Since 2020, Romania’s performance in the International Mathematical Olympiad (IMO) has been nothing short of amazing. In 2022, Romania came in fifth overall, fourth in 2023, and twelfth in 2024. In 2023, Romania placed fourth globally and first in Europe at the International Physics Olympiad, seventeenth globally and third in Europe at the International Olympiad in Informatics, sixth globally and second in Europe in the European Girls’ Mathematical Olympiad, first in the Balkan Mathematical Olympiad—which also included France, Italy, and the United Kingdom—and first in the Central European Olympiad in Informatics. Romania also performed well in the International Chemistry Olympiad and many others.

It’s an understatement to call Romania’s skill in Olympiads merely “overperformance”. Romania’s lackluster performance in international assessments and its relatively small population size of just over 19 million people makes the things they do in Olympiads downright miraculous.

Average Romanian educational performance is unimpressive. Romanian youth routinely perform below the average of OECD countries and near the bottom of the pack of European nations. Romania has a poor-to-mediocre showing whether you include or exclude migrants from the calculations, and its scores on assessments like the PISA aren’t low due to being tainted by bias in the examinations. Romania genuinely underperforms. But underperformance is not the impression you would get if you only knew of Romanian education from Olympiads.

One possibility is that Romanian students have more variable performance on international assessments than students in other countries. No dice: they aren’t much more variable than the student populations in other countries, and a handful of comparably-sized nations with worse Olympiad performance are more variable. Another possibility is that, for some reason, there’s a fat right tail in Romanian educational performance. If this is true, it just doesn’t show up in any existing data. Given the fact that international assessments indicate Romania’s sampling tends to be population-representative, we should have a strong prior against this possibility. Romanian test scores tend to be distributed along a symmetrical bell curve. 

Yet another possibility is that Romania has an undersampled ethnic group that overperforms, but whose schools aren’t tested very well. The only group this might be is Romanian Jews and using them as an explanation is problematic for two reasons. The first is that there are too few to realistically explain Romanian Olympiad performance. The second is that we know the identities of Olympiad participants from Romania, and they don’t seem to be Jewish.

Something else, something more mysterious, explains why Romania is such an outlier in international intellectual competitions. That thing is, in fact, the unique design of the Romanian educational system.

In the late 19th century, Romanian prince regnant Alexandru Ioan Cuza attempted to raise the status of the nation by instituting a mass literacy campaign centered around building free schools that children were compelled to attend. This effort was largely a failure, with literacy failing to break 50% by the 1930s. But World War II precipitated change. In 1948, Romania’s new governing communist party began to bring about serious educational reform at a breakneck pace.The Education Law of 1948 was passed to provoke a military-grade offensive against illiteracy, involving the mass participation of the literate from all walks of life in uplifting the poor, the abandoned, and those who simply shunned education. By the end of the 1950s, illiteracy was practically eradicated among Romania’s youth.

The education system that existed in Romania’s communist period was modeled on the system in place in the Soviet Union, and it included a fair helping of political propaganda in addition to physical labor. The system also overproduced schools, resulting in shoddy but widely available facilities dotting the country. Like the Soviet school system, Romania’s was marked by increasing lengths of compulsory education, poor availability of qualified teachers and educational supplies, high budgetary costs, and an extreme level of credential inflation.

After the fall of communism, the new democratic government went on to shutter many of these schools and to immediately lower compulsory schooling requirements to put an end to the bureaucratic nightmare that Soviet influence had saddled the country with. In the following years, how Romania wished to ration scarce governmental resources for education was a matter of intense debate, and out of that debate came a strong sentiment that, whatever the system, Romanian education would be structured competitively.

Nowadays, the most prestigious Romanian high schools are the National Colleges, or Colegiu Național. These schools are often international and frequently uphold old educational traditions sometimes dating back more than a century. Below these schools are the Liceu Teoretic, which are the norm, offering standard educations. Romania also has three military colleges—Colegiu Militar—managed directly by the Ministry of National Defense. There are also schools focused on service, technical schools, vocational schools, and apprenticeship programs. The brightest students get their pick among these schools after they take the national placement test, the Evaluarea Națională, when they are graduating the 8th grade around ages fourteen to fifteen.

The high school placement test is a standardized test covering Romanian language and literature as well as mathematics. Performance on the examination is reported publicly when students are issued a score on a one-to-ten scale with precision to two decimal places. A student who receives a high grade—say 9.65—would have their pick from most any school, whereas a student scoring 5.00 or below would usually be constrained to a less academically-focused form of education like a vocational program. Most students elect to go to the best school they are able to test into, and so the degree of sorting across schools is very high. To make this setup even more extreme, there is also often—but not universally—sorting within schools, as students select into educational tracks. This is done directly when applying to schools.

At the end of the Romanian high school experience, there is a graduation test, the Bacalaureat, or bac. This test is marked like the entrance examination and, to pass, students must obtain a score of at least five in the subjects they have elected to take. This testing includes written and oral examinations, assessments of foreign language and computer skills, and, for ethnic minorities, assessment of their skill with their maternal language other than Romanian. The need for a given score on this examination can range from requiring just passing to requiring a high score, depending on the university one intends to attend, if that is their goal.

