The Tragedy of a Bad Idea

    By Kevin E. Noonan --

Some (bad) ideas are a long time a-dyin'.  One of the most persistent in biotechnology patent law is the concept of the "tragedy of the anticommons."  An intriguing idea when first proposed by Heller and Eisenberg in 1998, it was also frightfully naïve about the "anticommons" effects of circumstances much more germane to most academic molecular biologists (like priority and career advancement).  The core idea was that permitting genes (particularly and parochially human genes) to be patented would create a "tragedy of the anticommons," where private ownership (however limited in duration) would somehow prevent genetic science to advance.  Of the many limitations of the idea, it neglected to distinguish the difference between commercial exploitation of isolated nucleic acids to produce otherwise unavailable therapeutic and other products (such as erythropoietin, human growth hormone, human insulin, tissue plasminogen activator, and a host of others now comprising the more than 400 biotechnology drugs under development) and the information from which genetic science has garnered the greatest advances in our understanding in history.

As it has turned out, the concept of the "tragedy of the anticommons" has been shown time and time again to be ephemeral to the point of non-existence.  (It is tempting to speculate that the idea persists because the phrase is so clever.)  Scientific studies have clearly established that there is no tragedy, and no anticommons.  These peer-reviewed studies include reports from the U.S. (Walsh et al., 2003, "Science and the Law:  Working Through the Patent Problem," Science 299: 1020; Walsh et al., 2005, "Science and Law:  View from the Bench: Patents and Material Transfers," Science 309: 2002-03), Germany (Straus, 2002, Genetic Inventions, Intellectual Property Rights and Licensing Practices), Australia (Nicol et al., 2003, Patents and Medical Biotechnology:  An Empirical Analysis of Issues Facing the Australian Industry, Centre for Law & Genetics, Occasional Paper 6) and Japan (Nagaoka, 2006, "An Empirical Analysis of Patenting and Licensing Practice of Research Tools from Three Perspectives," presented in OECD Conference in Research Use of Patented Inventions, Madrid) finding that "patent thickets" (Shapiro, 2001, "Navigating the Patent Thicket:  Cross Licenses, Patent Pools, and Standard Setting," In: Innovation Policy and the Economy 1: 119-50) or the "anticommons" (Heller & Eisenberg, 1998, "Can Patents Deter Innovation?  The Anticommons in Biomedical Research," Science 280: 698-701) rarely affects the research of academic scientists.

The paucity of the concept is demonstrated once again in a paper by Fiona Murray (at left) and Scott Stern (below right), from the Massachusetts Institute of Technology and Northwestern University, respectively, and published in the Journal of Economic Behavior and Organization ("Do formal intellectual property rights hinder the free flow of scientific knowledge?  An empirical test of the anti-commons hypothesis," 63: 648-87), published in 2007 from data collected between 1997-1999.  (A hat tip to one of our favorite members of the anti-gene patenting crowd for overstating (and as we will show, misinterpreting) the paper's conclusions with sufficient (and typical) hysteria that it piqued our interest.)  The authors state their prejudices in the first section of the abstract, citing "a growing 'anti-commons' perspective [that] highlights the negative role of IPR [intellectual property rights] over scientific knowledge."  In the service of this idea the authors have performed a statistical study to try to assess whether citation of scientific papers declines after a patent on the technology is granted.  As one might expect, the answer is "Yes," but as the authors themselves recognize the effect is "modest" at best.  While the authors seem to have appreciated that their study was performed on a sample certain to reflect the effect if it existed, the weakness of their results with this dataset negates their thesis that there is an anticommons effect or that any perspective based upon that mythical effect is or should be growing (outside the world of academic journals).

The study was based on 169 patent-paper pairs (and a total of 340 articles), wherein there is first a scientific paper published in a journal (Nature Biotechnology) and later the technology disclosed in the paper is encompassed in a granted U.S. patent.  The authors put it this way:  "if the granting of IPR raises the cost of building on a specific piece of knowledge, then the citation rate to IPR-linked scientific publications should decline post-patent grant."  The authors then perform a statistical analysis using the formula:

CITESi,t = f (εi,t ; γi + δt-pubyear + βt + ψWINDOWWINDOWi,t

+ψPOST-GRANTPOST-GRANTi,t )

 "where γi is a fixed effect for each article, δt-pubyear captures the age of the article, βt is a fixed effect for each citation year, WINDOW is a dummy variable equal to one in the year in which a patent is granted and POST-GRANT is a dummy variable equal to one only for years after the patent grant year for an individual article."  The authors posit that since empirical evidence has established that academic researchers do not undertake proactive monitoring of IPR (citing Walsh et al., 2003 and Walsh et al., 2005), reductions in citation of articles having disclosure also encompassed in a patent "is driven by surprise arising from the IPR."

