Dale Tavris -- World News Trust
May 18, 2018
Ever since the U.S. election of 2004, national elections (Presidency, U.S. Senate and House of Representatives) have been characterized by widespread and large discrepancies between exit polls and official election results.
In other countries, to varying degrees, exit polls have been used to monitor the results of important elections. When this is done and large discrepancies are observed, recounts have often been done or the results have been overturned. But in the United States exit polls have never been used for that purpose. That is despite the well-known fact that our electronic voting system is very vulnerable to hacking and fraud and despite widespread demands from large minorities of American citizens that exit polls be used to monitor our elections. Yet, we are told by our national news media that exit polls are not useful for that purpose.
When a discrepancy is characterized by better official results for Republicans than what is predicted by exit polls, the discrepancy is called a “red shift.”Another way of saying this is that the exit polls pose a red flag that suggests the official results have been “shifted” (by fraud?) in the Republican direction. Such a finding might then be used as a rationale to conduct a recount, or if that is not possible (because many electronic voting systems are not associated with paper ballots that can be recounted) a re-election.
An election where the discrepancy is in the other direction -- the official results favor Democrats compared with the exit polls -- would be known as a “blue shift.” Yet, since election integrity groups began monitoring these discrepancies in 2004, significant blue shifts rarely -- if ever -- have been observed. The widespread and large discrepancies between exit polls and official results in national elections since 2004 have always been red shifts, not blue shifts.
Summary of massive “red shifts” in U.S. elections since 2004
Presidential election of 2004
Red shifts initially gained widespread notoriety in the United States on Election Night 2004, with the re-election of George W. Bush, with a whopping national red shift of 5.4 percent. Exit polls predicted a 2.6 percent national popular vote margin for John Kerry, while the official vote count showed a 2.8 percent national popular vote margin for Bush. Statistical analysis of the red shift showed that the likelihood against a red shift of the magnitude seen (5.4 percent) occurring due to random chance was about 450,000 to 1.
Red shifts were seen in the good majority of states throughout the country that night. The red shifts were statistically significant in about 20 states. There were no statistically significant blue shifts. Among the 11 swing states, the following red shifts were seen in 10 of them.
||0.0% (no shift)|
There were no blue shifts of any magnitude in swing states. The red shifts were statistically significant at the 1 in 20 level in New Hampshire, Ohio, Pennsylvania, Minnesota and Florida. If we assume that the exit polls were accurate there were four states that were red shifted from Kerry to Bush (i.e. states that exit polls predicted for Kerry but that Bush won): Ohio, Nevada, New Mexico, and Iowa. Of those, the results were statistically significant at the 1 in 20 level only in Ohio. Coincidentally (or not), Ohio was the only red shifted state that made a difference all by itself in who won the Electoral College. Exit polls predicted a Kerry victory of 4.2 percent in Ohio, but Bush won it by 2.5 percent.
Brief summary of red shifts in 2006 to 2014
The 2006 national House of Representative elections, though won by the Democratic Party, were so red shifted that what the exit polls predicted to be a Democratic landslide resulted in only modest Democratic gains; red shifting of the 2010 Congressional elections (including the strange Republican win of a special Senate election in liberal Massachusetts that prevented a Democratic filibuster-proof Senate) was severe enough to bring a radical right Republican House to power that has persisted to this day; red shifting in Wisconsin in 2011 and 2012 destroyed the attempt to recall a radical right wing governor and state senators, and; there was massive red-shifting in the 2014 national House, Senate and Governor races.
There was also significant red-shifting in the 2008 and 2012 Presidential elections, but not enough to prevent Obama victories.
Presidential election of 2016
As with all Presidential elections since 2004, there was a red shift in the national popular vote of the 2016 election -- but a relatively modest one: Exit polls predicted a Clinton margin of 3.2 percent nationally, while she won the popular vote by a little less, 2.1 percent. So the red shift for the national popular vote was only 1.1 percent. However, much larger red shifts occurred in four of the five largest swing states, and a fifth swing state (Michigan) demonstrated a smaller red shift that could have made the difference, as follows:
|State||Predicted by exit polls||Official vote count||Red shift|
|North Carolina||Clinton by 2.1%||Trump by 3.7%||5.8%|
|Pennsylvania||Clinton by 4.4%||Trump by 0.7%||5.1%|
|Wisconsin||Clinton by 3.9%||Trump by 0.8%
|Florida||Clinton by 1.3%||Trump by 1.2%||2.5%|
|Michigan||Even||Trump by 0.2%||0.2%|
Of the swing states, the red shifts were statistically significant only in North Carolina, Pennsylvania, and Wisconsin. But those three states accounted for 45 electoral votes. So if Clinton would have carried additionally only those three states she would have won the Electoral College by 9 Electoral votes. Florida and Michigan accounted for another 45 Electoral votes, and if Clinton would have carried additionally those two states she would have won the Electoral College by 99 votes.
