I am not here to argue over whether blue states do better than red states economically. What I do want to point out is how professors Hacker and Pierson use intellectual property – and in particular patents – in making their argument. Companies in blue states, they write, "do more research and development and produce more patents[]" than red states. Indeed, "few of the cities that do the most research or advanced manufacturing or that produce the most patents are in red states." How, they ask rhetorically, can conservatives say red states are doing better when most patents are being generated in California?*
Hacker and Pierson's reasoning, which is quite common, goes like this. Patents are an indicator of innovation. Innovation is linked to economic prosperity. Therefore, patents – maybe even all forms of intellectual property – are linked to economic prosperity.
In my new paper, Technological Un/employment, I cast doubt on the connection between intellectual property and one important indicator of economic prosperity: employment.
This post is based on a talk I gave at the 2018 Works-In-Progress Intellectual Property (WIPIP) Colloquium at Case Western Reserve University School of Law on Saturday, February 17.
The usual story is that IP is good for jobs. For example, in debates surrounding passage of the America Invents Act (AIA), prominent members of Congress and the White House told us that a stronger and more streamlined patent system would promote innovation and help inventors and entrepreneurs "create jobs."
A recent report by the U.S. Patent & Trademark Office (USPTO) and the Economics and Statistics Administration (both located in the Department of Commerce) appears to provide empirical support for this long-held assumption. The report is entitled Intellectual Property And The U.S. Economy. It was issued in 2012 and updated in 2016.
The report identifies 81 industries, from among 313 total, as “IP-intensive”, meaning they have more patents, trademarks, or copyrights per size. The report concludes that in 2014, these 81 industries supported 45.5 million jobs in the economy, "about 30 percent of all employment.” What is more, in 2014, the average weekly wages of workers in IP-intensive industries like computer systems design were forty-six percent (46%) higher than in non-IP-intensive industries.
On its face, the report flawed for measuring intellectual property’s impact on employment for two reasons – one of which is fatal.
The first flaw, which Lisa Ouellette and the authors of the report themselves have pointed out, is that its methodology does not measure, or even try to measure, the causal connection between the incentive effects of intellectual property, and the documented impact on jobs and wages.** In other words, it could be true that IP-intensive industries added more and better-paying jobs than other industries because they own and enforce a lot of intellectual property, but it could also be because these industries are responsible for a lot of really great products and innovations that are in high demand, and they hire really good people whose skills are comparatively scarce.***
But the report's methodological limitation should not be taken to falsify a hypothesis that otherwise makes sense. After all, why wouldn't intellectual property contribute to job growth, and even higher wages, at least within those industries that generate, license, and enforce intellectual property? Assuming we accept intellectual property's incentives theory, then patents, copyrights, and trademarks (and trade secrets, which are a bit harder to measure), facilitate higher profits by conferring exclusive rights. The purpose of the exclusive rights is to stimulate innovation, creativity, and information production. But the upshot is more money to spend on research and output, and, yes, on hiring and wages. (As I discuss in my paper, several studies have observed sharing of innovation rents with workers.)
It is the report's second flaw that bothers me: it misses the other half the economy. Even if it is true that intellectual property rights are causally related to more hiring within IP-intensive industries, the report does not measure intellectual property’s impact on people working outside of IP-intensive industries.
The basic insight of the paper is as follows. Intellectual property may be partly responsible for job creation for people who work in IP-intensive industries such as software and robotics. But a significant subset of the innovations protected by intellectual property, from self-service kiosks to self-driving cars, are labor saving, and in many cases also labor displacing, innovations.
For example, the USPTO report finds software publishers, such as Oracle and Adobe, added 300, 600 jobs in 2013. But it says nothing about the impact of the software these companies created in that same year for others. Oracle, for instance, is beginning to specialize in machine learning software, and just released a product it unabashedly calls the World's First Self-Driving Database. The self-professed benefit of the Autonomous Database is that it will require "no human labor," will be "half the cost," and "100x more reliable." Oracle's invention is clearly labor saving. It is potentially "labor displacing" to the extent that it leads to a significant reduction in the amount of paid human labor required to collect and analyze data.
Speaking of self-driving things, autonomous vehicles provide another striking example of a labor displacing innovation. Companies like Alphabet, Uber, and Tesla are developing cars that can drive and navigate without human drivers. The result is greater profits for owners of intellectual property covering self-driving vehicles, and professional athlete-level salaries for roboticists, engineers, and A.I. specialists whose skills are necessary to generate this intellectual property. But self-driving vehicles, if widely adopted, could spell the end of paid employment for taxi drivers, Uber drivers, truck drivers, and millions of other people whose jobs entail driving for a living.
Self-driving cars may represent today's "Luddite moment."
Intellectual property is, or should be, a part of this discussion. A search for the term “labor saving” in Google Patents reveals over 90,000 results, such as labor saving long arm gardening shears , a labor saving materials dispenser , and a labor saving consolidated checkout system. The term “automation” yields over 400,000 results, including several recent patents involving “sales force automation” and “home automation system[s].” The term “autonomous vehicle” alone yields over 40,000 results, several of which are owned by Uber Technologies.
Take NCR Corp’s patent for a labor saving consolidated checkout system—the self-service checkout terminals we can now use at the grocery store and the pharmacy. Noting that “the largest expenditures” in the retail industry besides “the cost of the goods sold” are “the cost of labor expended,” the patent then discusses at length the invention’s goal to “reduce labor costs” associated with grocery and supermarket transactions by “reducing the number of occasions in which an employee of the retailer must intervene in the customer's transaction” relative to the prior art. In other words, the invention’s objective and major innovation is to reduce the amount of labor required to perform the task to as close to zero as possible.
To the extent the subject matter of intellectual property rights is labor displacing innovation, then intellectual property contributes to what economists call technological unemployment: job loss brought about by technological change.
I am not saying technology only eliminates jobs. Innovation also creates jobs. A lot of jobs. And sometimes those jobs are better quality and better paid. This is why, to be more accurate, I use a different term: technological un/employment. My title reflects the predominant view that advances in technology lead to both technological unemployment and technological employment.
The dark side of this story is job loss brought on by technological change. The big mover here is automation: use of machines to perform work otherwise done by paid humans. Self-Driving cars, machine learning that enables production of software, drones used to deliver packages are all examples.
Fortunately, economists have historically thought that the dark side is always going to be balanced out by the bright side: technological employment. There are two main ways technological employment happens. The first is "job generation," where technology generates new jobs requiring similar skill sets or new, unpredicted skills. The second mechanism is "demand-boosting." This is where labor saving technology leads to higher productivity, which permits lower prices, which boosts demand and thus hiring.
The ATM provides a compelling example. Although it permitted banking without going to see tellers, ATM's, B.U. professor James Bessen found, actually led to an increase in hiring of tellers. The reason, he theorizes was that ATM's lowered costs for banks and allowed opening new branches.
But the big concern today is that maybe this time is different. There are four reasons one might be worried. First, is the increasing amount and pace of automation. If machines get too good at what we do, there will be no jobs left for any human to do, period. Second, demand has limits. Especially if all workers are robots, who will do the spending? Third, I am skeptical of the quality of jobs left. How good will they be? Lastly, it is not at all clear who will be lucky enough to get those jobs. The probability is machines won't replace law professors, at least not for a while. They will replace poor people who didn't have the same education and job opportunities that we did. This is because the impact of technology on employment has historically been “skill-biased”—demand for high skills workers rises; demand for low skill workers falls.
So back to IP. I theorize that intellectual property has two effects on the process of technological un/employment: the Incentive Effect and the Distribution Effect. I don't think these two things are very easy to disagree with, but I call them hypotheses nonetheless.
