Exit Polls Versus Reported Vote Counts
By Theodore de Macedo Soares
According to the exit polls conducted by Edison Research, Biden easily won the presidency. As in past elections, the pattern of overwhelming discrepancies between the exit poll results and the unverified computer vote counts, always favoring the more politically conservative candidate, continues in this election. Incorporating the exit poll results in the New York Times interactive Electoral College map, Biden wins the presidency with a count of 328 electoral votes versus 210 for Trump. The actual electoral vote count was 306 for Biden and 232 for Trump.
Exit polls were conducted in 24 states. In 22 states the discrepancies between the exit polls and the vote count favored Trump. As the table below shows, in 11 of these states the discrepancies favoring Trump exceed the margin of error of the state’s exit poll.
As in the 2016 Democratic Party primaries and general election (and prior elections), the overwhelming discrepancies in the 2020 Democratic Party primaries and general election, always favoring the more politically conservative candidate, are a near statistical impossibility. The source of the problem is systemic. Either the exit polls and the pre-election polls have been improperly conducted or the vote counts are corrupt.
The exit polls conducted by Edison in this coronavirus pandemic year used the same methodology as in previous years. In 2016 absentee and early voting represented about 40% of the votes, this year it will exceed 60%. Early voters were submitted the exit poll questionnaire at their voting locations same as on election day. Telephone interviews were conducted of absentee voters as in previous years. See CNN article.
The possibility that our vote counts are corrupt cannot be dismissed off-hand or ignored. Computer vote counts are never verified by full hand counts and the vote counting software is proprietary—hidden from view and inaccessible to the public.
Election results in the United States are determined by vote counting systems supplied by three private corporations: Elections Systems & Software, LLC, Dominion Voting, and Hart Intercivics.[1] According to the Pew Research Center, 94% of all voters in 2016 voted on electronic voting machines or optically scanned ballots nationwide.[2] Members of Congress observe the fact that their proprietary software, inaccessible to elections officials and the public undermine trust in our elections.[3]
The possible means to corrupt the vote is not hard. As every vote counting machine has the vendor’s software custom updated for each state prior to every election, the votes can be corrupted by hacking/altering this software at the source and either the vendor or state functionaries will install this software in all the state’s machines.
The United States remains one of the few major democracies in the world that continue to allow computerized vote counting—not observable by the public—to determine the results of its elections. Countries such as Germany, Norway, Netherlands, France, Canada, United Kingdom, Ireland, Spain, Portugal, Italy, Denmark, Sweden, Finland and many other countries protect the integrity and trust of their elections with publicly observable hand-counting of paper ballots.
[1] Cohn, Jennifer. 2018. “Voting Machines: What Could Possibly Go Wrong?” New York Review of Books. Available at https://www.nybooks.com/daily/2018/11/05/voting-machines-what-could-possibly-go-wrong/
[2] DeSilver, Drew. 2016. “On Election Day, Most Voters Use Electronic Or Optical-Scan Ballots.” Pew Research Center. Available at http://www.pewresearch.org/fact-tank/2016/11/08/on-election-day-most-voters-use-electronic-or-optical-scan-ballots/
[3] Collier, Victoria & Ptashnik, Ben. 2016. “Members of Congress Call for End to Mass Voter Suppression and Insecure Elections.” Thruthout.org. Available at https://truthout.org/articles/members-of-congress-call-for-end-to-mass-voter-suppression-and-insecure-elections/
Exit polls downloaded from CNN’s website at poll’s closing: ALABAMA – ARIZONA – CALIFORNIA – COLORADO – FLORIDA – GEORGIA – IOWA – KENTUCKY – MAINE – MICHIGAN – MINNESOTA – MONTANA – NEVADA – NEW HAMPSHIRE – NEW YORK – NORTH CAROLINA – OHIO – OREGON – PENNSYLVANIA – SOUTH CAROLINA – TEXAS – VIRGINIA – WASHINGTON – WISCONSIN – NATIONAL
Please credit this article if you use these exit polls or figures derived from them in your publications.