Never in the history of the United States of America has there been such a large discrepancy between the popular vote and the Electoral College in elections decided by the Electoral College. We have had three elections in our history where the presidency was not fully decided by the Electoral College: 1820 (decided by the U.S. House of Representatives); 1876 (decided by a special judicial commission that overturned the Electoral votes in 4 states, to give Rutherford B. Hayes a one vote electoral margin); and 2000 (decided by a decision of the U.S. Supreme Court by halting a recount in progress in Florida, where Florida’s electoral votes would decide the winner). Not counting those three elections, prior to 2016 there was only one instance in U.S. history where the popular vote winner and the Electoral College winner did not coincide. That was 1888, when Grover Cleveland won the popular vote by 0.8 percent but lost the election in the Electoral College.
So 2016 was a very unique Presidential election. We have the losing candidate winning the popular vote by more than 2 percent, but with large statistically significant red shifts perhaps prohibiting her from winning at least 3 states (North Carolina, Pennsylvania, and Wisconsin) and the winning candidate winning 4 states (Pennsylvania, Wisconsin, Florida and Michigan, accounting for 75 Electoral votes) by such minute amounts (0.2 percent to 1.2 percent) that we see an unprecedented magnitude of discrepancy between the Electoral College and the popular vote.
Interpretation of red shifts
There are only three possible interpretations of red shifts or any discrepancy between exit polls and official vote counts:
So we have only two plausible explanations for the large and consistent red shifts that we’ve seen: Impaired election integrity or bias in the exit polls. For reasons that are not entirely clear, our national news media has decided to almost completely ignore the possibility that the red shifts may be due to impaired integrity of our elections. Keith Olbermann had been the only major news media exception to that, in his coverage of the 2004 Presidential election. Many believe that he was fired from his job solely because of that.
When people associated with the U.S. power structure have been pressured to respond to those who believe that election fraud is the explanation for the red shifts that we see so frequently, the only alternative that they can suggest is biased exit polls. They go so far in their claims of biased exit polls that they tell us that exit polls have no legitimate place in our country for the monitoring of the integrity of elections. They continue to utilize exit polls for producing reports on voting by a large range of demographic and other variables (i.e. by sex, race, age, income, education, religion and much else) and to assist in the early calling of elections on TV. Presumably, they think that exit polls are accurate enough to use for those purposes, but not for the only purpose that would really be helpful to our democracy -- the monitoring of the integrity of our elections.
Regarding the 2004 Presidential exit polls, these people had a special name for what they claimed to be exit poll bias: the “reluctant Bush responder” -- more commonly known as the “shy Bush voter." In other words, they claimed that the observed red shifts were the result of the fact that for some unknown reason, Bush voters were less likely to want to admit to voting for their candidate than were Kerry voters.
Is it possible that such a bias phenomenon could have been at work in the vast majority of U.S. states? Perhaps, although it doesn’t sound very likely to me. So additional statistical testing by groups that were very concerned about such large red shifts in a U.S. Presidential election were performed to delve into this possibility. There were many such tests, and as a group they suggested not only that what bias might have been present didn’t contribute to red shifts, but that what bias there was probably was more likely to result in a blue shift than in a red shift. For example, one might reasonably presume that if a “shy Bush voter” syndrome was causing bias in the exit polls that such a phenomenon would have been least prevalent in precincts that were characterized by large numbers of Bush voters (where it would seem that voters would be less shy about admitting to voting for Bush.) But the statistical analysis showed just the opposite. Precincts with large amounts of Bush voters exhibited less red shifting, not more.
Furthermore, as noted above, since 2004 red shifting of our elections has been found to be widespread, large, and consistent from election to election. So now, instead of using the “shy Bush voter” syndrome to explain bias resulting in red shifting we have think of a way to explain why red shifts are so consistent from one election to another. Perhaps we could postulate a “shy conservative voter” syndrome to explain it. Personally, the idea that conservative voters are so much less likely to respond to exit polls and admit to voting for their candidate than are other voters, regardless of who the individual candidates are, and that this has happened consistently from election to election over 12 years seems far-fetched to me.
Statistical “adjustment” of exit polls by our national news media
Most people are familiar with the so-called exit poll results that are presented to us on TV following U.S. elections. These reports are presented to us to provide a breakdown of voting patterns by demographic and other variables (for example, age, race, sex, income, education, religion, etc.) Before these reports are presented to us on national television, the results are statistically “adjusted” to match the official election results. For example, if exit polls in a two-candidate race show that candidate A was predicted to win 51 percent of the male vote and 51 percent of the female vote, but there was a red shift so that candidate B received 55 percent of the total vote, the “adjusted” exit polls would show that candidate B received 55 percent of the vote of both sexes, even thought the unadjusted exit polls show candidate B winning only 49 percent of the vote of both sexes. In that way, the exit poll reports produced by our news media reflect an overall outcome consistent with the official vote count rather than the exit polls which they purport to be reporting on. The relative proportion of the vote for each variable is obtained for the reports from exit poll data (since there is no official voting data that reflects this), whereas the total vote in these reports reflects the official vote count, not the exit polls.
But our news media NEVER tells us this when they present exit poll reports.
Why is that? Why do our news media present data that they call exit poll data when in fact the total vote count is obtained from the official results, not the exit polls, and often or usually contradicts the exit polls because it is red shifted? Many people who I have talked with about this, who are more concerned with the integrity of our elections than with demographic presentation of voting patterns, voice the opinion that this represents outright fraud on the part of our national news media.
I really don’t know if they have fraudulent intent in doing this. I suppose it can be rationalized by a basically honest person by making the assumption that whenever the exit polls and official vote count contradict each other, even by large margins, one should always make the assumption that the official vote count is the true measure of voter intent (i.e. no fraud) and that the exit polls are wrong because they are biased. If one makes that assumption, then the statistical adjustments that are made to ensure that the “exit poll report” matches the official vote count makes sense. And just think of the embarrassment that would be caused by presenting the actual, unadjusted exit poll data. If candidate A is presented as winning 51 percent of both the male and female vote (as shown by the unadjusted exit polls), and yet candidate B won the election and the popular vote, what are people going to think?
But still I find it less than totally honest, to say the least, that the data is presented in this way without even telling the public that it’s being done. How can our news media just assume that the exit poll results are wrong and the official results are right? How can they believe that exit polls are valid for presenting relative voting percentages by demographic variables and calling elections early, but not valid for assessing total vote counts (i.e. monitoring the integrity of elections). And how can they justify calling these reports exit poll reports when the total vote count in the report doesn’t reflect the exit polls at all.
Why do we see so many huge red shifts and no significant blue shifts?
This an important question that we would do well to consider. Our power structure attributes this to exit poll bias, and they label anyone who takes the idea of widespread election fraud seriously as “conspiracy theorists.”Personally I do not feel insulted by being called a “conspiracy theorist." Dark conspiracies with widespread terrible consequences are commonplace throughout our world’s historical record. Most people have a difficult time recognizing them at an early stage, and that’s why so many large scale conspiracies have been successful enough to do widespread and terrible damage. But when the term “conspiracy theorist” or “conspiracy theory” is used by our national news media it is always meant in a very derogatory sense -- as if there is something dishonorable or crazy about thinking about or investigating things that look like conspiracies.
So why do we see so many red shifts in our elections and no significant blue shifts? The truth is that the Republican Party is the party of the wealthy and powerful. In economic matters they uniformly espouse and vote for policies that favor the wealthy over the poor and middle class. Consequently, they receive much more financial support from the wealthy than do Democrats (though unfortunately, in recent years the Democratic Party has been moving to the right in order to obtain some support from these people.) Furthermore, all of the electronic voting companies are owned by Republicans. With the good majority of powerful and wealthy people on the side of Republicans is it any wonder that fraudulent elections are used to help maintain their wealth and power?
Misleading and invalid national news media characterization of exit polls
As noted above, the power structure in our country either totally ignores the issue of election fraud and the potential for exit polls to monitor them by demonstrating red shifts or blue shifts or, when pressured to rebut the accusations of election integrity organizations they provide arguments that are disingenuous at best.
The best example I can think of is an article by Nate Silver titled “Ten Reasons Why You Should Ignore Exit Polls." A major reason why I say that this is the best example that I can think of is that Nate Silver is perhaps the most well-known and respected pollster in our country. Consequently, when most people read this article they take it at face value. Indeed, I have heard it cited by many people in arguing that exit polls are totally useless for the purpose of monitoring elections. Yet none of the 10 reasons for ignoring exit polls that Silver cites in his article is, in fact, a valid reason for ignoring exit polls or for suggesting that they aren’t of value in monitoring elections.
