Anne Marie Knott joined me on the Acquisition Talk podcast to discuss her book How Innovation Really Works. She is a professor teaching strategy and innovation at Washington University at St. Louis, and is a former researcher at Hughes. Anne Marie tackles a lot of innovation dogma, including whether:
- Smaller companies more effective at R&D than large
- US federal research has actually declined
- Decentralized R&D beats out central labs
- Scientists have picked the low hanging fruit
- Firms are not spending enough on R&D
In the episode, Anne Marie discusses her measure for R&D productivity called the research quotient (RQ). You can think of it as the relationship between R&D spending and revenues for a given firm, controlling for other factors including capital (from the balance sheet), labor (from the number of employees), and a couple other variables. The more responsive revenues are to changes in R&D spending, the more productive a firm’s R&D is relative to other firms.
Anne Marie draws a number of conclusions from looking at firm RQs across sector and time. She finds a 65% decline in average R&D productivity over the past five or so decades. Two-thirds of firms are actually spending too much on R&D, and could increase revenues by cutting R&D. But she’s also seen maximum RQs increasing over time particularly in newly formed industries, so there seems to be this divergence in outcomes. Anne Marie takes an optimistic view. There are ways firms can improve their R&D productivity, lessons that can benefit defense policy makers.
Is science slowing down?
There’s the big debate about whether science is slowing down, pointing at facts like it has taken 24 times more scientists to maintain the same rate of growth in computation represented by Moore’s Law. Each scientist is apparently much less productive. In the podcast, Anne Marie breaks the issue down using Paul Romer’s endogenous growth model. She says that the folks like Chad Jones who argue scientific productivity has slowed down have not demonstrated that:
What are they doing in the paper? They are taking specific examples of technologies or fields and they’re showing decline. Now we know for all time… the norm is that when a new technology comes out, it’s slow to it’s slow to produce anything. Then we’ve got this dramatic rise. And then we get a point of diminishing returns in that technology, which is where we are with respect to with respect to Moore’s law and with respect to all the things that they look at in this paper.
Now, the really hopeful thing is, and, we know this from past sociology of science, et cetera, is that as a new technology becomes exhausted it becomes replaced by another technology, right? So there’s this renewal.
I think the problem of what to measure is a deep problem and can sometimes lead researchers in the wrong direction. The Chad Jones et al paper was in many ways really careful. They weren’t just looking at Moore’s law, but something more general to include performance, cost, and time: the number of computations per second per dollar.
But perhaps what matters becomes the qualitative use of those computations, such as using AI/ML. Or perhaps the measure is no longer relevant with quantum computing, or it provides totally new attributes of performance we care about like encryption. We live in a world where the particulars matter, and cannot be abstracted into general measures without losing something.
DoD as a monopsony
There was an interesting exchange where we talked about the government as a monopsony (single buyer) and what affect that had on the structure of industry. Anne Marie pointed out how, using Clayton Christensen’s framework, firms which serve a monopsony buyer specialize and continue to focus on that customer at the expense of disruptive innovation which usually starts by serving niche needs. Anne Marie recommends that DoD look at commercial equivalents:
The closest commercial equivalent that I can think of is Toyota. So Toyota is a huge customer. They could be the equivalent monopsony for many of their suppliers, taking like 80% of their output. And yet they still continue to maintain multiple suppliers for most of their goods. So it’s interesting that they’re able to do that. But that might be a good model for the DOD.
One could say that the auto market is much bigger than just Toyota, whereas DoD truly is a monopsony. But military service decisions can be quite independent of each other, and American allies can also increase the multi-buyer nature of defense. More of this deliberate competition on the buyer side is probably a good thing. Even though it’s full of politicking and incentives, foreign military sales to me are the best indicator of weapon system value because other countries chose to buy the product over all other alternatives.
Of course, another difference is that Toyota does design and assembly of its cars whereas DoD doesn’t.
Productivity of outsourcing R&D
Here’s Anne Marie Knott discussing her findings from National Science Foundation data on types of company R&D:
Do you want my big finding from that? The output elasticity of outsourced R&D is zero. Whereas on average for internal R and D a 1% increase in R&D gets you 0.1% increase in revenues. For outsourced R&D it gets you no increase in revenues.
I asked her about the implications of outsourcing R&D in government or at major primes. She responded that “you want the body that’s going to benefit most from those spillovers to be conducting the activity.” For example, an engine would likely be subcontracted for an aircraft unless perhaps it’s a new type of engine that requires tight integration with the aircraft.
Decline in development
One thing Anne Marie likes to push back against is the claim that there’s been a decline in federally funded research. She finds no evidence for that in the National Science Foundation data. All of the decline is in development, not research, and that’s contributing to the valley of death problem:
We don’t need more ideas. We don’t need more technology. We need people to actually either want to buy that technology at the end, which gives firms the incentives to make their own development expenditures, or we need to fund more development so that the amount of, the market size that we need isn’t as big as it’s currently to justify investment.
That’s an important point about sources of R&D funding. If companies are going to invest their own private funds into military R&D — or integrating commercial tech into military missions — there needs to be some credible signal that DoD will buy it on the back end if it is superior to existing alternatives. Palantir had a heck of a time with that, going so far as to sue the federal government for not buying a lower-priced commercial alternative.
Currently, funding priorities remain with major programs and their incumbent contractors. The primary reason new entrants had some success was a rapidly growing defense budget over the last few years. Going forward with expectation of flat or declining budgets, it will be hard to increase opportunities for new entrants.
Another problem that needs to be resolved for private funding is that the resulting product cannot be priced at marginal cost (as is customary for government funded R&D). Companies that sink not just R&D, but marketing and G&A, funding for years into a project need to have those costs reimbursed. But they often don’t have the accounting systems to properly track costs. And in any rate, R&D costs are expensed and so wouldn’t be an allowable costs in the price of the final product. By contrast, incumbent contractors have IR&D costs reimbursed by the government through G&A expensed to its large portfolio of existing contracts. That does not set up a fair playing field.
The alternative is that DoD provide more frequent R&D contracts to the new entrants instead of requiring private funding. While there’s been a lot of good intention, this hasn’t been super successful even in a time of growing budgets. With the return of a fiscally constrained environment, that means increasing development comes the expense of other budget titles.
Thanks Anne Marie Knott!
I’d like to thank Anne Marie for joining me on the Acquisition Talk podcast, and to Jaymie for introducing us. Be sure to pick up a copy of her book How Innovation Really Works, and check out her HBR article, The Trillion Dollar R&D Fix. Listen to her on the Behind the Markets podcast and the Edgewise podcast. Watch a short YouTube video and her speaking with Rich Makadok.
Leave a Reply