The Incentive Effect predicts that the incentives generated by intellectual property’s right-to-exclude magnify and accelerate the pace of technological un/employment. The Incentive Effect works as follows. The chance to obtain an exclusive right increases the incentive to invent and commercialize any given innovation. Within the entire universe of innovation, at least some will be labor saving innovations. At least some of these labor saving innovations will end up being labor displacing—meaning they eliminate or significantly reduce the amount of labor required to complete a task that would otherwise be performed by a paid human worker. Therefore, the existence of effective intellectual property laws should make it more likely that any given labor displacing innovation will be invented, commercialized, and adopted in industry.
The Distribution Effect is an outgrowth of the Incentive Effect. The Distribution Effect has two parts. First, intellectual property increases returns for the owners of intellectual property by giving them a right to exclude, and thereby increases demand and wages for people who possess the skills necessary to generate intellectual property (“IP-generators”). Second, because at least some of this same intellectual property involves labor displacing innovations, this contributes to lower demand and wages for people whose core skills are more easily replaced by machines. The upshot is that intellectual property magnifies the division between the owners and generators of intellectual property, and the workers whom their innovations replace.
This state of affairs raises several normative questions. Should law intervene? What should law do? What role does IP law play, if any?
As to the first question, I do think the law should intervene. There are a few reasons. As an initial matter, it is possible to frame this as a story about negative externalities. Automation, some have argued, can be conceptualized as a negative externality: a cost imposed on some workers (such as, for instance, drivers and factory workers) by innovators, who do not take those peoples' futures into account when they choose to invent things (like self driving cars or factory robots) that make those peoples' skills completely obsolete in a short period of time. One can dispute this characterization at a technical level. But I think, at the least, we can see pervasive automation as a temporal externality: the inventions humans make today could end up meaning there are no jobs left for their grandchildren.
Many might be more convinced, though, by appeals to distributive justice. Think back to Hacker and Pierson's argument, that blue states are "generally doing better" than red states, and that California, New York, Texas, and Washington – and within those states Palo Alto, Manhattan, Austin, and Seattle – have a lot more patents than other places. To me this is disturbing. If studies like the USPTO report (and many others) are correct that intellectual property is correlated with more education, better employment, and higher wages, and yet intellectual property is concentrated in only some areas, this seems imbalanced and unfair. It is all the more so if my hypotheses are correct, and government-sanctioned intellectual property rights are contributing to unemployment and unequal distribution of wealth between IP owners, and everyone else.
I understand the big objection here. Productivity, the story goes, is the crux of a thriving economy. And innovation is crucial to increasing productivity. Cost saving, whether through replacement of humans or augmentation of humans, is what makes businesses successful. If we make it harder to automate, we'll make businesses less efficient and put ourselves – the American economy, that is – at a disadvantage. This would be totally contrary to the policies behind intellectual property and sound economics. As Milton Friedman said, "If it’s jobs you want, then you should give these workers spoons, not shovels." (Thanks, T.J. Chiang, for that quote.)
But I think that a limited penalty on innovation is not too high price a pay to correct some of the inequality in the job market that we already see and will continue to see more of as automation becomes more common.
To be clear, I am not against innovation. I am a huge supporter of innovation. I think it is good in the long run, both with respect to jobs and with respect to welfare generally.***** That is why the title and entire premise of the paper is "technological un/employment." Technology does both. And this is why any solutions adopted (a tax, for instance) must be small enough to do no more than redistribute and potentially slow— not stop— technological progress
We lastly come to the question of how should law intervene. Without playing policymaker myself, I identify the big options, and give guidance on how policymakers can decide among them.
1. Bans. Bans are bad. They are most likely to interfere with market choices and put too much rubber on innovation and productivity. They also don't help workers much. They put nothing in anyone's pocket.