The results are, as the authors acknowledge, modest:

Table 3 compares the means of patented and unpatented articles within the sample.  A few notable differences stand out.  First, the average rate of citation is relatively similar across the two groups, with the patented articles receiving, on average, just over an additional citation per article-year over the sample.  However, this 10 percent average difference masks more substantial differences that manifest themselves over time.  In Fig. 3, the average FORWARDCITATIONS are plotted by AGE (years since publication).  During the year of publication and in the subsequent 3 years, PATENTED articles have a significant citation advantage, equivalent to nearly a 20 percent "boost" over the citation rates for non-patented articles.  However, in the fourth and fifth years after disclosure in the literature, patented articles converge to the citation rate associated with non-patented articles.  As we explore further in the next section, it is during these later years in which patented articles are in the post-grant phase . . . .  These data suggest that the number of citations and the article characteristics vary across the margin of whether or not an article is part of a patent-paper pair.  While the drop-off in the citation advantage associated with patented articles is consistent with the presence of an anti-commons effect, such an effect could result from differences in the characteristics of articles represented in the different age-cohort categories, which we explore in our more detailed empirical analysis.

In the empirical analysis section of the paper, the authors show that:

Three key findings stand out:  while patented articles enjoy approximately a 20 percent overall citation boost, a patent grant is associated with an insignificant (though negative) impact on citation in the year in which the patent is granted.  However the post-grant effect is associated with a statistically significant 19 percent decline in the expected citation rate.  In other words, the initially higher citation rate for patent-paper pairs is erased in the years after a patent is granted. . . .  Finally, in fourth column of Table 5, we report a differences-in-differences estimate, including a separate fixed effect for every article, as well as a complete set of fixed effects for age and citation year; as such, these estimates are identified exclusively from the within-article contrasts between pre-grant and post-grant citation levels (after accounting for the impact of article age and year).  According to this specification, the estimated post-grant decline is over 10 percent (and is significant at the 5 percent level).  Moreover, these baseline results are robust to alternative specifications and sample definitions.

The decline was found to be more pronounced with the number of years since the date of the patent grant, and interestingly "is particularly salient for articles authored by researchers with public sector affiliations," because the authors hypothesize that acquisition of IPR rights by inventors from public institutions are more likely to be a "surprise" to other academic researchers (although they provide no evidence for this except a correlation with more frequent reduction in citations of papers in patent-paper pairs assigned to public institutions).

While the authors confidently "reject the null hypothesis" that patents have no impact on the diffusion of scientific knowledge, there are a number of flaws in the analysis.  The first is that there is no evidence that the comparison considered the claims of the patents as opposed to the disclosure.  Since the claims define the patentees' "right to exclude," finding that the patents associated with these papers had claims of limited scope would severely compromise any conclusions drawn about the effects of the patent grant on forward citation (and presumably create an impediment to using the disclosed technology).  The most pronounced flaw is that the journal chosen as the source for the patent-paper pairs is simply not the journal where cutting-edge scientific discovery in molecular biology is published.  Several general and specialty publications (including Nature, Science, Cell, the Proceedings of the National Academy of Science USA, even the Public Library of Science (PLoS) website) serve that role.  There are several journals devoted to disclosing the types of scientific discovery associated with the "technology" of biotechnology, and Nature Biotechnology may be one of the preeminent journals of this kind.  But the types of articles published in this journal are precisely those that are intended to be associated with patents, diminishing any "surprise" anyone (particularly academic researchers disinterested in IPR) might encounter.

Moreover, the types of articles published in such journals have by nature a shorter "shelf-life" than basic research articles.  These papers are not generally fundamental discoveries, but rather articles describing technology having direct and evident biotechnology applications.  So the effects on "scientific knowledge" are minimal since the "knowledge" encompassed by these articles amount, in general, to research tools rather than basic scientific discoveries.

Finally, there are counter-examples galore regarding citations to articles that are later associated with patents.  For example, the fundamental papers on the polymerase chain reaction (PCR) have been heavily cited in the biotechnology literature since they wee published in the mid-1980's, yet PCR is associated with extensive IPR (first by Cetus and later by Hoffman-LaRoche).  So using forward citation as a metric can be confounded by instances where technology is developed and disseminated more broadly as the result of IPR supporting commercialization.

