Tuesday, October 30, 2018

Measuring the Role of Attorney Quality in Patenting

It stands to reason that better attorneys are better at turning patent applications into patents. Theoretically, better arguments about overcoming prior art, for example, will be more likely to lead to granted claims. But what about the quality of inventions? Maybe better patent attorneys just get better patent applications, so of course they have better success rates.

Measuring this is hard, but Gaétan de Rassenfosse (Ecole Polytechnique Fédérale de Lausanne) and four co-authors from University of Melbourne and Swinburne University of Technology think they have found the answer. Examining 1.2 million granted and refused patent applications in the US, Europe (EPO), China, Japan, and South Korea, they think they have the answer. They have posted a draft on SSRN, and the abstract is here:
Failure to obtain a patent weakens the market position and production chain of enterprises in patent-intensive technology domains. For such enterprises, finding ways to maximise the chance to obtain patent protection is a business imperative. Using information from patent applications filed in at least two of the five largest patent offices in the world between 2000 and 2006, we find that the ability to obtain patent protection depends not only on the quality of the invention but also on the quality of the patent attorney. In some cases, the latter is surprisingly more important than the former. We also find that having a high-quality patent attorney increases the chance of getting a patent in less codified technology areas such as software and ICT.
They use a clever approach with their multi-country methodology to separate attorney quality from invention quality. By estimating grant rates across countries for the same inventions as well as for the same attorneys, they are able to estimate the marginal value added by the invention versus the attorney. For example, if different attorneys have differing results in two countries with the same invention, then attorney quality is likely at play. However, if the same attorney has differing results in two countries with the same invention, then invention quality is more likely at issue. Of course, this doesn't work with a couple of patents, but with more than one million patent applications, the estimates likely trend toward a reasonable measure of each type of quality unaffected by the other.

Using this index of quality, they then estimate the effects of attorney and invention quality. They find that, as expected, invention quality matters. But they also find that attorney quality matters--sometimes a lot. More interesting, they find that attorney quality matters more where there is more wiggle room (my term, not theirs), such as in software (as opposed to chemistry). In other words, they find empirical evidence to support the intuition that attorneys who can mold claims to avoid rejections are more likely to wind up with issued patents.

Now, the benefit isn't unbounded. Because grant rates are really high (like 85%), the marginal benefit is unlikely to be huge. One marginal estimate they make is that moving from the 10th percentile to the 90th percentile in quality increases the grant probability by 13 percentage points (e.g, from 79% to 92%). What this means is that invention quality is very important, because even the least successful attorneys are successful most of the time. Of course, this is a potential criticism, because if everything is granted, then how can we measure quality based on grant rates? The paper addresses this, arguing that there is a lot of variability, even within the same patent family. They also test on other measures of patent quality and find similar results.

The authors also make other interesting findings: external counsel increase grant rates, attorney quality doesn't seem to have an effect on foreign v. local probabilities, and attorney quality is of less importance in PCT applications. There is a lot more to the paper - including effects in different offices, on different technologies and other great information. I found this to be a really interesting and useful read.

Tuesday, October 23, 2018

Patents and the Administrative State

When Justice Gorsuch was confirmed to the Supreme Court, many commentators, well, commented that he was wary of administrative overreach. But it turns out he was really active in patent cases, writing opinions in all the patent cases he saw last term. Who knew he was so interested in patents? He does have some IP chops; his opinion in Meshworks remains one of my favorite copyright cases, not the least of which because it validates a legal argument I made about virtual reality copyright some 25 years ago. I was able to cite that case in a recent book chapter on the same subject.

