Drawing Blanks

Here is another attempt to test the social conformity theory with 2011 Russian Elections, using Stephen Coleman’s methodology. I’m going to present basically the same data that I presented here but with more statistical rigor. I also show that after removing all ballots cast for United Russia from the valid vote count, the remaining data still fits the entropy curve very nicely. It would also be interesting to see what happens if  the count of registered voters is accordingly adjusted (as if those voters were not present in the population). I’ll look into that when I have more time.

If you’d like to understand what this is about (entropy of choice and expected correlation between the party choice and the turnout choice) I encourage you to read these Coleman’s works, they are short and very accessible: Russian Election Reform and the Effect of Social Conformity on Voting and the Party System: 2007 and 2008 http://mpra.ub.uni-muenchen.de/14304/ (final published version here:  Coleman, Stephen. 2010. “Russian Election Reform and the Effect of Social Conformity on Voting and the Party System: 2007 and 2008.” Journal of the New Economic Association (Moscow), 5: 72-90.  In Russian as “Реформа российской избирательной системы и влияние социальной конформности на голосование и партийную систему: 2007 и 2008.”) and A Test For Conformity In Voting Behavior http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.31.4615

Some more background. 7 parties participated in the 2011 Russian parliamentary elections. The United Russia party (the ruling party) won by a landslide. There is evidence (eyewitnesses, photo, video) of massive fraud that took place during voting. There is evidence (http://ruelect.com/en/ which I also mentioned here) of massive fraud that took place during vote counting. There are also tons of blog and press publications of attempts (mostly amateurish, imho) to detect and quantify the fraud purely by statistical analysis. Most of them are based on the assumption that in honest elections party support and voter turnout are not correlated. (Here is a published example of this approach: http://vote.caltech.edu/drupal/files/working_paper/vtp_wp62.pdf) I personally find that assumption a huge oversimplification, and Stephen Coleman has shown http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.31.4615 that it doesn’t hold in many different settings.

Even more background. Here I demonstrated that the Russian elections data is highly inhomogeneous, and here I dissected it into 3 categories which have very distinct statistics: C1 – ethnic outskirts, C2 – smaller precincts, C3 – larger precincts.

So here are Party Entropies vs. Turnout, fitted by Turnout Entropy, with distributions of residuals. NOT weighted by precinct size.

C3 (larger precincts).

“Coleman Factor” CF=2.146. Mean(H(P)) = 2.01.

Residuals:
Min         1Q         Median       3Q         Max
-1.75576 -0.09590  0.04891  0.16276  2.42235

C2 (smaller precincts).

CF = 1.98. Mean(H(P)) = 1.66

Residuals:
Min       1Q            Median       3Q        Max
-1.97287 -0.14239  0.04657  0.26603  2.80735

C1 (ethnic outskirts).

CF = 2.14. Mean(H(P)) = 2.01

Residuals:
Min         1Q          Median       3Q        Max
-1.75576 -0.09590  0.04891  0.16276  2.42235

General population.

CF= 2.07. Mean(H(P)) = 1.7

Residuals:
Min          1Q         Median       3Q        Max
-2.07171 -0.13714  0.06212  0.22727  2.80735

General population with all United Russia ballots removed from count.

CF=2.067. Mean(H(P)) = 1.7

Residuals:
Min        1Q       Median      3Q     Max
-2.0668 -0.1418  0.0337  0.2076  2.4531