In one of the recent posts on corruption, commentator AP wrote:
Kids from Moscow are having trouble getting into universities now because entrance, based on exam results, skews the chances of acceptance in favor of those students from corrupt regions where they can buy better results. Moscow is less corrupt than, say, Dagestan so Dagestani students perform much better on entrance exams.
Is this true? Seeing as how the Russian state doesn’t release Unified State Exam (USE) results by region, probably due to PC considerations, at first this assertion might appear to be unanswerable. However, there is a way to get round the problem.
(1) We know the PISA-derived IQ’s of some 43 Russian regions (which account for about 75% of its school-age population).
(2) The Russian government DOES release the the numbers of maximum scores in the USE tests by region. In this post we will consider the data for 2012. Furthermore, we know that at least at the federal level, these results tend to form bell curves.
(3) One of the primary “proofs” of electoral fraud in the Russian elections was the presence of spikes at convenient increments of 5%. In the case of USE fraud, we only have access to data for 100% scores and measuring the fatness of that tail should give us a clue as to its relative magnitude. (While it is possible and even likely that school administrators and regions would take care not to create too many maximum marks on the notoriously hard USE tests, far from everybody will follow said precautions. After all, if many regions didn’t even bother to smoothen the spikes to conceal fraud in the elections, is it realistic to posit that they’d take greater care around trifles like exams?).
(4) We know the number of 16 year old’s per Russian region from the 2010 Census, who would have participated in the 2012 exam season.
(5) We know the normal distribution.
The blue bars below show the number of top-scoring exams per region as a multiple of Russian 18 year olds there with an expected IQ of 130 or more, based on the region’s average PISA scores and a standard deviation of 15. The red bars show the same thing, with the major exception that an average IQ of 96 – that is, the national average – is assumed for ALL Russian regions.
As we can see above, the most suspicious results are mostly from ethnic Russian oblasts such as Stavropol, Kaluga, Rostov, Perm, and Vladimir, with the two big exceptions being Mari El and Chuvashia. To the contrary, Dagestan – the biggest Caucasian Muslim republic – has very few top scores relative to the number of very bright people we can expect to find there relative to most other Russian regions.
Finally, the reason that the red bar is a lot higher than the blue bar in Moscow, and to a lesser extent Saint-Petersburg, probably doesn’t have anything to do with foul play, but with the fact that their average IQ’s are about 106.6 and 102.6, respectively (i.e. considerably higher than the national average of 96). So while they generate a relatively disproportionate number of top USE scores, that is presumably because they attract the bulk of Russia’s most intellectual families (the so-called “cognitive clustering” effect).
Of course one problem is that we don’t have PISA data for all Russian regions. Maybe the Chechens do all the cheating then?
Probably not. Chechnya only had a total of five top scored USE results (for comparison, Moscow had 654 top results). In the graph below I produced results for all Russian regions, but with an unavoidable concession: In the case of those regions with no results from PISA, I had to make do with assuming a regional IQ of 96 (as per the Russian national average).
In so doing, yet another major region of likely fraud crops up: Bryansk. This oblast, along with Vladimir, produces as many top USE results as a percentage of its 18 year old population as does the intellectual capital, Moscow. Kalmykia, Kirov, and Lipetsk also join the list of Russian regions with suspiciously good USE results (probably not entirely coincidentally, Lipetsk and Kalmykia – along with Ingushetia – were the three regions whose USE results raised suspicions to the extent that they were rechecked).
He also makes the comment:
The schools with the top math students in the country stopped winning Olympiads, while private schools with politically connected kids started to win them…
No obvious way to statistically analyze this, but what we can say with some confidence is that there is no major ethnic angle to this:
As we can see above, the Central and North-West regions of Russia, which contain the cognitive hotbeds of the two capitals, massively surpass the number of people from the North Caucasus in the share of “Olympians” (basically students who did really well and get benefits) in the annual university cohort.
This is pretty much what we can expect on the basis of the average IQ differentials between these regions.