But it turns out that his interest in patents may be one and the same as his concern about the administrative state. We suspected as much with Oil States, but what about the others? To answer this, Daniel Kim and Jonathan Stroud (both of Unified Patents) have an article forthcoming in the Chicago-Kent Journal of Intellectual Property called Administrative Oversight: Justice Gorsuch’s Patent Opinions, the PTAB, and Antagonism Toward the Administrative State. A draft is posted on SSRN, and the abstract is here:
In his first term, Justice Neil Gorsuch has made a surprisingly forceful impact on, of all things, patent law—and even more unlikely, the United States Patent and Trademark Office’s adjudicatory arm, the Patent Trial and Appeal Board. Was there any way to predict, from his 10th Circuit opinions below, that he would author opinions in all three patent cases in his first term? Was this attention the result of deeply submerged but long-felt opinions on patent law, or rather a result of his sharp distrust of administrative overreach? We analyze 10th Circuit and Supreme Court opinions authored by Justice Gorsuch, and conclude his unforeseen interest springs from his desire to limit agency power rather than from any particular concern with patents. Still, his opinions—intentionally or by happenstance—will reverberate through our patent law for years.
This article is straightforward and illuminating. It begins with the justice's background and some comments on his writing style. It then examines his Tenth Circuit opinions in IP and tribal immunity (there are really not that many, which the authors attribute to the backwater location of the 10th Circuit devoid of any innovation, which I'm sure the folks in Denver will be happy to hear).

What's interesting about the article is that its analysis of the cases is not about IP outcomes, but about methods and the administrative state. Returning to Meshwerks, for example, the authors focus primarily on how it traces the history and purpose of IP rather than the important holding that realistic renditions of physical objects lack creativity.

This analysis bears fruit, though, because they show how these same methods and concerns about the administrative state drive Gorsuch's patent opinions, including those where he is dissenting or breaking from other conservative justices.

I found this article an interesting and insightful read, and I thought it was especially well done given the authors' own admission that they do not agree with Justice Gorsuch's judicial perspective. In other words, this wasn't a fan piece, but instead really useful commentary about how judicial philosophy and concerns about the administrative state may be brought to bear on the IP system in unexpected ways.

Friday, October 19, 2018

What impact do government grants have on small tech firms?

Most academic writing on direct government spending as an innovation policy tool focuses on how this mechanism compares with other policies rather than on the policy choices within the "direct spending" box. For example, in Beyond the Patents–Prizes Debate, Daniel Hemel and I considered a single category of "government grants—a category that includes direct spending on government research laboratories and grants to nongovernment researchers"—with a focus on the similarities among these direct spending mechanisms, and what makes them all different from the other tools in our four-box framework (R&D tax incentives, patents, and inducement prizes).

But we noted that there is variation within each policy box, and that in practice the boxes form a spectrum rather than discrete choices. And it is certainly worth diving within each of the four boxes of our framework from Beyond to dissect these policy tools. There is of course an extensive literature already on optimizing within the "patent" mechanism, but legal innovation scholars pay far less attention to the other boxes, including grants.

Even if one focuses on the most typical grants to academic scientists, there is some interesting research on the effect of different ways of awarding this funding, such as this paper by Azoulay et al. on NIH vs. HHMI grants. But the federal government also provides many other types of direct finding: in 2013, almost one-quarter of federal R&D expenditures went to for-profit firms. How does the theory behind this substantial expenditure of taxpayer funding differ from that for academic research, and what impact does it have in practice?

A recent study from Aleksandar Giga, Andrea Belz, Richard Terrile, and Fernando Zapatero at USC and NASA's Jet Propulsion Lab at Caltech provides some data on the Small Business Innovation Research (SBIR) program as administered by NASA. They find that compared with firms that do not receive these grants, "microfirms" (1-5 employees) with SBIR grants are twice as likely to produce patents and generate twice as many patents. They argue that this is unlikely to be due to a selection effect. They also find that the program does not show the same effect for larger firms, and they suggest that the size limitations for the program should be reconsidered.

Giga et al.'s work focuses on just one corner of the extensive field of direct government science funding, but I hope legal scholars will incorporate empirical work like this to provide a richer understanding of this type of innovation policy.

Tuesday, October 16, 2018

Equitable Servitudes and Post-Sale Restrictions

I have continued to find the issue of post-sale restrictions vexing. On the one hand, I think that there are sound economic reasons for them. On the other hand, I really don't like them, especially when they limit what should otherwise be reasonable and free activities.

The Supreme Court's recent cases in this area have made it more difficult to enforce such restrictions, but they have done so in a way that leaves open the possibility that some restrictions might apply while also not giving much guidance about when.

