« Kerry Leads by 4 in New FL Poll | Main | Inside Scoop on Rove's View of the Election »

Fasten Your Seat Belts, Turbulent Polls Ahead

By Alan Abramowitz

Fasten your seat belts and make sure that your tray tables and seat backs are in their fully upright and locked positions. The 2004 presidential campaign is coming in for a landing and it's going to be a bumpy ride. You're going to see some strange poll results over the next two weeks and the most important thing to keep in mind is that if a poll's results don't make sense--if they seem to be out of line with most other polls or if they show a big difference between registered voters and likely voters--don't believe them.

We've already pointed out that during the final week of the 2000 campaign, 39 out of 43 national polls showed George Bush leading Al Gore. A look back at the state polls released during the final two weeks of the 2000 presidential campaign shows that while most of the polls were reasonably accurate, there were some that were way off the mark.

The final Zogby tracking poll in California showed Al Gore with a razor thin 1 point lead over George Bush. Maybe that's why Bush ran off to California during the final few days of the campaign. Gore won the state by 12 points.

In Delaware, which was considered a swing state by many analysts in 2000, a late Mason-Dixon Poll had Bush leading Gore by 4 points. Gore won the state by 13.

In Illinois, which many pundits did not consider safe for Al Gore either, a Mason-Dixon Poll had Gore with only a 3 point lead and a Fox News/Opinion Dynamics Poll had Bush with a 3 point lead. Gore carried the state by 12.

In Michigan almost all of the polls showed Al Gore with the lead. But a Market Research Group Poll completed on October 26 had Bush ahead by 6. Gore carried the state by 5. Market Research Group was a Republican polling organization and this year we're seeing a lot more of those Republican polls, a lot of them from an outfit called Strategic Vision. Take their results with a very large grain of salt.

In New Jersey, three late polls, two by Quinnipiac University and one by the Gannett newspapers, all showed Al Gore with a fairly narrow lead--ranging from 5 to 8 points. Gore carried the state by 16. This year we're again seeing some state polls showing that New Jersey is close. We'll see if they're any more accurate this time than they were four years ago.

In Ohio, two late polls, one by the University of Cincinnatti and one by The Columbus Dispatch, gave George Bush a 10 point lead over Al Gore. On Election Day, Bush's margin was only 4 points.

Finally, in Wisconsin, a swing state in 2000 and in 2004, the final Zogby tracking poll and a late poll by Market Shares had Bush ahead of Gore by 6 and 5 points respectively. Al Gore carried the state by less than 1 percentage point.


maybe we should get rid of polling on elections. they seem to be being used to try to exercise undue influence on the system.

Ruy raises a good point, which is that everyone is saying Zogby's the best because of the way he called the last election. In fact there is plenty of evidence to indicate that Zogby polls are highly fallible. I think one of his recent polls had undecides at 25%, which is way over what every other poll is telling us. He also has Bush and Kerry essentially tied among young singles, undecides, you get the idea. I just think the best approach is to go with the average of all the polls, excluding outliers, and then add 1-2 points for the suspected Kerry undercount.

An excellent cautionary message. I am still a little fixated on these recent Zogby numbers, and I certainly accept your explanation for them. But my recollection is that he has pretty consistently shown the race a lot closer than many of the other polls have--for instance, the ABC/Post Poll (and certainly the Newsweek and Time polls right after the GOP convention). So I am puzzled that now, after the debates, when intuitively you would think Kerry had strengthened his position vis a vis, that suddenly Bush seems to have pulled into a fairly good lead. And on top of that, we have ABC/Post now showing the race just about dead even. So I guess my question would be: why has Zogby suddenly seem to have veered off the road?

That was a nice run-down of the wacko poll results from 2000 -- I wasn't aware of those (or I forgot in the horror of what subsequently happened). I am left wondering, however, if there were also poll results that showed Gore with healthy leads in states that he actually lost. Can you give us examples?

The reason I ask is that one conclusion that can be drawn from your interesting post (aside from: don't believe every poll result) is that the 2000 polling results were uniformily skewed against Gore. Was that the case?

I agree that Zogby's state polls are pretty suspect.(as other pollsters).

His national polls for the presidential race were dead on in 2000 and 1996. I think tthats why we all give his polls so much weight. I'm not all that concerned about the fact kerry is down 4 in the zogby poll now, these jumps happen, i fully suspect that in a few days the polls may show kerry up by 3 or 4.

I read the description of the methodology on the Fairleigh Dickinson University poll (the one that now shows Kerry only 2 points ahead in New Jersey). There was no mention of any weighting procedure. If they don't weight, it means that Kerry is almost certainly much stronger than the numbers indicate.

How much did the late DUI revelation factor into the Gore last minute surge and variance from the polls? Rove estimated 1 million votes.

To what extent was Gore's improvement over the last polls due to DUI and to what extent was it turnout or polling erro?

