3 ms·

> Is the US spending billions to educate them One PhD student costs a professor about $100K - $200K per year. A third of that is stipend and the other two thir

> Is the US spending billions to educate them One PhD student costs a professor about $100K - $200K per year. A third of that is stipend and the other two thirds is... well, in CS I'm not even sure. A good bit of administrative welfare AFAICT. (Universities don't pay FICA taxes for grad students and grad student healthcare premiums are actually more of a subsidy for the risk pool of the aging faculty/admins than an actually half-reasonable benefit for the grad students.) The US graduates 2K PhDs per year, of which >60% are foreign nationals. So we spend something roughly like (2000 * .6 people) * (150000 $/person) = $180 million per year. So a couple billion per decade. Now... we do get folks with CS BS degrees for 1-5 years at $30K/yr, which is a fucking steal, and at least a great universities they are often from the upper quarter or so of the undergrad distribution. So maybe the students are also paying in the form of subsidized labor. But we do spend hundreds of millions per year in any case. All of that said, as long as universities are staffing mathematics departments with shit paid ad juncts it's kind of hard to argue that our country is failing to compete due to a lack of phds...

abeppu 5y ago
Is looking at the per-student cost and multiplying by the number of students a good way to evaluate the systemic cost? Or are there other factors separate from the marginal cost to a professor of an additional grad student which should be included (e.g. are there institutional-level resources which are significantly supporting this workforce?)
- throwawaygh 5y ago
  > Or are there other factors separate from the marginal cost to a professor of an additional grad student which should be included (e.g. are there institutional-level resources which are significantly supporting this workforce?) No. At least, not in CS. What sorts if things would you be referring to? Compute? Honestly you're better off with a trivial cloud grant than trying to use the slurm cluster with a dozen ancient gpus for hundreds of faculty to share. Office space? Your average dilapidated cube farm is LUXURY compared to the sardine situation grad students put up with. Robotics hardware? Doesn't really scale like that. And... those are the only three things needed for like 90% of CS research. In CS, precious few dollars other than the stipend go toward anything useful. As far as a faculty member is concerns, that money money disappears into a pit. (What's going on is that the PhD stipend is fixed by the funding agency, and the university squeezes as much as possible away from the professor/student. Just armies upon armies of assistant deans doing not much of anything while the grad students toil away doing the only useful work in the place for rent and pizza.)
  - abeppu 5y ago
    Ohh I just accepted your previous statement of 2k/yr overall, but of course you're only talking about CS. > The US graduates 2K PhDs per year, of which >60% are foreign nationals. It looks like across fields, it's ~55k PhDs/yr, so the "Billions" quote was probably right.
    - throwawaygh 5y ago
      Yeah, sorry, I meant just CS. (Other fields get a lot more complicated pretty quickly, but most everything boils down to "stem phd students get paid an almost living wage".)

abeppu 5y ago

Is looking at the per-student cost and multiplying by the number of students a good way to evaluate the systemic cost? Or are there other factors separate from the marginal cost to a professor of an additional grad student which should be included (e.g. are there institutional-level resources which are significantly supporting this workforce?)

throwawaygh 5y ago
> Or are there other factors separate from the marginal cost to a professor of an additional grad student which should be included (e.g. are there institutional-level resources which are significantly supporting this workforce?) No. At least, not in CS. What sorts if things would you be referring to? Compute? Honestly you're better off with a trivial cloud grant than trying to use the slurm cluster with a dozen ancient gpus for hundreds of faculty to share. Office space? Your average dilapidated cube farm is LUXURY compared to the sardine situation grad students put up with. Robotics hardware? Doesn't really scale like that. And... those are the only three things needed for like 90% of CS research. In CS, precious few dollars other than the stipend go toward anything useful. As far as a faculty member is concerns, that money money disappears into a pit. (What's going on is that the PhD stipend is fixed by the funding agency, and the university squeezes as much as possible away from the professor/student. Just armies upon armies of assistant deans doing not much of anything while the grad students toil away doing the only useful work in the place for rent and pizza.)
- abeppu 5y ago
  Ohh I just accepted your previous statement of 2k/yr overall, but of course you're only talking about CS. > The US graduates 2K PhDs per year, of which >60% are foreign nationals. It looks like across fields, it's ~55k PhDs/yr, so the "Billions" quote was probably right.
  - throwawaygh 5y ago
    Yeah, sorry, I meant just CS. (Other fields get a lot more complicated pretty quickly, but most everything boils down to "stem phd students get paid an almost living wage".)