IP harmonization is bad for IP science

Intellectual Property Rights and Innovation: Evidence from Health Care Markets (summary) describes two investigations by the author, Heidi Williams, each attempting to overcome challenges to characterizing the impact of IP on R&D investment and on follow-on innovation. Namely, measuring R&D and innovation is difficult, as is finding variation in patent strength (referring to patent policy, eg duration and scope, not the quality of individual patents).

Williams’ investigations don’t quite produce evidence for or against patent strength abetting or hindering R&D or follow-on innovation (due to other unknown variables and using ad hoc contractual IP rather than patents for variation respectively; read the paper) but does open up some promising lines of inquiry and generates some interesting observations along the way. I’m going to comment and extrapolate on mostly the latter for the remainder of this post.

First, and taken as the post title (IP harmonization is bad for IP science), I can’t believe this hadn’t occurred to me before:

finding any variation (much less “clean” or quasi-experimental variation) in patent protection is difficult. There are few or no equivalents of clinical trials that have randomized patent protection across technological areas. On paper, the patent system is uniform, providing a twenty-year term for all inventions. Although historically (and, more recently, for developing countries) there exists some cross-country variation in patent protection, this type of cross-country variation seems likely to be under-powered to test how patents affect research investments given that many inventions are created for a global market. That is, a relatively small market like Austria extending its patent term from 17 years to 20 years may reasonably be expected to induce a relatively small change in global research incentives.

Given the paucity of direct evidence for whether patents ‘work’, global harmonization of patent strength is premature, to say the least. For the good of the science of IP (preferably taken as innovation policy, though of course I’d prefer ‘IP scholars’ to re-imagine themselves as commons scholars), variation should be encouraged! I suspect commons-favoring policy such as funder mandates for openness will turn out to be a rich source of variation in effective patent (and other intellectual property) strength for scholars who choose to tap it. Yet another reason to prioritize such policy in the reform agenda above tweaks that merely curb the worst abuses or slightly increase efficiency.

Second, Williams’ take on the purpose of IP is at once typically constrained, but pushing in a better direction:

The more effective are patents in inducing research investments, the stronger the case for longer or broader patents; on the other hand, the larger the costs of patents in terms of hindering follow-on innovation, the weaker is this case.

In other words, the only objective of innovation policy is more innovation. Other impacts, say on freedom, equality, and security, don’t count. But at least counting patents shouldn’t either:

Large literatures in economics, law, management, finance, and related fields have used patent-based outcome variables as measures of research investments. However, I argue that patent-based measures are usually not a useful outcome variable in studies where researchers wish to analyze how variation in patent protection affects research investments.

The concern that arises with this type of study is that such changes in patent policy change inventors’ incentives to file for patent protection on existing research investments. In the absence of some outside (non-patent based) measure of research investments, seeing that patent counts increased in response to TRIPS-induced patent term extensions is difficult to interpret.

Even better:

ideally we would continue to make progress towards measuring more welfare-relevant outcomes. That is, if a paper was able to document that longer patent terms induced more research investments, that would be very useful, but ideally we would like to trace out how those research investments in turn affect welfare-relevant outcomes such as prices and health outcomes.

Third, the investigation of R&D investment impact does seem to show an under-investment in early stage cancer interventions relative to investment in late stage interventions, attributed to greater commercialization lag for the former, as it takes longer to show efficacy (increased survival) when survival rates are higher. Williams produces one very rough estimate of a cost associated with this under-investment: $89 billion for cancer patients diagnosed in the U.S. in 2003 and urges changes which could reduce this lag such as finding surrogate measures for survival (symptom improvement) and using those for regulatory approval. Although Williams couldn’t draw direct conclusions concerning patent policy, and if she could have, presumably they would have been in the direction of increased duration for early stage cancer therapies, but the conclusion drawn for shorter time to commercialization is very helpful considering the very tight relationship of regulatory approval and patents and the need for greater regulatory flexibility to take advantage of precision medicine.

Finally on this point two speculations. One, if commercialization lag is a disincentive for R&D on early stage cancer interventions, could it not be an even worse disincentive for R&D on even longer term preventative interventions, at the extreme, aging therapies? Or perhaps cancer is atypical in defaulting to survival as a measure of efficacy. Two, does favoring late stage interventions play a role in increasing overall health care costs?

Fourth, the investigation of impact on follow-on innovation produces striking results: even a very brief (1 year) period of restriction results in substantially less follow-on knowledge and innovation. Again, this particular investigation can’t say anything about patent strength, as it looked at a contractual restriction on access to data. Indeed, a potential takeaway is that more patenting should be allowed (in this case of genes) in order to promote data availability over secrecy. However, patents are only one way to stimulate data sharing, and an indirect way with many bad effects. It seems to me this result should encourage construction of regimes that have data sharing built in, many of them broadly commons-favoring regimes.

Fifth, and related to the first point, Williams notes various measures of R&D and follow-on innovations that do not involve counting patents (or in one case takes invalidated patents as a source of variation). Highly relevant, subject of a forthcoming list/wiki topic.

I noticed this paper at http://marginalrevolution.com/marginalrevolution/2015/06/do-ip-rights-matter-for-health-care-markets.html

The author of this paper, Heidi Williams, was just awarded a MacArthur Fellowship commonly known as a “genius grant”.