In the early 1900s, competition among hospitals for interns and among
medical students for good internships led to increasingly early offers of
internships to students. By the 1940s, appointments were often made as early
as the beginning of the junior year of medical school. Hospitals thus had
little information about students' performance, and students frequently had
to make a final decision to accept or reject an offer without knowing which
other offers might be forthcoming. From 1945 through 1951, efforts were made
to enforce a uniform date for accepting offers. However, students were still
faced with offers having very short deadlines, compelling them to accept or
reject offers without knowing what other offers might be forthcoming. Hospitals
often had to scramble for available students, since if an offer was rejected,
it was often too late for them to reach their next preferred candidate. A
centralized clearinghouse was thus developed as a way of alleviating this
chaos and allowing a larger role to the preferences of both students and hospitals.
This evolved into the current matching program, whose algorithm continues
to be updated to take account of changing needs of applicants, such as growth
in the number of couples who seek 2 positions in the same vicinity.
The year 2002 marked 50 years of the resident match. Much has been learned,
both about the kinds of problems the market for interns experienced in the
first half of the 20th century and about the operation of clearinghouses like
the resident match.
The chief symptom that something was amiss in the early market for interns
was that hospitals began to try to hire interns earlier than their competitors,
so medical students often could only consider offers from one hospital at
a time, without knowing their prospects at other hospitals. The situation
in the 1920s is conveyed in a letter from the dean of the Columbia College
of Physicians and Surgeons1:
For a number of years attempts have been made to defer the appointment
of hospital internes until towards the close of the fourth year. The Association
of American Medical Colleges, the Council on Education of the American Medical
Association, and the American Hospital Association have all passed resolutions
favoring this idea. The difficulty has been in persuading someone to take
This is to inform you that it has been decided to defer the appointments
of internes at the Presbyterian Hospital in the City of New York until some
time in April.
It is earnestly hoped that other hospitals and schools will be able
to act in a similar manner.
That hope was in vain. A decade later, a survey of hospitals by Reginald
Fitz2 at Boston University found appointments
spread over the first semester of the students' senior year. He goes on to
Nearly a year ago the third year classes of the Harvard Medical
School and of Tufts Medical College wrote letters to the Boston Committee
suggesting, in effect, that it would be highly desirable from these students'
viewpoint if some arrangement could be established by which intern appointments
could be made in various hospitals at about the same time. . . . As one student
put it, there are very few men who have the conceit to pass up a very good
appointment in one locality offered early simply on the gamble of competing
for a somewhat more desirable appointment made later in another locality.
The problem worsened until, in 1945, there was an attempt to establish
and enforce a uniform time for intern appointments. In proposing the new plan,
Joseph Turner summarized the current state of internship appointments3:
Twenty-five and more years ago, the selection of internes by
most hospitals took place in the last half and even the last quarter of the
senior year. That selection has now been advanced on the school calendar to
the beginning of the junior year and, indeed, inquiries now come to me even
from sophomores. The dates of examinations and selection have been pushed
farther and farther back, through the efforts of some hospitals to get ahead
of others in the choice of candidates, for hospitals can exercise pressure
on the selected candidates by requiring acceptance of offers of internship
at once or within a short time. The student's dilemma is understandable; if
the first offer of this kind comes from a hospital of his second or third
choice, he loses out entirely if he declines and is not selected later by
the hospital of his first choice.
In response to this situation, it was proposed that medical schools
would not release information about students before an announced date.3 This "Cooperative Plan," adopted by the Association
of American Medical Colleges, achieved some uniformity in appointment times:
appointments for 1946 internships were largely made in the summer of 1945,
and in subsequent years the dates at which information was released by medical
schools was moved later into the senior year, and the date at which offers
were made followed in step.
However, over the next few years, students had to make increasingly
prompt decisions. In 1945, offers were to remain open for 10 days. By 1949,
a deadline of 12 hours was rejected as too long. Hospitals were finding that
if an offer was rejected after even a brief period of consideration, it was
often too late to reach their next most preferred candidates before they had
accepted other offers. Hospitals thus often pressured students to reply immediately;
offers conveyed by telegram were often followed by telephone calls requesting
an immediate reply.
The establishment of a clearinghouse, along the lines of what became
the resident match, was proposed as a way of continuing to reap the benefits
of uniform appointment dates while relieving pressure and congestion near
the deadline. In his preface to the proposal for a clearinghouse, F. J. Mullin,
dean of students at the University of Chicago School of Medicine, described
the shortcomings of the Cooperative Plan4:
The most frequently voiced objections to the present Cooperative
Plan are the following: Even when telegrams are filed early, the offices cannot
really release them all at once and the distribution gives much unfairness
and inequality. . . . Many hospitals have resorted to phoning the students
directly and putting pressure on students to make immediate decisions over
the phone. . . . Students sometimes get panicky and accept poor internships
way down on their lists because they have not heard from a higher position
on their order of preference. . . . Students have resented pressure for immediate
decisions put on them by phone communication from hospitals. Some hospitals
have felt that other hospitals have violated the principles of the Cooperative
Plan and have notified students early or have put undue pressure on students
for immediate decisions.
Mullin outlined how a clearinghouse would work: rank-order lists would
be solicited from students and hospitals, and used to produce a match.4 (Clearinghouses had earlier been tried on a regional
level, eg, in a Philadelphia Pool Plan5 and
a Boston Pool Plan [letter from William Castle to Reginald Fitz of Harvard
Medical School, September 28, 1951, courtesy of N. C. Webb, MD].) Mullin further
It should be made clear that under the proposed modification
of the Cooperative Plan hospitals and students would still be completely free
in making contacts and getting information about each other and in expressing
their choice in selection of placement and applicants. The proposal calls
for the establishment of a central clearing agency to act only as a mechanical
facilitation in the final step in the final process of intern selection.
At the annual meeting of the Association of American Medical
Colleges in October, 1950, this plan was discussed and it was voted to make
a trial run for the present year without influencing the procedures already
Following the trial run, it was resolved that, for 1952 internships,
a centralized match would be used to finalize internship appointments. Mullin
and Stalnaker,6 announcing the centralized
match, summarized its anticipated benefits as follows:
Under the plan the student will not be required to make a decision
on the basis of a telephone call or within a very limited period of time.
A last minute scramble, with its many uncertainties, is eliminated. No student,
under the plan, will receive telegraphic offers by a number of hospitals and
wonder if he will receive other offers later. Hospitals will not send out
telegraphic offers to many students only to receive no replies or negative
ones, thus requiring them to send out additional offers at a later time to
students who may, in the meantime, have taken another internship although
they preferred the hospital involved.
This plan was implemented, with one crucial change. The algorithm outlined
by Mullin and Stalnaker for turning rank-order lists into appointments met
with objections from students. W. Hardy Hendren, then a student at Harvard
Medical School, recounts how, after learning of the proposed algorithm, he
organized the National Student Internship Committee, which proposed a different
algorithm (oral communication). Hendren and his fellow students noted that
under the originally proposed algorithm, a student could suffer by submitting
a rank-order list that listed as first choice a position he or she was unlikely
to obtain. This is worth describing in more detail, since the choice of matching
algorithm had a large effect on the operation of the match.
Mullin and Stalnaker6 had described a
clearinghouse in which students ranked individual hospitals and hospitals
ranked students in groups, with "1" being reserved for the most preferred
students up to the number of available positions, "2" for the next most preferred
group, and so forth. The proposed algorithm first matched all hospitals and
students that were each others' first choices (1-1 rankings). Then, hospitals
would be matched with students in their second group if those students had
ranked the hospital first (2-1 rankings), followed by matches of hospitals'
first choices with students' second choices (1-2 rankings), and so on (2-2,
3-1, 3-2, 1-3, 2-3 . . . ). The intention appears to have been to give an
advantage to students, since when preferences conflicted, students' first
choices were considered earlier than hospitals' first choices.
However, consider a student who listed as his first choice a hospital
to which he did not match, but whose second-choice hospital included him among
its first choices. That hospital might fill all its positions in the 1-1 and
2-1 steps of the algorithm and have no position available for the student,
a 1-2 match. Thus, it was possible for a student to suffer by ranking first
a hospital to which he or she could not match; the student could end up at
a hospital he liked very little, even though his second-choice hospital had
ranked him first. Mullin and Stalnaker7 discussed
. . . There was dissatisfaction caused by the student fear of
being penalized for taking a "flyer." Following a meeting in New York, an
ad hoc student committee made proposals involving a change in the procedure
of matching which was supported by other student groups. The National Interassociation
Committee, after consideration of the suggested changes known as the Boston
Pool Modification, adopted them as the official method to be used in the matching.
The Boston Pool algorithm updated rank-order lists as it went along,
tentatively matching students to hospitals that presently ranked them in the
first group and deleting a student from a hospital's list only when that student
was tentatively matched to a hospital the student preferred (at which point
initially lower-ranked students could move into the first group on the hospital's
list). Thus, a student who ranked first a hospital to which he or she did
not match could nevertheless be assured that if the second-choice hospital
did not fill with students it preferred, he or she would get a position there.
This Boston Pool algorithm is equivalent to a "deferred acceptance" algorithm,
which can be interpreted as one in which hospitals make offers to applicants,
starting at the top of each hospital's rank-order list, and each applicant
holds on to the best offer he or she has received so far but can later reject
it if a better offer is forthcoming.8
The change in algorithms was fortunate for the longevity of the match,
which became the National Resident Matching Program (NRMP), because the Boston
Pool algorithm had another property the Mullin and Stalnaker6 algorithm
lacked. It produced outcomes that were stable, in
the sense that no applicant and hospital who were not matched with one another
preferred each other to their assigned matches.
The importance of stability has since become clear. If an algorithm
produces unstable outcomes, then there are applicants and hospitals who would
both prefer to be matched to one another than to accept the results of the
match (as in the example described earlier). This creates mutual incentives
for the unhappy pairs to circumvent the match.
For example, when the British market for interns experienced increasingly
early appointments in the 1960s, each region of the British National Health
Service devised its own centralized clearinghouse. Several used algorithms
very similar to that of Mullin and Stalnaker. These all failed and were abandoned
after interested applicants and hospitals learned to circumvent them. In contrast,
clearinghouses that produced stable outcomes succeeded and remained in use.9,10 This and related evidence strongly
suggest that, had the originally proposed match algorithm not been replaced,
we would not now be looking back on 50 years of operation of the NRMP.
In the 50 years since the inception of the match, changes in medicine
have been reflected in the demands on the match and in the design of the algorithm.
Fifty years ago, the vast majority of US medical graduates were men; today,
the match accommodates couples who may submit rank-order lists of pairs of
positions, to obtain jobs together. Fifty years ago, most internships were
nonspecialized rotations; today, many specialties require more than 1 residency,
and the match accommodates applicants who need to combine 2 positions by allowing
them to submit a primary rank-order list for (typically) second-year positions
and supplemental lists for first-year positions to be applied if a second-year
position is obtained.
The object of the most recent redesign of the algorithm, used since
1998, was that it should yield a match as favorable as possible to applicants
while producing a stable outcome that accommodated these contemporary requirements.11,12 The current (Roth-Peranson) algorithm
is a deferred acceptance algorithm that can be viewed as a process in which
applicants offer to come to residency programs, starting with the residency
program at the top of the applicant's rank-order list, rather than one in
which the programs make offers to applicants, starting at the top of the program's
Today, there are also matches for fellowship positions, many initiated
in the 1980s and 1990s, after fellowship markets exhibited the unraveling
of appointment dates that characterized the intern market before the resident
match. So, 50 years after the inception of the resident match—years
that have included significant changes and some controversy—it continues
to serve the function for which it was designed.
This article has focused on the match process, which has occurred in
the context of larger changes in the organization of US medical education.
Historical background information has been provided by Starr13 and
Ludmerer.14 Gale and Shapley15 and
Roth and Sotomayor16 provide further reading
on stability and matching. Roth and Xing17 describe
other markets that have experienced unraveling of transaction times; see also
my more formal contemporary overview.18
Some tools below are only available to our subscribers or users with an online account.
Download citation file:
Web of Science® Times Cited: 32
Customize your page view by dragging & repositioning the boxes below.
Enter your username and email address. We'll send you a link to reset your password.
Enter your username and email address. We'll send instructions on how to reset your password to the email address we have on record.
Athens and Shibboleth are access management services that provide single sign-on to protected resources. They replace the multiple user names and passwords necessary to access subscription-based content with a single user name and password that can be entered once per session. It operates independently of a user's location or IP address. If your institution uses Athens or Shibboleth authentication, please contact your site administrator to receive your user name and password.