Federal Review Composite Poll and EV Prediction: Bush 48.9 | Kerry 48.9 -- April 13, 2004

Federal Review ^ | April 13, 2004 | Darth Reagan

[Darth Reagan]

A 2000 Repeat?  Bush 48.9  Kerry 48.9
April 13, 2004 The race for president is currently a tie, with each candidate pulling 48.9% of the vote in the fully allocated April 13 Federal Review Composite Poll. Bush has kept it close in a tough news week for him and the country, largely due to a 4 point lead in the Gallup poll, a substantial lead in the Iowa Electronic Market and the results of last week’s Composite Poll in which Bush held a slight lead. Other polls, however, including Newsweek and FoxNews show Kerry with a 4 and 1 point lead, respectively.

This Week’s Polls
FoxNews / Opinion Dynamics
Newsweek
Gallup

All of the polls were taken prior to Condoleezza Rice’s testimony and release of the Presidential Daily Briefing, so it remains to be seen how those event will effect public opinion.

In addition, an analysis of Bush’s job approval numbers (52 approve, 45 disapprove), together with the favorable/unfavorable opinions of Bush and Kerry led to an allocation of undecided voters in Kerry’s favor.

The Federal Review Electoral College Prediction, which is based on the Composite Poll results as well as results in recent state polls indicate a Bush electoral victory. On the strength of positive polls for Bush in Florida (8), Wisconsin (6) and Pennsylvania (6), Bush leads the electoral vote projection 299-239. However, removing toss-up states, where a candidate leads by less than 5% in our analysis, the race remains close. Bush 198, Kerry 199 and Toss-ups 141. The Battleground States, with each candidate’s margin of projected victory, follows.

Battleground States
Florida (Bush 3.7)
Iowa (Kerry 0.9)
Missouri (Bush 2.9)
New Hampshire (Bush 1.6)
New Mexico (Kerry 0.6)
Ohio (Bush 1.9)
Oregon (Kerry 1.7)
Pennsylvania (Bush 0.6)
Virginia (Bush 3.7)
Washington (Kerry 3.7)
West Virginia (Bush 1.5)
Wisconsin (Kerry 0.1)

Since March 8, when we began this analysis, the largest lead by any candidate was on March 29, when Bush led 49.7 to 47.5. Kerry has led the electoral projection once, on March 8, 291 to 247. Bush’s largest electoral lead occurred on March 15 and 22, 338-200.

Date	Bush	Kerry
9-Mar	49.0	49.1
16-Mar	50.3	48.4
23-Mar	49.2	48.6
30-Mar	49.7	47.5
6-Apr	48.8	49.0
13-Apr	48.9	48.9

Electoral Vote Results

	With Toss Ups			Without Toss Ups
Date	Bush	Kerry	Toss	Bush	Kerry
9-Mar	133	189	216	247	291
16-Mar	199	86	253	338	200
23-Mar	184	110	244	338	200
30-Mar	194	110	234	335	203
6-Apr	198	182	158	278	260
13-Apr	198	199	141	299	239

[Darth Reagan]

This Composite Poll has grown since its inception in 2000 when I started it as an attempt to make some sense out of polls coming out the same week that had wildly different results. It went from an average, to a weighted average, and has now only gotten more complex. For the methodology, see here.

The new thing for 2004 is the attempt to project the electoral vote. This is different from Dales' invaluable Electoral College Breakdown in that it is heavily reliant on math and current national poll numbers, although local state polls are taken into account and the author's judgment does come into play, but not as heavily as in Dales' Breakdown. Thus, I let the math do most of the heavy lifting.

My attempt is to make this as reasonable as possible and any judgment calls are usually made in favor of Kerry in order to eliminate or at least mitigate my pro-Bush bias. The analysis will be available at Federal Review on Tuesdays, and I'll try to post it here in order to get the invaluable comments and advice from other Freepers.

I hope you enjoy.

[dawn53]

I'm impressed that the President is in a tie considering the news spin from the 9/11 Commission and the Iraqi situation.

And Kerry "hid" most of the week. I've noticed if Kerry gets much press time his poll numbers go down, so he's been operating below the radar.

