Saturday, March 7, 2009

Optimal Number of PhD Students?

I want to conquer the world. Should I take on more PhD students? I am too busy. Should I drop some of them? These are fundamental questions in the life of a professor.

The students, of course, have similar questions but much less control over this: "My advisor has no time for me, and they are taking ANOTHER student?" or "Man, I wish I wasn't the only student this crazy guy has; that way he'd leave me alone for longer."

So, I decided to start a poll here to determine what people think is the optimal number of PhD students that an advisor should have. (I have six.)

Most of you read this blog via some sort of blog reader like Google Reader. Unfortunately, you'll have to visit the blog directly to vote.

Addendum: After a week of voting, ~300 people responded as follows:
0 students: 3%
1 student: 4%
2-4 students: 63%
5-8 students: 20%
9 or more: 7%

Apparently I have too many students :)


  1. Lets assume it takes 5 years to complete a PhD. If you had 2-4 students, then chances are that you could end up with a time when most of your students graduate and you need to cultivate a new crop of students from scratch.

    The best research happens when you can function as a group, with varying levels of expertise and value. A new PhD student can benefit from helping a close-to-graduating student and the benefit to the latter is obvious.

    So smoothing out the curve and planning for the long-term viability and continuity of group research should be factored into the ideal number of students.

    In my opinion, assuming a 5-year cycle, that ideal number of students is about 5-7 (which would account for variations that in some year you may not add any, in some year get only one new student, and other years there may be two great students who are dying to work with you).

  2. I tend to like the idea of "one per year." One problem with this is how much it costs to support 5 students. At CMU, I have to raise about $75k per PhD student per year. With 5 students (assuming no fellowships), that's $375k per year every year.

  3. Luis, you need to spend less time watching TV and more time writing grants. Or to get your students to spend less time watching TV and more time writing grants for you.

    That said, I think 2-4 is about right, especially if not all of your work is with them. If you do no work on your own, then maybe 5 or 6, but I feel like even that starts stretching you to the point of being a professional manager, not a researcher.

  4. As always, it depends. Different types of research require different amounts of attention for each student. And different students require different amounts of attention.

    I would say you should be willing to spend at least one hour per week with each student. Some weeks it will be more, but it should always be at least one. If you can't afford that, then you have too many. I had an adviser who scheduled half-hour weekly meetings, but that person was rarely available for even that meeting. There were around 11 students in my group.

    I like the comment about working as a team. If your students can work well together, then that helps both you and your students. I had a bad experience in a group with little to no communication between students.

    I'm currently in a group with up to 5 people who work with my adviser (depending on how you look at it). It's a lot better than it was before.

  5. as a Phd student, I wish I was in a group of 3-5, preferably all of us at different stages. Enough to have a team, but not too many to get together easily. (and at my school, almost all the students are funded through TAs.)

  6. "At CMU, I have to raise about $75k per PhD student per year."
    Is this true???

  7. It never ceases to amaze me how my colleagues in fields like bioengineering manage to run a lab with 8+ PhD students *and* 5-6 postdocs. A lot of labs have more postdocs than students. Given that a postdoc costs 2x a grad student, they must write a LOT of grant proposals.

  8. Another way to look at things, is what students become after graduating. After all, the reason to undertake students is supposed to be less about getting cheap labor force, and more about forming future academics.

    If most PhD students hope to get an academic job afterward, and if most professors retire after 30 years of research (hence creating one new position), and if most professors graduate on average one student every year after the five first ones, then the number of academics should increase geometrically with a factor of 25-1 every 30 years: won't work till we get out of this litle crowded blue sphere, and even then...

    Maybe the right amount of students to undertake is how many you can reasonably hope to get into a good job after graduating, whether in academia or not...

  9. "At CMU, I have to raise about $75k per PhD student per year." Is this true???

    100% true.

  10. Given that a postdoc costs 2x a grad student, they must write a LOT of grant proposals.

    This is not quite true. At least in most private universities, postdocs cost about the same as students. For example, the cost of a student at CMU is Stipend + Tuition + Overhead. Tuition in private universities ranges between 30k and 40k per year. On the other hand, the cost of a postdoc is just Salary + Overhead.

    I do agree with your amazement that people in biology-related fields have huge armies working for them because they have to raise millions of dollars per year to support this.

  11. Maybe the right amount of students to undertake is how many you can reasonably hope to get into a good job after graduating, whether in academia or not...

