three variables that most directly affect the standard of living of large numbers of people. Over the past several decades, the United States has been doing especially poorly on productivity growth and income equality relative to its past domestic performance, to the performance of the major economies of Europe and Asia, and to what our own resources ought to permit. Relative to these indicators, U.S. performance remains poor even after five straight years of reasonable economic growth, despite the widely held perception that America's mid-1980s problems with competitiveness have been solved. While many U.S.-owned firms are indeed prospering again on world markets, the nation's long-run economic prospects remain troubled. Persistent income inequality and slow productivity growth threaten to undermine the nation's relative standard of living and ultimately its political influence abroad, while simultaneously setting the stage for domestic social unrest.
Essentially all economists agree that productivity growth is the key to doing better over the long term, but they can neither explain why productivity growth has slowed in the U.S. nor what to do to make it grow faster. 4 Most would agree that the answer lies in some combination of a higher level and altered composition of investment -- investment in capital formation (including infrastructure), in people (e.g., training and education), and in technical progress (including new technologies and corresponding new ways of organizing industrial activities). Of these variables, better technology is usually deemed the most significant. Even economists skeptical about technology policy admit that "technological progress is a vital source of economic growth and R&D a vital source of technological progress." . According to the widely cited growth-accounting literature, traditional factor inputs like capital and labor can not account for at least one-quarter and perhaps as much as one-half of the total U.S. growth rate since the end of World War II. This large residual is attributable to advances in technical know-how.
Other strands of the economics literature also emphasize the relationship of technological progress to economic performance. "New growth theory" emphasizes that the rate of economic growth is driven by the total stock of human capital -- the collection of knowledge or innovative "ideas" held at any one time by people in businesses, universities, and governments. This approach contends that new ideas are the source of growth because they lead to technological innovation and hence to productivity improvements. Thus, if too few resources are dedicated to education and scientific research and development (which both increase the stock of human capital or new ideas), then the rate of economic growth will be lower than it could be.
Compounded over generations, a 1 or 2 percent reduction in the overall growth rate could be the difference between the standard of living merely doubling or surging five-fold over a hundred year period. For countries with similar standards of living today, small differences in the rate of growth could lead to very different economic outcomes in the future. In addition, studies by economic historians and theorists of increasing marginal returns suggest that, once an economy sets out on a high-growth or low-growth path, both high growth and low growth may be self-reinforcing over time. For example, Argentina and the United States had roughly similar levels of economic performance during the 1860's, but the U.S. managed to launch itself on a high-growth path while Argentina became mired in low-growth which reinforced itself over time. Thus, rather than try to evolve from low growth to high growth through rapid industrialization, a strategy which would typically focus on the development of basic industries and a capacity for low-cost mass production, a relatively poorer country might do better to invest resources in higher education and high-tech R&D - that is, try to jump quickly onto a high growth path. Indeed, that appears to be the new strategic focus of developmental strategies in emerging Asian economies like Malaysia, the region around India's Bangelore, and the coastal provinces of mainland China.
Finally, a recent review of efforts to measure the rates of return on investments in new technology found that the private rates of return to firms performing R&D often vary from 20 to 30 percent in a variety of industries. In comparison, the average rate of return to investment in the business sector as a whole is about 10 percent. Estimated rates of return from R&D to society as a whole, due to beneficial spillovers from an initial R&D investment to consumers and to other firms, vary from 20 percent to well over 100 percent in a variety of industries, with an average near 50 percent. The channels of diffusion of the spillovers vary considerably, and their effects on productivity growth are sizable. These results also suggest that private firms may substantially underinvest in R&D because they can not internalize the significant returns to others, including competitors.
In sum, various economic approaches that use disparate methods all acknowledge a strong link between advances in knowledge, technical progress, and long-term growth in productivity and GDP. There is also general consensus that high rates of investment in broad-based R&D across a wide technical frontier are essential, and that a modern technical infrastructure and a skilled, technically competent work force are crucial complements to achieving sustained productivity growth.
Beyond the consensus on goals, however, economists do not agree on how to craft policies to meet these goals. First, there is broad disagreement about the composition of investment, whether public or private. There is no consensus about what the proper balance of investment ought to be between industry, government and academia. There is no agreement about whether government's role should include direct R&D dollars or be limited to investment incentives for private R&D spending (as with R&D tax credits). There is no agreement about whether public funding should be limited to cases of private market failure (as when private returns to R&D spending are small relative to social returns, providing little incentive for investment by private market actors) -- or if it should include a focus on government missions (e.g., defense) and social needs (e.g., health).
Second, there is substantial contention, highly contingent on period studied, over whether new technologies displace or create jobs on balance -- although the bulk of evidence suggests that over the long term, new technologies are capital saving, require the labor force to become more skilled, and thus create more challenging, higher paying jobs.10 In general, economists presume that new technology will generate more jobs than it eliminates, as it leads to new products and services, lower prices, and expanded markets. Unfortunately, there may be prolonged lags between job losses and new job creation, and the new jobs may not be appropriate for those displaced by technological change. In that adjustment process, income inequality can be greatly exacerbated. For example, the Industrial Revolution in Great Britain eventually produced a large and unprecedented middle-class, but only after creating a decline in the real wages of most workers - and tremendous resistance to change - during its first half-century. The consensus answer to the time lag problem - and it is the Clinton administration's answer as well - is that the compensating demand effects that offset job loss from technological change come more quickly when overall economic growth is strong and when markets for both labor and capital are flexible. Government can help by avoiding recessions and making workers more adaptable through improvements in education and training.
