Twitter | Search | |
Arseny Khakhalin
There was a vote in Russia last week, on making Putin president for life. Sadly, it passed. BUT the govt made a mistake & put all data online😄 Naturally, ppl scraped & analyzed it, as "lib" in stands for Liberty!! csv: +Analysis thread(1/7)
Reply Retweet Like More
Arseny Khakhalin Jul 6
Replying to @ampanmdagaba
Above is the main plot (after Klimek 2012). As both turnout and support were faked, we see a corner-blob emerge from the natural blob in the middle. In RU case, it's also striated, as cheaters go for round percent values. Some by-region histograms below (see 85%, 90% & 95%) (2/7)
Reply Retweet Like
Arseny Khakhalin Jul 6
Replying to @ampanmdagaba
Some other interesting regions include Tatarstan (a semi-independent republic in Central Russia speaking a Turkic language: what's the deal with this super-blot in the middle?), and Chechnya (well, you know this one! :) (3/7)
Reply Retweet Like
Arseny Khakhalin Jul 6
Replying to @ampanmdagaba
Also note the difference between cities that had active anti-Putin opposition plant themselves at voting stations & document everything (Moscow), and cities where it didn't quite happen (St. Petersburg). Putin would have still won, but there's a huge difference! (4/7)
Reply Retweet Like
Arseny Khakhalin Jul 6
Replying to @ampanmdagaba
I also tried to find biggest cheaters by measuring the combness of histograms (compared the freq of %% slightly-above and slightly-below a whole % number), and the prevalence of "lucky" ballot counts that yield a round %. (H0: % were noisified with σ=0.005, n=50). It kinda works?
Reply Retweet Like
Arseny Khakhalin Jul 6
Replying to @ampanmdagaba
Not surprisingly, the higher Putin-support, the closer the %% are to a nice round %. (Or rather: the closer it is to the roundest % one can get by dividing two integers, for a given voter list. Say, for 1111 voters, you can't get exactly 95%, but you can 1055/1111=0.9496) (6/7)
Reply Retweet Like
Arseny Khakhalin Jul 6
Replying to @ampanmdagaba
Finally, another fun visual is plotting "Pro-Putin turnount" (YES votes / voter list length) against polling place size. Tiny polls are all pro-Putin (either coz they are fake, or coz pollers actually visited old people in their homes). But large poll places go bimodal! (7/7)
Reply Retweet Like
Arseny Khakhalin Jul 6
Replying to @ampanmdagaba
Bonus links: 1) For Moscow ppl actually created a BROWSER (!!!), to see now their polling station compares to the rest! Wow data activism! 2) The turnout historams are already turned into nerdy oppositionary merchendise :)
Reply Retweet Like
Arseny Khakhalin Jul 6
Replying to @ampanmdagaba
That's all. The citation for the main plot: All analyses notebooks:
Reply Retweet Like
Arseny Khakhalin Jul 8
Replying to @hippopedoid
Updated data attribution: Original data comes from Sergey Shpilkin See for example his early post: The latest (better, cleaner, updated) version of the data can be found in this repo by :
Reply Retweet Like
Gennady Rudkevich Jul 8
Replying to @ampanmdagaba
I just added up all the Yes and No votes and ended up with 63,871,402 total votes. The official number of votes is over 73 million. Any idea why there's a disparity? I checked a few polling stations in Adygea & Moscow and they seemed accurate. Suggests some stations are missing.
Reply Retweet Like
Arseny Khakhalin Jul 8
Replying to @grudkev @hippopedoid
Hmm. No idea, honestly! The data is originally by Sergey Shpilkin (not on Twitter), found via github of . In the CVS I posted, I removed some NANs, but it should not have affected the total... idk! Maybe Shpilkin wrote about it on telegram or facebook at some point
Reply Retweet Like