I think la is correct in that the best idea at this juncture is to do the unthinkable- ignore the polls b/c they aren't going to help one bit. Over on Daily Kos you have people fretting over every little shift inthe polls- ie, Kerry was up by 2 last week, and down by 1 this week, and vice versa, and they look at the negative one as gospel. I think as I have said in other places on here this all feels like a Rorshach's test to me- people are seeing what they want to see. Right now I spy a tie of around 47 to 48 percent so what does that mean? The same as always and I will say it like a broken record convince everyone, their mama and grandmama to volunteer, to phonebank and drive up the GOTV.

I think all the polls this year are way off due to the under polling of the 18-30 year olds.

Considering the newly registered voters must be largely from this demographic and their reliance of cellular phones for (nearly) all their telecommunications, these voters will hold a huge surprise impact on this election.

I have two daughters, one 29 and one 25 and I can tell you there is a tremendous difference in their participatory interest.

The older one, going through high school while Reagan and Bush were in office, is pretty much ambivalent about politics.

The younger one, going through high school while Clinton was president, is very much involved and in tune with political activism.

The older one will not vote whereas the younger one is definitely voting for Kerry.

The youth vote will go strong and early for Kerry and this element is what all the major polls will miss.

This effect will become known as the "Cell-Phone Effect" but is really the "Clinton Youth Effect".

One question is: these are all examples of the polls UNDERestimating the Democrats' strength. What about OVERestimations too? Like in 2002 and surely SOME states in 2000, n'est-ce pas?

And then there's the reminder also to be mindful of LV/RV distinctions. I see that the new Newsweek poll is out. By LV's, Kerry is down 6%. By RV, just 2%, and just 1% in head to head RV.

I'd be willing to bet it's the 6% that hits the media...

Ruy does well to point out turbulence. Stay calm and focused all.

What are the chances of Bush occupying the Gore role from 2000? All the poles are close, and many not to be trusted, so how do we know they won't break for the President instead of for Kerry? Is there something from the internals that would indicate this?


Typically undecideds break for the challenger. I've seen polls suggesting that the undecideds, by a large majority, favor a change and have views closer to Kerry's.

Anything's possible, of course, but I'm optimistic about them at this point. The Rove quote from Abramowitz, above, fits with this point of view.

Late rapid drifts in the polls can be substantial.

This year, there is an additional complication relative to 2000 that may tend to bias the polls toward Bush relative to the election. In 2000, Bush, Gore and Nader were on the ballot in every (for Nader, almost every) state, Buchanan's campaign did not catch fire in that conservatives largely were happy with Bush, and Browne's campaign was ineffective for reasons I outlined in my book Funding Liberty (Third Millennium Press http://3mpub.com/~phillies if anyone is interested.)

This time, Nader is not even on the ballot in many states, so a poll that includes CA or TX voters and asks them if they will vote for Nader is just plain wrong. (and those wrong votes came in fair part from Kerry). This time, the Libertarian campaign (Michael Badnarik) is on and off polling around even with Kerry, is on the ballot in every state except NH and OK, and has a largely unified party and a campaign staff that is campaigning actively. (If that last bit sounds odd as something to mention, see my book.) The Peroutka campaign, with positions many Democrats will find a bit odd, is also firing on at least some cylinders. However, Peroutka and Nader are almost never polled except by Rasmussen. The Nader Peroutka Badnarik flaws may cost Kerry in the polls a percent of the vote, and may give Bush one percent that he will not have in reality.

However, if you are going to drop polls as outliers, you should keep a list of the outlier polls, lest you analyse yourself into rejecting all the good polls and keeping only the bad ones.

Perhaps the most important number, both nationally and in many states that Bush seems to be carrying, is that the first digit of his numbers is a "4". That's a losing number. Democrats should not lose heart, should heed the wise advice of the article on which we are commenting, and could consider reminding their conservative and small government friends about Mr Badnarik.

The undecideds cannot possibly break for the President. If they were going to support Bush they'd be in his column before now. He's the known quantity. He's been consistent for the last 4 years and also during the debates. Anyone who is really undecided at this point has already resisted being swayed by Bush for 4 years and the entire campaign. That's why there are so few real undecideds. What Kerry had to do and did during the debates is show he's capable of governing.

George Phillies wrote:

Perhaps the most important number, both nationally and in many states that Bush seems to be carrying, is that the first digit of his numbers is a "4".

Ah, through all the buzz, a statement of clarity.

Bush is losing and has been losing a long time. If polls are viewed generally for the past several months, there are only limited moments in which Bush has pierced the "50", and that only if you actually believe those polls.

Sabato said two months ago it would take a miracle for Bush to win, and he was right.

Bush has lost the election. I'm sure their major game plan is still to steal it. Florida and Ohio are where they plan to try to steal it.