“Entry and Patenting in the Software Industry,” A. Cockburn & M. MacGarvie (2010)

My understanding is that lawyers and politicians both generally believe patents to be necessary for inducing innovation. I do not believe that most economists believe the same to be true. See, for example, Levine and Boldrin’s book “Against Intellectual Monopoly” for many examples, or articles like Cohen et al (2000) which survey firms in many industries and find that the vast majority rely on methods other than patents to recover the fixed costs of R&D, such as pure first mover advantage, trade secrets, etc. Given limited benefits, then, you might worry about the costs of patents. In particular, many firms in the software industry, Adobe most famously, were vigorously against allowing software patents. The worry was that any new piece of software necessarily uses so many ideas and components of previous software that a “patent thicket” can develop whereby it is essentially impossible for firms to enter a market because the cost of licensing from hundreds of firms is very high. This thicket then allows incumbents to extract rents from monopoly beyond that which are the traditional raison d’etre of patents themselves.

Cockburn and MacGarvie use a great dataset of firms involved in 27 areas of software from 1990 to 2004 to estimate empirically whether patent thickets exist. They find that markets with more relevant patents held by incumbents see substantially less entry (elasticity of -.3 to -.8). This is true even when controlling for firm age, experience in related markets, the number of incumbents, and the quality of patents as measured by forward citations. When potential entrants hold relevant patents of their own, the anticompetitive effect is diminished, reflecting the ability to “trade licenses” which is common in software. There is a control for possible endogeneity whereby firms take out patents in order to head off potential future competition: software patents became available at different times in different submarkets. For instance, business algorithms were only patentable (in some ways) after the State Street case in 1998, whereas software for running machines was available for the entire sample period.

Beyond this paper, there is a running theme in empirical work on IP: it is very, very difficult to show that patents, copyrights, and other government granted monopolies are somehow good for welfare. This ought give us pause, since I can think of no (non-minor) way in which IP has been deliberately weakened in developed countries over the past half century, but many ways in which it has been strengthened; the only case I know of is the 1996 limitation of doctor liability which essentially makes patents on surgical techniques much less valuable. If I had to guess, I would guess that the welfare benefits of moving from the current regime to the optimal IP regime are more substantial than the welfare benefits of moving from the current trade regime to the optimal one. So why then do economists write a ton about trade rules but so little about WIPO, the PTO, and the deleterious effects widespread in innovation policy?

(Two more quick remarks. First, I really wish economists would restrict the use of the term “intellectual property” to the legal monopolies granted by copyright, patents, and the like, reserving an alternative term – I like “knowledge goods”, but “zero marginal cost goods” is fine – for the actual creative works themselves. The Harvard law professor William Fisher III has a very nice article which documents how the term “intellectual property” was rare until the 1970s, and was deliberately introduced in order to induce linguistic equivalence between knowledge and physical property. Second, economists really need to get the idea of the Romantic Age “heroic lone inventor” out of their intuition. There is little evidence that there ever was such a thing, and among historic discourse about invention, worldwide, only 19th and 20th century Europe and America seem to devote substantial time to creative works of geniuses. Alford’s book “To steal is an elegant offense” discusses this concept as it concerns Chinese laws on knowledge goods.)

http://www.econ.kuleuven.be/msi/_docs/workshops/2009-12-10.pdf (Working paper – final version forthcoming shortly in Management Science)


5 thoughts on ““Entry and Patenting in the Software Industry,” A. Cockburn & M. MacGarvie (2010)

  1. Robbie Clarken says:

    Nice article. One thing I have never understood regarding intellectual property laws is that when copyright terms are extended, the extension is applied to existing works. For example, the 1998 Copyright Term Extension Act applies to works published back in 1923.

    Even if you believe that longer copyright terms will promote more creative works, I see no benefit to consumers in extending the monopoly on creative works already in existence. Am I missing something?

  2. Francisco says:

    Forward citations as a proxy for patent quality? A patent doesn’t receive a forward citation for its quality (novelty and non-obviousness), but for its relevance to the patentability of another patent.

    Revocation rate is better in my opinion.

    • afinetheorem says:

      How would you propose using revocation rate to measure the value of an individual patent?

      I think Bronwyn Hall’s research is the reason you see forward citations so often as a measure of value. She has a lot of evidence that, for instance, firm value is affected by the quality of the patent portfolio as measured by forward citations – http://www.nber.org/papers/w13426.

  3. Francisco says:

    As a disclaimer, I’m not an economist myself, but a former patent examiner 🙂

    A working definition of patent quality could be:

    “A high-quality patent meets all statutory and regulatory requirements and will withstand legal challenges by competitors or other third parties.”


    Following that definition, if a patent is revoked it means that it had low quality.

    On the other hand, the fact that a patent *withstands* an opposition/revocation procedure signals that has *value* and *quality* (a competitor unsuccessfully tried to invalidate the patent)

    Regarding citations, a patent (“the first patent”) receives a citation in another patent (“the second patent”) if the first patent is relevant to the patent examiner for assessing the novelty and obviousness of the second patent. Basically, the first patent is cited in the second patent if the first patent takes away the novelty of the second patent. As a result and in my opinion, the citation itself has not a direct connection with the *quality* or *value* of the first patent. That’s the reason for my skepticism. Besides this paper also alerts of the inconvenients of forward citations:


Comments are closed.

%d bloggers like this: