Borjas and Doran 1

Prizes and Productivity:

How Winning the Fields Medal Affects Scientific Output

George J. Borjas
Kirk B. Doran

Abstract: Knowledge generation is key to economic growth, and scientific prizes are designed

to encourage it. But how does winning a prestigious prize affect future output? We compare the

productivity of Fields medalists (winners of the top mathematics prize) to that of similarly

brilliant contenders. The two groups have similar publication rates until the award year, after

which the winners’ productivity declines. The medalists begin to “play the field,” studying

unfamiliar topics at the expense of writing papers. It appears that tournaments can have large

post-prize effects on the effort allocation of knowledge producers.

George  J.  Borjas  is  the  Robert  W.  Scrivner  Professor  of  Economics  and  Social  Policy  at  the  

Harvard  Kennedy  School  and  Research  Associate  at  the  National  Bureau  of  Economic  

Research;  [email protected].  Kirk  B.  Doran  is  an  Assistant  Professor  of  Economics  at  

the  University  of  Notre  Dame;  [email protected].    We are grateful to Noam Elkies, William

Evans, Richard Freeman, David Galenson, Daniel Hamermesh, David Harbater, Larry Katz,

Curtis McMullen, Kannan Soundararajan, Wilfried Schmid, Bruce A. Weinberg, Yoram Weiss,

and Trevor Wooley for comments and suggestions on earlier drafts of this paper, and to Brian

Brinig for excellent research assistance. This paper uses confidential data from the MathSciNet

data archive maintained by the American Mathematical Society. The data can be obtained by

filing a request directly with the American Mathematical Society. The authors would be happy to

provider guidance to other researchers pursuing use of these data.

 Borjas and Doran 2

Prizes and Productivity: How Winning the Fields Medal Affects Scientific Output

George J. Borjas and Kirk B. Doran

I look forward to proving more theorems. I hope the weight of this prize doesn't
slow me down.
—Stanislav Smirnov, Fields Medalist, 2010

I. Introduction

The production of knowledge is central to long-term economic growth. Yet little is

known about how knowledge is produced, making it difficult to predict which types of incentives

are most effective in eliciting effort from knowledge producers. Prizes are a common incentive

for knowledge production; hundreds of scientific prizes are awarded throughout the world and

across all scientific disciplines. Although these prizes are frequently awarded with the explicit

goal of inspiring more and better scientific work (Scotchmer 2006), a question remains: are they

effective?

A voluminous theoretical and empirical literature examines how the presence of potential

future rewards (whether a promotion to CEO or winning a golf tournament) elicits optimal

efforts from the tournament participants in their effort to win the contest.1 This literature

emphasizes the incentive effects of the prize prior to the completion of the tournament. But what

happens to the productivity of tournament winners after they win the prize?2 Standard models of

labor supply suggest that the post-prize impact of a big win could be significant, especially when

the award is granted at a young age.

This paper examines the impact of winning the Fields Medal on the post-medal

productivity and research choices of mathematicians.3 The Fields Medal is the most prestigious

award in all of mathematics, awarded every four years to mathematicians under the age of 40.
 Borjas and Doran 3

Established by the Canadian mathematician John Charles Fields, the medal is often thought of as

the “Nobel Prize of Mathematics.”4 Inspired by the desire to promote mathematical cooperation

and research around the world, Fields used his position as Chairman of the Organization

Committee of the International Mathematical Congress to bring his idea to fruition. In a memo

appended to the January 12, 1932 minutes of the Committee of the International Congress at

Toronto, Fields described his vision:

It is proposed to found two gold medals to be awarded at successive International

Mathematical Congresses for outstanding achievements in mathematics. Because
of the multiplicity of the branches of mathematics and taking into account the fact
that the interval between such congresses is four years it is felt that at least two
medals should be available (Tropp 1976, p. 173).

In the same document, Fields explained the motivation for the award: “while [the medal]

was in recognition of work already done it was at the same time intended to be an

encouragement for further achievement on the part of the recipients and a stimulus to renewed

effort on the part of others” [emphasis added]. In other words, not only would the existence of

the prize solicit effort from the participants in this elite tournament, it would also encourage

additional (that is, post-medal) effort on the part of the winners.

Using administrative data from the American Mathematical Society (AMS) and the

Mathematics Genealogy Project (MGP), we examine the shape of the age-productivity profile of

these exceptional mathematicians along a number of dimensions, including the number of papers

published, citations received, and students mentored. Our empirical analysis exploits the fact that

only a subset of the great mathematical contributions in the past 80 years resulted in Fields

medals, and that this subset was partly determined by arbitrary factors such as the quadrennial

timing of the award, the age restriction, and subject-level biases.

 Borjas and Doran 4

Our main set of results uses the set of winners from a broader set of prizes for great

mathematical achievement (awards which are themselves good predictors of winning the Fields

Medal) to construct a representative group of brilliant mathematicians who can be thought of as

“contenders” for the medal. Our analysis compares the research output of the medalists with that

of the losing contenders. The age-productivity profile of the two groups is similar until a

particular mathematician wins the Fields Medal (or does not win it). Remarkably, the

productivity of the Fields medalists declines noticeably relative to that of the contenders in the

post-medal period, regardless of whether we measure productivity in terms of papers published,

citations received, or students mentored.

The award of a Fields Medal enhances a mathematician’s professional reputation far

more than any other potential event in the mathematics profession. Although the prize money is

itself negligible, the medal can certainly lead to a substantial increase in “wealth” (both in terms

of dollars and in intangible measures of academic prestige) if the mathematician chooses to

leverage the award in the marketplace. The neoclassical labor-leisure model predicts that the

expansion in the opportunity set should induce the medalists to consume more leisure in the post-

medal period. Moreover, the wealth effect can also lead to a shift in the research strategy pursued

by the Fields medalists: they are now free to “play the field” and pursue topics in different areas

of mathematics (or even outside mathematics) that they may find interesting or worthwhile and

have a high consumption value.

We employ the notion of “cognitive mobility” (Borjas and Doran 2014) to capture the

transition in the space of ideas as knowledge producers move from one research topic to another

over the course of a career. The AMS data classifies each published paper into one of 73 specific

and narrow mathematical fields. It turns out that there is a crucial link between a
 Borjas and Doran 5

mathematician’s propensity for cognitive mobility across mathematical fields and the awarding

of a Fields Medal. Specifically, while medalists and contenders have similar cognitive mobility

rates initially, the medalists exhibit a far greater rate of mobility in the post-medal period.

Because cognitive mobility is costly (for example, additional time is required to prepare a paper

in an unfamiliar field), the increased rate of cognitive mobility reduces the medalists’ rate of

output in the post-medal period. The data suggest that about half of the decreased rate of output

is due to the increased propensity for “trying out” unfamiliar fields, often outside pure

mathematics.

Every four years, the greatest mathematicians in the world gather to select and honor new

medalists and to remind them that the Fields Medal is meant to encourage their future

achievement. In fact, the medal reduces the rate of publication and the likelihood that its winners

produce great achievements in pure mathematics. At least in this context, it seems that the net

productivity impact of selecting winners on the basis of a tournament depends crucially on what

happens as the winners adjust their behavior to take advantage of the post-prize expansion in the

opportunity set.

II. Historical Background

The first Fields Medals were awarded soon after Fields’ 1932 memo. The medals are

traditionally awarded during the opening ceremony of the quadrennial International Congress of

Mathematicians (ICM). In 1936, the medals were awarded to two mathematicians. Because of

World War II, the medals were not awarded again until 1950, when they were again given to two

mathematicians. Since 1950, the Fields Medal has been awarded quadrennially, to two, three, or

four mathematicians in each cycle.

 Borjas and Doran 6

The initial moneys available to fund the medals were the result of an accidental surplus of

funds left over after the 1924 ICM. These funds, accompanied by the bequest of Fields himself,

allowed for the granting of two medals. In 1966, an anonymous donor made additional funds

available allowing four medals to be awarded in each of the next two cycles (ICM 1966). As a

result, the number of Fields Medals awarded in any given 4-year cycle was not mainly

determined by how many mathematicians had made fundamental advances in the relevant time

period. Instead, the number often depended on how much income had accumulated in the Fields

Medal account, on the availability of private anonymous donors, and on an upper limit of

(initially) two or (later) four medals to be awarded by any particular Congress (ICM 2006).

As noted earlier, the Fields Medals were designed partly to promote future mathematical

achievement on the part of the recipients. This goal is sufficiently important that it has been

repeated verbatim and expounded upon at nearly every award ceremony. For example, in the

1954 Congress, eminent mathematician Hermann Weyl spoke movingly to the winners: “The

mathematical community is proud of the work you both have done. It shows that the old gnarled

tree of mathematics is still full of sap and life. Carry on as you began!”

From its inception, the committees have interpreted Fields' desire for future

encouragement to mean that the medal should be awarded to mathematicians who are “young”

(ICM 1936), and the word “young” has consistently been interpreted to mean that the medal may

only be awarded to mathematicians under the age of 40 (ICM various issues).5 Most recently, the

2006 committee explicitly stated its requirement that a mathematician qualifies for the Fields

Medal only if he has not yet turned 40 as of January 1 of the year in which the Congress meets

(ICM 2006).
 Borjas and Doran 7

The restrictions on the number and age distribution of the Fields medalists introduce

arbitrary variation in which subset of great mathematicians of the past eighty years received the

award and which did not. As a result, many mathematicians who are widely perceived as “great

architects of twentieth-century mathematics,” even for work done at an early age, did not receive

the Fields Medal (Tropp 1976). There are numerous such examples. The American Mathematical

Society said of mathematician George Lusztig: “[His work] has entirely reshaped representation

theory and in the process changed much of mathematics” (AMS 2008, p. 489). Although

“Lusztig's exceptional mathematical ability became evident at an early stage of his career,” and

"it can be no exaggeration to say that George Lusztig is one of the great mathematicians of our

time," he did not receive the Fields Medal (Carter 2006, pp. 2, 42).

Similarly, the Norwegian Academy of Science and Letters (Solholm 2010) cited John

Tate for “his vast and lasting impact on the theory of numbers,” claiming that “many of the

major lines of research in algebraic number theory and arithmetic geometry are only possible

because of [his] incisive contributions and illuminating insights.” Nevertheless, Tate also did not

receive a Fields Medal.

