« National Tracking Poll Roundup | Main | Kerry's Lead Among Independents Makes Bush's Situation Extremely Dire »

About the Final Pew Poll

By Alan Abramowitz

The Pew Research Center has just released its final numbers on the 2004 presidential race and they show George Bush with a 3 point lead among "likely voters" but John Kerry with a 1 point lead among registered voters. Not to worry. There are good reasons for skepticism about those likely voter results.

First of all, here's Andrew Kohut's track record at Pew and, previously, at Gallup. In 2000, the final Pew Poll had George Bush leading by 3 points. We all know what actually happened. In 1996, Pew had Bill Clinton ahead of Bob Dole by 14 points--the actual margin was 9. In 1988, Kohut, then at Gallup, had George H.W. Bush ahead of Michael Dukakis by 12 points--the actual margin was 8. And in 1980, Gallup had Ronald Reagan ahead of Jimmy Carter by only 3 points in their final poll. Reagan won by 10. Notice that in all three of the elections involving incumbents, the final poll overestimated the incumbent's support.

In addition, their equal allocation of undecided voters between the incumbent and challenger is contrary to the normal tendency of undecided voters to break in favor of the challenger by a wide margin.

And there's more. The current Pew sample of likely voters consists of 37.3 percent Republicans and 33.6 percent Democrats. That's right--a 3.7 percent Republican identification advantage. Of course we know that every national exit poll since 1988 has had a Democratic identification advantage of between 3 and 5 points. It is highly unlikely that in a very high turnout election in 2004, Republicans will outnumber Democrats.

Finally, here are a couple of numbers from the Pew Poll that are revealing. John Kerry is leading among independents by 48 percent to 44 percent and Kerry is also leading among those who have already voted by 48 percent to 47 percent.


Among thost who have already voted, Pew has Kerry ahead by just 1 pt. But beneath the surface this number looks a lot more encouraging for Kerry. Most of the states that have early polling are red states and a few are battleground states. Assuming Bush is doing well in red states, Kerry must be doing really well in the battleground states.

This will be my final post before the election. Thank you Ruy and Alan and the rest for all the great information and insight into this crazy polling business.

Regarding the Pew poll, in the final analysis, I'm forced by hard reality to become a skeptic of weighting polls by party id (sorry Alan, I know you feel strongly on this one). MysteryPollster has an excellent and convincing refutal of the contention that polls should be weighted by party ID. I urge you all to read it.

However, there are real factors working in Kerry's favor that aren't fully reflecting in the polls; party-ID misweighting is simply not one of them (my humble opinion).

So what are the real polling unknowns in these final days?

The first is voter turnout. It would seem to favor the Democrats that turnout is expected to match or exceed 1992 (based on days of trolling the blogs and news cites, I predict it will exceed 1992). (Hence the sudden spike in Republican voter supression tactics in key battleground states - these are very real, very ugly, and very disgusting). The pollsters admit their models can't effectively factor in the results of heavy turnout.

The second is possible underrepresentation of Kerry supporters in the poll samples and LV models. It's probably a very small cumulative effect (1-2%) but in an election this close it could have an impact (well, obviously it will have an impact). I hate to be a wet blanket, but I tend to agree with the pros that the "unreachable cell phone" voter is given greatly more significance on this site than he/she deserves, at least for this election. The wingnut blogs entertain a fantasy that there are also millions of invisible Bush voters out there, but there's really no good evidence to back this up. Invisible voters tend to be minority, young, mobile, phone-screeners, busy, politically unmotivated (in most election cycles), etc. They are probably more Democrat than Republican. Probably.

The last factor that may be skewing the national polls is Red states that are top-heavy with Bush supporters. Bush's entire campaign has been aimed at energizing his base, which is very energized indeed and it concentrated in the (election 2000) Red states. When Fox stopped oversampling from Red states in their tracking polls, Bush's support dropped seven points in four days, to +2 Kerry today. Kerry's support in the battleground states is 2% or more higher than it is nationally.

When these three factors are combined, Kerry has a decent chance of winning the election even when the average of the final polls show Bush ahead by 2 or 3 points, which seems to be how it's shaping up.

I don't believe the predictions of a Kerry blowout (or a Bush blowout, either). It's going to be very, very close, folks.

It's been great blogging with all of you. I'll now recede and await the outcome on Wednesday morning. God bless you all and God save the Republic (this coming from an atheist - that's how nervous I am).

Very interesting. My only observation is that there are now more self identifingrepublicans then in the examples cited(especially in the midwest swing states). 9/11 had an impact that will be measured on tuesday.

My worry is that the early voters are (likely to be, and this is just a guess) disproportionally students and the elderly, in which case wouldn't we expect a bigger Kerry lead?


It's not going to be close, coldeye. It's going to be Kerry in a big win, and the undervaluing of Dems in the polls will be identified as one of their compelling failures. Their inherent design defect will be another.

Kerry by 4-5 million, and by at least 40 points in the EC.