Third, although there is basic agreement that the vast bulk of social benefits from technology flow from its application and widespread diffusion, it is unclear whether an economy the size of the U.S. must be a leading-edge producer of a new product or a new piece of production equipment in order to reap the full benefits from its use. Technology policy proponents typically argue that the initial establishment of a dominant position in markets for an advanced technology can lock in control of a long stream of follow-on product and process innovations, making market entry much harder for technology "followers." This locked-in position constitutes what game theorists and business analysts refer to as a "first mover advantage." This means that a temporary market advantage can turn into a more enduring technological advantage. A more conventional argument holds that it might actually be to the economic advantage of manufacturers to be second -- to absorb the spillovers from investments made initially by competitors (foreign or domestic), and thus start production further along the technological learning curve. In practice, there are good examples of both strategies succeeding (and failing). Such issues are likely to remain unsettled. For the foreseeable future, there is no algorithm or formula that can identify the best balance of choices and optimize the economic returns. Where one stands is ultimately a normative judgment about the role of government, the virtue of certain ends, and the relative efficacy of different means. Thus, the current debate pits both conservatives, who oppose an aggressive technology policy as too interventionist, and liberals, who think government should intervene to serve different ends, against various groups of so-called moderates, economic nationalists and self-interested industry trade associations who favor it. These political stances should not be permitted to obscure the real stakes in the debate over technology policy. In light of the probably central role technical progress plays in long-term economic performance, much is at risk either if the conservatives' hands-off prescriptions are followed or if liberals do not include technological development among the objectives for which they believe government should intervene in the economy. If such intervention is either forbidden outright or simply sloughed off in favor of other worthy goals, we will risk a significant sacrifice of opportunities for long-term national economic growth and productivity advance. We will risk, that is, a permanent inability to extract ourselves from precisely the conditions we are now experiencing.
And there is surely cause for long-term concern. Both Europe and Japan project significant increases in civilian R&D over the next few years, with Japan proposing to double R&D spending between 1995 and the year 2000. U.S. spending, particularly federal spending, may be headed in the opposite direction. Recent proposals, both from congress and the president, would cut federal R&D spending between 10 and 30 percent over the next seven years. Other proposals would skew federal spending toward basic R&D while eschewing an active technology policy. The United States could enter the next century spending less on technological innovation than its major competitors - less in absolute dollars as well as in percentage of GDP - for the first time in the postwar era.
The public debate on technology policy is typically truncated into the issue of picking winners and losers, leading to two facile conclusions: First, that markets do that most effectively; and second that porkbarrel politics is more likely to support the losers anyway. This neat two-step eliminates from the role of technology policy everything for which government is institutionally well-suited, from infrastructure building and investment incentives to support of skills training. It then notes that what is left is, of course, institutionally more appropriate for the market. The argument is legitimated simultaneously by our ancient faith in markets and our recent cynicism about politics.
In fact, even accepting the critic's definition of the issue, there are limiting cases in which the reductionist conclusion about picking winners and losers is not defensible. The most important is the development of new technologies, for which markets are not entirely adequate institutions. As previously noted, empirical evidence suggests that as a result of spillovers of all kinds, the social returns to R&D spending on new technologies far exceed the private returns, perhaps by as much as 50 to 100%. Appropriability problems lead to over-investment in some technologies and under-investment in others relative to the social optimum.
Markets also deal inadequately with technological progress because of the highly contingent nature of innovation. Rather than being preordained by scientific logic, technology development is contingent upon the actions of developers, producers and users, as they perform their respective roles, interact, and accrue different kinds of know-how over time. The contingent nature of technical progress means that perfect information is impossible; neither innovators nor the private capital markets that fund them are fully capable of accurately evaluating the risks involved. Therefore, private capital markets and innovators alike must misallocate their investment and effort. Some bets will pay off big; some not at all. Winners and losers can only be positively identified in the revealing gaze of hindsight.
This is as true for private as public investment. For every Apple Macintosh there are normally several Altairs and Amigas. For every IBM there are several GEs and RCAs whose technological bets on mainframe computers failed to pay off. For every Intel there are defunct Molectros and AMEs. For every winner in a venture portfolio, there are untold losers that get nowhere near the publicity. Indeed, there is absolutely no evidence, beyond the economist's leap of faith, that private investment is any more capable than public investment of separating the winners from the losers before the fact. The major difference is that private losers exit the market, while publicly backed losers are held to the higher standard of wasting taxpayers' money.