2. Tax. Tax is a great contender. I like tax. So does Bill Gates, who has stated in an interview that government should consider a "robot tax." “You ought to be willing to raise the tax level and even slow down the speed” of automation, he said. Of all the options, this is probably the best type of solution that can be pursued, with my preference being to spend the proceeds on education or, if there's eventually no work left at all (unlikely), a universal basic income. But why tax the businesses that adopt robots? Why not tax the owners of the technology embedded in the robots – the people who are profiting? To me it makes more sense to tax, say, Oracle, than the mom-and-pop shop that adopts Oracle's software to save on costs. Robert Reich, for instance, has proposed a tax on underlying intellectual property to even out the distribution of returns from innovating. ("Researchers estimate that almost half of all U.S. jobs are at risk of being automated in the next two decades. ... One answer: A universal basic income – possibly financed out of the profits going to such labor replacing innovations, or perhaps even a revenue stream off of the underlying intellectual property.")
We lastly come to the question of how should law intervene. Without playing policymaker myself, I identify the big options, and give guidance on how policymakers can decide among them.
1. Bans. Bans are bad. They are most likely to interfere with market choices and put too much rubber on innovation and productivity. They also don't help workers much. They put nothing in anyone's pocket.
2. Tax. Tax is a great contender. I like tax. So does Bill Gates, who has stated in an interview that government should consider a "robot tax." “You ought to be willing to raise the tax level and even slow down the speed” of automation, he said. Of all the options, this is probably the best type of solution that can be pursued, with my preference being to spend the proceeds on education or, if there's eventually no work left at all (unlikely), a universal basic income. But why tax the businesses that adopt robots? Why not tax the owners of the technology embedded in the robots – the people who are profiting? To me it makes more sense to tax, say, Oracle, than the mom-and-pop shop that adopts Oracle's software to save on costs. Robert Reich, for instance, has proposed a tax on underlying intellectual property to even out the distribution of returns from innovating. ("Researchers estimate that almost half of all U.S. jobs are at risk of being automated in the next two decades. ... One answer: A universal basic income – possibly financed out of the profits going to such labor replacing innovations, or perhaps even a revenue stream off of the underlying intellectual property.")
3. Intellectual property. A final option is to disallow IP on labor displacing innovations. Without going into details (as I do in the paper), the basic idea is to deprive the innovator of the benefit of an exclusive right. Believe it or not, Queen Elizabeth thought this was a good idea. She denied William Lee a patent on his spinning loom, which reduced the amount of human labor needed to spin cloth, because she feared the implications for employment of her subjects. In Queen Elizabeth's time, no patent meant no permission. Today, the effect would be less dire – no patent just means no exclusivity; it doesn't mean no permission to practice at all. Still, if we accept the Incentive Effect, this would still dampen incentives to automate. In the paper, I end up rejecting this option. I think it's not the one that make most sense, not least because it requires the USPTO to do too much heavy lifting.
To conclude, the big picture take home I''d like to impart is that the story about IP and jobs, that IP is good for jobs, is misleading. IP is good for some people's jobs, and may be good for economic growth in the long run. IP is not good for everyone's job, especially if your job is being an Uber driver.
* California was responsible for 40,196 patents granted in 2015. Mississippi, for example, was responsible for only 138 – a markedly lesser number even when population size is accounted for. California has around 39.25 million people. Mississippi has around 2.989 million. This means California has 0.00103066666 patents per capita. Mississippi has 0.00004616928 patents per capita.
** The 2016 update to the USPTO report concedes that “[w]e find differences in employment, wages, value added, and other outcomes that are correlated with IP use…[But] our methodology does not permit us to attribute those differences to IP alone.” As Lisa put it, with respect to the initial 2012 report, the report "did not inquire into how IP relates to growth: It simply quantified the economic contribution (not the IP- specific contribution) of an extremely broad list of “IP-intensive” industries as five trillion dollars.”
*** The computer systems design industry includes firms like IBM that design computer systems, including hardware and software, and help people install and implement those systems.
****It is also worth noting that "IP-intensive" includes trademark-intensive industries, which include, inter alia, grocery stores. According to the report, grocery stores and grocery related products accounted for around 3.4 million (3,352,500) jobs in 2013.
*****Why else would I be obsessed with IP?