Covid-19 Innovation – Are We on the Right Track?

I never discuss my own research on this website – it’s more fun (for me at the very least!) to dive in to the great results the rest of the economics community produces. So I hope you’ll forgive me for breaking this rule today, as I want to show a few interesting, very time-sensitive results Jorge Lemus, Guillermo Marshall and I have developed about Covid-19 innovation.

Many of us in the innovation economics world have been asked by governments how they should handle R&D right now. The basic problem is clear. There is a pandemic. Stopping this has enormous economic benefits – a vaccine that arrived tomorrow would literally be among the most economically valuable inventions ever made. Treatments which allow normal economic activity are incredibly valuable as well. As always, governments have limited knowledge about who is able to invent what. There is tremendous uncertainty about how various R&D projects will pan out. Should government be running huge prizes for specific inventions? General subsidies for medical R&D? Precommitments to buy certain remedies? Perhaps laissez faire alone is enough to induce this invention? Should patents be stronger, to increase the returns to R&D, or should they be weaker, to encourage the WHO’s global access to remedies?

These are very challenging questions. Let’s instead narrow down to a simpler one: given existing policy, is the rate and direction of Covid-19 R&D worrying in any way? Our basic finding is that the rate of Covid innovation is incredibly rapid, but competitive forces are pushing that research in a very short-term direction. The policy implications are subtle – many ideas that we normally think of as useful for R&D, especially on global health issues, may actually be counterproductive.

Here we compare the rate of therapies somewhere in the pipeline, and the number of academic publications, related to Covid-19 compared to other epidemics like Zika, Ebola, and H1N1, and compared to breast cancer, the most heavily-funded long-run disease. The pipeline data is from BioMedTracker, a standardized commercial research database that independently validates reports of new projects on a given indication. Note two things. First, the rate of Covid research far exceeds the long-term average for breast cancer or the post-epidemic rate of research on Ebola, H1N1, or Zika. Second, this gap grows even larger after the globalization of the pandemic in early March 2020, as indicated by the vertical line in the figures. Covid therapies are entering the pipeline at a rate 15 to 80 times faster than any previous epidemic, with over 4 new therapies entering the commercial pipeline every single day. The number of these therapies in clinical trials within four months of the early December beginning of the epidemic exceeds the entire first-year number of trials for H1N1, Zika, and Ebola combined. This figure tracks through April 22, but the rate of new drugs in the pipelines since that date has continued at nearly the same pace. The expected return on Covid innovation is high enough to induce an incredible amount of entry.

Alas, there is a downside. Let’s split the pipeline into vaccines vs. other drugs, and repurposed drugs vs. novel compounds. The relative share of “short term” solutions – non-vaccines and repurposed drugs – is unusually high. 23 percent of Covid therapies are vaccines, versus at least half for the previous three recent less severe epidemics. Over 60 percent of Covid therapies are repurposed drugs, versus no more than a quarter of those for Ebola, Zika, or H1N1. The short-term share is particularly strong after Covid explodes globally in early March – the rate at which new vaccines enter the pipeline is essentially the same in February and April! Broadly, as the epidemic gets worse, a greater share of R&D goes to projects that can be developed quickly.

How should we interpret this? We need a model to understand what the direction of invention “should be”. Theoretically, let there be firms of different sizes considering paying a fixed cost to enter the market for drug therapies on a particular disease. After entering, these firms choose whether to work on short-term therapies, which can be developed quickly but are not as valuable, or long-term ones, which take time to invent but are quite valuable. Once some firm invents a remedy, the marginal value of other remedies change: for instance, a vaccine is more valuable than a treatment, but the marginal value of the vaccine if a reasonable treatment exists is lower than it would have been otherwise.

Let’s model an epidemic intensifying by a multiplicative increase in the payoff to all inventions related to that disease. This increased payoff for successful invention, holding the number of firms constant, increased each firm’s expected payoff from R&D. The higher expected payoff induces more firms to enter, particularly firms with limited specialized research capacity who otherwise wouldn’t bother with a disease outside their wheelhouse.

Increased entry means a more fractured market for R&D, with many small firms doing research instead of just a few big ones. Assume it would be efficient for most firms to try to invent a vaccine. A small firm – say, one which represents 1% of total research capacity in the industry – will reason as follows. “I can try to invent the vaccine, but the Sanofis and the Modernas of the world are likely to get their vaccine way before I do. However, since their projects will take many months to develop and validate, I can instead try to quickly develop a marginally useful treatment. The existence of that treatment, once invented, lowers the marginal payoff to working on vaccines, but who cares – I am very unlikely to invent a vaccine in any case.” The increased entry driven by huge payoffs to any successful Covid therapies causes entrants to inefficiently race toward lower-value therapies. If enough small firms begin to race in this way, even the large firms that otherwise would have worked on vaccines will give up. And note that this pattern appears empirically: more severe epidemic leads to more entry by small firms, more work on short-term projects, and a decreasing share of long-term projects even from large firms as that competitive racing gets worse.

The policy implication is somewhat worrying. Normally, we worry in global public health that there isn’t enough incentive to work on a disease class at all, since firms worry that poor countries would either be unable to afford, or would expropriate ex-post, any useful invention. Things like the advance market commitments supported by last year’s Nobel winner Michael Kremer, have richer governments precommit to buying successful therapies for diseases like malaria. In the case of Covid, however, the main problem isn’t the rate of innovation, but the direction – the expected financial return on innovation may be too high for short-run, partial solutions, yet too low for vaccines or novel therapies.

So how can society both encourage a ton of Covid-19 R&D so we can escape this pandemic, while also discouraging racing behavior toward minor solutions? We propose three solutions. First, limited antitrust enforcement on research joint ventures can help by causing firms to partially internalize the racing externality they would otherwise have imposed on their joint venture partner. Second, targeted subsidies specifically for vaccines or promising novel therapies reshift the balance back toward the social optimum. It’s amazing how little of this is being done: a recent Guardian roundup of leading vaccine candidates proudly notes that 40 million pounds of support has been given to the Imperial College London team. 40 million in support for a vaccine that, if successful, probably has a net present value of over a trillion pounds! Third, we suggest AMCs can be useful as long as the value of the AMC is based on the ex-ante value of the therapy being developed, rather than the ex-post value – doing so allows firms with ability to work on long-term, high-value solutions to ignore the racing being done by their smaller rivals.

April 2020 Draft: “Innovation During a Crisis: Evidence from Covid-19”. We worked very hard – literally many all nighters – to get an analysis done quickly which nonetheless uses the best available Covid R&D data, and provides policy advice based on rigorous innovation theory. In my work with an entrepreneurship program here at the University of Toronto since we finished the draft, where we consult with very experienced biotech VCs, I’ve actually noticed our proposed mechanism in action – everyone is aware of the competitive nature of Covid research, and the first-order questions are always whether a firm has the capacity to work at high speed, and whether their therapy could be modified into something less significant but quicker to develop. I truly do not believe the externality we suggest is merely a theoretical curiosity.

How We Create and Destroy Growth: A Nobel for Romer and Nordhaus

Occasionally, the Nobel Committee gives a prize which is unexpected, surprising, yet deft in how it points out underappreciated research. This year, they did no such thing. Both William Nordhaus and Paul Romer have been running favorites for years in my Nobel betting pool with friends at the Federal Reserve. The surprise, if anything, is that the prize went to both men together: Nordhaus is best known for his environmental economics, and Romer for his theory of “endogenous” growth.

On reflection, the connection between their work is obvious. But it is the connection that makes clear how inaccurate many of today’s headlines – “an economic prize for climate change” – really is. Because it is not the climate that both winners build on, but rather a more fundamental economic question: economic growth. Why are some places and times rich and others poor? And what is the impact of these differences? Adam Smith’s “The Wealth of Nations” is formally titled “An Inquiry into the Nature and Causes of the Wealth of Nations”, so these are certainly not new questions in economics. Yet the Classical economists did not have the same conception of economic growth that we have; they largely lived in a world of cycles, of ebbs and flows, with income per capita facing the constraint of agricultural land. Schumpeter, who certainly cared about growth, notes that Smith’s discussion of the “different progress of opulence in different nations” is “dry and uninspired”, perhaps only a “starting point of a sort of economic sociology that was never written.”

As each generation became richer than the one before it – at least in a handful of Western countries and Japan – economists began to search more deeply for the reason. Marx saw capital accumulation as the driver. Schumpeter certainly saw innovation (though not invention, as he always made clear) as important, though he had no formal theory. It was two models that appear during and soon after World War II – that of Harrod-Domar, and Solow-Swan-Tinbergen – which began to make real progress. In Harrod-Domar, economic output is a function of capital Y=f(K), nothing is produced without capital f(0)=0, the economy is constant returns to scale in capital df/dK=c, and the change in capital over time depends on what is saved from output minus what depreciates dK/dt=sY-zK, where z is the rate of depreciation. Put those assumptions together and you will see that growth, dY/dt=sc-z. Since c and z are fixed, the only way to grow is to crank up the savings rate, Soviet style. And no doubt, capital deepening has worked in many places.

Solow-type models push further. They let the economy be a function of “technology” A(t), the capital stock K(t), and labor L(t), where output Y(t)=K^a*(A(t)L(t))^(1-a) – that is, that production is constant returns to scale in capital and labor. Solow assumes capital depends on savings and depreciation as in Harrod-Domar, that labor grows at a constant rate n, and that “technology” grows at constant rate g. Solving this model gets you that the economy grows such that dY/dt=sy-k(n+z+g), and that output is exactly proportional to capital. You can therefore just run a regression: we observe the amount of labor and capital, and Solow shows that there is not enough growth in those factors to explain U.S. growth. Instead, growth seems to be largely driven by change in A(t), what Abramovitz called “the measure of our ignorance” but which we often call “technology” or “total factor productivity”.

Well, who can see that fact, as well as the massive corporate R&D facilities of the post-war era throwing out inventions like the transistor, and not think: surely the factors that drive A(t) are endogenous, meaning “from within”, to the profit-maximizing choices of firms? If firms produce technology, what stops other firms from replicating these ideas, a classic positive externality which would lead the rate of technology in a free market to be too low? And who can see the low level of convergence of poor country incomes to rich, and not think: there must be some barrier to the spread of A(t) around the world, since otherwise the return to capital must be extraordinary in places with access to great technology, really cheap labor, and little existing capital to combine with it. And another question: if technology – productivity itself! – is endogenous, then ought we consider not just the positive externality that spills over to other firms, but also the negative externality of pollution, especially climate change, that new technologies both induce and help fix? Finally, if we know how to incentivize new technology, and how growth harms the environment, what is the best way to mitigate the great environmental problem of our day, climate change, without stopping the wondrous increase in living standards growth keeps providing? It is precisely for helping answer these questions that Romer and Nordhaus won the Nobel.

Romer and Endogenous Growth

Let us start with Paul Romer. You know you have knocked your Ph.D. thesis out of the park when the great economics journalist David Warsh writes an entire book hailing your work as solving the oldest puzzle in economics. The two early Romer papers, published in 1986 and 1990, have each been cited more than 25,000 times, which is an absolutely extraordinary number by the standards of economics.

Romer’s achievement was writing a model where inventors spend money to produce inventions with increasing returns to scale, other firms use those inventions to produce goods, and a competitive Arrow-Debreu equilibrium still exists. If we had such a model, we could investigate what policies a government might wish to pursue if it wanted to induce firms to produce growth-enhancing inventions.

Let’s be more specific. First, innovation is increasing returns to scale because ideas are nonrival. If I double the amount of labor and capital, holding technology fixed, I double output, but if I double technology, labor, and capital, I more than double output. That is, give one person a hammer, and they can build, say, one staircase a day. Give two people two hammers, and they can build two staircases by just performing exactly the same tasks. But give two people two hammers, and teach them a more efficient way to combine nail and wood, and they will be able to build more than two staircases. Second, if capital and labor are constant returns to scale and are paid their marginal product in a competitive equilibrium, then there is no output left to pay inventors anything for their ideas. That is, it is not tough to model in partial equilibrium the idea of nonrival ideas, and indeed the realization that a single invention improves productivity for all is also an old one: as Thomas Jefferson wrote in 1813, “[h]e who receives an idea from me, receives instruction himself without lessening mine; as he who lights his taper at mine, receives light without darkening me.” The difficulty is figuring out how to get these positive spillovers yet still have “prices” or some sort of rent for the invention. Otherwise, why would anyone pursue costly invention?

We also need to ensure that growth is not too fast. There is a stock of existing technology in the world. I use that technology to create new innovations which grow the economy. With more people over time and more innovations over time, you may expect the growth rate to be higher in bigger and more technologically advanced societies. It is in part, as Michael Kremer points out in his One Million B.C. paper. Nonetheless, the rate of growth is not asymptotically increasing by any stretch (see, e.g., Ben Jones on this point). Indeed, growth is nearly constant, abstracting from the business cycle, in the United States, despite a big growth in population and the stock of existing technology.

Romer’s first attempt at endogenous growth was based on his thesis and published in the JPE in 1986. Here, he adds “learning by doing” to Solow: technology is a function of the capital stock A(t)=bK(t). As each firm uses capital, they generate learning which spills over to other firms. Even if population is constant, with appropriate assumptions on production functions and capital depreciation, capital, output, and technology grow over time. There is a problem here, however, and one that is common to any model based on learning-by-doing which partially spills over to other firms. As Dasgupta and Stiglitz point out, if there is learning-by-doing which only partially spills over, the industry is a natural monopoly. And even if it starts competitively, as I learn more than you, dynamically I can produce more efficiently, lower my prices, and take market share from you. A decentralized competitive equilibrium with endogenous technological growth is unsustainable!

Back to the drawing board, then. We want firms to intentionally produce technology in a competitive market as they would other goods. We want technology to be nonrival. And we want technology production to lead to growth. Learning-by-doing allows technology to spill over, but would simply lead to a monopoly producer. Pure constant-returns-to-scale competitive production, where technology is just an input like capital produced with a “nonconvexity” – only the initial inventor pays the fixed cost of invention – means that there is no output left to pay for invention once other factors get their marginal product. A natural idea, well known to Arrow 1962 and others, emerges: we need some source of market power for inventors.

Romer’s insight is that inventions are nonrival, yes, but they are also partially excludable, via secrecy, patents, or other means. In his blockbuster 1990 JPE Endogenous Technological Change, he lets inventions be given an infinite patent, but also be partially substitutable by other inventions, constraining price (this is just a Spence-style monopolistic competition model). The more inventions there are, the more efficiently final goods can be made. Future researchers can use present technology as an input to their invention for free. Invention is thus partially excludable in the sense that my exact invention is “protected” from competition, but also spills over to other researchers by making it easier for them to invent other things. Inventions are therefore neither public nor private goods, and also not “club goods” (nonrival but excludable) since inventors cannot exclude future inventors from using their good idea to motivate more invention. Since there is free entry into invention, the infinite stream of monopoly rents from inventions is exactly equal to their opportunity cost.

From the perspective of final goods producers, there are just technologies I can license as inputs, which I then use in a constant returns to scale way to produce goods, as in Solow. Every factor is paid its marginal product, but inventions are sold for more than their marginal cost due to monopolistic excludability from secrecy or patents. The model is general equilibrium, and gives a ton of insight about policy: for instance, if you subsidize capital goods, do you get more or less growth? In Romer (1986), where all growth is learning-by-doing, cheaper capital means more learning means more growth. In Romer (1990), capital subsidies can be counterproductive!

There are some issues to be worked out: the Romer models still have “scale effects” where growth is not constant, roughly true in the modern world, despite changes in population and the stock of technology (see Chad Jones’ 1995 and 1999 papers). The neo-Schumpeterian models of Aghion-Howitt and Grossman-Helpman add the important idea that new inventions don’t just add to the stock of knowledge, but also make old inventions less valuable. And really critically, the idea that institutions and not just economic fundamentals affect growth – meaning laws, culture, and so on – is a massive field of research at present. But it was Romer who first cracked the nut of how to model invention in general equilibrium, and I am unaware of any later model which solves this problem in a more satisfying way.

