Oct. 30, 2012
My comments in bold follow selected paragraphs from Chapter 1 of a new text Exit Polls:Surveying the American Electorate, 1972-2010 by Samuel J. Best, University of Connecticut and Brian S. Krueger, University of Rhode Island.
The authors write:
“Despite the unique insights that exit polls can provide about the composition and preferences of voters, they are seldom used after the days immediately following an election. Once media organizations have tapped the exit polls for explanations of electoral outcomes, they often disappear from the public eye. Some scholars may use them over the next year or two to explore the voting behavior of certain subgroups, such as Hispanics, women, or young people, but for the most part they recede into memory, rarely used beyond the next national election.”
“Unfortunately, few efforts are made to consider the behavior of voters over time. Historical context typically centers on comparing an election to its most recent predecessor, such as contrasting the 2008 presidential election with the 2004 contest. Rarely are exit poll responses tracked and analyzed over time, leaving many important questions understudied. For example, how have various subgroups in the electorate evolved over time? Have their relative sizes in the active electorate increased or decreased? Have their voting patterns grown increasingly partisan or independent? Which subgroups in the electorate behave similarly through the years?”
In 2010, I wrote Proving Election Fraud: Phantom Voters, Uncounted Votes and the National Exit Poll. My new book was published on Oct. 27, 2012: Matrix of Deceit: Forcing Pre-election and Exit Polls to Match Fraudulent Vote Counts
I created the 1988-2008 State and National Presidential Exit Poll Spreadsheet Database based on the Roper (University of Connecticut) election data archive.
In the 1988-2008 presidential elections there were 274 state exit polls – and 226 red-shifted to the Republican. The probability is ZERO (3.7E-31). Of the 274 polls, 126 exceeded the margin of error; only 14 would be expected to do so at the 95% confidence level. Once again, the probability is ZERO (8E-75). Of the 126 polls which exceeded the margin of error, 123 red-shifted to the Republican. The probability is ZERO (5.4E-106).
“In the weeks and months that follow, exit polls are used time and again to give meaning to the election results. Newly elected officials rely on them to substantiate policy mandates they claim to have received from voters. Partisan pundits scrutinize them for successful and failed campaign strategies. Even political strategists use them to pinpoint key groups and issues that need to be won over to succeed in future elections.”
But what if the final, adjusted exit polls can be shown to be mathematically impossible? If so, the fact that they must be adjusted to conform to the recorded vote indicates that the recorded vote must also be impossible.
“Unfortunately, these same exit poll results are not easily accessible to members of the public interested in dissecting them. After appearing in the next day’s newspapers or on a politically oriented website, they disappear quickly from sight as the election fades in prominence. Eventually, the exit polls are archived at universities where only subscribers are capable of retrieving the data. But nowhere is a complete set of biennial exit poll results available in an easy-to-use format for curious parties.”
That is why I created the 1988-2008 presidential state and national exit polls using Roper as the source.
This graph summarizes the discrepancies between the1988-2008 State Exit Polls vs. the corresponding Recorded Votes
“Second, and far more troublesome for the reputation of the exit polls, the preliminary exit poll results showed a partisan skew. They overstated Bill Clinton’s share of the vote by 2.5 points in the 1992 presidential race and understated George H. W. Bush’s share by 2.5 points, giving the impression that Clinton won by a far greater margin than the officially tabulated votes indicated.”
“The raw exit poll data had never been deemed “accurate” in the past prior to being weighted to the actual results, but with the release of early results, observable, but correctable, sampling errors gave the impression that the numbers were off.”
One very plausible reason that they were “off” were the 10 million net uncounted votes, the majority from minority precincts that are 90%+ Democratic. The voters were polled, but their votes were not counted. Clinton may have lost millions of other votes due to switched and stuffed ballots. In order to match the 1992 recorded vote, the Final National Exit Poll required that 119% of living Bush 1988 voters turned out in 1992.
“VRS claimed the Democratic overstatement in the raw exit poll data was due to partisan differences in the willingness of voters to complete the exit poll, not to a poor selection of precincts or differential response rates by age, race, or gender. Republicans simply refused to participate at the same rates as Democrats, resulting in there being fewer Republicans in the raw exit poll results than there should have been. Mitofsky speculated that the disparity was due to different intensities of support for the candidates—Democratic voters were just more excited about voting for Clinton than Republican voters were about voting for Bush and, as a result, were more motivated to communicate this message by filling out the exit poll questionnaire; others thought it was due to Republicans in general having less confidence in the mass media.”
