Polls' Accuracy

People shouldn’t get too excited/discouraged based on the polls – at least insomuch as they don’t show a huge move in one direction or the other. The polls are just too reliant on estimates, and too prone to inaccuracies.

I found this interesting information this morning – it looks at the final poll results released prior to the 2000 election by the various pollsters. Overall, pretty darn bad job in terms of accuracy - and that’s just in terms of polling for the national popular vote:


The final result of the 2000 election: 48 percent for Gore, 48 percent for Bush, 4 percent other. Note that the DUI appeared to have a very late-breaking effect on Bush?s level of support.

Zogby underestimated Bush by 2, had Gore accurate.
Washington Post had Bush accurate, Gore underestimated by 3.
Pew underestimated Bush by 2; underestimated Gore by 5.
Newsweek underestimated Bush by 3; underestimated Gore by 5.
NBC/Wall Street Journal underestimated Bush by 1; underestimated Gore by 4.
Marist overestimated Bush by 1, underestimated Gore by 4.
Harris underestimated both Bush and Gore by one point.
Gallup had Bush accurate, Gore underestimated by 2.
Fox News underestimated both Bush and Gore by 5 percentage points each.
CBS News underestimated Bush by 4, Gore by 3.
Battleground overestimated Bush by 2, underestimated Gore by 3.

The raw data:

A rundown of every final national poll conducted since 10/31. ^ indicates totals included the allocation of undecideds. Order is Bush, Gore, Nader.

1 Battleground (11/5-6; 1,000 LVs; +/- 3%)^ 50 45 4
2 Bullseye (10/31-11/2, 11/5; +/- 3.1%) 45 42 4
1 CBS News (11/4-6; 806 LVs; +/- 3%) 44 45 4
1 Democracy Corps (D) (10/30-31; 881 DVs; +/- 3.1%) 45 45 5
2 Fox News (11/1-2; 1,000 LVs; +/- 3%) 43 43 3
1 Gallup (11/5-6; 2,350 LVs; +/- 2%)^ 48 46 4
1 Harris (11/3-5; 1,348 LVs; +/- 3%) 47 47 5
0 ICR (11/1-5; 1,141 LVs; +/- 3%) 46 44 7
2 Marist Inst. (11/1-2; 623 LVs; +/- 4.5%) 49 44 2
1 NBC/WSJ (11/3-5; 1,026 LVs; +/- 3%) 47 44 3
2 Newsweek (10/31-11/2; 808 LVs; +/- 4%) 45 43 5
0 Pew (11/1-4; 1,307 LVs; +/- 3%) 46 43 3
1 TIPP (11/4-6; 1,292 LVs; +/- 2.8%) 48 46 4
2 Wash. Post (11/4-6; 1,801 LVs; +/- 2%) 48 45 3
1 Zogby Int’l (11/5-6; 1,200 LVs; +/- 3%)^ 46 48 5

UPDATE: Gerry Daly observes
that one pollster ?called? over 29% of the 2002 Senate and/or Gubenatorial races for the wrong candidate, despite polling more races than all but one other company.

The average for everybody else was getting about 13% of the races wrong, by comparison?

This pollster? Zogby.

Another pollster polled more 2002 races than any others, but only got one wrong. This pollster? Mason-Dixon.

This is why I stick with the futures markets, which have been proven more accurate than the polls:




There is no question that the democrats leaking that Bush DUI two days before the election had a dramatic effect on Bush’s numbers.

I have read that up to 4 million (give or take) Evangelical Christians did not vote for Bush, and they normally would have.

I wonder what Terry has in mind as a last minute dirty trick this year?

On the other hand I wonder what Rove has as a last minute dirty trick? Afterall, turn about is fair play, and all is fair in love, war and politics, so it seems.

Zeb –

It amuses me that the Democrats are always talking about Karl Rove, dirty tricks and an “October Surprise,” when the best example from the last election cycle of that sort of tactic was the Bush/DUI float.

On the subject of dirty tricks, this one seems, from the info at hand, to be particularly egregious: fake voter-registration fraud, blame other side…


Eh, forgot to put in the excerpt before I posted that – here ya go:

UPDATE: The Kerry Spot despises efforts to mess with the ballot box, whoever is doing it. A former employee of Voters Outreach of America claims that this private voter-registration company is throwing out non-Republican voter registration forms.
This former employee has gone to the FBI. The feds ought to investigate, and if this organization did what its former employees charged, then they should get the book thrown at them.