While the authors are properly "cautious in the interpretation of [their] findings," their intentions are clearly to show an anticommons effect.  This is not to say that they have not performed their experiment or analysis properly.  It is evident, however, that the authors found what they expected to find, although even they must admit that the effects are modest.  The only reasonable conclusion is that the modesty of the effects provides the most damning evidence that there isn't much tragedy in those anticommons.  And one day maybe everyone will be willing to admit it.

Comments

[Bartmans]

You name PCR as a counter-example of heavily patented technology that still is cited very much. Another example could be the work on RNAi, first published by Fire et al., who themselves have filed a patent on the technology. Also this field is heavily patented nowadays, but has led to an avalnche of scientific research and research papers.

[Dan Feigelson]

The Emperor has no clothes!

Thanks, Kevin.

[saddlepack maker]

Kevin,

I love your reference to the one that shall not be named. I too read the statement regarding Prof. Murray's paper and so I download that paper and a few others she has that will soon publish. I was surprised to see this sudden change of direction by Prof. Murray. She has had a long line of pro-entrepreneurial pro-biotech papers. Just currious.

[David Boundy]

The real tragedy here is that Murray & Stern measured the wrong thing. A far more important thing to measure is "How many ideas get translated into products, treatments, diagnostic assays" or at least into commercial research projects? Let's compare patented vs. unpatented ideas on that dimension...

All the journal papers in the world don't mean a hill of beans if they don't get turned into Economic Behavior and Organization.

[EG]

Kevin,

Nice piece. I really like the analogy to a "greek tragedy" with your symbol. May the "anti-commons" rhetoric perish as it should.

[Ron Katznelson]

Kevin,
Thanks for pointing out this topic.

As other academics had, it appears that Murray and Stern had made fundamental mistakes in their premises and analysis when it comes to patents and the diffusion of information disclosed in patents. I believe that their results do not show “evidence for a modest anti-commons effect”, as they claim, but rather evidence that patents and patent applications serve as a conduit for publication and knowledge dissemination that DISPLACE the need to cite the scientific journal paper. Here is where they commit the simple and glaring error:

On the one hand they acknowledge that scientific knowledge is “disclosed as a patent-paper pair”. On the other hand, they analyze and count forward citations only in ONE COMPONENT of that pair – the journal article. They posit that the citation rate [in scientific papers] to the scientific publication disclosing that knowledge should be lower than for scientific publications with no IP and should fall after formal property rights are granted. But the SAME observation is consistent with the simple fact that where IP does exist, patents are published and forward citations of the subject matter may be made citing THE PATENT and not the original journal paper. I have encountered many situations in published papers where the author preferred to cite the patent rather than the counterpart journal paper because the patent disclosure contained more detail.

Thus, the Murray and Stern study is flawed at its core because it fails to count the TOTAL citations to the “patent-paper pair” by counting only the citation to the paper and not the patent. Their “control” analysis is equally destined to find false support for the modest “anti-commons effect” because the relatively higher citation rate to the “control” papers is due to the lack of a patent that can displace some of these forward citations away from the paper.

Murray and Stern used the ISI citation index database. They could have easily done the right thing by using ISI’s patent citation section after they had gone to great length to identify the counterpart patent numbers. They apparently also failed to study the “patent-paper pair” citation literature, specifically discussing patents role in the article citation displacement phenomena. See for example: M. Bregonje, “Patents: A Unique Source For Scientific Technical Information In Chemistry Related Industry?”, World Patent Information, 27(4), pp. 283-368, (2005) (showing that patents’ share as a first information source to citing articles increased over two decades since 1980 at a rate of approximately 8%-10% per decade).

Murray and Stern also appear to fully ignore the intricate aspects of multiple points in time of patent and continuation publications and fail to take into account the effect of the 18-months application publication. They failed to identify which fraction of the papers had counterpart patent applications (including continuations and divisions) that published before the patent issued. Their whole experimental design predicated on the grant date of the patent (which they find to have been 2-4 years after filing) is counterfactually classified for their tested hypothesis in an unknown number of cases. They say at page 650 “[t]o the extent that a patent grant comes as a “surprise” to at least some potential follow-on researchers, this difference allows us to ask how does the grant of formal patent rights over such knowledge influence the trajectory of forward citations and therefore the impact of the scientific research findings in the public domain”. But in many cases where an application was published long before a patent grant, the “surprise” is discovered much earlier than their analysis is configured to investigate. They appear to give short shrift to the elaborate and diligent analysis and determination made by R&D organizations prior to making development investments based on rival’s published applications well before a patent is granted.

