Harold Winslow

Against the Medical Match

Doctors go to medical school for four years and then do additional training in a speciality for ~3-7 years, depending on the specialty. This period is known as residency. Doctors in residency work long hours at a teaching hospital and get paid a small fraction of the salary they will eventually make - generally something like $60K - $70K in residency vs $200K - $400K afterwards.

The labor market for doctors who have just finished medical school is quite odd. In college admissions, and in the labor market at large, a person can apply to as many institutions as they like, potentially receive several offers, and ultimately choose the one that suits them best. But when a soon-to-be MD is seeking admission to a residency program - a sort of apprenticeship in which they provide useful labor but also deepen their skillset - they confront a different process. The student will still apply to many programs (often upwards of 30), but there the similarities end. Instead of directly receiving offers and picking among those, each student is asked to submit a ranked list of programs to a central service. Likewise, each residency program submits a ranked list of students to the service. The service then follows an algorithm to pair applicants to programs. This process is colloquially known as The Match. The Match has some nice theoretical properties – it is in everyone’s best interest to report their true preferences (the system provably cannot be gamed), and the resulting match will always be stable in the sense that no resident-hospital pair will ever want to be together more than their current pairings. The match is also a hallmark of the economic field of market design and was created in its current form by Nobel Prize-winner Al Roth.

Nevertheless, it is an awful way to run a labor market, as you shall soon see. Since the match is such an odd institution, a word about why it exists in the first place. Once upon a time hospitals wanted to poach the best young doctors they could find, as this is generally good for prestige and patient care. So they started making employment offers earlier and earlier. Eventually offers were being made in the first half of medical school, before students had a chance to do any clinical rotations or take major board exams. At the same time, hospitals started making exploding offers – e.g. offering a student with top grades employment but requiring a response within 12 hours. These practices put students in poor positions to make career decisions and too many contracts were being broken. It was time for a change.

Sidenote: why were early and exploding offers common in medicine but not in other fields? I think the main factors are commitment on the part of the potential employee and specificity and stability in the needs of the employer. Medical students are far enough along in their education that they've credibly committed to a particular profession. This is not the case for, say, college sophomores - no hiring manager is going to plan on having a sophomore start two years from now if only because that sophomore may very well change their major, drop out, etc before their start date. In addition, hospitals know exactly what type of candidate they're looking for when hiring, and can easily screen for such candidates, which means hospitals can seek out candidates directly instead of waiting for candidates to respond to a job posting. Hospitals can also expect these needs to remain stable for years to come. Neither of these features are typical of most employers.

The change (which doctors initially came up with themselves) was to centralize the labor market – i.e. to force all decisions to be made at the same time following a standardized procedure. This would eliminate the incentive and possibility for hospitals to make early and/or exploding offers of employment.

And in this way the match has been successful – no medical student today would think of trying to privately negotiate an employment contract before submitting their preferences through the match. That is largely because they've made it exceedingly difficult to do so.

The difficulty of circumventing the match

As per the National Residency Match Program contracts, the outcome of the match is binding:

Programs and applicants are not authorized to release each other from their binding commitment. Once a party has matched […], a waiver of the binding commitment may be obtained only from the NRMP.

Students who would be interested in deviating from the match would also have to do so without much support from their school officials:

It shall be deemed a violation of this Agreement for the school official, school administrator, or any other individual or entity to engage in any communication (a) concerning SOAP-eligible or SOAP-ineligible unmatched senior students or graduates prior to contact from directors of unfilled programs […]

Plus you can’t try to give yourself a safety net by entering the match and simultaneously shopping around: Applicants who register for the Main Residency Match but who do not withdraw from the Match before the Rank Order List Certification Deadline are prohibited during the period between the Rank Order List Certification Deadline and 3:00 p.m. eastern time on Monday of Match Week from applying for, discussing, interviewing for, or accepting a position that would run concurrent with positions offered in the Main Residency Match.

So the incentives are pretty strongly stacked against students unilaterally attempting to subvert the match. To legally do so they could not enter the match at all. And since the vast majority of residency positions are filled via the match (a small fraction of students go unmatched and find jobs in the post-match “scramble”), this student would need to start marketing themselves to hospitals early on. The main incentive for doing so would be to earn a higher wage. Would any hospital be willing to pay a resident more than the market (or match) rate? In a more conventional labor market, sure. The match facilitates sorting according to talent/prestige: the ‘best’ students match with the ‘best’ hospitals. That means that hospitals with second-tier reputations generally only have access to second-tier talent, and such hospitals might be willing to pay more for better candidates if they could. But the NRMP doesn’t allow trifling – you either fill all of your positions through the match or you fill none:

The sponsoring institution agrees that prior to the release of the results of the Main Residency Match, all programs sponsored by the institution, including those that do not participate in the Match, will select sponsored applicants only through the Match or another national matching plan. If any position is offered to sponsored applicants outside the Main Residency Match or another national matching plan, including a preliminary position for a program that participates in another national matching plan, the institution will be in breach of this Agreement and may suffer the penalties described in Section 10.0.

