Category Archives: Sequential Invention

“Collective Invention,” R. Allen (1983)

Who invents? Standard theories usually deal just with R&D-performing firms and individual inventors. But enormous amounts of invention come as a byproduct of everyday firm investment. This type of invention tends to be incremental, and tends to be neither patentable nor held secret by the inventor. Robert Allen, in this famous paper from the early 1980s, refers to such invention as “collective invention.”

Consider the British blast furnace industry in the mid 1800s. There was certainly no meaningful corporate R&D at the time, as the world’s first corporate R&D labs were only just appearing then (in the German chemical industry). Yet the blast furnace industry in Cleveland changed enormously over a couple decades, nearly doubling the height of blast furnaces, and more than doubling temperatures. Such changes were greatly beneficial for reducing fuel consumption.

No single firm made these drastic changes overnight. Rather, furnace heights were increased incrementally by some firms when they built a new factory. Benefits in terms of lower fuel use were then made publicly available through personal correspondence, industry gatherings, and journal publications. Two factors were critical in this shift. First, the industry was rapidly adding capital. If a new plant is being built, experimentation has low costs: the cost of adding a foot to the chimney is that efficiency might be harmed slightly, and the benefit is that efficiency may be helped slightly. When an industry is not accumulating capital, this sort of minor experimentation is much more costly, since the only experimentation involves building an entirely new, not-yet-necessary factory. The second critical factor is some reason to avoid secrecy. In the blast furnaces, secrecy was more or less impossible. Builders and workers were frequently moving from plant to plant, and could simply tell their new employer what they learned. Since information is leaking out anyway, it may be an equilibrium to share information in the hopes that others will have useful information to share with me: work by von Hippel, previously discussed here models this sort of sharing in more detail.

The reason we tend to ignore this type of public, incremental innovation is because of a bias, in popular culture and in policy, toward big technological advances. A paper of mine, which I hope to have ready to share here soon, argues that the patent bias toward technological achievement and away from incentivizing the nexus of inventions which lead to a commercially viable product can be seriously harmful. The importance of minor inventions is more than the importance of the famous ones, they shout from the rooftops!

An interesting update of Allen would be in the context of China. To the extent that industries accumulating capital quickly throw off, as a byproduct, incremental inventions, there can be rapidly increasing cost efficiency in even developed industries when some shock causes the industry to switch to a new region with little capital. Peter Hessler, National Geographic’s man in China and a great chronicler of that nation, tells a great story about technology transfer and incremental growth in his book Country Driving. I’m also curious to see how one would distinguish learning-by-doing in aggregate statistics from learning-by-sharing at the plant level.….collinvent.pdf (Final JEBO copy. The only nongated version I can find is this Google cached article.)

“Sequential Innovation, Patents and Innovation,” J. Bessen & E. Maskin (2009)

Back in the 70s and 80s, you could study the economics of invention as if it were static. X is out there and we want to invent it, so how can we incentivize people to do the R&D? The famous tradeoff in patenting is found in the literature: if I give a patent, I generate a deadweight loss triangle from monopoly (plus the “inefficiency rectangle” if you like your firm dynamics with an evolutionary flavor), but if I don’t give a patent, no one has any reason to invest in the fixed cost of R&D. The second part of that argument weakens a bit if firms can keep their invention from being copied for a short while while they extract rents – it takes time to reverse engineer – but the basic tradeoff is still there. (What about the tradeoff between secrecy and disclosure, you may be wondering; my reading of the literature is that the requirement that patents disclose how to make an invention is essentially a meaningless restriction, since firms can choose to rely on trade secrets and since, even when they patent, firms can be obtuse enough in describing certain technical aspects of the invention that it may not be possible for competitors to imitate.)

Recent work on patents, though, takes a much more dynamic attitude toward invention: “If I have seen farther, it is because I have stood…” and all that. When invention is cumulative, the relevant tradeoffs are much more subtle. For instance, if you need A to make B, and both A and B have fixed costs, then who should get the patent to B? The firm who did the R&D work on A that made it possible? The firm that actually invented B? Bessen and Maskin follow this line of reasoning to a surprising conclusion: in some industries, not only do patents decrease social welfare, but they aren’t even in the interest of any of the firms in that industry! I also like this paper as well because it’s nice evidence that (at least some) economists could care less about credentials: James Bessen is a lecturer at a law school with no formal graduate degree of any kind, while Eric Maskin is a Nobel prize-winning economic theorist at Princeton’s Institute for Advanced Study. This paper actually needs them both. Bessen has done a lot of really interesting work on innovation policy, while the result I’ll show shortly relies on some non-obvious game-theoretic reasoning about imperfect information, an area right in Maskin’s wheelhouse.

The basic model is simple. There are two firms, who can either do research for free or at cost c. If one firm does research, the next invention in a sequential line is invented with probability p. If both firms do research, the next invention is found with probability q, where p<q<2p. Each invention in the line has identical value v, drawn from a known distribution, where v represents the incremental value of each new invention. After an invention is made, any firm can imitate; after innovation, any firm can imitate costlessly, in which case each firm gets payoff sv, where s<1/2 is a "share" of the full value v. Patents make imitation impossible. The social planner, of course, would have firms always do research when it is free, and have either one or both of the firms work if they have a cost of research. The choice of zero, one or two firms working when research is not free depends on a cutoff rule on the value of each invention.

Sequential invention makes firms more likely to innovate in the absence of patents than static innovation. The reason is simple: though I am still upset that an imitator may take some of the revenue from an invention I invest in, I also know that if I don’t make that invention, the next product in the sequential line will never be possible. So I need less value from the present invention to make R&D worthwhile. And patenting can make us worse off. The problem is asymmetric information: I don’t know whether the other firm is a high cost or a low cost of R&D firm. When I invent product 1 in a line, the model assumes my patent can keep you from inventing and selling product 2. Of course, I can offer you a license. If I think you have cost 0 R&D, I will set the license equal to exactly the surplus you expect to gain from inventing the next product in a sequence. Then we’ll both work on the next product, and no matter what, I will reap all the surplus. But it may be costly for you to do R&D. In order to get you to do work on the next product, then, I would offer you a licensing fee that just lets you cover the cost of your R&D, but otherwise gives me all the surplus. It turns out that in a range of values, the firm that invents product 1 would rather just get the higher licensing fee always from the zero cost guys, meaning that it never licenses to the high cost guys, even when product values are high enough that social welfare would want to have both firms working on research, and where both firms would indeed do research if there were no patents. Ex-ante, meaning before firms have any patents yet, the firms may well prefer the world without patents to the world with patents for this reason (though, of course, once they have a patent, the firm can only be made worse off by losing it).

One final result, relevant to software and other industries: if the value of inventions is large, and if s is close to one-half (meaning that imitation is not “too” dissipating), then in both the patent and no patent case, firms are better off having a competitor who may imitate/become a licensee. Having a competitor means, basically, the industry is able to invent more quickly. Even when Lotus and Microsoft keep ripping each other off, the fact that they are both inventing new features expands the size of the market, and makes both better off than in the monopolistic world where only one firm did all the R&D.

File these results under “patents can sometimes harm” and “someone tell the law & econ guys that informational frictions matter”. (2006 Working Paper. The final version is in RAND 2009. If you have access, read the RAND and not the working papers – this paper floated around for a decade or so before publication, and some early versions have a number of mathematical mistakes.)

%d bloggers like this: