Identifying Technology Spillovers and Product Market Rivalry

One of the greatest papers ever written

“Identifying Technology Spillovers and Product Market Rivalry” by Nicholas Bloom, Mark Schankerman, and John Van Reenen is one of the best papers I have ever read, on plausibly the most important topic in the world. In the long-run, growth comes from new ideas. There are limits to how much we can grow simply from accumulating capital; we must invent new technologies, or else have our savings rate be equal to depreciation.

Ideas are special. They are non-rival, meaning that anyone can use the ideas which another person makes without reducing the ability of others to use them. They have a marginal cost of reproduction of approximately zero, and cost some non-zero amount of money to create. In order for anyone to look for an idea, they must be able to get some amount of excess profits. In competition, however, price is driven down to the marginal cost of zero. If competition is too high, then nothing is invented at all. You can find this basic insight in the Aghion-Howitt model of creative destruction, which is a simple two-period model. Its failure mode of no growth is when everyone expects their investments in the first period will be made obsolete by copying in the second period.

Ideas do get invented even without formal protection, though, largely because firms are competing in general equilibrium. It is conceivable that the technology is stolen by their competitors, but that they are then all collectively advantaged compared to other firms in the economy, which are unable to adopt the new technological discoveries. In this case, technological discoveries by competitors have positive spillovers, raising the value of both themselves and other firms. Bloom, Schankerman, and Van Reenen set out to investigate whether these positive spillovers are bigger than the negative spillovers, or in other words, whether society’s spending on research and development is too small or too large.

Their main result is that the amount of spending on research and development is much too small. Technological discoveries by rivals raise the value of competing firms. They argue that the social optimum would be to spend at least twice as much on research and development, which is in line with other estimates. Jones and Williams (1998) estimate that the optimal level of research spending is two to four times higher than observed, and (Ben) Jones and Summers (2020), who take into account more ways in which benefits can spillover, estimate the optimal amount as some very large multiple of our present amount.

Bloom, Schankerman, and Van Reenen have a panel dataset of firms over 20 years, combining Compustat data with records of three million patents. The value of patents being approximated by the number of citations they later get. The authors conceptualize competition as when firms tend to file patents in, and sell products in, the same “space”. This is fairly literal, although it is difficult to visualize when there are many dimensions. There are many bins in which the patent office categorizes patents, and they are able to find the degree of similarity between different bins by how much the patents of firms colocate.

They differentiate product and technological competition, as firms might do similar research but sell very different products. Their favored example is IBM, Apple, Motorola, and Intel, which all research similar things but sell different products. We should expect, and they find that, competing in technology space has positive spillovers, and competing in product market space has negative spillovers.

Meanwhile, it isn’t enough simply to show that the value of firms in competition covary. It is quite possible that they are simply exposed to common shocks, such as an increase in demand for the products which the companies sell, or the technological discoveries being easier to make. They use an instrumental variable approach, utilizing changes in how different governments treat research and development for tax purposes. Yes, these policy changes could be endogenous to the conditions of the particular firms. I agree with the authors that this is unlikely, and that the identification is valid. Their results are robust to whatever they throw at it, including changing datasets and constructing different distance measures.

A striking implication is that the research done by smaller firms tends to have lower spillovers than those done by bigger firms. Smaller firms occupy smaller niches, with fewer other firms close by. I think we do far too much to try and support small businesses in America – no, I am not convinced by the “liquidity constraints” argument in America, not if we can’t even get substantial misallocation from finance in developing countries – so it jumped out to me.

I have not included everything that is in this paper. It is truly exhaustive. Nor is this a complete discussion of the projects which the authors have worked on in this space, for which I highly suggest you check out the new version of Nicholas Bloom’s Wikipedia page. I view encouraging technological growth as the cause which lies above all causes. To me, everyone who is worse off because of ideas which could have been found, but were not, is as harmed as everyone who is deprived of an idea invented. While, on the margin, it may be best to help the world by distributing what we can at little cost to the poorest parts of the world, we must not lose sight of the importance of technology. For every bad thing, and for every quarrel over a limited resource, there is a technology which would make that bitterness a relic of the past. So I encourage you, the reader – go work at those things which you are both capable of, and which will have positive spillovers.