Evidence-based federal research patent policy

The Commission on Evidence-Based Policymaking is holding a public hearing in Chicago in January for “any interested stakeholders” to provide input. Given that the commission’s statutory mandate is more toward database access and security, I’m not sure that the lack of data pertaining to Bayh-Dole is all that close to their interests. Here’s what I’d present to the Commission. Maybe I’ll submit something for the West Coast meeting.


The federal government provides $40 billion per year to fund research at American universities. The Bayh-Dole Act, a part of federal patent law, governs the disposition of patentable inventions made in the course of this research. Such inventions, defined as “subject inventions” by Bayh-Dole, represent a significant portion of the research output of federally sponsored university research. It is important that the federal government has the ability to assess the effects of the Bayh-Dole Act.

Patents grant their owners for 20 years from the date of application the right to exclude all others from the practice of a patented invention. Federal patent law does not require that the owner of a patent practice the  patented invention. Nor is the patent owner required to allow any others to do so.

Bayh-Dole Act authorizes the creation of standard patent rights clauses that federal agencies must use in funding agreements for research conducted at universities and by small businesses. The primary standard patent rights clause permits small businesses and nonprofits to retain title to subject inventions assigned to them by inventors, subject to various conditions.

The Bayh-Dole Act directs that patents on subject inventions be used to promote the use of those inventions. The standard set forth is that of “practical application”—use with public benefits on reasonable terms. In addition to use, patent rights on subject inventions are subject to a number of federal restrictions that do not apply to other patents, including restrictions pertaining to assignment, exclusive licenses, use of royalties, preference for small businesses, special notices in patent applications, periodic reporting, and possible federal agency intervention for nonuse or unreasonable use among other causes of action.

Although Bayh-Dole does provide that federal agencies may request information from the owners of patents on subject inventions (and their assignees and exclusive licensees), including the date of first commercial sale or use, there is no requirement that they do so. Furthermore, the Bayh-Dole Act exempts from the Freedom of Information Act all information received by federal agencies in response to such requests.

There is, then, no publicly available data on the fundamental statutory goal of the Bayh-Dole Act. In turn, we have no direct way of assessing the benefits of Bayh-Dole in relationship to the costs of compliance and potentially adverse effects on scientific research and innovation. Nor can we readily determine whether federal agencies are complying with Bayh-Dole or enforcing the terms of the standard patent rights clause authorized by Bayh-Dole and administered by NIST under delegation from the Department of Commerce.

A number of authorities assert that the Bayh-Dole Act has been an impressive success. However, there are no reliable data to establish the truth of the claim. The Association of University Technology Managers (AUTM), a professional organization of university licensing officers, publishes an annual survey of licensing activity, including the number of inventions reported, patent applications filed, patents issued, license agreements signed, patenting expenditures, and start up companies formed. The AUTM licensing survey does not break out subject inventions for other inventions and does not report the fundamental objective of Bayh-Dole—that practical application has been established for any given subject invention, nor the date on which such practical application has been achieved. These two measures are essential to any assessment of the success of the Bayh-Dole Act.

The AUTM survey suffers from other defects as well—data are self-reported and not subject to audit. The data cannot be established as  uniform , accurate, or verifiable. Despite attempts at definitions and guidance from AUTM, university officials are left to decide, for instance, what constitutes a “commercial license” or counts as a “start up” formed to receive a license to an invention. Since each university reports independently and reports only activity totals, inventions owned jointly by two or more universities (a routine occurrence) are counted multiple times in aggregate totals, inflating measures of activity such as the number of patent applications filed, patents issued, and licenses granted.

The absence of activity data and outcome specific to subject inventions and the objectives of Bayh-Dole mean that federal patent policy governing research and innovation activities has operated without primary evidence from which to reason. The scattered instances in which data on invention commercialization have been made available by universities points to a rate of practical application below 1% of inventions claimed. The University of California recently estimated in an internal document a rate of 0.5%. Stanford University, in assessing 36 years of invention management, has presented figures that suggest a rate of 0.8%. These licensing programs are regarded among the best in the country.

By comparison, universities and their affiliated research foundations reported commercialization rates of 25% to 33% for inventions not made with federal support. Under the Institutional Patent Agreement program operated by HEW from 1968 to 1978—a forerunner to Bayh-Dole that allowed federally supported inventions to be patented and managed by universities and their designated licensing agents—universities reported a commercialization rate of just over 4%. The 4% rate was comparable to the rate reported at the time for inventions owned and managed directly by federal agencies, leading some observers to wonder if university patent licensing practice offered any meaningful advantage over management of subject inventions by the federal government.

In presentations by advocates for Bayh-Dole, however, the federal government was and has been continued to be criticized for its past rate of commercialization—that such low rates of commercialization hold back valuable inventions, block technological innovation, and prevent economic prosperity. Yet it appears that Bayh-Dole, hailed as successful, now produces commercialization rates that are 1/10th those of the federal government or under the IPA program. The lack of federal data to contribute to public policy discussions of the place of patent management in the conduct of science and innovation leads to the mistaken impression that robust activity measures must be evidence of robust commercialization rates. However, if practical application is being achieved no better than 1 in every 100 or 200 subject inventions, then robust activity measures indicate inefficiency and potential hazard.

