On Monday, actor Tim Robbins caused a stir when he tweeted out a Facebook meme, charging that CNN and The New York Times are blind to a massive conspiracy going on right beneath their noses. It had close to 1,000 retweets when Robbins apparently deleted it.
A quick glance is enough to know that there are problems with the meme. The exit poll numbers are wrong. In Massachusetts, for example, CNN reported that exit polls showed Clinton winning by 2 points, which is very close to her 1.4 percent margin in the final results. In Alabama, CNN reports the exit polls showing Clinton with a 57-point margin, the Facebook meme claims it was 44.7 points, and the final result was 60.4 percent.
But where did Tim Robbins come up with these numbers? I decided to do a bit of reporting, and I ended up chasing this Facebook meme down a rabbit-hole of misinformation and conspiracism. It offers a pretty good case-study of how bullshit can come to dominate our online discourse.
The meme was created by Lee Camp, a political comedian who hosts a weekly show on RT, the Russian foreign news network. It has over 2,000 shares on Facebook as of this writing.
Via email, I asked Camp for his source, and he pointed me to a post on Reddit by a user who goes by the handle “turn-trout.” Turn-trout, who didn't respond to a message seeking comment, claims that these are unadjusted exit polls, and links to a spreadsheet purportedly showing wide discrepancies between the raw data and the final results.
The spreadsheet was created by Richard Charnin, who writes a blog devoted to “JFK conspiracy and systemic election fraud analysis.” Charnin's spreadsheet appears to be the basis of a broad swath of viral Internet content alleging widespread election theft during the 2016 primaries, including the work of Free Press editors Harvey Wasserman and Bob Fitrakis. Charnin seems to think that exit polls can reveal that virtually all our elections have been rigged, writing, “in the 1988-2008 presidential elections, the Democrats won the exit polls by 52-42%; they won the recorded vote by just 48-46%, an 8% discrepancy.”
After the New York primary, which was a total mess, Charnin sent an email to his list with a detailed statistical analysis based on the assumption that 5 percent of New York voters had been disenfranchised, and that 75 percent of them would have voted for Sanders if they'd had the chance. Given that pre-election polls showed Clinton leading the state by 13.5 percent -- and that New York City, where the biggest irregularities occurred, was a Clinton stronghold -- this is what statisticians call “pulling an assumption out of your ass.”
I exchanged some... interesting correspondence with Charnin. After calling me “very biased and misleading” for a recent piece, he claimed that “ALL exit polls are forced to match the recorded vote.” I asked him whether the exit poll data in his spreadsheet were unadjusted, and he said that they were the data released by major media organizations. He then told me that “the mainstream media won't dare touch the Third Rail - ELECTION FRAUD,” but it's cool because “Tim Robbins just talked about it.” Finally, the truth emerges.
Virtually all of these claims are based on the idea that exit polls are a telltale sign of fraud. In a follow-up tweet, tim Robbins explained that, “exit polls are historically pretty accurate,” and “are a heads-up on vote tampering.” Turn-trout agrees, writing, “Exit polls have historically and throughout the world been used as a check against, and rough indicator of, the degree of election fraud.” This is also the basis of claims by Wasserman and Fitrakis – who point to the precision of German exit polling to emphasize the point – and Steven Freeman, a Penn State psychologist who authored the book, Was the 2004 Presidential Election Stolen?: Exit Polls, Election Fraud, and the Official Count.
I asked Joe Lenski, executive vice president of Edison Media Research, about all of this. Edison has conducted all of the exit polls for major U.S. media organizations since 2003, and Lenski has also done exit polling in Azerbaijan, Georgia and Venezuela.
As for using his results to suss out fraud, he says that American exit polls are “just not designed for that type of precision. They're surveys, and like any other survey, they have a margin of error. The precision that a lot of these people are talking about just doesn't exist with our polls.”
In emerging democracies, says Lenski, “the exit polls are designed specifically to catch any manipulations of the vote count, and also to bring some transparency so voters can trust the vote count. They have a lot more locations in the sample, they do a lot more interviews and they use a much, much smaller questionnaire. In some cases, they just ask, 'who did you vote for?'” A brief questionnaire, he explains, increases the response rate. “The more interviews you do, the more locations you cover and the shorter the questionnaire, the higher response rate you'll get, and that all leads to a much smaller margin of error.”
The news organizations that sponsor our exit polls are just looking for a sense of who voted, and what motivated them to vote. They use longer questionnaires – typically with about 20 questions – and it takes a little longer for voters to fill them out. Lenski says that while the methodology hasn't changed much since exit polling was first introduced in 1967, the fact that news organizations post preliminary data in real-time leads to “a lot of commentary on social media.”
When people claim that early exit poll data are “unadjusted,” they're wrong. Edison adjusts its data throughout the day to compensate for “non-response rates and other sampling issues that come up when we conduct the survey.” When someone is approached to take a survey and says 'no thanks,' Edison's pollster notes that individual's gender, race and approximate age. They then adjust the raw data to match the demographics of the people they saw voting at that location during the day.
The early data news organizations post on their websites has already been adjusted. In states with large numbers of absentee voting, the data are also merged with phone interviews conducted before Election Day. If the number of absentees is significantly higher or lower than expected, they adjust for that difference as well. And then, finally, the data are adjusted again to match not the results of the election statewide, but the vote tallies at the specific polling places they surveyed.
Lenski stresses that pre-election polls are also adjusted to conform their samples to what pollsters know about the populations they're trying to measure. The irony of all of this is that the adjusted data are far more accurate than the raw data.
“Remember the process here,” says Lenski. “In most of these states, we're talking about having interviewers at several dozen locations. The vote returns at those locations are used in the model, and the vote counts at the county level are used in the model, so if there were some sort of fix involved, there would have to be hundreds or thousands of people manipulating the data at all of those locations to make it work.”
So there you have it. They say a lie can travel halfway around the world before the truth can get its boots on, and that's especially true of the internet. Here we have an example of an actor citing a comedian who picked up a claim from an an anonymous Reddit user citing preliminary exit poll data put together by a JFK conspiracy theorist. Bringing it all full circle is The Hill, which ran a story titled, “Actor Tim Robbins blames Sanders losses on ‘voter fraud,’” which will no doubt be shared thousands of times on Facebook and Twitter.
I asked Lee Camp if he was “interested in the fact that this is factually inaccurate and really misleading? I mean, can a meme be retracted? Is that something that would interest you?” He didn't respond.
