Do we see knowledge spillovers in real life? How do they change over time? originally appeared on Quora: the place to gain and share knowledge, allowing people to learn from others and understand the world better.
Answer by Evgenii Fadeev, assistant professor at Duke Fuqua, op Quora:
Great questions! Someone reading this answer and (hopefully) learning from it is an example of a knowledge spillover. In general, knowledge spillovers are characterized by a situation in which the creators of knowledge do not have full control over how their knowledge is used and may not always be compensated for it. For example, many answers posted on Quora can lead to knowledge spillovers. The answer to the first question is therefore a resounding yes.
The answer to the second question is more nuanced. While the existence of knowledge spillovers is obvious, quantifying and tracking them over time is challenging. Paul Krugman noted that such spillovers “are invisible; they leave no paper trail by which they can be measured and tracked” (Krugman 1991). However, this inherent invisibility has not deterred researchers!
In 1993, Adam Jaffe, Manuel Trajtenberg, and Rebecca Henderson (Jaffe et al. 1993) introduced the idea of using patent citations as a means to measure knowledge spillovers. The patent system operates on a ‘quid pro quo’ basis, where inventors reveal their ideas to the public in exchange for exclusive rights to their invention. Although replication of the patented idea is prohibited due to the inventor’s exclusive rights, others are free to learn from the revealed idea and develop new technologies based on it. Paul Romer highlighted patents as a source of knowledge spillovers in his influential article on economic growth (Romer 1990):
“[I]Inventors are free to spend time studying the patent application for the widget and gain knowledge that will help in designing a wodget. The inventor of the widget does not have the ability to prevent the inventor of a widget from learning from the design of a widget.”
According to the rules of the patent system, if patent B is developed using knowledge from patent A, it must cite patent A. Jaffe et al. (1993) used patent citations to trace the spread of knowledge spillovers across different geographic locations. Since their seminal paper, the use of patent citations has expanded to various applications, such as estimating models of economic growth and providing policy recommendations.
A recent study by Fadeev (2023) highlights a decrease in the number of companies citing a typical patent, which could be interpreted as a decrease in the spread of knowledge spillovers. However, this interpretation faces a significant problem. The article argues that citations may not accurately reflect knowledge spillovers in the ways we discussed earlier. This shows that the majority of citations come from business partners of the patent holders, often from just one partner. For example, IBM’s patent 5877043 has 218 citations, yet 94% of these come from a single input supplier, Amkor Technology.
The article states that companies often do not disclose all their knowledge in patent applications, leaving certain aspects secret. An example of this is the complexity of the mRNA technology used in COVID-19 vaccines. Simply reading the patent files does not reveal the full technology, as many trade secrets and technical know-how are not included in the patents (e.g. Price II et al. 2020). As a result, it becomes challenging for others to fully learn from the patent without access to these undisclosed secrets. Fadeev (2023) suggests that citations mainly come from a few partners because only they have access to the confidential information surrounding the patent. Although the patent system imposes certain disclosure requirements, companies have developed ways to circumvent these requirements, as noted by some legal scholars (Roin 2005).
Patent citations are not the only method to measure knowledge spillovers. Nicholas Bloom, Mark Schankerman and John Van Reenen (Bloom et al. 2013) have introduced an alternative approach. Imagine that two companies, A and B, each file patents in a similar field. If company B’s productivity responds positively to company A’s R&D expenditure, this could indicate a knowledge spillover from A to B. (2019) showed that the extent of knowledge spillovers, as measured in this way, over the course of time hasn’t changed. However, a recent article by Arqué-Castells & Spulber (2022) challenges this view. Their research shows that the companies commonly used in such estimates (A and B) often enter into licensing agreements. As a result, the observed improvement in B’s productivity may not be an incidental benefit; instead, firm B could pay firm A for this benefit.
In summary, common measures of knowledge spillovers may actually include collaboration and intentional knowledge sharing between business partners. While it is possible that the extent of knowledge spillovers has changed over time, the observed changes in citation patterns are indicative of the decline in intentional collaboration between business partners rather than a decline in knowledge spillovers.
The distinction between unintended spillovers and intentional knowledge sharing has significant implications for several economic issues. For example, the rationale behind R&D subsidies is often anchored in the concept of knowledge spillovers: companies may underinvest in R&D if they cannot fully exploit the value of their inventions. However, if companies can control the spread of their knowledge, for example by selectively sharing trade secrets with certain partners, the effectiveness of R&D subsidies is questioned.
This question originally appeared on Quora – the place to gain and share knowledge, allowing people to learn from others and understand the world better.