A recent article takes this on. Tim Scott (King & Spalding) has published The Availability of Post-Sale Contractual Restrictions in the Wake of Impression Products, Inc. v. Lexmark, 581 U.S. 1523 (2017) in Les Nouvelle. It is available on SSRN as well, and the abstract is below. I write about this article in part because the abstract is uninspiring when compared to the high quality analysis in the article. Don't be fooled.
In Impression Products, Inc. v. Lexmark International Inc., 581 U.S. ___, 137 S. Ct. 1523 (2017), the United States Supreme Court reaffirmed the patent exhaustion rule; i.e., patent rights are exhausted upon the first sale of the patented item such that the patentee has no rights to impose any post-sale conditions or limitations on the use of the product, at least under the patent laws. Id. at 1529. In doing so, the Court left open the question of whether such conditions or limitations could be imposed as a matter of contract law. Thus, the “restrictions in Lexmark’s contracts with its customers may have been clear and enforceable under contract law, but they do not entitle Lexmark to retain patent rights in an item that it has elected to sell.” Id. at 1531 (emphasis added). In summarizing its decision in Quanta Computer, Inc. v. LG Electronics, Inc., 533 U.S. 617 (2008) ruling that the sale of computer components exhausted the plaintiffs patent rights in those components, the Court noted that it reached that conclusion “without so much as mentioning the lawfulness of the contract.” Lexmark, 137 S. Ct. at 1533. And it later summarized its holding by stating that “whatever rights Lexmark retained are a matter of the contracts with its purchasers, not the patent law.” Id. (emphasis added).
This is my kind of article: it's short, it gets right to business, and it is thorough despite its brevity. The article introduces the cases, discusses contract v. property rights, discusses equitable servitudes, surveys the literature on equitable servitudes on personal property (including pros and cons), proposes alternatives to equitable servitudes (and points to critiques), and then discusses how the alternatives might have applied to Lexmark's activities in Impression Products.

Not bad for eight journal pages. This is recommended reading for anyone who wants a quick background on the state of post-sale restrictions.

Friday, October 12, 2018

Mike Andrews on Historical Patent Data

Mike Andrews is a postdoc at NBER, and I recently came across his PhD dissertation, Fuel of Interest and Fire of Genius: Essays on the Economic History of Innovation. He presents some interesting new results from historical patent records:

I already described the work in chapter 1 in my post on the NBER Summer Institute; in short, he compares U.S. counties that received new colleges in the period 1839-1954 with finalist sites that were not chosen for plausibly exogenous reasons. He finds that counties that received a college had 33% more patents per year, mostly due to increases in population rather than the colleges' graduates and faculty.

Chapter 2 looks at the effect of statewide alcohol bans on counties that previously set their own alcohol policies. Statewide prohibition reduced patents by 15% per year in previously wet counties relative to previously dry counties, and there is a larger decline for men than for women. Andrews suggests this decline is due to a disruption of information social interactions in saloons.

Chapter 3 matches 1870–1940 patents with census data and finds that patentees are consistently more likely to be older, white, male, and living in a state other than the one in which they were born. Establishment of a historically black college increased representation of black inventors, but the effect largely disappears after controlling for a county's black population, suggesting it is driven by concentration rather than the college itself. Extension of the franchise to women did not seem to increase the representation of women among inventors.

Finally, chapter 4 compares patent historical datasets. For those considering historical work with patent data, this is probably a good place to start. One should always be cautious about generalizing from the innovation institutions of a century ago to the ones that exist today—e.g., the effect of universities on the patent system has changed significantly in the past few decades—but it is still interesting to understand how patents worked in a particular historical context.

Tuesday, October 9, 2018

Do Patent Laws Affect the Location of R&D?

One of the common complaints about weakening patent protection is that it causes reduced R&D in the country with weakened protection. I've always been skeptical of this claim in the modern era, because one can develop anywhere and import into a location with better protection. As a result, one would expect that patent protection is unrelated to R&D offshoring.

