FederalReview Composite Poll and E.V. Prediction, July 20, 2004, Bush 47.5%-246 | Kerry 49.8-292

Federal Review ^ | July 20, 2004 | winston

[Darth Reagan]

THE BOUNCE HITS THE ELECTORAL VOTE 

Bush 47.5% – 246 EV | Kerry 49.8% – 292 EV
EV without Toss Up states (under 2% margin): Bush 216| Kerry 250| Toss 72

July 20, 2004 Florida finally slips over into the Kerry column after weeks of hanging out with Bush despite the popular vote swing toward Kerry. But, as predicted last week, the Edwards bounce has arrived in the electoral college, where Kerry now leads 292-246.

You'll also notice a change in the above graphic. Despite the convenience (and now cliché) of referring to "Red States" and "Blue States", I've decided to adopt what I understand is the traditional color coding standard. The incumbent is blue and the challenger is red. That suits me, because I think red suits Kerry better.

But before we get to the numbers, and our eye catching new way of presenting them, let's talk about trends and forecasting. Thanks to the wonderful versatility of Excel and my own spreadsheet OCD, I've run some predictions based on the last 4 months of numbers. Since I've always said that I think this analysis works more for showing trends in the race and not for showing a specific daily snapshot or actual prediction of where things will be in November, I ran a least squares linear regression analysis to forecast the November 2 results. Basically, if the prevailing trends in the race over the last 4 months continue until election day, then this forecast will be accurate. Of course, that's an untrue assumption, isn't it. Dynamics will change. But I thought the results are interesting, especially as everything is swinging back toward Kerry.

So, the popular vote forecast for election day is Bush 46.8 | Kerry 48.7. Forecasting the electoral vote, state by state according to our weekly analysis, says Bush will win 278-260 - the same spread (adjusted for the census) that Bush beat Gore, but some of the states are different. But I'll tell you right now that if Bush loses the popular vote by almost 2 percent, he is not going to win the electoral vote.

Now, about the oft repeated notion that the undecides break for the challenger. In analyzing the weekly polls that this analysis uses (not including the Iowa electronic markets, which is not a poll), the Undecideds come in at an average of 6.85%. I use a convoluted method analyzing job approval and favorability ratings to determine how to allocate the undecided. Right now, they break down in favor of Kerry, who gets 56%-44%. But Dick Morris tells you they break much more heavily for the challenger. I'm not so sure.

I analyzed the numbers since 1976 reported by the National Council on Public Polls. Since that election, undecides have gone in favor of the challenger 55%-34%. Not too far off my number, but more favorable for the challenger. But the thing with statistics, you can find contradictory numbers.

Is challenger status the only thing that may effect how undecided voters break? Perhaps Democrats are generally more attractive to late deciders. So, I ran the numbers by political party. Generally, undecides break toward the Republican, 48-40. That's not a big advantage, but certainly different than the challanger/incumbent analysis. Why should one be more correct than the other?

Well, what about Republican Incuments versus Democrat challengers? Undecides break for the Democrat challenger 42%-31%. So, I took all those measures that are consistent with this race, Challenger vs. Incumbent, Republican vs. Democrat and Republican Incumbent vs. Democrat Challenger, and averaged them. The result, if this is actually predictive, is that undecides in this race will break for Kerry 55%-45%. Consistent with my convoluted job approval / favorability analysis that today allocates the undecided 56%-44%. Does all of this analysis mean that my numbers are the best predictions of the trends in this race? I leave it up to you to decide.

Now, some more numbers:

This Week's Polls
CBS/New York Times (Bush 42 | Kerry 45)
Investor's Business Daily (Bush 43 | Kerry 47)
Rasmussen Week Average (Bush 45.7 | Kerry 47.1)
Iowa Electronic Market (Bush 52.4 | Kerry 48.2)

Last Week's Polls
AP/Ipsos (Bush 49 | Kerry 45)
CNN/USAToday/Gallup (Bush 45 | Kerry 50)
Zogby (Bush 45 | Kerry 47)
Newsweek (Bush 44 | Kerry 47)
Rasmussen Week Average (Bush 45.7 | Kerry 47)
Iowa Electronic Vote Share Market (Bush 50.9 | 47.5)



Bush State Gains
None.

Kerry State Gains
Florida
Wisconsin

Toss Up States (Slight Lead of <2%)
Arkansas *NEW* (Bush)
Florida (Kerry)
New Hampshire (Bush)
Ohio (Bush)
West Virginia *NEW* (Kerry)
Wisconsin (Kerry)

See the fancy New Data Page.

[Darth Reagan]

Just so you know, I'm the author of this analysis. Some don't realize that and are surprised after flaming me over my obviously biased liberal pro-Kerry analysis.

