Rerunning the 1979 MekhMat Admissions
In the summer of 1979, two Moscow mathematics teachers, Boris Kanevsky and Valery Senderov, tracked every graduate of six elite math high schools (Nos. 2, 7, 19, 57, 179 and 444) who applied to MekhMat, the mathematics faculty of Moscow State University. They split the applicants into two groups: 47 with no Jewish parent or grandparent, and 40 with at least one. Their samizdat table, reprinted by Alexander Shen in The Mathematical Intelligencer 16(4), 1994, is precise enough to run a counterfactual on:
| 1979, six schools | Group 1 (n=47) | Group 2 (n=40) |
|---|---|---|
| Olympiad winners | 14 | 26 |
| Multiple winners | 4 | 11 |
| Total olympiad prizes | 26 | 48 |
| Admitted | 40 (85%) | 6 (15%) |
Group 2 won 1.2 olympiad prizes per student against Group 1’s 0.55, and was admitted at 15% against 85%, an odds ratio of about 32. If the 46 places had been a random draw from the 87 applicants, the chance of Group 2 getting six or fewer is about 2×10⁻¹¹. Counterfactual one, equal rates: at the pooled rate MekhMat would have taken about 21 Group 2 students instead of 6, so about 15 are missing. Shen’s figures for MIFI (26 of 54 vs 3 of 29) and MFTI (39 of 53 vs 4 of 32) add about 7 and 12 more; some students applied to several institutes, so 34 is an upper bound, but it is still roughly a whole cohort from six schools in one year. Counterfactual two, merit by the regime’s own measure: the Soviet state ran and celebrated the olympiads. There were 40 winners in this pool and 46 places. Admit every winner first and Group 2 gets at least 26 places, so the gap is 20, not 15. Both numbers are lower bounds, because, as Shen notes, “in later years, the potential victims, aware of the barriers, didn’t try to apply.”
The mechanism lived in the oral exam. Shen records an answer, “x = 1 or x = 2,” marked wrong because the examiner wanted “x = 1; 2”; an oral that ran 5.5 hours; appeals that had to be filed within one hour. One problem set by named examiners had been the hardest problem in the second round of the 1985 All-Union Olympiad, solved by 6 of 100 contestants. When Senderov brought such a list to the 1975 Soviet IMO team, Tanya Khovanova recalls in a paper with Alexey Radul, the team “solved only half of them” in a month. Applicants got these one after another until they failed one; even at a 50% solve rate, three in a row passes one time in eight.
Where did the missing fifteen go? The tempting answer is “somewhere good.” Mark Saul’s 1999 Notices piece tells of “Edik” from Kolomna, failed at MekhMat in 1984 after more than four hours, partly for defining a circle without the word “all.” His examiner met him in the elevator and suggested the Gubkin Institute of Petrochemical and Natural Gas Industry — Kerosinka — because “they take people like you there.” Edik studied there and climbed the fence into the university to sit in on Fuchs and Feigin; the details match Edward Frenkel’s own published account. From 1978 there was also the “Jewish People’s University,” run from Bella Subbotovskaya’s apartment. George Szpiro’s account counts 21 unpaid teachers, including Fuchs, Feigin, Zelevinsky and Shen himself, about 350 students and about 100 “graduates.” It ended after the KGB questioned Subbotovskaya and she was killed by a truck in September 1982. Stories like these make the counterfactual look small, as if the talent were rerouted rather than lost. Shen, who taught in that apartment, disagrees: its non-MekhMat participants “were very gifted, but very few of them became professional mathematicians.” Saul’s Kerosinka alumni list points the same way, with more than half in technical fields. A Frenkel needed a fence to climb and two mathematicians giving up their evenings; that does not scale to fifteen a year, and Anatoly Vershik’s companion essay puts the system-wide loss at “hundreds, and possibly thousands.”
So the rerun comes out like this: with a fair exam, MekhMat’s 1979 intake from these six schools would have had roughly 15–20 more students, drawn from its strongest olympiad performers, and the shadow system recovered a famous few at about the scale of one Saturday seminar. The claim is testable. The original samizdat named its applicants, so the 87 could in principle be traced to their careers. My prediction is that the 34 rejected Group 2 students produced fewer research mathematicians than the 40 admitted Group 1 students, despite winning nearly twice as many olympiad prizes. If that holds, the measurable cost of the filter sits in the median turned-away olympiad winner who became a petrochemical engineer — and the only clean record of it exists because two schoolteachers kept a table.