« Charlie Cook Shares His Thoughts | Main | Who Is Insider Advantage, and What Do They Know About Florida? »
Friday, October 15, 2004
General Election Cattle Call, September 15
Posted by Chris BowersNational Popular Vote Projection
Bush: 49.1
Kerry: 48.9
Other: 2.0
Status: Toss-Up
Polls Included: Rasmussen, TIPP and Zogby
Electoral Projection
Kerry: 311, 197 solid
Bush: 227, 153 solid
States changing hands from 2000: FL, NH and OH to Kerry
The Battleground
AZ: Bush +4.5
AR: Bush +2.9
CO: Bush +3.6
FL: Kerry +3.5
IA: Kerry +1.0
ME, State: Kerry +4.7
ME, CD-2: Kerry +2.5
MI: Kerry +6.8
MN: Kerry +4.9
MO: Bush +2.5
NV: Bush +0.9
NH: Kerry +0.2
NM: Kerry +5.9
NC: Bush +4.2
OH: Kerry +1.1
OR: Kerry +4.1
PA: Kerry +3.4
VA: Bush +5.3
WA: Kerry +7.6
WV: Bush +5.4
WI: Kerry +4.7
Sorry about the one-week absence. I have spent the last several days examining polling methodologies, in order to determine which polls I would include in my dataset. In the end, I have decided to go with only the Rasmussen, TIPP, and Zogby tracking polls. The advantage of using only these three is that they are updated every day, so I have no need to ever change the size of the dataset. Further, I pretty much agree with the methodology used by all three. ABC isn't as bad as I thought it was, but it is not good enough for me to add it. At least it isn't Gallup.
This system should, overall, do a very good job of accounting for the Incumbent Rule (by allocating 80% of undecideds to Kerry), Party ID (all three polls I use weight by Party ID, albeit somewhat differently), and historical voting patterns (because I use the partisan index). What this system will not do a very good job of measuring is the way this election might be different from previous elections, especially concerning turnout. My projections are so deeply mired in polling history that the further 2004 deviates from past elections, the more inaccurate I will be.
Posted at 02:32 PM in General Election Cattle Call | Technorati
Comments
Well, a higher turnout necessarily favors us, so if you are off about that, it will only be in our favor.
Posted by: DavidNYC at October 15, 2004 02:37 PM | Permalink | Edit Comment | Delete Comment
Am I correct in saying you take the average of the three national tracking polls, compare it to the 2000 national results, then calculate the impact on each state based on the 2000 results?
The problem with that is that it assumes each state's position in the relative D/R ratings will be the same as in 2000. That doesn't account for demographic shifts, disproportionate increases in minority registration, and other trends.
Although it is the only figure backed by an actual vote rather than a poll, the 2000 partisan index is four years out of date. I'd deem it less useful than current state polls.
Posted by: Tyrone at October 15, 2004 02:39 PM | Permalink | Edit Comment | Delete Comment
I can't seem to track down the link to the last methodology post. Chris, perhaps you can re-post it?
Posted by: DavidNYC at October 15, 2004 02:40 PM | Permalink | Edit Comment | Delete Comment
Do you really think Kerry could be up in Ohio? You're percentages seem to favor Kerry in many of the close ones....I know this whole thing will come down to turnout. I strongly believe that the Democratic base is more fired up and determined to get to the polls no matter how much Bush plays to his base. They seem content, not angry. And in my opinion, in American politics in the last 10 years, nothing motivates voters to get to the polls like anger has.
Advantage to Dems, or so I hope. GET OUT THERE AND CANVASS, TALK TO VOTERS, PUT UP SIGNS, WORK YOUR ASS OFF PEOPLE!
It's how this will be won, and if you don't get your ass out there, you didn't do your part.
Posted by: Doug in Virginia at October 15, 2004 04:10 PM | Permalink | Edit Comment | Delete Comment
"What this system will not do a very good job of measuring is the way this election might be different from previous elections, especially concerning turnout. My projections are so deeply mired in polling history that the further 2004 deviates from past elections, the more inaccurate I will be."
If only every pollster was so honest!
Posted by: Delicious Pundit at October 15, 2004 04:32 PM | Permalink | Edit Comment | Delete Comment
Folks, you are cherry picking. Particularly in the battleground state polls. Pick a cut-off for the date of the poll, in October for example, and look at your data. Trying to interpret a difference of 1.0% just ain't meaningfull, and I won't even comment on NH and 0.2%. Try again.
If you are going to cherry-pick anyway, and most of us can't resist, then look at the data that normally doesn't favor your side. In my case, that would be Zogby, the Chicago Tribune, ARG, and anything paid for by the Dems. WI looks like it might be tough for the GOP, but IA might be closer. I don't believe any GOP poll with a 1 or 2% margin for Bush. That "lead" will evaporate. Likewise, I would suggest that a 1 or 2% margin for Kerry by Zogby or ARG would be ... suspect.
Posted by: MarkOlsen at October 15, 2004 04:46 PM | Permalink | Edit Comment | Delete Comment
I distinguish between a pollster explicitly paid for by Dems or Repubs, and one who is independent, but who has been pro-Dem or pro-repub.
I see ARG and Zogby as being pro-Dem, Mason Dixon a being pro-Repub, Ramsussen and Survey USA as beign all over the map.