Nordhaus and the Economic Solution to Pollution

So we have, with Romer, a general equilibrium model for thinking about why people produce new technology. The connection with Nordhaus comes in a problem that is both caused by, and potentially solved by, growth. In 2018, even an ignoramus knows the terms “climate change” and “global warming”. This was not at all the case when William Nordhaus began thinking about how the economy and the environment interrelate in the early 1970s.

Growth as a policy goal was fairly unobjectionable as a policy goal in 1960: indeed, a greater capability of making goods, and of making war, seemed a necessity for both the Free and Soviet worlds. But by the early 1970s, environmental concerns arose. The Club of Rome warned that we were going to run out of resources if we continued to use them so unsustainably: resources are of course finite, and there are therefore “limits to growth”. Beyond just running out of resources, growth could also be harmful because of negative externalities on the environment, particularly the newfangled idea of global warming an MIT report warned about in 1970.

Nordhaus treated those ideas both seriously and skeptically. In a 1974 AER P&P, he notes that technological progress or adequate factor substitution allow us to avoid “limits to growth”. To put it simply, whales are limited in supply, and hence whale oil is as well, yet we light many more rooms than we did in 1870 due to new technologies and substitutes for whale oil. Despite this skepticism, Nordhaus does show concern for the externalities of growth on global warming, giving a back-of-the-envelope calculation that along a projected Solow-type growth path, the amount of carbon in the atmosphere will reach a dangerous 487ppm by 2030, surprisingly close to our current estimates. In a contemporaneous essay with Tobin, and in a review of an environmentalist’s “system dynamics” predictions of future economic collapse, Nordhaus reaches a similar conclusion: substitutable factors mean that running out of resources is not a huge concern, but rather the exact opposite, that we will have access to and use too many polluting resources, should worry us. That is tremendous foresight for someone writing in 1974!

Before turning back to climate change, can we celebrate again the success of economics against the Club of Rome ridiculousness? There were widespread predictions, from very serious people, that growth would not just slow but reverse by the end of the 1980s due to “unsustainable” resource use. Instead, GDP per capita has nearly doubled since 1990, with the most critical change coming for the very poorest. There would have been no greater disaster for the twentieth century than had we attempted to slow the progress and diffusion of technology, in agriculture, manufacturing and services alike, in order to follow the nonsense economics being promulgated by prominent biologists and environmental scientists.

Now, being wrong once is no guarantee of being wrong again, and the environmentalists appear quite right about climate change. So it is again a feather in the cap of Nordhaus to both be skeptical of economic nonsense, and also sound the alarm about true environmental problems where economics has something to contribute. As Nordhaus writes, “to dismiss today’s ecological concerns out of hand would be reckless. Because boys have mistakenly cried “wolf’ in the past does not mean that the woods are safe.”

Just as we can refute Club of Rome worries with serious economics, so too can we study climate change. The economy affects the climate, and the climate effects the economy. What we need an integrated model to assess how economic activity, including growth, affects CO2 production and therefore climate change, allowing us to back out the appropriate Pigouvian carbon tax. This is precisely what Nordhaus did with his two celebrated “Integrated Assessment Models”, which built on his earlier simplified models (e.g., 1975’s Can We Control Carbon Dioxide?). These models have Solow-type endogenous savings, and make precise the tradeoffs of lower economic growth against lower climate change, as well as making clear the critical importance of the social discount rate and the micro-estimates of the cost of adjustment to climate change.

The latter goes well beyond the science of climate change holding the world constant: the Netherlands, in a climate sense, should be underwater, but they use dikes to restraint the ocean. Likewise, the cost of adjusting to an increase in temperature is something to be estimated empirically. Nordhaus takes climate change very seriously, but he is much less concerned about the need for immediate action than the famous Stern report, which takes fairly extreme positions about the discount rate (1000 generations in the future are weighed the same as us, in Stern) and the costs of adjustment.

Consider the following “optimal path” for carbon from Nordhaus’ most recent run of the model, where the blue line is his optimum.

Note that he permits much more carbon than Stern or a policy which mandates temperatures stay below a 2.5 C rise forever. The reason is the costs to growth in the short term are high: the world is still very poor in many places! There was a vitriolic debate following the Stern report about who was correct: whether the appropriate social discount rate is zero or something higher is a quasi-philosophical debate going back to Ramsey (1928). But you can see here how important the calibration is.

There are other minor points of disagreement between Nordhaus and Stern, and my sense is that there has been some, though not full, convergence if their beliefs about optimal policy. But there is no disagreement whatsoever between the economic and environmental community that the appropriate way to estimate the optimal response to climate change is via an explicit model incorporating some sort of endogeneity of economic reaction to climate policy. The power of the model is that we can be extremely clear about what points of disagreement remain, and we can examine the sensitivity of optimal policy to factors like climate “tipping points”.

There is one other issue: in Nordhaus’ IAMs, and in Stern, you limit climate change by imposing cap and trade or carbon taxes. But carbon harms cross borders. How do you stop free riding? Nordhaus, in a 2015 AER, shows theoretically that there is no way to generate optimal climate abatement without sanctions for non-participants, but that relatively small trade penalties work quite well. This is precisely what Emmanuel Macron is currently proposing!

Let’s wrap up by linking Nordhaus even more tightly back to Romer. It should be noted that Nordhaus was very interested in the idea of pure endogenous growth, as distinct from any environmental concerns, from the very start of his career. His thesis was on the topic (leading to a proto-endogenous growth paper in the AER P&P in 1969), and he wrote a skeptical piece in the QJE in 1973 about the then-leading theories of what factors induce certain types of innovation (objections which I think have been fixed by Acemoglu 2002). Like Romer, Nordhaus has long worried that inventors do not receive enough of the return to their invention, and that we measure innovation poorly – see his classic NBER chapter on inventions in lighting, and his attempt to estimate how much of how much of society’s output goes to innovators.

The connection between the very frontier of endogenous growth models, and environmental IAMs, has not gone unnoticed by other scholars. Nordhaus IAMs tend to have limited incorporation of endogenous innovation in dirty or clean sectors. But a fantastic paper by Acemoglu, Aghion, Bursztyn, and Hemous combines endogenous technical change with Nordhaus-type climate modeling to suggest a middle ground between Stern and Nordhaus: use subsidies to get green energy close to the technological frontier, then use taxes once their distortion is relatively limited because a good green substitute exists. Indeed, since this paper first started floating around 8 or so years ago, massive subsidies to green energy sources like solar by many countries have indeed made the “cost” of stopping climate change much lower than if we’d relied solely on taxes, since now production of very low cost solar, and mass market electric cars, is in fact economically viable.

It may indeed be possible to solve climate change – what Stern called “the greatest market failure” man has ever seen – by changing the incentives for green innovation, rather than just by making economic growth more expensive by taxing carbon. Going beyond just solving the problem of climate change, to solving it in a way that minimizes economic harm, is a hell of an accomplishment, and more than worthy of the Nobel prizes Romer and Nordhaus won for showing us this path!

Some Further Reading

In my PhD class on innovation, the handout I give on the very first day introduces Romer’s work and why non-mathematical models of endogenous innovation mislead. Paul Romer himself has a nice essay on climate optimism, and the extent to which endogenous invention matters for how we stop global warming. On why anyone signs climate change abatement agreements, instead of just free riding, see the clever incomplete contracts insight of Battaglini and Harstad. Romer has also been greatly interested in the policy of “high-growth” places, pushing the idea of Charter Cities. Charter Cities involve Hong Kong like exclaves of a developing country where the institutions and legal systems are farmed out to a more stable nation. Totally reasonable, but in fact quite controversial: a charter city proposal in Madagascar led to a coup, and I can easily imagine that the Charter City controversy delayed Romer’s well-deserved Nobel laurel. The New York Times points out that Nordhaus’ brother helped write the Clean Air Act of 1970. Finally, as is always true with the Nobel, the official scientific summary is lucid and deep in its exploration of the two winners’ work.