The design of Romania’s educational system makes it perhaps the most stratified educational system in the world. The fact that they have a centralized repository containing all student and teacher educational data makes their system perfect for a high-powered evaluation of exactly what happens when a country opts to hyper-stratify education.

One of the cruel parts of the Romanian system is that, though sorting is nationally available, students do not have equal opportunities to sort. Students located in smaller towns have fewer high school options to select from unless they’re among the few who opt into a military academy, which means joining the military. The extent of sorting is far more intense in areas with larger numbers of schools. In a recent paper, the Romanian economist Andrei Munteanu provided an illustration of how this works: essentially, the fewer schools in a locale, the more each individual school contains students with a wider range of ability and, the more schools in a locale, the more each individual school will be stratified into low, middle, or high ability. 

This combined sorting between schools and tracks means that low-ability students get stuck with other low-ability students, and high-ability students are surrounded by other high-ability students. In effect, peer groups throughout high school are extremely homogeneous. This matters because then low-performing students drag down low-performing students, and high performers cause each other to rise. Romania’s educational system has causal peer impacts on student performance on the graduation test that are very large in both directions, but primarily where there are opportunities for sorting to take place.

But peer effects are not everything to Romania’s exceptional Olympiad performance; they are just the fertile ground in which exceptional performance is fostered. The next part has to do with teachers. Like students, Romania’s teachers must take tests to be able to do what they want to do. Teachers naturally prefer to lecture smarter students, and the smartest teachers have their pick of the schools, and even of the tracks. In a paper with extremely robust results, researchers from the last decade described this as such:

[Teachers] with higher certification standards are more likely to work at better-ranked schools. This sorting persists even within schools as one moves from a weaker to a stronger track, and even within tracks as one moves from a weaker to a stronger class.

The best teachers also opt into towns with more schools. It’s apparent, then, that teachers prefer teaching in the highest-achieving places they can be, both within and between towns. The effect of teacher-student ability pairing is accentuated even more by incentives to compete. The government of Romania is not unique in providing monetary rewards for those who win Olympiads, those who teach winners of Olympiads, or those schools Olympiad winners attend, but they are unique in having all the previously-mentioned institutional characteristics on top of providing comprehensive monetary incentives for Olympiad achievement. 

Romania’s immense success in Olympiads and the widely recognized importance of Olympiad wins for signaling student human capital has also spawned a small number of private schools that advertise their prominence and tutoring capabilities. Many teachers also recommend to parents that they obtain additional tutoring for their brighter pupils, and tutoring services are commonplace. The commonality of tutoring for Olympiad winners is a global constant, whereas the things distinguishing Romania are not.

Two notable factors do not increase performance in the same direction. These are very slight decrements in funding allocated to the highest-ability schools, and when parents reduce the time they spend helping their students with homework, conditional on their kids matching into better schools. Another potential factor that militates against the synchrony of resource allocation in Romania is that children in more selective schools report feeling marginalized because they realize that they’re not as strong of students as they believed. The decrements in funding are likely to be unproblematic, because higher-scoring schools tend to be larger and more urban, lending them economies of scale. Due to this, they may have effectively more funding.

With all the pieces on the board, the key to Romania’s Olympiad success is three-fold: put the best students in the same classrooms, put the best teachers with the best students, and then incentivize schools, teachers, and students each to win Olympiads.

This system has proved amazingly fruitful. Given its underlying human capital, the poverty from its communist legacy, and its modest population size, Romania should not perform the way it does in academic Olympiads. And yet it does. The trade-off for Romania, however, is palpable.

Large portions of Romania’s Olympiad winners leave the country. Because Romania is a member state of the European Union, the people the country has put great effort into training and credentialing are easily able to leave the country and acquire jobs elsewhere.

Losing the right tail to brain drain is damaging for many countries, but it’s arguably worse for Romania because its educational system is so zero-sum: the top performers do better, while the low-performers do worse. This sorting does not “lift all boats,” as it were. In Romania, the system makes for an incredibly well-trained right tail and a neglected left tail, and that left tail might hurt more than the right tail is helped, if effects on test scores are any indication. On its own, Romania’s system might be a stellar boon to the country. But with free movement of talent between countries, Romania ends up subsidizing talent discovery for other countries with less apt educational systems. 

Most of the growth we see around us is due to the innovations of the right tail, and if they do better, we all do better. Though I doubt Romania’s schooling raises the intelligence of the right tail, even raising aptitude is worth something, because we must get capable people to the frontiers of their respective fields in order to innovate, and Romania has fostered a system that seems to do just that. Moreover, even if Olympiad training does not make those on the right tail more capable but instead simply prepares them better, then it can still have large, socially beneficial effects simply through providing Romania’s highly capable people with a means of having their talents recognized internationally. 

But these benefits are returned only very indirectly to Romania, if at all on net. Rather than changing Romania’s educational system or closing the borders, the right solution is for more nations to choose to be like Romania, getting a lot more juice out of their smart kids by designing a system just for them.