Finally, I submit that any serious and credible attempt to study the anti-commons effect (alleged suppression of downstream developments due to “blocking” patent rights) must include evaluation of established indicators of development activities and not only the scientific paper activities, as Murray and Stern attempted. To demonstrate suppression or deterrence due to “blocking” patents one must show evidence that indicators most tightly linked to commercial activity - patenting activities, have been suppressed or deterred. Thus, at the very least, in the citation indicator domain, the “patent-paper pair” should be truly taken up both ways: as a full unit of development activity both as the focal source for forward citation (discussed above) and as the downstream target from which citations are analyzed. If anti-common effects exist, it should be possible to show that U.S. patents cite disproportionately more scientific papers that belong to the “no patent counterpart” class than the citations of patents or papers belonging to the “patent-paper pair” class. Murray and Stern cannot burry their head in the sand and ignore the very profound linkage between patents issued for downstream development activities and the scientific papers in their selection. To appreciate the magnitude of the downstream picture missed by Murray and Stern, see F. Narin, K.S. Hamilton and D. Olivastro, “The Increasing Linkage Between U.S. Technology And Public Science,” Research Policy, 26(3), pp. 317-330, (October 1997) (showing that references in U.S. patents to U.S.-authored research papers have tripled over a six-year period, from 17,000 during 1987–1988 to 50,000 during 1993–1994, a period in which the U.S. patent system grew by only 30%); Analysis of citations in patents of the broader world scientific publications is summarized in: National Science Board, ‘Science and Engineering Indicators 2006.” Ch. 5, Outputs of S&E Research: Articles and Patents, (2006), at: http://www.nsf.gov/statistics/seind06/c5/c5s3.htm , (showing in Figure 5-58 that U.S. patent citations of science and engineering articles on an average per patent basis rose from 1987 to 2004 by more than a factor of 5). Thus, in addition to the points raised above, a proper study by Murray and Stern would have also included the forward citations found IN PATENTS, citing papers in their selected paper ensemble.

To sum up, I’ll believe in the anti-common effect when I see evidence for it. Thus far, I have seen no credible evidence – only folklore and flawed empirical biased frameworks from those who are predisposed to find such effects.

Ron Katznelson

[Kevin E. Noonan]

Dear Ron:

Thanks for your contribution to the discussion. I have one small quibble with your analysis. The papers chosed for the study were published between 1997-1999, prior to the time that US patent applications were published. On the other hand, some fraction of these applications were also published abroad (PCT, EP, etc.) so the "surprise" factor is diminished accordingly.

The authors use the decrease in citations of the paper as an indication that the patent inhibits further scientific inquiry. For the reasons in your comment, and the reasons in the post I think they are wrong. Indeed, it might be a more interesting study to look at journals disclosing more fundamental science to see if there is any effect (I think not, but I'd like to see the data). Unfortunately, when a hypothesis is negated in political waters like these, the study demonstrating it rarely sees the light of day.

Thanks again for the comment.

[Ron Katznelson]

Kevin,
Thanks for your response. Your comment addresses a secondary point I made and I assume that you agree with the analysis of the primary and basic flaw – ignoring patent citations. As to the secondary point, although the scientific PAPERS were published before the 18-month publication rules were put in place under the 1999 patent act, it is unclear what fraction of counterpart patent APPLICATIONS were in fact filed later than or during late 2000, when the publication rules were already in place. Given the 1-year grace period under Section 102 and the fact that continuations and divisions are filed months or years later, there must have been an unknown fraction of applications that fell under this category but the authors failed to identify those. Furthermore, as you correctly pointed out, there were some PCT applications and applications claiming foreign priority that would have been published earlier than the U.S. patent grant. This is why I wrote that the authors have counterfactually classified an UNKNOWN number of cases for testing their hypothesis. Do you agree that they should have used for each paper the earliest publication date of a counterpart patent application rather than the US patent grant date?

Ron Katznelson

[6]

I tend to agree with you guys on this whole anti commons thing probably not existing. But if it does, and they were god and thus they shoved infinite evidence in your faces, I doubt any of you would change your stance.

And that is sad.

[Kevin E. Noonan]

Dear Ron:

I certainly agree that at least some of the applications were published one way or another, and that the level of "surprise" was probably close to non-existent.

