« The "How Can Gallup........" Game (Florida Edition) | Main | Latest ARG Polls: Kerry Leading By 3% in Florida, 2% In Ohio, 3% in Pennsylvania. »

Gallup Vs. Everyone Else

By Alan Abramowitz

It's not just Gallup's recent national polls that appear to be out of line with the results of most other recent polls. Some of their state polls have also produced rather bizarre results. Two notable examples are Gallup's recent polls in the key battleground states of Florida and Wisconsin.

In Florida, Gallup just released a poll showing George Bush leading John Kerry by 9 percentage points among registered voters. Wow! No other poll in the past month has given Bush a lead of more than 3 percentage points in Florida. In fact, of the 12 other Florida polls released in the past two weeks, 6 have John Kerry leading while only 3 have Bush leading, and 3 have the race tied. On average, in these 12 polls, Kerry held a lead of 0.6 percentage points. Quite a difference.

In Wisconsin, Gallup just released a poll showing Bush leading Kerry by 8 percentage points among registered voters. No other poll this month has given Bush a lead of more than 3 percentage points in Wisconsin. Of the 5 other Wisconsin polls released in the past two weeks, 3 have Kerry leading, 1 has Bush leading, and 2 have the race tied. On average, in these 5 polls, Kerry held a lead of 1.5 percentage points.

If this keeps up, when this election is over, the folks over at Gallup are either going to look like idiots or geniuses. I'll leave to you to guess which one is more likely.


Their huge pro-GOP sample bias in Florida certainly speaks for itself. However, what about the latest Gallup poll for Ohio, which was released a day after their latest Wisconsin poll:


It showed Kerry up 48-47 among LVs, and a startling 50-44 Kerry lead among RVs! Huh? Most Ohio polls are looking pretty good for Kerry, but 50-44? Isn't that a bit much?

Read The Vulcans. There is chapter about Rumsfeld's rise during the Nixon Administration. One of his duties during Dick's re-election was to game the Gallup brothers. While Rummie wouldn't get the Gallup poll to lie, he would get advance warning of the poll's findings.

Seems this year, he is getting the Gallup poll to lie.

Actually they will look like geniuses or SHILLS. I don't think there's a lot of doubt about which, but we'll know in a week.

Gallup is really walking the plank on this election. But a lot of the blame rests on CNN and USATODAY for spreading the baloney poll results around and paying Gallup a lot of $. An important goal is to try to get CNN and USATODAY to dissociate themselves from Gallup. It's impossible for this election, but possible for the future.

Ok, it's only a week before the elction and my paranoia is running rampant. Could it be that the Gallup poll is providing cover for rampant electronic voting fraud?

Happy Halloween!


CNN & USA Today must fire Gallup

If this election proves that the Gallup methodology is flawed and their flaws show a tilt to one side then we should compell CNN & USA Today to fire Gallup. Recently CNN interviewed Gallup and asked him about the poll discrepencies...his answer was that their LV model was better. When asked about the 'bandwagon effect' (do polls effect how people view the candidates and do they become a Self-fulfilling Prophecy Frank actually said "would that be a bad thing?..people should factor polls into their decisions".

Therefore if they turn out to be 'idiots' we should immediately do to CNN & USA Today what we did to Sinclair and force them to drop Gallup

Isn't the best evidence that this is total horse----- the fact that Bush is campaigning frenetically in these states. I would have thought that 9% would mean going somewhere really close.

The Bushites are scared. Bush visited WI again today. Even the Rep that was on NPR admitted that WI appears to be trending towards Kerry Edwards.

Question: Has any pollster come up with a methology for correlating the intensity of support for a candidate or party with final results?

In 1994 polls showed a fairly close Congressional election, but there was great intensity among Republicans (remember the "revenge of the white men?") and they won Congress.

In 1986, however, there was greater intensity among Democrats, something not found in polls, and you had some huge upsets, like Wyche Fowler in Georgia.

My own guess on intensity shows it off the charts on the Democratic side, slightly elevated on the Republican. This points toward a huge Democratic swing, absolute control of Congress and a landslide for Kerry.

But I'm not a statistician, and I don't know how you could measure it with a poll.

Any thoughts?

Okay, call me paranoid, but maybe Gallup is deliberately"gaming its polls to give cover to the Republicans who appear to be doing their dead-level best to steal the election in Florida and in Ohio. If Bush "miraculously" pulls ahead by four or five points on election day, the media will simply point to these convenient Gall-up polls, announce Bush would have won anyways---look at those Gallup polls--tell Democrats that they should just Get Over It.

I'm predicting a Kerry victory---I think just too many Democrats will show up in the polls for the GOP to scare off or intimidate or spin away. And I also predict that Gall-up won't be short of clients. CNN will continue to buy their services. Or (worst-case scenario) Gall-up will lose a few media contracts, but be more than compensated by all the new right-wing think tank business.

