Elaborating and making more explicit some of the other models here, I propose this alternative explanation which I don’t think you’ve ruled out (and which I’m sympathetic to).
1. PhDs have no causal impact on research productivity.
2. PhDs, for the sort of person who does groundbreaking impressive original research, have substantial positive expected personal value. You get social legibility and status, you get higher pay, and it is a chance to do funded research for a few years while building useful connections. “PhDs are fun” is not a popular view in 2020, but I’m enjoying mine.
Now, I’d be surprised if this strong model is entirely true. The social legibility and status make it easier to spend more time on research, the presence potentially pushes people away from less interesting but more profitable problems, etc. But your current analysis does not allow us to distinguish between this model and its inverse (The entire observed effect of PhDs on research productivity is causal, and none of it is omitted variable bias).
I’m aware this is correlative, and I tried to address this in my post. My model is that PhDs are some mix of useful and attractive to unproven geniuses.
The reason they’re useful and helpful is an interesting issue of its own. My main goal, though, was to rule out the idea that they’re not actually useful or attractive as a platform for innovative work relative to either saving and self-funding or going straight into industry.
Elaborating and making more explicit some of the other models here, I propose this alternative explanation which I don’t think you’ve ruled out (and which I’m sympathetic to).
1. PhDs have no causal impact on research productivity.
2. PhDs, for the sort of person who does groundbreaking impressive original research, have substantial positive expected personal value. You get social legibility and status, you get higher pay, and it is a chance to do funded research for a few years while building useful connections. “PhDs are fun” is not a popular view in 2020, but I’m enjoying mine.
Now, I’d be surprised if this strong model is entirely true. The social legibility and status make it easier to spend more time on research, the presence potentially pushes people away from less interesting but more profitable problems, etc. But your current analysis does not allow us to distinguish between this model and its inverse (The entire observed effect of PhDs on research productivity is causal, and none of it is omitted variable bias).
I’m aware this is correlative, and I tried to address this in my post. My model is that PhDs are some mix of useful and attractive to unproven geniuses.
The reason they’re useful and helpful is an interesting issue of its own. My main goal, though, was to rule out the idea that they’re not actually useful or attractive as a platform for innovative work relative to either saving and self-funding or going straight into industry.