In short, picking winners and losers is the wrong metaphor to characterize the government's socially useful and necessary activity of supporting the process of innovation. Government is actually placing bets on our collective future. From the public standpoint, the magnitude of the potential social gains is sufficiently large to provide a comfortable margin for error in choosing among technologies to back.
The case against aggressive technology policy thus falls short of damning a significant government role in support of new technology development. But this failure does not by itself justify government support. Two related rationales, one political and the other economic, accomplish that. First, the government is a significant consumer of technology as it provides for our common needs. In areas ranging from national defense to infrastructure, the government must open our wallets to get the technology it needs. Very often that means sponsoring research and procurement that launch new industrial capabilities.
As consumer, the government's demands are usually determined through the political process rather than the market. Those who acknowledge all of the flaws of politics but none of the market are extremely leery of this. They dismiss it by labeling it porkbarrel politics. To be sure, this process can involve substantial, time-honored earmarking of public funds to favored projects in favored Congressional districts, with little regard for overall social benefits. As we discuss later, however, it can also be organized in ways that limit earmarking and maximize the likelihood for public gain. As a process in its own right, American politics may fail to satisfy the economist's dream of perfect efficiency, but the government can hardly fail to respond to constituents' demands.
It is quite possible that politics does effectively what the market does not, namely aggregate the demand of numerous dispersed customers (citizens) who would otherwise have no other way of expressing their collective influence over technological development. In that way, a broader portfolio of socially useful technologies is undoubtedly explored and screened than the market would ever normally permit. Those who see this process as porkbarrel are often lamenting not the economic inefficiency but the lack of expert influence.
The pure economic case in favor of public support to new technologies must rest on the disproportionate importance to economic well-being of high-tech industries that produce substantial value-added domestically. Industries may be strategic for economic welfare in at least three ways: They may contribute a major share of the technological progress that is central to long-term growth; they may provide a higher return to factors of production than could be earned elsewhere in the economy; or they may provide externalities like technological spillovers that broadly benefit the rest of the economy. High-tech industries do all three. First, as suppliers of producer goods (and service inputs), high-tech industries are primary carriers of technological progress. Second, high-tech industries fund a disproportionate amount of industrial R&D, offering innovations that pervasively spill over to the economy as a whole. In the early 1990s, high-technology industries accounted for only about 20 percent of the nation's manufacturing output and 24 percent of its manufacturing value-added, but nearly 60 percent of its private industrial R&D. Third, high-technology industries are high-productivity industries that pay higher compensation than other manufacturing industries. By the early 1990s, value-added per worker in all high-technology industries was one-third higher than the average for all manufacturing and two-thirds higher if only production workers are included. These differentials are significant and persistent. In short, the production of a dollar's worth of silicon chips really is more important than that of potato chips for many principal determinants of economic well-being such as wages, skill formation, productivity, investment, and R&D.
Equally important, the pervasive economic and technological spillovers generated by high-tech industries appear mostly to accrue locally: Domestic production is necessary if the nation is to enjoy many of their economic benefits. While some technical knowledge is footloose -- embodied in products, blueprints, or open technical forums like published research -- much is generated only in the processes of development and production. That kind of technical knowledge accumulates in specialized local assets like labor pools and supplier networks. It is embodied in them and does not diffuse easily.
When U.S. PC assemblers go to Taiwan for design and development of notebook computers, when U.S. disc drive assemblers go to Singapore for processes and volume manufacturing, when IBM moves microsystem development out of the U.S. to Japan, they are seeking access to such specialized and local know-how in components and microsystems' design, processes, and manufacturing. Such domestic capabilities are the probable basis for product differentiation and new technology generation. They help to attract footloose technological know-how originating abroad, and ensure that it can be exploited domestically. In other words, without such domestic capabilities, an economy has no enduring potential to operate at the technological frontier, with all that this implies for maintaining national well-being.
This localization of technology's economic benefits is strongly reinforced by the imperfect nature of technological competition. Modern high technology markets are characterized by extreme scale economies, oligopoly, persistent entry barriers, and often, strong first-mover advantages. Firms or nations that establish an initial advantage -- whether through private competence or government support -- can enjoy those advantages long enough for the economic benefits to accrue domestically rather than abroad. This happened with Japan's concerted efforts to dominate semiconductor memory and display technologies, and Europe's Airbus program.
A bigger national share of global high-tech output can thus mean a bigger share of good jobs and a higher level of economic well-being at the expense of others. That is why international competition in technology-intensive industries often generates beggar-thy-neighbor trade disputes. An aggressive technology policy that aims at sponsoring the development and launch of new technology industries may be a necessity for the U.S. merely to counter the efforts of other nations both to eliminate U.S. technological leadership and to force U.S. firms to remove domestic capabilities from the domestic economy and transplant them abroad.
How effective has the U.S. government's sponsorship of new technologies been? In the postwar period, federal support to new technology crystallized around defense, the development of nuclear energy and later space exploration. The spending model was premised on the belief that pouring in investments in science at the front end of the development pipeline would produce technology out the other end. Public spending supported the enormous development costs of relevant new technologies. Initial applications were developed for and procured by the military, and later "spun off" into commercial use. For example, U.S. defense spending promoted the rapid development of jet aircraft and engines, silicon chips, computers and operating systems, complex machine tools, data networks, data compression, optoelectronics, and advanced ceramic and composite materials.
In these cases, government underwrote the relevant basic science research at universities and labs, direct R&D contracts accelerated the development of the technology, and defense procurement at premium prices constituted a highly effective initial launch market. Very often, the military funded varied technological approaches to the same goals, prudently spreading its bets under conditions of uncertainty. The successful approaches were judged according to strict cost-performance criteria, and then were launched through procurement and strongly supported. A variety of mechanisms, ranging from patent pooling and hardware leasing (e.g., machine tool pools) to loan guarantees for building production facilities, helped to lower entry costs, diffused technology widely among competitors, and set the stage for commercial market penetration. Aspects of this support model were adapted for government investment in other sectors, notably for public health, and produced similarly beneficial results in the form of new drugs, diagnostics, medical equipment and biotechnologies.
The overall key to the successful cases of government sponsorship was the successful launch and diffusion of a technology development path -- a trajectory -- whose characteristics corresponded with the requirements of the commercial marketplace. For example, when the military pushed silicon chips toward high reliability, miniaturization, high performance, and low costs, it was helping to create a trajectory that the commercial computer industry could ride. Similarly, when it turned to the national community of scientists and engineers in their roles as users to define the characteristics for the ARPANet, the Defense Department was launching a data networking trajectory that would also meet that community's commercial needs, as the ARPANet metamorphosed into today's Internet.
The historical experience strongly suggests that the U.S. government's direct R&D sponsorship has often been far less important for commercial success than its support to diffusion and use. Its procurement of new technologies and other indirect supports for application, launched the fledgling technologies and helped to diffuse them into widespread use. Although some of the winners generally credit their parallel civilian R&D efforts for the relevant technological advances, they all acknowledge the benefit of procurement, of know-how spilling over from defense R&D, of defense funding of graduate education and research in the relevant technical disciplines, of funding of prototype systems that demonstrated the efficacy of new technologies, and of the variety of other mechanisms that supported diffusion and use.
The strategy of public support was not a simple stepchild of the technological successes of World War II. Indeed, before the recent ideological purity set in, the U.S. government was an occasionally active public risk-taker. For example, government support to aeronautics began in earnest with the creation of NASA's precursor, the National Advisory Commission on Aeronautics (NACA). NACA was a vital source of the R&D and testing during the 1920s and 30s that led to the moder passenger airliner. Similarly, RCA grew out of Woodrow Wilson's concern that British dominance of radio technology would limit America's commercial rise, and was created to establish a commercial U.S. presence in radio. With the guarantee of Navy contracts providing R&D funding, the launch market, and lure, RCA was formed as a patent-pooling consortium among the Navy, GE, and eventually AT&T, Westinghouse, and United Fruit.
However, the most elaborate, and arguably most successful U.S. program of public support to commercial innovation is the Agriculture Research and Extension System, a network of interdependent institutions from the federal to the local level, including land-grant colleges, the state experiment stations, and research and extension services. Dating from the Morrill Act in 1862, the evolving system has provided focused education and training, long-term R&D, and widespread diffusion of new technology to America's farms. Although not without controversy -- e.g. its neglect of organic farming and pest-control methods -- it is still widely credited with a major role in making American agriculture the world's most productive.
While such successes are suggestive, there is as much to learn from the failures. These include outright flops like the supersonic transport, synfuel plants and the fast breeder reactor, as well as more ambiguous cases like the development of numerical control (NC) for machine tools or of photovoltaics. For example, the Air Force sponsored the development of NC technology for machine tools to build advanced aircraft. The programming language proved too complex for general commercial use; diffusion was slow and civilian application costly. The resulting technological development path produced only a commercially vulnerable U.S. industry that was squeezed by Japanese competitors from the low end and German firms from the high. Similarly, the supersonic transport (SST) failed because the commercial airliner market was aiming at short-haul and wide-bodies rather than supersonic speeds. The fast breeder reactor and synfuel programs were far more expensive than commercial alternatives, particularly after the oil shocks abated. In each case there were problems of both conception and execution: performance objectives were narrowly construed and alternative technological paths were not sufficiently explored. Demonstrations and pilots proceeded despite experimental evidence of failure. In some cases, like photovoltaics, political considerations killed development prematurely.
Public support to technology thus runs into trouble mostly when it pursues development paths that diverge from commercial market cost-performance requirements, particularly when it over-specifies an exotic technical solution in the form of a particular product. Thus, the first requirement of government technology development is that it must be sensitive to the needs of commercial market diffusion, like the need for manufacturability of designs or for customer-defined cost and performance. Second, it must support multiple experimental approaches to achieve cost-performance goals, letting the specific technical solution emerge from the resulting competition. Such requirements should not be difficult to build in to future programs, particularly if the public risk is shared jointly with private investment, as it has been with more recent forays into industrial policy like Sematech, the semiconductor industry's manufacturing technology program.