Bad Turnout Assumptions = Bad Polling Assumptions

Gallup’s LV results, and their use as headline facts by USA Today and CNN, may be 21st century equivalents of 1936’s Literary Digest poll, which pronounced Alf Landon our next president. If Gallup sticks with its LV screening method, a bad call on election day and de-legitimization of this method seem almost inevitable. Certainly the partisan and racial biases in the method that Ruy has discussed loom large as problems for Gallup’s credibility. But so do Gallup’s assumptions about turnout, which have not been given the scrutiny they deserve. For instance, as Ruy wrote (October 19):

“Gallup asks each [RV] respondent seven LV screening questions, and gives each person an LV score of 0 to 7. [Assuming a turnout of 55 percent], the top 55% are classified as likely voters….All other voters are discarded from the sample.….the demographics of Gallup's LV sample are not adjusted in any way.”

Calibrating turnout in this way, Gallup has mismatched samples and populations. Its 55% turnout assumption is a reasonable, high turnout by recent American standards if the relevant population is the voting age population (VAP). But Gallup’s initial sample is restricted to RVs, thus excluding ineligibles (non-citizens and felons) as well as eligible non-registered citizens in the VAP. According to the Census Bureau’s 2000 Election Report, these groups, collectively, represented 36% of VAP (8% and 28% respectively) in the last presidential election. Assuming similar proportions for 2004, and adding them to the 45% of initially sampled RV’s that Gallup cut from its LV sample (i.e., .45 x .64 or 29% of VAP), we see that Gallup’s LV sample excludes approximately 65% of VAP, thus representing a VAP turnout of about 35%, not 55%. If Gallup were to appropriately match samples and population, its LV sample should include between 82-90% of its initial RV sample, that is, the range of RVs that actually voted from 1972-2000, rather than 55%. Since Democrats meet Gallup’s likely voter criteria less frequently than Republicans, especially because they vote less often, it’s no wonder Gallup’s overly restrictive LV screen over-samples Republicans.

Note: All data in the last paragraph was drawn or calculated from U.S. Census Bureau, “Voting and Registration in the Election of November 2000.” P20-542, issued February, 2002.

Excerpt from James Mann, Rise of the Vulcans:

"We have decided that we'll try Rumsfeld working with Gallup. He went to school with George [Gallup] Jr. at Princeton," Colson told the president in July 1971. Nixon and Colson were eager to try to influence the results of major pollsters, notably Gallup and Harris, perhaps getting them to phrase their questions or to present their results in ways that were helpful to Nixon. "I mean, if the figures aren't up there, we don't want them to lie about it," Nixon explained to Colson at one point. "They can trim them a little one way or another." [Note 40: Nixon phone call to Colson, July 23, 1971, conversation 6-197, Nixon tape collection, National Archives.] ...

RUMSFELD: Say, I want to just report, sir, about my conversation with George Gallup [Jr.].

NIXON: Oh yeah, you went to school with him, didn't you?

RUMSFELD: I did. And I kind of want to be awful careful about telling people around the building that I'm talking to him. Because all he's got in his business is his integrity.

Rumsfeld then informed Nixon an upcoming Gallup Poll would show that the president's popularity had recently gone up. [Note 41: Nixon conversation with Rumsfeld, October 19, 1971, conversation 11-135, Nixon tape collection, National Archives.]


The intensity on both sides is intense. I doubt that there will be a Democratic majority in either the House or the Senate.

But the intensity on the side of the Democrats is a good thing. The Republicans have had good turnout for years, but not the Democrats. I'm hoping that there will be some surprises on Nov 2nd.

Dana, you speak of the love that I dare not speak its name - the SWEEP.

I hope it is coming, and I agree with your logic.

I've been telling anyone who will listen we are going to win by 5 million votes.

I hope we have enough to take back the House. We have to stop those crazy rightwingers by rooting out their nest.

Gallup Pollsters may be taking into account Gov. Jeb Bush’s wholesale attempt to purge the Florida voter roles, once again, of black voters. A Miami Herald article, “Bush urged to ‘pull the plug’ on voter purge” (October 16, 2004) will get readers up to speed on this outrage. People for the American Way Foundation (see web site at: http://www.pfaw.org/pfaw/general/default.aspx?oid=17265 ) has posted a pdf of the memo alluded to in the article. They also have a link to the letter their president, Ralph G. Neas, wrote to U.S. Attorney General John Ashcroft beseeching him to launch an immediate investigation into the “scandal”. I fear that the cynics and the paranoids among us are on the right track (sad as that may be). The good news is that there are armies of volunteers down here prepared to do everything within their power to protect the vote. . . and to help the Gallup folk join the ranks of the unemployed on November 3.

While Gallup is finding Bush comfortably ahead of Kerry in Wisconsin, Sen Russ Feingold (D-WI) is ahead by more than 20 over his GOP challenger. I know we're famous for ticket-splitting, but it's not very likely that we'd go overwhelmingly for a liberal like Feingold AND go for Bush. Gallup is no longer worthy of reading or reporting.