In fact, considering the number of mathematicians who are regularly lauded by the

various National Academies of Sciences and Mathematical Societies for (re)inventing new

subfields of mathematics, it is clear that the 52 Fields Medals that have been awarded (as of

2013) are insufficient to cover even half of all the great achievements that have made modern

mathematics possible. Hence it should not be particularly surprising that Robert Langlands, a

mathematician whose work specifically inspired and made possible the contributions of at least

two Fields Medalists (Laurent Lafforgue and Ngô Bảo Châu), and who founded the most
 Borjas and Doran 8

influential program connecting number theory and representation theory, did not receive the

Fields Medal himself.

Historians of the Fields Medal have also documented the “bias” that causes some fields

and styles of mathematics to be better represented among winners (Monastyrsky 2001). For

example, Langlands (1985, p. 212) wrote of mathematician Harish-Chandra: “He was considered

for the Fields Medal in 1958, but a forceful member of the selection committee in whose eyes

Thom [one of the two Fields medalists that year] was a Bourbakist was determined not to have

two. So Harish-Chandra, whom he also placed on the Bourbaki camp, was set aside."6

Similarly, the arbitrary age cut-off and the four-year periodicity of the award work

together to exclude mathematicians who obviously should have received the medal. The New

York Times obituary of Oded Schramm states: “If Dr. Schramm had been born three weeks and a

day later, he would almost certainly have been one of the winners of the Fields Medal…But the

Fields Medals, which honor groundbreaking work by young mathematicians, are awarded only

once every four years and only to mathematicians who are 40 or under. Dr. Schramm was born

on Dec. 10, 1961; the cutoff birth date for the 2002 Fields was Jan. 1, 1962. Wendelin Werner, a

younger mathematician who collaborated with Dr. Schramm on follow-up research, won a Fields

in 2006” (Chang 2008).

In short, while it is tempting to claim that the 52 Fields medalists are in a class by

themselves, and that there are no losing contenders with equivalent or better early achievements,

this view does not correspond with what mathematicians themselves have written. As the ICM

noted: “we must bear in mind how clearly hindsight shows that past recipients of the Fields’

medal were only a selection from a much larger group of mathematicians whose impact on

mathematics was at least as great as that of the chosen” (ICM 1994). The arbitrariness in the
 Borjas and Doran 9

number, timing, and field distribution of Fields medalists means that a similarly great group of

“contenders” should exist that can be contrasted with the winners in a difference-in-differences

strategy to determine how winning the medal influences productivity and research choices.

III. Data

To measure the life cycle productivity of elite mathematicians, we use the comprehensive

data contained in the AMS MathSciNet archives. The AMS provided us with a database that

reports the number of papers published by every mathematician in the world, by field and year,

since 1939. The AMS professional staff assigns each publication in mathematics to one of the

many fields that make up the discipline (and this information will prove useful below). Our

database contains the author-year-field information at the two-digit field level, classifying every

publication over the 1939-2011 period into one of 73 different fields. The database also contains

information on the number of citations received by the papers. It is important to note, however,

that the AMS citation data is incomplete. In particular, it only counts citations in a limited

number of journals (which include the most important journals in mathematics), and only reports

the post-2000 citations received by a paper (regardless of when the paper was published).

We wish to determine what the post-medal career path of Fields medalists would have

looked like had they not been awarded the medal. Because of the capricious events affecting the

selection of the subset of great mathematicians who received the medal, we conjecture that there

should exist a comparison group of mathematicians who did similarly path-breaking work before

the age of 40, but who did not receive the medal and can serve as a control group.7

We use a systematic and easily replicable method for constructing the set of

“contenders.”8 Specifically, our construction of the control group starts out by including the
 Borjas and Doran 10

winners of six other major mathematical awards with roughly similar goals as the Fields Medal.

It turns out that winning any one of these prizes is a good predictor for receiving a Fields Medal.

First, we consider the two most prestigious general mathematics prizes (after the Fields

Medal), which tend to be given closer to the end of a mathematician’s career. Both the Abel

Prize and the Wolf Prize cover the entire breadth of the mathematics discipline and are only

given to mathematicians who have made extraordinary contributions. The Abel Prize, which has

a significantly higher monetary value than the Fields Medal (nearly $1 million versus $15,000),

began to be awarded in 2003 to one or two mathematicians a year.9 The Wolf Prize has been

awarded annually since 1978, typically to two mathematicians (although no prize has been

awarded in some years). Any mathematician who won either of these prestigious awards (and did

not win the Fields Medal) is clearly a key formulator of modern mathematics and automatically

becomes part of our group of potential contenders.

In addition to these two general prizes, there are a number of prestigious area-specific

prizes in mathematics. Specifically, we consider the four most prestigious area-specific awards

for: algebra (the Cole Prize of the AMS); analysis (the Bôcher Prize of the AMS); geometry (the

Veblen Prize of the AMS), and the study of Fourier series (the Salem Prize). We add into our

group of potential contenders any mathematicians who won one of these four area prizes before

the age of 40 (and did not win the Fields Medal).10

This algorithm yields the names of 92 potential contenders who contributed significantly

to at least one of the key subject areas of mathematics or to mathematics as a whole, but who did

not receive the Fields Medal. There is a very strong correlation between winning any of these

prizes and winning the Fields Medal: 52 percent of the Fields medalists also won at least one of

these prestigious awards. The predictive power of each prize is as follows: five out of the 13
 Borjas and Doran 11

Abel Prize winners also won the Fields; as did 13 out of the 54 Wolf Prize winners; three out of

the 26 Cole Prize winners; four out of the 32 Bôcher Prize winners; four out of the 29 Veblen

Prize winners; and seven out of the 48 Salem Prize winners.

Our empirical strategy requires us to determine if a mathematician is eligible for the

Fields Medal in any particular cycle, so that we need to observe the mathematician’s date of birth.

Although the AMS data does not provide this information, we ascertained the birth date (through

internet searches for each mathematician’s curriculum vitae or personal contact) for all Fields

medalists and for almost all of the potential contenders. The systematic archival of publications

by MathSciNet started in 1939, and some mathematicians in our sample published in their

teenage years, so we restrict the study to those born in or after 1920. Further, we exclude the six

potential contenders for whom we could not confirm a date of birth. This leaves us with a sample

of 47 medalists and 86 potential contenders. Appendix 1 presents the combined list of all winners,

a list that includes all the mathematicians mentioned in our historical survey.

There is obviously a great deal of variation in the mathematical significance and timing

of the work of the potential contenders. For example, the narrowness of the area prizes suggests

that the contribution of some of these winners, although very important in that particular area,

may not have the “breadth” required to generate sufficient interest in the broader community of

mathematicians. Similarly, some of the contenders (who perhaps went on to win one of the

general prizes) may have produced their best work after their eligibility for the Fields Medal

ended. Hence we whittle down the list of 86 potential contenders by examining how often other

mathematicians cite the work that the contenders produced during the years they were eligible

for the Fields Medal.

 Borjas and Doran 12

A recipient of the Fields Medal cannot have turned 40 after January 1 of the year in

which the medal is awarded. For example, the 2010 medal cycle would have been the last cycle

for a person born anytime between January 1, 1970 and December 31, 1973. Even though the

contenders born in this time frame did not win in their last shot at the medal, the incentives for

“impressing” the Fields Medal committee ended in 2010. Hence we assign the year 2010 as the

“medal year” for these contenders to separate the pre- and post-medal periods. We used a similar

exercise to ascertain the medal year for all the contenders in our sample. A mathematician’s

eligibility period is then given by the years between the mathematician’s first publication and the

medal year.

We calculate the annual rate of citations generated by a potential contender during his

eligibility period by dividing the total number of citations received by papers published in this

period (cumulative as of October 2011) by the number of years in the eligibility period. We then

define the final set of contenders as the 43 mathematicians in this group whose annual eligibility-

period citation rate is above the median. In other words, our final group of contenders represents

“la crème de la crème” of mathematicians who did widely recognized work during the eligibility

period and who did not win the Fields Medal.11

Table 1 reports summary statistics for the sample of Fields medalists as well as for the

control group (using both the final list of contenders with above-median citations, as well as the

group of all 86 prize winners).12 The table also reports comparable statistics (when available) for

a sample of “professional mathematicians,” which we define as the group of mathematicians in

the AMS archive whose first and last published papers span at least a 20-year period.

Obviously, both the Fields medalists and the contenders publish much more and receive

many more citations than the average mathematician. There is, however, relatively little
 Borjas and Doran 13

difference in measured productivity between the final group of contenders and the Fields

medalists. The medalists published 3.1 papers per year during their career, as compared to 3.6

papers for the contenders. The typical paper published by a medalist received 21.0 citations, as

compared to 17.5 citations for the contenders. The average mathematician in both groups was

born around 1950, and they each published their first paper at the early age of 23 or 24.

The table also summarizes the rate of output by age, calculating the average number of

papers published annually by the medalists and the contenders between the ages of 20-39 and 40-

59. The data reveal suggestive differences. The medalists and the contenders published

essentially the same number of papers per year in the early part of the career (3.4 papers), but the

medalists published 1.2 fewer papers per year after age 40. This striking pattern presages the

nature of the empirical evidence that will be documented in subsequent sections.

IV. The Fields Medal and the Age-Productivity Profile

As noted earlier, despite the plethora of important prizes that a brilliant mathematician

can potentially receive, the prestige of the Fields Medal is substantially greater than that of any

other prize. In fact, the ICM Fields Medal announcement emphasizes that the prestige effect is

far greater than the accompanying monetary award: “The Fields Medals carry the highest

prestige of all awards in mathematics. This prestige does not derive from the value of the cash

award, but from the superb mathematical qualities of the previous Fields Medal awardees” (ICM

2010).

Nevertheless, it is obvious that the financial impact of the Fields Medal on a

mathematician’s lifetime wealth is not limited to the $15,000 monetary prize. Fields medalists

are likely to see a substantial expansion in their opportunity set, in terms of high-quality job

offers, additional research funding, and many other career opportunities. It is conceivable,
 Borjas and Doran 14

therefore, that the wealth effect (which includes money as well as the additional opportunities

allowed by the substantial increase in professional prestige) could be sizable and could alter the

medalist’s post-medal behavior.13

The neoclassical labor-leisure model suggests that the wealth effect should increase the

consumption of leisure by the Fields medalists relative to that of the contenders. As a result, we

should not be surprised if the “weight of the prize” does indeed slow the Fields medalists down.

Moreover, the wealth effect might influence the mathematician’s choice of research topics, either

because the mathematician can now afford to explore topics that are essentially “consumption

goods” or because the medalist feels that he can pursue “riskier” topics. These shifts in research

interests, discussed in more detail below, also have productivity consequences.14

We initially measure the productivity of the elite mathematicians by the number of papers

published in each year. Figure 1 illustrates the life cycle trend in the average number of papers

published by both the medalists and the control group composed of the subset of contenders who

have above-median citations in the eligibility period. Specifically, the figure plots the average

number of papers published per year by the medalists and the contenders at the prime of their

career, relative to the medal year. The Fields medalists are plotted relative to the year they

actually received the medal (a zero on the x-axis represents the year of the prize); the contenders

are plotted relative to the last year of their eligibility for the medal (a zero on the x-axis

represents the last year of their eligibility).

It is evident that the medalists and the contenders had very similar age-productivity

profiles during the eligibility period, publishing around three to four papers per year. The figure

also shows, however, a dramatic drop in the annual rate of output for the medalists that coincides
 Borjas and Doran 15

with their receipt of the Fields Medal. A decade or two after the Fields medal, the average

medalist published around 1.5 fewer papers per year than the average contender.15

Of course, these differences could be due to factors that cannot be controlled for by the

graphical analysis, including individual fixed effects, calendar-year effects, and age differences.

We stack the annual data in our panel of medalists and contenders, and estimate the regression

model:

(1) yit = δi + δt + α Ti + β(Ti × Fi) + Zi γ + ε,

where yit gives the number of papers published by mathematician i in calendar year t; δi and δt

are vectors of individual and calendar-year fixed effects, respectively; T is a dummy variable

indicating if the observation refers to the post-medal period; F is a dummy variable indicating if

mathematician i won the Fields medal; and Z is a set of background characteristics that includes

the mathematician’s age (introduced as a fourth-order polynomial). The data panel contains one

observation for each mathematician for each year between the year of the first publication and

the most recent year of potential activity (if alive) or the year of death. The coefficient α

measures the difference in the annual rate of publication between the post- and pre-medal periods

for the contenders, while the coefficient β measures the relative change in this gap for the Fields

medalists.

Table 2 reports the estimated coefficients using a variety of alternative specifications.

The first two columns report coefficients when the control group is formed by the sample of

contenders with above-median citations in the eligibility period. Row 1 reports the simplest

regression model. The estimate of α is small, suggesting no substantial difference in the average
 Borjas and Doran 16

annual product of the contenders in the pre- and post-medal periods (after controlling for age and

period effects). The estimate of β is negative and around -1.4, indicating a (relative) drop of

more than one paper per year in the post-medal period for the Fields medalists. In other words,

even after controlling for individual-specific productivity differences and aging effects, there is a

sharp decline in the productivity of the medalists after they were awarded the medal.

The specification changes reported in the remaining rows of Table 1 corroborate this

finding. Row 2 uses the log number of papers per year as the dependent variable, but excludes

from the regression those (relatively few) years where the elite mathematicians did not publish at

all. The log papers regression shows a 24 percent decline in productivity in the post-medal period.

Row 3 reports the coefficient from a quantile regression where the dependent variable is the

median number of papers per year (using bootstrapped standard errors clustered at the

mathematician level). The regression shows a decline of -0.7 papers per year in the post-medal

period. Finally, the last two rows of the table use alternative methods for ascertaining the “medal

year” in the sample of contenders: either at age 36 (the median age at which Fields medalists

actually receive their medal) or age 40 (the maximum age of eligibility). The estimate of the

coefficient β is robust to these alternative definitions.

The last two columns of the table report the estimated coefficients when the control group

includes all 86 members of the sample of contenders (that is, all the prize winners without any

quality cutoff). Regardless of the method used to define the sample of contenders, the regression

coefficients are similar. The data reveals that the Fields medalists produce between 0.9 and 1.4

fewer papers per year in the post-medal period (or roughly a 20 percent decline in productivity)

than would be predicted either from their previous output or from the output of other great

mathematicians who did not win the highly coveted prize.16

 Borjas and Doran 17

The observed decline in the annual number of publications cannot be attributed to either

the effect of mechanical mean reversion or to an “expectation bias” among the members of the

award committee. Suppose that a mathematician’s observed productivity at a point in time has a

transitory component. As Lazear (2004) notes, the population of mathematicians whose

productivity is above some bar at time t will always show an average decline in productivity after

time t due to mean reversion in the transitory component of output. To avoid this type of

contamination, we restricted our analysis to contenders who had reached similarly high

productivity levels in the Fields Medal eligibility period. As a result, any mean reversion should

operate equally on both groups.

Second, the award committee members, containing some of the best mathematicians in

the world, surely observe future predictors of productivity among the contenders that we cannot

measure in a publication database, and they may be swayed by this private information in their

discussions. Given the future expectations bias, therefore, it would not be surprising if the

tournament winners do better in the post-tournament period. This bias would imply that our

regression coefficients understate the true post-medal productivity decline.

We suspect, however, that this bias is less likely to be important in the context of the

Fields Medal than in comparable tournaments in “softer” sciences, such as the John Bates Clark

Medal in economics. First, publication lags for ground-breaking papers in pure mathematics can

be considerably shorter than they are in economics. Second, many competing groups of

mathematicians are often working on the next great unsolved problems at once. Not surprisingly,

they often wait until the proof is sufficiently complete before discussing their techniques openly.

Finally, even the private information that a mathematician is out to prove Hilbert's Eighth

Problem is unlikely to influence the Fields Medal committee, as the resulting proof will either be
 Borjas and Doran 18

correct or not and the committee cannot determine the validity of the proof in advance. In

contrast, an interesting and fertile research agenda in applied economics can often produce

exciting papers regardless of the direction in which the data points.17

The AMS data also allows us to examine other output effects of the Fields Medal. Table

3 re-estimates the basic regression model using alternative dependent variables. The dependent

variable in row 1 is the probability that a mathematician publishes at least one paper in a given

year. The relative probability of publishing a paper falls by about 11.8 percentage points for the

medalists in the post-medal period. Row 2 uses the number of citations generated by papers

written in year t as the dependent variable. Although the AMS data only reports the post-2000

citations for a paper regardless of when the paper was published, the calendar year fixed effects

included in the regression model should control for the variation in citations between older and

newer papers.18 The coefficient β is again negative and significant, suggesting a decline of about

44 citations annually for papers produced in the post-medal period.

Part of the decline in citations is attributable to the fact that the medalists are less likely to

publish (and publish fewer papers when they do publish). Row 3 uses the number of citations per

paper published in a given year as the dependent variable (excluding years when the

mathematician did not publish at all). The material published by the medalists in the post-medal

period is, on average, less citation-worthy than the material published by the contenders. Even if

the typical post-medal paper written by a medalist generates fewer citations, the medalists may

be just as likely to hit a “home run.” We calculated a vintage-specific citation cutoff for papers

published each year using the universe of publications in the AMS database. By definition, a

mathematician hits a “home run” if the number of citations per paper published that year was

above the 99.5th percentile for all mathematicians in the AMS database. Row 4 shows that the
 Borjas and Doran 19

(relative) probability that a medalist hits a home run in the post-medal period declines by 15.6

percentage points. At the other extreme, a mathematician may “strike out” and write papers that

are never cited. Row 5 shows that the medalist’s (relative) probability of striking out rises by 5.3

percentage points in the post-medal period.

Finally, many of these elite mathematicians devote considerable time and effort to

training the next generation of mathematicians. In fact, biographies and laudations of their

achievements emphasize the training and mentoring of students as evidence of their long-lasting

impact on mathematics. We therefore also examine the impact of the Fields Medal on the

medalists’ mentoring activities.

We obtained access to the data in the Mathematics Genealogy Project (MGP), and we

merged the genealogy data with the AMS publication data. The MGP data identifies the

intellectual progeny of the renowned mathematicians in our sample, as well as the year in which

those students received their doctoral degree. We were able to match 104 of the 133

mathematicians using an MGP-AMS match provided by the administrators of the MGP. For 28

of the remaining 29 unmatched mathematicians, we were able to obtain information about the

graduation years and names of their students from name-based searches of their curriculum vitae,

obituaries, or unmatched online MGP entries. The merged data also provides information on the

research output of the students in their post-doctoral career.19 The merged data, therefore, allows

us to examine not only the impact of the Fields Medal on the number of students produced in the

post-medal period, but also the impact on the quality of the students.

We are interested in the relation between the timing of the year in which a student

becomes an elite mathematician’s mentee and the year in which the mathematician receives (or

does not receive) the Fields Medal. The mentoring agreement typically occurs two to four years
 Borjas and Doran 20

before the student obtains his or her doctoral degree. We lag the MGP degree date by three years

to approximate the year in which the mentoring relationship began.20

It turns out that Fields medalists are not only publishing fewer papers in the post-medal

period, and that those papers are relatively less important, but they are also accepting fewer

mentees under their wing. Row 6 of Table 3 reports the relevant coefficients when we estimate

the regression model using the number of mentees as the dependent variable. The regression

shows a (marginally significant) relative decline in the number of mentees accepted by the Fields

medalists of about 0.1 students per year. The last two rows estimate the regression model using

the mentee’s total number of publications and citations over their career to date as dependent

variables. The results show a pronounced decline in the quality-adjusted student output of Fields

medalists in the post-medal period.

An  Alternative  Control  Group  

It is important to establish that the evidence reported in Tables 2 and 3 cannot be

attributed to idiosyncratic properties of the method we used to construct the control group of

contenders. In fact, it is easy to show that the results are almost identical if we pursued a

drastically different method for constructing the sample of contenders. In addition to the various

prizes discussed above, the mathematics profession honors a select number of elite

mathematicians with invitations to be “plenary speakers” at the quadrennial International

Congress of Mathematicians. The qualifications required for receiving such an invitation are

clearly stated in the memo detailing the program committee’s responsibilities (ICM 2007):

Plenary lectures are invited one-hour lectures to be held without other parallel
activities…The lectures should be broad surveys of recent major developments,
aimed at the entire mathematical community. Plenary speakers should be
outstanding mathematicians and good lecturers. It is the privilege of the PC
[Program Committee] to select the plenary speakers.
 Borjas and Doran 21

The program committees in charge of organizing the plenary lectures have typically

scheduled around 20 plenary lectures at each ICM in recent decades. There have been 269

plenary speakers since the 1950 ICM, and these speakers include 34 of the 50 mathematicians

awarded the Fields medals during this period. We again obtained the lifetime publication record

for each of the invited plenary speakers from the AMS archives, and were able to determine the

year of birth for all but eight of the speakers. We then re-estimated the basic regression model in

equation (1) using the group of plenary speakers (who did not win the Fields medal) as the

control group.21

Table 4 summarizes the key results from our replication of the analysis. We illustrate the

robustness of the evidence by estimating alternative specifications of the regression model,

including using the group of all plenary speakers as contenders or only the group of plenary

speakers who received an invitation to present a lecture by age 44.22 In addition, the table

illustrates the robustness of the results when these samples are further restricted to only include

mathematicians whose citation rate during the eligibility period placed them in the top half of the

group. Regardless of the exact definition of the control group, the evidence summarized in Table

4 is similar to that discussed earlier (both qualitatively and quantitatively). Fields medalists

publish around one fewer paper annually after they receive the medal; the papers they do publish

get cited less often; and they have a lower probability of hitting a home run and a higher

probability of striking out. Because the evidence is robust to the definition of the control group,

the analysis presented in the remainder of this paper reverts to the simpler definition of a control

group that uses the sample of mathematicians who have received one of the six prestigious

awards noted earlier.

 Borjas and Doran 22

V. Cognitive Mobility

The post-medal productivity of Fields medalists, in terms of the number of papers

published, citations generated, or students mentored, is lower than would have been expected. A

question immediately arises: what exactly are the medalists doing with their time in the post-

medal period?

One obvious possibility is that the expansion in a mathematician’s opportunity set

implied by the Fields Medal introduces incentives to consume more leisure—along the lines of

the wealth effect in the neoclassical labor-leisure model. As long as leisure is a normal good, the

increase in the opportunity set associated with the Fields medal (which includes the value of the

additional prestige, job offers, grant opportunities, etc.) could lead to the medalists behaving in

the predicted fashion and increasing their consumption of leisure. The increased leisure leaves

less time for writing papers and supervising students.

In fact, the neoclassical labor-leisure model has a second implication: the Fields Medal

should increase his consumption of all normal goods. A medalist could respond by increasing his

consumption of “enjoyable research” in fields outside of pure mathematics, and perhaps begin to

dabble in such disciplines as biology and economics.23 Moreover, the medalist may now

perceive a freedom to pursue research topics that lead to riskier outcomes than he would have

pursued otherwise.24 These shifts in research interests may also affect productivity.

We apply the concept of “cognitive mobility” to analyze the choice of post-medal

research topics by the elite mathematicians in our sample. As noted in Borjas and Doran (2014),

knowledge producers who are conducting research on a particular set of questions may respond

to changed opportunities by shifting their time, effort, and other resources to a different set of
 Borjas and Doran 23

questions. Cognitive mobility then measures the transition from one location to another in idea

space.

We compare the cognitive mobility rates of the medalists and contenders in the post-

medal period. As noted earlier, the AMS data provides information not only on the annual output

(as measured by papers and citations) of mathematicians, but also categorizes each paper into

one of the 73 fields that make up the discipline of mathematics and related subjects. Because of

the large number of fields, it is obvious that we need to reduce the dimensionality of the space of

ideas in order to operationalize the concept of cognitive mobility in the current context.

Assume that a mathematician’s career begins the year he publishes his first paper. We

can then examine the distribution of a mathematician’s research topics in, say, the first x years of

his career. The AMS data allows us to determine the modal field of the papers published in those

years, as the mathematician was getting his career started and signaling his “quality” to the rest

of the profession.

Although mathematics is composed of 73 fields, some of these fields are intellectually

close to the modal field, while others are unrelated. The notion of cognitive mobility, therefore,

should incorporate the fact that a move between the modal field and any other field may be

“cognitively close” or “cognitively far.” To determine the cognitive distance between any two

fields, we calculated a matrix with elements [fij] showing the fraction of references made by

papers published in field i to papers published in field j.25 To illustrate, suppose that the modal

field was Partial Differential Equations. The three most closely related fields (with the three

largest values of fij) are Partial Differential Equations itself, Global Analysis, and Fluid

Mechanics. These three fields account for 72 percent of all references made by papers published
 Borjas and Doran 24

in Partial Differential Equations. At the other extreme, papers published in Partial Differential

Equations never referenced papers published in either General Algebraic Systems or K-Theory.

It turns out that we typically do not need to expand the definition of “cognitively close”

beyond 15 fields to capture almost all the references made by papers published in field i. For

example, 77.6 percent of all references made in Partial Differential Equations are to the top 5

fields, 87.4 percent are to the top 10, and 92.5 percent are to the top 15. This clustering around a

very small number of fields is quite representative of the discipline of mathematics. In particular,

93.1 percent of the references in papers published in the median field of mathematics are made to

papers published in only 15 other fields (the respective statistics for the 10th and 90th percentile

fields are 87.6 and 97.5 percent).

Of course, it is not uncommon for elite mathematicians to move within a small (and often

related) set of fields in the early part of their career. To capture this oscillation, we expand the

definition of the “modal field” to include either the most common or the second most common

field in the early part of a mathematician’s career. For each of these two modes, we then

constructed the set of the 15 most related fields. Our cognitive mobility variable then indicates if

the mathematician moved outside the two modal fields and all related fields (in other words, if

the mathematician moved out of the potential maximum of 30 fields that broadly define his

initial research interests or “comfort zone”).

To easily illustrate the trends in cognitive mobility, we first define the “early career

period” as the eligibility period for the Fields Medal (that is, the years before the Fields medalist

won the medal or the years in which contenders were eligible for the medal). We then calculate

the probability that papers published in each year of a mathematician’s career are outside the

comfort zone (or in a different field than the two modal and related fields). Figure 2 plots this
 Borjas and Doran 25

measure of cognitive mobility. As before, Fields medalists are plotted relative to the year they

actually received the medal, and the contenders are plotted relative to the last year of eligibility.

The probability that either the medalists or contenders strayed from their comfort zone

prior to the medal year is small, around 5 percent a year for mathematicians in either group. This

similarity, however, breaks down dramatically in the post-medal period. The rate of cognitive

mobility doubled to 10 percent for the contenders, but rose dramatically for the medalists,

quintupling to 25 percent. In short, the data reveal that the awarding of the Fields Medal is

associated with a strong increase in the likelihood that a mathematician tries out fields that are

very distant from those fields that established his reputation.

To determine if this correlation persists after controlling for individual and period fixed

effects, we use the AMS data to construct a panel where an observation represents a paper

published by each mathematician. In particular, let pint be an indicator variable set to unity if the

field of the nth paper published by mathematician i (and published in year t) differs from that of

the modal and related papers in the baseline period. We then estimate the regression model:

(2) pint = δi + δt + α Ti + β(Ti × Fi) + Zi γ + ε.

Table 5 reports the relevant coefficients (α, β) using several alternative specifications. As

with the illustration in Figure 2, the first row of the table uses the publications in the eligibility

period to define the set of fields that make up the mathematician’s comfort zone. We illustrate

the robustness of our results by using either the two modal fields (and up to 30 related fields), or

just simply the modal field (and its 15 related fields). Regardless of the specification, the

awarding of a Fields medal substantially increases the rate of cognitive mobility. Even after
 Borjas and Doran 26

controlling for individual-specific fixed effects, the awarding of the Fields Medal increases the

probability of a move by between 16 and 21 percentage points.

The next two rows of the top panel conduct sensitivity tests by using alternative

definitions of the “early career” period used to construct the mathematician’s comfort zone. Row

2 uses the first three years of the career, while row 3 uses the first five years. Similarly, the

regressions in Panel B use the entire sample of contenders (without any quality cutoff) to

estimate the model. Regardless of the specification, the awarding of the Fields Medal has a

positive and significant impact on the probability that a mathematician engages in cognitive

mobility.26

The freedom to try out new things, however, does not come cheap. Cognitive mobility,

like any other type of move, can be costly. The mathematician is exiting a field where he has

remarkable technical skills and attempting to prove theorems in areas where his intuition may not

be as strong and where the proofs may require a new set of tools. It would not be surprising,

therefore, it if takes longer to produce a paper after the mathematician has engaged in cognitive

mobility.

Define the duration of a “preparation spell” as the length of time elapsed (in years)

between any two consecutive papers in a mathematician’s career. We estimated a regression

model to measure the relation between the length of the preparation spell for paper n (n > 1) and

cognitive mobility:

(3) πint = δi + δt + λF (Fi × pint) + λC (Ci × pint) + Zi γ + ε,

 Borjas and Doran 27

where πint gives the length of the preparation spell required to write paper n; pint is the indicator

variable set to unity if paper n involved a cognitive move, and Ci is a variable indicating if

mathematician i is a contender (Ci = 1 – Fi). Table 6 summarizes the estimates of the vector (λF,

λC) using alternative specifications of the model. It is evident that cognitive mobility is

associated with a longer preparation spell for both the medalists and the contenders, and the

effect is numerically important. A cognitive move increases the length of the preparation spell by

between 0.16 and 0.23 years.

In sum, the data indicate that the Fields medalists engaged in more cognitive mobility in

the post-medal period and that cognitive mobility imposes a cost; it takes longer to produce a

paper. This behavior, therefore, will inevitably result in a reduced rate of publication for the

medalists in the post-medal period.

The regression coefficients can be used to conduct a back-of-the-envelope calculation

that determines how much of the observed decline in productivity was due to cognitive mobility.