“The First Patent Litigation Explosion,” C. Beauchamp (2016)

There has been a tremendous rise in patent litigation in the past decade (Bessen and Meurer 2005). Many of these lawsuits have come from “non-practicing entities” – also known as patent trolls – who use their patents to sue even as they produce no products themselves. These lawsuits are often targeted at end-users rather than directly infringing manufacturers, supposedly on the grounds that end-users are less able to defend themselves (see my coauthor Erik Hovenkamp on this point). For those who feel the patent system provides too many rights to patent-holders, to the detriment of societal welfare, problems like these are case in point.

But are these worries novel? The economics of innovation and entrepreneurship is, like much of economics, one where history proves illuminating. Nearly everything that we think is new has happened before. Fights over the use of IP to collude by incumbents? See Lampe and Moser 2010 JEH. The importance of venture capital to local ecosystems? In the late 19th century, this was true in the boomtown of Cleveland, as Lamoreaux and her coauthors showed in a 2006 C&S (as to why Cleveland declines as an innovative center, they have a nice paper on that topic as well). The role of patent brokers and other intermediaries? These existed in the 19th century! Open source invention in the early days of a new industry? Tales from the rise of the porter style of beer in the 18th century are not terribly different from the Homebrew Computer Club that led to the personal computer industry. Tradeoffs between secrecy, patenting, and alternative forms of protection? My colleague Alberto Galasso shows that this goes back to Renaissance Italy!

Given these examples, it should not be surprising that the recent boom in patent litigation is a historical rerun. Christopher Beauchamp of Brooklyn Law School, in a 2016 article in the Yale Law Journal, shows that all of the problems with patent litigation mentioned above are not new: indeed, the true heyday of patent litigation was not the 2010s, but the late 1800s! Knowing the number of lawsuits filed, not just the number litigated to decision, requires painstaking archival research. Having dug up these old archives, Beauchamp begins with a striking fact: the Southern District of New York alone had as many total patent lawsuits filed in 1880 as any district in 2010, and on a per patent basis had an order of magnitude more lawsuits. These legal battles were often virulent. For instance, Charles Goodyear’s brother held patents for the use of rubber in dentistry, using attractive young women to find dentists using the technique without a license. The aggressive legal strategy ended only when the Vulcanite Company’s hard-charging treasurer was murdered in San Francisco by a desperate dentist!

These lawsuits were not merely battles between the Apples and Samsungs of the day, but often involved lawsuits demanding small license payments from legally unsophisticated users. Iowa Senator Samuel Kirkwood: patentholders “say to each [farmer], ‘Sir, pay me so much a mile or so much a rod for the wire…or you must go to Des Moines…and defend a suit to be brought against you, the cost of which and the fees in which will in themselves be more than I demand of you…[O]ur people are paying day by day $10, $15, $20, when they do not know a particle more whether they owe the man a dollar or a cent…but paying the money just because it is cheaper to do it than to defend a suit.” Some of these lawsuits were legitimate, but many were making claims far beyond the scope of what a court would consider infringement, just as in the case of patent troll lawsuits today. Also like today, farmers and industry associations formed joint litigation pools to challenge what they considered weak patents.

In an echo of complaints about abuse of the legal system and differential costs of filing lawsuits compared to defending oneself, consider Minnesota Senator William Windom’s comments: “[B]y the authority of the United States you may go to the capital of a State and for a claim of $5 each you may send the United States marshal to a thousand men, or ten thousand…and compel them to travel hundreds of miles to defend against your claim, or, as more frequently occurs, to pay an unjust demand as the cheapest way of meeting it.” Precisely the same complaint applies to modern patent battles.

A question of great relevance to our modern patent litigation debate therefore is immediate: Why did these scattershot individual lawsuits eventually fade away in the late 1800s? Beauchamp is equivocal here, but notes that judicial hostility toward the approach may have decreased win rates, and hence the incentive to file against small, weak defendants. Further, the rise of the modern corporation (see Alfred Chandler’s Scale and Scope) in the late 19th century changed the necessity of sublicensing inventions to local ligitating attorneys, rather that just suing large infringing manufacturers directly.

Of course, not everything historic is a mirror of the present. A major source of patent litigation in the mid-1800s involved patent reissues. Essentially, a patent would be granted with weak scope. An industry would rise up using related non-infringing technology. A sophisticated corporation would buy the initial patent, then file for a “reissue” which expanded the scope of the patent to cover many technologies then in use. Just as “submarine” patents, held secretly in application while an industry grows, are a major problem recently, patent reissues led to frequent 19th century complaints, until changes in jurisprudence in the late 1800s led to greatly decreased deference to the reissued patent.

What does this history tell us about modern innovation policy? As Beauchamp discusses, “[t]o a modern observer, the content of the earlier legal and regulatory reactions can seem strikingly familiar. Many of the measures now proposed or attempted as solutions for the ills of modern patent litigation were proposed or attempted in the nineteenth century as well.” To the extent we are worried about how to stop “patent trolls” from enforcing weak patents against unsophisticated end-users, we ought look at how our 19th century forebears handled the weak barbed wire and well patents filed against small-town farmers. With the (often economically-illiterate) rise of the “Hipster Antitrust” ideas of Lina Khan and her compatriots, will the intersection of patent and antitrust law move from today’s “technocratic air” – Beauchamp’s phrase – to the more political battleground of the 19th century? And indeed, for patent skeptics like myself, how are we to reconcile the litigious era of patenting of 1850-1880 with the undisputed fact that this period was dead in the heart of the Second Industrial Revolution, the incredible rise of electricity and modern chemicals inventions that made the modern world?

Full article is in Yale Law Journal, Feb. 2016.

“Patents as a Spur to Subsequent Innovation: Evidence from Pharmaceuticals,” D. Gilchrist (2016)

Many economists of innovation are hostile to patents as they currently stand: they do not seem to be important drivers of R&D in most industries, the market power they lead to generates substantial deadweight loss, the legal costs around enforcing patents are incredible, and the effect on downstream innovation can be particularly harmful. The argument for patents seems most clear cut in industries where the invention requires large upfront fixed costs of R&D that are paid only by the first inventor, where the invention is clearly delineated, where novelty is easy to understand, and where alternative means of inducing innovation (such as market power in complementary markets, or a large first mover advantage) do not exist. The canonical example of an industry of this type is pharma.

Duncan Gilchrist points out that the market power a patentholder obtains also affects the rents of partial substitutes which might be invented later. Imagine there is a blockbuster statin on patent. If I invent a related drug, the high price of the existing patented drug means I can charge a fairly high price too. If the blockbuster drug were off patent, though, my competitors would be generics whose low price would limit how much I can charge. In other words, the “effective” patent strength in terms of the markup I can charge depends on whether alternatives to my new drug are on patent or are generic. Therefore, the profits I will earn from my drug will be lower when alternative generics exist, and hence my incentive to pay a fixed cost to create the new drug will also be lower.

What does this mean for welfare? A pure “me-too” imitation drug, which generates very little social value compared to the existing patented drug, will never enter if its class is going to see generics in a few years anyway; profits will be competed down to zero. That same drug might find it worthwhile to pay a fixed cost of invention and earn duopoly profits if the existing on patent alternative had many years of patent protection remaining. On the other hand, a drug so much better than existing drugs that even at the pure monopoly price most consumers would prefer it to the existing alternative priced at marginal cost will be developed no matter what, since it faces no de facto restriction on its markup from whether the alternatives in its drug class are generics or otherwise. Therefore, longer patent protection from existing drugs increases entry of drugs in the same class, but mainly those that are only a bit better than existing drugs. This may be better or worse for welfare: there is a wasteful costs of entering with a drug only slightly better than what exists (the private return includes the business stealing, while social welfare doesn’t), but there are also lower prices and perhaps some benefit from variety.

I should note a caveat that really should have been noted in the existing model: changes in de facto patent length for the first drug in class also affect the entry decision of that drug. Longer patent protection may actually cause shorter effective monopoly by inducing entry of imitators! This paper is mainly empirical, so no need for a full Aghion Howitt ’92 model of creative destruction, but it is at least worth noting that the welfare implications of changes in patent protection are somewhat misstated because of this omission.

