# More on matches

EDIT: Just so people don’t get the wrong idea, I’ll mention that I’m not suggesting this because of some personally traumatic experience I’ve had with job searching;  in fact, the last difficult career decision I made was when I was 18.  I just think it’s an idea worth considering, and one worth hearing other people’s input on.  END EDIT.

As I’ve read more about the medical resident match, I’ve recently become a lot more convinced that a match for mathematics jobs makes a lot of sense. Fundamentally, the point of a match world be that schools and candidates wouldn’t ever have to play mind games with each other. Everyone would just make a list, and the computer would make sense of them.

I feel the benefits of such a scheme are obvious. What about the objections?

• How random is it? In practice, not very. While one can come up with pairs of preference lists with lots of very different stable matchings, these seem to be very unrealistic. Roth and Peranson did an analysis of the medical resident match, and found that roughly 1 in a 1000 residents had different matches that were stably possible.
• What about couples? Luckily, the medical resident match has already dealt with this problem. They use a modified match algorithm which allows couples to rank pairs of positions, including “unmatched.” One could imagine having this with collaborators, as well as romantic partners.
• What about negotiating salaries? This is one can of worms that the medical resident program has stayed out of is the question of differential salaries; each resident program has a single salary and benefit package which they publish beforehand. But I don’t think this is strictly necessary. Schools could publish the possible different combinations of salary they would consider and people could rank those, so you could decide
school X at \$70,000 a year > school Y at \$70,000 a year > school X at \$65,000 a year
and so on. Schools would simply have separate preference lists for the 65K and 70K positions. Similarly, I don’t think it would damage the algorithm much for a school to say “we would hire 3 people if they were all on this list, but only 2 otherwise.” This would just involve replacing a 3 opening job with 3 different jobs, one of which has a shorter preference list.
• Won’t it be extremely difficult for schools and candidates to rank that much stuff? Yeah, you got me there. That wouldn’t be fun. On the other hand, each choice would be much lower stakes. Hopefully it would be easier to do interdepartmental horse trading when the question is “how many spots do we move so-and-so up” rather than being offered a binary choice “do we give a offer to X or Y?” In fact, one could imagine simply having a departmental vote using STV.
• Wouldn’t university bureaucracies be completely resistant to such an idea? Oh, Boldface, why must you burst my bubble?

## 60 thoughts on “More on matches”

1. anon says:

Regarding issue 2, do couples in which one partner is a non-mathematician (and possibly even a non-academic) exist in your universe?

2. Now that there has been some experience with the math jobs wiki, I like to emphasize that it would be even better to have a reformed job market that somehow obviated the wiki. Should it be a match that looks like the medical school match? Should it perhaps look like Operation Postes in France? I don’t really know what is feasible or desirable. But I am conviced that a more centralized job market of some suitable kind would be a big improvement.

Certainly MathJobs.Org is already a big improvement over what came before, the printed “AMS Cover Sheet” stuffed into 75 envelopes. Even that old vehicle is better in some ways than what still exists in some disciplines other than mathematics; even though it wasn’t electronic, it was at least uniform.

3. @1: actually I’m a member of such a relationship. For such people, I think the existence of a match makes little difference. It would be nice if it did, but I don’t regard the lack as a flaw.

4. anon says:

My concern is that a member of such a relationship would likely need to make their list before having a definite idea of their partner’s ability to find suitable employment in the various places that s/he is considering, and then would be *contractually obligated* to accept whichever single place s/he gets matched to. This is different (and, at least from the standpoint of the selfish interests of this hypothetical couple, worse) than the status quo, where there is at least the possibility of turning down an offer if employment in the area doesn’t work out for the partner.

5. Allen Knutson says:

Is there any reasonable way to make this work without having (almost) all schools participate? I’m imagining something like mathjobs, where schools can slowly succumb to the steamroller, and it’s still useful before they’ve done so.

6. anon: I am also in such a relationship, and what I found was that the current system isn’t that much better. I was able to research the general job markets in the cities we might move to, and get promises from schools of various amounts of help, but the deadlines were still too early to actually lock in a nonacademic job for my wife by time I had to make a decision. There’s no reason that I couldn’t do that sort of research earlier in the year.

I would strongly prefer a system such as Ben is suggesting. I was applying to very good schools, all of which I would have been happy to work for, but it still drove me nuts that I had to try convince every school they were my first choice. (Even though I would be perfectly glad to be a third or tenth choice, and in some cases believed that I was.) I’m sure if I were a department chair, it would drive me equally nuts to try to convince every candidate that they were my favorite. Under the stable marriage algorithm, there is no incentive to mislead.

I’d like to hear how more mathematicians would feel about losing the ability to negotiate their contract. Would people feel cheated if they lost the ability to fight for course reductions and raises based on each of their accomplishments? I wouldn’t, and I wonder whether a lot of us misevaluate our ability to negotiate a contract, compared to a chair who has negotiated dozens.

7. My wife is also an academic, but in a different field, and it doesn’t sound like this would work at all for such situations. Our job cycles are roughly the same and under the current system there’s a decent chance of deciding at the same time. Getting such two-body problems to work is already more than difficult enough!

I’m sure there are also certain types of non-academic jobs for which something similar is true.

8. Alexander Woo says:

A comment from the other world…

A matching system really would not make much of a difference for liberal arts colleges, remembering that the most important qualification is the ability to teach our students.