The results in Table 5 indicate that the awarding of the Fields Medal increased the probability of

cognitive mobility for a paper published in the post-medal period by around 15 percent. Both the

medalists and the contenders published four papers per year at the time the medal was awarded

(see Figure 1). Using this rate of output as the baseline, the regression coefficient in Table 5

indicates that the awarding of the medal led to a 0.6 increase (or 0.15 × 4) in the number of

papers published annually in an unfamiliar field.

At the same time, Table 6 shows that cognitive mobility increases the length of a

preparation spell by about 0.2 years. Putting these results together implies that the increased

incentive for cognitive mobility in the post-medal period and the longer preparation spell reduces

the amount of “effective” time available in a given year by about 0.12 years (or the 0.6 papers
 Borjas and Doran 28

published in an unfamiliar field times the 0.2 longer years it takes to produce such a paper). In

rough terms, therefore, we expect a 12 percent decline in the number of papers that a medalist

published annually in the post-medal period simply because cognitive mobility diverts 12 percent

of his time to other uses (such as learning new skills). As we saw in Table 2, there was a 24

percent decline in annual output. The increased experimentation exhibited by Fields medalists in

the post-medal period can account for about half of the decline in productivity.

It is important to emphasize that the decline in productivity resulting from the wealth

effect that increases leisure is conceptually different from the decline induced by the increased

experimentation. Although the cognitive mover publishes fewer papers, those papers may

provide a social benefit. The medalist is applying his talents to unfamiliar questions, and may

generate important insights in areas that were previously under-served by exceptional

mathematical talent.

In fact, among the great architects of late twentieth century mathematics in our sample,

there are three well-known examples in which a Fields medalist who made extraordinary

contributions to a specific area of pure mathematics went on to mathematize a distant applied

subject later in their career. René Thom (and his development of singularity/catastrophe theory),

David Mumford (and the mathematics of vision and pattern theory), and Stephen Smale.

Smale’s experience is particularly illuminating. Gleick (1987, p. 45) recounts what

happened soon after Smale proved a pure mathematics result (the Generalized Poincaré

conjecture) that helped earn him the Fields Medal.

Smale [was]. . . already famous for unraveling the most esoteric problems of
many-dimensional topology. A young physicist, making small talk, asked what
Smale was working on. The answer stunned him: “Oscillators.” It was absurd.
Oscillators—pendulums, springs, or electrical circuits—were the sort of problem
that a physicist finished off early in his training. They were easy. Why would a
great mathematician be studying elementary physics?
 Borjas and Doran 29

Even if a young physicist considered Smale’s new choice of topic simplistic and absurd, the

enormity of the mathematician’s previous achievements insulated him from any real loss of

prestige. Smale’s post-medal experimentation built the mathematical foundation of chaos theory.

In fact, Smale went on to make important contributions in biology, astronomy, and even in

theoretical economics.

VI. Summary

A vast literature explores the impact of tournaments, contests, and prizes on the

productivity of tournament participants, analyzing the implications of pre-award productivity

effects for the efficient design of incentive mechanisms. A working assumption in this literature

is that the labor supply consequences of actually winning a tournament are minimal. This paper

studies the impact of winning a tournament on the productivity and effort choices of tournament

participants in the post-tournament period.

We examine how winning the Fields Medal affects the post-medal productivity and

research choices of mathematicians. The Fields Medal is the most prestigious award in

mathematics, awarded every four years to mathematicians under the age of 40. Using archival

data from the American Mathematical Society and the Mathematics Genealogy Project, we

document the shape of the age-productivity profile of these exceptional mathematicians along a

number of different dimensions, including the number of papers published, citations received,

and students mentored. We find that the age-productivity profile of the Fields medalists and of

the losing contenders is similar until the year in which a particular mathematician wins the Fields

Medal (or does not win it). Remarkably, the rate of output of the Fields medalists declines

noticeably in the post-medal period.

 Borjas and Doran 30

We also show that the medalists exhibit a far greater rate of cognitive mobility in the

post-medal period, pursuing topics that are far less likely to be related to their pre-medal work.

Because cognitive mobility is costly (additional time is required to prepare a paper in an

unfamiliar field), the increased rate of mobility reduces the medalists’ rate of output in the post-

medal period. The data suggest that about half of the decreased productivity in the post-medal

period can be attributed to the increased propensity for experimentation.

Hundreds of scientific and technical prizes are awarded around the world. Our evidence

suggests that the post-prize productivity impact of winning a prestigious award can be substantial,

affecting both the quantity and type of research the winners produce. Although some Fields

Medalists may have taken Hermann Weyl’s words to heart and “carried on as they began,” this

was not the typical outcome. The data instead reveal that the increased opportunities provided by

the Fields Medal, in fact, discouraged the recipients from continuing to produce the pure

mathematics that the medal was awarded for, while encouraging time-consuming investments in

ever more distant locations in the space of ideas.

 Borjas and Doran 31

Appendix 1. List of Fields medalists and contenders

Anantharaman, Nalini, S Jones, Peter, S Rosenlicht, Maxwell, C

Arnold, Vladimir, W Journé, Jean-Lin, S Roth, Klaus, F
Artin, Michael, W Keller, Joseph, W Sato, Mikio, W
Aschbacher, Michael, W,C Kenig, Carlos, S Schoen, Richard, B
Astala, Kari, S Kirby, Robion, V Schramm, Oded, S
Atiyah, Michael, F,A Klartag, Boáz, S Seidel, Paul, V
Avila, Artur, S Kontsevich, Maxim, F Serre, Jean-Pierre, F,A,W
Baker, Alan, F Konyagin, Sergei, S Shelah, Saharon, W
Beckner, William, S Körner, Thomas, S Shishikura, Mitsuhiro, S
Bombieri, Enrico, F Lacey, Michael, S Simons, James, V
Borcherds, Richard, F Lafforgue, Laurent, F Sinaĭ, Yakov, W
Bott, Raoul, W Lang, Serge, C Singer, Isadore, A
Bourgain, Jean, F,S Langlands, Robert, W Smale, Stephen, F,W,V
Brown, Morton, V Lax, Peter, A,W Smirnov, Stanislav, F,S
Caffarelli, Luis, W,B Lindenstrauss, Elon, F,S Soundararajan, Kannan, S
Calderón, Alberto, W Lions, Pierre-Louis, F Stallings, John, C
Carleson, Lennart, A,W Lovász, László, W Stein, Elias, W
Chandra, Harish, C Lusztig, George, C Sullivan, Dennis, W,V
Cohen, Paul, F,B Margulis, Grigory, F,W Swan, Richard, C
Connes, Alain, F Mazur, Barry, V Szemerédi, Endre, A
Dahlberg, Björn, S McMullen, Curtis, F,S Tao, Terence, F,B,S
David, Guy, S Melrose, Richard, B Tataru, Daniel, B
de Jong, Aise Johan, C Meyer, Yves, S Tate, John, A,W
Deligne, Pierre, F,A,W Milnor, John, F,A,W Taubes, Clifford, V
Donaldson, Simon, F Montgomery, Hugh, S Thiele, Christoph, S
Drinfeld, Vladimir, F Mori, Shigefumi, F,C Thom, René, F
Faltings, Gerd, F Moser, Jürgen, W Thompson, John, F,A,W,C
Fefferman, Charles, F,B,S Mostow, George, W Thurston, William, F,V
Feit, Walter, C Mumford, David, F,W Tian, Gang, V
Freedman, Michael, F,V Naor, Assaf, B,S Tits, Jacques, A,W
Furstenberg, Hillel, W Nazarov, Fedor, S Tolsa, Xavier, S
Gowers, Timothy, F Ngô, Báo Châu, F Varadhan, S.R., A
Green, Ben, S Nikishin, Evgeniĭ , S Varopoulos, Nicholas, S
Griffiths, Phillip, W Nirenberg, Louis, B Venkatesh, Akshay, S
Gromov, Mikhael, A,W,V Novikov, Sergei, F,W Villani, Cédric, F
Grothendieck, Alexander, F Okounkov, Andrei, F Voevodsky, Vladimir, F
Hacon, Chistopher, C Ornstein, Donald, B Volberg, Alexander, S
Herman, Michael-Robert, S Ozsváth, Peter, V Werner, Wendelin, F
Hironaka, Heisuke, F Perelman, Grigori, F Wiles, Andrew, W
Hirzebruch, Friedrich, W Petermichl, Stefanie, S Witten, Edward, F
Hochster, Melvin, C Piatetski-Shapiro, Ilya, W Wolff, Thomas, S
Hörmander, Lars, F,W Pichorides, Stylianos, S Wooley, Trevor, S
Hunt, Richard, S Pisier, Giles, S Yau, Shing-Tung, F,W,V
Jones, Vaughan, F Quillen, Daniel, F,C Yoccoz, Jean-Christophe, F,S
Zelmanov, Efim, F
 Borjas and Doran 32

Notes: The superscripts indicate the prize awarded to the mathematician; F = Fields Medal; A =

Abel Prize; W = Wolf Prize; C = Cole Algebra Prize; B = Bôcher Prize; V = Veblen Prize; and S

= Salem Prize.
 Borjas and Doran 33

References

American Mathematical Society. 2008. "Leroy P. Steele Prize for Lifetime Achievement, 2008
Steele Prizes." Notices American Mathematical Society 55(4): 486-90.

Anderson, Rolph E., Alan J. Dubinsky, and Rajiv Mehta. 1999. “Sales Managers: Marketing’s
Best Example of the Peter Principle?” Business Horizons 42(1): 19–26.

Arrow, Kenneth J. 1965. Some Aspects of the Theory of Risk Bearing. Helsinki: Yrjö Jahnssonin
Säätiö.

Azoulay, Pierre, Joshua S. Graff Zivin, and Gustavo Manso. 2011. "Incentives and Creativity:
Evidence from the Academic Life Sciences." The RAND Journal of Economics 42(3): 527-54.

Barmby, Tim, Barbara Eberth, and Ada Ma. 2012. “Incentives, Learning, Task Difficulty, and
the Peter Principle: Interpreting individual output changes in an Organisational Hierarchy.”
Labour Economics 19(1): 76-81.

Borjas, George J., and Kirk B. Doran. 2012. “The Collapse of the Soviet Union and the
Productivity of American Mathematicians.” Quarterly Journal of Economics 127(3): 1143-1203.