Mitofsky may have “speculated” but there is no evidence that Democrats were more responsive to the exit pollsters. In fact, since 2000 response rates in GOP strongholds were higher than comparable Democratic rates. GOP exit poll and vote shares were positively correlated (.25) to state exit poll response. The average Democratic correlation was -0.26. Bush vote shares increased as response rates increased. In 2004, exit poll precinct data showed that response rates were higher in partisan Bush precincts.
“Despite the source of the partisan bias in the raw results, the exit polls were able to characterize accurately the voting patterns of demographic subgroups and partisan constituencies once they were weighted to match the official returns. The problem was that the data could not be corrected until the official results began coming in. As a result, the exit polls were susceptible to inaccurate vote projections on election night, especially early in the evening right after poll closings. Nonetheless, the cautious analysts at VRS still called all the races correctly in the 1992 election.”
The data could not be corrected until the official votes came in? Or was it that the data could not be rigged until the official votes came in? Of course the cautious analysts called the winner correctly – Clinton won easily – but they did not call the vote shares correctly. Clinton won by a much bigger margin than they said he did.
The 2000 Election Debacle
“Network competition to call winners culminated in the disastrous 2000 presidential election, when these systems of race projections broke down, and the networks wound up retracting their calls for the winner in Florida and presumptively the election, not once, but twice on election night. The trouble began early in the evening, when VNS alerted the networks around 7:50 p.m. that their statistical models predicted Al Gore the winner in Florida and that the networks should consider calling the state for Gore. This prediction took place even though only 4 percent of the actual vote had been counted and numerous precincts in the Florida panhandle, which happened to be in the central time zone, remained open until 8 p.m.”
If the exit polls show a clear winner – as they did in Florida – the fact that just 4% of the votes were recorded is irrelevant. The exit polls were completed by 7:50pm – and panhandle precincts were exit polled throughout the day. Calling the race 10 minutes before the polls closed was of no consequence. Gore won the Florida exit poll (1816 respondents) by a whopping 53.4-43.6%, far beyond the 3% margin of error.
“Less than ten minutes later, the decision desks at all the networks and the AP agreed with VNS and announced Gore the winner in Florida. Over the next hour-and-a-half, VNS discovered that vote-count data from Duval County had been entered incorrectly, making Gore appear as if he had many more votes than he actually did. After fixing this error, the statistical models used by VNS and decision desks at all the networks showed the race could no longer be projected safely for either candidate. By 10:18 p.m., all the networks announced they were moving the state back to the undecided category, prompting Jeff Greenfield of CNN to quip, “Oh waiter, one order of crow.””
Of the 185,000 spoiled ballots in Florida, 113,000 were double and triple punched – and Gore’s name was punched on 75% of them. Almost 30,000 overpunched ballots were in Duval County which has a large black population. Could the spoiled ballots have been the cause of the Duval adjustments?
“At 2:15 a.m., Fox News called Florida and the presidency for Bush. Within five minutes, NBC, CNN, CBS, and ABC followed suit, announcing that Bush would be the forty-third president of the United States. Meanwhile, VNS and the AP chose not to call the race in Florida a second time, wary of the volatility in the data with the contest that close.”
“During the next couple hours, new errors were discovered. VNS had underestimated the number of votes remaining to be counted. Two counties—Volusia and Brevard—had mistakenly entered their vote totals in favor of Bush. Once these mistakes were corrected, the race narrowed considerably, so much so that Bush’s lead was inside the margin of error.”
What about the -16,022 Gore votes in Volusia? The media commentators called it a computer “glitch”. They always do. They never consider that it could have been the result of malicious coding.
“An embarrassment early in the evening had turned to a humiliation by the end, leading NBC News anchor Tom Brokaw to remark, “We don’t just have egg on our face; we have an omelet.”
“Despite the resulting indignation, the exit polls were not responsible for the erroneous second call. In fact, the exit polls were at that point no longer part of the estimation models, having been replaced by actual vote counts—incorrect as they were in some cases—over the course of the evening.”
Replaced by actual vote counts? That is what the perpetrators wanted to do all along. The media never reported that Gore won the unadjusted state exit polls by 50.8-44.5% (5.5 million votes) – way beyond the MoE. Or that he won the unadjusted National Exit Poll 48.5-46.2%, a 2.5 million margin. There were 5.4 million net uncounted votes. The True Vote Model indicates that he had 50.7%.
