We investigate the effect of mobility of R&D workers on the total patenting activity of their employers. Our study documents how mobile workers affect the patenting activity of the firm they join and the firm they leave. The effect of labor mobility is strongest if workers join from patent-active firms. We also find evidence of a positive feedback effect on the former employer's patenting from workers who have left for another patent-active firm. Summing up the effects of joining and leaving workers, we show that labor mobility increases the total innovative activity of the new and the old employer. Our study which is based on the population of R&D active Danish firms observed between 1999 and 2004 thus provides firm-level support for the notion that labor mobility stimulates overall innovation of a country or region due to knowledge transfer.The abstract doesn't really do the paper justice. My comments are after the jump.
The first thing to say about this study is that the methodology is improved over similar studies. Most studies look at patent filings to find mobile workers. But this leaves out mobile workers that don't file patents at their new jobs, and thus creates a selection effect. This paper, on the other hand, links two separate sets of data: mobility of all workers and patenting data. This is a big deal, because it eliminates many typical objections to studies of this kind.
The study has three primary findings:
1. Firms that hire workers have a substantially higher growth rate in patenting.
2. Firms that lose workers also have a substantially higher growth rate in patenting.
3. Point 2 holds only when the mobile worker joins a high patenting firm. If the mobile worker's new firm doesn't patent, then the losing firm doesn't see growth either.
Points 2 and 3 are the real surprises here for worker mobility. Theoretical and empirical studies have long shown that there is knowledge spillover that aids new firms. But few studies have so strongly supported a view that firms losing workers get a spillover effect from former employees. This is an important and interesting finding that should be further studied.
But it also leads to my skepticism. First, as a general matter, I'm always wary of studies that equate patenting with R&D. While the two are no doubt linked, it is not 1:1, nor is it necessarily a two-way street. Spending on R&D might lead to more patenting, but it's not clear to me that more patenting always means more R&D spending or even more innovation generally. Then again, the authors here find on average a 30% increase in patent applications. That's nothing to sneeze at.
More important, there may simply be other factors that lead to increased patenting associated with mobility. We might, for example, expect to see hot markets in growing areas. The more new growth, the more need for new labor, and the more likely old firms will patent even if people leave. In other words, the effect may be that R&D growth causes mobility, and not vice versa. The feature of the findings that most bolsters my skepticism is that it is the patenting activity of the new firm, not the mobile individual, that correlates with high patenting in the old firm. I have trouble buying a story that a worker leaving suddenly causes full spillover by all the employees of the new firm to the one new person's old firm without some further evidence.
The authors answer this skepticism in a couple ways. First, they show correlation in patenting growth with citation to the patents between the firms, which implies knowledge spillover and feedback in both directions. This is helpful, but it may be that firms in the same area cite each other; we also don't know if the citations come from patent examiners or from the inventors. Second, they show that the results hold even if they correct for industry growth and lag general labor mobility (to add a time element to their correlation). This, I think, is a more persuasive answer to my concerns. I look forward to future research that teases these effects out.