In a draft article called Offshoring Patent Law, Gregory Day (Georgia Business) and Steven Udick (Skiermont Derby LLP) consider this question. Their article is forthcoming in the Washington Law Review and a draft is on SSRN. Here is the abstract:
Legislators and industry leaders claim that patent strength in the United States has declined, causing firms to innovate in foreign countries. However, scholarship has largely dismissed the theory that foreign patents have any effect on where firms invent, considering that patent law is bound by strict territorial limitations (as a result, one cannot strengthen their patent protection by innovating abroad). In essence, then, industry leaders are deeply divided from scholarship about whether innovative firms seek out jurisdictions offering stronger patent rights, affecting the rate of innovation.
To resolve this puzzle, we offer a novel theory of patent rights — which we empirically test — to dispel the positions taken by both scholarship and industry leaders. Since technology is generally developed in one country, the innovation process exposes the typical inventor to infringement claims only in that jurisdiction. In turn, we demonstrate that inventors have powerful, counterintuitive incentives to develop technology where patent rights are weaker and enforcement is cheaper. Given that it typically costs more to defend a patent infringement claim in the United States than to lose one in another country (the cost to litigate a patent in the United States averages around $3.5 million and royalty awards have surpassed $2.5 billion), our empirical research contributes to the theoretical understanding of patent rights by shedding new light on the important, yet largely dismissed, dimension of where innovation takes place.
We received invaluable support from international research organizations and patent attorneys working for top-tier law firms. Notably, the Global IP Project, which is a multinational research group spearheaded by the leading global intellectual property (“IP”) law firm, Finnegan, Henderson, Farabow, Garrett & Dunner, LLP, as well as Darts-ip, an international organization dedicated to the study of global IP litigation, provided proprietary data, enabling us to explore whether firms optimize value by placing research and innovation in countries with “better” patent laws. To verify our models, we interviewed notable patent attorneys practicing in the United States, Europe, and Asia.
The primary takeaway from their approach is that not only might the strength of the laws matter, but also the costs of defense. To tell this story, they use Marvell as an example, but that was actually a rare case where the R&D and sales process itself constituted infringement to trigger worldwide sales. I would expect that companies can usually design in the U.S., send designs overseas (see Microsoft v. AT&T), and ship from there. Thus, the more important complaint is that patent enforcement causes manufacturing to move offshore, not R&D.

That said, the article performs a regression on R&D and several variables that might affect R&D like tax rates and human capital density, and finds that costs of defense and damages awards are negatively correlated with R&D, while strength of enforcement is positively correlated. This is all reasonable enough, but I'm concerned that the empirical model is incomplete. Though the word "cost" appears dozens of times in the article, not once is it mentioned with respect to the cost of R&D. Might the reason R&D gets offshored be that it's cheaper? And could cheaper R&D also correlate with lower enforcement of IP? My guess is yes, based on the studies I've read over the years. I would have liked to have seen some analysis and discussion of this point.

While I think this is an interesting paper, I think that the model is underdeveloped in two ways. The first is the focus on costs in only half of the equation. The second is the neglect of trade secret enforcement. Unlike patent law, trade secret laws can affect R&D in the country in which the R&D takes place because the developer can lose value without ever selling into that country. Studies by Lippoldt and Schultz and also by Png demonstrate this pretty well.

For those interested in this topic, I recommend this article, and I recommend a contrast with Bilir, Patent Laws, Product Life-Cycle Lengths, and Multinational Activity, in the American Economic Review. Bilir develops a similar model, but bases it on location of companies (which covers some of the manufacturing issues), and considers the life-cycle of R&D (long term v. short term protection) as well as trade secrets. Bilir does not directly consider costs of defense, so it would be interesting to see how that notion from this new article would overlay onto Bilir.

Saturday, October 6, 2018

Language Barriers and Machine Translation

One of the more expensive parts of acquiring global patent protection is having a patent application translated into the relevant language for local patent offices. This is typically viewed simply as an administrative cost of the patent system, though my survey of how scientists use patents suggested that these translation costs may improve access to information about foreign inventions. As I wrote then, "[t]he idea that patents might be improving access to existing knowledge through mandatory translations and free accessibility is a very different disclosure benefit from the one generally touted for the patent system and seems worthy of further study." E.g., if researchers at a U.S. firm publish their results only in English but seek patent protection in the United States and Japan, then Japanese researchers who don't speak English would be able to read about the work in the Japanese patent.

