Patents the Innovation Landscape for Scientific Research

1. Patents & the Innovation Landscape for Scientific Research Fiona Murray* Associate Professor in Management MIT Sloan School March 2009 joint with Garth Sheldon-Coulson MIT & Harvard Law School

2. How should we conceptualize & analyze the role of patents in scientific research? Traditional view (post Bayh-Dole) University funding & reputation-based rewards provide adequate incentives for scientific research in academia Patents (over the research outputs) provide key incentives for investments in follow-on innovation

3. Conceptually Research as a single �line� with ideas from one stage needing significant investment in commercialization to move to the next stage�

4. Observed rapid rise in patenting across (US) universities

5. To put it another way�

6. Unintended consequences If innovation is a cumulative step-by-step process along a research line [not simply a matter of investing to commercialize] with patents at each stage then�

7. Unintended consequencesReduced incentives for follow-on innovation Upstream patent holders may limit the incentives of downstream innovators (Scotchmer 1987; Green & Scotchmer 1990; Anton & Yao 1994)

8. Academic example A123 Licensed material technology for a battery from MIT Material less effective than anticipated � requires complex additional research � newly adapted material patented Firm invests in plug-in hybrid battery development with new material Still pays �reach thru� to MIT for initial patent

9. But� Single �line� perspective on cumulative innovation at odds with recent theorizing about nature of innovation landscape: Creativity & innovation as the combination of disparate ideas (Johansson 2004) Innovation as recombination (Romer 1990; Fleming 2007) Innovation as movement in design space or fitness landscape (Levinthal 1997; MacCormack 2007)

10. Re-conceptualizing the cumulative innovation landscape Multiple possible research �lines� There are many biological pathways via which to attack Alzheimer�s disease There are many ways to develop an alternative vehicle that meets emissions & efficiency requirements! Single step can require multiple inputs (n -> 1) Single step can lead to multiple outputs (1 -> n) Nature of steps vary - some are more focused on exploitation while some are exploratory & open up new arenas (March 1991)

11. Innovation Landscape

15. Overlaying the patent landscape on the innovation landscape Single vs. multiple patents (Shapiro 2001) Patent breadth (Lerner 1995) Patent ownership fragmentation (Ziedonis 2006) E.g. 4,382 of the 23,688 genes in the human genome are claimed in ~ 4270 granted U.S. patents

18. Unintended consequencesReduced incentives for single follow-on line If innovation is cumulative but requires multiple [patented] inputs for subsequent stages then�follow-on innovators may be deterred by a patent �thicket� (Shapiro 2001) Face expropriation, transaction costs & uncertainty - lower innovation may result leading to an anti-commons effect (Heller & Eisenberg 1998) � gridlock (Heller 2008)

19. What happens when we impose gridlock? Evidence from Huang & Murray (2009) Significant decrease in rate of follow-on research (publications) after patents on gene sequences have been filed & granted Decline amplified by: Fragmentation of patent ownership Number of salient patents on a gene Strength of evidence for the gene in cancer or human disease more broadly (OMIM)

21. Unintended consequencesReduced incentives for multiple follow-on lines Mouse engineering technologies make many different mice All patented � claiming use of technology & mice made with it IP on Cre-lox & Onco exclusive to DuPont 1990s Required researchers to �acknowledge company's rights to the animals, share any money that may be made on discoveries from the technology, & distribute the animals only to other researchers whose institutions have agreed to these terms� (Science 1997) 1998 NIH-DuPont make Cre-lox & Onco subject to a �simple� license & no reach-thru

22. What happens when we remove roadblocks? Evidence from Murray et al. (2009) Significant increase in rate of follow-on research after the NIH agreements (compared to rate for unaffected knock-out mice) Boost concentrated in: Contributions by �new� authors or institutions More diverse types of research Follow-on boost in �lines� depends on type of change: Broad, speculative new exploration-oriented lines for Cre (not accessible prior) Narrow, applied exploitation-oriented lines for Onco (accessible but with RTL prior)

25. Concluding thoughts Broader framework within which to identify & consider wide variety of cases e.g. gridlock & roadblocks & others� Re-orients attention to relationship between patent landscape & innovation landscape & two critical characteristics: Multiple paths to the same goal� Limits the impact of a patent along a particular line but focusing attention to opportunities for inventing around Highlights potential costs of patents whose breadth is high relative to the opportunities for inventing around in the innovation landscape (see Acemoglu 2009) Suggests other policy remedies in the face of broad IP e.g. additional research funding, patent purchases etc. Downstream branching� Amplified the impact of a patent particularly those that block many branching lines Highlights potential costs of patents licensed by upstream holders who ignore the benefits of branching in their licensing strategies Suggests policy intervention for tools, GPTs etc. focused on access