Patent statistics as economic indicators: a survey
Griliches offers a very broad review on patents as economic indicators. Stretching back to seminal work by Schmookler in his book on Invention and economic growth, patents have been a core means of measuring innovation econometrically. As a major participant in this literature, Griliches attempts to provide a broad survey on the use of patents in econometric analysis. Griliches take-away is (perhaps unsurprisingly) that, "patent statistics are interesting despite all the difficulties that arise in their use and interpretation."
Griliches first gives a detailed description both in patents and into patent statistics. In particular, he explains that patents and R&D expenditure are both indicators of innovative activity but that they don't necessary track each other. Griliches arguments that there are two major problems with using patents for economic analysis:
- Classification of patents into meaningful categories that match industries: patents are categorized but in a very function way that is a poor match for industry (e.g., a holy water dispenser and a water pistol are in the same the category)
- Intrinsic variability: Some patents are extremely valuable while most are not.
Griliches discusses at some length the important question of what exactly patents are an indicator of. He argues that economists want patents to be an indicator of output indicator of inventive activity but that patents are much more of an input -- the major observation that Schmookler made in his own work. Griliches describes, in depth and with figures, a very clear model of knowledge production and places patents as an output of knowledge production.
Griliches then offers a detailed description of the relationship between patents and R&D and, in particular, discusses in-depth the question of whether there are diminishing returns to R&D expenditure and whether (and how) patents can help us get a hold of the question. Griliches concludes that, "at the moment, the evidence is suggestive [of diminishing returns] but no conclusive." Even this statement is tempered in the later part of his paper where he concludes that the question remains unresolved.
The survey continues to a discussion of patent value (mean of $600,000 but median of $25,000) and means of measuring patents through abandonment in mandatory renewal processes. He also discusses patents and stock market values (with very tentative conclusions). He talks about patent spillovers (citing Jaffe's work).
The end of the paper is focused on using aggregate patent data to talk about business and bureaucratic cycles. In particular, he focuses on a dip in the number of patents filed in 70s and 80s. Griliches convincingly argues that this was a reflection of a decrease in funding in the patent office which led to an inability to grant patents as quickly and was not representative of a decrease in innovative activity.
In summary, Griliches concludes with a reflection on the "frustration with our inability to really answer the 'big' questions" but with excitement about the possibilities of using patent data -- especially by doing more than simple counting and looking at citations, geography, and more.
Theoretical and practical relevance:
Griliches 1990 has been cited nearly 3,000 times, primarily in the economics and innovation literatures. It is very broad in scope and, as a result, has been cited widely by authors working patents and economics more broadly.