Borjas, George J., and Kirk B. Doran. 2014. “Cognitive mobility: Labor market responses to
supply shocks in the space of ideas.” Journal of Labor Economics. Forthcoming.

Bricongne, Jean-Charles. 2014. "Do Prizes in Economics Affect Productivity?" Unpublished

Manuscript, Bank of France.

Carter, R. W. 2006. “A Survey of the Work of George Lusztig.” Nagoya Mathematical Journal
182: 1-45.

Chan, Ho Fai, Bruno S. Frey, Jana Gallus, and Benno Torgler. 2013. “Does the John Bates Clark
Medal boost subsequent productivity and citation success?” Working Paper No. 2013-0. Zurich:
CREMA.

Chang, Kenneth, 2008. “Oded Schramm, 46, Mathematician, Is Dead.” New York Times,
September 11.

Ehrenberg, Ronald G., and Michael L. Bognanno. 1990. "Do Tournaments Have Incentive
Effects?" Journal of Political Economy 98(6): 1307-24.

Gleick, James. 1987. Chaos: Making a New Science. New York: Viking-Penguin.

International Congress of Mathematicians, Proceedings of the International Congress of

Mathematicians (1936, 1954, 1966, 1970, 1974, 1978, 1986, 1994, 1998, 2006, 2010).
 Borjas and Doran 34

International Congress of Mathematicians, 2007. “Guidelines for the Program Committee (PC)
and the Organizing Committee (OC).” November 21; www.mathunion.org/fileadmin/IMU/PC-
OC-Guidelines.pdf.

Knoeber, Charles R., and Walter N. Thurman. 1994. “Testing the Theory of Tournaments: An
Empirical Analysis of Broiler Production.” Journal of Labor Economics 12(2): 155-79.

Langlands, R. P. 1985. “Harish-Chandra. 11 October 1923-16 October 1983.” Biographical

Memoirs of Fellows of the Royal Society 31: 198-225.

Lazear, Edward P. “The Peter Principle: A Theory of Decline.” Journal of Political Economy
112(1): S141-S163.

Lazear, Edward P., and Sherwin Rosen. “Rank-Order Tournaments as Optimum Labor
Contracts.” Journal of Political Economy 89(5): 841-64.

Levhari, David, and Yoram Weiss. 1974. “The Effect of Risk on the Investment in Human
Capital.” American Economic Review 64(6): 950-63.

Levin, Sharon G., and Paul E. Stephan. 81. “Research Productivity Over the Life Cycle:
Evidence for Academic Scientists.” American Economic Review 81(1): 114-32.

Main, Brian G. M., Charles A. O'Reilly III, and James Wade. 1993. “Top Executive Pay:
Tournament or Teamwork?” Journal of Labor Economics 11(4): 606-28.

Monastyrsky, Michael. 2001. “Some Trends in Modern Mathematics and the Fields Medal.”
Notes of the Canadian Mathematical Society 33(2): 3-5 and 33(3): 11-13.

Rosen, Sherwin. 1986. "Prizes and Incentives in Elimination Tournaments." American Economic
Review 76(4): 701-15.

Scotchmer, Suzanne. 2006. Innovation and incentives. Cambridge, MA: MIT Press.

Solholm, Rolleiv. 2010. "The Abel Prize to John T. Tate," Norway Post, March 25.

Sparshott, Jeffrey. 2014. "Do Prizes Dent Productivity?" Real Time Economics, Wall Street
Journal, April 14; blogs.wsj.com/economics/2014/04/18/do-prizes-dent-productivity/.

Stern, Scott. 2004. "Do Scientists Pay to Be Scientists?" Management Science 50(6): 835-53.

Stiglitz, Joseph E. 1969. “The Effects of Income, Wealth, and Capital Gains Taxation on Risk-
Taking.” Quarterly Journal of Economics 83(2): 263-83.

Tropp, Henry S. 1976. "The Origins and History of the Fields Medal." Historia Mathematica
3(2): 167-81.
 Borjas and Doran 35

Zuckerman, Harriet. 1996. Scientific Elite: Nobel Laureates in the United States. New
Brunswick, NJ: Transaction Publishers.
 Borjas and Doran 36

 
ENDNOTES

1. Lazear and Rosen (1981) and Rosen (1986) give the classic presentations of the tournament

model. Empirical evidence on the productivity effects includes Ehrenberg and Bognanno (1990),

Knoeber and Thurman (1994), and Main, O’Reilly, and Wade (1993).

2.  Some recent studies address this question in the context of job promotions. Lazear (2004)

offers an important discussion of the statistical problems introduced by mean reversion in the

transitory component of productivity when measuring post-promotion productivity effects; see

also the related empirical work in Anderson, Dubinsky, and Mehta (1999), and Barmby, Eberth,

and Ma (2012).  

3. Zuckerman (1996) documents that the research output of Nobel Prize winners declines after

winning the prize. Her descriptive evidence, however, is likely contaminated by the late age of

the winners and the possibility of mean reversion because the comparison group is less

productive prior to the awarding of the prize. Chan et al. (2013) and Bricongne (2014) present

related studies of the impact of the John Bates Clark Medal on the productivity of economists

and reach conflicting results. The productivity impact of the Clark medal is discussed in more

detail below. Finally, Azoulay, Graff-Zivin, and Manso. (2011) explore the impact of funding at

the start of a scientist's career on subsequent productivity.

4. Partly due to jealousy and conflict between Alfred Nobel and the Swedish mathematician

Magnus Gotha Mittag-Leffler, Nobel famously left mathematics out of his list of recognized

disciplines when he founded the prize that bears his name (Tropp 1976). Ironically, Fields and

Mittag-Leffler were close friends.

 Borjas and Doran 37

 
5. The age restriction has been applied consistently over time. For example, the 1998 ICM

stated: "As all the Committees before us, we agreed, . . . to follow the established tradition and to

interpret the word 'young' as 'at most forty in the year of the Congress’" [emphasis added].

6. The goal of the French Bourbaki group was to write down all of mathematics as a linear

development from general axioms. The arbitrariness of the decision to exclude Harish-Chandra

is doubly ironic: “Harish-Chandra would have been as astonished as we are to see himself

lumped with Thom and accused of being tarred with the Bourbaki brush, but whether he would

have been so amused is doubtful, for it had not been easy for him to maintain confidence in his

own very different mathematical style in face of the overwhelming popular success of the French

school in the 1950s” (Langlands 1985, p. 212).

7. An alternative to the difference-in-differences approach would be the use of instrumental

variables. The obvious choice of an instrument for winning the Fields Medal is given by the

combination of the quadrennial timing of the award and the age cut-off: some mathematicians

have almost four more years to compete for the Fields Medal than others. While the resulting

variation in the maximum number of “eligible work years” is positively related with winning the

Fields Medal in the sample of contenders constructed below, the Angrist-Pischke multivariate F-

tests of excluded instruments show that the relationship is not sufficiently strong to make it a

useful instrument by itself.

8. An alternative empirical strategy would be to rely on statistical matching based on papers per

year or citations per paper to construct a sample of contenders from the universe of all

mathematicians. We did not pursue this alternative approach for two reasons. First, it would be

difficult to operationalize because a key variable in the matching algorithm would be year of

birth (which determines eligibility for the Fields Medal). The birth year would need to be
 Borjas and Doran 38

 
uncovered one at a time through archival research or one-on-one contact for the universe of

potential matches in the AMS data. It turns out that approximating year of birth by year of first

publication leads to a very poor approximation of the end of the eligibility period. Second, we

prefer the approach of actually constructing a relevant group of contenders with “real” people

because it uses the valuable information embedded in the profession’s willingness to publicly

acclaim a particular person’s contributions with a scarce award (or, as discussed below, with an

invitation to present a prestigious lecture). This type of construction likely leads to a better

approximation of the “short list” of contenders considered by the prize committee.

9. The Abel Prize creates a multi-stage tournament for mathematicians. As in Rosen (1986),

Fields medalists may wish to keep participating in the tournament in order to receive the sizable

monetary award associated with the Abel Prize. The changed incentives, however, are unlikely to

affect our results because the Abel Prize began late in the sample period.

10. There are many other mathematical prizes around the world, but they are far less prestigious

or worse predictors of winning the Fields Medal. For example, none of the winners of the AMS

Cole Prize for Number Theory has ever gone on to win the Fields Medal.

11. An alternative way of defining the final set of contenders would be to calculate the annual

rate of papers published during the eligibility period, and select the 43 mathematicians whose

annual rate of output is above the median. The evidence reported below is similar if we used this

alternative definition. We also accounted for the fact that the nature of the AMS citation data

could imply that the number of citations received by more recent mathematicians may be greater

than the number received by mathematicians active in the 1950s and 1960s. We defined the final

set of contenders based on the above-median ranking of the residual from a regression of the per-
 Borjas and Doran 39

 
year number of citations in the eligibility period on the calendar year of first publication, and the

results are very similar to those reported in the next two sections.

12. We ignore the posthumous publications of the Fields medalists and contenders.

13.  Data  on  the  impact  of  the  Fields  medal  on  a  winner’s  financial  wealth  are  obviously  very  

limited.  Nevertheless,  there  is  some  evidence  suggesting  that  the  effect  may  be  quite  large.  

In  particular,  it  is  possible  to  examine  the  salary  history  of  Fields  medalists  employed  at  a  

few  public  universities  (where  the  salary  information  is  publicly  released).  Our  

examination  of  the  salary  history  of  one  such  winner,  employed  at  a  large  public  university  

in  the  western  United  States,  indicates  that  the  medalist  experienced  a  $120,000  salary  

increase  (equivalent  to  a  67  percent  raise)  between  the  academic  years  preceding  and  

subsequent  to  the  award.  The  medalist’s  salary,  in  fact,  continued  to  increase  rapidly  in  

subsequent  years  and  was  five  standard  deviations  above  the  mean  salary  of  full  professors  

in  the  department  six  years  after  the  award.  This  episode  suggests  that  the  impact  of  the  

Fields  medal  on  the  present  value  of  lifetime  wealth  (even  when  income  is  restricted  only  

to  academic  earnings)  can  be  substantial,  totaling  in  the  millions  of  dollars.  

14.  There  are  two  additional  types  of  productivity  effect  that  may  be  important,  though  

hard  to  measure.  First,  winning  the  Fields  medal  “raises  the  bar”  in  terms  of  what  a  

medalist  perceives  to  be  publishable  output  (and  also  in  terms  of  what  the  mathematical  

community  expects  from  him).  In  particular,  the  medalist  may  feel  that  he  has  a  brand  

name  to  protect  and  is  unwilling  to  devote  his  time  or  effort  to  smaller  topics  that  could  

well  lead  to  publishable,  but  not  seminal,  papers.  Second,  the  medalist  is  likely  to  be  

distracted  by  competing  offers  for  advice,  speeches,  etc.  As  Cédric  Villani,  a  Fields  medalist  

in  2010,  puts  it,  “Every  Fields  Medal  winner  knows  how  much  the  productivity  decreases  
 Borjas and Doran 40

 
after  the  award.  .  .  Just  because  you  are  so  much  invited  and  wanted  by  everybody—for  

juries,  ceremonies,  public  speeches,  TV,  radio,  newspapers”  (Sparshott  2014).  

15.  Figure  1  seems  to  show  a  small  decline  in  productivity  among  the  Fields  Medalists  for  a  

few  years  prior  to  the  receipt  of  the  medal.  This  decline  is  partly  an  artifact  of  the  three-­‐

year  moving  average  used  in  the  figure.  In  fact,  the  only  unusually  low  year  of  output  

actually  occurs  exactly  one  year  before  receipt  of  the  Fields  Medal.  The  number  of  papers  

published  two  years  before  receipt  of  the  Fields  Medal  and  during  the  year  of  the  medal  

itself  is  either  on  or  above  the  trend.  We  further  examined  the  pre-­‐trend  issue  by  

estimating  a  regression  specification  that  included  a  variable  “years-­‐since-­‐medal”  (both  

before  and  after)  and  interacted  this  variable  with  the  indicator  of  whether  the  

mathematician  won  the  Fields  medal.  We  aggregated  the  years-­‐since-­‐medal  variable  into  

three-­‐year  intervals  to  obtain  a  sufficient  sample  size  for  each  interaction.  The  regression  

revealed  that  only  one  of  the  pre-­‐medal  interactions  was  significant  (and  it  was  the  one  

indicating  the  period  at  the  beginning  of  the  career),  and  none  of  the  others  were  even  

marginally  significant.  In  contrast,  three  of  the  post-­‐medal  interactions  are  significant,  with  

four  more  being  marginally  significant.  The  data,  therefore,  strongly  suggest  that  the  small  

decline  in  productivity  observed  just  before  the  awarding  of  the  Fields  medal  is  not  part  of  

a  consistent  pre-­‐treatment  trend  of  low  output  for  the  treated  group.  

16.  We also examined the robustness of our results to the use of the Abel and Wolf prizes in the

construction of the sample of contenders in two alternative ways: by excluding from the sample

of contenders those mathematicians who won the Abel or Wolf prizes (yielding a sample of 35

contenders) or by only including in the sample of contenders those mathematicians who did win
 Borjas and Doran 41

 
the Abel or Wolf prizes (8 contenders). The interaction coefficient β  is  -­‐0.908  (with  a  standard  

error  of  0.458)  in  the  first  case  and  -­‐2.553  (2.015)  in  the  second.  

17.  During  the  two-­‐year  period  in  which  we  prepared  this  study,  two  related  working  

papers  examining  the  productivity  effect  of  the  Clark  medal  were  also  produced.  The  

evidence  in  these  studies,  however,  is  mixed.  Chan  et  al.  (2013)  use  a  statistical  matching  

approach  and  find  a  favorable  productivity  effect,  while  Brincogne  (2014)  constructs  a  

control  group  composed  of  prominent  economists  and  finds  that  the  favorable  productivity  

effect  disappears  once  the  regression  controls  for  individual-­‐specific  fixed  effects  in  

productivity,  suggesting  that  the  net  productivity  effect  of  the  medal  may  be  slightly  

negative.  Even  if  the  evidence  were  less  ambiguous,  there  are  a  number  of  issues  that  

complicate  the  interpretation  of  the  results  from  the  Clark  medal  studies.  In  particular,  the  

Clark  medal  is  not  the  last  stage  of  the  tournament;  there  may  be  well  be  “forward  

expectation  bias”  influencing  the  choice  of  the  next  Clark  medalist;  and  the  additional  grant  

funding  that  a  Clark  medalist  will  likely  generate  will  be  an  input  in  the  production  function  

for  subsequent  research  (grant  funding  is  a  much  less  important  input  in  the  production  of  

pure  mathematics).  

18.  The  fact  that  our  citation  data  only  includes  citations  made  after  2000  (regardless  of  

when  the  paper  was  published)  prevents  us  from  addressing  the  possibility  that  receipt  of  

the  Fields  medal  may  itself  cause  the  medalists'  pre-­‐medal  work  to  become  more  highly  

cited  in  the  post-­‐2000  period  than  it  otherwise  would  have  been.  A  detailed  analysis  of  this  

issue  would  require  much  more  detailed  (and  contemporaneous)  citation  data.  

19. For students whose AMS identification numbers are listed in the MGP database, we use the

AMS data to calculate their career papers and citations. Many mathematics doctorates, however,
 Borjas and Doran 42

 
do not publish a single paper in their career (and the mode for those who do is a single

publication with zero citations; see Borjas and Doran 2012). The absence of a publication

implies that the student will never appear in the MathSciNet database. We assume that the

students who do not have an AMS identification number have zero lifetime publications and

citations.

20.  The results are similar as long as we lag the degree date by two or more years.  

21.  It  is  interesting  to  note  that  39  of  the  86  contenders  produced  by  the  method  based  on  

receipt  of  mathematical  prizes  have  also  given  a  plenary  lecture,  as  have  21  of  the  43  “final”  

contenders  with  an  above-­‐median  citation  rate  in  the  eligibility  period.  

22.  By  restricting  the  sample  to  plenary  speakers  who  gave  their  lecture  before  age  44,  we  

are  indirectly  attempting  to  construct  a  sample  of  outstanding  mathematicians  that  

includes  those  who  “just  missed”  their  chance  for  a  Fields  medal  because  of  the  interaction  

between  the  age  restriction  and  the  quadrennial  timing  of  the  award.    

23.  Levin and Stephan (1991) and Stern (2004) suggest that some scientists, particularly

theoretical ones, derive consumption value from doing research they enjoy.  

24.  Since Arrow (1965) and Stiglitz (1969), it is well known that a wealth increase may prompt a

utility-maximizing agent to undertake riskier investments. Levhari and Weiss (1974) extended

this insight to the human capital framework. As 2010 Fields medalist Stanislav Smirnov notes,

the pursuit of riskier investments raises the possibility that the medalist may be “doing  

something  very  important  in  the  end"  (Sparshott  2014).  

25. It is not possible to estimate this matrix with the data that the AMS provided us. We instead

purchased citation data from the ISI Web of Science to calculate these distance measures; see the
 Borjas and Doran 43

 
Data Appendix to Borjas and Doran (2012) for details. The calculation of the matrix uses all

publications in the 1979-2009 period.

26. We also examined the probability that a mathematician conducts research outside pure

mathematics. Specifically, we constructed a variable indicating if a paper was in any of the

following “applied” areas: “History and biography,” “Statistics,” “Computer science,”

“Geophysics,” “Game theory, economics, social and behavioral sciences,” “Biology and other

natural sciences,” “Information and communication, circuits,” or “Mathematics education.” The

estimated regression coefficient suggests that the (relative) probability of a Fields medalist

publishing in one of these applied areas in the post-medal period rose by 75 percent.
 Borjas and Doran 44

Table 1
Summary Statistics

Contenders with
All Fields above-median All
Variable: mathematicians medalists citations contenders
Lifetime papers 31.8 116.5 126.4 106.4

Papers per year 0.9 3.1 3.6 2.7

20-39 years old --- 3.3 3.4 2.7
40-59 years old --- 2.9 4.1 3.0

Lifetime citations 93.6 2451.9 2213.5 1640.4

Citations per year 2.5 64.0 56.0 39.1

Citations per paper 1.8 21.0 17.5 15.6

Year of birth --- 1949.5 1953.7 1946.2

Year of first publication 1972.6 1972.6 1977.7 1970.8

Age at first publication --- 23.1 24.0 24.6

Deceased (percent) --- 10.6 14.0 20.9

Age at death --- 74.0 60.5 66.3

Number of mathematicians 72,140 47 43 86

Notes: The summary statistics for “all mathematicians” are calculated using the group of

mathematicians in the AMS database who had at least 20 years of experience before ending their

publication career; in other words, those whose most recent publication is at least 20 years after

their first publication. The group of “contenders” is composed of persons who were awarded at

least one of six other mathematics prizes (the Abel, Wolf, Cole Algebra, Bôcher, Veblen, and

Salem Prizes), but were not awarded the Fields Medal. The group of contenders with “above-

median citations” is composed of the contenders who had above-median per-year citations
 Borjas and Doran 45

during the eligibility period for the Fields Medal. The years included in the "papers per year"

calculations across different age groups begin with the year of the first publication and end at age

59 or the year of death (whichever comes first).