The Clinton Administration came into office committed to an aggressive technology policy in the service of long-run economic growth. The scope and ambition of the effort, if not the rhetoric defending it, were soon scaled back in deference to the growing popularity of efforts to eliminate the federal budget deficit.  Many of the Clinton Administration's higher-profile technology efforts had actually been launched by a reluctant Bush Administration under pressure from the Congress. Nevertheless, even though, as expected, the new Administration gave the initiatives more resources and more public exposure, the combined efforts continued to represent a minuscule claim on the federal R&D budget. In fiscal 1994, when Democrats still controlled both the Congress and the White House, the technology initiatives taken over or launched by the Clinton Administration accounted for less than $3 billion (roughly 4 percent) out of a total federal R&D budget of $71 billion. Defense ($38.3 billion), Health ($10.4 billion) and NASA ($9.4 billion) retained the largest share of federal R&D spending in their traditional programs. Despite the creation of two overarching policy coordination councils at the White House - the National Economic Council (NEC) and the National Science and Technology Council (NSTC) - it has been difficult for the Clinton administration to graft its expanded agenda for technology policy onto entrenched prerogatives of the traditional agency-based R&D programs. And as David Hart's chapter implies, this persistent difficulty has undermined the effectiveness of the Administration's overall efforts.
As the Clinton administration's technology investment strategy was put in place early in 1993, policy designers consciously attempted to prevent new instances of "government failure" -- the public sector analog of the market failures these projects were trying to correct - by building several programmatic features into the projects to ensure that they would not dampen market signals. First, to avoid backing the wrong technologies, the Clinton technology initiatives relied on industry to participate in the design of the research agenda for each project. This ensured that project awards would target technologies that were also thought likely to be commercially viable once technical bottlenecks were overcome. Second, investments were made in an array of technical fields to ensure that the champions of any particular technology or industry did not exercise undue influence. The programs also required grant applicants to compete in teams (e.g., defense and commercial firms with universities or government labs) for a finite flow of federal funds. The competitions were to be judged on technical and economic criteria by a panel of government technical experts or by some other independent peer-review process. Applicants were typically required to provide evidence that the technology at hand could be commercially sustained within five years without further federal funding.
Third, to prevent the subsidies from reducing the efficiency-inducing effects of competition, private sector participants were required to cover at least 50 percent of the project's costs. (Later, to make it easier for small companies to participate, companies were allowed to use federal funds from other government programs, such as Small Business Innovation Research grants, to pay for part of their "cost share"). To prevent the creation of technology pork barrels, government program managers were committed at the outset to rigorous program evaluations and could typically make only time-limited grants. Technical milestones and other performance metrics were established up front. It was also felt that the 50 percent matching requirement would make private sector partners eager to abandon technological approaches that were not working.
After four years, the various Clinton technology initiatives have demonstrated progress: as documented in the chapters by Hill, Shapira, and Guston, the Clinton efforts have fostered new industry-led R&D partnerships in a number of technical fields and have encouraged defense and commercial firms to work together on the commercial development of a number of military-relevant technologies. By playing midwife to consortia or teams of companies, universities, and national labs, these initiatives appear to be facilitating more rapid technology transfer and innovation, though at a small scale. In some cases, the government's involvement appears to have helped public and private research and development performers to overcome the "collective action" problems that otherwise prevent them from exploiting potentially significant economic and technological opportunities. Indeed, many recipients of the awards -- and even some teams that failed to win --report that the programs have facilitated beneficial organizational relationships that would not have existed had the programs not existed. In effect, as suggested in the chapter by Fountain, the Clinton efforts have helped to build "social capital."
Significantly, the features these programs incorporated to avoid government failure also appear to have worked. It is too early to judge the effect of time-limited grants, but other features of these programs -- government-industry cost-sharing, competitive selection, and the requirement that applicants be made up of industry-led teams -- have combined to render these efforts nearly free of political pork.
Ironically, the very success of those features has reduced opportunities for supporters to cultivate stable political constituencies for these programs, never mind expanding them. For example, three of the major technology initiatives begun under Bush and expanded under Clinton have been scaled back or essentially eliminated: the Commerce Department's Advanced Technology Program (geared toward promoting private-sector competitiveness and economic growth and discussed in the chapter by Hill); the Pentagon's Technology Reinvestment Project (geared toward promoting the commercial development of technologies with both civilian and military applications and discussed in the chapter by Cohen), and the Department of Energy's severely curtailed program of Cooperative Research and Development Agreements (CRADAs) between DOE's national laboratories and industry. In contrast, the Clinton administration's National Flat Panel Display Initiative, discussed in the chapter by Roos, and the "Clean Car" program (officially the Partnership for a New Generation of Vehicles) have survived Congressional scrutiny so far, down-sized but intact, due to the focused efforts of their specific and therefore readily-organized industrial constituencies. More tellingly, the Commerce Department's Manufacturing Extension Partnership (MEP) which, as Shapira describes in his chapter, fortuitously placed more than 70 manufacturing extension centers in all 50 states (and thus in at least that many congressional districts), is actually being expanded by the Republican-led Congress.