Empirically, Gilchrist shows clearly that the beginning of new clinical trials for drugs falls rapidly as the first drug in their class has less time remaining on patent: fear of competition with generic partial substitutes dulls the incentive to innovate. The results are clear in straightforward scatterplots, but there is also an IV, to help confirm the causal interpretation, using the gap between the first potentially-defensive patent on the fulcrum patent of the eventual drug, and the beginning of clinical trials, a gap that is driven by randomness in things like unexpected delays in in-house laboratory progress. Using the fact that particularly promising drugs get priority FDA review, Gilchrist also shows that these priority review entrants do not seem to be worried at all about competition from generic substitutes: the “me-too” type of drugs are the ones for whom alternatives going off patent is most damaging to profits.

Final published version in AEJ: Applied 8(4) (No RePEc IDEAS version). Gilchrist is a rare example of a well published young economist working in the private sector; he has a JPE on social learning and a Management Science on behavioral labor in addition to the present paper, but works at robo-investor Wealthfront. In my now six year dataset of the economics job market (which I should discuss again at some point), roughly 2% of “job market stars” wind up outside academia. Budish, Roin and Williams used the similar idea of investigating the effect of patents of innovation by taking advantage of the differing effective patent length drugs for various maladies get as a result of differences in the length of clinical trials following the patent grant. Empirical work on the effect of patent rules is, of course, very difficult since de jure patent strength is very similar in essentially every developed country and every industry; taking advantage of differences in de facto strength is surely a trick that will be applied more broadly.

“Scale versus Scope in the Diffusion of New Technology,” D. Gross (2016)

I am spending part of the fall down at Duke University visiting the well-known group of innovation folks at Fuqua and co-teaching a PhD innovation course with Wes Cohen, who you may know via his work on Absorptive Capacity (EJ, 1989), the “Carnegie Mellon” survey of inventors with Dick Nelson and John Walsh, and his cost sharing R&D argument (article gated) with Steven Klepper. Last week, the class went over a number of papers on the diffusion of technology over space and time, a topic of supreme importance in the economics of innovation.

There are some canonical ideas in diffusion. First, cumulative adoption on the extensive margin – are you or your firm using technology X – follows an S-curve, rising slowly, then rapidly, then slowly again until peak adoption is reached. This fact is known to economists thanks to Griliches 1957 but the idea was initially developed by social psychologists and sociologists. Second, there are massive gaps in the ability of firms and nations to adopt and quickly diffuse new technologies – Diego Comin and Burt Hobijn have written a great deal on this problem. Third, the reason why technologies are slow to adopt depends on many factors, including social learning (e.g., Conley and Udry on pineapple growing in Ghana), pure epidemic-style network spread (the “Bass model”), capital replacement, “appropriate technologies” arriving once conditions are appropriate, and many more.

One that is very much underrated, however, is that technologies diffuse because they and their complements change over time. Dan Gross from HBS, another innovation scholar who likes delving into history, has a great example: the early tractor. The tractor was, in theory, invented in the 1800s, but was uneconomical and not terribly useful. With an invention by Ford in the 1910s, tractors began to spread, particularly among the US wheat belt. The tractor eventually spreads to the rest of the Midwest in the late 1920s and 1930s. A back-of-the-envelope calculation by Gross suggests the latter diffusion saved something like 10% of agricultural labor in the areas where it spread. Why, then, was there such a lag in many states?

There are many hypotheses in the literature: binding financial constraints, differences in farm sizes that make tractors feasible in one area and not another, geographic spread via social learning, and so on. Gross’ explanation is much more natural: early tractors could not work with crops like corn, and it wasn’t until after a general purpose tractor was invented in the 1920s that complementary technologies were created allowing the tractor to be used on a wide variety of farms. The charts are wholly convincing on this point: tractor diffusion time is very much linked to dominant crop, the early tractor “skipped” geographies where were inappropriate, and farms in areas where tractors diffused late nonetheless had substantial diffusion of automobiles, suggesting capital constraints were not the binding factor.

But this leaves one more question: why didn’t someone modify the tractor to make it general purpose in the first place? Gross gives a toy model that elucidates the reason quite well. Assume there is a large firm that can innovate on a technology, and can either develop a general purpose or applied versions of the technology. Assume that there is a fringe of firms that can develop complementary technology to the general purpose one (a corn harvester, for instance). If the large firm is constrained in how much innovation it can perform at any one time, it will first work on the project with highest return. If the large firm could appropriate the rents earned by complements – say, via a licensing fee – it would like to do so, but that licensing fee would decrease the incentive to develop the complements in the first place. Hence the large firm may first work on direct applications where it can capture a larger share of rents. This will imply that technology diffuses slowly first because applications are very specialized, then only as the high-return specialties have all been developed will it become worthwhile to shift researchers over to the general purpose technology. The general purpose technology will induce complements and hence rapid diffusion. As adoption becomes widespread, the rate of adoption slows down again. That is, the S-curve is merely an artifact of differing incentives to change the scope of an invention. Much more convincing that reliance on behavioral biases!

2016 Working Paper (RePEc IDEAS version). I have a paper with Jorge Lemus at Illinois on the problem of incentivizing firms to work on the right type of project, and the implications thereof. We didn’t think in terms of product diffusion, but the incentive to create general purpose technologies can absolutely be added straight into a model of that type.

“Inventing Prizes: A Historical Perspective on Innovation Awards and Technology Policy,” B. Z. Khan (2015)

B. Zorina Khan is an excellent and underrated historian of innovation policy. In her new working paper, she questions the shift toward prizes as an innovation inducement mechanism. The basic problem economists have been grappling with is that patents are costly in terms of litigation, largely due to their uncertainty, that patents impose deadweight loss by granting inventors market power (as noted at least as far back as Nordhaus 1969), and that patent rights can lead to an anticommons which in some cases harms follow-on innovation (see Scotchmer and Green and Bessen and Maskin for the theory, and papers like Heidi Williams’ genome paper for empirics).

There are three main alternatives to patents, as I see them. First, you can give prizes, determined ex-ante or ex-post. Second, you can fund R&D directly with government, as the NIH does for huge portions of medical research. Third, you can rely on inventors accruing rents to cover the R&D without any government action, such as by keeping their invention secret, relying on first mover advantage, or having market power in complementary goods. We have quite a bit of evidence that the second, in biotech, and the third, in almost every other field, is the primary driver of innovative activity.

Prizes, however, are becoming more and more common. There are X-Prizes for space and AI breakthroughs, advanced market commitments for new drugs with major third world benefits, Kremer’s “patent buyout” plan, and many others. Setting the prize amount right is of course a challenging project (one that Kremer’s idea partially fixes), and in this sense prizes run “less automatically” than the patent system. What Khan notes is that prizes have been used frequently in the history of innovation, and were frankly common in the late 18th and 19th century. How useful were they?

Unfortunately, prizes seem to have suffered many problems. Khan has an entire book on The Democratization of Invention in the 19th century. Foreign observers, and not just Tocqueville, frequently noted how many American inventors came from humble backgrounds, and how many “ordinary people” were dreaming up new products and improvements. This frenzy was often, at the time, credited to the uniquely low-cost and comprehensive U.S. patent system. Patents were simple enough, and inexpensive enough, to submit that credit for and rights to inventions pretty regularly flowed to people who were not politically well connected, and for inventions that were not “popular”.

Prizes, as opposed to patents, often incentivized the wrong projects and rewarded the wrong people. First, prizes were too small to be economically meaningful; when the well-named Hippolyte Mège-Mouriès made his developments in margarine and garnered the prize offered by Napoleon III, the value of that prize was far less than the value of the product itself. In order to shift effort with prizes, the prize designer needs to know both enough about the social value of the proposed invention to set the prize amount high enough, and enough about the value of alternatives that the prize doesn’t distort effort away from other inventions that would be created while relying solely on trade secrecy and first mover advantage (I discuss this point in much greater depth in my Direction of Innovation paper with Jorge Lemus). Providing prizes to only some inventions may either generate no change in behavior at all because the prize is too small compared with the other benefits of inventing, or cause inefficient distortions in behavior. Even though, say, a malaria vaccine would be very useful, an enormous prize for a malaria vaccine will distort health researcher effort away from other projects in a way that is tough to calculate ex-ante without a huge amount of prize designer knowledge.