The only way to really evaluate our candidates is to bring them to campus and have them give a colloquium and actually talk to our students. (While they are here, we might as well have most of the faculty talk to them one-on-one.)

We have the budget to bring three people to campus. A fourth might be possible if it’s cheap.

If all three or four turn us down, the search is failed. (My (2-year) job and the search this year is the result of last year’s failed search, so it does happen.)

So under a matching system, we would have exactly three candidates on our list (or two in the rare event that one of the candidates invited to campus really was not acceptable).

Some of our candidates are in teaching positions at other colleges, so they can only take so much time out of their jobs to interview.

The incentive to mislead in this system is at the Joint Meetings Interview, since the scarce resource is the on-campus interview, not the offer.

I don’t think the situation is all that different for most liberal arts colleges.

9. David writes: it still drove me nuts that I had to try convince every school they were my first choice. (Even though I would be perfectly glad to be a third or tenth choice, and in some cases believed that I was.) I’m sure if I were a department chair, it would drive me equally nuts to try to convince every candidate that they were my favorite. Under the stable marriage algorithm, there is no incentive to mislead.

This seems strange to me; I was on the job market not so many years before you (and interviewing at many of the same places) and I never felt called upon to give department X the impression that I was their first choice. I certainly think they wanted the impression that I would seriously consider accepting an offer from them — but that was true in every case, as long as you’re conscientious about withdrawing yourself from jobs you know you wouldn’t take. On the other side, I was on the hiring committee last year and never felt the need to convey to a candidate that they were our first choice (even when it was true) — all I felt we had to convey was that we would be very happy if the candidate were to join our department, which is again clearly true of anyone who makes it to the interview stage.

As for the match: let me play boldface and suggest at least one more problem, which is that it would require people to make judgments in advance about their preferences among departments. I was lucky enough to interview in several places, and it’s certainly the case that the process of visiting departments, meeting potential colleagues, looking at places I might live, etc., changed my internal rankings a lot.

10. which is that it would require people to make judgments in advance about their preferences among departments.

No, it wouldn’t. In the medical resident match, the rankings are done after interviews.

11. Alex-

Are you saying that liberal arts schools never miss out on candidates that preferred them, but got offered a job at another school first? Because to me, that’s a big advantage of the match.

12. Alexander Woo says:

Ben,

I am saying that happens a lot less than one might think, because schools interview (on campus) relatively few candidates and most candidates interview (on campus) at relatively few schools.

Schools interviewing particularly late do run into a problem. But these schools might be missing the match anyways.

What might be more useful is a matching system for on campus interviews, run after the Joint Meetings.

13. Andy-

It’s true that some people trying to find a job with a partner would be worse off, but I expect a number of them would be better off. I think you may be underplaying some of the advantages of a match. One of these is that you know exactly when you will find out about your job. You would never have to worry about whether you might get a better offer later.

I mean when you say “Our job cycles are roughly the same” you mean that you will get job offers which will stay open for a couple of weeks at some point during the same 4 month period. It’s true that you might get all your job offers at the same time, but maybe one of you will get offered a job in city X in January, and the other one will get an offer there in April.

In part, how well this works depends on how accommodating other departments are. If they’re willing to make offers timed so that your partner knows about them before the match, and can decide afterwards then things will work great. If not, there certainly would be problems.

14. Mark Meckes says:

For what it’s worth, the medical resident match is, as far as I know, only applied to match medical students to medical residencies. That would seem to correspond to a similar system for math grad students seeking postdoc positions. And that highlights one of the difficulties in importing the system: medical students all do residencies (I think; feel free to correct me if I’m wrong), whereas many math grad students do postdocs but many others go directly to tenure-track positions.

15. Hidden Costs says:

Medical students are expected to pay to travel to their interviews: this can cost ~\$10k, and students often take out additional loans to cover these costs.

16. Hidden Costs-

This has nothing to do with the match. It’s a separate issue that seems very unlikely to change in mathematics. Medical students are expected to pay for visits to medical schools before matriculating and to pay tuition, both of which mathematicians are very rarely required to.

17. Ben,

I’m having trouble seeing many benefits. This whole “you know exactly when you will find out about your job” business sounds great, but it seems essentially to be a convenience; likewise for any other benefit I can come up with. The “you would never have to worry about whether you might get a better offer later” isn’t some magic: it’s that the computer more or less plays out various rounds of offers and you simply don’t get any time to negotiate or wait for a partner to hear something during any of the rounds.

In terms of any sort of concurrent decision with a partner: as you point out, these are quite difficult anyways. A match system seems to make this strictly worse. For simplicity let’s say both our fields use this system, though the case where one does is not that different (it’s some intermediate case). If our matches resolve on separate days, then it’s as though we applied in different years and the best we can do is guess for the first person and have the second person have their top preference be where the first person got a job. (If they resolve on the same day, then we’re completely blind.)

It seems couples currently have various ways to win (e.g. concurrent offers from an acceptable place, departments that are willing to work together, etc) and the match system reduces the winning options to one.

Perhaps you can explain what these other benefits are that I’m missing. Perhaps also you can give an example of how this system would help couples in a different sort of situation (you made some claim here that a match system would be worse for me but better for certain others).

In any case, the issue of spouses each with careers (academic or not) has become quite an important one to many people. I can’t see moving to a system that is any worse for this than it is now.

