The Brookings patent report is highly questionable

Brookings has a new report out by Jonathan Rothwell, José Lobo, Deborah Strumsky, and Mark Muro that “examines the importance of patents as a measure of invention to economic growth and explores why some areas are more inventive than others.” (p. 4) Since I doubt that non-molecule patents have a substantial effect on growth, I was curious to examine the paper’s methodology. So I skimmed through the study, which referred me to a technical appendix, which referred me to the authors' working paper on SSRN.

The authors are basically regressing log output per worker on 10-year-lagged measures of patenting in a fixed effects model using metropolitan areas in the United States.

$$\ln y_{i,t} = \begin{multline}c + \beta_{1} \ln ( patenting_{i,t-10}) + \beta_{2} \ln ( Population_{i,t-10}) + \\ \beta_{3} \ln ( y_{i,t-10}) + \beta_{4} \ln ( \text{predicted productivity}_{i,t-10}) + \\ \beta_{5} \ln ( \text{educational attainment}_{i,t-10}) + \text{place and dummy variables} + \varepsilon_{i,t}\end{multline}$$

The model is structured in this relatively standard way to reduce endogeneity—there might be more patents filed where labor productivity is highest, rather than higher labor productivity where the most patents are filed. And if the only concern were reverse causality, then it would be a good way to study the question of patents and innovation.

The authors find positive coefficients on the patenting variables and conclude that patents drive economic growth both in local areas and in general.

This report documents how a strong national innovation system plays out across a dispersed array of U.S. metropolitan areas, contributing to economic growth in both local places and across a large and diverse country.

Clear in these pages is the continued vibrancy of the U.S. innovation as well as the general utility of the nation’s patenting system. (p. 28, emphasis added)

These conclusions are unwarranted given the model and findings expressed in the paper. To see that this is the case, assume temporarily that patents do nothing to incentivize real innovation, and that they merely transfer wealth from consumers at large to the patent holder through firm profits. If this were the case, then we would find that measured output per worker was higher in metropolitan areas with more patents—exactly what the authors found!—because they are gaining profits at the expense of consumers in metropolitan areas with fewer patents. In other words, the authors could be laboring under a fallacy of composition. Just because patents enrich the MSAs that generate them doesn’t mean that they are a source of prosperity for the nation as a whole or that they increase social welfare.

Alternatively, assume temporarily that patents do nothing to incentivize real innovation, but that firms that produce valuable innovations must defensively patent them to avoid being taken to court for using their own inventions. If this were the case, then patents would correlate with real innovation, and therefore with output per worker, but they would not cause an increase in productivity. In addition, at least some of the measured increase in output would come from an influx of highly-paid intellectual property attorneys, which by assumption does not represent real added productivity. Note that the top-patenting MSA in the study is Silicon Valley, the part of the country where people are most concerned about defensive patenting. But the word “defensive” does not appear even one time in the report, the appendix, or the working paper.

The authors have done nothing to identify the effect of patents on productivity, which is to say, nothing to rule out either of the possible assumptions above. They are simply relying on the assumption that more patents means more innovation.

This flaw in the paper makes all of their policy conclusions suspect. For example, if patents represent a mere transfer, then encouraging patent-generating institutions is socially destructive. It might nevertheless be rational for a single MSA to encourage such institutions, because residents of the MSA would enrich themselves at the expense of other MSAs. In this case, we should adopt federal policies to discourage patent-generating institutions. If patents merely correlate with innovation due to defensive patenting in some domains, then the U.S. patent system is not working as intended, which is again the opposite of what the authors conclude.

On point, the Winter 2013 Journal of Economic Perspectives is out this week, featuring a four-paper symposium on patents. The lead article is by Boldrin and Levine, entitled “The Case Against Patents.” Here is the first paragraph:

The case against patents can be summarized briefly: there is no empirical evidence that they serve to increase innovation and productivity, unless productivity is identified with the number of patents awarded—which, as evidence shows, has no correlation with measured productivity. This disconnect is at the root of what is called the “patent puzzle”: in spite of the enormous increase in the number of patents and in the strength of their legal protection, the US economy has seen neither a dramatic acceleration in the rate of technological progress nor a major increase in the levels of research and development expenditure.

Petra Moser’s article does a historical comparison of countries with strong and weak patent laws and concludes:

Overall, the weight of the existing historical evidence suggests that patent policies, which grant strong intellectual property rights to early generations of inventors, may discourage innovation. On the contrary, policies that encourage the diffusion of ideas and modify patent laws to facilitate entry and encourage competition may be an effective mechanism to encourage innovation. (emphasis in original)

I hope that policymakers don’t rely on Brookings’s strong reputation and infer that our patent system is the strong engine of economic growth that Rothwell et al. suggest it is.