Equally significant, the Clinton projects suffer from two important political deficiencies. One is the lack of programmatic objectives tied to clearly defined government missions, as opposed to more amorphous goals like "competitiveness" and "growth," which most Americans assume to be primarily the responsibility of the private sector. Success may be possible where competitiveness is clearly defined in terms of a specific, tangible mission such as the broader diffusion of technological advances and best practice to small manufacturers, a main goal of the MEP. But historically, competitiveness -- comparatively long-run productivity growth and a rising standard of living -- is best pursued indirectly: It results when federal spending develops new technology industries through sponsored research and through procurement in mission areas where government itself can serve as the initial launch market.
The second flaw of the Clinton initiatives is their lack of sufficient scale and scope: it is fine to fund small-scale projects, but only so long as they are clearly connected, as a group, to a larger set of public purposes. And the public purposes to which they are connected must be understood and approved broadly enough to inspire strong support from a wide range of political constituencies. Instead, the Clinton initiatives spread limited resources across too many projects that, taken in the aggregate, are still too small in overall scale to have a major impact on most large industry sectors or any pressing national problems. In addition, the projects have typically failed to exploit the potential for more stable budgetary support that could have been provided had they been designed in to the regular R&D programs of the various mission-oriented agencies of the Federal government. The lack of scale is largely a consequence of President Clinton's political inability to sell his "investment" program in the political debates over deficit reduction early in his first term. The lack of scope reflects the Clinton administration's tactical preference for funding stand-alone initiatives that are easier to "market" and for which it is easier to claim political credit when they go well, but that are also more vulnerable to budget cutters both in and outside the administration. This tactic has diverted the administration's attention from the more arduous and important work of integrating the expanded reliance on industrial partnerships for technology development into the standard operating procedures of the traditional mission agencies.
Difficult as it will continue to be to create political and budgetary support to enlarge technology policy efforts in scale and scope to the point where they would have significant impacts in launching new technology industries, this would be a most inopportune moment for the United States to confine its role only to R&D instead of comprehensive technology sponsorship, or to rely blindly on the invisible hand. Until the 1980s, when the absolute lead U.S. industry enjoyed in most high-technology sectors began to evaporate, the federal government could be certain that the domestic economy would enjoy the lion's share of the broad social gains generated by its vast R&D budget. As the strongest and most advanced economy, the U.S. was always the launch market for the new technologies fostered by public spending. U.S. industry typically commercialized and produced the innovations at home, and then exported abroad. Initial and leading customers -- those who shaped the new technology's initial development and its path of diffusion -- were also typically domestically based. Local R&D, production, and advanced use meant that most of the spillovers that generated the broad social benefits would occur within U.S. borders.
During the past decade, however, several trends have converged to challenge the easy identity between federal R&D and domestic spillovers. Foreign competitors have caught up with and in some cases surpassed U.S. producers. Foreign governments followed the U.S. lead to sponsor high and rising levels of R&D spending. Foreign markets became effective launch markets for their own new technologies, such as the Europe's Airbus's pioneering of fly-by-wire and other aeronautical innovations. Available time for spillover from U.S. defense spending shrank as foreign producers caught up with U.S. innovation. And as international competition intensified, so did the costs and risks of private R&D investment, so that even U.S. firms chose to spread them across global markets by producing abroad and finding foreign partners.
As a result, technologies pioneered in the U.S. now flow rapidly across the national borders, sometimes to be commercialized, produced, and exploited more effectively there than in the U.S. Conversely, more and more innovations now originate abroad, but because foreign economies are rarely as open as the U.S., a reverse flow of innovation into the U.S. has not fully materialized. Indeed, international technological specialization is increasing -- new technical skills that are essential to the commercialization of innovation are arising in new places around the globe, especially in Asia, and are not readily duplicated back in the U.S.
Thus the relationship between federal R&D spending and the capture of its domestic spillovers and social benefits has been significantly. This makes it all the more important for government to focus its own sponsorship in areas where spillovers are most likely to be generated and captured locally. Because know-how developed through production and use, and embedded in local assets like labor pools, supplier networks, and infrastructure is less likely to diffuse readily across borders, the government should focus on helping to bring such assets into being. Such assets will be created whenever comprehensive government sponsorship of new technologies helps to develop the domestic market as the principal launch market for those technologies.
In the current era of tight budget constraints, this means that government must focus its scarce resources, not squander them in piecemeal sponsorship of small projects with at best modest impacts. It also means that the government cannot focus on the amorphous goal of directly supporting commercial competitiveness -- for in most cases, the market dynamics of commercial industries are already developed, and policy intervention is unlikely to lead to alter them significantly in ways that create new domestic capabilities. Rather, federal sponsorship can most effectively launch local capabilities only by focusing on wholly new technological possibilities linked tightly to a government mission (so that the government itself becomes the initial launch market for new technologies for which commercial markets have yet to develop).
Two prime possibilities are environmental stewardship and infrastructure. The environmental opportunity, explored in greater detail in the chapter by Banks and Heaton, is to move beyond existing efforts aimed at regulating waste reduction and mandating clean-up. Sponsorship should instead be directed to replacing existing industrial production with technologies that generate no waste or pollution in the first place. Government procurement, from automobile fleets to office supplies, should favor industrial processes that boost pollution prevention, resource sustainability, and efficient resource usage. Policy should set performance standards only, inviting different competitive approaches to determine the most effective means of meeting those standards.
Similarly, there is an acknowledged need to rebuild much of the nation's eroding networks for transportation, power, sewage, and water, and to upgrade the infrastructure for communications. Sponsorship of innovation to meet modern infrastructure needs would spur a host of new technologies from low-maintenance concretes to optical control systems. Emphasis would be on seeding and then procuring new technological approaches that, while more costly up-front, held the promise of reducing total life-cycle costs.
U.S. public support for government investments such as these remains weak, even though they are critical to raising long-run living standards. To many Americans, the Clinton administration's investments in technological innovation have appeared to benefit only the multibillion-dollar corporations and high-tech professionals who are already doing well in the global, information age economy. Most middle-class voters, concerned about the impact of new computer-based technologies on their own jobs, uncertain and impatient about the economic future, see no evidence that these policies are actually working, let alone that larger scale projects might work.
Nevertheless, a broad consensus both about national needs and the legitimacy of government's role persists in two areas -- national security and public health. A set of small technology development and demonstration projects connected to these and other larger purposes, aggregated, that is, into a set of large, high-profile efforts to address easily demonstrated national needs - protecting the environment, improving mass transportation - might finally attract a broader, more stable constituency for government investments in technological progress.
Where possible, government procurement should help to provide launch markets for new technologies that provide promising solutions to national needs, as long as deadlines and benchmarks for commercialization and the eventual withdrawal of government subsidy are established up front. But government technology development programs, to which such procurement should be tied, must incorporate the features that have proved critical to ensuring that recent initiatives such as ATP and TRP remained responsive to commercial market demand - industry must be involved in setting research agendas, applicants must compete for individual grants, and there must be strict time-limits on government support. The necessity of involving profit-seeking firms in the development of commercially sustainable technologies means that the public is always likely to remain ambivalent about government's proper role. Nevertheless, a set of programs modelled on ATP but established within the various mission agencies (e.g., the Department of Transportation, the Environmental Protection Agency, the Department of Energy) might help establish a standard mechanism (and a set of political constituencies) to ensure that the U.S. can routinely explore technological innovation to address national needs, while at the same time enlisting the private sector in partnerships to accelerate the invention and eventual commercialization of any novel technologies that result.
 From the poem, The Hollow Men, by T.S. Eliot (1925).
 If productivity growth is accomplished simply by corporate "downsizing," then at least in the short run it may come at the expense of progress on income inequality and unemployment.
 Income inequality and slow productivity growth are equally important factors in determining the standard of living, but because the way that new technology enables workers to earn higher real wages is by first making them more productive, we focus here on the direct link between technological progress and productivity growth.
4 Prior to his recent vituperative attacks on popular concepts of relative economic performance, Paul Krugman made this point most succinctly for non-economist, policy audiences in his The Age Of Diminished Expectations (Cambridge, MA. and London: The MIT Press, 1990).
 The quote is from Linda R. Cohen and Roger G. Noll, The Technology Pork Barrel (Washington, D.C.: Brookings, 1991), p.11.
 See, e.g., Edward F. Denison, Accounting for United States Economic Growth, 1929-1969 (Washington D.C.: Brookings, 1974). The role of technology and R&D has been studied by economists for more than forty years. The earliest studies almost stumbled upon the importance of technology-spawned productivity improvements as an explanation for economic growth. One, which examined the U.S. economy over the period 1909-1949 when gross output per household doubled, estimated that only 12.5 percent of this increased output was due to increased use of capital (i.e. more machines). More importantly, the residual growth of 87.5 percent could only be explained by technical change, i.e. new machines and better ways of organizing industrial activities. (Robert M. Solow, "Technical Change and the Aggregate Production Function," Review of Economics and Statistics, No. 39 (1957), pp. 312-320. Attempting to overcome many restrictive assumptions of these initial studies, a recent study by Boskin and Lau examined economic growth in the five largest industrial economies and found that, consistent with the earlier work, technological progress is by far the most important source of economic growth (Michael J. Boskin and Lawrence J. Lau, "The Contribution of R&D to Economic Growth: Some Issues and Observations," American Enterprise Institute, conference paper, October 3, 1994. For a nice summary of the growth accounting and return on investment literature, see Gregory Tassey, Technology and Economic Growth: Implications for Federal Policy (Washington DC: U.S. Department of Commerce, Technology Administration, October, 1995).
 Paul M. Romer, "Increasing Returns and Long-Run Growth," Journal of Political Economy, Vol. 94, No. 4 (October 1986), pp. 1002-37, and "Endogenous Technological Change," Journal of Political Economy, Vol. 98, No. 5 (1990), pp. S71-S102.
