A Thought Experiment by Jones and Summers
One good way to think of R&D. My experience is that research and development follow two different statistics and are hard to think of as one continuous process. Development is more predictable - outcomes are more gaussian. Research is unpredictable and driven by power-law statistics. There is an average growth rate since one only has a finite amount of historical data. But the average is driven by the big power-law returns not the smaller development returns.
If the returns of R&D are so high the why don't we pour the whole budget into R&D? Power-law returns are driven by two primary dimensions: quality of each investment and the number of investments. If you make more quality investments your chances of a big power-law return goes up, not the case if you make more non quality investments. There comes a point where investing more money it R&D is funding lower quality opportunities and the costs out weight the probability of hitting a power-law winner.
There is a saying one cannot pick winners - true; however, I think you can spot the bottom ~20-30% non winners with fairly high probability.
Great questions posed and insightful conclusions.
This is super clear and interesting, thank you!
Another reason policymakers might not want to invest in R&D is that there can be negative externalities to R&D, both to society as a whole (e.g., AI could be used in divisive or oppressive ways) or to specific segments who will be negatively effected by innovation (e.g., beef farmers might not want lab-based meat R&D). Especially in the latter case, lobbying by entrenched interests could really increase the downside policymakers see in prioritizing certain types of R&D
The Jones and Summer paper seems vulnerable to several criticisms, chiefly their definition of R&D which is skewed towards government definitions of R&D, their limits to technological R&D and their omission of other kinds of thinking, research and experiments which fall outside their approach. Oh, and their reliance on GDP. And the apparent belief that the R&D/growth relationship can be analysed within one country. Their idea would have worked up to about 1990. Not now.
The RoI argument is simple and sound. But yeah, what's counted in stats? Not tinkering, deployment/transition, art/design. Moreover, the value created isn't always economic. More-moreover, cultural change can alter how much value we place on tech. Let's not limit "R&D" to S&E.