18. The big problem with match systems is how they interface with other deadline based systems, especially for people considering more than one possible job, or people with a partner in another deadline based career track.

As another example making many of the same points as Andy is making, but without any academics, I know friends where one was applying for jobs in the justice dept. and regular lawyering jobs, while the other was starting a residency. The problem is that the deadlines go: make your list, hear back from justice, get matched. Their priorities were more along the lines of: 1) Justice+residency near DC, 2) Pick the best residency and find a lawyering job wherever that happens to be. There’s no way to deal with this well in a match system that doesn’t include all “young professional jobs with strict hiring cycles” together.

Matching systems are totally fine (and possibly better) for: two academics both in the same field using the same matching system, or one academic and someone in a profession that has a less structured hiring cycle.

19. As for the issue of needing to communicate to schools that they would be a high choice, my impression is like David’s, that this is a significant problem.

With this round of postdocs at least once I ran into the problem of not being offered a job because I’d chosen Columbia for my NSF so they didn’t think they’d be able to hire me. As it turns out they were right, I did get an NSF. But if I hadn’t gotten an NSF, my impression is that Columbia would not have hired me, and then I’d have missed out on an excellent opportunity just because it wasn’t my first choice.

The existence of the NSF requires you to put on paper your first choice. For tenure-track there’s no such requirement. Without the existence of the NSF I would have felt a lot of pressure to give multiple places the impression that they were my first choice. There seemed to be a strong incentive to be less than totally truthful.

Perhaps this year’s job market had a different feel though because of these tough economics times?

20. One revision: I was a bit hasty dismissing all benefits of match systems. Essentially they enforce some regulation (e.g. no rogue University can go and make offers with really early deadlines). Perhaps this issue is better addressed head-on: have universities agree to stricter regulations than they do now. This is something that could happen incrementally (like use of MathJobs or, more relevantly, the fairly wide agreement to have the first deadline be after the NSF gets back for postdocs).

With appropriate regulations/agreements, my statement that a match system does little more than play out various rounds of offers (but too quickly) could be much closer to the truth.

21. Resistant To Change says:

I agree with Andy: a matching system makes complicated negotiations such as those regarding two-body problems almost impossible. Basically a matching system has you attempt to program in advance how to respond in any possible situation. It seems clear to me that this will not work as effectively as normal negotiations, where 1) you can negotiate more options that cannot be pre-programmed in advance, and 2) you only have to consider the options you have or might have (e.g. for scheduling interviews) instead of trying to consider everything conceivable.

I don’t really see any “obvious” benefits. Yes, in the current system it takes months for employers and employees to get paired up. So what? Proper negotiations require time.

Also, the whole idea doesn’t make much sense (as Mark Meckes points out) because you would have to pick a particular set of math jobs to match, you couldn’t possibly match all different math jobs at once. (The idea of matching, say, tenured jobs is simply absurd. I think the only reason why one might consider matching postdocs is because these are temporary so the consequences of a poor match won’t last forever.) But it is not clear what this set of jobs would be and how one would apply both to jobs in this set and not in this set.

22. Anonymous says:

Perhaps this issue is better addressed head-on: have universities agree to stricter regulations than they do now.

I think the current system (with a common deadline for first round postdoc offers) is already adequate, with one exception, namely the few universities that refuse to go along. For example, Caltech, at least as of a few years ago.

The difficulty is that there’s a big recruiting advantage from having an early deadline. In some ways it’s amazing that so many institutions are willing to agree to the common deadline.

If enough grad students care, they could eliminate this difficulty by publicly agreeing not to accept any postdoc offer with a deadline before the common deadline. If enough students from the top universities signed up, then there’d be great pressure on universities not to set early deadlines. Of course, the agreement wouldn’t be binding on the students, but it would be humiliating to break it, and chances are that enough students would honor it for it to be effective.

23. There’s a common deadline now? When did this happen?

24. No, it wouldn’t. In the medical resident match, the rankings are done after interviews.

Oh, so this is quite a different thing — are you envisioning a system in which departments would choose their interview candidates more or less as they do now, and then would rank only those, while candidates would rank only those schools where they got interviews?

25. JSE- Yes. I think that is how all currently functioning matches work, and it seems like the right choice to me.

26. John-

Anonymous…misspoke. There’s a (mostly) agreed upon day before which schools will not require people to answer offers, roughly so one won’t have to accept a job before one knows about the NSF fellowship.

“Anonymous”-

Having an actual common deadline without a match is a disaster. It basically removes completely schools ability to make a second round of offers. Having a common deadline was one of the stages in the unravelling of the medical resident market that led to the institution of the match.

27. it’s that the computer more or less plays out various rounds of offers and you simply don’t get any time to negotiate or wait for a partner to hear something during any of the rounds.

With an incredibly important difference: the offers made in the successive rounds are provisional, so schools lose nothing by making an “offer” to someone who won’t accept it.

Partly, the costs of not having a match are mostly hidden, they’re in people not getting offers they would have accepted because the school thinks they won’t, or people not getting offers they would have accepted because they had already taken another job that year.

Perhaps you can explain what these other benefits are that I’m missing

Consider your own situation this year. At least my understanding is that your wife was already established with a job in Boston, and so your highest priority was finding a job in Boston (if this was not the case, consider a hypothetical world where it was; certainly it’s a common enough situation).

