On December 28th, the WSJ published an article on “Russia’s Dubious Election” by Gregory White and Rob Barry (it’s behind a paywall, but you can read it here). In it they described the most famous argument for the 15% Club (i.e., the purported scale of fraud in the 2011 Duma elections) – namely, that of Sergey Shpilkin. A brief description of his approach: Observe that a higher turnout means more votes for United Russia; make a blanket assumption that all these extra votes are suspect, remove them as “irregularities”, and voila! United Russia’s plummets from 49% to about 34%! (Neither he nor the WSJ, to their credit, claim that it proves fraud; they use the more qualified phrase “cast doubt”). In the process, not only the elections are discredited but pretty much the entirety of Russian opinion polling and exit polling (a reminder: all the pre-elections polls gave United Russia 50% or more, and the most comprehensive exit poll, FOM, was 6% lower than its official tally).
What other Russian bloggers have pointed out is that a whole lot of other countries – Germany, the UK, Israel – have similar voting tendencies. There, more turnout means more votes for their conservative parties (Christian Democrats, Tories, Kadima, respectively). So since most readers would agree that those countries have clean elections, the “more turnout and more votes for one party MUST MEAN fraud fraud fraud!!!” thesis can’t exactly be universally valid.
This linear relation between more turnout and more votes for United Russia further makes sense because, whereas a party like the Communists has a hard core of supporters who tend to turn out reliably (with proportional representation, their votes aren’t “lost” even though the party has no real chance of winning), United Russia’s electorate is much more apathetic, a “silent majority” according to economics blogger Sergey Zhuravlev. More turnout means it manages to mobilize more people to go out and vote; naturally, a greater turnout means more votes for the party of power. This is a constant in Russian politics that stretches back to the 1990’s (recall the 1996 election when Yeltsin was appealing to Russians to go out and vote to forestall the Communist victory that would have resulted had they remained at home in large numbers).
The WSJ did not mention these counterarguments to Shpilkin, neither did they outline any of the numerous alternate methods, of which there are legion, from either the 0% Club or (especially inexcusably) from the 5% Club. Then again, if you wanted balanced Russia coverage the WSJ shouldn’t exactly be on your reading list anyway. So why am I bothering with this post?
Ah, the plagiarism! Or more specifically, non-attribution. The WSJ wrote this:
For its analysis, The Wall Street Journal designed a computer program to assemble this month’s official voting totals from the 95,228 electoral precincts across Russia. A subsequent statistical analysis revealed phenomena that scholars who study vote data say are suggestive of vote-rigging. …
There is no reliable way to use the statistical analysis to calculate how many votes were falsified. But a rough calculation that eliminates the unusually high levels of support for United Russia at the precincts with unusually high turnout raises questions about as many as 14 million of the 32.4 million votes that United Russia claimed.
Sounds very close to Shpilkin’s estimate. In fact, I publicly said as much in response to a comment by Jeremy Putley yesterday: “I suspect the WSJ method, which gives 14 million falsified votes, is basically Shpilkin redux.” I was far more correct than I realized.
In their brief section on methodology, the WSJ wrote this:
Russian election authorities post official vote results on the Internet, but not as a single database. To obtain the data for individual precincts, The Wall Street Journal wrote a computer program that downloaded 2,957 web pages posted on Russia’s Central Election Commission website.
Using another program, reporters mined the pages for precinct-level data, extracting outcomes for 95,228 precincts spread across 2,745 electoral commissions. The largest precinct, in Derbent, Dagestan, reported 3,470 votes. The smallest was one vote in Kaspiisk, Dagestan. The average precinct size was 690 votes.
Is it fair to say that all this text very strongly implies that it was the WSJ itself that came up with these models?
As well they should. In a post after that article’s publication, Shpilkin called the Wall Street Journal the “motherland of elephants” (translation: Takes all the credit for itself, not matter how implausibly).
… After the elections, I was contacted by the Moscow bureau of the WSJ, who requested a consultation on fraud calculations. They said that they wanted to repeat my calculations independently and write an article on it.
I described my methods at length, including formulas, preliminary estimates, and the rules of calculating turnout from the protocols. I naively assumed that this a consultation would merit a mention as one of the sources in the article. To my surprise, there were no links to me on their article. Furthermore, I later found out that they collectivized borrowed from not other sources too – for instance, the picture with the spikes at nice percentages for United Russia were field published by Maxim Pshenichnikov, and the [theoretical proof for the irregularity] of those peaks was provided by Dmitry Kobak.
I expressed my bewilderment in a letter to the head of the WSJ’s Moscow bureau, but he replied, “I had hoped to include you and everyone else we talked to in the story but there simply wasn’t space, particularly because we had to include the surkov news, as well.”
So if you’ve got any further questions – it’s all Surkov’s fault.
Now as far as I can see, the WSJ article makes at least two major violations of journalistic ethics and integrity here.
(1) One-sided coverage. As far as I can see, this is not an op-ed, but a regular news item. But no efforts are made to cover the numerous alternate methods or counter-arguments to Shpilkin’s methodology that have been mentioned on this blog. His argument is reproduced exactly and with all the flaws that have already been picked up by other Russian elections analysts.
(2) Non-attribution at best, plagiarism at worst. Shpilkin himself edges away from using the P-word in his public complaint letter to the WSJ, describing it as non-attribution, but I think the line is a fine one here. The WSJ replicates his exact method after consulting him. It does not cite him once in its article, making it clearly and convincingly evident to its readers that it was specifically the Wall Street Journal that “designed” and “wrote” the computer program (i.e. after getting all of Shpilkin’s formulas).
A few hours after the WSJ started getting complaints from the Russian elections blogging community, they did give Shpilkin a mention in a new blog post on the WSJ’s Emerging Europe blog – i.e., not published in print, and far less widely read – but if anything it adds insult to injury by describing him as a “Russian amateur” (in contrast to Western professionals, presumably) who only makes “preliminary estimates” (as opposed to the WSJ, which presumably has a monopoly on “confirming” them).
What an insipid, insidious rag. But can we really expect anything better from a publication that proclaims a new era of brain drain from Russia just as it is – back in the world of facts and statistics – coming to an end?