Skip to main content

In any statistical study, the best data must first be collected. The following election fraud analysis is based on the 1988-2008 Unadjusted State and National Exit Poll Spreadsheet Database.

The data source is the Roper Center Public Opinion Archives. Exit polls are available for 274 state presidential elections, 50 in each of the 1992-2008 elections and 24 in 1988.

Exit polls are surveys conducted in selected voting precincts that are chosen to represent the overall state voting population demographic. Voters are randomly selected as they leave the precinct polling booth and asked to complete a survey form indicating 1) who they just voted for, 2) how they voted in the previous election, 3) income range, 4) age group, 5) party-id (Democrat, Republican, Independent), 6) philosophy (liberal, moderate, conservative), and many other questions.

In this analysis we consider the most important question: who did you vote for? Having this information, we calculate the discrepancy between the state exit poll and the recorded vote count.

Note that respondents are not asked to provide personal information. There is no excuse for not releasing exit poll/voting results for each of the 1400+ exit poll precincts. There is no privacy issue.

Key results
 - Republican recorded presidential vote shares exceeded the corresponding unadjusted exit poll shares in 226 (82.4%) of the 274 state elections for which there is exit poll data. One would normally expect approximately 137 (50%). The probability is virtually ZERO.

- The exit poll margin of error (described below) was exceeded in 126 (46%) of the 274 polls. The statistical expectation is that the margin of error (MoE) would be exceeded in 14 (5%). The probability is ZERO.

- 123 of the 126 exit polls in which the MoE was exceeded moved to the recorded vote in favor of the Republican (the “red shift”). Just 3 moved in favor of the Democrat (” the blue shift”). There is a ZERO probability that this one-sided shift was due to chance. It is powerful evidence beyond any doubt of pervasive systemic election fraud.

- The Republicans won the recorded vote in 55 states in which the Democrats won the exit poll. Conversely, the Republicans lost the recorded vote in just two states (Iowa and Minnesota in 2000) in which they won the exit poll. If the elections were fair, the number of vote flips would be nearly equal. The probability of this disparity is virtually ZERO.

Basic Statistics and the True Vote Model
 The True Vote Model (TVM) is based on current and previous election votes cast (Census), voter mortality and returning voter turnout. Published National Exit Poll (NEP) vote shares were applied to new and returning voters. The TVM closely matched the corresponding unadjusted exit polls in each election. It shows that the exit poll discrepancies were primarily due to implausible and/or impossible adjustments required to force the NEP to match the recorded vote. The exit polls were forced to match the recorded votes by adjusting the implied number of returning voters from the previous election. These adjustments are clearly indicated by the percentage mix of returning voters in the current election..

The bedrock of statistical polling analysis is the Law of Large Numbers. As the number of observations in a survey increases, the average will approach the theoretical mean value. For instance, in coin flipping, as the number of flips increase, the average percentage of heads will approach the theoretical 50% mean value.

The Normal distribution is considered the most prominent probability distribution in statistics (“the bell curve”). It is used throughout statistics, natural sciences, and social sciences as a simple model for complex phenomena. For example, the observational error in an election polling is usually assumed to follow a normal distribution, and uncertainty is computed using this assumption. Note that a normally-distributed variable has a symmetric distribution about its mean.

The Binomial distribution distribution calculates the probability P that a given number of events (successes) would occur in n trials given that each trial has a constant probability p of success. For instance, the probability of flipping heads (a success) is 50%. In a fair election, the probability that the exit poll would flip from the Democrat to the Republican (and vice-versa) is also 50%.

The Poisson distribution calculates the probability of a series of events in which each event has a very low probability. For instance, there is a 5% probability that the recorded vote share will differ from the exit poll beyond the MoE.

The Binomial distribution converges towards the Poisson as the number of trials (n) goes to infinity while the product np remains fixed (p is the probability). Therefore the Poisson distribution with parameter λ = np can be used as an approximation to the Binomial distribution B(n,p) if n is sufficiently large and p sufficiently small. The approximation is good if n ≥ 20 and p ≤ 0.05, or if n ≥ 100 and np ≤ 10.

The margin of error is a function of the number of respondents and exit poll “cluster effect” (assumed as 0.30). The Margin of Error Calculator illustrates the effects of sample size and poll share on the margin of error and corresponding win probability.

Ohio 2004 presidential election
In the exit poll, 2020 voters were sampled in approximately 40 precincts, of whom 1092 said they voted for Kerry (54.1%) and 924 for Bush (45.7%). Bush won the recorded vote by 50.8-48.7% (119,000 vote margin). There was a 10.6% margin discrepancy. Given the exit poll data, we can calculate the probability of a) Kerry winning the election and b) of Bush getting his recorded vote share.

The Ohio exit poll MoE was 2.8%. There is a 95.4% probability that the True Vote lies within 2.8% of Kerry’s 54.1% exit poll share. The probabilities are:
 - 95.4% that Kerry’s share was between 51.3 and 56.9%
 - 97.5% that Kerry had at least 51.3%
 The Normal distribution calculates the probability P that Kerry won Ohio.
 P = 99.8% = Normdist (.541,.500,.028/1.96, true)

Bush won Ohio with a 50.8% recorded share – a 5.1% increase (red-shift) over his 45.7% exit poll share. The probability of the increase is 1 in 4852 (.02%). Which is correct, the poll or the recorded vote? How could there be such a wide disparity?

1988 presidential election
As indicated above, 24 state exit polls are listed for 1988 on the Roper Center site. These states accounted for 68.7 (75%) of 91.6 million national recorded votes. Dukakis led the 24-poll aggregate by a 51.6-47.3%, but Bush won the corresponding recorded vote by 52.3-46.8%, a 9.8% margin discrepancy. The exit poll margin of error was exceeded in 11 of the 24 states – all in favor of Bush (see the summary statistics at the bottom).

Dukakis also won the unadjusted National Exit Poll by 49.8-49.2% – but Bush won by 7 million votes, 53.4-45.6%. According to the U.S. Census, 102.2 million votes were cast and 91.6 million recorded, therefore a minimum of 10.6 million ballots were uncounted. Dukakis had approximately 8 million (75%) of the uncounted votes (see below). Of course, voters whose ballots were uncounted were interviewed by the exit pollsters. That may be one of the reasons why Dukakis won the state and national exit polls and lost the recorded vote.

Calculating the probabilties
 The probability P that 55 of 57 exit polls would flip from the Democrats in the exit polls to the Republicans in the recorded vote is given by the Binomial distribution: P= 1-Binomdist(54,57,.5,true)
 P= 1.13E-14 = 0.000000000000011 or 1 in 88 trillion!

The probability that the exit poll margin of error would be exceeded in any given state is 5% or 1 in 20. Therefore, approximately 14 of the 274 exit polls would be expected to exceed the margin of error, 7 for the Republican and 7 for the Democrat.

Given the relationship between the exit poll, margin of error and corresponding win probability, we compare the 274 state exit polls to the corresponding recorded votes. The Republicans did better in the recorded vote than in the exit polls in 226 (82.4%) of the 274 elections. The probability of this one-sided red-shift is 3.7E-31 or 1 in 2.7 million trillion trillion.

The MoE was exceeded in 123 exit polls in favor of the Republican – and just 3 for the Democrat. The simple Poisson spreadsheet function calculates the probability P:
 P = 5E-106 = Poisson (123, .025*274, false)
 P = 1 in 1.8 billion trillion trillion trillion trillion trillion trillion trillion trillion.
 The probability is ZERO. There are 106 places to the right of the decimal!
 P = .0000000000 0000000000 0000000000 0000000000 0000000000 0000000000 0000000000 0000000000 0000000000 0000000000 000005

Sensitivity Analysis
 Sensitivity analysis is an important tool for viewing the effects of alternative assumptions on key results from a mathematical model.

In pre-election polls, the margin of error (MoE) is based strictly on the number of respondents. In exit polls, however, a “cluster factor” is added to the calculated MoE. Therefore, the number of states in which the MoE was exceeded in 1988-2008 (and the corresponding probabilities) is a function of the cluster effect.

The MoE was exceeded in 126 of 274 exit polls assuming a 30% “cluster factor” (the base case). Although 30% is the most common estimate, political scientists and statisticians may differ on the appropriate cluster factor to be used in a given exit poll. Therefore, a sensitivity analysis worksheet of various cluster factor assumptions (ranging from 0% to 200%) is displayed in the 1988-2008 Unadjusted Exit Poll Spreadsheet Reference. The purpose is to determine the number of exit polls in which the MoE was exceeded over a range of cluster factors.