I've also been interested in the proliferation of machine translation tools for patents—which can make patents even more accessible, but which also might limit this comparative advantage of patents over scientific publications if machine translation of journal articles becomes commonplace.

I don't know of much data on the actual economic impact of any of these translation tools, so I was intrigued to spot a new empirical study about the benefits of machine translation for international trade: Does Machine Translation Affect International Trade? Evidence from a Large Digital Platform. Three researchers from MIT and Washington University, Erik Brynjolfsson, Xiang Hui, and Meng Liu, found that the introduction of eBay Machine Translation increased international trade on the platform by 17.5%. They conclude: "Our results provide causal evidence that language barriers significantly hinder trade and that AI has already begun to improve economic efficiency in at least one domain."

Of course, this trade benefit of machine translation is different from the effect on patent disclosure, but the study made me wonder if a similar methodology could be applied by the hosts of patent translation tools (e.g., WIPO, EPO, SIPO) to study the increase in access to patent documents from different countries. Such a study could be a nice complement to survey-based work about how researchers in different countries access information about foreign work, and how machine translation fits into this picture. I'm not currently planning any of this work myself, but if the topic interests you, feel free to email—it seems like a fruitful area for a number of studies, and I'd love to share more thoughts and advice.

Tuesday, October 2, 2018

Valuing Wikimedia Commons Images

Several years ago, both Lisa and I wrote about Heald, et al.'s study that attempted to value public domain photographs as used on Wikipedia. While I liked the study a lot, two of my chief critiques were small sample size and unclear value of hits on Wikipedia pages.

A new paper extends their study, and provides even more evidence of the extensive use of Wikimedia Commons photos. In What is the Commons Worth? Estimating the Value of Wikimedia Imagery by Observing Downstream Use, Kris Erickson (University of Leeds), Felix Rodriguez Perez (Independent), and Jesus Rodriguez Perez (University of Glasgow), have attempted to generalize the findings from the prior study. The paper is published in an ACM conference proceeding, but is available without a paywall on SSRN. The abstract is here:
The Wikimedia Commons (WC) is a peer-produced repository of freely licensed images, videos, sounds and interactive media, containing more than 45 million files. This paper attempts to quantify the societal value of the WC by tracking the downstream use of images found on the platform. We take a random sample of 10,000 images from WC and apply an automated reverse-image search to each, recording when and where they are used ‘in the wild’. We detect 54,758 downstream uses of the initial sample and we characterize these at the level of generic and country-code top-level domains (TLDs). We analyze the impact of specific variables on the odds that an image is used. The random sampling technique enables us to estimate overall value of all images contained on the platform. Drawing on the method employed by Heald et al (2015), we find a potential contribution of USD $28.9 billion from downstream use of Wikimedia Commons images over the lifetime of the project.
In one fell swoop, the authors have answered my two concerns. The random sample is much larger, and their search went far beyond Wikipedia, to commercial and non-commercial uses. It turns out that the images were used a whopping 5.4 times each on average, which is a lot of usage when extrapolated to the millions of images in the Commons.

As with the prior study, estimating the value is a bit back of the envelope. Assuming that every commercial (and non-commercial) user would have paid the Getty Images fee is a big assumption, as many might have substituted to homegrown photos or maybe no photo at all. The authors note that this is a big assumption. Another issue is that not every item in the commons is within copyright, and my have been findable by other means.

That said, I do not think the assumption detracts from the value of the Wikimedia Commons for two reasons. First, they report Getty having revenues of nearly $1 billion per year, so finding $28 billion value over the lifetime of the WC is perhaps not far-fetched. Second, even if people would not pay the full amount, they might have been willing to pay less than the Getty fee (which also includes some public domain items). In the absence of WC, the differences between what they would have paid and what they get (either nothing or homegrown or search costs) is deadweight loss.

I frankly had no idea that Wikimedia Commons was used so much, but I'm glad that there's competition in the stock photo market. I'll finally note that the discussion about which images get used is an interesting one. It turns out-just like Netflix, Facebook, and Twitter-the stuff that gets curated for you is the stuff you wind up seeing and using.