But even this is not definitive. First, a small fraction of firms--even of those with patents--get venture funding, so it is unclear what role patents play. Second, causality is notoriously hard to show, especially where unobserved factors may lead to both patenting and success. Third, timing is also difficult; many have answered my simple data with the argument that it is the funding that causes patenting, and not vice-versa. Fourth (and contrary to the third in a way), signaling theory suggests that the patent (and even the patent application) signals value to investors, regardless of the value of the underlying invention.
Following my last post, I'll discuss here a paper that uses granular application data to get at some causality questions. The paper is The Bright Side of Patents by Joan-Farre Mensa (Harvard Bus. School), Deepak Hegde (NYU Stern School of Business), and Alexander Ljungqvist (NYU Finance Dept.). Here is the abstract:
Motivated by concerns that the patent system is hindering innovation, particularly for small inventors, this study investigates the bright side of patents. We examine whether patents help startups grow and succeed using detailed micro data on all patent applications filed by startups at the U.S. Patent and Trademark Office (USPTO) since 2001 and approved or rejected before 2014. We leverage the fact that patent applications are assigned quasi-randomly to USPTO examiners and instrument for the probability that an application is approved with individual examiners’ historical approval rates. We find that patent approvals help startups create jobs, grow their sales, innovate, and reward their investors. Exogenous delays in the patent examination process significantly reduce firm growth, job creation, and innovation, even when a firm’s patent application is eventually approved. Our results suggest that patents act as a catalyst that sets startups on a growth path by facilitating their access to capital. Proposals for patent reform should consider these benefits of patents alongside their alleged costs.The sample size is large: more than 45,000 companies, which the authors believe constitute all the startups filing for patents during their sample years. For those not steeped in econometric lingo, the PTO examiner "instrument" is a tool that allows the authors to make causal inferences from the data. More on this after the jump.
An instrumental variable allows researchers to overcome the omitted variable bias that I mentioned above (the second problem). Because patents are more or less randomly assigned to examiners, varying grant rates can be used to model patent application quality in a way not previously available. This is going to be a key method of performing research; it is the second paper I know of to use this technique. The authors here follow Sampat & Williams discussed in a prior post here. As the paper explains: "The quasi-random allocation of applications to examiners thus results in the assignment of some applications to examiners who are more likely to grant patents and others to examiners who are less likely to do so. We use this variation in individual examiners’ approval rates to instrument for the probability that a given startup’s first patent application is approved, which allows us to isolate the effect of exogenously granted patent rights on startups’ subsequent growth and success."
The paper also seeks to answer the other three concerns I list above. They consider not just patents, but patent applications (granted or not), and delays between application and patent. The paper also considers timing of venture capital. It finds that 92.5% of the companies had obtained no venture capital before the first office action. Because the office action can come months or years after the application, even fewer were likely to be capitalized prior to application. This answers the old question of whether patent generally follows investment or vice versa.
The authors find that patent approval improves chances of receiving venture capital funding between 1.2 and 2.8 percentage points depending on how long after the patent grant. This sounds small, but venture capitalization hovers around 5% in their data, which means that patent approval can bump the likelihood of funding by 50%. The authors conclude that this is a causal mechanism of reduced information friction (as opposed to, say, signaling) because it is based on approval.
These are just some of the highlights - there is much, much more to this paper, including looking at which firms were more likely to benefit from patents.
It should be noted that the authors had the benefit of using internal USPTO data (and not the publicly available data), which allowed them to see rejections and non-issued, non-published applications. They report that the biggest difference for their set from the public data was patent applications abandoned before first publication (and therefore not public). This was 15% of them (!) While there is no reason to doubt the veracity of the data, it does make replication more difficult.