 Borjas and Doran 46

Table 2
Impact of the Fields Medal on the number of papers published per year

  Sample of contenders  
  Contenders with above-  
median citations   All contenders  
Post-medal   Post-medal
Post-medal period × Post-medal period ×
Specification:   period   Fields Medal   period   Fields Medal  
1. Number of papers   0.119   -1.378     -0.157   -0.918  
  (0.548)   (0.676)     (0.291)   (0.435)  
2. Log number of papers   -0.074   -0.244     -0.093   -0.173  
  (0.089)   (0.105)     (0.065)   (0.090)  
3. Papers, quantile regression   -0.421   -0.665     -0.314   -0.756  
  (0.283)   (0.256)     (0.150)   (0.161)  
Number of papers:            
4. Contenders’ post-medal   0.160   -1.395     -0.166   -0.914  
period begins at age 36   (0.494)   (0.656)     (0.279)   (0.435)  
5. Contenders’ post-medal   0.266   -1.440     0.069   -0.987  
period begins at age 40   (0.545)   (0.681)     (0.286)   (0.436)

Notes: Standard errors are reported in parentheses and are clustered at the individual level. The

regressions using the sample of contenders with above-median citations have 3,269 observations

(2,719 observations in the log papers regressions); the regressions using the sample of all

contenders have 5,213 observations (4,109 observations in the log papers regressions). In rows 1-

3, the contenders’ post-medal period begins the year they are no longer eligible to receive the

Fields Medal.
 Borjas and Doran 47

Table 3
Impact of the Fields Medal on other annual measures of productivity

  Sample of contenders  
  Contenders with above-  
median citations   All contenders  
Post-medal   Post-medal
Post-medal period × Post-medal period ×
Dependent variable:   period   Fields Medal   period   Fields Medal  
1. Published at least one paper   0.030   -0.118     -0.002   -0.092  
  (0.041)   (0.047)     (0.030)   (0.044)  
2. Number of citations   -1.146   -44.182     5.448   -44.493  
  (13.480)   (14.645)     (8.826)   (12.631)  
3. Citations per paper   2.204   -11.000     2.560   -10.716  
  (3.351)   (3.145)     (2.739)   (2.866)  
4.  Probability  of  a  “home  run”   0.062   -0.075     0.054   -0.075  
  (0.035)   (0.034)     (0.027)   (0.031)  
5.  Probability  of  a  “strikeout”   0.009 0.053   0.022 0.058
  (0.024) (0.028)   (0.021) (0.026)
6. Number of mentees   -0.016   -0.126     -0.017   -0.137  
  (0.078)   (0.074)     (0.054)   (0.058)  
7. Number of papers   1.430   -3.981     0.883   -4.932  
published by mentees   (3.194)   (2.522)     (2.364)   (1.828)
8. Number of citations -12.676 -56.309   -17.963 -60.682
generated by mentees (44.334) (37.395)   (29.630) (29.787)

Notes: Standard errors are reported in parentheses and are clustered at the individual level. The

regressions using the sample of contenders with above-median citations have 3,269 observations;

the regressions using the sample of all contenders have 5,213 observations. The sample sizes for

the regressions reported in rows 3-5 are 2,719 and 4,109 observations, respectively; and the

sample sizes for the regressions reported in rows 5-7 are 2,999 and 4,797, respectively. A “home

run” occurs when the number of citations per paper published in a given year is above the 99.5th

percentile for all mathematicians in the AMS database; a “strikeout” occurs when the number of

citations per paper published in a given year is zero.

 Borjas and Doran 48

Table 4
Impact of the Fields Medal Using ICM Plenary Speakers as Control Group

  Sample of contenders  
    Plenary speakers with
All plenary speakers   above-median citations  
Post-medal   Post-medal
Post-medal period × Post-medal period ×
Specification:   period   Fields Medal   period   Fields Medal  
A. Contenders invited to be  
plenary speakers at any age
1. Number of papers   0.096   -­‐1.076     -­‐0.091   -­‐0.955  
  (0.189)   (0.356)     (0.291)   (0.432)  
2. Log number of papers   -­‐0.014   -­‐0.239     -­‐0.059   -­‐0.183  
  (0.043)   (0.080)     (0.059)   (0.088)  
3. Citations   8.998   -­‐50.058     3.749   -­‐47.642  
  (7.305)   (9.698)     (11.742)   (14.952)  
4. Citations per paper   1.217   -­‐10.950     0.257   -­‐10.012  
  (2.011)   (2.709)     (3.035)   (4.013)  
5. Probability of a “home run” 0.030   -­‐0.078     0.034   -­‐0.066  
(0.019)   (0.029)     (0.029)   (0.032)  
6. Probability  of  a  “strikeout”   0.006   0.073     0.011   0.048  
  (0.152)   (0.024)   (0.018)   (0.023)  
   
B.  Contenders  invited  to  be        
plenary  speakers  by  age  44  
1. Number of papers   0.039 -0.906   -0.029 -1.130
  (0.338) (0.448)   (0.584) (0.685)
2. Log number of papers   -0.063 -0.161   -0.092 -0.188
  (0.067) (0.089)   (0.095) (0.107)
3. Citations   0.587 -39.978   -8.096 -37.961
  (10.427) (12.928)   (14.276) (15.344)
4. Citations per paper   0.386 -8.993   0.672 -9.019
  (2.756) (2.876)   (3.912) (3.468)
5. Probability of a “home run” 0.040 -0.065   0.060 -0.072
(0.026) (0.038)   (0.035) (0.035)
6. Probability  of  a  “strikeout”   0.011 0.030   0.023 0.023
(0.017) (0.026)   (0.024) (0.025)

Notes: Standard errors are reported in parentheses and are clustered at the individual level. The

sample sizes for the regressions reported in rows 1 and 3 of Panel A are 9,011 and 5,133

observations, respectively; and the sample sizes for the regressions reported in all other rows are
 Borjas and Doran 49

7,505 and 4,396, respectively. The sample sizes for the regressions reported in rows 1 and 3 of

Panel B are 4,745 and 3,171 observations, respectively; and the sample sizes for the regressions

reported in all other rows are 3,966 and 2,655, respectively. A “home run” occurs when the

number of citations per paper published in a given year is above the 99.5th percentile for all

mathematicians in the AMS database; a “strikeout” occurs when the number of citations per

paper published in a given year is zero.

 Borjas and Doran 50

Table 5
Impact of the Fields Medal on the probability of cognitive mobility

  Baseline field defined using Baseline field defined using

two modes     one mode  
Post-medal   Post-medal
Post-medal period × Post-medal period ×
Sample and specification:   period   Fields Medal   period   Fields Medal  
A, Using sample of contenders          
with above-median citations  
1. Baseline = eligibility period   -0.049   0.163     -0.065   0.213  
  (0.031)   (0.036)     (0.038)   (0.039)  
2. Baseline = first 3 years   -0.064   0.132     -0.049   0.135  
  (0.038)   (0.039)     (0.043)   (0.042)  
3. Baseline = first 5 years   -0.066   0.158     -0.025   0.147  
  (0.033)   (0.036)     (0.036)   (0.039)  
           
B. Using sample of all          
contenders  
1. Baseline = eligibility period   -0.016   0.146     -0.020   0.183  
  (0.025)   (0.033)     (0.030)   (0.040)  
2. Baseline = first 3 years   -0.042   0.121     -0.034   0.134  
  (0.030)   (0.036)     (0.033)   (0.040)  
3. Baseline = first 5 years   -0.029   0.135     -0.010   0.144  
  (0.028)   (0.034)     (0.030)   (0.038)

Notes: Standard errors are reported in parentheses and are clustered at the individual level. The

“baseline field” is defined by the set of the (one or two) modal fields and all related fields in

which the mathematician published during the baseline period (either the entire eligibility period,

the first three years, or the first five years of his career). The dependent variable is a cognitive

mobility indicator set to unity if the field of publication for each paper during the

mathematician’s career is not in the baseline field. The regressions in Panel A have 10,911

observations; the regressions in Panel B have 14,628 observations.

 Borjas and Doran 51

Table 6
Cognitive mobility and the duration of the preparation spell

  Baseline field defined Baseline field defined

using two modes     using one mode  
Fields   Fields
Sample and specification:   medalist   Contender   medalist   Contender  
A, Using sample of contenders          
with above-median citations  
1. Baseline = eligibility period   0.213   0.229     0.173   0.212  
  (0.033)   (0.038)     (0.043)   (0.035)  
2. Baseline = first 3 years   0.202   0.196     0.183   0.164  
  (0.035)   (0.039)     (0.036)   (0.039)  
3. Baseline = first 5 years   0.193   0.201     0.212   0.183  
  (0.030)   (0.034)     (0.036)   (0.033)  
           
B. Using sample of all contenders            
1. Baseline = eligibility period   0.211   0.263     0.171   0.214  
  (0.032)   (0.035)     (0.040)   (0.036)  
2. Baseline = first 3 years   0.200   0.208     0.181   0.192  
  (0.036)   (0.039)     (0.037)   (0.035)  
3. Baseline = first 5 years   0.193   0.224     0.209   0.194  
  (0.031)   (0.033)     (0.036)   (0.034)

Notes: Standard errors are reported in parentheses and are clustered at the individual level. The

length of the preparation is the length of time elapsed (in years) between any two consecutive

papers in a mathematician’s career. The “baseline field” is defined by the set of the (one or two)

modal fields and all related fields in which the mathematician published during the baseline

period (either the entire eligibility period, the first three years, or the first five years of his career).

The dependent variable is a cognitive mobility indicator set to unity if the field of publication for

each paper during the mathematician’s career is not in the baseline field. The regressions in

Panel A have 10,821 observations; the regressions in Panel B have 14,495 observations.
 Borjas and Doran 52

Notes: The group of “contenders” is composed of persons who were awarded at least one of six

other mathematics prizes (the Abel, Wolf, Cole Algebra, Bôcher, Veblen, and Salem Prizes) and

have above-median per-year citations during the eligibility period for the Fields Medal, but were

not awarded the Fields Medal. We  smooth  out  the  trend  by  using  a  3-­‐year  moving  average  

centered  on  the  middle  year  in  the  interval.

Figure 1
Average number of papers published annually by the Fields medalists and the contenders (three-
year moving average)
 Borjas and Doran 53

Notes: The group of “contenders” is composed of persons who were awarded at least one of six

other mathematics prizes (the Abel, Wolf, Cole Algebra, Bôcher, Veblen, and Salem Prizes) and

have above-median per-year citations during the eligibility period for the Fields Medal, but were

not awarded the Fields Medal. Cognitive mobility indicates if a paper published at any point

during the mathematician’s career differs from the “baseline fields” in the eligibility period for

the Fields Medal. The “baseline field” is defined by the set of the two modal fields and all related

fields in which the mathematician published during the eligibility period. We  smooth  out  the  

trend  by  using  a  3-­‐year  moving  average  centered  on  the  middle  year  in  the  interval.

Figure 2
The probability of cognitive mobility for the Fields medalists and the contenders (three-year
moving average).