There is a more serious problem with prizes. Because the cutoff for a prize is less clear cut, there is more room for discretion and hence a role for wasteful lobbying and personal connection to trump “democratic invention”. Khan notes that even though the French buyout of Daguerre’s camera patent is cited as a classic example of a patent buyout in the famous Kremer QJE article, it turns out that Daguerre never actually held any French patent at all! What actually happened was that Daguerre lobbied the government for a sinecure in order to make his invention public, but then patented it abroad anyway! There are many other political examples, such as the failure of the uneducated clockmaker John Harrison to be granted a prize for his work on longitude due partially to the machinations of more upper class competitors who captured the ear of the prize committee. Examining a database of great inventors on both sides of the Atlantic, Khan found that prizes were often linked to factors like overcoming hardship, having an elite education, or regional ties. That is, the subjectivity of prizes may be stronger than the subjectivity of patents.

So then, we have three problems: prize designers don’t know enough about the relative import of various ideas to set price amounts optimally, prizes in practice are often too small to have much effect, and prizes lead to more lobbying and biased rewards than patents. We shouldn’t go too far here; prizes still may be an important part of the innovation policy toolkit. But the history Khan lays out certainly makes me more sanguine that they are a panacea.

One final point. I am unconvinced that patents really help the individual or small inventor very much either. I did a bit of hunting: as far as I can tell, there is not a single billionaire who got that way primarily by selling their invention. Many people developed their invention in a firm, but non-entrepreneurial invention, for which the fact that patents create a market for knowledge is supposedly paramount, doesn’t seem to be making anyone super rich. This is even though there are surely a huge number of inventions each worth billions. A good defense of patents as our main innovation policy should really grapple better with this fact.

July 2015 NBER Working Paper (RePEc IDEAS). I’m afraid the paper is gated if you don’t have an NBER subscription, and I was unable to find an ungated copy.

“Editor’s Introduction to The New Economic History and the Industrial Revolution,” J. Mokyr (1998)

I taught a fun three hours on the Industrial Revolution in my innovation PhD course this week. The absolutely incredible change in the condition of mankind that began in a tiny corner of Europe in an otherwise unremarkable 70-or-so years is totally fascinating. Indeed, the Industrial Revolution and its aftermath are so important to human history that I find it strange that we give people PhDs in social science without requiring at least some study of what happened.

My post today draws heavily on Joel Mokyr’s lovely, if lengthy, summary of what we know about the period. You really should read the whole thing, but if you know nothing about the IR, there are really five facts of great importance which you should be aware of.

1) The world was absurdly poor from the dawn of mankind until the late 1800s, everywhere.
Somewhere like Chad or Nepal today fares better on essentially any indicator of development than England, the wealthiest place in the world, in the early 1800s. This is hard to believe, I know. Life expectancy was in the 30s in England, infant mortality was about 150 per 1000 live births, literacy was minimal, and median wages were perhaps 3 to 4 times subsistence. Chad today has a life expectancy of 50, infant mortality of 90 per 1000, a literacy of 35%, and urban median wages of roughly 3 to 4 times subsistence. Nepal fares even better on all counts. The air from the “dark, Satanic mills” of William Blake would have made Beijing blush, “night soil” was generally just thrown on to the street, children as young as six regularly worked in mines, and 60 to 80 hours a week was a standard industrial schedule.

The richest places in the world were never more than 5x subsistence before the mid 1800s

Despite all of this, there was incredible voluntary urbanization: those dark, Satanic mills were preferable to the countryside. My own ancestors were among the Irish that fled the Potato famine. Mokyr’s earlier work on the famine, which happened in the British Isles after the Industrial Revolution, suggest 1.1 to 1.5 million people died from a population of about 7 million. This is similar to the lower end of the range for percentage killed during the Cambodian genocide, and similar to the median estimates of the death percentage during the Rwandan genocide. That is, even in the British Isles, famines that would shock the world today were not unheard of. And even if you wanted to leave the countryside, it may have been difficult to do so. After Napoleon, serfdom remained widespread east of the Elbe river in Europe, passes like the “Wanderbucher” were required if one wanted to travel, and coercive labor institutions that tied workers to specific employers were common. This is all to say that the material state of mankind before and during the Industrial Revolution, essentially anywhere in the world, would be seen as outrageous deprivation to us today; palaces like Versailles are not representative, as should be obvious, of how most people lived. Remember also that we are talking about Europe in the early 1800s; estimates of wages in other “rich” societies of the past are even closer to subsistence.

2) The average person did not become richer, nor was overall economic growth particularly spectacular, during the Industrial Revolution; indeed, wages may have fallen between 1760 and 1830.

The standard dating of the Industrial Revolution is 1760 to 1830. You might think: factories! The railroad! The steam engine! High Britannia! How on Earth could people have become poorer? And yet it is true. Brad DeLong has an old post showing Bob Allen’s wage reconstructions: Allen found British wages lower than their 1720 level in 1860! John Stuart Mill, in his 1870 textbook, still is unsure whether all of the great technological achievements of the Industrial Revolution would ever meaningfully improve the state of the mass of mankind. And Mill wasn’t the only one who noticed, there were a couple of German friends, who you may know, writing about the wretched state of the Working Class in Britain in the 1840s as well.

3) Major macro inventions, and growth, of the type seen in England in the late 1700s and early 1800s happened many times in human history.

The Iron Bridge in Shropshire, 1781, proving strength of British iron

The Industrial Revolution must surely be “industrial”, right? The dating of the IR’s beginning to 1760 is at least partially due to the three great inventions of that decade: the Watt engine, Arkwright’s water frame, and the spinning jenny. Two decades later came Cort’s famous puddling process for making strong iron. The industries affected by those inventions, cotton and iron, are the prototypical industries of England’s industrial height.

But if big macro-inventions, and a period of urbanization, are “all” that defines the Industrial Revolution, then there is nothing unique about the British experience. The Song Dynasty in China saw the gun, movable type, a primitive Bessemer process, a modern canal lock system, the steel curved moldboard plow, and a huge increase in arable land following public works projects. Netherlands in the late 16th and early 17th century grew faster, and eventually became richer, than Britain ever did during the Industrial Revolution. We have many other examples of short-lived periods of growth and urbanization: ancient Rome, Muslim Spain, the peak of the Caliphate following Harun ar-Rashid, etc.

We care about England’s growth and invention because of what followed 1830, not what happened between 1760 and 1830. England was able to take their inventions and set on a path to break the Malthusian bounds – I find Galor and Weil’s model the best for understanding what is necessary to move from a Malthusian world of limited long-run growth to a modern world of ever-increasing human capital and economic bounty. Mokyr puts it this way: “Examining British economic history in the period 1760-1830 is a bit like studying the history of Jewish dissenters between 50 B.C. and 50 A.D. At first provincial, localized, even bizarre, it was destined to change the life of every man and women…beyond recognition.”

4) It is hard for us today to understand how revolutionary ideas like “experimentation” or “probability” were.

In his two most famous books, The Gifts of Athena and The Lever of Riches, Mokyr has provided exhausting evidence about the importance of “tinkerers” in Britain. That is, there were probably something on the order of tens of thousands of folks in industry, many not terribly well educated, who avidly followed new scientific breakthroughs, who were aware of the scientific method, who believed in the existence of regularities which could be taken advantage of by man, and who used systematic processes of experimentation to learn what works and what doesn’t (the development of English porter is a great case study). It is impossible to overstate how unusual this was. In Germany and France, science was devoted mainly to the state, or to thought for thought’s sake, rather than to industry. The idea of everyday, uneducated people using scientific methods somewhere like ar-Rashid’s Baghdad is inconceivable. Indeed, as Ian Hacking has shown, it wasn’t just that fundamental concepts like “probabilistic regularities” were difficult to understand: the whole concept of discovering something based on probabilistic output would not have made sense to all but the very most clever person before the Enlightenment.