Thanks for the comment.

[Kevin E. Noonan]

Dear 6:

Of the many, many things that are sad, I think your doubts about my faith in empirical evidence are what is truly sad. I don't need God to prove there is a tragedy of the anticommons - just some reliable data that proves the point.

The earth isn't flat, the sun doesn't revolve around the earth, and (so far) there is no evidence of any tragedy of the anticommons.

Thanks for the comment.

[breadcrumbs]

Understandably, this forum focuses on the biotech sector of the patent world.

Would the thoughts concerning tragedy of the anticommons expressed here be transferable to another patent sector?

Specifically, I am thinking of the sector where I have seen the most application of the anti-commons philosophy: the computer arts (and even more specifically, the computer software arts).

The Software sector seems to have the most virulent anti-patent movement, where rhetoric and unsubstantiated findings often replace rational consideration. I would be interested in hearing from the learned Drs. Katznelson and Noonan.

[Sean Brennan]

Another tragedy, in my experience, is that the media appear more interested in perpetuating speculative "Chicken Little" scenarios that in presenting facts and opinions contrary to the "anticommons" dogma. I recently submitted the following letter to the editors of Science magazine, who declined to publish it. A similar letter to the San Francisco Chronicle, was similarly declined.

To: Editors, Science magazine
The news article “Lawsuit Challenges Legal Basis for Patenting Human Genes” (Science 22 May 2009, pp. 1000-1001) quotes Hans Sauer of the Biotechnology Industry Organization as saying that there is “essentially ‘no evidence’ that patents have impeded research or medicine.”

The common misapprehension that patents deter progress in medicine and research appears to be based primarily on speculation, rather than evidence (1). However, not only is there no evidence to support the assertion that patents block research; evidence does exist that patents do not impede research. John P. Walsh, of the University of Illinois at Chicago, and his colleagues have reported, in these pages, that “. . . almost none of our respondents reported worthwhile projects being stopped because of issues of access to IP rights to research tools.”(2) Walsh et al. have concluded that “[o]ur results offer little empirical basis for claims that restricted access to IP is currently impeding biomedical research . . . ”(3)

Unfortunately, blame for certain unpopular business practices appears to have been mistakenly directed at the United States’ patent system.

1. See, for example, Heller and Eisenberg (1998) Science 280:698-701.

2. Walsh et al. (2003) Science 299:1021. See also Walsh et al. (2003) “Effects or Research Tool Patents and Licensing on Biomedical Innovation.” In Patents in the Knowledge-Based Economy (W.M. Cohen & S.A. Merrill, eds.), The National Academies Press, Washington, D.C., pp.285-340.

3. Walsh et al. (2005) Science 309:2002-2003

[Kevin E. Noonan]

Dear Bread:

I hesitate to speak extensively on this, since it isn't my area. But it seems that there are at least the following differences in subject matter between the sectors.

On the biotech side, the idea is that genetic information is a commons, like a communal grazing field, and the tragedy is that permitting private ownership of the commons (i.e., converting it into an "anticommons") will prevent full exploitation of it.

Software seems a little more like expressive speech, which permits the user to "play" a computer like an instrument. The open source movement would like all of this code to be open for future improvement, but it is hard to characterize what programmer A does as "commons" if it is his composition.

But there is a parallel in that programmer B usually builds on programmer A's code (I think), adding new applets and other things I don't understand to provide new functionalities.

Finally, I think the the reason there is no tragedy in the biotech commons is that patenting promotes disclosure, and then researchers go to a naturally-occurring cell or a database and create anew in a non-infringing way. Most of the dustups over the anticommons outside academia involves someone (like a university or hospital) wanting to infringe without liability.

Thanks for the comment. Maybe Ron will be able to add something more pertinent to the discussion.

[Ron Katznelson]

Breadcrumbs,

My comments with respect to software patents are the same comments I would make with respect other arts: many assertions and allegations of patents suppressing downstream development are accompanied with absolutely no evidence to back them up.

A classical example is Jaffe & Lerner’s celebrated 2004 book on the “broken patent system”, in which they discuss on page 201 the views of software patent opponents. J&L make “the straightforward observation that software innovation was flourishing before the 1980’s, when the CAFC clarified and broadened the patentability of software. This seems to show that patents for software are not necessary.” They go on to say: “More fundamentally, it is argued that software development is, by its nature, so cumulative that it is impossible to parse out the contribution of one developer sufficiently to grant patent rights, and it is counterproductive to try to do so because subsequent development will be hampered.”

