America's Industrial Resurgence: How Strong, How Durable?
The U.S. economy responded successfully to the challenges of the 1980s, but this is no time for complacency.
Reports in the late 1980s painted a gloomy picture of U.S. industrial competitiveness. The report of the MIT Commission on Industrial Productivity, perhaps the best known one, opined that "American industry is not producing as well as it ought to produce, or as well as it used to produce, or as well as the industries of some other nations have learned to produce...if the trend cannot be reversed, then sooner or later the American standard of living must pay the penalty." The commission criticized U.S. industry for failing to translate its research prowess into commercial advantage.
Since that report's publication, overall U.S. economic performance has improved markedly. What is the true nature of that improvement? Is it a result of better performance in the industries analyzed by the MIT Commission? Is it a development that holds lessons for public policy?
Economy-wide measurements actually paint a mixed picture of industry performance and structural change since the early 1980s. The trade deficit has grown, hitting a record high of $166 billion in 1997. Although nonfarm business labor productivity growth rates have improved since 1990, they remain below the growth rates achieved between 1945 and 1980. Unemployment and inflation are significantly lower than in the 1970s and 1980s, but not all segments of the population have benefited equally: Households in the lowest quintile of national income have fared poorly during the past two decades, whereas the top quintile has done well.
Other indicators suggest that the structure of the U.S. R&D system changed significantly beginning in the early 1980s and that this structural change has yet to run its course. Industrially financed R&D has grown (in 1992 dollars) by more than 10 percent annually since 1993, but real industrial spending on basic research declined between 1991 and 1995. Recent growth in industrially financed R&D is dominated by spending on development.
Aggregate performance indicators thus are mixed, although broadly good. Moreover, much of the improvement is the result of developments in the economies of other nations. For example, severe problems hobbled the Japanese economy for much of the 1990s, weakening many of the Japanese companies that were among the strongest competitors of U.S. companies during the 1980s. Thus, the relationship between this improved aggregate performance and trends in individual industries, especially those singled out for criticism by the MIT Commission and other studies, remains unclear.
A new study by the National Research Council's Board on Science, Technology and Economic Policy, U.S. Industry in 2000: Studies in Competitive Performance, assesses recent performance in 11 U.S. manufacturing and nonmanufacturing industries: chemicals, pharmaceuticals, semiconductors, computers, computer disk drives, steel, powdered metallurgy, trucking, financial services, food retailing, and apparel. Its first and most striking conclusion is how extraordinarily diverse their performance has been since 1980.
Some, such as the U.S. semiconductor and steel industries, have staged dramatic comebacks from the brink of competitive collapse. Others, including the U.S. computer disk drive and pharmaceutical industries, have successfully weathered ever-stronger foreign competition. For the nonmanufacturing industries included in the study, foreign competition has been less important, but deregulation and changing consumer preferences have increased domestic competition.
This diversity partly reflects the industries' contrasting structures. Some, such as powdered metallurgy and apparel, comprise relatively small companies with modest in-house capabilities in conventionally defined R&D. Others, such as pharmaceuticals and chemicals, are highly concentrated, with a small number of global companies dominating capital investment and R&D spending. In semiconductors, computer software, and segments of computer hardware, by contrast, small and large companies complement one another and are often linked through collaborative R&D. Similar diversity is apparent within the three nonmanufacturing industries. Although entry barriers appear to be high and growing higher in some industries, such as chemicals and computer disk drives, in others a combination of technological developments and regulatory change is generating new competitors.
Despite this diversity, which is compounded by differences among industries in the indicators used to measure their performance, all of these industries have improved their competitive strength and innovative performance during the past two decades. Improvements in innovative performance have not rested solely on the development of new technologies but also on the more effective adoption and deployment of innovations.
The definition of innovation most relevant to understanding the improved performance of U.S. companies in these industries thus must be broad, including not just the creation of new technology but also its adoption and effective deployment. Yet the essential investments and activities associated with this definition of innovation are captured poorly, if at all, in public R&D statistics. Even the broader innovation surveys undertaken by the National Science Foundation (NSF) and other public statistical agencies omit many of these activities.
