• John
  • Felde
  • University of Maryland
  • USA

Latest Posts

  • USLHC
  • USLHC
  • USA

  • James
  • Doherty
  • Open University
  • United Kingdom

Latest Posts

  • Andrea
  • Signori
  • Nikhef
  • Netherlands

Latest Posts

  • CERN
  • Geneva
  • Switzerland

Latest Posts

  • Aidan
  • Randle-Conde
  • Université Libre de Bruxelles
  • Belgium

Latest Posts

  • TRIUMF
  • Vancouver, BC
  • Canada

Latest Posts

  • Laura
  • Gladstone
  • MIT
  • USA

Latest Posts

  • Steven
  • Goldfarb
  • University of Michigan

Latest Posts

  • Fermilab
  • Batavia, IL
  • USA

Latest Posts

  • Seth
  • Zenz
  • Imperial College London
  • UK

Latest Posts

  • Nhan
  • Tran
  • Fermilab
  • USA

Latest Posts

  • Alex
  • Millar
  • University of Melbourne
  • Australia

Latest Posts

  • Ken
  • Bloom
  • USLHC
  • USA

Latest Posts

Flip Tanedo | USLHC | USA

View Blog | Read Bio

The game theory of the postdoc market: why today is a stressful day

If you know any particle theory graduate students who have applied for postdoctoral positions this year, today might be an especially stressful time for them. While this is still a couple of years away for me, I’ve been watching with fascination as many of my friends and colleagues go through this process. [Note that while this holds primarily for the particle theory community, I imagine a similar process occurs for other fields.]

A post-doc is a 2 to 3 year academic position in between graduate school and an assistant professorship. It’s a time to develop one’s independent research without the teaching requirements of a faculty position. Postdoc applications are typically sent out in fall and offers start trickling back in December.

In a month or so things will be all sorted out, but early January is when applicants get to see how the sausage is made, so to speak. The process can be a bit rough primarily because of the small size of specialized research communities like particle theory. Unlike undergrad admissions where there are thousands of accepted applicants for a flexible number of positions, most research groups can only hire a few postdocs (often just one) and have no wiggle room. This means that if there’s only funding for one job, a group cannot afford to make multiple offers at a time because it would be a disaster if more than one person accepted. Making postdoc offers becomes a non-trivial multiple-stage process that requires some strategy.

To keep make the playing field fair to the applicants, just about all departments have agreed to the particle theory postdoc deadline agreement, which states that no offer can be made that requires a response before January 7th. (That’s tomorrow!) This is effectively a deadline for the first round of offers and protects applicants from offers that try to force a commitment before other universities can make offers.

But now there’s still a lot of ‘game theory’ involved in the process. As is often the case in theoretical physics, a simple “toy model” is sufficient to demonstrate the phenomenon. Suppose you have postdoc applicants Alice, Ben, and Chris,  and departments at X, Y, Z. For simplicity let’s assume that these lists are ordered by status: A > B > C and X > Y > Z. Thus universities want to hire A while applicants want to go to University X. Let me pause and say that this is a gross simplification: usually rankings depend on particular research interests and can be confounded by all sorts of external factors (e.g. spouses).

So here are some examples of what could happen:

  • Scenario 1: Every department makes an offer to Alice, so Ben and Chris have to wait until after the first round deadline to get offers. They’re sweating bullets because they don’t know if they’ll have a job next year. Alice ends up going to X, and then by a similar process, Ben ends up at University Y in the second round, and finally Chris ends up at University Z in the third round.
  • Scenario 2a: Now consider the case where University Z wises up a little. They know that applicant Alice is out of their range, so instead of making a first round offer to her, they go straight to Ben. Now in the first round Alice can choose between Universities X and Y, but Ben has an offer from University Z and no longer has to worry about getting a job. Now Ben has until the first round deadline to accept Z‘s offer. He thinks that maybe he’s good enough for University Y, but he can’t be sure and he doesn’t want to gamble with his career so he accepts Z. In this scenario, then, the third-ranked university was able to snag the second-best applicant. We’ll say that University Y “fell through the cracks.”
  • Scenario 2b: could have turned out differently: Maybe Alice decides immediately that she wants to go to X. Then she can inform Y ahead of time (though she’s not obligated to do so) that she’s taking another offer, and Y can move on to make an offer (essentially a second-round offer) to Ben before the first-round deadline. In this case we end up with the same matching as Scenario 1. Note that it is often the case that it’s very difficult to choose between X and Y, so Alice ends up using the entire first-round period to mull over her choices and this secnario has to happen at the last minute (e.g. the day before the deadline — today!), or might not happen at all.
  • Scenario 3: another twist: This time, maybe University Y decides that it wants Chris (for one of multiple valid reasons) and University Z has  ‘resigned’ itself to being the third-best choice, so it cuts straight to the chase and also makes an offer to Chris. (e.g. so that the fourth university, W, doesn’t steal him like scenario 2a.) Now Alice goes to X, and Chris can choose between Y and Z, but Ben has no first round offer, even though Z would have been happy to have him. In this case the second-best applicant has fallen through the cracks. [e.g. maybe he is forced to accept a first round offer from the fourth best university.]

