What happens when you get more Ph.D.s?

Following the fall of the USSR, hundreds of world class mathematicians emigrated to the USA. Intuitively, this should have made American mathematics stronger. Did it?

Borjas and Doran examined the problem. Their starting point was the realization that the expertise of Soviet mathematicians differed from the expertise of American mathematicians. For historical reasons, these two communities did different mathematics. So we can examine side-by-side areas of mathematics impacted by the arrival of the new mathematicians and areas unaffected. Their verdict is that the influx of Soviet mathematicians was unhelpful to academic production:

It is (…) difficult to find convincing quantitative evidence that there was an improvement in the overall “output” of either the pre-­‐existing American workforce or of that community combined with the Soviet émigrés.

What happened? Basically, the number of research jobs, including tenure-track positions, is mostly independent from the supply of Ph.D.s: Harvard will not hire more professors next year even if all of the Ph.D.s in the world move to Boston. The number of new hires depends of factors such as government funding for research and the number of undergraduate students in research universities. So the new researchers just took the jobs that would have gone to other researchers. Young American researchers had to drop out of research for lack of a position.

This study offers an important policy lesson. Training more Ph.D.s in some targeted areas might fail to improve research output in these areas. In this instance, supply-side economics fails. It might be preferable to create new research jobs instead and attract the Ph.D.s with better salaries.

For example, imagine that the government wants to help cancer research. Providing more scholarships to graduate students either directly, or through grants to professors, sounds sensible. However, it might not have the desired effect at all, according to this study. It would be preferable to use the money to create more research jobs pertaining to cancer research.

Credit: Thanks to Larry Larbear for the reference to this study.

Daniel Lemire, "What happens when you get more Ph.D.s?," in Daniel Lemire's blog, February 20, 2012.

Published by

Daniel Lemire

A computer science professor at the University of Quebec (TELUQ).

7 thoughts on “What happens when you get more Ph.D.s?”

  1. @Jeremy

    I think you can ask the question without any specific measure of “research output”.

    How do you speed up research on cancer? Train more researchers, hoping they will generate research positions, or create more research positions, thinking that it will entice people to train for these positions?

    One approach pulls another pushes. One generates demand, the other generates supply.

    Take this question, apply whatever metric you like and let us see what the correct answer is!

  2. Very interesting research, thanks for the pointer.

    I disagree with your conclusion, though:
    “Training more Ph.D.s in some targeted areas might fail to improve research output in these areas. In this instance, supply-side economics fails.”

    The Soviet emigres published and were cited more than there peers pre-collapse, as economics would predict: Eliminating positions in the USSR pushed out marginal producers, leaving a more talented cohort.

    Where we diverge from classical economic thinking is that Americans in “Soviet” fields declined. However, the paper suggests older, tenured professors were less affected: it was the young mathematicians that got edged out. Instead of competing on any metric of quality, Americans competed on age, and this seemed to produce a lower quality cohort.

    Thus I don’t think you can conclude that overproducing young mathematicians will lower the output quality: My read of the paper is that it’s specifically an influx of older, well-established applicants edging out the young that causes the problem.

  3. @Paul

    It should not be surprising that economics fails to predict outcome in this type of market, because it is very far from a free market. This means that it can be very hard (at least to me) to make sense of it.

    I’m not going to argue specific points here, but I’ll just say that there is not a lot of difference between adding already trained foreign Ph.D.s or training new Ph.D.s.

  4. @Daniel

    Agreed on the trickiness of predicting non-free markets. And the USSR situation is doubly tricky, because it was a discontinuity: a one-time sudden large influx of talent.

    Your last point is actually what I find most interesting about this topic. What is the difference between training a new Ph.D. and hiring an older, established researcher? My own impression of the world is that very different dynamics can exist when you pull in all senior, established people, vs. a start-up of exclusively young, vs. a mixed group, especially in a very long term endeavor like science.

  5. @Paul

    • Very often, government programs have a limited duration. For example, the program can decide to sustain research on computer security or AI for the next 5 years, and then it stops. So there are many discontinuities.
    • Academic departments tend to have a fixed number of positions allocated by the University. Sometimes they can hire at any rank. However, they are usually encouraged (and often required) to hire younger professors, mostly because older professors are more expensive. That is, there is open age discrimination in academia against older people. As far as I know, there is no planning to ensure that you have a mix of young and older professors except maybe that if several professors are due to retire, the department may get to open a few positions earlier to avoid a big and sudden turnover. Of course, government research laboratories have more flexibility when they hire. They also don’t have tenure so it is possible, in theory, to let go older and less productive researchers. I would think that many research laboratories would try harder than academic departments to strike a balance between older and younger researchers. Note also that professors are typically expected to do research more or less independently. In research laboratories, co-workers are more likely to be collaborators.

Leave a Reply

Your email address will not be published.

You may subscribe to this blog by email.