UPDATE: Hmm. The KLAS-TV story linked above says the group, “Voters Outreach of America, AKA America Votes…” America Votes is linked to:

ACORN AFL-CIO AFSCME America Coming Together (ACT) American Federation of Teachers Association of Trial Lawyers of America Brady Campaign to Prevent Gun Violence United With the Million Mom March Clean Water Action Defenders of Wildlife Action Fund Democracy for America EMILY's List Environment2004 The Human Rights Campaign League of Conservation Voters The Media Fund MoveOn.org Voter Fund Moving America Forward Music for America NAACP National Voter Fund NARAL Pro-Choice America National Education Association National Jewish Democratic Council National Treasury Employees Union Partnership for America's Families People for the American Way Planned Parenthood Action Fund SEIU Sierra Club USAction Voices for Working Families Young Voter Alliance 21st Century Democrats

ACORN, the first group on that list, has been tied to voter fraud investigations in Colorado
, Ohio

and Pennsylvania

Just what the heck is going on here?

Here’s a fraud story from Florida – man, if this is close, it’s going to get seriously, seriously ugly… This sort of thing is bad – this story happens to be about tossing Republican registrations, but I am equally against tossing Dem registrations. I want to win, but I want to win fairly.


Rumors of vote fraud rampant


TALLAHASSEE – The U.S. Justice Department and state police are investigating widespread allegations of criminal vote fraud in Florida, days before the state closes registration for the presidential election.

Allegations range from phony voter registrations to forged party-affiliation change cards and absentee ballots.

“Several law enforcement investigations are under way,” wrote Elections Division Director Dawn Roberts in a Friday memo to election supervisors.

Agency spokeswoman Alia Faraj said the Justice Department has agreed to review the allegations of vote fraud in Florida. She could not further discuss the active state and federal investigations.

A field director for one of the many national partisan organizations trying to drum up votes in Florida admits to routine efforts to rig the outcome. They include submitting thousands of invalid voter registration cards, as well as failing to turn in boxes of cards filled out to register Republicans.

“There was a lot of fraud committed,” said Mac Stuart, former Miami-Dade field director for ACORN. Among his allegations – that ACORN “quality control” workers routinely kicked back Republican voter registrations while paying for Democratic ones. “They said they had enough,” he said.

ACORN is spearheading both a minimum wage ballot initiative and a voter registration drive. Its top two Florida directors failed to return telephone calls Friday.

Stuart is listed as a plaintiff in a notice of intent to sue ACORN and others in a discrimination class-action lawsuit. “The voter registration project has been operating illegally since it started,” the intent-to-sue filing asserts.

In Leon County, elections supervisor Ion Sancho said he found nearly 1,000 apparently fraudulent party-change forms. All were to re-register African American Democrats at Florida A&M University as Republican.

The suspect forms came to his office last week from the Florida Division of Elections office in Tallahassee, where employees said organizations conducting statewide voter registration drives are turning in thousands of records a day to beat Monday’s deadline.

“They didn’t have a clue where those forms came from,” Sancho said. He said he is concerned the state is forwarding similarly suspect forms to other counties.

Citing such “reports of irregularities” around the state, the Division of Elections on Friday afternoon sent a memorandum to county election supervisors, asking them to scrutinize “all voter registration agencies” in their counties and report suspect activity.

Active investigations include one in South Florida involving ACORN.

Stuart said he has been interviewed twice by an agent for the Florida Department of Law Enforcement. That FDLE agent declined Friday to confirm his investigation. Florida law makes it a third-degree felony to interfere with someone’s right to register as well as to pay a per-card solicitation fee for gathering registrations. Stuart says ACORN did both.

Stuart said ACORN officials at state headquarters in Tampa were aware of what was going on, and discouraged him from talking about it. He said he was ultimately fired as “a loose cannon.”

While Republican registrations were ignored, Stuart said those of convicted felons were eagerly sought, even though by law they are ineligible until they are granted clemency by the state. Stuart set up registration tables outside the Miami police department and Dade County jail.

“We targeted them because ACORN had a goal: 103,000 new registrations from Dade County,” Stuart said.