I have read many of Silver’s other articles, and they all point to the fact that he is highly intelligent and has a great deal of statistical expertise. So why would such a highly respected pollster write such nonsense? I can only conclude that he was “persuaded” to do so by some very powerful people. I have no idea what form this persuasion took, but I can think of no other reason why he would risk his reputation among knowledgeable people by writing such nonsense.
Rather than discuss each of his points separately I will group his points into categories because some of his individual reasons are highly duplicative with others. I guess he had to do that in order to reach ten reasons. Bear with me. I will try to explain this in a way that any thinking person, including those who lack any formal statistical knowledge, will be able to understand the basics of what I’m saying:
In making an argument, “circular reasoning” is the (totally invalid) process whereby you begin with the assumption that your argument is correct in your provision of background information to support your argument. If you start with that basic assumption and make it a pre-condition of the conclusion that you are arguing for, then indeed your argument will always show that your pre-specified conclusion is valid.
Silver’s reasons #2, #3, #5, and #10 are good examples of blatant circular reasoning. Reason #2 states “Exit polls have consistently favored the Democratic share of the vote.” That statement is clearly based on a comparison between past exit polls and official results. In the past Democrats have done better in exit polls than they have done in the official vote count. I guess that Silver expects his audience to assume automatically that when past exit polls have varied substantially from the official vote count that the exit polls were wrong and the official vote count was right. But that is what the whole argument is about. A proper interpretation of those past instances would be to say that the exit poll deficiencies of the past could indicate either that the exit polls were wrong (exit poll bias) or that the official vote count was wrong (election fraud).
Reason #3 states that “exit poll discrepancies were particularly bad in this year’s primaries” (He’s referring to 2008, when his article was written). This is almost a total reiteration of reason # 2, except that it uses different examples. It needs no further discussion.
Reason #5 states that “Democrats may be more likely to participate in exit polls." This is the same principle as reasons #2 and #3. If you assume that exit polls are always wrong when they differ from the official vote count, then you have to explain why. The only possible reason would be that the exit polls are biased. The most likely root reason for the bias by far is that the supporters of one candidate are more likely to participate in the exit poll than supporters of the other candidate. So if you assume that exit polls are always wrong compared to the official vote count, then anytime you see a red shift you will conclude that Democrats are more likely to participate in the exit poll than Republicans. Again, Silver is asking you to make the same assumption as in reasons #2 and #3.
Reason #10 states that “You’ll know the actual results soon enough." This statement willfully ignores why almost all, if not all election integrity activists, and many other Americans believe that exit polls need to be done: to monitor our elections for signs of fraud. Again by referring to the actual results that we will soon know when we hear the official vote count he is asking us for the fourth time to assume the falsity of the other side’s argument.
Problems with easy fixes
Another three of Silver’s reasons involve problems that can be so easily fixed that it is hard to imagine why he would bother mentioning them at all if not because he was pressured to find a way to discredit exit polls.
Reason #1 states that “Exit polls have a much larger margin for error than regular polls." That is because of the cluster sampling that is used related to the precincts that are sampled in exit polls. what’s the big deal? Use appropriate statistical tests to measure the margin of error. Problem solved.
Reason #7 states that “Exit polls may miss late voters." Silver makes no statement about how frequently this occurs or why it would happen. Common sense tells us that the only reason that exit polls may miss late voters would be if the pollsters quit working while the polls are still open. If that does happen (Silver doesn’t say that it happens, only that it may happen), the problem could be easily fixed by telling the pollsters to keep working until the polls close.
Reason #8 states that “’Leaked’ exit polls may not be the genuine article." Silver clarifies this statement by saying that sometimes exit polls are leaked early and are therefore incomplete. He ought to know that the election integrity people who leak exit polls to the public always note the time of the exit poll, and the only ones that they use in their analyses are those that are captured very late in the day. Why is it a problem that exit polls may sometimes be leaked early, as long as those exit polls are not the ones that are used to assess the integrity of the election?
Problems that are similar to or worse in pre-election polls than in exit polls
The other three reasons that Silver mentions have to do with methodological issues that are similar, worse or much worse for pre-election polls. This is important to note because Silver’s career and reputation are based on his analysis of pre-election polls, for which he has an excellent record of accuracy (until the 2016 Presidential election, which stymied virtually all pollsters). Would he tell people to ignore pre-exit polls for the very same reason that he tells us to ignore exit polls?
