Part of the “canon” for researchers in innovation is that patenting is only really used by a small subset of firms in a small subset of industries (large pharma firms, in particular). In many other industries, many inventions use no real protection at all, or else attempt to rely on trade secrets laws – Google was well known for its small number of patents until the Motorola purchase. Richard Levin’s “Yale Survey” in the 1980s, and the “Carnegie Mellon Survey” by Cohen, Nelson and Walsh talked to R&D heads to figure out what was going on, and essentially came back with the same conclusion: complementary capabilities, speed of research and many other factors tend to be more important than patent protection in expropriating gains from R&D. This is particularly worrying since a huge majority of research and policy concerning invention has to do with formal protection like patents: if a researcher makes some claim about the innovation process and provides only evidence from patent data, have we really learned much?
But how robust is this finding historically? It may be that we are in a particularly patent-heavy, or patent-light, era. Do the results that pharma and chemicals rely heavily on patents remain even in the past? There are good theoretical reasons to believe so, but little empirical data. Tom Nicholas reminds us that the U.S. National Research Council, beginning in 1921, has run periodic surveys sent to firms “which by a liberal interpretation do any research work.” The survey was also run every few years through the 1980s, so there’s clearly much more to mine from that dataset. NRC gives you the number of research workers, firm size, location, etc., and the firm name can be matched (in what is surely an arduous process) to firm names in patent data. About a third of US patents from the 1920s and 1930s were from firms in the NRC data. We can then examine propensity to patent overall, by industry, for publicly traded firms, and many other nice cross-sections.
It appears that firms with research capabilities (sometimes only one R&D worker) had a higher propensity to patent in the 20s and 30s than firms today: 59% of all the NRC firms, and 88% of the publicly traded NRC firms, were granted a patent between 1921 and 1938. The article attempt to sell this as a considerable difference today, but the data from NRC surveys and more modern surveys are so different in range that I would find it difficult to make such a strong assertion: nonetheless, the data do suggest that patenting was probably more important for a subset of firms in the early part of the century than today.
Two other results are worth mentioning. First, patenting behavior, even within industries, appears quite similar between the 1920s and 1930s; if you know the work of Alexander Field, who suggests that the 1930s were super-important for new inventions, this might be surprising. A Field-style result is clear in terms of the size of research departments in the surveys, though: during the 1930s, firms doing research more than doubled their research staffs, both at the mean and the median. This does not appear to be spurious data, as the number of firms surveyed is roughly similar at the start and end of the 30s. Second, a Levin style result on differences in propensity to patent across industry also held in the 20s and 30s. Indeed, industries that say they rely heavily on patenting in modern surveys also had a high propensity to patent in the 20s and 30s, with medical technology being a big exception. I am sure some historian must have looked at the professionalization of medical research during this interval; a pointer to a citation here would be greatly appreciated!
(A final, important, note: I find it strange to deal with this question while omitting any discussion of the fact that patents, or patents per capita, or patents per firm, are overall much, much higher today than in the past. So clearly somebody is patenting much more. The fact that the dependent variable here is a binary “did you or didn’t you patent?” is misleading in some ways. Even after reading this paper, I would be shocked to find out that propensity to patent, properly measured, was lower today than in the past.)
http://people.hbs.edu/tnicholas/RD_pat.pdf (Final version; published in the Journal of Economic History December 2011)
A Fine Theorem 
Shouldn’t it be “DID R&D FIRMS USE TO PATENT?”?
I usually don’t even read articles where the title is grammatically incorrect.
I think it’s fine…use to is probably better because of the “did” but both are perfectly standard English now.