be linked to enable a simultaneous examination of worker characteristics and firm characteristics. In addition, there are databases such as patent files and indexes to scientific publications that can be searched by individual. Some government research agencies, universities, and professional associations track the employment and work product of individuals affiliated with them; and, of course, scholars construct their own data sets for particular purposes.
SESTAT data show several trends in the deployment of PhD scientists and engineers in the United States—the increased role of industry as an employer relative to universities and government, the increased importance of the service sector relative to manufacturing industries, and the movement from the laboratory into non-R&D positions. These trends illustrate broader changes in the innovation process—the reduced role of government-financed R&D, the emergence of the service sector as a locus of much innovation, and the integration of R&D into other functions of the firm—strategy, management, and marketing. Perhaps the best examples of productive uses of HR data are research on the growth of biotechnology, which has no standard industry classification category, and research on alliances among firms and collaborations between firms and university researchers.
Much can be done to enhance the utility of human resource data. More precise information on employers and their locations would enable analysts to relate individual and firm characteristics and help compensate for the lack of business unit detail in the national R&D expenditure data. More information on what scientists and engineers do in firms would help illuminate the relationship of research to other functions. Information on scientists’ and engineers’ outputs (e.g., publications, conference proceedings, and patents) would help illuminate R&D spillovers between firms and across industries. And information on people in science and engineering (S&E) occupations who nevertheless lack a baccalaureate degree would help us understand a prevalent pattern in information technology.
Important advances in the analysis of innovation would come from linking human resource data with other data sets. For example, matching the name and location of respondents to the NSF surveys with Census establishment data could illuminate the relationship between firm size and innovation and between internal and external sources of innovation.
Finally, there is a need for new data on how people actually spend their time. Knowing what share of industrial scientists’ time is spent in the laboratory, managing other employees’ research, assessing the capabilities of other firms, collaborating with professionals outside the firm, or in production engineering would tell us a good deal about how innovation is taking place.