Agrawal Henderson (2002) - Putting Patents In Context Exploring Knowledge Transfer From Mit

Jump to navigation Jump to search
Has bibtex key
Has article title Putting Patents In Context Exploring Knowledge Transfer From Mit
Has author Agrawal Henderson
Has year 2002
In journal
In volume
In number
Has pages
Has publisher
©, 2016


  • Agrawal A., and R. Henderson (2002), "Putting Patents in Context: Exploring Knowledge Transfer from MIT", Management Science. pdf


In this paper we explore the degree to which patents are representative of the magnitude, direction, and impact of the knowledge spilling out of the university by focusing on the Massachusetts Institute of Technology (MIT), and in particular, on the Departments of Mechanical and Electrical Engineering. Drawing on both qualitative and quantitative data, we show that patenting is a minority activity: a majority of the faculty in our sample never patent, and publication rates far outstrip patenting rates. Most faculty members estimate that patents account for less than 10% of the knowledge that transfers from their labs. Our results also suggest that in two important ways patenting is not representative of the patterns of knowledge generation and transfer from MIT: patent volume does not predict publication volume, and those firms that cite MIT papers are in general not the same firms as those that cite MIT patents. However, patent volume is positively correlated with paper citations, suggesting that patent counts may be reasonable measures of research impact. We close by speculating on the implications of our results for the difficult but important question of whether, in this setting, patenting acts as a substitute or a complement to the process of fundamental research.

Keywords: Patents; University Science; Knowledge Transfer; Technology Transfer

Short Summary

Patenting and Licensing data have become important because:

  • Patenting requires inventor names, dates, assignee locations and other crucial information to be recorded, and this information is rarely recorded elsewhere
  • Patented innovations are, by definition, expected to be commercially useful
  • Patenting data is available in machine readable formats
  • Licensing data has recently become more readily available (thanks to the Association of University Technology Managers, AUTM).

Public support for university research is based upon the belief that the private sector will underfund basic (or fundamental) research, because the results are difficult to appropriate.

  • (EE) Basic research has large positive social externalities and small private benefits, as well as large costs.
  • Focusing more on patenting might substitute or complement basic research - the finding is one of complementarities.
  • The Bayh-Dole Act (1980) gave universities (and other entities partially or fully funded by federal funds) control over their inventions, and therefore acts as an incentive to create non-basic (commercializable) research.

Professors (and universities) transfer knowledge through:

  • Mentoring students (who may do research or be recruited, etc)
  • Giving conference presentations
  • Free publication of ideas in refereed journals
  • Consulting
  • Collaborative research
  • Patenting


  1. Is patented research representative of all research?
  2. Is it just the mode of transfer that is different?
  3. Does patented work have a different impact?

The paper considers the supply side of the research market. Patenting is NOT the only source of knowledge flow and is not representative. Focusing on patenting in isolation will significantly misrepresent the impact of university generated knowledge on the economy. Studies that estimate this impact must include other channels.


  • Data from faculty who published or patented at MIT Mechanical Engineering (ME) or Electrical Engineering and Computer Science (EECS) from 1983-1997.
  • 236 faculty, 640 patents, 5132 papers. Citations on papers and patents. Interviews with 68 faculty who had patented.
  • 35% from ME, 65% from EECS
  • Problem of selection bias as only 'successful' faculty survive


General findings:

  • Patenting output by universities (MIT, CalTech, Carnegie Mellon, Columbia, and Stanford) has increased approximately four fold in the sample period.
  • MIT increased from 23 (1983) to 130 (1997) - approximately 5-6 fold.
  • Av. faculty publishes 1.5-2 papers per year
  • Av. faculty patents 0.25 patents per year
  • Ratio of patents to pubs rose from 0.11 to 0.13 over the period (but c.f. Life Sciences)
  • Selection bias (checked against population patent-to-pubs ratio) is rejected
  • Distributions are skewed:
    • 44% never patent, <15% have >5 patents, <6% have >10 patents.
    • Pubs dist is less skewed: 14% have pub'd >35 papers, 5% > 100 papers

Specific Findings:

  • Only a small number of faculty patent each year (10-20%)
  • Most faculty publish every year (60%)

Patenting as One Mechanism among many

  • Patents are not a major source of knowledge transfer:
    • Faculty estimate that only 6-7% of knowledge transfer occurs through patents (c.f. Cohen et al.'s 11-12%)
    • Pubs account for about 18% (same as Cohen et al.)
    • Informal channels (consulting and conversation) account for about 31%

Different Firms, Different Channels

The paper looked at firms that collaborated on papers and on patents, as well as the set of firms that cite papers and patents produced by professors in the sample.

  • The set of firms that colloborate on patents are different from those that collaborate on papers
  • The set of firms that cite patents differs from that which cites papers

Patenting as predictor of publishing

Patenting has a slight negative correlation, not statistically significant, with publishing. However, patents are positive significant predictors of citations (i.e. patents predict research quality). The resutls are estimated using three one-year lag variables. Granger causality tests (in both directions) suggest that patenting and publishing are independent.

Substitute or Complement

Generally basic research is considered a substitute for more applied work in commercial settings. The same is feared in academic settings. However, neither patenting or publishing were generally found to be the motivation for research agendas in interviews. Although a small group of faculty do patent much more proportionately than their peers, on average there is no negative correlation. The correlation between patents and citations (to pubs) might reflect that the research can be immediately applied. But citation counts include both academic and commercial, so the paper concludes that patenting "may actually be a complement to fundamental research".