Bayh-Dole has dropped commercialization rates from 25% to 0.5%: what more can we expect?

University licensing programs appear to have about a 0.5% commercialization rate. That is, of all the assets reported to them which they claim, only 1 in 200 (or less) actually results in a commercial product (without regard to the “success” of that product). This commercialization rate appears to be consistent across the last forty to fifty years of effort, despite changing laws and policies, a huge increase in budget, and a specific focus on “commercialization.” One might say that in terms of commercialization rates, Bayh-Dole has done nothing at all. One might also say that in light of all the other consequences of Bayh-Dole, it has failed as innovation policy, unless the purpose of such a policy is to create new opportunities for patent administrators and patent attorneys. Then Bayh-Dole has been a grand success.

In the late 1960s, Harbridge House, Inc. did an extensive study of federal agency patent practices. It’s well worth the read. One of the issues that the study considered was commercialization rates for contractors with various backgrounds. Here’s the money table, so to speak:


Here, a contractor is any contractor–commercial or otherwise. The commercial use rates for inventions in which the contractor holds title and has prior experience are near 25%. Compare that with the current university rates at best around 0.5%–companies with expertise and ownership were doing 50x better than present university practice.

Look in particular, then, at what happens when the contractor has no prior experience. With title, the rate is 6.6%. Without title, 2.2%. And with prior experience but without title, 13.3%–about half the rate of a contractor who holds title.

One might see that the Kennedy patent policy–which under section 1(b) allowed contractors with capability and established commercial positions to hold title in federally supported inventions–was pretty astute. Those contractors had a 1 in 4 chance of using what they had invented.

In 1978, Thomas Jones, then the vice president for research at MIT, testified at a Senate hearing that studies by HEW and Northwestern University had found licensing rates for inventions made with HEW support were about 24%. Jones cites a SUPA (now AUTM) survey from 1977 that claims 50% of all patents managed by universities over the prior decade had been licensed. Howard Bremer, in summarizing “50 Years of Technology Transfer” at WARF, gives the following figures:


The selectivity rate can’t be rightly calculated, because we can’t assume that there’s only one patent application per disclosure. But the upper end for the selectivity rate would be 24%. The commercialization rate–inventions producing royalty income–is 2.5% of the reported inventions, and 16% of issued patents (though even here, the number is merely an estimate, since there’s no indication whether the licensed inventions are the same as licensed patents, and whether some inventions with pending patent applications have also been licensed). It’s also not clear whether an invention “producing” a “royalty” includes any up-front licensing fees or whether those royalties are only ones earned on sales of commercial product. Obviously, anything but royalties arising from sales can’t be counted as a sign of commercialization. Of course, commercial use might also generate a royalty (if the licensee is silly and allows that sort of invasive monitoring of its activity). But the overall argument is that over fifty years, WARF has licensed on the order of one fifth of the patents it has obtained. In line with the figures given by MIT’s Jones.

Notice, however, WARF’s IPA record. Bremer reports 0 inventions obtained under the Wisconsin IPA have produced a royalty. The IPA rate of commercialization at 7 years is 0%. Over 10 years, the SUPA rate reported by Bremer is 50%. Clearly, the IPA approach for WARF is not functioning at anywhere close to the same level as university licensing generally or at WARF in particular. Is that a function of the conditions placed on IPA licenses? Or does it have to do with who does the selecting of inventions? In the general case, for many universities, inventors chose whether to patent, while under the IPA, university administrators apparently decide. Perhaps who chooses an invention for patenting work–and where that patenting work will be done–has something to say about the future likelihood that an invention will be “commercialized.” In the same way, Picasso and I may both have used cars for sale. Amazingly, (let’s say) Picasso drove the same model 1974 Dodge Colt that I did. Which car will go for the better price? Sooner? Oh, you sold your invention to a university administrator? Hmmm.

Now consider what Bayh-Dole does. It allows contractors with no established commercial position–universities–and no “prior experience” in the area of commercial application to acquire title to inventions. We are at the 6.6% commercialization rate immediately. But there’s more. The university then expects to license, not assign, the invention to a for-profit business. That would be the equivalent of a contractor without title–at best, 13.3% (of 6.6%!–or 0.9%). The simple requirement in Bayh-Dole that forbids assignment (unless the agency approves) to a company (other than one set up to manage inventions) results in an expected empirical reduction in commercialization rate of 50% (using Harbridge House figures).

There are qualifications to make on all of this, of course. The federal licensing practice was predominately non-exclusive. We don’t know what the rate of exclusive licensing might be like. Would it be comparable to that of contractors that owned their inventions? Not likely. Would it be better than situations in which contractors obtained just non-exclusive licenses? I don’t know the answer. It also may be that lots of federal non-exclusive licensing (and dedication of inventions to the public) may have made the exclusive licensing opportunities even more effective when they were offered–less in the way of fragmented patent ownership, less gridlock of competing claims, less involvement by nonprofit patent administrators each wanting a cut of the action in the form of royalties or threatening litigation for infringement).

But consider this variation. The university licenses its patent exclusively to a startup company–that is, to a company that has neither prior experience nor title. We are at the lowest rate of commercialization–2.2% (or 2.2% of 6.6% or 0.1%). Except that current university practice is now at 0.5% or worse (but look on the bright side–it’s 5x better than the Harbride House figures might indicate!). For whatever else a university might profit from startup licensing–cash for licensing fees, equity in the startup, a funnel of research dollars from venture investment–the university operates at 4x worse rates than the worst rates under the Kennedy patent policy.