Notes:
[1] Exit polls (EP) conducted by Edison Research and published by CNN shortly after the closing of polls for the state and downloaded by TdMS. Edison Research conducted one national EP and EPs in 24 states. Exit poll results are derived from the gender category–the proportion of men and women voting for each candidate. As these first published exit polls are altered/adjusted to conform with the unverified computer vote counts, the discrepancies shown above are adjusted to near zero in the final EPs.
The states marked by an asterisk are states where the vast majority of the state’s polls close an hourly earlier than the rest. Edison Research uses their access to these vote counts to better match their exit polls to the unverified computer vote counts. The discrepancies, therefore, between the exit polls and the vote counts may be greater than shown above.
[2] New York Times reported vote counts. https://nyti.ms/3mN2ujW
[3] The margin columns subtract Trump totals from Biden’s. A Biden win is shown by a positive sign and a Trump win by a negative sign.
[4] Note that the Margin of Error (MOE) is for the differences between the two candidates (at 95% CI). This MOE is about double the usual MOE for each candidate. MoE calculated with multinomial formula discussed in sections 2 and 4 in: Franklin, C. The ‘Margin of Error’ for Differences in Polls. University of Wisconsin, Madison, Wisconsin. October 2002, revised February 2007. Available at: https://abcnews.go.com/images/PollingUnit/MOEFranklin.pdf
Could you comment on this article about exit polls in The Nation? https://www.thenation.com/article/archive/reminder-exit-poll-conspiracy-theories-are-totally-baseless/
In the article, they interview Joe Lenski, VP of Edison Media Research. The article describes the methodology Lenski says Edison uses, and goes on to comment on “raw data”:
“In most states, Edison conducts phone interviews before Election Day to capture absentee and early voting. Then, on Election Day, they send staff to between 15 and 50 polling places per state, and they ask between 500 and 3,000 voters to fill out questionnaires indicating which candidate they voted for and what issues are important to them. In order to account for those voters who refuse to fill out a questionnaire, exit pollsters have to adjust their survey data. Lenski says that about 50–60 percent refuse to participate. When someone says no, the pollster notes the person’s rough age, race, and gender. They then weight their data to match the population that voted at that location.
Some media outlets post preliminary data when the polls close—that’s the supposedly raw data that, according to the conspiracy-minded, reveal the fraud. But those data have already been merged with the results of those telephone interviews, and they have already been adjusted throughout the day (the interviewers send in their survey results in three waves). Unadjusted data are never released. (If you Google “exit polls adjusted New York,” you’ll get back dozens of posts claiming that the “unadjusted exit polls” varied significantly from the final results. All of those posts are dead wrong, as none of their authors have any idea what the unadjusted data looked like.)”
This means that the preliminary data posted by CNN and other media outlets is *already* adjusted to match the demographics of the people who respond to the poll, presumably to correct for various forms of non-response bias. Then later it’s adjusted based on the final results, which I agree is an issue in using exit polls from after votes have been counted (which you don’t do).
I’m wondering how the adjustments that are (according to Lenski) already present when the first exit polls are released affect your methodology.
Hi Ethan. Thanks for linking the 2016 article by Joshua Holland in the Nation. The comments section well reveal the faults in the article. No one claims that the first published exit polls, as used here, are the “raw” data. All pollsters necessarily adjust their polls to the demographics, such as gender, age, education, etc., of the respondents, to ensure accuracy.
Joshua confuses these necessary adjustments with the alteration of the exit polls to match the incoming vote totals. These vote counts have NO demographic information so they cannot be used to increase the accuracy of the exit polls. They can only be used to match the exit polls to the vote counts—the exit polls are thus effectively trashed by Edison Research, their creator.
To answer your question, the first published exit polls used here provide the most accurate information available as to who won the election and why.
Please keep in mind, that this alteration of the exit polls to match the vote count, already occurred with the states marked by an asterisk in the table above. In Florida, for example, much of the state, save for a sliver in the panhandle, the polls in the earlier time zone closed an hour earlier than the tiny section.
As communicated to me by Mr. Lenski, the Executive Vice President of Edison Research, they incorporate the vote returns from the earlier time zone section in their exit poll before they are first published. This means that where, as here, their exit polls show Trump leading by less than one percent, it is quite possible that Biden had in fact a significant lead in the state in the exit polls prior to incorporating the possibly corrupt vote totals in their exit polls.