For the record, Section 10.0 basically just says that the NRMP will tell everyone that you violated the match agreement.

These constraints imply that to circumvent the match one would need to get an entire hospital to opt out of the match one year. But no hospital would want to opt out of the match without a reasonable pool of candidates who might apply to their program. And other candidates are not going to opt out of the match without having multiple residencies to apply to...it's theoretically viable to have some portion of candidates/residencies operating within the match and some outside of it, but to get to that equilibrium from the current one would require something more akin to revolution than incremental change.

Does this matter? Is there a reason why students or residency programs (hospitals) would want to escape the match?

Unintended consequences of centralizing the labor market

A resident's salary appears to be far below the marginal product of their labor. There are a few reasons to believe this. First, a resident's salary may triple from one month to the next simply by virtue of finishing residency. Meanwhile, the nature of their work changes little - by the end of residency they are already operating mostly independently. An increase of 3x seems exorbitant.

Second, residents can make substantially more than their base salary by moonlighting. Although moonlighting work isn't exactly the same as working under an attending physician and periodically receiving feedback and instruction, it does demonstrate that residents are capable of more remunerative work. In addition, residents routinely perform tasks which nurses cannot do and nevertheless make less money while working longer hours than nurses. And economic theory predicts that institutions like the match will result in wage suppression.

At least some wage suppression is expected because residents value the training they are receiving and the prestige of the sponsoring institution and are willing to implicitly pay for these benefits. And some people think that this implicit tuition accounts for most of the wage suppression.

Suppose that low wages are driven primarily by this implicit tuition effect. That would mean that residents are willing to forgo tens to hundreds of thousands of dollars in salary in exchange for prestige. This is possible - medical students are nothing if not type-A. On the other hand, it seems somewhat absurd. Pursuing a prestigious medical school makes some sense - people will routinely ask where you went. But most people don't know enough to ask about residency or evaluate the prestige of residency programs, so it is hard to believe that most doctors would be willing to take a substantial pay cut if they had the choice.

The only way to conclusively rule out this possibility is to decentralize the labor market and find out if wages rise. But there is one remaining piece of evidence which suggests that it is the match that drives wage suppression rather than implicit tuition fees. Some residents were so outraged at the terms of The Match that they sued the Association of American Medical Colleges. If things had gone their way, The Match would have been abolished or severely restricted. But hospitals were so against such an outcome that they lobbied against it. In the end, the match was condoned: it is now explicitly exempted from anti-competitive statutes. Presumably they know that they wages they provide are artificially low and that if they are forced to compete with each other then they will end up paying much more.

Even if I'm wrong about the size of the tuition effect, it still can't account for the relatively stagnant wages that residents see over their four years of training. A fourth year resident is nearly an independent doctor, receiving far less supervision and instruction than a first year, so they should not need to pay nearly as much implicit tuition and their wage should be much higher - in the general ballpark of what they'll make later. And yet wages only rise on the order of $15K between the beginning and end of residency.

To summarize thus far: the algorithm that matches residents to hospitals probably allows hospitals to keep resident wages artificially low. It would be nice to change this, but there strong incentives and legal structures present which prevent any individual entity from precipitating the change. This is frustrating, but there is still hope.

Any proposal for reform must address the issues that caused the labor market to unravel in the first place: early and exploding offers. One incremental change that might address while also opening up competition is to award each resident with two matches. This idea has a lot of theoretical appeal - it's logistically close to the status quo but plausibly quite a lot better. There are other ways to upgrade the match as well. Another alternative is for medical schools to place a collective embargo on student grades until the fourth year, thereby eliminating the viability of early employment offers. Exploding offers are harder to prevent, but they are less of a concern when made in a student's fourth year than when made in their second.

Regardless of your preferred alternative, you might wonder what could force a change to a new system. Residents do have a union. Unfortunately they don’t seem very powerful – their benefits page cites Costco coupons as an example. Costco is already a discount store, people. Nevertheless, they report a membership of 17,000 residents nationwide. A strike would draw a lot of attention and likely force change of some kind. The only other mechanism for reform that I can think of to start a hospital from scratch and commit to paying residents reasonably from the start. This strategy hinges on finding the right person – someone with enough expertise to run a hospital and sufficient gall to pierce the anti-competitive veil.