If there was no cluster effect, the margin of error was exceeded in 157 of 274 exit polls. In the base case (30% cluster), 126 exceeded the MoE.

Note: MoE is the average margin of error for the 6 elections, CF is the cluster factor, N is the number of exit polls in which the MoE was exceeded.
 MoE CF...N..Probability
 2.5% 0% 157 2.0E-106 ZERO
 3.2% 30% 126 8.0E-75 ZERO (base case cluster factor)
 3.7% 50% 113 1.4E-62 ZERO
 5.0% 100% 76 1.5E-31 ZERO (1 in 7 million trillion trillion)
 6.2% 150% 50 2.5E-14 (1 in 40 trillion)
 7.0% 180% 35 6.6E-7 ( 1 in 1.5 million)
 7.5% 200% 25 1.9E-03 (1 in 500)

Even with extremely conservative cluster factor assumptions, the sensitivity analysis indicates a ZERO probability that the margin of error would be exceeded in the six elections. Were the massive discrepancies due to inferior polling by the most experienced mainstream media exit pollsters in the world? Or are they further mathematical confirmation of systemic election fraud – beyond any doubt?

Overwhelming Evidence
 The one-sided results of the 375,000 state exit poll respondents over the last six presidential elections leads to only one conclusion: the massive exit poll discrepancies cannot be due to faulty polling and is overwhelming evidence that systemic election fraud has favored the Republicans in every election since 1988.

Fraud certainly cost the Democrats at least two elections (2000, 2004) and likely a third (1988). And in the three elections they won, their margin was reduced significantly by election fraud.

To those who say that quoting these impossible probabilities invites derision, that it is overkill, my response is simply this: those are the actual results that the mathematical functions produced based on public data. The mathematical probabilities need to be an integral part of any election discussion or debate and need to be addressed by media pundits and academics.

Media polling pollsters, pundits and academics need to do a comparable scientific analysis of historical exit polls and create their own True Vote models. So-called independent journalists need to discuss the devil in the details of systemic election fraud. They can start by trying to debunk the analysis presented here.

Presidential Summary

Election.. 1988 1992 1996 2000 2004 2008 Average
 Recorded Vote
 Democrat.. 45.7 43.0 49.3 48.4 48.3 52.9 47.9
 Republican 53.4 37.4 40.7 47.9 50.7 45.6 46.0

Unadjusted Aggregate State Exit Polls (weighted by voting population)
 Democrat.. 50.3 47.6 52.6 50.8 51.1 58.0 51.7
 Republican 48.7 31.7 37.1 44.4 47.5 40.3 41.6

Unadjusted National Exit Poll
 Democrat.. 49.8 46.3 52.6 48.5 51.7 61.0 51.7
 Republican 49.2 33.5 37.1 46.3 47.0 37.2 41.7

1988-2008 Red-shift Summary (274 exit polls)
 The following table lists the
 a) Number of states in which the exit poll red-shifted to the Republican,
 b) Number of states which red-shifted beyond the margin of error,
 c) Probability of n states red-shifting beyond the MoE,
 d) Democratic unadjusted aggregate state exit poll share,
 e) Democratic recorded share,
 f) Difference between Democratic exit poll and recorded share.

Year RS >MoE Probability.... Exit Vote Diff
 1988* 20. 11... 5.0E-11..... 50.3 45.7 4.6 Dukakis may have won
 1992 44.. 26... 2.4E-25..... 47.6 43.0 4.6 Clinton landslide
 1996 43.. 16... 4.9E-13..... 52.6 49.3 3.3 Clinton landslide
 2000 34.. 12... 8.7E-09..... 50.8 48.4 2.4 Gore win stolen
 2004 40.. 22... 3.5E-20..... 51.1 48.3 2.8 Kerry landslide stolen
 2008 45.. 36... 2.4E-37..... 58.0 52.9 5.1 Obama landslide denied
 Total 226. 123. 5.0E-106.... 51.7 47.9 3.8
 * 274 exit polls (24 in 1988, 50 in each of the 1992-2008 elections)