 W. Brian Arthur, Increasing Returns and Path Dependence in the Economy (Ann Arbor, MI: University of Michigan Press, 1994). See, in particular, Chapter 7, pp. 111-132.
 M. Ishaq Nadiri, "Innovations and Technological Spillovers," National Bureau of Economic Research, Working Paper No. 4423 (August 1993).
10 Richard M. Cyert and David C. Mowery, editors, Technology and Employment: Innovation and Growth in the U.S. Economy (Washington, D.C.: National Academy Press, 1987); Paul Krugman, "Technology's Revenge," Wilson Quarterly (Autumn 1994), pp. 57-64; David R. Howell, "The Skills Myth," The American Prospect, Number 18, (Summer 1994), pp. 81-89; "Technology and Unemployment," The Economist (February 11, 1995), pp. 21-23.
 U.S. Department of Commerce, Building the American Dream (August, 1996).
 In addition to Nadiri, "Innovations and Technological Spillovers," see Martin Neil Baily and Alok K. Chakrabarti, Innovation and the Productivity Crisis (Washington, D.C.: Brookings Institution, 1988), Edwin Mansfield "Social Returns from R&D: Findings, Methods, and Limitations," Research and Technology Management (November-December, 1991); and Zvi Griliches, "The Search for R&D Spillovers," Scandinavian Journal of Economics, vol. 94, supplement (1992), pp. 29-47.
 For a concise summary, see Linda R. Cohen and Roger G. Noll, The Technology Pork Barrel at pp.18-22; and Richard R. Nelson, editor, Government and Technical Progress (New York: Pergamon Press, 1982), pp. 2-5 and pp. 480-481.
 Gene M. Grossman, "Promoting New Industrial Activities: A Survey of Recent Arguments and Evidence," OECD Economic Studies, No. 14 (Spring 1990), pp.87-125.
 Cohen and Noll, The Technology Pork Barrel, are illustrative.
 The latter two ideas originated in the so-called new trade theory literature, see, e.g., Paul R. Krugman, editor, Strategic Trade and the New International Economics (Boston, MA.: MIT Press, 1986).
 This and the following data are drawn from Laura D'Andrea Tyson, Who's Bashing Whom: Trade Conflicts in High Technology Industries (Washington D.C.: Institute for International Economics, 1992).
 See, e.g., William T. Dickens and Kevin Lang, "Why It Matters What We Trade: A Case for Active Trade Policy," in William T. Dickens, Laura D'Andrea Tyson, and John Zysman, editors., The Dynamics of Trade and Employment, (Cambridge MA.: Ballinger, 1989).
 For a fuller elaboration of the successes and failures of this technology development model, see the Chapter by Jay Stowsky in Wayne Sandholz, Michael Borrus, et al., The Highest Stakes: The Economic Foundations of the Next Security System, (New York: Oxford University Press, 1992), pp. 114-140.
 For details see Eric Barnouw's classic history of broadcasting, A history of broadcasting in the United States (New York: Oxford University Press, 1966).
For an evaluation, see the chapter by R.E. Evenson in Nelson, editor, Government and Technical Progress.
 See Stowsky, The Highest Stakes, pp. 122-126.
 For details, see the relevant chapters in Cohen and Noll, The Technology Pork Barrel.
Even in those cases there are likely to be important technical spillovers, especially when generic research is funded as part of the program.
 The decision to scale back the technology initiatives was made even before the new President officially assumed his office, in a series of internal debates detailed by journalist Bob Woodward, in The Agenda (New York: Simon and Schuster, 1994). The initiatives were scaled back further in bruising, early-term battles with Congress over spending priorities and deficit reduction.
 Figures from American Association for the Advancement of Science, Congressional Action on Research and Development in the FY 1997 Budget, special table, "Trends in R&D, FY's 1994-97," and National Science Foundation, Science and Engineering Indicators, 1996. Text table 4-5.
 For additional discussion of the political dynamics underlying TRP and the Pentagon's other dual use technology initiatives, see Jay Stowsky, "The Dual-Use Dilemma," Issues in Science and Technology Vol. XIII, No. 2 (Winter 1996-97), pp. 56-64.
 For a discussion of the history and fate of CRADA's at the DOE labs, see "DOE to Industry: So Long, Partner," Science (Vol. 274), October 4, 1996. For more on the National Flat Panel Display Initiative, see Kenneth Flamm, "Flat-Panel Displays: Catalyzing a U.S. Industry," Issues in Science and Technology Vol. XI, No. 1 (Fall 1994). For more on the Clean Car Initiative (officially, the Partnership for a New Generation of Vehicles), see Daniel Sperling, "Rethinking the Car of the Future," Issues in Science and Technology Vol. XIII, No. 2 (Winter 1996-97), pp. 29-34.
 For evidence, see e.g. Daniele Archibugi and Jonathan Michie, "The Globalization of Technology: Myths and Realities," Cambridge Journal of Economics No. 19 (1995), and Keith Pavitt and Parimal Patel, "The International Distribution and Determinants of Technological Activities," Oxford Review of Economic Policy No. 4 (1988), p.35-55.