Makes sense, thanks for your prompt response!
It could also be the other way around though, couldn’t it: data incorporated from early time zones could skew their polls towards the other candidate as well? Does this mean Edison’s exit polls are exceptionally unreliable for states like South Dakota, Idaho, Kentucky, and Tennessee which have huge swaths of the state cut up by time zones? This could be a problem if there is any sort of regional bias towards one party or candidate over the other across the time zones in these states.
Shouldn’t states close all their polls at the same time? Has any proposed shifting poll hours forward by 1/2 hour in the western time zone and backward by 1/2 hour in the eastern – or is that too rational?
Most states with two time zones close their polls at the same universal time. It is curious that some states close their polls according to their time zone. They certainly won’t change their practice to prevent fudging of exit poll results!
Thank you for this important work, as usual the media is ignoring the problem of likely vote flipping.
Given tinyurl.com/palast2020, it’s very likely that the red shift is mostly, if not wholely, due to uncounted/rejected ballots.
Which is not to say we should be using fraud friendly voting machines.
How curious. Of my 3 posts, only this one made it past “moderation”.
The one suggesting changing polling hours so that all polls in a state open and close simultaneously, and the one suggesting checking how much of the red shift is accounted for by rejected/uncounted ballots were apparently immoderate.
Thanks for your comments. A bit late but they are up!
Hi John. Just watched the video you linked. It is excellent. I was astonished to learn that 1/3 of the mail-in ballots in some Georgia counties were not counted!
Is it possible to add a column that shows the make/model of the computerized vote counting equipment? We know that almost all votes are counted on ES&S, ex-Diebold, and Hart systems. It’s hard (but not impossible) to imagine that all three with their different software can flip in one direction.
This writer is aware that vote flipping can just as easily be done on optical scan systems. Touch screen systems (DREs), in fact, might be harder to flip because once a voter inserts a ballot into a scanner the voter has no idea how it’s tabulated.
Unfortunately, the public is lured into thinking that the possibility for fraud is eliminated by simply having paper ballots.
Unfortunately polling place equipment tends to not be uniform across any give state. That stuff is usually decided at a county level. Verified Voting has a quality breakdown of equipment used by county if you’re interested:
https://verifiedvoting.org/verifier/
I would look at KY for a granular look at Voting equipment disparities.
That state has counties that use all hand-marked and most of the eastern part of the state still uses DRE’s. That is where I would check for patterns that don’t match exit polling. In the Senate race in particular. McGrath was polling well in KY. She lost by a lot,
Good idea!
Thanks for doing this important work TDMS! I look forward to following your updates and analyses. Keep up the great work.
In 2016 general election you also monitored many important Senate races and created a table of results. Looking at your table for Biden vs Trump, some states with big deviations, might have suspect Senate Races, did you get to monitor any senate races, and if so will you put a table of results on your site?
thanks again,
Lindsay V
Hi Lindsay, the table on the Senate races will be published tomorrow. Same story. Oops, November 23, 2020–still working on it!
As someone who does statistical hypothesis testing all the time in my line of work, I have always been rather uncomfortable with the apparent standard of measuring the plausibility of election fraud by the number of percentage points outside the “margin of error” for a poll. This is a strange number to use because the actual probability associated with this number depends heavily on the sample size and the expected proportion deviations are being tested from. It’s even more bizarre to me that the US state department uses this as its metric for crying election fraud in foreign countries, rather than something more standardized. A p-value is a crystal-clear, standardized value that can be interpreted universally and we really ought to be obtaining p-values from simple hypothesis tests for any exit poll vs. reported count comparisons. It’s clear many of these deviations are extremely improbable and once you obtain p-values from them you can really get a real sense of JUST HOW improbable.