[Darth Reagan]

I think most of the polls were done before Rice's testimony and release of the PDB. I guess we'll see the effects of those and the President's news conference in next week's poll.

[Darth Reagan]

Ping.

[Common Tator]

It is not a tie in the electoral college. It is a tie in the popular vote.

The question is what turn out numbers are the pollsters using. In 2000 they tended to look at turn outs that resembled 1988. After all 1992 and 1996 had been three man races with Ross Perot clouding the results. There were about 90 million voters in 1988, then 1992 the turn out went all the way up to 105 million, in 1996 it dropped back down to 95 million and in 2000 it went up to 105 million again.

So turn out is important. A 1988 style turn out may give bush a big win. A 2000 turn out may make it a too close to call race. Much depends on turn out.

I do know this. In good times it is very hard to get a good Democratic turn out. Democrats only agree on the economy, and if the economy is good the lower demographic democrats don't come to the polls. In good times their votes have to be bought with money and booze.. and the dead must rise to go to the polls.

I would say that if Bush wins PA, Ohio, And Florida it would be next to impossible for Kerry to win. It is quite likely that as we approach the election the Kerry support will fall.

Remember as much as half the independents will vote for the most likable candidate. The more they see of Kerry the less they will like him.

We also need to understand why the Bush administration wants to turn Iraq over to the Iraqi governing council by June. Once they are in control their lives and their families lives will depend on them retaining control. They are not likely to put up with armed rebellion. They are apt to be lot more vindictive against the rebelious clerics than we are.

When they crack down on their opponents the kick will be that Bush put thugs worse than Saddam in charge. But it is quite likely the violence will stop by the end of july.

If the killing is surpressed, then Iraq will be a big pro in the Bush Column come november.

If the insurgents thought it would be easier to take over Iraq after the transfer of power they would wait two and a half months to act. They fact they are doing it now speaks volumes.

[MattinNJ]

Excellent work.

[Darth Reagan]

I had considered reviewing state by state trends from 1988, 1992 and 1996 to compare the margin of victory per state against the national popular vote margin, and then with consideration of the turnout level. That would be a long and involved analysis, and maybe I'll get around to it.

Remember as much as half the independents will vote for the most likable candidate. The more they see of Kerry the less they will like him.

This is already worked into the analysis by the way the undecided vote is allocated between Bush and Kerry. Currently, Kerry gets more of the undecided because his negatives are not as high as Bush's. I agree, this will change.

[Darth Reagan]

Bump.

[Dales]

Outstanding work.

I think it would be helpful for everyone if you could go through the methodology in a particular state, especially one where the poll numbers vary significantly from what your analysis has. Two such states that come to mind are Iowa, where your analysis has it pretty close despite the recent polls showing Kerry way ahead, and New Jersey (which I assume you have Kerry winning although the most 'recent'-- using the word loosely since they aren't that recent-- polls show Bush ahead).

In both cases, I tend to think your slotting of them is more accurate, btw.

[Darth Reagan]

I'll revise the methodology explanation. Currently, I don't use state polls that are over a month old, hence New Jersey. When there are multiple polls, I tend to take an average, or something close to that. This is the real punditry part. If I see multiple polls for a state, I tend to take the one that seems most reasonable to me.

I haven't seen a recent Iowa poll and haven't included one in my analysis. Your analysis has been invaluable in helping me find recent state polls.

Perhaps this is a bias from 2000, but I ignore Rasmussen's polls altogether (unless it shows a Kerry lead in a state and other polls show a Bush lead, then I temper the Bush lead to help eliminate my pro-Bush bias).

Here's Iowa (positive numbers mean Bush lead, negative mean Gore/Kerry):

Two party vote margin in 2000: -0.0032
After Nader/Buchanan Allocation: -0.86
After accounting for party registration changes: -0.92
Latest Composite Poll shows a tie (or actually is is 48.98 to 48.96 in favor of Kerry, so margin on April 13: -0.94.

[Constitution Day]

Very impressive, Darth. BTTT!

[Darth Reagan]

In my eager attempt to make sure that my prognostication is reasonable, I got someone to link it over at Democratic Underground, where they are also aware of Dale's Electoral College Breakdown. I think the analysis depressed some of them, though that wasn't the intent.