Furthermore, he doesn’t balance his discussion by noting some very substantial advantages that exit polls have over pre-election polls -- far more substantial than the reasons he gives for ignoring them. The very substantial advantages that exit polls have over pre-election polls are:
Reason #4 states that “exit polls challenge the definition of random sample." As noted above, this problem is far worse with pre-exit polls.
Reason #9 states that “A high-turnout election may make demographic weighting difficult." This problem to a large extent overlaps with reason #4 because demographic weighting only becomes important to the extent that a sample is not random. As noted above, exit polls are far more random than pre-election polls for the simple reason that the voters are so much easier to identify than in pre-election polls, where the voters must be reached by phone.
Reason #6 states “exit polls may have problems calibrating results from early voting." Silver acknowledges that attempts are made to account for early voting in exit polls for most states. But he doesn’t tell us that the states where no attempt is made to account for early voting are the ones where the amount of early voting is considered too small to make a difference. This objection does have some validity to it because of the difficulty in estimating the number of early voters relative to Election Day voters, but this problem pales in comparison with the difficulties with pre-election polling.
Summary of Silver’s reasons for ignoring exit polls
Silver’s resorting to circular reasoning (#2, #3, #5, and #10) in four of his 10 reasons for ignoring exit polls is very perplexing. Surely he is intelligent and knowledgeable enough to realize the total invalidity of such reasoning.
His including of problems with such easy fixes that they don’t appear to be worth mentioning (#1, #7, and #8) is likewise perplexing. This seems like a desperate attempt to include 10 reasons in his article, regardless of how weak they are.
The other three reasons (#4, #6, and #9) have some validity to them. They all pose problems in obtaining a perfectly random sample, which does introduce the potential for some bias, even though those problems are substantially less than the problems in obtaining a random sample in the performance of pre-election polls, which Silver makes a living analyzing.
But to say that exit polls cannot obtain a perfectly random sample does not provide adequate reason for ignoring them as a means of monitoring elections, especially when our electronic voting process is so vulnerable to hacking and fraud. Silver does not even mention the potential problems with our election integrity. He merely assumes in his article that election fraud is not an issue worthy of consideration, and he subtly asks his readers to assume that our elections are free of fraud in accepting his reasons for ignoring exit polls. Yes, if fair elections were an unassailable fact of life in our country then I would agree with him that exit polls would have little or no value.
One should wonder why, given the well-known vulnerability of our electronic voting machines, so little is done to monitor the integrity of our elections.
One should wonder why, at least since 2004, exit polls of national U.S. elections have consistently shown widespread red shifts of great magnitude.
One should further wonder why our national news media, over the span of 12 years since this has been happening, has said almost nothing about the re-occurring red shifts, and when they are pressured to talk about it at all, are so intent on discrediting the exit polls that show it to be happening.
One also has to wonder why, when election integrity organizations have asked to investigate the voting machines that produce highly suspicious results (i.e. vary substantially from the exit polls), they are not allowed access. They are told that the machines are “proprietary." In other words, the voting machines that are responsible for democracy in our country are considered private property, so the processes by which they count our votes are secret.
Election integrity people and organizations do not ask that elections be invalidated when exit polls contradict official election results. They ask that such situations be thoroughly investigated with a hand recount of the vote and that we stop using voting methods that cannot be recounted. As noted by the U.S. election integrity organization, US Counts Votes, in their National Election Data Archive Project, “Without auditability or transparency in our election systems, the role of exit polls as a trigger for further scrutiny is of paramount importance." Seriously monitoring our elections with exit polls is a small price to pay for an election system that all Americans can have confidence in.
Dale Tavris has worked as a public health physician/epidemiologist for 40 years, with state departments of public health, the U.S. Air Force, the Medical College of Wisconsin, and the Food and Drug Administration. In that capacity, he has authored 39 publications in peer-reviewed medical or public health journals.
Since 2004 he has been actively involved in the national election reform movement, serving in a volunteer capacity with the Election Defense Alliance for a few years as their data coordinator. He has written dozens of online articles about election fraud. In 2007 he co-authored a journal article on election fraud: “Fingerprints of Election Theft: Were Competitive Contests Targeted.” He has written and published three books, including two of a political nature: “The Unfulfilled Promise of the American Dream: The Widening Gap between the Reality of the United States and its Highest Ideals,” 2011; and “Democracy Undone: Unequal Representation, the Threat to our Election System, and the Impending Demise of American Democracy,” 2012.