The existence of tinkerers with access to a scientific mentality was critical because it allowed big inventions or ideas to be refined until they proved useful. England did not just invent the Newcomen engine, put it to work in mines, and then give up. Rather, England developed that Newcomen engine, a boisterous monstrosity, until it could profitably be used to drive trains and ships. In Gifts of Athena, Mokyr writes that fortune may sometimes favor the unprepared mind with a great idea; however, it is the development of that idea which really matters, and to develop macroinventions you need a small but not tiny cohort of clever, mechanically gifted, curious citizens. Some have given credit to a political system, or to the patent system, for the widespread tinkering, but the qualitative historical evidence I am aware of appears to lean toward cultural explanations most strongly. One great piece of evidence is that contemporaries wrote often about the pattern where Frenchmen invented something of scientific importance, yet the idea diffused and was refined in Britain. Any explanation of British uniqueness must depend on Britain’s ability to refine inventions.

5) The best explanations for “why England? why in the late 1700s? why did growth continue?” do not involve colonialism, slavery, or famous inventions.

First, we should dispose of colonialism and slavery. Exports to India were not particularly important compared to exports to non-colonial regions, slavery was a tiny portion of British GDP and savings, and many other countries were equally well-disposed to profit from slavery and colonialism as of the mid-1700s, yet the IR was limited to England. Expanding beyond Europe, Dierdre McCloskey notes that “thrifty self-discipline and violent expropriation have been too common in human history to explain a revolution utterly unprecedented in scale and unique to Europe around 1800.” As for famous inventions, we have already noted how common bursts of cleverness were in the historic record, and there is nothing to suggest that England was particularly unique in its macroinventions.

To my mind, this leaves two big, competing explanations: Mokyr’s argument that tinkerers and a scientific mentality allowed Britain to adapt and diffuse its big inventions rapidly enough to push the country over the Malthusian hump and into a period of declining population growth after 1870, and Bob Allen’s argument that British wages were historically unique. Essentially, Allen argues that British wages were high compared to its capital costs from the Black Death forward. This means that labor-saving inventions were worthwhile to adopt in Britain even when they weren’t worthwhile in other countries (e.g., his computations on the spinning jenny). If it worthwhile to adopt certain inventions, then inventors will be able to sell something, hence it is worthwhile to invent certain inventions. Once adopted, Britain refined these inventions as they crawled down the learning curve, and eventually it became worthwhile for other countries to adopt the tools of the Industrial Revolution. There is a great deal of debate about who has the upper hand, or indeed whether the two views are even in conflict. I do, however, buy the argument, made by Mokyr and others, that it is not at all obvious that inventors in the 1700s were targeting their inventions toward labor saving tasks (although at the margin we know there was some directed technical change in the 1860s), nor it is even clear that invention overall during the IR was labor saving (total working hours increased, for instance).

Mokyr’s Editor’s Introduction to “The New Economic History and the Industrial Revolution” (no RePEc IDEAS page). He has a followup in the Journal of Economic History, 2005, examining further the role of an Enlightenment mentality in allowing for the rapid refinement and adoption of inventions in 18th century Britain, and hence the eventual exit from the Malthusian trap.

“Identifying Technology Spillovers and Product Market Rivalry,” N. Bloom, M. Schankerman & J. Van Reenen (2013)

How do the social returns to R&D differ from the private returns? We must believe there is a positive gap between the two given the widespread policies of subsidizing R&D investment. The problem is measuring the gap: theory gives us a number of reasons why firms may do more R&D than the social optimum. Most intuitively, a lot of R&D contains “business stealing” effects, where some of the profit you earn from your new computer chip comes from taking sales away from me, even if you chip is only slightly better than mine. Business stealing must be weighed against the fact that some of the benefits of knowledge a firm creates is captured by other firms working on similar problems, and the fact that consumers get surplus from new inventions as well.

My read of the literature is that we don’t have know much about how aggregate social returns to research differ from private returns. The very best work is at the industry level, such as Trajtenberg’s fantastic paper on CAT scans, where he formally writes down a discrete choice demand system for new innovations in that product and compares R&D costs to social benefits. The problem with industry-level studies is that, almost by definition, they are studying the social return to R&D in ex-post successful new industries. At an aggregate level, you might think, well, just include the industry stock of R&D in a standard firm production regression. This will control for within-industry spillovers, and we can make some assumption about the steepness of the demand curve to translate private returns given spillovers into returns inclusive of consumer surplus.

There are two problems with that method. First, what is an “industry” anyway? Bloom et al point out in the present paper that even though Apple and Intel do very similar research, as measured by the technology classes they patent in, they don’t actually compete in the product market. This means that we want to include “within-similar-technology-space stock of knowledge” in the firm production function regression, not “within-product-space stock of knowledge”. Second, and more seriously, if we care about social returns, we want to subtract out from the private return to R&D any increase in firm revenue that just comes from business stealing with slightly-improved versions of existing products.

Bloom et al do both in a very interesting way. First, they write down a model where firms get spillovers from research in similar technology classes, then compete with product market rivals; technology space and product market space are correlated but not perfectly so, as in the Apple/Intel example. They estimate spillovers in technology space using measures of closeness in terms of patent classes, and measure closeness in product space based on the SIC industries that firms jointly compete in. The model overidentifies the existence of spillovers: if technological spillovers exist, then you can find evidence conditional on the model in terms of firm market value, firm R&D totals, firm productivity and firm patent activity. No big surprises, given your intuition: technological spillovers to other firms can be seen in every estimated equation, and business stealing R&D, though small in magnitude, is a real phenomenon.

The really important estimate, though, is the level of aggregate social returns compared to private returns. The calculation is non-obvious, and shuttled to an online appendix, but essentially we want to know how increasing R&D by one dollar increases total output (the marginal social return) and how increasing R&D by one dollar increases firm revenue (marginal private return). The former may exceed the latter if the benefits of R&D spill over to other firms, but the latter may exceed the former is lots of R&D just leads to business stealing. Note that any benefits in terms of consumer surplus are omitted. Bloom et al find aggregate marginal private returns on the order of 20%, and social returns on the order of 60% (a gap referred to as “29.2%” instead of “39.2%” in the paper; come on, referees, this is a pretty important thing to not notice!). If it wasn’t for business stealing, the gap between social and private returns would be ten percentage points higher. I confess a little bit of skepticism here; do we really believe that for the average R&D performing firm, the marginal private return on R&D is 20%? Nonetheless, the estimate that social returns exceed private returns is important. Even more important is the insight that the gap between social and private returns depends on the size of the technology spillover. In Bloom et al’s data, large firms tend to do work in technology spaces with more spillovers, while small firms tend to work on fairly idiosyncratic R&D; to greatly simplify what is going on, large firms are doing more general R&D than the very product-specific R&D small firms do. This means that the gap between private and social return is larger for large firms, and hence the justification for subsidizing R&D might be highest for very large firms. Government policy in the U.S. used to implicitly recognize this intuition, shuttling R&D funds to the likes of Bell Labs.

All in all an important contribution, though this is by no means the last word on spillovers; I would love to see a paper asking why firms don’t do more R&D given the large private returns we see here (and in many other papers, for that matter). I am also curious how R&D spillovers compare to spillovers from other types of investments. For instance, an investment increasing demand for product X also increases demand for any complementary products, leads to increased revenue that is partially captured by suppliers with some degree of market power, etc. Is R&D really that special compared to other forms of investment? Not clear to me, especially if we are restricting to more applied, or more process-oriented, R&D. At the very least, I don’t know of any good evidence one way or the other.

Final version, Econometrica 2013 (RePEc IDEAS version); the paper essentially requires reading the Appendix in order to understand what is going on.

“Competition, Imitation and Growth with Step-by-Step Innovation,” P. Aghion, C. Harris, P. Howitt, & J. Vickers (2001)

(One quick PSA before I get to today’s paper: if you happen, by chance, to be a graduate student in the social sciences in Toronto, you are more than welcome to attend my PhD seminar in innovation and entrepreneurship at the Rotman school which begins on Wednesday, the 7th. I’ve put together a really wild reading list, so hopefully we’ll get some very productive discussions out of the course. The only prerequisite is that you know some basic game theory, and my number one goal is forcing the economists to read sociology, the sociologists to write formal theory, and the whole lot to understand how many modern topics in innovation have historical antecedents. Think of it as a high-variance cross-disciplinary educational lottery ticket! If interested, email me at kevin.bryanATrotman.utoronto.ca for more details.)