Consider, for instance exclusive patent licenses. An exclusive license conveys rights from the patent owner to another entity. In the case of university licensing, typically the licensee is a company, and often a startup company. On the face of it, such activity points to commercialization. However, the fact of a license does not ensure practical application. Nor does investment in the company (which may be used to develop a product not based on the invention licensed from the university). Nor does licensing income received by the university licensor (which reflects the value of the patent license but does not mean that a product has been successfully made and offered for sale).

Exclusive licenses granted to startup companies that fail to achieve practical application are among the most adverse outcomes possible. Not only is the patent unavailable to others for use, but investors in the start-up may use the patent to attack others in industry. The president of the Licensing Executive Society, in a recent briefing in defense of Bayh-Dole for legislative officials, asserted that if a startup company fails, its investors can “monetize” the patent license obtained from the university. One “monetizes” such a license by demanding payments from others who may be practicing the invention without a license, or by selling the rights to demand such payments to a company that specializes in such business. Rather than promoting the use of the subject invention, a patent used in this way serves to suppress practical application and operates as a barrier to rapid innovation—the opposite of the central policy and objective of the Bayh-Dole Act.

Activity measures without outcomes are necessarily ambiguous and may easily mislead. Imagine reporting the number of hits and errors in a baseball game without reporting the runs scored. For Bayh-Dole, we have information on the number of the number of pitches thrown, the number of fly balls, and the total number of baseballs used. From this information we are to decide whether our team has won.

Where activity measures are presented by organizations financially interested in retaining a federal benefit—such as offered by Bayh-Dole—the absence of outcome measures is cause for concern. Self-reporting may be selective, may be in error, may be duplicative of reports made by others, and may be deliberately misleading, as has been documented in the case of two major universities in their efforts to obtain additional funding for their research and innovation programs from state legislatures. Without reporting with accountability and independent review, universities may easily slip from providing data useful for evidence-based policy making to data designed to recruit or mislead or distract policy making from its proper role. The result of such behaviors is the preservation of institutional privilege, not effective science and innovation policy.

It is essential that data used for policymaking has been validated. Sophisticated reasoning with regard to flawed data results in flawed recommendations having the appearance of quantitative evidence–worse, perhaps, than no recommendation at all.

Under the influence of the Bayh-Dole Act, university patent policies have shifted from agent-based models of invention management to portfolio models. In an agent-based model, an agent is responsible for successfully achieving practical application for each invention the agent takes under management. Under an agent model, an agent is highly selective, focuses attention on those inventions that are suited to the agent’s methods and industry connections, and returns any invention that does not timely show indications of successful licensing. If an agent fails to succeed a significant percentage of the time, the agent’s reputation suffers and it loses business or fails entirely. A sports agent will lose clients if the word gets out that the agent is rarely successful in negotiating a contract.

Under the agent approach in place before Bayh-Dole, universities reported commercialization rates of 25% to 33%, high enough to support an argument that agents were a better approach than the approach preferred by the federal government, which was to dedicate inventions to the public or to license patents non-exclusively and generally royalty free. In a portfolio-based model, by contrast, an organization accumulates and holds many inventions, anticipating that some few may become valuable. Since the organization is not selective, it does not have to determine which inventions might be value but rather it works to prevent any inventions from falling outside its claim of ownership.

The secret to financial success is to assume that any inventor who wished to avoid the institution’s claim of ownership must have a valuable invention, which therefore must also be claimed. Commercialization rates under a portfolio model are typically much lower than for the agent model, but a portfolio business model requires only one significant patent license every twenty years to appear financially successful, and a portfolio model can persist for a decade or more lacking a significant licensing event before being called into question. The likelihood of a multi-million dollar license arising from any one of hundreds if not thousands of retained patents appears to increase with an increase in the number of patents under management.

Such probabilities are statistical illusions. A greater number of patents under management means greater expense, greater licensing complexity,  less ability to focus, and less chance of achieving practical application as the result of a patent license. A high number of unworked, unlicensed, or licensed and failed patents is indicative of a portfolio that will make its money, if ever, by asserting patent rights against companies that have adopted the invention without the need for a patent license to justify the effort and expenditure.

It is essential that federal agencies obtain and provide for policy review outcome information specific to Bayh-Dole. For each subject invention claimed by a university, whether that subject invention has achieved practical application and if so, with what date of first commercial sale or use. These two pieces of information are necessary to any determination of the effectiveness of the Bayh-Dole Act. Other data would provide additional insight, but without these data, no assessment of the Bayh-Dole Act is based on reliable, meaningful evidence.

The United States has over the past 35 years entrusted hundreds of billions of dollars of research funding to American universities following a patent and innovation policy that it has never evaluated based on the evidence of outcomes consistent with the expectations of the law that has formalized that policy. Instead, policy makers have relied upon the impression created by activity measures that lack validation.

The commission should recommend that federal agencies be required to request data on the practical application of each subject invention claimed by a university and the date of first commercial sale or use, if any, for each. That information should be aggregated into a public database and published for analysis and policy discussions directed at the conduct of science, technology development, and innovation using the inventive results of federally funded research.


This entry was posted in Bayh-Dole, Metrics, Policy. Bookmark the permalink.