Using Human Resource Data to Track Innovation: Summary of a Workshop (2002)

The National Academies’ Board on Science, Technology, and Economic Policy (STEP) conducted a workshop on November 23, 1999, to consider how more systematic exploitation of data on professionals—their training, mobility and career paths, functions in corporations, relationships across sectors, and productivity—could improve understanding of the process of innovation.

The premise of the workshop was that innovation—the invention, commercialization, and diffusion of new products, processes, and services—is an important determinant of economic growth, productivity, and welfare, and we want to understand better the factors that govern it. Ultimately, we want to use that understanding to adjust public policies to stimulate innovation and influence its direction to improve public health, enhance national security, and protect the environment as well as to foster economic development. Although it is a long way from a better understanding of the characteristics of innovation and how they are changing to policy prescriptions, measures informed by this understanding could take a variety of forms—increased public investment in research and training of scientists and engineers in particular fields, modifications of regulations that impede innovation, or adjustments to tax rates or intellectual property policies.

Innovation’s role in improving the competitive performance of U.S. industry in the 1990s has been a principal focus of the STEP Board’s attention, along with that of other economic analysts. Among other activities, STEP

commissioned 11 industry case studies to determine the extent and sources of U.S. industrial resurgence from the mid-1980s to the latter half of the 1990s. Summarizing the findings regarding sectors as diverse as steel and semiconductors and food retailing and banking, David Mowery observed:

Ideally, the collection of data on innovation should be guided by a solid theoretical understanding of the process and its impact. Such understanding of causes and effects, while being developed, is not very far advanced. It is possible, however, to elaborate a conceptual framework that encompasses direct indicators of innovation, what we believe to be the principal influences on innovation, and its effects on economic performance as a way of cataloging innovation indicators and data. The broad categories of innovation information include

• influences on innovative activity, ranging from market conditions to investments in R&D and human resource capacity to the ways knowledge is communicated and used (inputs);

• innovation characteristics as evidenced by new products, processes, and services introduced in the market (outputs); and

• the effects of innovation on firms, workers, regions, and the economy as a whole (outcomes).

The pattern, determinants, and effects of innovation involve variables that change over time, sometimes remarkably quickly and radically. At no time has that been more apparent than in recent years. At the workshop Paula Stephan, in introducing her commissioned paper, described several changes that are widely assumed to be occurring rapidly, to be substantial, and to have significant consequences for economic performance in the short term or in the longer run. These include (1) shifts in the industry and technology distribution of innovative activity, (2) shifts in the time horizon of innovative effort and investment, (3) changes in the

organizational structure of innovative activity, and (4) changes in the location of innovative activity both within the United States and globally.

The presumed changes and some speculative concerns about their implications are as follows:

• Distribution. The sectoral and technological distribution of U.S. industrial research and innovation is shifting toward nonmanufacturing and new emerging industries and technologies. This is of concern to the extent that some industries are becoming less capable of attaining and sustaining long-term competititive advantage.

• Orientation. U.S. firms have been conducting less fundamental research with longer-term payoffs, focusing instead on more incremental innovative efforts with clear market applications and generally shorter time horizons. This may lead to a weakening of U.S.-located innovative capacity in the long term, although it may have enhanced firms’ profitability in the near term.

• Organization. In-house R&D activities have been decentralized as firms have shifted away from the central R&D laboratory model. Decentralization of intrafirm innovative activity may have resulted in greater integration of R&D with corporate strategy and with other business units (product design, marketing, etc.), but unique technical resources of central corporate laboratories may have been dissipated in the process. Another organizational change is the increase in collaboration and outsourcing of research and innovation, not only among firms but also between firms and universities and government laboratories. This may improve the efficiency of innovative activities by reducing redundancies and accelerating implementation and diffusion, but it may entail a “hollowing out” of companies’ innovative capacity, with possible negative effects on long-term competitiveness.

• Location. Firms in some industries have shifted R&D capacity abroad and entered into alliances with foreign firms, possibly gaining access to talent, technology, and markets but also somewhat reducing the likelihood of locating future activity in the United States. At the same time, foreign firms have invested more in U.S.-based R&D activities. As firms continue to outsource many of their innovation-related activities and work closely with universities and other public and private institutions, successful innovation appears to depend increasingly on geographically clustered networks of related organizations. Regions such as Silicon Valley, Research Triangle Park, and Austin, Texas prosper while regions lacking key infrastructure may experience economic stagnation.