These statements are utterly nonsensical, with no empirical evidence to back them up. First, J&L did not even suggest an objective measure to the degree software “flourishes.” Second, a more sober “straightforward observation” by one measure would show that software products actually started to “flourish” by TWO ORDERS OF MAGNITUDE in the mid 1980’s compared to the levels in the late 1970’s. J&L totally ignore the fact that the development of the 8-bit PC after 1980 was predominantly governed by software systems innovations – not hardware innovations. The microprocessor was on the scene a decade earlier. I have used Intel’s 8-bit microprocessor (8008) back in 1973. Of course, these PC software system developments might have taken place without the software patent protection boost that J&L refer to – we will never know. However, there is some evidence that patenting trends generally were correlated with the emergence of the PC. See my discussion of such evidence in pages 49-50 of my 2007 paper on patenting trends available at http://works.bepress.com/rkatznelson/3/ ). It is nevertheless clear that when the “anti-commoners” even refuse to accept the existence of evidence contrary to their “factual observations”, they are unlikely to engage in disciplined and scientifically-correct hypothesis testing.

Second, J&L’s purported ipse dixit pronouncement that “it is impossible to parse out the contribution of one developer” in software is nonsensical, supported by no rationale or evidence. It is also self-contradictory: how can “subsequent development” be “hampered”, if it is indistinguishable from a prior development? Any such new indistinguishable development must therefore be cabined within the prior art and as such would be non-infringing and thus “unhampered.” J&L and anybody holding such views has the burden of showing THE SPECIFIC attributes of software design and embodiments, as opposed to other engineering arts, that renders a determination of the metes and bounds of a software invention indistinguishable from the prior art. If this were true, it would have been impossible to draft claims for software inventions. What J&L are in fact saying is that all software patents are invalid. Just because they wish it, does not make it so.

Ron Katznelson

[breadcrumbs]

Thank you Ron and Kevin.

[breadcrumbs]

I am still digesting your responses, but thought I'd offer a second course from the ranks at Patently-O.

Ciaran appears to be a new actor on that site and has provided a link to a self-professed "end software patents" site:
http://en.swpat.org/wiki/Studies_on_economics_and_innovation

I offer this as a complement to Ron's post and wonder if this site has any information that Ron hasn't seen or anything that would distinguish itself from Ron's comments.

I appreciate your replies.

[breadcrumbs]

My previous post (not yet up) had a link to a Stop Software Patents type of website. I posted the link prior to reviewing the contents. I offer my apologies. The site is a virulent anti-patent site.

The agenda is so pronounced that I hold little hope that a balanced and thoughtful discussion would come from anyone associated with that site. It does serve as an example of what I noted in my August 27, 2009 09:34 AM post.

[Ron Katznelson]

Dear Breadcrumbs:

Despite the “Stop Software Patents” site’s anti-software patent stance, it cites a paper by respectable scholars such as Bessen and Hunt. Therefore, their quote deserves a substantive critical answer.

The web site quotes B&H’s 2004 paper “An Empirical Look at Software Patents”. The empirical data they collect is of considerable interest but, as explained below, not their conclusions:

“The very large increase in software patent propensity over time is not adequately explained by changes in R&D investments, employment of computer programmers, or productivity growth. [...] We find evidence that software patents substitute for R&D at the firm level; they are associated with lower R&D intensity.”

The web site also attributes the following quote to B&H’s paper’s conclusion:
“For industries like software or computer, there is actually good reason to believe that imitation becomes a spur to innovation, while strong patents become an impediment." This nonsensical baseless conclusion is nowhere to be found in B&H’s paper as posted on SSRN. It might have been there in a previous draft version of the paper. However, B&H conclude with an equally unsupportable conclusion on R&D and innovation: “Our results are difficult to reconcile with the traditional incentive theory—that granting more patents will increase R&D investments. Rather, if legal changes have encouraged strategic patenting, the result might well be less innovation.” It is unclear by whose “tradition” they presume their “patent incentive theory” that leads them to conclude that if patenting intensity is somehow negatively correlated with ALL R&D expenditures, it must necessarily mean that innovation is suppressed. Their leap is breathtaking and spectacularly flawed.

This is one of many papers written by economists who compare the number of issued patents to economic measures such as TOTAL R&D spending, GDP or employment levels. For some reason, none appear to compare the number of patents to the number of new products, for example. In many such studies, the authors’ inability to adequately explain the correlation between the growth in patent numbers and that of other economic metrics lead them to highly questionable inferences.