The Democrats led the 1988-2008 vote shares as measured by:
 1) Recorded vote: 47.9-45.9%
 2) Exit Pollster (WPE/IMS): 50.8-43.1%
 3) Unadjusted State Exit Polls: 51.7-41.6%
 4) Unadjusted National Exit Poll: 51.6-41.7%

True Vote Model (method based on previous election returning voters)
 5) Method 1: 50.2-43.4% (recorded vote)
 6) Method 2: 51.6-42.0% (allocation of uncounted votes)
 7) Method 3: 52.5-41.1% (unadjusted exit poll)
 8) Method 4: 53.0-40.6% (recursive True Vote)

The Democrats won the exit poll but lost the recorded vote in the following states. The corresponding decline in electoral votes cost the Democrats to lose the 1988, 2000, 2004 elections:

1988 (7): CA IL MD MI NM PA VT
 Dukakis’ electoral vote was reduced from 271 in the exit polls to 112 in the recorded vote. The U.S. Vote Census indicated that there were 10.6 million net uncounted votes in 1988. Since only 24 states were exit polled, a proxy equivalent was estimated for each of the other 26 states by allocating 75% of the uncounted votes to Dukakis. The average 3.47% MoE of the 24 exit polls was assumed for each of the 26 states. Four of the 26 flipped to Bush: CO LA MT SD.

The rationale for deriving the estimate is Method 2 of the 1988-2008 True Vote Model in which 75% of uncounted votes were allocated to the Democrat. The resulting 51.6% average Democratic share (see above) exactly matched the unadjusted exit polls (TVM #3). This article by Bob Fitrakis provides evidence that uncounted votes are heavily Democratic.

1992 (10): AK AL AZ FL IN MS NC OK TX VA
 Clinton’s EV flipped from from 501 to 370.

1996 (11): AL CO GA IN MS MT NC ND SC SD VA
 Clinton’s EV flipped from 464 to 379.

2000 (12):AL AR AZ CO FL GA MO NC NV TN TX VA (Gore needed just ONE to win)
 Gore’s EV flipped from 382 to 267.

2004 (8): CO FL IA MO NM NV OH VA (Kerry would have won if he carried FL or OH)
 Kerry’s EV flipped from 349 to 252.

2008 (7): AL AK AZ GA MO MT NE
 Obama’s EV flipped from 419 to 365.

Exit Poll Fraud Analysis

1988-2008 Unadjusted Exit Poll Spreadsheet

EMAIL TO A FRIEND X
Your Email has been sent.
You must add at least one tag to this diary before publishing it.

Add keywords that describe this diary. Separate multiple keywords with commas.
Tagging tips - Search For Tags - Browse For Tags

?

More Tagging tips:

A tag is a way to search for this diary. If someone is searching for "Barack Obama," is this a diary they'd be trying to find?

Use a person's full name, without any title. Senator Obama may become President Obama, and Michelle Obama might run for office.

If your diary covers an election or elected official, use election tags, which are generally the state abbreviation followed by the office. CA-01 is the first district House seat. CA-Sen covers both senate races. NY-GOV covers the New York governor's race.

Tags do not compound: that is, "education reform" is a completely different tag from "education". A tag like "reform" alone is probably not meaningful.

Consider if one or more of these tags fits your diary: Civil Rights, Community, Congress, Culture, Economy, Education, Elections, Energy, Environment, Health Care, International, Labor, Law, Media, Meta, National Security, Science, Transportation, or White House. If your diary is specific to a state, consider adding the state (California, Texas, etc). Keep in mind, though, that there are many wonderful and important diaries that don't fit in any of these tags. Don't worry if yours doesn't.

You can add a private note to this diary when hotlisting it:
Are you sure you want to remove this diary from your hotlist?
Are you sure you want to remove your recommendation? You can only recommend a diary once, so you will not be able to re-recommend it afterwards.
Rescue this diary, and add a note:
Are you sure you want to remove this diary from Rescue?
Choose where to republish this diary. The diary will be added to the queue for that group. Publish it from the queue to make it appear.

You must be a member of a group to use this feature.

Add a quick update to your diary without changing the diary itself:
Are you sure you want to remove this diary?
(The diary will be removed from the site and returned to your drafts for further editing.)
(The diary will be removed.)
Are you sure you want to save these changes to the published diary?

Comment Preferences

  •  this is what happens when electronic (1+ / 0-)
    Recommended by:
    Andrew F Cockburn

    polling machines are made by republicans

    •  An interesting (if tedious) question to answer (2+ / 0-)
      Recommended by:
      HudsonValleyMark, Bob Love

      would be whether there is a different distribution of results based on method of voting. If there is no difference between machine voting vs paper ballot, then the explanation must be somewhere else in the process.

      Of course if paper ballots give the result the diarist expects and machines are the source of the bias...

      •  extensively discussed in 2004 (2+ / 0-)
        Recommended by:
        Bob Love, Andrew F Cockburn

        "Machine voting vs paper ballot" may not be the right dichotomy, since optical scanners are subject to hacking.

        The hand-count precincts that year did have the smallest average Within Precinct Errors, but because there were few of them -- and because voting method is confounded with other variables -- it's unlikely that the difference was statistically significant.

        There were very large average discrepancies in lever machine precincts. A lot of those precincts were in New York, where the official results made more sense than the exit poll results.

        See, for starters, page 40 of the Edison/Mitofsky evaluation report.

  •  Accounting for possible non-fraud factors? (4+ / 0-)

    Analyses such as you present here are certainly interesting and worth pursuing. However, as a long time statistician involved with the analysis of huge and messy data sets I will just point out  there are other factors besides fraud that may affect the discrepancies in the exit polls and recorded votes  in the direction that you describe here.  

    That is, even if no fraud were present, these other factors would lead to some systematic difference between the exit polls and the reported votes. Thus, the p-values you calculate (before any appropriate adjustments) cannot be taken at face value. In other words, in the presence of these other factors, the null hypothesis isn't really that there is no difference between the exit polls and the reported votes.

    In fact you point out one factor that I would guess is fairly important i.e. the large difference between the number of voters and the votes counted.  Different methods of recording votes are known to result in different percentages of 'spoiled' or uncounted ballots and from what I've read, many of the areas with the highest proportion of spoilage tend to vote for Democrats. Thus, you could add a component to your analysis that would estimate the distribution Dem/Rep of these uncounted ballots and then re-assess the significance of the how discrepant. Note, I'm not saying that you should figure out the size of this effect by attributing all of the difference to this factor!  

    If, for example,  1/2 of the observed discrepancies could be ascribed to this factor then the p-values for the remaining differences certainly would be much larger than in your original calculation.

    Other factors:  

     1) Absentee ballots
       I'm assuming that these voters do not get interviewed so any tendency for them to vote more Republican would need to be considered.

     2) Refusal to be interviewed
        I'm not going to raise the issue of the 'grouchy' Republican here but there certainly some factors (e. g. age) that are related to both the likelihood of agreeing to be interviewed and which party they vote for.

    •  I don't know about Republicans in general, (2+ / 0-)
      Recommended by:
      HudsonValleyMark, 207wickedgood

      but I suspect that the libertarian types are much less likely to respond to any sort of poll. My father, who was strongly libertarian, refused to answer any questions from any pollster. Whether there were enough people like him to account for the difference I don't know.

    •  this writer consistently ignores non-fraud factors (1+ / 0-)
      Recommended by:
      207wickedgood

      (I'm referring to Richard Charnin, not the person who posted the diary.)

      There probably isn't enough spoilage to matter much. In 2004, the exit poll discrepancy in New York (measured by best model estimate using exit poll data only) was almost 14 points on the margin. I don't remember if there is a solid residual vote estimate, but it would not be very large. Few states had residual vote rates over 1%, but the overall "red shift" was around 6 points or more, depending on measure.

      Also, some spoilage could be attributable to fraud.

      There is some phone interviewing of early/absentee voters, but that's a tricky business.

      Non-response bias is widely believed to be a biggie.

  •  this is a complete copy-and-paste, yes? (1+ / 0-)
    Recommended by:
    Bob Love

    This is an odd situation. You appear to be reposting content in toto for a banned user. At the end, you link to the source, but you don't indicate that in fact, all(?) the text is from that link.

    It's not uncommon for people to cross-post their own writing, but to copy and paste an entire diary as if it were your own writing is uncommon. At least one commenter is, naturally, assuming that this is your work.

Subscribe or Donate to support Daily Kos.

Click here for the mobile view of the site