In the computer industry, for example, innovation relies in part on "co-invention," a process in which the users of hardware and software contribute to its development. Similar examples can be drawn from other industries. In still others, specialized suppliers of logistics services, systems integration, and consulting services have been essential.
Another factor in improved performance is the efficient adoption of technologies from elsewhere. In many cases (for example, finance, apparel, pharmaceuticals, and computers), the adoption of new technologies (including new approaches to managing innovation) has required significant changes in organizational structure, business processes, or workforce organization.
The intersectoral flow of technology, especially information technology, also has contributed to stronger performance in many of these industries. The importance of this flow underscores the fallacy of separating "high" technology from other industries or sectors in this economy. Mature industries in manufacturing (such as apparel) and nonmanufacturing (such as trucking) have rejuvenated performance by adopting technologies developed in other industries. The effects are most apparent in the nonmanufacturing industries of trucking, food retailing, and financial services, all of which have undergone fundamental change as a result of adopting advanced information technologies. Moreover, management of the adoption process and effective absorption of technology from other sectors are themselves knowledge-intensive activities that often require considerable investment in experimentation, information collection, and analysis.
An excellent illustration of the importance of these relationships among sectors is the importance to U.S. firms in the computer and semiconductor industries of their proximity to demanding, innovative users in a large domestic market. In addition, the rapid growth of desktop computing in the United States was aided by imported desktop systems and components, which kept prices low. It also propelled adoption of this technology at a faster pace than in most Western European economies or in Japan, where trade restrictions and other policies kept prices higher. The rapid adoption of desktop computing contributed to the growth of a large packaged software industry, which U.S. companies continue to dominate.
This virtuous circle was aided further by the restructuring and gradual deregulation of the U.S. telecommunications industry that began in the 1980s. The result was the entry of numerous providers of specialized and value-added services, which created fertile terrain for the rapid growth of companies supplying hardware, software, and services in computer networking. This trend benefited the U.S. computer industry, the U.S. semiconductor industry, and the domestic users (both manufacturing and nonmanufacturing companies) of products and services produced by both. These and other intersectoral relationships are of critical importance to understanding U.S. economic and innovative performance at the aggregate and industry-specific levels.
Diffusion of information technology, which has made possible the development and delivery of new or improved products and services in many of these industries, appears to be increasing the basic requirements of many jobs that formerly required minimal skills. These technologies place much greater demands on the problem-solving, numeracy, and literacy skills of employees in trucking, steel fabrication, banking, and food retailing, to name only a few. Trucking, for example, now relies heavily on portable computers operated by truck drivers and delivery personnel for monitoring the flow and content of shipments. Workers in these industries may have adequate job-specific training, but they face serious challenges in adapting to these new requirements because of weaknesses in the basic skills now required.
But the adoption and effective implementation of new technologies also place severe demands on the skills of managers and white-collar workers. Not only do managers need new skills, including the ability to implement far-reaching organizational change, but in industries as diverse as computing or banking, they face uncertainty about the future course of technologies and their applications.
Nontechnological factors such as trade and regulatory policy, the environment for capital formation and corporate governance, and macroeconomic policy all play important roles in industrial performance too, especially over the long run. One of the most important is macroeconomic policy, which affects the entire U.S. economy yet rarely figures prominently in sectoral analyses. Both monetary and fiscal policy have been less inflationary and less destabilizing during the 1990s than during the 1980s. Although the precise channels through which the macroeconomic environment influences the investment and strategic decisions of managers are poorly understood, these "micro-macro" links appear to be strong. They suggest that a stable noninflationary macroeconomic policy is indispensable for improved competitive performance.
Another common element that has strengthened competitive performance, especially in the face of strong foreign competition, is rapid adaptation to change. U.S. companies in several of these industries have restructured their internal operations, revamped existing product lines, and developed entirely new product lines, rather than continuing to compete head-to-head with established product lines. Many of the factors cited by the MIT Commission and other studies as detrimental to U.S. competitiveness, such as the entry of new companies into the semiconductor industry or pressure from capital markets to meet demanding financial performance targets, actually contributed to this ability to change. In some cases, efforts by U.S. companies to reposition their products and strategies were criticized for hollowing out these enterprises, transferring capabilities to foreign competitors and/or abandoning activities that were essential to the maintenance of these capabilities. To a surprising degree, these prophecies of decline have not been borne out.