Posted by: erg at October 15, 2004 05:09 PM | Permalink | Edit Comment | Delete Comment
Good news in latest Washington Post poll.
Bush 48, Kerry 48, Nadar 1%
13 battleground states
Bush 43%, Kerry 53
A total of 1,555 registered voters were interviewed Tuesday through Thursday nights, including 1,203 likely voters.
Posted by: DFuller at October 15, 2004 05:15 PM | Permalink | Edit Comment | Delete Comment
And to add to that, there was responses ont he mary Cheney flap. While 64 % disapproved, including 4 in 10 of Kerry's supporters, only half of uncommited voters disapproved.
While its clearly preferrable that he hadn't made the comment, it seems it did not go over as badly as I feated., Even among uncommitted voters. And very few of those small groups are likely to change their opinion based on this.
Ok, my negativism is officially on hold (hopefully forever) on this point. [ Incidentally, I believe that if Kerry had used the word gay, rather than lesbian, the flap would ahe been much smaller]
Posted by: erg at October 15, 2004 06:00 PM | Permalink | Edit Comment | Delete Comment
One problem I see in your methodology is the 80-20 split for undecideds. Apart from 80% being rather high (others seem to go with something like 2-1), I also wonder whether it holds at this stage already. The closer one gets to the election, the less voters will be undecided. Has the campaign already gone far enough that we only have the 'ultimate undecided's left?
Posted by: Andre at October 15, 2004 07:22 PM | Permalink | Edit Comment | Delete Comment
My current analysis:
Of the important swing states, Maine (except 2nd district), Minnesota, New Hampshire, New Mexico, Pennsylvania are leaning Kerry. This gives 246 EVs.
The main battlestates beyond that are Florida, Iowa, Nevada, Ohio and Wisconsin (plus the 2nd district from Maine, and the 4 votes of the Colorado amendment). To get 24 EVs here, Kerry would need either Florida alone, or Ohio and one other state (WI+IA+NV would do with the CO amendment, but I _really_ hope the CO amendment is not going to matter, or we get 2000 back squared). The "winner is the win who gets two states from FL/OH/PA"-prediction still stands.
Posted by: Andre at October 15, 2004 07:42 PM | Permalink | Edit Comment | Delete Comment
The horse race:
Three day tracking polls Today (Yesterday)
Zogby
Bush 48 (46)
Kerry 44 (45)
Unsure 7 (7)
ABC
Bush 48 (48)
Kerry 48 (48)
TIPP
Bush 47 (46)
Kerry 44 (42)
Unsure ? (12)
Rasmussen
Bush 49 (49)
Kerry 46 (46)
Unsure 3 (3)
Average
Bush 48 (47.3)
Kerry 45.5 (45.3)
Bush +2.5 (+2.0)
Posted by: DFuller at October 15, 2004 07:59 PM | Permalink | Edit Comment | Delete Comment
We often hear about the candidates' "internal polls", which are allegedly more accurate and more expensive. Does anybody know what sort of polls these are? Are they simply bigger samples? (Although a factor of 10 in sample size only lowers the error by a factor of 3.).
Posted by: science at October 15, 2004 09:25 PM | Permalink | Edit Comment | Delete Comment
Wisconsin poll data shows that Michael Badnarik will help Kerry win the state.
47%-Bush
48%-Kerry
1%���Badnarik
1%���Nader
1%���Some other candidate
2%���Not sure
Badnarik is running ads in Wisconsin to pull disgruntled republicans.
Posted by: VoteBadnarik at October 15, 2004 09:33 PM | Permalink | Edit Comment | Delete Comment
I suspect the candidate polls use far more questions, especially to see what policies are more popular or not.
As for size, also remember that internal polls are probably very interested in sub-areas of a state and ethnic groups, and other classifications. Raising the overall size may not decrease the overall error rate that much, but for subgroups, it could lead to much better sample size.
Re: Badarnik -- Ive always thought that the Libertarian party could be useful to Dems in NM, NH, NV, possibly AZ. Even WI. True Libertarians wont care for the Patriot Act or the IRaq war.
Posted by: erg at October 15, 2004 09:56 PM | Permalink | Edit Comment | Delete Comment
Sorry for being so trolly, but l-e-s-b-i-a-n. I like the ring to it! Did anyone watch that late night show, Mark Kimmel Live? His commentary showing Kerry and Edwards using the L word, LESBIAN, was absolutely hillarious. ;) Using the L word, LESBIAN, was brilliant strategy. Brilliant! You have to love controversy.
It looks like Ohio and NH are trending in Kerry's direction:
http://www.cnn.com/2004/ALLPOLITICS/10/15/electoral.map/index.html
Posted by: Shar at October 15, 2004 10:03 PM | Permalink | Edit Comment | Delete Comment
The horse race:
Three day tracking polls Today (Yesterday)
Zogby
Bush 48 (48)
Kerry 44 (44)
Unsure 7 (7)
ABC
Bush 50 (48)
Kerry 47 (48)
TIPP
Bush 47 (47)
Kerry 44 (43)
Unsure 10 (11)
Rasmussen
Bush 48 (49)
Kerry 46 (46)
Unsure 3 (3)
Average
Bush 48.3 (48)
Kerry 45.3 (45.3)
Bush +3.0 (+2.7)
Posted by: DFuller at October 16, 2004 10:54 PM | Permalink | Edit Comment | Delete Comment