What would you have done if you had gotten an offer from Stanford in January and several schools in Boston had told you were high on their short list, but they couldn’t make an offer yet?

I feel like in part, we’re estimating the probabilities of things differently. I’m pretty skeptical about one’s ability, especially as a fresh-faced postdoc, to seriously negotiate with schools, but know tons of people who have gotten stuck in a situation like the one above, where they have an offer from one school and are unsure about whether to wait for a better one. I’ve been lucky enough to avoid such situations, so this is not a personal axe I have to grind, but I feel like I see it constantly.

28. Much of this thread may be putting the cart before the horse. If North America had a centralized math job market, then a steering committee could choose among various protocols and rules to decide what fits our needs. For instance, Operation Postes is the centralized math job market of France. They have various reasons to do things in a specific way, which in the end has some key differences from the medical internship match. Some of these reasons and choices won’t be evident until you have centralization.

In fact, MathJobs.Org is already a solid amount of centralization. I think that math jobs wiki is also another turn of the screw — because you can start to see the job market as a whole instead of as separate pseudo-confidential parts — even if it’s not nearly as important as MathJobs.Org. The hardest part is to figure out how to push for further centralization, not to decide what to do with centralization if you had it. Plausibly the most realistic path forward it to persuade the AMS and the MathJobs people to pursue further centralization.

29. (The idea of matching, say, tenured jobs is simply absurd. I think the only reason why one might consider matching postdocs is because these are temporary so the consequences of a poor match won’t last forever.)

Because people never switch tenure track jobs? I’ve got no statistics, but it sure seems like people switch tenure track jobs a lot nowadays, in large part because of the two-body issue. It’s almost as though our current system sometimes makes poor matches.

30. Much of this thread may be putting the cart before the horse.

Of course, before you get around to hooking the cart up to the horse, you have to decide where you want to take the cart. It’s hard to centralize if people haven’t had a chance to throw around ideas for what a desirable centralization would look like.

31. I have to agree that centralization of senior appointments seems premature. I concede that some of the issues are the same, but that level of the job market still does resemble a collection of separate plays rather than one huge mosh pit.

The real issue is the job market from grad school to postdocs, and the job market from postdocs to tenure-track jobs. It is true that that second level does include some people who are leaving tenure-track positions, and some people who are taking tenured positions. But there is also a third senior level which looks rather different.

As Ben suggests, you could say that the current system already is a matching algorithm, and not a very good one. As I said, it seems premature to debate the details of some future centralized math job market. But if the opposition position is to reject all proposals as unworkable, I think that France’s Operation Postes is a clear counterexample to that.

One impulse would be to argue that France is too different from the US and we have little to learn from their practices. But that is what many mathematicians said about the physics arXiv before the math arXiv existed. In both cases, yes there are some differences, but not enough to sink the idea. Maybe they are as different as singular cohomology and etale cohomology. They are not as different as Calabi-Yau spaces and Frechet spaces.

32. Greg, do you understand Operation Postes? There seems to be a serious dearth of English language information on it. There needs to be a good wikipedia article or something.

33. Anonymous says:

Having an actual common deadline without a match is a disaster. It basically removes completely schools ability to make a second round of offers.

I’m not sure what you mean by a common deadline (do you consider the current situation an actual common deadline?). I should have specified that I meant one for first-round offers. That’s not quite true currently, but the bound for how early a deadline can be set turns into a default deadline for first-round offers. This system certainly doesn’t seem to destroy the ability to make a second round in practice. It does cause problems if someone waits until the deadline to turn an offer down, but most decisions happen much earlier. People seem to hold onto multiple offers only when they are genuinely conflicted, which happens but is not common.

Much of this thread may be putting the cart before the horse. If North America had a centralized math job market, then a steering committee could choose among various protocols and rules to decide what fits our needs.

I actually think centralization should come only after there’s a fairly clear proposal. The problem is that it will have to be voluntary. In the medical case, the profession is already vastly more centralized, with organizations like the AMA having far more power than any math society could ever plausibly have (for example, being the sole accreditor of US medical schools). If a handful of organizations like the AMA demand a certain system, then no medical school or hospital can resist. By contrast, in mathematics no central body has the power to compel change.

In mathematics, I don’t think very many departments would unilaterally delegate their hiring practices to a centralized steering committee. I certainly wouldn’t be in favor of it unless I already had a clear picture of what the committee would do and why it would be a definite improvement. Even then, I can’t imagine granting the committee any actual long-term power. I.e., I can see using a centralized system, while reserving the right to opt out next year if necessary, but I do not believe we would ever make a binding decision for more than a year at a time, and I bet many departments would react similarly. In effect, all large-scale changes must happen by consensus.

How does Operation Postes work in France? One key difference may be the role of government.

34. Alexander Woo says:

France is very different from North America.

In France, there is basically one kind of job: m.d.c. at research universities not all that different from each other except in prestige, location, and specifics of individual people already there.

In North America, there are several overlapping job markets (industry, research universities, liberal arts colleges, regional comprehensives) which are quite distinct, with fairly separate candidate pools, but some candidates have overlapping preferences between them.

It’s not uncommon for someone to have preferences such as Oregon > Macalester > Miami > Wake Forest > Google > San Fran. State > Northeastern. How do you deal with that? Does it make sense to try to have a common market here when SFSU and Miami might only have this single serious applicant in common?