U.S. disk drive manufacturers, for example, shifted much of their production off shore, but the shift has not damaged their ability to compete. Nor has the withdrawal of most U.S. semiconductor manufacturers from domestic production of DRAM (dynamic random access memory) components severely weakened their manufacturing capabilities in other product lines. In many U.S. industries, the post-1980 restructuring has been associated with the entry of new companies (such as specialty chemical companies, fabless semiconductor design companies, package express companies, or steel minimills). In other cases, restructuring has been aided by the entry of specialized intermediaries (systems integration companies, consultants, logistics companies, or specialized software producers).
Restructuring is not always successful. In financial services, for example, many mergers and acquisitions ended by diminishing shareholder value. But in some industries (notably steel, disk drives, and semiconductors) European and Japanese companies were slow to respond to the new competition, often because their domestic financial markets were less demanding than those in the United States. This financial environment also has facilitated the formation of new companies in such U.S. industries as semiconductors and biotechnology.
At least two issues remain unresolved. First, if U.S. companies' restructuring in the 1990s was an important factor in their improved performance, why did it take so long to begin? Second, will restructuring be only occasional in the future, or will it be a continuing process? Moreover, rapid structural change has significant implications for worker skills and employment, an important policy issue that has received little attention in most discussions of industrial resurgence.
Change in the structure of innovation
Since 1980, innovation by companies in all 11 of the industries examined in U.S. Industry in 2000: Studies in Competitive Performance has changed considerably. The most common changes include 1) increased reliance on external R&D, such as that performed by universities, consortia, and government laboratories; 2) greater collaboration in developing new products and processes with domestic and foreign competitors and with customers; and 3) slower growth or outright cuts in spending on research, as opposed to development.
Beginning in the 1980s, a combination of severe competitive pressure, disappointment with perceived returns on their rapidly expanding investments in internal R&D, and a change in federal antitrust policy led many U.S. companies to externalize a portion of their R&D. Large corporate facilities of pioneers of industrial R&D such as General Electric, AT&T, and DuPont were sharply reduced, and a number of alternative arrangements appeared. U.S. companies forged more than 450 collaborations in R&D and product development, according to reports they filed with the Department of Justice between 1985 and 1994 under the terms of the National Cooperative Research Act. Collaboration has become much more important for innovation in industries as diverse as semiconductors and food retailing.
U.S. companies also entered into numerous collaborations with foreign companies between 1980 and 1994. Most of these international alliances for which NSF has data link U.S. and Western European companies. Alliances between U.S. and Japanese companies also were widespread. But these were outstripped by "intranational" alliances linking U.S. companies with domestic competitors. Both kinds of alliances are most numerous in biotechnology and information technology. In contrast to most domestic consortia, which focused on research, a large proportion of U.S.-foreign alliances focused on joint development, manufacture, or marketing of products. In addition to seeking cost sharing and technology access, U.S. companies sought international alliances in order to gain access to foreign markets.
U.S. companies in many of these industries reacted to intensified competitive pressure and/or declining competitive performance by reducing their investments in research. These reductions appear to have accelerated during the period of recovery despite significant growth in overall R&D spending. During 1991-95, total spending on basic research declined, on average, almost 1 percent per year in constant dollars. This decline reflected reductions in industry-funded basic research from almost $7.4 billion in 1991 to $6.2 billion in 1995 (in 1992 dollars). Real federal spending on basic research increased slightly during this period, from $15.5 billion to almost $15.7 billion. Industry-funded investments in applied research grew by 4.9 percent during this period, and federal spending on applied research declined at an annual rate of nearly 4 percent. In other words, the upturn in real R&D spending that has resulted from more rapid growth in industry-funded R&D investment is almost entirely attributable to increased spending by U.S. industry on development, rather than research.
