Monday, November 3, 2008

Brief preview of final polling

Here is a brief preview of final state polling, and which polls (IMO) to look for:

Ohio: The university of Cincinnati poll is usually counted as the gold standard for Ohio. The SUSA poll nailed the primary results so look out for their final poll and perhaps also PPP which came very close and also will have a large sample size. (The university of Cincinnati also came very close). Just for some variability also look to the Rasmussen poll which weighs it’s sample by party-id.

Pennsylvania: There is no one poll which I would emphasize in the keystone state but probably an average of Rasmussen, SUSA and the Muhlenberg/Morning Call tracker should give a good prediction. I would normally add PPP but this was the one state which they got badly wrong during the primary. They were open about this and have apparently corrected the problem, hopefully this will be shown on Tuesday.

Florida: Mason-Dixon has a good reputation in the sunshine state. It is also below the mason-dixon line (Mason-dixon, at least in the east, tend to poll well below the line but badly above), also look out for Rasmussen and perhaps Quinnipaic (large sample).

SUSA was the closest to the actual results in 04 with Mason-dixon a close second so look out for those polls. Perhaps also keep an eye out for the final rocky mountain news poll although they were out somewhat in favour of Bush in 04. PPP have already released their final CO poll showing Obama up by 10 with a heavy lead among those already voted.

Virginia: This is a little bit of guess-work as in the past there has only been infrequent polling of the presendential race. Times-Dispatch was the closest in 04 followed by mason-dixon, unfortunately this year Mason-Dixon did the polling for Times-Dispatch and during the primary they missed the result by over 10 points however they did nail the 06 Webb/Allen race. To make matters worse the final Mason-Dixon results this year was 47O-44M with a far too high 9% undecided. SUSA who came closest to the primary result should be taken seriously. (Their final poll is 50O-46M). Perhaps the average of PPP, Mason-dixon, Rasmussen and SUSA might be the best bet.

North Carolina:
The in-state PPP should be the first call. They both have the most experience and the most at stake in their home state, not to mention being closest to the primary result. The other in-state pollster Civitas should also be looked at (although it is affiliated with the republican party). SUSA for some reason tends to lean to McCain in North Carolina so anything equal to or better than a tie would be excellent for Obama.

Missouri: Rasmussen, SUSA and Mason-Dixon all came very close to the 04 result so taking the average of these polls would seem reasonable here.

Indiana: This should be an interesting one considering how many years it has been since Indiana has been a swing state. SUSA was about the only pollster to poll Indiana in the last week in 04 and came very close. In the primary PPP did very well. So perhaps an average of these two would be in order. Also if Selzer does another poll before Tuesday that should be a very good indicator.

Iowa: Well SUSA’s final poll has it at Obama +15, the oracle of Iowa (Selzer) has it at +17. McCain’s internals apparently have it at +1. No prize for guessing which is right.

New Hampshire:
UNH tracker, Mason-Dixon and Rasmussen all came very close to the actual result in 04. So an average of these should give a reasonable result. All polls missed the primary results by a large margin. (Fingers crossed that does not happen ever again).

New Mexico: Unlikely to be close especially looking at the early voting but if a prediction is required then an average of SUSA, Rasmussen, PPP and the Alberquerque Journal would be seem reasonable.

Nevada: Hard state to poll, but Rasmussen did best in 04. Also note the early voting has been especially heavy in this state giving the Democrats a very strong lead.

Georgia: With African-American turnout (looking very high in early voting) and support for Barr being very difficult to predict this year, any poll could be a long way off. The in-state Insider Advantage tends to predict a higher turnout for African-Americans so perhaps this is the poll to look out for, but only if they (unlike usually) have a decent sample size.

Montana: There is a expected relative high third party vote in the Treasure state, so look out for polls which show this.

North Dokata: One of the problems with expanded the map is a complete lack of polling history. With nothing really to base results on I would suggest just averaging all (if there are any) poll from the last week.

Arizona: Recently added due to tightening polls. Again not much history to run on, in 04 the result was exactly in the middle of Rasmussen and SUSA, so perhaps an average of these.

Notes: Do not rely too much on any one poll (especially one with low sample size) as even the best poll will suffer from sampling error.

Mason-Dixon did very well in 2004 but are using a conservative model in terms of turnout especially in terms of youth vote and African-american (and Hispanic) make-up of the electorate, so there is a good possibility of this pollster being off by 2-3 point this year.

Do not take anything zogby interactive or ARG show seriously.

No comments: