Who Wins Nobel Prizes?

The most prestigious scientific achievement is likely the Nobel Prize, which is awarded every year to “those who confer the greatest benefit to mankind” in the fields of physics, medicine, and chemistry (of course, there are also Nobel Prizes for literature and peace, as well as a Nobel Prize for economics which was added much later). Because the prize has been awarded for so long (since 1901), and because the bar for winning them is so high, the history of prize winners gives us an interesting glimpse into the history of scientific progress. Looking at the backgrounds of Nobel Prize winners can tell us about where some of the best scientific work is being done, and the sort of people and institutions that are doing it.

Putting together the data

To review the history of scientific Nobel Prize winners, I started with two datasets. One is a list of prize winners, along with various biographical facts about them, provided by the Nobel Foundation and available at Opendatasoft. The other is a dataset of publications by Nobel Prize winners in physics, chemistry, and medicine assembled by Li, Yin, Fortunato, and Wang, available at Harvard Dataverse.

This second dataset was particularly valuable: it lists nearly every publication authored or co-authored by a prize winner (tens of thousands of publications in total), and it flags which publications were for Nobel-winning work. For each publication, the dataset lists where it was published, and (crucially) the institutional affiliation of the author at the time of publication. This lets us see not just who the Nobel Prize winners are, but where prize-winning work gets done.¹

This second dataset needed some repairing. Around a third of the institutional affiliations were missing, so I used Claude 3.7 Sonnet to fill in the blanks. When checking Claude’s work, I was surprised at how accurate it was. Of the roughly 275 entries Claude filled in, there were errors in fewer than 10 of them, and the errors that it did make were often borderline (i.e: a publication coming out in a given year and an author moving institutions that same year, or an author having multiple affiliations).

In fact, Claude’s answers proved to be more accurate than the original dataset, at least for institutional affiliation. Spot-checking the original dataset revealed numerous errors, so I also ended up using Claude to make corrections to the original dataset. Altogether Claude (with me checking) fixed probably ~100 errors in categories like institutional affiliation and publication date.

I supplemented these two datasets by using Claude to add information on where prize-winners attended graduate school (I checked Claude’s answers here as well.)

The final dataset contains information on 877 Nobel Prize-winning publications by 545 prize winners for prizes in physics, chemistry, and medicine between roughly 1915 and 2016.² It includes the journal published, the year of publication, the year and type of prize, the author, and the author’s background (country of birth, year of birth, educational background, etc.) If you’re interested in this dataset, I’ve made it available on github here.

This isn’t a perfect dataset. For one, pinning down institutional affiliation can be difficult. Prize winners will often have multiple affiliations, or take seasonal research appointments, or do some work in a major lab (like CERN) while being employed elsewhere. And “affiliation” doesn’t necessarily mean an institution had a meaningful contribution to the work: Einstein’s affiliation at the time of his prize-winning work was the Swiss Patent Office, but it wasn’t supporting his research in any real way.

There are similar (though less severe) categorization problems with some other fields. A prize winner may have multiple graduate degrees, or transfer schools part way through their degree, and so on.

It’s also hard to decide which papers should be credited as prize-winning work. A prize winner may have put in years of work gradually chipping away at a problem before publishing what are considered their prize-winning publications. (Richard Feynman famously won his Nobel Prize for work that began with him informally trying to figure out the mathematics of plates that wobbled when tossed in the air.)

None of this precludes analysis of the data, but it does mean that we should regard the numbers as somewhat approximate.

Where Nobel Prize-winning work gets done

To start, let’s look at the countries where prize-winning research is done. This table shows number of Nobel Prizes by country of the institution where the research was done:³⁴

The US is far and away the leader in number of prizes, followed by the UK, which punches far above its weight. (“Prizes per capita” is tough to calculate when populations change over time, but the UK would certainly be above the US on this metric.) Germany is next, followed by France, Switzerland, and Japan. Other European countries, along with Canada, round out the top 10.

Other than Japan and Israel, there are almost no countries outside Europe and the Anglosphere. Argentina has two prizes (Bernardo Houssay in 1947 and Luis Federico Leloir in 1970), China has one (Tu Youyou in 2015), and India has one (C.V. Raman in 1930).⁵

We can also look at how countries of prize-winning work have changed over time. The chart below shows prizes per country in 10-year buckets (the number of prizes won between 1915 and 1924, 1925 and 1934, etc.)

Early in the 20th century, very little Nobel Prize-winning work was done in the US (I noted this previously in my essay comparing China and the 19th century US, where I pointed out that the US was fairly weak in science even as it was becoming a major world power). Between 1915 and 1925, the leading Nobel-winning countries were Germany and the UK, with 6 and 5 prizes respectively. By the early 1930s the US had reached parity with Germany and the UK, but it wasn’t until after WWII that the US began to dominate the field (this is also when German prizes began to decline).

The US’s fraction of prizes peaked in the late ‘80s and early ‘90s. Between 1995 and 2015, the American share of prizes declined, mostly due to increasing prizes for French and Japanese work.

One other notable pattern here is that credit for prizes is getting more widely distributed over time. Prizes are much more likely to be shared by multiple recipients today than they were in the early 20th century.

Now let’s take a look at the institutions where Nobel Prize-winning work gets done. Overall prize winners were affiliated with over 200 institutions. But distribution is highly asymmetric. Over 130 institutions have only one prize apiece, and another 42 have only two prizes apiece. The top 39 institutions are responsible for more than 50% of the prize-winning work, and the top 10 institutions have close to 30%. Here are the top 40 institutions by number of prizes, which includes every institution that has three prizes or more.⁶

The UK’s science chops also show in the list of top institutions, with Cambridge’s 32 prizes leading the pack. Number two is Harvard (no surprise) with 22 prizes, followed by Columbia with 13.

While the top three institutions weren’t surprising to me, the institutions farther down the list were more surprising. I wouldn’t have guessed CalTech as coming in at number four, or Cal Berkeley as number six. Number five, Rockefeller University, I hadn’t even heard of until looking at this data. (Per Wikipedia, Rockefeller University is “a private biomedical research and graduate-only university in New York City.”) I’m also mildly surprised that Oxford doesn’t rank higher (it’s tied for number 22 with a bunch of other universities that have four prizes).

If we look at trends in prize-winning institutions over time, we see a few interesting patterns. Rockefeller University and University of Chicago peak in the 1950s to 1970s. MIT doesn’t start winning prizes until the late ‘70s and early ‘80s (surprising to me given the importance of the Rad Lab during WWII). Harvard starts winning prizes mostly after WWII, and Cambridge has always been strong.

What about the type of institution where prize-winning work gets done? Here are prizes broken down by type of institution: universities, government labs (like Brookhaven National Lab), private labs (like the Pasteur Institute), and corporations (like Bell Labs).⁷

We see that roughly three fourths of Nobel Prize-winning work is done at universities, with government labs, private labs, and corporations sharing the remainder. This isn’t particularly surprising, given the list of top prize-winning institutions above, which is dominated by universities.

We can also see how these proportions have changed over time. Interestingly, the fraction of prizes going to universities has actually slowly declined, with the balance mostly taken up by government labs. (It would be interesting to see if this trend continues through the 2015-2025 period).

For prize-winning work done at corporations, the winners are mostly electronics and communication companies (Bell Labs, General Electric, Standard Telephone Labs) and pharmaceutical and chemical companies (Cetus, Dupont, Merck).

Who wins Nobel Prizes

Now let’s look at the prize-winners themselves. The table below shows the country of birth for all each prize winner (for countries that don’t exist any more, I’ve tried to use modern equivalents. For countries whose borders have changed, I’ve generally used border extents at date of birth).

Overall 43 countries have birthed a prize winner in physics, medicine, or chemistry, but once again the distribution is highly asymmetric. Slightly more than a third of prize winners have been born in the US, and the US, UK, Germany combined have produced just under 60% of all science Nobel Prize winners.

If we look at prize winners by gender, we see that prize winners are overwhelmingly male. Between 1915 and 2016 there have been only 16 female science prize winners, though this has trended upwards more recently (between 2016 and today there have been eight more female science prize winners).

If we look at the educational background of prize winners, we see a similar pattern to where the prize-winning work gets done. Prize winners attended 159 different graduate schools, but 81 of those trained just a single prize winner, and 26 trained just two. The top 10 schools trained over 1/3rd of all prize winners, and these schools are mostly the same ones where the prize-winning work gets done: Harvard, Cambridge, Columbia, etc.

Most of these institutions are in a small number of countries. 70% of prize winners went to graduate school in either the US, the UK, or Germany.

So overall there’s a high degree of concentration in the institutions responsible for Nobel Prize winning research. A small number of elite institutions produce a disproportionately large share of Nobel Prize winners, both in terms of where they go to graduate school and where they do their prize-winning research. However, there’s also a long tail at work; there’s a long list of countries that have birthed at least one Nobel Prize winner, and a long list of institutions that have trained them, or been the location of prize-winning work. While Nobel Prizes are most likely to be won by folks born in the US, UK, or Germany, working and studying at elite institutions, they can come from almost anywhere.

When is prize-winning work done, and when is it rewarded

Now let’s take a look at how long it takes prize-winning work to get rewarded. The graph below shows the age of Nobel prize winners over time.

The average age of prize-winners has steadily trended upward over time, from around 45 in the 1920s to around 65 in the 2010s. And here's the time between when winners were awarded a Nobel and when they did their prize-winning work.⁸

This has also trended up over time. In the 1920s it took just over 10 years on average between prize-winning work and actually winning the prize. Now the average is closer to 25 years, though there remains a great deal of variation.

If the average age of prize winners when they did their prize-winning work has remained roughly flat over time.