Of all the things a university might do with patents on federally supported inventions, licensing them to startups appears to be the worst possible outcome in terms of commercialization rates. Of course, if one is looking only at total financial return over the portfolio, all one needs is one startup to “go big time” and the money outcome will appear to be successful for decades. Stanford licensed the page rank algorithm to Google, took an equity position, sold that equity position when Google went public, acquired a new equity position at the same time, and has done very, very, very well. But that financial success does not mean that Stanford’s overall commercialization rate, invention by invention, is high. Stanford reports that it doesn’t license 80% of what it claims. So, a 20% licensing rate. That’s impressive. But the commercialization rate (taking $100K of income as a threshhold) is much lower–0.45%.

It would be wonderful to see a new report on the model of Harbridge House. What are the subject invention claim rates, licensing rates, and commercialization rates for universities? for small businesses? for large businesses? That would be a way to see the effect of changes in policy, in markets, in the focus of resources. Teasing out what part Bayh-Dole has played would take some attention–but at least we would know better where we stand. As it is, it would appear that the federal government would do substantially better in terms of commercialization rates by requiring universities to assign claimed inventions to for-profit concerns than to license the inventions “exclusively.” Further, the federal government would do better by licensing inventions non-exclusively to companies that have prior experience than by allowing universities to license inventions to their own not-arm’s-length startups.

If universities were held accountable for commercialization rates, universities would be much more selective of what they claimed to own. Their claim rates would go from 70% or more to 10% or less. The volume of work would drop dramatically. They would focus on what they had selected, not on “processing” everything and billing it out as “potential.” They also would lose the incentive to license to their own startups. They would have to seek out companies with established experience and try their best to get those companies to adopt. Money and legalistic fuss would have to take a back seat–yeah, there will still be fussiness, but fussiness wouldn’t be doing the driving. They would spread their risk and license whenever they could to multiple companies–let them race to the market, let them share development resources if they want (a commons! standards! interoperability! pelaton competition!). License exclusively only where someone actually commits to significant commercial development and even then also license for internal use wherever–make it a race between internal commercial use and commercial product development. It’s the commercial use rate that matters, not whether that use depends on purchase of a product.

The money position for a patent owner might be in a royalty on commercial sales. Or it might be–for particularly useless but grand-sounding inventions–in speculation on the future value of a patent (as often happens with startups based on a patent license). But if the condition on which a university patent program has access to title depends on the commercialization rate, then everything changes. Patents are acquired for use, not as a threat against use. Applications are public from the start. Industry is consulted before applications are even filed. “If we file, will you work with us to learn how to use this invention?” Research use and internal use are open. Sales rights would be non-exclusive except in certain industries for certain uses–pharma, say, where nothing ever changes.

How would such a change come about? One way would be, for nonprofits, to make Bayh-Dole’s title rules conditional on the nonprofit’s commercialization rate. To implement, just add before the disposition of principal rights:

Each nonprofit organization, if it has an established practical application rate greater than the minimum determined by the Secretary of Commerce to reflect the public interest, but in no event less than 5%, or small business firm may, within a reasonable time after disclosure as required by paragraph (c)(1) of this section, elect to retain title to any subject invention: Provided, however, . . . .

Where “practical application” is a defined term in the law and serves as a useful proxy for “commercialization.” There, that would do it. Nonprofits that can’t show a decent rate would not have the right to acquire any more patents on subject inventions. They would have to clean out their portfolios and adopt something closer to an agent model of management. They would also adopt licensing for use ahead of most licensing for commercial product development, and they would prefer licensing non-exclusively for commercial product development than granting a one-time monopoly license. When they license non-exclusively, they might give early adopters an advantage, such as royalty-free rights when there’s still risk and which lock-in the moment there’s practical application (as defined by Bayh-Dole). Anyone coming later might pay a reasonable royalty, reflecting the reduction in risk brought about by those companies already conditioning the market and suppliers to the new product. And when nonprofits license exclusively, they might prefer to offer shorter exclusive terms, and might limit the field of use so that there might be others developing products for other markets. When they license exclusively, they may also be more ready to terminate exclusivity if a company stops developing an invention–and certainly terminate the license for those parts of a licensed invention that an exclusive licensee shows no sign of developing (for example, in an exclusive license for a class of compounds of which the licensee develops only one of hundreds of compounds–why should the use of the rest be suppressed in favor of just one?). Anything to keep that practical application rate above the minimum.

As the new programs for licensing develop at universities, the Secretary of Commerce might then raise the minimum threshold to ensure that universities keep up with best practices. After all, it’s still a long way from 5% to the 25% rate that company contractors were reporting in the 1960s. In fifty odd (very odd) years, federal policy led by patent attorneys has managed to increase the volume of patents acquired and in doing so lowered the commercialization rate by two orders of magnitude. Isn’t it time to reverse the direction and push for rates to move up? And then to deal with all the inventions that ought not be handled by university patent brokers and ought not be handled by patents, or if with patents, then not with a profit motive? Then we’d be making some progress on both university formal patent licensing programs and the informal invention and NIPIA (non-IP intangible asset) transfer programs.

This entry was posted in Bayh-Dole, Commons, History, Metrics, Policy and tagged , , , , . Bookmark the permalink.