One fundamental false premise that such scholars appear to hold is that the economic value or SCOPE of the patenting unit that they measure (an issued patent), is an INVARIANT that should be meaningfully accounted for by the mere COUNT of the patents. They treat this number as one would count the number of real dollars spent on R&D, GDP, or the number of employees. However, one must question what makes the NUMBER of issued patents a definitive variable of any import for comparison with the chosen economic indicators? What is the economic and legal meaning of this patenting unit? Does it have a common meaning across patentees like the value of a real dollar? More importantly, has that meaning or value changed over time? I have yet to see any of these scholars seriously attempt to address these simple questions, let alone answer them. It is clear that without addressing the fundamental “invention currency unit”, any findings of a relationship between the number of issued patents and other economic variables are meaningless and can lead to nonsensical conclusions.

Suppose, for example, that using their “empirical” methods to track growth trends, B&H had used the number of scientific and technical papers published or the number of journal titles published over the years, instead of the number of patents issued. This measure is arguably just as relevant, as it relates to the R&D intensity, R&D employment and discovery of new knowledge. By analyzing such data they would have found and pronounced that “the very large increase in paper publication propensity over time is not adequately explained by changes in R&D investments, employment of scientists and engineers, or productivity growth” - using the words of B&H above. This is because the number of published journal titles grows exponentially, corresponding to doubling approximately every decade. [J. Meadows, The Growth Of Journal Literature: A Historical Perspective, In: B. Cronin and H. Barsky-Atkins, (eds.), The Web of Knowledge: A Festschrift in Honor of Eugene Garfield, pp. 87-107, American Society for Information Science & Technology Monograph Series, Medford, NJ, (September, 2000) (See Table 5.7)].

With these findings on journal publications, any reasonable scholar would naturally question the quantitative meaning of a published paper per se, as a research and development “output unit” or a “knowledge unit” to be compared with other economic indicators. Given certain fluctuations and historical declines in real R&D spending, I doubt that reasonable scholars would come to the conclusion that the relative increase in publication propensity means that there exists an effective SUBSTITUTION of R&D activity by journal publication activity. Rather, they would probably recognize that a journal paper publication unit is NOT an invariant over time and that the ever-expanding fields of knowledge and technology and the narrowing specialization must be at play. In other words, reasonable scholars would conclude that the possibility of a shrinking “publication currency” must be studied. Why don’t Economists come to similar simple conclusions when they count patents? I believe it is because they have already arrived at their conclusions on patents prior to their studies, which they construct in order to support their preconceived conclusions.

Indeed, there is substantial empirical evidence suggesting that over the last 35 years, patent claims have been issued with progressively diminishing scope. I discuss this evidence in Section 4.3 on “The Patent Scope Erosion” starting on page 35 of my 2007 patenting trend paper linked above (Figure 6 and accompanying text). Could this phenomenon be a much more plausible explanation for the increase in the number of issued patents? Is it not simply a result of patentees’ efforts to appropriate equivalent returns from their inventions and secure patent protection for rapidly changing new product offerings?

Other problematic areas in B&H’s framework are some of their underlying premises. On closer inspection of B&H’s empirical framework, one learns that they would have found that patents “substituted” R&D not just in the software patents area, but also in any new emerging-growth area undergoing relative increases in patent filings. Although some elements of B&H’s conclusions may be supportable on other grounds, the empirical framework they analyze is simply not equipped to provide support for their “finding” of patents’ “substitution” of R&D. It is remarkable that in one place they admit it and yet in another they contradict themselves. At page 38 they say that their “empirical technique cannot identify the CAUSAL relationships” for observed correlations between patent numbers and R&D expenditures but they hasten to declare in their abstract that “[w]e find evidence that software patents SUBSTITUTE for R&D at the firm level”. Can any reasonable reading of their conclusion on “substitution” be read in any way other than a CAUSAL finding? If they are not concluding a CAUSE, how can they infer that one SUBSTITUTED the other?

B&H’s mathematical model is rendered moot by the underlying flawed premise they hold. The main problem here is B&H’s underlying premise on the workings of patents, which appears to have been prevalent in many of Bessen’s other writings. It is indicative of a profound divorcement from the real patent world. They explain their premise as follows:

“[T]he patent incentive theory is akin to saying that patents and R&D are complements. Greater use of patents should be associated with greater use of R&D. And if patents become more cost effective—that is, if their quality-adjusted price falls—then the share of resources allocated to R&D relative to other factors of production should increase.”