Brian Kettenring, ACORN’s state director in Florida, and Frank Houston, who spearheaded the affiliated Project Vote registration drive, did not return calls Friday.

The questions surrounding ACORN’s voter registration efforts were preceded by problems with initiative petition signatures the group submitted earlier in the year.

A Hillsborough County election official in July found some 800 apparently fraudulent signatures among the minimum wage petitions turned in by ACORN. Elections Supervisor Buddy Johnson said he merely notified the organization, and ACORN agreed to police its own workers.

Such criminal allegations are giving initiative opponents a foothold for proposed amendments beyond those sponsored by ACORN.

Opponents of the slot-machine initiative contacted 5,278 individuals in Broward County who allegedly signed the gambling petition. Of those, 3,587 said the signature on the petition was not theirs. Another 33, the opponents allege, were dead.

In a lawsuit filed last week in Leon County circuit court, slot-petition opponents assert Floridians for a Level Playing Field paid Arno Political Consulting as much as $6.50 per petition signature.

The business practice, they contend, “invited the submission of fabricated signatures.”

The opponents include The Humane Society, Grey2K USA and Floridians Against Expanded Gambling.

The initiative to repeal high-speed rail is also being challenged in court for relying on fraudulent petitions.

With the federal investigation into registration fraud, such questions about faked petition signatures grow to fears of efforts to hijack November’s vote.

“We’re extremely concerned, with so many groups that came into Florida,” she said. “While it’s a great thing (to register voters) we encourage voters to get to know the group, and follow up with a phone call to the (election) supervisor” to make sure they are, indeed, registered.

Here’s another comparison of final polls against results, this time from 2002 Senate races (courtesy of Jim Geraghty at The Kerry Spot):

Let’s compare the final polls of 2002 to the final results of 2002 for the major Senate races. These were the states that had the nation’s eyes on it, with the polls conducted in late October or even up to the day before Election Day. Pollsters should have been at their best, with every advantage.


In Minnesota, Zogby had Walter Mondale over Coleman by 6 points. The Minneapolis Star had Coleman down 5 in its last poll.

Final: Coleman 50, Mondale 47. Coleman by 3.


In Colorado, Zogby had Strickland over Allard by 5 points, but at the end of October he had Strickland by 9. The Rocky Mountain News had Strickland by 4. The Denver Post had Strickland by 1.

Final Allard 51, Strickland 46. Allard by 5.

New Hampshire

In New Hampshire, the Concord Monitor had Democrat Shaheen over Republican Sununu by one, FPC/WNDS-TV had Shaheen by 5, University of New Hampshire had Shaheen by 4, then had Sununu by 1.

Final: Sununu 51, Shaheen 47. Sununu by 4.


In the Texas Senate race (not that Texas is a swing state) Zogby’s final poll had Republican Cornyn over Democrat Kirk by 4 points, and a day earlier had Cornyn up one. The Dallas Morning News had Cornyn up 9, Survey USA had Cornyn up 8.

Final result: Cornyn 55, Kirk 43. Cornyn by 12.


In the Georgia Senate race the final four polls were Zogby showing Democrat Cleland up by 2, Zogby showing a tie, Atlanta Journal-Constitution showing Cleland up 3, and Mason-Dixon showing Cleland up 6.

Final result: Not even close. Republican Chambliss 53, Democrat Cleland 46, Chambliss by 7.

North Carolina

In the North Carolina Senate race the final five polls were Zogby showing Republican Dole up by 10; Zogby showing Dole up 6, Carolina Poll showing Dole up 7, Survey USA showing Dole up 4, and Mason-Dixon showing Dole up 6.

Final result: Dole 54, Bowles 45. Dole by 9. Only Zogby’s last-second “correction” poll was close.


In Iowa, the Des Moines Register had Democrat Harkin with a moderate 9-point lead, but SurveyUSA had Harkin beating Republican Ganske by an overwhelming 22 points.

Final result: Harkin 54, Ganske 44. Harkin by 10.

There are a few exceptions to this rule. In Missouri, the late polls had Talent up by 8, and up by 4. He won 50 percent to 49 percent.

Note that many of these states are swing states this year. Also note that these pollsters were wrong by 8, 9, 10 points, and they were underestimating support for the Republican candidates by 8, 9, 10 points.