Universities' share of total U.S. R&D performance grew from 7.4 percent in 1960 to nearly 16 percent in 1995, and universities accounted for more than 61 percent of the basic research performed within the United States in 1995. By that year too, federal funds accounted for 60 percent of total university research, and industry's contribution had tripled to 7 percent of university research. The increased importance of industry in funding university research is reflected in the formation during the 1980s of more than 500 research institutes at U.S. universities seeking support for rserach on issues of direct interest to industry. Nearly 45 percent of these institutes involve up to five companies as members, and more than 46 percent of them receive government support.
The Bayh-Dole Act of 1980 permitted federally funded researchers to file for patents on their results and license those patents to other parties. The act triggered considerable growth in university patent licensing and technology transfer offices. The number of universities with such offices reportedly increased from 25 in 1980 to 200 in 1990, and licensing revenues increased from $183 million to $318 million between 1991 and 1994 alone. During the 1980s, U.S. universities nearly doubled their ratio of patents to R&D spending, from 57 patents per billion in constant dollars spent on R&D in 1975 to 96 per billion in 1990, even though U.S. patenting by universities declined steeply overall, from 780 per billion dollars of R&D in 1975 to 429 in 1990.
Another shift in the structure of innovation was the increased presence of non-U.S. companies in the domestic R&D system. Investment by U.S. companies in offshore R&D (measured as a share of total industry-financed R&D spending) grew modestly during 1980-95, from 10.4 percent in 1980 to 12 percent in 1995. But the share of industrial R&D performed within the United States and financed from foreign sources grew substantially, from 3.4 percent in 1980 to more than 11 percent in 1995.
Despite this growth, as of 1994 foreign sources financed a smaller share of U.S. industrial R&D than they did in Canada, the United Kingdom, or France. Increased foreign financing of U.S. R&D is reflected in a modest increase in the share of U.S. patents granted to foreign inventors, from 41.4 percent in 1982 to 44.9 percent in 1995. Foreign companies also formed joint research ventures with U.S. companies; this international cooperation accounted for nearly a third of research joint ventures between 1985 and 1994.
Finally, foreign companies doing R&D in the United States collaborated with U.S. universities. More than 50 percent of the Japanese R&D laboratories in the United States, more than 80 percent of the U.S.-sited French R&D laboratories, and almost 75 percent of German corporate R&D laboratories in the United States had collaborative agreements with universities.
Policy issues and implications
The restructured innovation process that has contributed to the resurgence of many U.S. industries emphasizes rapid development and deployment of technologies but places decreasing weight on the long-term scientific understanding that underpins future technologies. This shift has produced high private returns, but its long-term consequences are uncertain.
The changing structure of innovation also highlights the difficulty of collecting and analyzing data that enable managers and policymakers to assess innovative performance or structural change. As I noted earlier, many of the activities contributing to innovation are not captured by conventional definitions of R&D. They include investments in human resources and training, the hiring of consultants or specialized providers of technology-intensive services, and the reorganization of business processes. All of these activities have contributed to the innovative performance of the industries examined in the STEP study.
The STEP study focused primarily on industry-level changes in competitive performance, rather than public policy issues. But the study raises a number of issues for public policy. They include 1) the ability of public statistical data to accurately measure the structure and performance of the innovation process; 2) the level and sources of investment in long-term R&D; 3) the role of federal regulatory, technology, trade, and broader economic policies in these industries' changing performance; 4) the importance and contributions of sector-specific technology policies to industry performance; and 5) worker adjustment issues posed by structural and technological change.
Data currently published by NSF provide little information on changes in industrial innovation. R&D investment data, for example, do not shed much light on the importance or content of the activities and investments essential to intersectoral flow and adoption of information technology-based innovations. Indeed, all public economic data do a poor job of tracking technology adoption throughout the U.S. economy. Moreover, in many nonmanufacturing industries that are essential to the development and diffusion of information technology, R&D investment is difficult to distinguish from operating, marketing, or materials expenses. For example, these data do not consistently capture the R&D inputs provided by specialized companies to supposedly low-technology industries such as trucking and food retailing. Without substantial change in the content and coverage of data collection, our portrait of innovative activity in the U.S. economy is likely to become less and less accurate.