These alleged changes in innovation patterns, even if real and substantial, are not permanent. Others will be identified in due course and

become the focus of attention. Nevertheless, the changes described are the source of much of the current discussion of the need for improvements in industrial science and technology indicators and data. Consequently, they are useful partial tests of how well current innovation data serve our analytical purposes.

In part because they are the most consistently collected data and represented the best data time series related to innovation, R&D expenditures are often taken as the best or even surrogate indicator of innovation. In the conceptual scheme outlined above, R&D represent innovative effort and an important influence on whether new products and services and processes are developed and commercialized; but they are not a substitute for direct measures of innovation.

Paula Stephan and other workshop participants observed that the most commonly cited R&D data, from the National Science Foundation’s Industrial Research and Development (RD-1) Survey, have other limitations as a source of information on innovation, and particularly on recent changes in innovation processes (see also Cooper and Merrill, 1998).

• Distribution. An effort in recent years to include nonmanufacturing firms in the survey has confirmed that there is substantial and growing R&D activity in service industries. But the survey reports expenditures only at the firm level, classified in 2- or 3-digit Standard Industrial Classification (SIC) categories, obscuring the composition of activity within large diversified firms, and precluding analysis of activity in emerging technologies or industries lacking an SIC category (e.g., biotechnology).

• Orientation. NSF surveys of both public and private R&D expenditures classify activities as basic research, applied research, and development, which are defined in standard ways. (The industrial survey no longer asks for a short-term/long-term breakdown of expenditures.) Corporate basic research did decline in the mid-1990s but only temporarily and partly as a function of growth in overall private R&D activity. Besides, it is not clear what bearing the basic/applied/development classification has on the concern that corporate R&D has shifted to shorter term, more incremental objectives.

• Organization. The collection of R&D spending data at the corporate rather than the business unit level is a severe handicap in assessing whether there has been a redistribution of activity and tighter integration of R&D and business strategy within most firms. Business R&D data show modest growth in contracting out and in contributions to universities in the 1990s (and the revenues have clearly become more important to the recipients), but there has been little increase in their share of all corporate R&D. Likewise, counts of consortia creation, R&D joint venture agreements, Cooperative Research and Development Agreements (CRADAs)

with federal laboratories, and mergers and acquisitions, while in some but not all cases increasing over time, fail to the indicate the origin, longevity, intensity and value of the collaborations. Moreover, these data are confined to formal, usually contractual relationships, while there is a dearth of statistical data on individual and informal transactions—technical communications between firms, customer-supplier exchanges, participation in scientific and technical meetings, and transactions involving transfers of knowledge by consultants, accountants, systems integration firms, and a variety of other sources.

• Location. The industrial R&D survey includes private and foreign-owned as well as publicly held companies but covers activities carried out in the United States only. Reporting of R&D funds spent abroad is voluntary. Moreover, again because the survey reports expenditures at the firm level geographic detail on the distribution of domestic expenditures is limited.

Although in the terms of the classification outlined at the beginning of this report, there are other kinds of innovation input or activity data—for example, counts of patents, counts of contractual collaborations, and counts of inventions or innovations—none goes very far in overcoming these limitations.

Data on professionals’ skills, work, relationships, and productivity is an alternate or complementary source of information on industrial innovation. That is not merely because considerable data exist but because technological advance depends on human resources. In economic terms, technology is embodied in human as well as physical capital, and the interaction of scientists, engineers, and technicians is a principal means of technology diffusion. Changes in the distribution, orientation, organization, and location of innovative activity and capacity are reflected in changes in the training and mobility (education and career paths), the place and structure of work and affiliations (e.g., industry, occupation, allocation of time, consultancies, and informal collaborations), and the output and productivity of skilled personnel. Despite the obvious importance of human resources, national HR data have been underutilized and insufficently appreciated.

Discussion of uses of HR data at the workshop focused primarily on scientists and engineers with at least bachelor-degree-level training on whom there is fairly robust information from surveys (described in Chapter 2) sponsored by the National Science Foundation. But of course scien-

tists and engineers are not the only contributors to innovation. It would be useful to consider what other kinds of professionals are instrumental to innovation and what information about them exists or could be acquired, but these topics were largely beyond the scope of the workshop.

The workshop elaborated the following typology of professional characteristics primarily relevant to the science and engineering workforce that might have a bearing on innovation and on which survey data are or could be collected:

Training

• Educational qualifications and whether in or out of science and engineering

• In-service training or continuing education

Employment and Mobility

• Changing established firms or sectors

• Multiple institutions

• Domestic geographic movement

• International geographic movement

Structure of Work and Affiliations

• Allocation of time among functions (e.g., research, administration, information technology management, investment, analysis, etc.)

• Participation in interfirm or multifirm projects (consortia)

• Coauthorship of scientific or technical papers

• Consulting relationships

• Participation in other collaborations

Output/Productivity

• Patents applied for or received

• Publications

• Citations

• Salaries and other income (e.g., royalties, fees, etc.)

Various workshop participants suggested ways in which these characteristics might be used to illuminate changes in innovation processes. The examples are illustrative rather than exhaustive.

• Sectoral Distribution. Paula Stephan pointed out that the nonmanufacturing sector of the economy accounts for an even larger share of the industrial employment of scientists and engineers—about 45 percent—than its share of industrial R&D expenditures—about 25 percent. An examination of the roles of scientists and engineers in service sector firms would in all likelihood confirm that to a large extent innovative activity in that sector entails applications of information technology to customer service, marketing, inventory, and logistics rather than activities represented by standard R&D data. Similarly, the role of scientists, many of them academics, in forming and advising new firms has been critical to the emergence of industrial biotechnology and illustrates its close connections to publicly funded fundamental research. (See Chapter 3).

• Time Horizon Orientation. Subtle shifts in the focus of R&D and innovation within established firms might be illuminated by changes in the mix of personnel and their qualifications. For example, a reorientation toward short-term projects and incremental innovation might be associated with a change in the mix of scientists and engineers or of technical versus nontechnical professionals. Similarly, shrinking time horizons might be associated with technical professionals devoting less time to research and more to production engineering.

• Organization. The movement of technical personnel within firms could be a leading indicator of decentralization and the creation of teams of professionals with mixed qualifications and experience an indicator of the integration of R&D with other business functions—strategy, production, and marketing. The rise of outsourcing, alliances, and other collaborations is presumably associated with higher rates of copublication, coinvention, and consulting relationships. Another way to assess the incidence of collaborations and their importance to firms is to determine how senior technical personnel allocate their time between in-house activities and managing external relationships. Probing further, one could use data on individuals to shed light on the origins of collaborations and how they change over time.

• Location. Determining the distribution of scientists and engineers within and across firms and tracking changes over time would shed a good deal of light on the extent of regional clustering and globalization.

Creative uses of human resource data could also explore links among characteristics of professionals working in firms, innovations, and traditional measures of performance of firms—sales, profitability, employment, and productivity. Stephan cited the work on industrial biotechnology of Lynne Zucker and Michael Darby, who show that the extent of

collaboration between firm scientists and leading (“star”) academic scientists is a powerful predictor of firms’ success in terms of how many products they have on the market or in development, their sales, and their size and the value of their stock (Darby, 1999).

Donald Siegel of Nottingham University discussed a second important set of effects of innovation, on the workforce itself and how the workforce is managed. There is a strong theoretical case for believing that technological innovation is skill-biased; it increases the demand for highly educated and highly skilled workers because they have a comparative advantage in helping companies implement new technologies effectively (Siegel, 1999). Testing the hypothesis depends critically on data not only on specific technology adoptions but also on labor composition and relative compensation before and after the implementation of new technology. Siegel pointed out that a good deal of research has been on the manufacturing sector and the effects of process innovations, such as computer-aided design and manufacturing, robotics, flexible manufacturing systems, and just-in-time inventory systems. But much work remains to be done on the effects of service sector innovation and in distinguishing types of technologies with different workforce efforts, developing better measures of skill than educational level, and addressing intervening variables such as organizational and human resource management changes.

Workshop participants also observed that data on professionals’ career paths and productivity can supplement other means of evaluating government research and education programs. The link to government education grant, traineeship, fellowship, and research associateship programs is obvious. But federal research grants and contracts also indirectly support a significant share of graduate students in most fields of science and engineering. Government laboratories provide postdoctoral and training to scientists and engineers who leave to work elsewhere, often in industry. As a general matter, linking output and human resource data to government funding data provides an important avenue for measuring government performance. The Academies’ Committee on Science, Engineering, and Public Policy (COSEPUP), in a series of workshops and reports, has discussed the importance of using human resources output as a performance measure for research under the Government Performance and Results Act of 1993 (National Academy of Sciences et al., 1999, 2001).