“However, the partisan skew in the measure of aggregate vote choice was higher than in previous elections. The preliminary data overstated the difference in the George W. Bush-John Kerry vote on election night by 5.5 percentage points, predicting a 51- to 48-percent advantage for Kerry rather than a 50.5- to 48-percent win for Bush.”
Kerry won the unadjusted state exit poll aggregate by 51.0-47.9%. He won the unadjusted National Exit Poll by 51.7-47.0%. The True Vote Model indicates that he had 53.5%.
“This was the highest error in the preliminary results since the 1992 election and double the error found in the previous two presidential elections. The discrepancy between the preliminary exit poll findings and the final election results was even greater in the competitive states. The exit polls predicted a Kerry victory in four states—Ohio, Iowa, New Mexico, and Nevada—in which Bush won, and overstated Kerry’s support by 11 percentage points in Ohio, 9 points in Pennsylvania, and 8 points in Florida.”
“Considering the closeness of the election, the exit polls seemed to suggest that Kerry was capable of winning the 2004 election. Political observers used these differences between the preliminary exit polls and the final results to support allegations of vote rigging and fraud in precincts deploying electronic voting machines, particularly in Ohio, where the state’s twenty-seven electoral votes, enough to change the winner of the Electoral College from Bush to Kerry, was decided by 118,775 ballots.”
The adjusted National Exit Poll indicated that there were 52.6 million returning Bush 2000 voters. But in the 2000 election, Bush had just 50.5 million recorded votes. He needed a 110% turnout of living Bush 2000 voters to match the 2004 recorded vote. Clearly a physical and mathematical impossibility.
“Steven Freeman of the University of Pennsylvania calculated the odds of the exit polls in Ohio, Pennsylvania, and Florida being as far off the final outcome as they were as 662,000 to 1.”
Note: The state exit poll margin of error (MoE) includes a 30% cluster factor.
In Pennsylvania, there were 2107 respondents (2.75%).
Kerry won the poll by 56.6-42.9%, an 800,000 vote margin.
In Ohio, there were 2020 respondents (2.82%).
Kerry won the poll by 54.1-45.7%, a 450,000 vote margin.
In Florida, there were 2862 respondents (2.38%).
Kerry won the poll by 50.8-48.2%, a 200,000 vote margin.
“The National Election Data Archive, a nonpartisan group of mathematicians and statisticians promoting election reform, found that twenty-two of the forty-nine precincts in Ohio polled by Edison/Mitofsky had reported Kerry vote share results that had less than a 5 percent chance of occurring, based on the state’s exit polls.”
“Rep. John Conyers, D-Mich., even used the exit polls as the basis for holding congressional hearings on vote irregularities in Ohio. Edison/Mitofsky disputed these charges in a follow-up report, contending that precincts with electronic voting had virtually the same rates of error as those using punch card systems.”
“They again attributed the bias to within-precinct error—error due to a systematic bias in the selection of voters within a precinct—and not to bias in the selection of precincts themselves. Bush voters were more likely to refuse to participate in the exit polls than Kerry voters. They hypothesized that the result was a function of the disproportionate numbers of interviewers under age thirty-five who administered the exit poll. Young people had more problems securing participation from voters than older respondents, perhaps because they were correctly perceived to have been more likely to have voted for Kerry.”
The same old discredited and debunked Reluctant Bush Responder canard that was refuted by the exit pollsters own data. It showed that exit poll response was highest in partisan Bush precincts – and in strong Republican states.
“Edison/Mitofsky also found that voting patterns within electoral groups were accurate once they were weighted to the official results. They found no evidence that the distribution of presidential vote choices within various demographic groups was biased, despite the vote choice of exit poll respondents overall overstating Democratic support.”
The “overstating” of 56 Kerry respondents for every 50 Bush respondents was not due to differential response; it was due to the fact that Kerry won the election with about 53% of the vote.
“Since 2004, less controversy has surrounded the exit polls. No serious technical problems have surfaced during the last three elections, enabling the media to prepare analyses of the outcome in a timely manner. Leaks of early wave findings have been contained. The preliminary exit polls have continued to overstate support for Democratic candidates; however, the final vote counts have had such large winning margins that the projected outcomes were no different.”
