Empirical IP Research Conference, NYU

Plenary Session: How Do Patents Affect Innovation?

Facilitator: Katherine Strandburg (NYU)

Longer history of empirical work in patent, but still difficulty answering basic question of effect on innovation. In 1958, economist Fritz Machlup famously concluded that “none of the empirical evidence at our disposal” “either confirms or confutes the belief that the patent system has promoted the progress of the technical arts and the productivity of the economy.”  Are we in a better position today?

Panelists: Heidi Williams (MIT Dep’t of Economics)

Topic: Measuring the Effects of Patents on Downstream Innovation, Commenting on: Alberto Galasso & Mark Schankerman, Patents and Cumulative Innovation: Causal Evidence from the Courts (2013)

How to patents on existing tech affect follow-on innovation?  Two key challenges have hindered prior work: how can we construct an appropriate counterfactual?  Can’t randomize patents for some tech and not others. Selection bias is a problem.  Insight: lever CAFC judge assignment for variation in active patents; this experiment holds disclosure constant.  Second challenge: measurement. What paper trail can measure follow-on innovation?  Patent citations can’t be used to track innovation on tech that doesn’t receive patents.  Using patent invalidations makes citations a feasible outcome.

Does invalidation change investment in follow-on innovation?  No estimated effect on pharma cases, but Williams suggests that awareness of such cases and their likely outcome means that invalidation proceedings’ likelihood are already incorporated into investment decisions.

Economists have looked carefully at citations: hybrid corn patents, where yield of corn provides some measure of patent value.  What are the citations measuring? Inventors and examiners cite a core set of patents that establish patentable subject matter (PSM) in an area—citations establish many things.  Case could be cited more/less because it’s being used to define PSM—so that might not provide a measure of follow-on innovation. 

Authors are aware of this, but note the difficulty of finding a measure of innovation—tried to collect non-citation measures of follow-on innovation: clinical trials for pharma—when a patent is invalidated, do you see more clinical trials on that active ingredient?  Medical device registrations: if invalidated, do you see more medical devices measured by FDA registrations?  Consistently w/citation results, they don’t see impact on clinical trials (though since these cases involve ANDAs there might not be new information revealed anyway); medical device data are more nuanced—new devices/generic versions of existing devices: when there’s an invalidation, you get more generic device registrations—substantially equivalent to product whose patent is invalidated.  Not sure that this measures follow-on innovation; these don’t require licenses and may just reflect changes in competition.

Excellent empirical work; leaves open questions.

Stefan Bechtold (ETH Zurich)

Topic: The Effects of Patents on Scientific Research, Commenting on: Bhaven Sampat & Heidi L. Williams, How Do Patents Affect Follow-On Innovation? Evidence from the Human Genome

Try to identify variable correlated w/ explanatory variable but doesn’t suffer from causality/endogeneity problems of explanatory variable.  Existing literature: IP rights without disclosure may reduce follow-on academic research/commercial development; with disclosure IP rights may reduce downstream innovation in computers, electronics, medical instruments, but not in drugs, chemicals, or mechanical technologies.  Paper: do patents in human genes affect follow on innovation as measured by scientific research and product development?

Look at patent applications, both successful and unsuccessful.  Gene sequences extracted from patent claims and linked to data on publication citation, clinical trials, and diagnostic tests. Endogeneity: genes claimed in patent applications may have different characteristics than genes are never claimed; inventors may file for more valuable technologies.  Look then at successful v. unsuccessful applicants, as well as examiner variance.  Random variation? In fact assignment isn’t random, but within art units it’s random; key question is whether successful and unsuccessful applications are similar at the time of application.  Use leniency of individual patent examiner.

In general, no effect of gene patents on follow-on innovation.  No decline or increase in follow-on publications and clinical trials.

What’s the relevance of the economist v. legal scholars?  Lawyers know institutional details so they can figure out what might be causal.

Michael Meurer (Boston University)

Topic: Non-Patent Measures of Innovation, Commenting on: Petra Moser, Patents and Innovation: Evidence from Economic History, 27 J. Econ. Persp. 23 (2013)

Douglas North: emphasized secure property rights, rule of law for role in innovation in Industrial Revolution. Skeptics: with exception of James Watt, few prominent innovators profited from patents or even got them.  Aggregate data on growth and availability of IP rights: methods of measuring patent strength are fairly crude, and there are problems of endogeneity: which came first—innovation or strong patent rights?

Harmonization has been effective today: Chinese patent system differs very little from American system in relevant ways—that gives an advantage to 19^th c patent data where there’s a large amount of credibly exogenous variation in patent law; mostly domestic patenting prior to Paris Convention.

Besides aggregate data, there are case studies.  Much work on collective invention/knowledge sharing in blast furnaces, high pressure steam engines, weaving machinery (learning by doing): ways to appropriate value by patents, trade secrets, or some form of open sharing—when it’s not necessary to get a patent to drive innovation and invention.  Sewing machine, for example, was recombinant technology drawing on lots of previous innovations spread out over time.  Research: patent pool seemed to decrease innovation; pool members seemed to innovate less. Evidence of maximum stitch rate: growing before pool, leveled out during pool, grew again after pool. No pool in England; controlling for variations across inventions, found significant dampening of R&D especially by US pool members. Previous economists looking at patent pools didn’t pay much attention to outsiders, but they played a big role—effective threats of patent litigation from pool pushed them to invent around. Is that normatively desirable?  Wasteful duplication or better innovation? Others have explored patent patents, compulsory licenses, periodic table.

Comparability: Economists can be informed about institutional details with case studies that are contemporaneous; harder when your case studies are 150 years old. Archival research is also necessary.

Petra Moser (Stanford Dep’t of Economics)

Topic: Measuring the Effects of Patents on Private Incentives, Commenting on: James Bessen et al., The Costs and Benefits of United States Patents (2014)

Estimating private costs and benefits of US patents owned by publicly held firms. Event study approach to estimated losses suffered by alleged infringers 1984-2009; authors use market value regressions to estimate value of patent rents for publicly traded US firms. Findings: Total private costs exceed private benefits; excess costs increase over time surege in lawsuits (NPE, computers/communications patents, software and telecom); growth in private costs outstrips growth in lawsuits (lawsuits cost increases over time).

Lower bound estimates of costs according to paper: but that’s not precise enough for Moser—study measures effect of threat of filing law suit, but what happens when a firm wins or loses a lawsuit or unexpectedly has to defend?  Innovation incentives: even if costs outweigh rents, if rents are given to those who don’t bear the costs, then innovation incentives could still be huge.  Suggests looking at individual cases rather than aggregate evidence.

Differences across countries: use EPO patents to look at how differences in patent systems influence private costs and benefits. Considering US patent only: upper bound rent of $517,000 per patent 1979-2002, adding EPO patents gets you to $351,000 per patent.