    This is an excellent point!

  12. I have done observation in biomedical engineering labs, and yes, they have many many students and postdocs. One thought is that given the expense of running a lab, students and postdocs might be relatively cheap.

    The answer I'm more certain of is that these labs were much more focused than those I observed in computer science. That is, everyone in the lab was doing something relatively similar, and the lab culture was such that the more senior students helped the younger students learn what they needed to (how to culture cells, etc.) The PI of the lab was on travel around 150-200 days a year. The lab had to be self-sufficient.

    In contrast, in many CS labs each graduate student is working on something very different, and needs mentoring specifically from the advisor. The older students can't help them much, and this puts a time strain on the advisor.

    My solution is to try to have my lab so focused that I can offload some of the mentoring to the more senior students. I organize them somewhat hierarchically. Undergrads work for Ph.D. students, and junior undergrads work for the more senior undergrads. If I can do this, I can deal with many more Ph.D. students.

    That said, I'm still an assistant prof. We'll see how it works out over time.

  13. I'd agree with Jeremy that (once you have enough data to estimate it) the key limitation should be how many students will get good jobs. One crucial barrier is when you have two students graduating in the same year. You might get lucky, with the two students having totally different interests or career goals. However, there's a good chance they'll end up competing against each other for the same jobs. If one student is much stronger, it requires careful coordination of where they apply (to keep the stronger student from crowding out the weaker one for every job). Even if the students are comparable, your letter may be used as a tie-breaker and small phrases can be blown out of proportion to make everyone prefer the same student.

    The hard part is that it's not a stable equilibrium. Once someone has established a reputation as a good advisor, more students will want to work with them than they can handle, and there's a near-infinite supply of qualified applicants to grad school. I don't know how to solve this optimization problem: is it better to give a great experience to a select group of students, or an OK experience to a much larger group?

  14. My respect for my advisor has doubled since I read your post. $75K is a huge amount of money.

  15. If you get enough students (they don't all have to be your direct Ph.D. supervisees), they can start taking care of each other. So the answer varies dramatically if you're a prof in a much smaller pond than CMU.

    What happens if you get a scholarship student? What if the students work as RAs instead of for you?

    When I was a prof at CMU, the cost was a little less, but the problem was the same in CS. So we ran the computational linguistics program out of the Philosophy Dept because (a) that's where us linguists with teaching commitments like me and Chris Manning hung out, and (b) we only had to pay CMU 20% of the students' tuition (it was 100% in CS). The whole reason the comp ling program collapsed in Philosophy (it was reborn as the LTI) is that the Philosophy department couldn't tolerate the risk and CMU was run almost entirely at a departmental/college level.

    In practice, we always had too few students for the amount of money raised by folks in computer science and at private companies.

    Part of the reason that bio labs can be big is that biology pay rates (at least for post-docs) tend to be very very low.

  16. If the students are good, then you could potentially handle an arbitrary number. If the students are bad, then even one feels like too much.

  17. Luis, do you factor in undergraduates into your maximum number of advisees - I.e. 3 undergraduates = 1 PhD student ? Or do they fall into different categories?

  18. I don't know if I factor undergrads :) On the one hand, they require time, but on the other I don't have to pay for them. Also, sometimes graduate students can help advising undergrads.

  19. I heard from a friend of mine that a Prof. manages 100 Ph.D students successfully under him. In some countries we would need to do that. Any information / experience on this management paradigm?

  20. An advisor's capacity to assist PhD students is key and speak volumes as to his (or her) dedication to the field. The correct "As many as he (or she) can competently handle".

  21. "My respect for my advisor has doubled since I read your post. $75K is a huge amount of money."

    PhDs & Professors are also expected to respect Undergrads and Masters students for indirectly funding their research work (Yes Indirectly!!)

    We should not ignore undergrads or masters students. They are the most valuable resources to any professor. With little guidance and directions they are as productive as any other PhD student; in some cases, even better than PhD students.

    The worst part of the story is that many universities do not value undergrad/MS students' contribution to research. Undergrads/MS research assistants earn $9 per hour (95% of such RA positions are Part-time). These students contribute same as any other PhD student, but they get just $9/hr.. A MS student at CMU pays $75k tuition from his own pocket... Certainly, it is very "profitable business model" for universities, but what about those young brilliant minds?

    Does anyone care? :-(