35. Ben, the points you make are well taken, which I why I backpedaled a bit above. I’m still adamant that the match system “goes by too quickly” for couples or other complications. How would a “drawn-out” match system look?

To simulate a drawn-out match (I’ll use the University-makes-the-offers model since it’s closer to what we have now), you’d have to modify the current system such that:
1) Offers never ran out
2) Applicants could keep no more than one offer at a time

This is of course unrealistic. Let’s approximate this by:
1′) Offers last for at least time T_1
2′) Applicants who have an offer A and receive an offer B must reject one of A,B within time T_2

Basically, as long as T_1 >> T_2, it seems you’d have a reasonable approximation of the match system, but drawn out to give time for couples etc to figure things out. I’d guess T_2 shouldn’t be less than a week for this, and the appropriate minimum T_1/T_2 that “works” should be modeled. Hopefully the duration of such a system would be something feasible. Of course, there may be some terrible flaw I’m missing.

You could even incrementally reach such a system: universities could agree to enforce 1′ and 2′ among offers A,B from the group of universities that have agreed.

PS I too have no axe to grind other than in the abstract given that my wife was not applying as well this year (I also had none of the unfortunate issues Ben describes).

36. I would help tremendously to interview someone who has actually been through Operation Postes, or who helps run it, but I figured out a few things.

The page that announces positions is the one entitled “Concours MCF/PR”. “Concours” means “contest”, in the sense of “market”. MCF means associate professor and PR means Professor. 25 means mathematics, 26 means applied mathematics, and 27 means computer science. The numbers come from a French academic classification that somewhat resembles the MSC codes for mathematics. So in the “Concours” page, if you click on “MCF 25”, that lists openings for associate professorships in (pure) mathematics.

There is a complicated multi-stage application process with the first deadline in the fall. The final deadline is April 2. Without particularly simplifying the application system, Operation Postes (or maybe some other French administrative body) has at least synchronized it.

I think that the interviews come after April 2. I think that they have some round-by-round system which is still under way or has just been completed. At the moment, the home page of Operation Postes has a link that says “Les premières listes de classés sont sorties” which I think means “First-round rankings”. It has a link below that that says “Listes d’auditionnés”, which means “interview lists”, or in US job market terminology, short lists.

Somewhere on the site they have a PDF of a spreadsheet table with all of their hires last year.

A key difference between the French and US job market is that financially, the entire French university system is organized like a single state university system. By analogy, it would be easier for the University of California (or maybe UC plus Cal State) to synchronize and centralize its math hiring than for the entire US to do it. But since UC is not at all an island in job market space, it’s not clear that it should take such a step unilaterally.

37. I actually think centralization should come only after there’s a fairly clear proposal.

Yes, actual centralization should only come after there is a credible proposal. But the first step is to figure out how to petition the AMS. The AMS would want to make its own decisions for how to proceed. Proposing a specific set of rules can be useful as a way to argue that something should be tried. The fact that France already has such a system is another argument, in my view even more compelling. The success of MathJobs.Org is another compelling, related fact.

Organizations like the AMA having far more power than any math society could ever plausibly have

Yes, the AMA does seem more powerful than the AMS in some ways, but the AMS is hardly dead in the water. And I’m not saying that you should tap your foot and expect the AMS to solve our problems. What is true is that it’s much easier to make progress with help from the AMS than without it.

Remember, the AMS took the first step 20 years ago, when it made the AMS Cover Sheet. It was really a great thing in its day. The AMS also endorsed and now helps support the second step, MathJobs.Org, although that started at Duke.

38. France is very different from North America.

Right, Alex, in France no one would think to rank a job at Google France between tenure at Paris Sud and tenure at Clermont-Ferrand.

39. About Opération Poste: this is entirely a volunteer-run organization that centralizes information which is publically available but might be hard for an individual to get (typically because you would need to call 30 secretaries to find out when the hiring committees meets, then what their interview lists are, then what the ranked list is). Opération Poste _doesn’t_ create any synchronization: the various deadlines are fixed by the central Ministry of Higher Education and Research.

In any case, the synchronization can be hell for candidates: suppose you are currently a postdoc in a US university, and your teaching there ends May 15; it may well be that you will receive 10 interview offers where all the interviews _have_ to be held between April 25 and May 5 (because the last ranking meeting has already been scheduled for May 10). And all of the relevant expenses are for you to take care of (you won’t get a cent, even from the most prestigious place).

In any case, although it works quite nicely to make the hiring process more transparent in France, it can’t really be used as a comparison point for what is discussed in this thread, if only because negociating possibilities in France are essentially _nil_. (E.g., two body problems can be terrible in France, especially at the stage where you try to move from “Maître de conférence” to “Professeur”, which in 99% of cases, you can not do at the same place.)

(I should add that the system is supposed to change so that the universities can offer positions at any time when they become available; whether they will actually try to do something different than what currently exists is not clear).

40. anonymous says:

Why don’t people simply lie to Caltech? Why don’t they accept the Caltech offer, and then break it and go elsewhere? Are they afraid that Caltech faculty will retaliate in ways other than job offers?

Al Roth, who designs matching systems, says that this is an important step in incrementally adopting such systems: everyone in the system gets together and says that Caltech is cheating and so their agreements are not binding. (Mainly he talks about getting all the schools to agree to punish cheaters this way, where they definitely are cheating because they made this agreement ahead of time. But I think he advocates this more generally: he said that they can’t do this for clerkships because talking to a judge on the phone is legally binding in a way completely at odds with all other employment law.)

41. P.S. About Comment #38: I’m not sure I understand the meaning here; what is the implicit ranking between Google France, Clermont-Ferrand, and Paris Sud? There are very good mathematicians in Clermont-Ferrand (“Université Blaise Pascal”) , the rugby is much better, and a junior faculty member can actually rent a decent appartment for less that 3/4 of his salary.

42. I am not an expert on the French system, but from what I know of it, to describe “Maitre des Conferences” (MCF) as an associate professor is not right. An MCF cannot supervise graduate students, the pay scale is much lower than that of professors, they have very little administrative influence and get hired right after their PhDs. The only thing in common with the position of associate professor is that an MCF has tenure, but that is an artifact that academic jobs are government jobs and civil servants in France have job security. So “tenure” is better described as job security, not as academic tenure.

43. About Opération Poste: this is entirely a volunteer-run organization that centralizes information which is publically available but might be hard for an individual to get

Then I have grossly misunderstood the situation. It is an interesting point of comparison, but not in the way that I had in mind. As you describe it, Operation Postes actually sounds more similar to the math jobs wiki than to some kind of medical school match.

Maybe the medical school match is the best reference point after all. Although it is true that somehow the math job market in France is synchronized and there are rankings, as Operation Postes at least documents.

The fact that they don’t reimburse travel expenses for interviews is obnoxious, but that’s really separate from whether the deadlines are synchronized or not.

I’m not sure I understand the meaning here; what is the implicit ranking between Google France, Clermont-Ferrand, and Paris Sud?

Rest assured that I didn’t mean to denigrate Clermont-Ferrand as a mathematics department. In fact I picked it at random without knowing much about it. Let’s say that my contrived example might work for geographical reasons.

44. I am not an expert on the French system, but from what I know of it, to describe “Maitre des Conferences” (MCF) as an associate professor is not right.

Boy I really don’t have my facts this morning at all. Numerous Google searches told me that MCF is something like an associate professorship. But what you describe is somewhere between an assistent professrship with tenure and a postdoc with tenure.

My numerous mistakes are not my fault. I blame the Internet. :-)

45. Anonymous says:

Why don’t people simply lie to Caltech? Why don’t they accept the Caltech offer, and then break it and go elsewhere? Are they afraid that Caltech faculty will retaliate in ways other than job offers?

It’s not unethical for Caltech to be a jerk. Making offers with early deadlines (when they have never promised not to do so) is highly unfriendly but it’s still within their rights. Responding with a boycott could be reasonable, but there’s no moral principle saying it’s OK to punish jerks by treating them even worse in return.

P.S. I regret having brought up Caltech as a specific example. Several years ago they definitely did this (I know someone it happened to), but I do not know whether they still do today. In any case, they are not listed at http://www.ams.org/employment/postdoc-offers.html among the schools that have agreed to coordinate on timing, so they are free to do whatever they feel is best.

46. Yes, “Maître de Conférence” has no real equivalent that I know of in the US system (or in fact in any other academic system). The main comparison is with the “Professeur” positions in France, and the difference is — roughly — seniority. Typically people are first MCF, and after a few years and enough work, write an “Habilitation thesis”, after which they apply to Professor positions.

It is true that in France there is a ranked list of up to five names which is publicly available after the last meeting of the hiring commitees. At least in the current system, those have to be held before a certain date, and there is then a fixed period during which all candidates must indicate (on a central government web site) the position they select among those where they are ranked first or where other better-ranked candidates have declined already.
But the candidates do not, themselves, formulate any ranking beforehand.

I still wouldn’t recommend trying to get too much inspiration from this French model: everything about it was/is designed to makes things as simple as possible for the central administration only, and not to help the candidates or the hiring institutions.

47. I still wouldn’t recommend trying to get too much inspiration from this French model: everything about it was/is designed to makes things as simple as possible for the central administration only, and not to help the candidates or the hiring institutions.

I suppose that you’re right. In any case you clearly know better than I do.

Even if it does not work well as inspiration, I think that it is an interesting example. I’m convinced that centralization is the hard part of what Ben wants, and it is interesting that the French universities have it and make some real use of it. I take your point, though, that this is done for the convenience of the bureaucracy, and not to help either candidates or even search committees make better choices.

48. Peter says:

One problem with the current system seems to be an arms race in numbers of applications. A candidate who applies for more jobs has a significant advantage, and the _incremental_ cost is comparatively small; and this is only truer as the chances of success of each application get smaller. But as applicants apply to more jobs, each hiring committee is then choosing from a wider pool of applications, so can’t give each as much time for consideration…

So even if the ratio of jobs to applicants is stable, there’s a gradual increase in the time that applicants and hirers have to put in to the process (especially applicants), and a corresponding carefully either can find a fit. This is especially true lower down the academic ladder; in some fields I know people who’ve been advised that they need to apply to around a hundred jobs.

A matching system would presumably help this problem some.

49. When I brought up a suggestion like this on a philosophy blog last year, a few important points that came up in discussion stuck out.

One is that it deprives individuals of any bargaining power with respect to the institutions. Apparently, medical residents receive much lower pay than is really commensurate with their duties and training. Now perhaps post-docs are paid too much (so that there are too few post-doc positions to go around), but this really doesn’t seem clear, so instituting a change that would depress post-doc pay seems risky that way.