This is counterfactual and nonsensical on multiple grounds. First, where in the world has anybody established that the patent incentive means that the “greater use of patents should be associated with greater use of R&D”? I am aware of no basis for this premise, certainly not in “patent incentive theory”. Even on its face, it fails spectacularly as many inventions come about because of the lack of R&D resources and not because of their abundance. While it is true that patents may channel SOME R&D spending towards patentable inventions because a strong patent reduces the market risks for such investments, firms have other means for reducing their investment risks by investing in established and entrenched product lines. But B&H had not distinguished such R&D investments from those in patentable inventions. They lump ALL R&D expenditures together. The bulk of R&D spending comes from large corporations and market incumbents who spend mostly on product developments, improvements and selective adaptation of relatively mature technologies. Patenting intensity is more closely related to introduction of new products (whether or not they prove ultimately successful). Small businesses with smaller R&D budgets have a much larger share in this area. Therefore, it is not surprising that small business patenting entities produce 13-14 more patents per employee compared to large patenting firms. [U.S. Small Business Administration, “Small Serial Innovators: The Small Firm Contribution To Technical Change,” at http://www.sba.gov/advo/research/rs225tot.pdf, at12]. Can B&H explain this ratio by their “substitution” theory? Do small business patents substitute non-existent R&D budgets? Clearly, B&H’s conclusions on patenting intensity and R&D lack a basis in fact and are questionable even at the large firm level. For example, B&H’s findings on IBM’s high patenting intensity may have very little to do with R&D spending but more with factors that B&H had apparently ignored. The only patenting “substitution” that I can think of as taking place in IBM’s case is that of patents substituting for technical journal publications. IBM had always used defensive disclosure strategies in conducting its business. For years it published much subject matter in the "IBM Technical Disclosure Bulletin". It ceased publishing the "Bulletin" in 1998 and has since used the PTO as its publisher of defensive disclosures.

Second, the “patent incentive theory” B&H speak of is undefined and it is unclear whether it takes into account real-world patenting considerations. From the above discussion, it is evident that B&H failed to take into account real world “patent incentives.”

I apologize that my response turned out so long…. To sum up, as in other similar studies that are premised on a misunderstanding of the patent world, some of the conclusions of these author’s may be supportable on grounds other than that which they provide, and other conclusions remain manifestly unsupported and even outright illogical. It certainly fails to support any thesis that software patents suppress innovation. Rather, as the recent patent infringement judgment against Microsoft shows, software patents suppress infringement.

Ron Katznelson

[Locke]

Here's an interesting thought: The opponents of so-called "gene patents" try to occupy the moral high ground viz-a-viz the anti-commons by pretending to support the free and open exchange of scientific information. See, e.g., a quote by the American Medical Association on the ACLU’s blog supporting its attack on gene patents (http://blog.aclu.org/2009/08/28/how-the-patents-on-the-breast-cancer-genes-harm-physicians-and-patients/). Specifically, the AMA says “[t]he use of patents . . . to limit the availability of medical procedures places significant limitation on the dissemination of medical knowledge, and is therefore unethical.”

Never mind the fact the Journal of the American Medical Association (JAMA) is NOT open access and most articles must be paid for. I fail to see how making doctors and researchers pay $30 for access to medical knowledge is any less significant a limitation on the dissemination of medical knowledge or any less “unethical” than the $50-$60 the average patient pays out-of-pocket for getting her BRCA genes tested. Hypocrisy; pure and simple.

[Locke]

Then, of course, there's the fact scholarly articles mentioning the BRCA genes on PubMed number over 7800, more than other important genes including CFTR, pRb, and TP53. I realize this is anecdotal, but it is telling that these are precisely the genes whose patents are being attacked on the basis of supposed suppression of research and patient access.

[Ciaran O'Riordan]

@Ron Katznelson

Thanks for your observations about http://en.swpat.org and it's page of studies. I've made a note on the wiki article's Talk: page to say that your comments should be read and that that Bessen & Hunt article should be reviewed.

@breadcrumbs:

swpat.org isn't anti-patent, only anti-*software*-patent. What the car and bio industries do to each other, and discussions of when it's good or bad for the public, is of no interest to the site. However, patents that block software compatibility and block individuals from writing software: those shouldn't exist. Information about the patent system - including general problems - is gathered to give context only.