There was less controversy in 2008 only because Obama won by 9.5 million recorded votes. But the exit polls indicated that he won by nearly 23 million; the landslide was denied. The level of fraud was equivalent to 2004. Obama won the aggregate of the unadjusted state exit polls (82,388 respondents) by 58.0-40.5%. He won the unadjusted National Exit Poll (17,836 respondents) by 61-37%. He won the independent True Vote Model with 58.0%, exactly matching the state exit polls. He won the recorded vote by just 52.9-45.6%. How does one explain the massive discrepancy? It was surely not due to differential response.
Selection of Precincts
“National exit pollsters choose precincts by taking stratified probability samples in each of the states before drawing a national subsample from the state samples. This process involves sorting the precincts in each state into different categories or strata to guarantee that particular groups are represented adequately. To begin, precincts in each state are initially grouped into two strata according to their size to ensure the selection of smaller precincts.”
“Within each of these size strata, precincts are categorized by geographic region, usually between three to five regions in each state. For each state geographic region, precincts are ordered by their percentage vote for one of the major political parties in a previous election. Precincts are sampled from these strata with probabilities proportionate to the total votes cast in them in a prior election, so that every precinct has as many chances of being picked by pollsters as it has voters. The samples drawn in each state are then combined, and a national sample of precincts is selected from them using a previous presidential race to determine the relative number of precincts chosen from each state.”
Sampling voters in proportion to the recorded vote in prior elections is a persistent source of bias, since the recorded votes were fraudulent and favored the Republicans. So the sampled exit polled precincts were over-weighted for the GOP.
“Typically, the total number of precincts selected in the national exit poll is between 250 and 300. Ultimately, the number of precincts chosen represents a tradeoff between sampling error and financial constraints. Research by Edison/Mitofsky has shown that the number of precincts selected has not been responsible for the Democratic overstatements that have continually appeared in the exit polls.”
“For example, they found that for the 2004 election the actual distribution of the presidential vote in the precincts used in the exit poll samples did not differ significantly from the actual vote distribution nationwide. In fact, these precincts overstated support for the Republican candidate, George W. Bush, but only by 0.4 points, on average, across the states.”
Mitofsky believed that the exit poll precinct samples were perfect. But he also believed that 56 Democrats responded for every 50 Republicans – even though his own data indicates that response rates were higher in partisan Bush precincts.
“Refusal rates, or for that matter miss rates, are not necessarily problematic, as long as the propensity of different groups to participate does not vary. However, if one group is more or less likely than other groups to complete exit surveys, their responses will be over or under-represented, thereby biasing estimates for the overall electorate. For example, the partisan overstatement repeatedly found in the national exit polls over the past several decades appears to be due to the greater willingness of Democratic voters to complete the exit polls, compared with their Republican counterparts. However, once this discrepancy has been corrected by weighting the exit polls to correspond with the actual vote, there has been no evidence that the vote estimates within groups are biased.”
Greater Democratic willingness to be exit polled is a myth -not a fact. The exit pollsters own data shows otherwise. In 2000, 2004 and 2008, Republican exit poll shares and vote shares were positively correlated (.25) to state exit poll response. Bush vote shares increased as response rates increased, refuting the Reluctant Republican Responder hypothesis.
“National exit pollsters account for early/absentee voting by conducting telephone surveys in states where the rates of early voting are highest. VNS first incorporated early/absentee voting in 1996, surveying voters in California, Oregon, Texas, and Washington. By 2008, NEP was conducting telephone surveys in eighteen states, including Oregon, Washington, and Colorado, where the proportions of early voting were so high that no in-person exit polls were conducted on election day.”
Early voting data in the 2008 election indicates that Oregon, Washington, and Colorado had the lowest red-shifts. Was it just a coincidence that the states with the highest early voting rates were the ones which most closely matched the unadjusted exit polls?
Take the Election Fraud Probability Quiz.
Election Model Forecast; Post-election True Vote Model
2004 (2-party vote shares)
Model: Kerry 51.8%, 337 EV (snapshot)
State exit poll aggregate: 51.7%, 337 EV
Recorded Vote: 48.3%, 255 EV
True Vote Model: 53.6%, 364 EV
Model: Obama 53.1%, 365.3 EV (simulation mean);
Recorded: 52.9%, 365 EV
State exit poll aggregate: 58.0%, 420 EV
True Vote Model: 58.0%, 420 EV
2012 (2-party state exit poll aggregate shares)
Model: Obama 51.6%, 332 EV (Snapshot)
Recorded : 51.6%, 332 EV
True Vote 55.2%, 380 EV