Another point is that this might run afoul of anti-trust law. Apparently Congress granted some explicit waiver to medical schools for residency programs.

Also, one’s preferences among institutions often changes quite a bit during the course of the process. At the stage immediately after the conference interviews (and before the campus visits), my ranking among the places that eventually gave me offers probably would have been Texas, Wisconsin, Pittsburgh, USC; after I finally got all the offers (I was second choice or later at all of them, so I got some of the offers fairly late in the season) my ranking ended up USC, Wisconsin, Pittsburgh, Texas, with all of the places moving up in my estimation of them, as a result of getting to know faculty members, visit the cities, and think more about what it would actually be like to be there. Since preferences change in the course of the process, it doesn’t seem fair to lock people into their original ones.

Also, one further point I should mention is that my original post was pointing out that, despite the Marriage Theorems, the system may also be subject to Arrow-style impossibility results, so that “strategic voting” becomes an important factor in creating lists. Of course, it’s not immediately obvious what circumstances would favor strategic voting, and whether those would be common, but it’s worth thinking about.

50. anonymous says:

Kenny Easwaran:
One is that it deprives individuals of any bargaining power with respect to the institutions. Apparently, medical residents receive much lower pay than is really commensurate with their duties and training.

The economists who study this kind of thing claim that it doesn’t have an effect on salaries. I don’t know whether to trust them, but there is data: the situation has changed several times in several fields. Also, Al Roth proved that the Gale-Shapely algorithm encourages honesty, which is more than can be said for the current system.

The obvious reason why residents are paid poorly is because there is a cartel of hospitals and they can’t become doctors, except by going through them. Few complain that medical students are paid poorly (ie, negatively), but residents are hardly different from medical students.

Anonymous @45:
I used Caltech not (just) because you mentioned it, but because I know about it.

Yes, that answers my question (@40) about why people don’t unilaterally lie to Caltech (though my sense of ethics condemns jerks), but that doesn’t answer my question about why the AMS doesn’t condemn Caltech and endorse lying. But what I really want to respond to is:

there’s no moral principle saying it’s OK to punish jerks by treating them even worse in return.

That moral principle is the basis of law. But maybe it’s different for individuals and groups.

51. Nat Whilk says:

anonymous wrote:

“[Punishing jerks by treating them even worse in return] is the basis of law.”

I can think of lots of counterexamples.

“But maybe it’s different for individuals and groups.”

Let’s hope. If we’re dishonest when it’s convenient, do we have any moral standing to ask our students not to cheat?

52. Danny Calegari says:

At the risk of wading into a discussion whose tone is in some ways a little impolite, let me make a few comments about the “common deadline”, at least as it is viewed at Caltech. Let me state from the outset that I speak here only for myself, and not as a representative of the Caltech math department. I am also not in any way endorsing Caltech’s position (or, for that matter, rejecting it), just trying to portray it from a different perspective.

1. There is not universal agreement here at Caltech about the merits or otherwise of adhering to the common deadline; this is an issue that comes up every year, and tends to generate a lot of heated discussion. The faculty here are sensitive to the issue, and especially to its impact on young mathematicians.

2. One of the main arguments that is put forward that we should *not* adhere to the common deadline as it stands, is that the existence of a common deadline is very advantageous to big schools, and detrimental to small schools. A key advantage of a small school is that it can make decisions very quickly, and therefore when they get a rejection, they can turn around and make another offer very quickly, before their competitors. This is a very significant factor at all hiring levels, including the senior level, by the way; I think it would be a mistake to dismiss this argument out of hand.

3. I have never heard (in mathematics) of a job applicant accepting an offer, with the deliberate intention of breaking their commitment if a better offer comes along. I strongly caution any graduate students reading this thread not to pursue such a strategy (at Caltech, or anywhere). Mathematics is a very small world, and it is very easy to get an undesirable reputation (not to mention the fact that the legal status of a verbal/email acceptance is quite unclear to me, and university lawyers are likely to have all kinds of views on the subject).

Anyway, I understand that some people have very strong feelings about this issue, so I can see why people use words like “jerk” and “cheaters”, and I sympathize. However, I think it’s worth pointing out that, in the first place, the use of such terms is very unlikely to change anyone’s mind; and in the second place, “cheaters” has a very specific technical meaning in social contract theory, which is inapplicable in this case, because Caltech is very explicit that it does not adhere to the common deadline.

53. Anonymous says:

At the risk of wading into a discussion whose tone is in some ways a little impolite

I’m sorry for contributing to the impolite tone. What upsets me is that it looks from the outside like Caltech is deliberately trying to lock in postdoc candidates by setting inflexible deadlines before they are likely to have heard from other top schools. First, this doesn’t seem very collegial. It creates a bad impression for someone’s introduction to the profession to consist of having a university try to close off their other options and trap them into going there. I also think it is bad on utilitarian groups: the harm to the student may be high (if they have strong personal or professional reasons to prefer another job), while the gain to Caltech is unlikely to be as high.

I’m also genuinely puzzled by the motivation. I don’t see why someone would want to lock in candidates who would prefer to go elsewhere. I’d rather have colleagues who chose to be here, rather than, say, slightly smarter colleagues who sometimes came under duress. I don’t know whether the belief is that they’ll end up just as happy and productive at Caltech (it’s true that lots of people think they know what they want, but they actually lack sufficient perspective to know), or that it really doesn’t matter (that winning by attracting the best postdocs is the key thing, regardless of how the postdocs feel).

