By Alan Abramowitz
In a previous posting, I pointed out that there was no consistency in the day-to-day movements of the (then) four national tracking polls, suggesting that these shifts were basically random and that the underlying division of the electorate between Bush and Kerry was not changing. Since then, political junkies like me have discovered yet another set of tracking polls to obsess about–the Zogby 10 state tracking poll.
Now I respect John Zogby, even though he is considered something of a pariah in the polling community for, among other things, his use of a rigid party identification weighting scheme in his national polls. But four years ago, Zogby also undertook a series of state tracking polls in the final days of the campaign and the results were decidedly mixed. Some of Zogby’s final state polls were very accurate but some were wildly inaccurate. The most notable example of the latter was Zogby’s final poll in California which showed Al Gore leading George Bush by a meager 1 percentage point. On Election Day, Gore swamped Bush in the Golden State by 12 percentage points.
So based on his track record, there are reasons to be skeptical about the results of Zogby’s 2004 state tracking polls. For example, Zogby’s current tracking poll shows George Bush leading John Kerry by 9 points in New Mexico. This is simply not plausible. Al Gore narrowly carried New Mexico in 2000 and since then the Hispanic proportion of the New Mexico electorate has increased. For some reason, perhaps a lack of Spanish-speaking interviewers, the Zogby tracking poll is clearly off the mark in New Mexico.
But the main point that I want to make here is not that some of Zogby’s state polls are probably inaccurate, but that the day-to-day movements of these polls across the ten states are totally inconsistent. There have been rather dramatic shifts in the relative standing of the candidates in some of these states, but these shifts show no consistent pattern. For example, since Zogby released his first tracking poll results on October 24, George Bush has gained 9 points of margin in Colorado and Michigan, but John Kerry has gained 11 points of margin in Wisconsin and 5 points in Florida. In just the past two nights, Bush has gained 8 points in Ohio and 6 points in Colorado. Pretty scary, right? But in the same two nights, John Kerry has gained 5 points in Wisconsin, 4 points in Minnesota, and 3 points in Florida and Michigan.
The average correlation in the day-to-day movements between each state and the other nine states ranges from -.16 for Michigan to .02 for Pennsylvania. The average correlation across all 10 states is -.08.
Since October 24 the mean shift in the margin between Kerry and Bush across the ten states has been exactly zero. In the past 2 days, the mean shift has been 0.2 points toward Kerry.
Conclusion–there is nothing going on here. The day-to-day shifts in the Zogby state tracking polls, like the day-to-day shifts in the national tracking polls are basically random. So stop obsessing. Do something useful. Go carve a pumpkin and make sure that everyone you know who hasn’t already voted gets to the polls on Tuesday.

Yesterday, I mentioned two sites that do EV calculations based on polls–NowChannel.com and Electoral Vote Predictor 2004. And, of course, there are many, many others.
But, in the end, the estimates they provide cannot, by their very nature (they are driven by the latest polls and only the latest polls), incorporate the wide range of historical and campaign data that are relevant to thinking through a state’s probable electoral outcome this year. For that, we have to fall back on that old stand-by: the educated judgement of an actual person sifting through all available evidence.
Here’s a good example of what I have in mind. “mattb25” over the Daily Kos offers an assessment that, in his description:

1. Assigned Bush and Kerry their safe states
2. Briefly examine[d] five states that are pretty strongly leaning Bush but could conceivably go Kerry: Missouri, Nevada, Colorado, West Virginia & Arkansas
3. [Took] a closer look at the nine states that, in my view, are 99% certain to determine the election: Ohio, Florida, Pennsylvania, Wisconsin, Iowa, Minnesota, Michigan, New Mexico, & New Hampshire. Please note that I DO NOT consider these states toss-ups, like much of the ignorant SCLM does–but they are the states that the campaigns are playing for right now.

I strongly recommend you read his piece in full. He does an excellent job of combining a critical analysis of available polling data with reasonable judgements about the state of the campaign and underlying political trends and history in a given state. The result is a solid assessment that takes you a lot farther than any purely-poll-driven EV analysis possibly can.

We’ve discussed how uncorrelated the movements of the tracking polls have tended to be–they rarely seemed to show coherent movement in one direction. But it’s interesting to note that, even with the addition of a new tracking poll (Fox), the movement today was generally all in one direction: Kerry’s. (You can find all the relevant data and links at NowChannel.com.)
Here are today’s results:
Zogby: 47-46 Kerry from a 46-46 tie yesterday
Rasmussen: 47.9-47.1 Bush, from 48.7-46.7 Bush yesterday
WaPo LV: 49-48 Bush, from 50-47 Bush yesterday
WaPo RV: 48-47 Kerry, from 48-46 Bush yesterday
Fox 3way LV: 47-45 Bush, from 50-45 Bush yesterday
Fox 3way RV: 46-46 tie, from 48-46 Bush yesterday
The once exception to that pro-Kerry movement: the TIPP poll which (3way) was 46-44 Bush today, compared to a 46-46 tie yesterday.
Note: Today’s WaPo RV result makes 5 out of 6 days that Kerry has been ahead in their RV sample.
Note: At this stage of the election in 2000, Bush was ahead of Gore by 4 points in the Zogby tracking poll, by 4 points in the ABC/WP tracking poll and by 7 points in the TIPP tracking poll.

By Alan Abramowitz
RealClearPolitics.com is a popular website for tracking state and national polls, but it also has a real clear conservative slant–check out the sponsors and the bulk of the op-ed articles posted on the site. RealClearPolitics also includes some dubious polls in calculating their national and state averages–notably the robodialers Rasmussen and SurveyUSA. In addition, RealClearPolitics always posts results for likely voters rather than registered voters despite the questionable validity of the likely voter screens used by some olling porganizations. So one should take the results posted on the RealClearPolitics website with a grain of salt. What is interesting, however, is that realclearpolitics currently shows that John Kerry is leading in the polls in 8 of 16 key battleground states with 93 electoral votes (Ohio, Pennsylvania, Michigan, Wisconsin, Oregon, New Hampshire, Minnesota, and Maine) while George Bush is leading in 8 states with 75 electoral votes (Florida, Iowa, New Mexico, Missouri, Nevada, Colorado, Arkansas, and West Virginia). This would be an improvement of 12 electoral votes for John Kerry compared with Al Gore’s showing in these 16 states in 2000.
More importantly, though, George Bush is currently averaging only 48.8 percent of the vote compared with 46.4 percent for John Kerry in the 8 states that he carried in 2000. And Bush is averaging only 45.8 percent of the vote compared with 47.9 percent for John Kerry in the 8 states that Al Gore carried in 2000. That measn Bush is running below the critical 50 percent mark for an incumbent president even in the battleground states that he carried in 2000 and well below the 50 percent mark in the battleground states that Al Gore carried in 2000.
So one thing is “real clear”–no matter what website you use to follow the polls, George Bush is in serious trouble.

I know a lot of people like Electoral Vote Predictor 2004. And it is indeed a handsome site that provides much valuable information.
However, there are some problems with it that, for me, reduce its usefulness and reliablity.
1. It is only updated once a day and the data are frequently a bit stale.
2. It does not provide links to the original, full data from a given poll.
3. Its EV calculations do not favor RV data over the more problematic LV data.
4. Its EV calculations include polls from the notorious partisan polling firm, Strategic Vision.
5. Its EV calculations include polls conducted over the internet, which have yet to prove their reliability.
6. Its EV calculations mix tracking polls in with regular polls which, since they come out every day (e.g., Zogby’s ten battleground state tracking polls), are highly likely to be the most recent polls in a given state and hence push that state’s EVs in whatever direction the tracking poll’s going. Such an approach gives too much influence to the polling firm(s) doing the tracking and should be avoided.
So where can you find a site that corrects for all these problems? You’re in luck: just trot on over to Bob Poulsen’s 2.004k.com (recently renamed NowChannel.com) and dive right in. Poulsen’s site is updated frequently, always provides links to the original data and uses a solid methodology for its EV counts:

If results with “Registered Voters” are available, those results are used. If not, “Adults” are used, or, as last choice, “Likely Voters”.
Note: Pollsters’ methods of determining “Likely Voters” are arbitrary and subjective.
Not Included: Polls from partisan polling firm Strategic Vision
Not Included: Polls from online-based surveys, including WSJ/Zogby Interactive
Methodology change on October 10: mid-date of poll is now used to determine chronological order.
Trend/Margin/Leader, and Electoral College tally, based on the most recent poll (only).
Beware, this is an imperfect system, and a single outlier state poll can make a dramatic change if it is the most recent.
For the best concept of a possible outcome, one must skeptically study each state poll for each battleground state.
Q. Where are the current Zogby Battleground State Polls, and the Detroit News/Mitchell Research Polls?
A. At the Daily Tracking Poll page.
Q. Why aren’t the daily tracking polls included among all other state polls?
A. Because these polls are published daily, including them would give too much advantage to a single polling firm (since their polls would always be the most-recent poll).

Exactly. Note that, even though Poulsen doesn’t use the state tracking data for his EV calculations, he still provides all those data, neatly and crisply arrayed on his very nifty Daily Tracking Poll page, along with all the national tracking polls, also very neatly organized.
You’ll find navigating his site very easy to do (the latest data, both nationally and from the states are all available on one page, each state has its own page, etc.) and the original data from a poll are always just a mouse click away. And he puts his latest EV calculation on the bottom of each page, along with a link to the state data being used for the calculation.
It may not be as pretty as the Electoral Vote Predictor site. But the data are better and superbly organized. I urge you to make it a frequent stop as we wind down to election day.
Corrections: Votemaster over at Electoral Vote Predictor 2004 informs me that RVs are now favored over LVs in calculations and Strategic Vision is excluded from “averaged” poll readings for states.

Today the TIPP tracking poll was a tie; at this time before the 2000 election, it was +5 Bush.
Today the Zogby tracking poll was a tie; at this time before the 2000 election, it was +3 Bush.
Today, the WP/ABC tracking poll (LV) was +3 Bush; at this time before the 2000 election, it was also +3 Bush…where it stayed, with a brief detour to +4, until its final poll, thereby missing the actual popular vote margin by 3.5 percentage points.
Today, the Rasmussen poll was +2 Bush; no information available on where it was at this point before the 2000 election (and it wouldn’t be strictly comparable anyway, since Rasmussen has substantially changed their methodology since then), but it seems safe to say that Bush’s margin was far larger–their final poll, after all, had Bush winning by 9 points.

By Alan Abramowitz
Just to reinforce the point that I made in my earlier analysis of the day-to-day movements of the tracking polls this year, in the past three days the four tracking polls have produced the following changes in the margin between Bush and Kerry. A positive number means that Bush’s margin increased in the past three days; a negative number means that Bush’s margin decreased in the past three days.
TIPP – 5
Zogby – 3
Rasmussen + 2
WP/ABC + 5
Once again, we see that there is no consistency in the day-to-day movements of the tracking polls. While George Bush was gaining in the Rasmussen and Washington Post polls, he was slipping in the TIPP and Zogby polls. All four polls continue to show the race very tight, however, which does make sense.
So stop obsessing about the day to day movements of the tracking polls and do something useful with your time. Like canvassing, or leafletting, or phone-banking, or voting.

I’ve made no secret of my preference for RV over LV data during the course of this campaign, for reasons that have been outlined in detail in many posts.
But what about now? If LV data make any sense at all, it’s now, on the eve of the election, when polls are no longer, in effect, screening voters for participation in some mythical “snap” election substantially preceding the real election, but are getting close to screening voters for the real election itself. This was, after all, the original purpose of LV screens and models and perhaps these screens and models still deserve our trust this close to the election.
I’m not so sure. Just because these models can now be employed in a way closer to their original purpose doesn’t mean they are now particularly trustworthy. The same problems that have vexed LV samples throughout this election seem likely to persist, at least at some level, right up to election day. I have, of course, reviewed these problems and the spotty track record of LV predictions in detail, and readers are urged to go back and consult those posts if they wish. And for extra credit, I strongly urge you to consult Mark Blumenthal’s (of Mystery Pollster) epic seven part series on likely voter methods (soon to be major motion picture, I understand). You really should read the whole thing, but here’s the Classic Comics version:
1. Likely voter methodology is extremely complicated.
2. Everybody does it differently.
3. They don’t really know for sure their methods work–especially during the bulk of the campaign season–and there are all kinds of ways in which likely voter methods can produce biased samples and wild swings in results that are unrelated to voter sentiment.
4. It’s not clear that the likely voter models have superior accuracy, even at election time, in today’s political environment.
Which leads me to the following query: why bother? If these likely voter methods don’t work that well, are prone to fairly serious bias and volatility and everybody has a different approach so you can’t easily compare different polls, why go to all the trouble of drawing these samples anyway? Maybe good old-fashioned, low-hassle RV samples are just as good–in fact, better.
That said, we’re kind of stuck with LV samples in most polls today, so in many cases it’s either LV data or no data. As a mitigating factor, many of these polls that just report LVs have comparatively simple and modest screens–a question or two–that produce samples not too far off from a straight RV sample. But for the complicated LV approaches that use an elaborate series of questions and scores to ferret out the allegedly likely voters (think Gallup, but others like WP/ABC use similar systems), I’d still be inclined to take their RV data (assuming they’re still reporting it) a bit more seriously than their LV data. Or, if you want a compromise, average the two.
OK. Class dismissed. Back to obsessively following the latest polls…

By Alan Abramowitz
Florida, Pennsylvania, Ohio, and Michigan are the four most important battleground states in the 2004 presidential election. Together they hold 85 electoral votes. In the 2000 election, Al Gore carried Michigan and Pennsylvania while George Bush carried Florida (barely) and Ohio. An analysis of independent polls that have been conducted in these four states since October 15th reveals that John Kerry is now leading in Pennsylvania, Michigan, and Ohio and trailing only slightly in Florida. This analysis includes only independent media and academic polls. It excludes partisan polls and robo-dial polls such as those done by Rasmussen and Surveyusa and the Zogby tracking polls. However, including the robo-dial and tracking polls would not change the overall results appreciably in any of these states. Results here are for likely voters. In most cases, John Kerry does slightly better among all registered voters than among likely voters.
In Florida (27 electoral votes) there have been 9 polls since October 15th. George Bush led in 5 polls, John Kerry led in 3 polls, and 1 poll was tied. The average of the 9 polls was Bush 47.8, Kerry 45.7, Nader 1.2.
In Pennsylvania (21 electoral votes) there have been 8 polls since October 15th. Bush led in 1 poll, Kerry led in 6 polls, and 1 poll was tied. The average of the 8 polls was Bush 46.5, Kerry 48.8. Ralph Nader is not on the ballot in Pennsylvania.
In Ohio (20 electoral votes) there have been 8 polls since October 15th. Bush led in 2 polls, Kerry led in 6 polls. The average of the 8 polls was Bush 46.5, Kerry 48.0. Ralph Nader is not on the ballot in Ohio. Remember, no Republican has ever won a presidential election without carrying Ohio.
In Michigan (17 electoral votes) there have been 4 polls since October 15th. Kerry led in all 4. The average of the 4 polls was Bush 43.8, Kerry 48.0, Nader 1.0.
Based on the most recent independent polls conducted in these 4 key states, John Kerry appears to be poised to carry Pennsylvania, Michigan and Ohio. In addition, George Bush’s lead in Florida is far from secure. If undecided voters in Florida break toward Kerry, as most analysts expect, John Kerry could sweep these four states next Tuesday.
Source for all polling data is pollingreport.com.

Reported registration levels are the highest they’ve been since 1960-64. Most of this increased registration level is attributable to increased registration in the battleground states. Here’s Curtis Gans of the Committee for the Study of the American Electorate on the likelihood of increased turnout this year:

There is reason to believe that this year’s increased registration, coupled with heightened voter interest and intense feelings generated by the 2004 election campaign, will produce substantially higher turnout this year.

Speaking as an old turnout hand, Gans is usually right in this area, so I’d put considerable stock in this assessment.
Another piece of evidence: early voting is going through the roof. On to election day.