My Letter to Nate Silver on the Huge Exit Poll Discrepancies We’re Seeing in the Dem Primaries [View all]

Dear Nate:

I have followed your column with great interest from time to time. As an epidemiologist for over 40 years, I have a good amount of experience in statistics, and I have been very impressed with the great care you put into your statistical modeling, the way you explain it, and the accuracy of most of your predictions – Of course this Democratic primary season has been an exception to that accuracy, as it has frustrated and confused all pollsters, most notably in Michigan, but in many other states as well.

I have a big favor to ask of you, not on my behalf alone, but on behalf of our whole country: I have been struck by the huge discrepancies I have seen this year between exit polls and official election results in the Democratic Party primaries – most of them well outside of the margin of error. I believe that the following exit poll discrepancies in the Democratic primaries include all the primaries where exit polls have been taken this year:

Arkansas: 6.7 in favor of Clinton (official count compared to exit polls)
Alabama: 15.7 in favor of Clinton
Tennessee: 8.8 in favor of Clinton
Virginia: 4.5 in favor of Clinton
Georgia: 12.4 in favor of Clinton
Texas: 9.9 in favor of Clinton
Massachusetts: 7.8 in favor of Clinton
Oklahoma: 6.8 in favor of Sanders
Vermont: 0.9 in favor of Clinton
Mississippi: 10.4 in favor of Clinton
Michigan: 4.8 in favor of Clinton
Ohio: 10.2 in favor of Clinton
Florida: 3.2 in favor of Clinton
North Carolina: 1.8 in favor of Clinton
Illinois: 4.2 in favor of Clinton
Missouri: 4.0 in favor of Clinton
Wisconsin: 13.8 in favor of Clinton
New York: 12.0 in favor of Clinton

All of these discrepancies except one (Oklahoma) point in the same direction – They favor Clinton in the official count compared to the exit polls. In all but two of them, Vermont and North Carolina, they exceed the margin of error. I have not done formal statistical tests on this, but I’m sure you would agree that the odds against this happening by chance are astronomical.

I believe you would also agree that exit polls are likely to be far more accurate than pre-election polls: They measure how voters actually voted, rather than how they intend to vote at a later date; they do not rely on models (which can often be misleading) which estimate which voters are more likely to vote, and; it is far easier to get an accurate random sample because they do not rely on telephone samples, which are likely to misrepresent the population of actual voters. It has been pointed out that this last issue can also be a problem with exit polls when early voting /absentee ballots are taken into account. But that problem can be easily addressed by looking at exit poll discrepancies separately for early voting/absentee ballots vs. Election Day voting. And I believe that such an analysis would likely show some very interesting and informative results.

More importantly, exit polls are far more important than pre-election polls, in that they can be and are often used to monitor the integrity of elections (in other countries). Taken in the context of other extra-ordinary events happening during this Democratic primary season (I don’t know if the same thing applies to the Republican primaries because I haven’t been following them closely), the exit polls I quoted above make me extremely concerned that we are seeing here massive election fraud that threatens to destroy our democracy. The following items are the context that I’m talking about:

1) We have also seen massive voter disenfranchising in at least Arizona and New York. In Maricopa County, AZ (which constitutes about half the voters in the state), polling places were reduced from the previous election from about 200 to 60, with the result that voting lines reached as long as half a mile, and voters had to wait in line for several hours to vote. The result was that Election Day voters in that county (who voted heavily for Sanders) constituted less than 15% of total votes in the county (the rest being early voters, who voted heavily for Clinton). In both states, tens or hundreds of thousands of would-be voters who say that they were registered to vote in the Democratic primary found themselves to be purged from the voter rolls on Election Day. In Brooklyn alone, 70,000 would-be voters were purged from the voter rolls, and Mayor de Blasio (who endorsed Clinton) said “I am calling on the Board of Elections to reverse that purge”.

2) Sanders does far better than Clinton in caucus states and in primary precincts where ballots are hand counted rather than counted by machine. In both cases, massive election fraud would be much more difficult to perpetrate.

3) A public citizen observer, who attended a random precinct audit in Chicago, testified at a hearing that she observed that 21 Sanders votes were erased and 49 Clinton votes added to a hand count audit, in order for the audit to mimic the official results tabulated by the machine. As noted above, Illinois exhibited a 4.2% exit poll discrepancy, and if the official results were close to what the exit polls showed, Sanders would have won Illinois. We don’t know how many audited precincts in Illinois were characterized by apparent fraud (in both the initial count and the audit) and did not come to our attention because there was no vigilant public observer there to report her findings. I imagine that this finding by the public observer is just the tip of a very large iceberg.

This isn’t the first time this has happened, by any means. The most notable previous example is the Presidential election of 2004, which was characterized by substantial exit poll discrepancies nationally and in many states, most notably Ohio. Because the Electoral votes from Ohio determined the Presidency in 2004, it was thoroughly investigated. Even as early as January, 2005, the obvious “irregularities” in Ohio were so great that Senator Barbara Boxer officially objected to the results of the election, which required a public debate and vote in the U.S. Senate. Following numerous investigations by untold numbers of individuals and groups, eventually a hearing was to be held at which Michael Connell, Karl Rove’s “IT guru”, was to testify as to how he helped to orchestrate a massive electronic switching of votes in Ohio from John Kerry to George W. Bush on Election Day 2004. He had already signed an affidavit to that effect. Unfortunately, he died in a plane crash shortly before he was due to testify (just a coincidence?).

So let me now get back to my request of you. You are a very well-known and highly respected public figure. You discuss in your columns a great deal about your pre-election poll findings and methodology. But I see almost nothing in them about exit polls. I did read on the 538 Website, while the votes were being counted in Ohio and New York, a brief discussion of how your own final exit poll differed substantially from what you were seeing in the official results, and surprise over that fact, but the discussion was very brief, I have seen nothing on the subject from the 538 Website since that time, and I can no longer find that brief discussion at your website, though I saved the link.

So what I am requesting of you is that you begin some serious discussion on your website of the substantial exit poll discrepancies that we’re seeing in the Democratic primaries. I request that the discussion focus on the implications of those discrepancies and perhaps some further thorough analysis of them, on par with the analyses you devote to pre-election predictions.

I know that that is asking a lot of you. I realize that our national news media castigates anyone who dares to question the integrity of our election system. But our democracy and the fate of our country and the entire world depend on it. If election fraud is being perpetrated here to the extent that I believe it is, and if it is allowed to stand, our democracy is rapidly being destroyed. You can shine a lot of light on this issue with thorough and intelligent analysis and discussion on par with what you devote to pre-election polls. I know that that would result in serious risks to you and your career, and only a very brave person in your position would do this, but I am making the request because the fate of our country depends on shining a light on this issue and giving it a great deal more public attention than it has received.

Thanking you in advance,

Dale Tavris


Note to DUers:

I sent the letter to Nate earlier this afternoon.

Since then, for the sake of posting on DU, I’ve made some slight changes after I sent the letter, based on new information that came to me (additional exit polls) or something that I decided would be worth adding (Karl Rove’s electronic manipulation of the 2004 Presidential election vote, which gave George W. Bush the win in that election). If you agree with me that this is an issue of utmost and urgent importance and you are a member of other organizations where publicizing this issue would be helpful, please consider doing so.