Thank you, Jimbo, for your thoughtful comment. I agree with you; p-values would clearly show the improbability of the results. They would, I think, be best illustrated with probability graphs. Want to help create them? Ted
[…] polls are used for, at least internationally, is to serve as a check on the election results. As TDMS Research points […]
[…] polls are used for, at least internationally, is to serve as a check on the election results. As TDMS Research points […]
[…] are used for, not less than internationally, is to function a test on the election outcomes. As TDMS Analysis factors […]
As you update table of results, I hope you will archive (can be internally) each previous table, since in some states it might be telling, since each day forward might be showing a trend, because the portion of electronically at poll votes, in proportion to mail-in, becomes smaller in many states. Might help expose where the vote tampering pointed to by the large deviations from exit polls shown in your tables is occurring. As time allows I will try to remember to archive your postings.
Also, posting some Senate Races, in those 11 states with large presidential deviations, which also happened to have big deviations from election pre-polls on the Senate Races, might show something interesting.
thank you,
Thank you for the unadjusted Presidential exit polls. Will you have the Senate and Governors EPs soon?
Very Soon!
Nov 23, 2020 still working on it!
have you noticed that in 2016 and 2020, the steal from third parties. i compared the exit poll total percentages for most states for biden and trump and compres them with ap final results. almost every state besides california and maybe michigan the exit polls favor third parties compared to the vote count
Yes. Exit poll/vote count discrepancies also apply to third parties. The numbers in my tables, include such third party shifts usually going to the Republican candidate.
You are likely aware of Greg Palast’s reporting on voter disenfranchisement of likely left voters, for example, tinyurl.com/palast2020. Have you been in touch with him? Wouldn’t it be interesting to calculate how many uncounted ballots would bring the reported results in line with the unadjusted exit polls and compare that to the number of voters illegitimately purged?
Also, is there any effort being made to organize an independent exit poll, at least in swing states?
Exit polling interviews those who voted. As voter suppression efforts generally prevent people from voting (and therefore participating in exit polls) it would not affect exit poll accuracy. Those who manage to vote but with a provisional ballot may or may not get their ballots counted. The availability of data for uncounted provisional ballots is lacking.
There have been some independent exit polls conducted but not, to my knowledge, at a state level. It is a good idea but the will and associated funding for such endeavors are lacking.
…
Edit-Just watched the video you linked. It is excellent. I was astonished to learn that 1/3 of the mail-in ballots in some Georgia counties were thrown out as defective! This would definitely affect exit poll accuracy. Preventing people from voting, not counting the ballots of those who have, and suspect computer counts all, in my estimation, add up to what we have in the US.
I realize this is a lot of work, and thank you for it.
I would continue it for you, but you have the data exit poll data before it was forced to conform to the “actual vote”. In one of your responses in this discussion indicated you have some important Senate races exit poll data before conformity.
Out of my own pocket, given I think this is important enough, I am willing to pay you to post some of the Senate Races in the same manner you posted the Presidential race. It will important to see if there is similar shifts to republican candidates in important Senate races that had large shifts to republicans? Lindsay at [email protected] Thank you,
Thank you for your offer and appreciation of this work. Work on the US Senate tables have been unfortunately delayed. It should go up in a few days. The discrepancies are large and all in favor of Republican candidates. No need to pay me! Donations, however, are certainly always welcomed!
Thanks again, Ted
The national vote count discrepancy has been greatly reduced.
Yes it has. With the continued vote count, the discrepancy is now down to 2% but given the large number of respondents (21,573) to the poll the margin of error is small at 1.3%. The remaining discrepancy still exceeds the margin of error. I will update this table in a few days.
But there were a large number of absentee and early voters in this election. They could not be reliably sampled for an Exit poll.
A more reliable evidence for fraud will be substantial discripency in vote share from different voting or counting equipment, in mixed states.
Has it ever been determined how Edison Research receives its vote counts after the polls close? I mean the technical method that is employed. Do they have volunteers that constantly look at the states/counties’ voting results websites, their own proprietary software that “read” these same websites, or is there a direct data feed from the states/counties that give them the voting count data that they then send to their subscribers via NEP?
[…] ← Previous […]
[…] day 2020. The computerized vote counts has the Republican Party with a two seat advantage. Like the 2020 and 2016 presidential elections and the 2016 U.S. Senate elections the 2020 US Senate races show […]
[…] day 2020. The computerized vote counts has the Republican Party with a two seat advantage. Like the 2020 and 2016 presidential elections and the 2016 U.S. Senate elections the 2020 US Senate races show […]