The improved performance of many of the industries examined in the STEP study has occurred despite reductions in industry-funded investments in long-term R&D. This raises complex issues for policy. Specifically, should public R&D investments seek to maintain a balance within the U.S. economy between long- and short-term R&D? If so, how? Some argue for closer public-private R&D partnerships, involving companies, universities, and public laboratories. Yet most recent partnerships of this sort have tended to favor near-term R&D investment. There are few models of successful partnership in long-term R&D that apply across all industries.
A second issue concerns the treatment of the results of publicly funded R&D in the context of such partnerships. A series of federal statutes, including Bayh-Dole, the Stevenson-Wydler Act of 1980, the Technology Transfer Act of 1986, and others, have made it much easier for federal laboratories and universities to patent the results of federally funded research and license these patents to industrial partners. Proponents of licensing argue that clearer ownership of intellectual property resulting from federal R&D will facilitate its commercial application. Patenting need not restrict dissemination of research results, but restrictive licensing agreements may do so. For example, the science performed in U.S. universities, much of which was funded by the National Institutes of Health (NIH) during the postwar period, has aided the U.S. pharmaceuticals industry's innovative performance. If new federal policies limit the dissemination of research results, however, the industry's long-term performance could be impaired.
Industry's growing reliance on publicly funded R&D for long-term research and the increase in patenting and licensing by universities and federal laboratories create challenges that have received too little attention from industry and government officials. There is little evidence that these new arrangements are impeding innovation or limiting public returns on the large federal investments in R&D. But careful monitoring is required, because warning signals are likely to lag significantly behind the actual appearance of such problems.
Their impact varies, but federal intellectual property, antitrust, trade, and regulatory policies have affected the resurgence of many industries. These federal policies have been most effective where their combined impacts have supported high levels of domestic competition and opened U.S. markets to imports and foreign investment. Modifications of intellectual property, trade, and antitrust policy must not inadvertently protect companies from competitive pressure. For example, liberal policies toward foreign investment allowed U.S. companies to benefit from the management practices of foreign-owned producers of semiconductors, steel, and automobiles. The restructuring and deregulation of telecommunications, trucking, and financial services also have intensified pressure on U.S. companies to improve their performance.
The record of technology policy in the STEP industry studies is less clear. The studies suggest that the most effective technology policies involve stable public investment over long periods of time in "extramural" (that is, nongovernmental) R&D infrastructure that relies on competition among research performers. U.S. research universities are especially important components of this domestic R&D infrastructure. In some cases, as in federal support for biomedical research through NIH or the Advanced Research Projects Agency's support for computer science since the 1950s, these investments in long-term research have had major effects. U.S. competitive strength in pharmaceuticals, biotechnology, computers, and semiconductors has benefited substantially from federal investments in a robust national research infrastructure.
Sector-specific technology support policies, such as defense-related support for disk drive technologies or even SEMATECH, appear to have had limited but positive effects. This more modest impact reflects the tendency of such policies to be episodic or unstable, the relatively small sums invested, and the extremely complex channels through which any effects are realized.
Finally, attention must be paid to the effects of industrial restructuring, technology development and adoption, and competitive resurgence on U.S. workers, especially low-skill workers. Technology continues to raise the requirements for entry-level and shop-floor employment even in the nonmanufacturing sector. In addition, the very agility of U.S. enterprises that contributed to recent improvements in performance imposes a heavy burden on workers. Moreover, the perception that such adjustment burdens are unequally distributed can have significant political effects, revealed most recently in the 1997 congressional defeat of "fast-track" legislation to support continued trade liberalization. The United States and most other industrial economies lack policies that can help workers adjust to economic dislocation and compete effectively for better-paying jobs without increasing labor market rigidity. The political and social consequences of continuing failure to attend to these adjustment issues could be serious.