One of the main arguments that is put forward that we should *not* adhere to the common deadline as it stands, is that the existence of a common deadline is very advantageous to big schools, and detrimental to small schools. A key advantage of a small school is that it can make decisions very quickly, and therefore when they get a rejection, they can turn around and make another offer very quickly, before their competitors.

Where I am, we almost never make a junior offer without having a specific backup already in mind in case of rejection, so that if someone turns us down we can immediately make another offer. Is this uncommon? I would have guessed most places do this, but I don’t have broad enough experience to say. In any case, the common deadline doesn’t keep anyone from taking advantage of quick turn around time when an offer is rejected.

As for speed in making the first offer, that’s an arms race. If Caltech decides to start three weeks earlier than anyone else, we could all match it, with no net gain to anyone. It doesn’t seem worth competing in this. Some schools may have a little less bureaucracy, but I can’t believe there’s really a fundamental obstacle here.

which is inapplicable in this case, Caltech is very explicit that it does not adhere to the common deadline.

Certainly, there’s no ethical issue here (although if I were at Caltech I’d argue hard for doing things differently).

54. Danny Calegari says:

Hi anonymous 53 – I understand the points you make and where you’re coming from. As I mentioned above, I am not endorsing Caltech’s position, just stating it as I understand it (and with an insider’s view of how this position was reached). How to manage job offers is a complicated business, and I think public discussions like this are refreshing and valuable. I promise to bring up some of this discussion the next time we have a hiring meeting!

Best,

Danny

55. anonymous@40 says:

Danny Calegari,
The comment about big schools is interesting, but it fails to address the NSF, which is the nominal purpose of the AMS agreement. Maybe that’s just a cover story for a conspiracy of big schools–and continual failure of the NSF to live up to its part of the bargain is evidence against its relevance–but my impression is that people resent Caltech not for missing the opportunity to go to a slightly more prestigious “big school,” but for missing the NSF. This is a PR battle you could wage, if you were more public.

No, Caltech is not “very explicit”; Anonymous@45’s fear of falsely accusing Caltech is pretty good evidence of that.

The whole point of social contract theory is all that we’re bound by rules that we haven’t accepted. That you reject the rule doesn’t stop me from calling you a cheater. If I reject monogamy, most people are going to call me a cheater, too, even if they won’t prosecute here and now. They are moving conventions and I could be wrong, but I think most people do consider Caltech cheating, even if they’re not in favor of retaliation.

56. Dear Anonymous 40, 50, and 55,
Please make an effort to moderate your tone. We welcome anonymous comments, but I have little patience for name-calling (even in hypothetical contexts).

57. It’s worth pointing out that Caltech is hardly unique. Just from friends job searches this year I can think of a couple other examples. Danny thanks for your patience in not taking this thread too personally.

I’m curious what happens in practice when someone accepts a pre-deadline offer and then gets an NSF to another school. Does the school always allow the applicant (who has already accepted) to delay their start by a year?

Where I’d really worry about early deadlines causing problems is if schools do it with jobs that aren’t totally posh. For example the Miller Institute postdoc at Berkeley is such a nice gig that it hardly seems like trickery to have a super early deadline. I forget the details of the Caltech job, but my recollection is that it’s pretty nice (in terms of high pay and low teaching). But what I would think of as abusing the power difference between school and applicant would be to offer less choice postdocs (say 2-2 teaching load at 50K) with an early deadline. Behavior like that would be something where “punishing” by the community would be feel more appropriate to me.

I’m also somewhat surprised at Danny’s thought that someone changing after committing to an early deadline would get them a terrible reputation. At the senior level similar sorts of things seem to happen all the time (prof. at school X takes an offer at school Y but first spends a year at school Z who then give them an offer and so the person never ends up at school Y at all, person takes an offer but only goes for a semester and then goes back to their old school, etc.) and everyone seems to understand that you need to do what’s best for yourself. Obviously postdocs can’t take the sort of risks that senior people can, but still I’m surprised that it’d be thought of that negatively.

58. One must remember that mathematics is a world-wide activity (medicine is also, but the regulations about practicing medicine outside of the country where you got your diploms are much more stringent!), and institutions outside the US have no particular reasons to want to follow US deadlines (whether formal or informal). There seems to be no reason to speak of cheating if other countries are involved.

I think that this year our own ETH deadline was earlier than, but very close, to the “NSF” one, but that was a coincidence. I don’t know how the idea of aligning ourselves on it formally would be received (my own opinion is that our offers are good enough that someone who was considering ETH sufficiently seriously to apply and come to the interview should be able to decide to take it, or to take a chance on getting something that they feel is better — of course I can recognize that there are real-life issues like two-body problems that may complicate everything, but individual cases should be decidable on an individual basis).

59. It might be of interest to some to know that theoretical high-energy physics has a common deadline for postdocs. No offers will be made that require a decision from candidates before January 7. Their FAQ

This blog post talks about some of the game theory involved. (It’s a physics blog, but the issues are essentially the same.)

And for reference of those who might be wondering, the analogous AMS-sanctioned math postdoc deadline this year is February 5. (In both cases the deadlines are not universal.)