Back to Aghion et al. Let’s kick off 2015 with one of the nicer pieces to come out the ridiculously productive decade or so of theoretical work on growth put together by Philippe Aghion and his coauthors; I wish I could capture the famous alacrity of Aghion’s live presentation of his work, but I fear that’s impossible to do in writing! This paper is based around writing a useful theory to speak to two of the oldest questions in the economics of innovation: is more competition in product markets good or bad for R&D, and is there something strange about giving a firm IP (literally a grant of market power meant to spur innovation via excess rents) at the same time as we enforce antitrust (generally a restriction on market power meant to reduce excess rents)?

Aghion et al come to a few very surprising conclusions. First, the Schumpeterian idea that firms with market power do more R&D is misleading because it ignores the “escape the competition” effect whereby firms have high incentive to innovate when there is a large market that can be captured by doing so. Second, maximizing that “escape the competition” motive may involve making it not too easy to catch up to market technological leaders (by IP or other means). These two theoretical results imply that antitrust (making sure there are a lot of firms competing in a given market, spurring new innovation to take market share from rivals) and IP policy (ensuring that R&D actually needs to be performed in order to gain a lead) are in a sense complements! The fundamental theoretical driver is that the incentive to innovate depends not only on the rents of an innovation, but on the incremental rents of an innovation; if innovators include firms that already active in an industry, policy that makes your current technological state less valuable (because you are in a more competitive market, say) or policy that makes jumping to a better technological state more valuable both increase the size of the incremental rent, and hence the incentive to perform R&D.

Here are the key aspects of a simplified version of the model. An industry is a duopoly where consumers spend exactly 1 dollar per period. The duopolists produce partially substitutable goods, where the more similar the goods the more “product market competition” there is. Each of the duopolists produces their good at a firm-specific cost, and competes in Bertrand with their duopoly rival. At the minimal amount of product market competition, each firm earns constant profit regardless of their cost or their rival’s cost. Firms can invest in R&D which gives some flow probability of lowering their unit cost. Technological laggards sometimes catch up to the unit cost of leaders with exogenous probability; lower IP protection (or more prevalent spillovers) means this probability is higher. We’ll look only at features of this model in the stochastic distribution of technological leadership and lags which is a steady state if there infinite duopolistic industries.

In a model with these features, you always want at least a little competition, essentially for Arrow (1962) reasons: the size of the market is small when market power is large because total unit sales are low, hence the benefit of reducing unit costs is low, hence no one will bother to do any innovation in the limit. More competition can also be good because it increases the probability that two firms are at similar technological levels, in which case each wants to double down on research intensity to gain a lead. At very high levels of competition, the old Schumpeterian story might bind again: goods are so substitutable that R&D to increase rents is pointless since almost all rents are competed away, especially if IP is weak so that rival firms catch up to your unit cost quickly no matter how much R&D you do. What of the optimal level of IP? It’s always best to ensure IP is not too strong, or that spillovers are not too weak, because the benefit of increased R&D effort when firms are at similar technological levels following the spillover exceeds the lost incentive to gain a lead in the first place when IP is not perfectly strong. When markets are really competitive, however, the Schumpeterian insight that some rents need to exist militates in favor of somewhat stronger IP than in less competitive product markets.

Final working paper (RePEc IDEAS) which was published in 2001 in the Review of Economic Studies. This paper is the more detailed one theoretically, but if all of the insight sounds familiar, you may already know the hugely influential follow-up paper by Aghion, Bloom, Blundell, Griffith and Howitt, “Competition and Innovation: An Inverted U Relationship”, published in the QJE in 2005. That paper gives some empirical evidence for the idea that innovation is maximized at intermediate values of product market competition; the Schumpeterian “we need some rents” motive and the “firms innovate to escape competition” motive both play a role. I am actually not a huge fan of this paper – as an empirical matter, I’m unconvinced that most cost-reducing innovation in many industries will never show up in patent statistics (principally for reasons that Eric von Hippel made clear in The Sources of Innovation, which is freely downloadable at that link!). But this is a discussion for another day! One more related paper we have previously discussed is Goettler and Gordon’s 2012 structural work on processor chip innovation at AMD and Intel, which has a very similar within-industry motivation.

“Organizing Venture Capital: The Rise and Demise of American Research and Development Corporation, 1946-1973,” D. Hsu & M. Kenney (2005)

Venture capital financing of innovative firms feels like a new phenomenon, and is clearly of great importance to high tech companies as well as cities that hope to attract these companies. The basic principle involves relatively small numbers of wealthy individuals providing long-term financing to a group of managers who seek out early-stage, unprofitable firms, make an investment (generally equity), and occasionally help actively manage the company.

There are many other ways firms can fund themselves: issuance of equity, investment from friends or family, investment from an existing firm in a spinoff, investment from the saved funds of an individual, or debt loans from a bank, among others. Two questions, then, are immediate: why does anyone fund with VC in the first place, and how did this institutional form come about? VC is strange at first glance: in a stage in which entrepreneur effort is particularly important, why would I write a financing contract which takes away some of the upside of working hard on the part of the entrepreneur by diluting her ownership share? Two things are worth noting. VC rather than debt finance is particularly common when returns are highly skewed – a bank loan can only be repaid with interest, hence will have trouble capturing that upside. Second, early-stage equity finance and active managerial assistance appear to come bundled, hence some finance folks have argued that the moral hazard problem lies both with the entrepreneur, who must be incentivized to work hard, and with the VC firm and their employees, who need the same incentive.

Let’s set aside the question of entrepreneurial finance, and look into history. Though something like venture capital appeared to be important in the Second Industrial Revolution (see, e.g., Lamoreaux et al (2006) on that hub of high-tech, Cleveland!), and it may have existed in a proto-form as early as the 1700s with the English country banks (though I am not totally convinced of that equivalence), the earliest modern VC firm was Boston’s American Research and Development Corporation. The decline of textiles hit New England hard in the 1920s and 1930s. A group of prominent blue bloods, including the President of MIT and the future founder of INSEAD, had discussed the social need for an organization that would fund firms which could potentially lead to new industries, and they believed that despite this social goal, the organization ought be a profit-making concern if it were to be successful in the long run.

After a few false starts, the ARD formed in 1946, a time of widespread belief in the power of R&D following World War II and Vannevar Bush’s famous “Science: the Endless Frontier”. ARD was organized as a closed-end investment trust, which permitted institutional investors to contribute. Investments tended to be solicited, were very likely to be made to New England firms, and were, especially in the first few years, concentrated in R&D intensive companies; local, solicited, R&D heavy investment is even today the most common type of VC. Management was often active, and there are reports of entire management teams being replaced by ARD if they felt the firm was not growing quickly enough.

So why have you never of ARD, then? Two reasons: returns, and organizational structure. ARD’s returns over the 50s and 60s were barely higher, even before fees, than the S&P 500 as a whole. And this overstates things: an investment in Digital Equipment, the pioneering minicomputer company, was responsible for the vast majority of profits. No surprise, then, that even early VCs had highly skewed returns. More problematic was competition. A 1958 law permitted Small Business Investment Corporations (SBICs) to make VC-style investments at favorable tax rates, and the organizational form of limited partnership VC was less constrained by the SEC than a closed-end investment fund. In particular, the partnerships “2 and 20” structure meant that top investment managers could earn much more money at that type of firm than at ARD. One investment manager at ARD put a huge amount of effort into developing a company called Optical Scanning, whose IPO made the founder $10 million. The ARD employee, partially because of SEC regulations, earned a $2000 bonus. By 1973, ARD had been absorbed into another company, and was for all practical purposes defunct.

It’s particularly interesting, though, that the Boston Brahmins were right: VC has been critical in two straight resurgences in the New England economy, the minicomputer cluster of the 1960s, and the more recent Route 128 biotech cluster, both of which were the world’s largest. New England, despite the collapse of textiles, has not gone the way of the rust belt – were it a country, it would be wealthier per capita than all but a couple of microstates. And yet, ARD as a profitmaking enterprise went kaput rather quickly. Yet more evidence of the danger of being a market leader – not only can other firms avoid your mistakes, but they can also take advantage of more advantageous organizational forms and laws that are permitted or created in response to your early success!

Final published version, in Industrial and Corporate Change 2005 (RePEc IDEAS).