Putting Patents in Context: Exploring Knowledge Transfer from MIT
Management Science, 48, 1, (2002) pp. 44-60.

Ajay Agrawal (University of Toronto) and Rebecca Henderson (MIT)

FULL TEXT

- Summarized by Ajay Agrawal

Summary

Knowledge transfer from universities is important. A significant portion of productivity-enhancing innovations stems from discoveries made in university science and engineering labs, and it is useful to deepen our understanding of how knowledge flows from the academy to the economy. Unfortunately, knowledge flows are notoriously difficult to measure: the transfer of knowledge is largely invisible and leaves no paper trail. One important exception, however, is patent data.

Patent data has been used to study knowledge transfer for three reasons. The first is simple availability. Patent data offers a unique record of information concerning the characteristics of an invention, the identity of inventors, the identity of the inventors’ employer, the date of filing (an approximation of the date of invention), and other useful information. In addition, studies that also include patent licensing data can track the specific firms that develop a given invention and may even estimate the value created by measuring royalty income collected by the inventor. The second reason is commercial application. Since a wide variety of research is conducted at universities, much of which has little apparent commercial application, patent data offers a sub-sample of inventions that are more likely to be commercially viable. Third is growth in patenting activity. Arguably, university patenting is particularly interesting to study because it has been growing at a significant rate, especially since the early 1980s. This is at least partly due to the passage of the Bayh Dole Act, which permitted universities to retain title to inventions developed under federally-funded research programs, and due to the growth of biotechnology and computer science, both as fields of research and as markets.

Still, though patenting is the most measurable channel of knowledge transfer, it is certainly not the only one. University researchers also transfer knowledge to the private sector through a variety of other mechanisms, including publishing, conference presentations, and consulting. Since so much of past empirical research on university knowledge transfer has focused on patenting, we are interested in exploring how significant patenting is as a channel for university knowledge transfer, and how representative the patenting channel is of other channels. To examine these questions, we collected patenting and publishing data associated with 236 faculty members of the Mechanical Engineering and Electrical Engineering/Computer Science departments at MIT from 1983 to 1997 who were still at MIT in the year 2000. In addition, we conducted interviews with 68 of these researchers. We found several unexpected results.

First, the patent channel is surprisingly small. The average professor published approximately seven papers for every one patent he or she was issued. Viewing the data from a different perspective, we can say that only a small fraction of the faculty engaged in patenting activity. While more than 50 percent of our sample published at least one paper in any given year and less than 3 percent had yet to publish, nearly half the sample had never patented at all. This finding is further reinforced by our interview results. In each interview, we carefully reviewed the respondent’s CV, recording his or her perception of the impact of particular research projects and how the associated knowledge was transferred to the private sector. Our results concerning the relative importance of knowledge transfer channels are as follows: 1) consulting (26 percent of knowledge transfers), 2) publications (18 percent), 3) recruiting graduate students (17 percent), 4) collaborative research (12 percent), 5) co-supervising graduate students (9 percent), 6) patents and licenses (7 percent), 7) informal conversations (6 percent), and 8) conferences (5 percent). These results are quite similar to those reported by Cohen et al, who asked U.S. manufacturing firms how important they considered various knowledge transfer channels from the university to their industry.

Second, the patent channel is not broadly representative of other knowledge transfer channels. If it were—in other words, if the firms that collaborate with professors on patents and cite faculty patents were the same firms that engage in other channels of access—then a focus on patenting would offer a more useful lens through which to view the impact of the university on the economy. However, that is not the case. Of the 271 firms that collaborated with MIT faculty during the period under investigation, only 9 percent collaborated on both patented and published research, while 58 percent collaborated only on papers, and 39 percent only on patents. Similarly, a full 24 percent of the firms that cited MIT research cited only papers, not patents.

Finally, we examine whether patenting activity can be used as a predictor of publishing behavior. While we do find that past publishing activity predicts current publishing and past patenting activity predicts current patenting (the number of papers in year t is strongly correlated with the number of papers in year t-1, and similarly for patents), we do not find a significant correlation between these two activities. In addition to examining simple correlations, we perform a variety of regression analyses, but our core result holds: patenting activity does not appear to be significantly related to publishing behavior. We do, however, find a positive relationship between the cumulative number of times a paper is cited and the cumulative number of patents issued. So while patent counts are not good predictors of paper counts, or of the volume of a faculty member’s research, they do reflect a given paper’s impact.

In summary, while patent data is certainly a valuable tool for facilitating empirical research on knowledge flows, it has important limitations. Our results suggest that studies of university knowledge transfer based on patent data must be interpreted carefully, and generalizations should be viewed with caution. This study clearly highlights the pressing need for empirical research on other knowledge transfer mechanisms, perhaps most importantly consulting, publishing, and the mobility of graduate students. Such research will provide important complementary insights to those that patent data offers.

References

Cohen, W., R. Florida, L. Randazzese, and J. Walsh (1998) “Industry and the Academy: Uneasy Partners in the Cause of Technological Advance” R. Noll, ed. Challenges to the Research University, Brookings Institution, Washington, D.C.

This work is licensed under a Creative Commons License.