Now you can see how this can get a lot more complicated. There are maybe a hundred or so applicants and maybe a few dozen universities. You can expect that top-tier universities will target top-tier applicants and so forth, but it’s not clear where the boundaries are and it’s not clear who falls through a crack (as Ben did in scenario 2a). Maybe Alice does string theory while Ben does LHC physics and the top universities are currently looking for LHC physicists. Or maybe Alice has a spouse who refuses to live in X, so she won’t consider their offer. Maybe a university has multiple postdoc positions so they can afford to be more ambitious. At the end of the day, it gets really complicated.

What practically happens is that there are a lot of telephone calls in December as faculty members in charge of their local postdoc search call up their colleagues to ask about their students who are applying. (Like I said, it’s a small community.) These are usually to further assess how well that student would fit as a postdoc and how likely it is that the student would accept an offer were it made. Then by late December the first round offers are made and the lucky students have until January 7th to accept an offer. Often departments will inform students if they’re short-listed, which partially means that they’re waiting to see how the market turns out before committing to making an offer. When a department hears from an applicant that they’ll politely turn down an offer, they can immediately go to the next person on their list, hoping that this person hasn’t already accepted elsewhere. As you can imagine, January 6th can be a bit of scramble as departments try to make offers before applicants are forced to commit to an existing offer. It’s been suggested that proper etiquette requires one to inform institutions as soon as possible about one’s decisions, or even that it is only reasonable to hold on to no more than two offers, but currently such things are completely voluntary and nobody wants to decline an offer unless they’re 200% sure that something won’t change their mind down the road.

Since a solid postdoc is one of the keys to proceeding onward to a faculty job, this can be an extremely tense time for young scientists. (It’s actually a very good thing that they have the winter break to be with friends and family during this period.)

There have been two recent developments in the postdoc market that have changed the game a little bit. The first one is the existence of an unofficial postdoc “gossip” page where postdoc offers can be self-reported. It is the only way to get a semblance of the status of the postdoc market. I have to admit that I keep up with this the same way that basketball fans keep up with the NBA draft.

The second development was just rolled out this year, a centralized system called AcademicJobsOnline to organize postdoc applications (“officially” endorsed by the HEP community). Like the Common Application for undergrad admissions, this makes it much easier for recommenders to upload one letter (instead of many dozens) and for an applicant to avoid filling out the same data on different forms. I’ve heard unofficially that this has led to a big increase in the number of applications to some institutions, which is something of a minor annoyance to prestigious institution but can be a big boon for ‘diamond in the rough’ departments in lesser-known universities.

As many of my colleagues bemoan the uncertainty of January 6th, there is another conversation which keeps popping up every year: why can’t we do things the way the medical doctors do? The National Resident Matching Program is a ‘magical computer program’ that matches med students to 25,000 residency positions. The system is a bit mysterious, but it pairs up students with a residency in a way that somehow maximizes the desires of the medical program and the applicant (after interviews). The general statement from the people who wrote the common deadline agreement is that the NRMP’s large administrative overhead makes it difficult to implement in academia.

While it’s always true that it’s hard to shift to a new system, there is certainly some merit to having some kind of matching algorithm where ranked preferences from institutions and applicants can be taken into account to make postdoc pairings. Because the theoretical physics postdoc community is so much smaller than the medical resident community (by factors of tens of thousands), I suspect the overhead can be significantly trimmed. The program could be written to simulate the process as it exists today, with institutions making offers and applicants choosing between them based on the preference lists. Multiple rounds of matching can be done automatically without the threat of “falling through the cracks.” This way applicants don’t have to feel like they’re having their choice taken away from them. Unlike the NRMP, preference lists and the computer code can be made to be completely transparent to ensure that there are no secret back-room deals. In fact, now that applications are beginning to be centralized through AcademicJobsOnline, there already exists a natural framework to implement such an automated system.

I’m a bit naive about these things, but the actual implementation seems simple: Applicant submit an ordered list of jobs and, afterward, institutions submit an ordered list of people they’d like to hire. Then what follows is an optimization algorithm that can be tuned depending how one wants to break “ties.” This requires some choices that the community has to agree upon, but it is still more reliable than of whether or not someone officially declines an offer before the Jan 7 deadline.

Every year this discussion must pop up at informal lunchtime at different universities, and every year people start out being very skeptical about radical changes… until January 6th, when the stress of the current postdoc market catches up to applicants and they worry that they might fall through the cracks (e.g. University Y in scenario 2a or Ben in scenario 3 above) and they wish that a more certain system were in place. Then a few months later everything works out, people are excited about their new jobs, and everybody forgets about the postdoc market again. Hopefully the community can work something out that avoids the shortcomings of the current system.

This post is dedicated to all of my friends who are holding their breaths for postdoc offers on this “day-before-Jan 7.” Good luck, everyone!

Flip, US/LHC

Share

Tags: , ,