The resurgence of U.S. industry during the 1990s was as welcome as it was unexpected, given the diagnoses and prescriptions of the 1980s. Indeed, this recovery was well under way in some industries at the very time when MIT Commission presented its critique. Moreover, in at least some of the key industries identified as threatened by the MIT study and others, factors singled out in the 1980s as sources of weakness became sources of competitive strength in the 1990s. After all, the competitive resurgence of many if not most of the industries discussed in the STEP study reflects their superiority in product innovation, market repositioning, and responsiveness to changing markets rather than dramatic improvements in manufacturing. Manufacturing improvements in industries such as steel or semiconductors were necessary conditions for their competitive resurgence, but they were not sufficient.
This argument raises a broader issue that is of particular importance for policymakers. Observers of industrial competitiveness must accept the reality that performance indicators have a very low signal-to-noise ratio: data are unavailable, unreliable, and often do not highlight the most important trends. Uncertainty is pervasive for managers in industry and for policymakers in the public sector. Government policies designed to address factors identified as crucial to a particular performance problem may prove to be ineffective or even counterproductive when the data turn out to be inaccurate. Improvements in the collection and analysis of these data are essential. But in a dynamic, enormous economy such as that of the United States, these data inevitably will provide an imperfect portrait of trends, causes, and effects. In other words, policy must take perpetual uncertainty into account. Ideally, policies should be addressed to long-term trends rather than designed for short-run problems that may or may not be correctly identified.
Is our present state of economic grace sustainable? A portion of the improved performance of many of these U.S. industries reflects significant deterioration in Japan's domestic economy. Japan's recovery may take time, but eventually the outlook will improve for many of the companies that competed effectively with U.S. companies during the 1980s.
Prediction is an uncertain art, but it seems unlikely that U.S. companies have achieved permanent competitive advantage over those in other industrial and industrializing economies. The sources of U.S. resurgence are located in ideas, innovations, and practices that can be imitated and even improved on by others. Global competition will depend more and more on intellectual and human assets that can move easily across national boundaries. The competitive advantages flowing from any single innovation or technological advance are likely to be more fleeting than in the past. Economic change and restructuring are essential complements of a competitive industrial structure.
Some relatively immobile assets within the U.S. economy will continue to aid competition and innovation. The first is the sheer scale of the U.S. domestic market, which (even in the face of impending monetary unification in the European Union) remains the largest high-income region that possesses unified markets for goods, capital, technology, and labor. Combined with other factors, such as high levels of company formation, this large market provides a test bed for the many economic experiments that are necessary to develop and commercialize complex new technologies.
Neither managers nor government personnel are able to forecast applications, markets, or economic returns from such technologies. An effective method to reduce uncertainty through learning is to run economic experiments, exploring many different approaches to innovation in uncertain markets and technologies. The U.S. economy has provided a very effective venue for these experiments, and the growth of new, high-technology industries has benefited from the tolerance for experimentation (and failure) that this large market provides.
A second important factor is a domestic mechanism for generating these experiments. Here, the postwar U.S. economy also has proven to be remarkably effective. Success has been influenced by large-scale federal funding of R&D in universities and industry, as well as a policy structure (including the financial and corporate-governance systems and intellectual property rights and competition policies) that supports the generation of ideas as well as attempts at their commercialization and supplies the trained scientists and engineers to undertake such efforts.
Both of these assets are longer-lived and more geographically rooted than the ideas or innovations they generate. They contribute to high levels of economic and structural change that are beneficial to the economy overall, while imposing the costs of employment dislocation or displacement on some groups and individuals.
The current environment of intensified international and domestic competition and innovation is a legacy of an extraordinary policy success in the postwar period for which the United States and other industrial-economy governments should claim credit. Trade liberalization, economic reconstruction, and economic development have reduced the importance of immobile assets (such as natural resources) in determining competitive advantage.
These developments have lifted tens of millions of people from poverty during the past 50 years and are unambiguously good for economic welfare and global political stability. Nevertheless, these successes mean that competitive challenges and, perhaps, recurrent crises in U.S. industrial performance will be staples of political discussion and debate for years to come. This economy needs robust policies to support economic adjustment and a world-class R&D infrastructure for the indefinite future.
David C. Mowery is Milton W. Terrill Professor of Business Administration at the Walter A. Haas School of Business at the University of California, Berkeley.