For prizes in the 1920s, the average age of prize winners when they did their work was just over 38. From 2007 to 2016, it was just over 41.

So while the age of prize winners when they receive their award has steadily increased, and the time between when the work is done and when it’s rewarded is increasing, the age when they do their prize-winning work has remained roughly constant. (This is true if we look at year of prize, but not true if we look at age against year of actual work. See update below the conclusion.) Prizes are thus increasingly going to older work done in the past. We see something consistent with this pattern if we look at decades when prize-winners did their work, with work done in the 1950s through early 1980s making up a huge fraction of prizes.⁹

There are a few possible explanations of this. One is that scientific progress has slowed down, modern discoveries are less ground-breaking, forcing the prize committee to increasingly reward older work (which maybe wasn’t originally considered prize worthy). Another possible explanation is that there is lots of prize-worthy modern work, but the folks doing it have to wait in line to get their prize behind older folks whose “turn” it is to win.

Without wading too deeply into the debate around the nature of recent scientific progress, my guess is that there’s some of both phenomena going on. The “waiting your turn” for Nobel Prizes, and prizes being viewed as at least partially a lifetime achievement award, are both, as I understand it, real phenomena. But it also seems likely that there’s fewer groundbreaking discoveries today, at least in the fields the prize rewards. In the early days of the Nobel prize it wasn’t even clear what an electron was, quantum mechanics hadn’t yet been discovered, and heavy elements like plutonium hadn’t been synthesized. Much of the fundamental nature of reality still hadn’t been discovered, which seems much less true today.

Speculation as to what an electron might be from a 1920 book about steel making.

Conclusion

To sum up, there’s a large degree of concentration in Nobel Prizes. Most prize-winning work has been done in a small number of countries — the US, UK, and (historically) Germany — and most prize-winners get their degrees and do their prize-winning work at a small number of elite institutions in those same countries. But there’s also a long tail at work: while most prize winners are born in the US, UK, or Germany, 43 countries have produced a scientific Nobel Prize winner. And while most prize-winning work is done at places like Cambridge, Harvard, and Columbia, over 200 institutions have produced prize-winning work.

Prize winner’s age at the time they receive their prizes is steadily increasing over time, though the age when they do their prize-winning work isn’t, suggesting Nobel’s are increasingly being awarded to past work. This could be due to a slowdown of groundbreaking discoveries in the fields the Nobel rewards, or it could be due to the sociology of how prizes get awarded, with my guess being a combination of both.

3/31/25 update - age at time of prize-winning work

I previously noted that the age of prize winners when they did their prize-winning work has remained roughly flat, and that this differed from a famous paper by Ben Jones which found that the age when prize-winning work is done has increased over time.

It is true that age is roughly flat, if you look at the age when prize-winning work was done against the year the prize was awarded. However, if we look at age against year of publication, we find that, similar to Ben Jones, age when prize-winning work is done is steadily increasing over time:

For prize-winning work done between 1900 and 1910, the average age of the winner was just under 37. For prize-winning work done between 1995 and 2005, the average age was just over 45. (Thanks to Ben Jones for pointing this out to me.)

1. NobelPrize.org lists institutional affiliation at time of award, but not when the work was actually done.

2. The year range is “roughly” 1915–2016 because the dataset contains some earlier entries, but data before 1915 isn’t complete.

3. For countries that don’t exist any more, or whose borders have changed, I used my judgment to either categorize them as their modern equivalents, or by the territorial extents that existed at the time. Austria-Hungary gets categorized as Austria or Hungary, USSR gets categorized as Russia, Prussia gets categorized as Germany, etc.

4. The table counts unique countries per prize. If there are multiple winners for a prize who all did their work in the US, that gets counted as one prize for the US. If a prize has three winners, two from the US and one from the UK, that counts as one for the US and one for the UK. If the same winner does work across multiple countries (say the US and Israel), that gets counted as one for the US and one for Israel.

5. For each of these countries, additional winners were born there but did their prize-winning work elsewhere.

6. Credit for a prize is assigned to institutions in a similar fashion to countries. If there are multiple winners for a single prize from the same institution, that gets counted once. If there are multiple winners at say, Harvard and Columbia, they each get counted once, etc.

7. I assigned hospitals to a category based on the sort of organization that funded or operated them. So hospitals attached to medical schools were classified as universities, government hospitals were classified as government labs, private hospitals were classified as private labs, etc.

8. This uses a simple average of the publication year for each prize-winning publication listed in the dataset. So if a prize winner has 10 publications equally spread between 1925 and 1935, that would give an average year of 1930.

9. Unlike previous graphs of prizes by country (which showed unique countries per prize), because this is measuring year of work, and different winners from the same country may have done their work at different times, this will “double count” countries in some cases. For instance, two of the winners for the 2016 Nobel Prize in physics, Haldane and Kosterlitz, each did their work in the US but at different times and places, so on this graph that would be counted as two prizes for the US.