Personally, I still hope and believe that Bush will win with over 300 electoral votes, that the "undecideds" won't vote and that some of Kerry's support is so soft it will switch to Bush.

I hope.

[areafiftyone]

Already posted HERE

[Caipirabob]

Cool, at least I have one poll I can look to. As far as a numbers, well, we have ground to cover...

[Wonderama]

Darth, this is a very intersting poll. Thanks for all your work. It looks like it's going to be a horse-race all the way to Nov.2

[My2Cents]

I trust your analysis, but what I don't trust are the poll findings you're using in your analysis. I just find it hard to believe that nearly 50% of the voters, nationwide, are prepared to vote for that opportunistic, flip-flopping, absentee, leftist.

[austinrepub]

I agree -- well, I used to until 2000. It's equally unthinkable that so many people voted for Gore.

[Republic If You Can Keep It]

Darth --- I haven't seen a single poll showing Kerry leading in either Va., Mo. or Col.

I understand that you are looking at trends rather than current polls. Still, I can't see Bush losing any of these 3 states.

[Dales]

I am struck by how similar the results of your methods and mine are right now-- despite the methods being very, very different!

If you go by the ECB "with tossups", there are only two differences: I have WI and NH as complete tossups, and you have the former going slightly for Kerry and the latter slightly for Bush.

If we expand it out to the next level, I have 3 more states designated for each guy-- Virginia, Arizona and Colorado to Bush, and Michigan, Oregon and Washington to Kerry. But I think that those are the 6 leaning states most likely to flop, so I can't argue with your designations with them at all.

I like where we stand right now.

[My2Cents]

I think a good number of people voted for Gore because they were genuinely concerned about Bush's inexperience. In four years on the job, those doubts are legitimately erased, but Kerry seems to be getting the same level of support Gore received.

Frankly, I think the only thing these polls prove is that about half of America is braindead, and morally challenged.

[Dales]

Er, not Arizona. Nevada.

[Redmen4ever]

What is very interesting in your analysis is the tilt you impute to the electoral college, so that Bush might win with as much as a 2 point deficit in the popular vote.

Given the current numbers, you have three states with 0.2 or less difference in the vote (FL, NH and WV). If the election were held today, these states would be absolutely unpredictable, and the election would go whichever way Florida went.

I would add two considerations to your analysis of how to allocate the undecideds (which, by the way, I think you have done well). With a close race, some of those favoring third-party and independent candidates will shift to their acceptable second choice. To illustrate, part of Gore's surge in 2000 came from last minute shrinkage of Nader's vote. If I am right about this, the 2 point tilt in the electoral college might turn out to be a smaller tilt, perhaps 1/2 to 1 point.

Second, Bush's approval ratings, the right direction / wrong direction number for the country, and Bush's numbers in the head-to-head might all improve by election day. We have reason to expect that they will improve because of the momentum built into the economy. Thus, the Fair model points to a compfortable win for Bush. (Of course, in 2000, the Fair model pointed to a compfortable win for Gore.)

This year, we can suspect that the unpleasant realities of the war in Iraq will be considered by voters, in addition to the economy (just as, in 2000, Clinton's personal shortcomings were considered along with the economy). This means that the Fair model might be off. However, in terms of the dynamics of the race from now through election day, I think we can say that the economic news will tend to be favorable for Bush. on the other hand, the news from Iraq from now through election day is unpredictable.

[Darth Reagan]

That's a valid point regarding the Nader vote fleeing to Gore at the last minute in 2000, which tells us that maybe a dead even race is really a slight Kerry advantage on election day. I hope it doesn't come to that.

And I hope the Fair model is right!

[Darth Reagan]

I am also struck by the similarities. When I constructed my spreadsheet, which makes all the calls for me except for whether to include a poll (I didn't use Marist because there aren't many and I like to keep with the usual polls, and I don't use Zogby Interactive because you really convinced me that we can't rely on them (yet)), I intended for it to make what I would think is a reasonable estimation of both the popular vote and the electoral college. But the numbers will control and I won't have to make the tough calls you do and I think I'd be less objective that you (I'd find it almost impossible to believe that Bush could win Wisconsin, for instance, or that Kerry could win Nevada).

Thus, when I see Virginia falling into the toss-up range (within 6 points - the margin of error on most polls, as you know), I ask if that is reasonable. If not, then I would have to change the methodology that I employ.

In fact, I did make one change this week. I changed my weighting of state by state polls, because I noticed that early June polls showing Bush up 7 and 10 in Florida were weighted too highly, because they still resulted in putting the state in the Bush column despite 4 other, more recent, polls, 3 showing Kerry up and one with Bush up 1. So, I increased the weights of polls taken within the last month relative to older ones. Thus, a more reasonable picture of where I see the race today.

And, where you see the race today.

[Darth Reagan]

I did a search but didn't see it. Note to self: Improve thoroughness of posting searches.