Two questions about the state polls (0.00 / 0)

1.  Why do they seem to lag the national polls?
2.  Why do you say that some of the swing state polls “are afflicted by age?”

---

by: ThomasPaine @ Wed Sep 17, 2008 at 13:38:12 PM CDT

 

some of the state polls (4.00 / 1)

were in the field with data collection during McCain’s bounce which ran from Palin’s RNC speech to about Sept. 14. So their old or “aged.” His “real” (today) support is lower than it was just a few days ago.

John McCain or Barack Obama?

---

by: tremayne @ Wed Sep 17, 2008 at 13:44:34 PM CDT

[ Parent ]  

True, but a State Poll is a Better Indicator (0.00 / 0)

If the data is fresh since the election turns on EVs. It matters little how big Obama’s margin in New York, or McCain’s in Oklahoma gets as long as they win. The problem is that this election has been changing swiftly, especially now.

During July when Obama’s lead was steady the state and national polling lined up.

But it’s really difficult to rely on polling right now since there are big differences between all the polls.

Ex: Rasmussen uses BOTH weighting for Party ID and a Likely Voter screen. That means they weight their sample based on the most recent polling they do on what the total number of Democrats versus Republicans should be.

Gallup uses neither. Research 2000 uses both, but their sample weighting is different so they produce very different results.

Rasmussen adds a likely voter model that tries to determine how likely voters are to vote. This is highly problematic throughout most of the election we had almost ZERO idea who a “likely voter” is!

Starting sometime around the beginning of October Likely voter models begin to work better than registered voter models. RV models like Gallup generally have a slight Democratic bias because some registered Democrats don’t vote.

If all Democrats registered and voted at the same level as Republicans they would never lose an election, but they don’t. And Democratic party loyalty is a lot lower than Republican loyalty right now too.

For instance, Research 2000 polls for DailyKOS uses a partisan ID of Democratic +9% (Dem: 35%, Rep.: 26%, Indep./unknown: 39%.

Rasmussen is far more likely right in it’s estimate of Democratic +5.1% right now, but that too has varied. In July it was Dem. +10. Even 5.1% would be a rout, if Obama wasn’t getting about 82% of Democrats compared with McCain’s around 90% of Republicans. That’s what’s kept the election close — Obama’s inability to rally the same 89% of Democrats that Kerry got in 2004. If he could do that he’d blow the election wide open.

---

by: Cugel @ Wed Sep 17, 2008 at 18:12:34 PM CDT

[ Parent ]  

that is why last week at this time the state polls favored obama (0.00 / 0)

slightly.  It takes about a week for the state polls to “catch up” to the national polls.  This time next week (2 days before the next debate) it will be Obama by 3-5 points and then he has to close in the first debate.  if he closes/convincingly presents his ideas and himself, I think it will start to shift his way in places like VA, OH, FL, MI, MN, more so than it is right now.  

---

by: georgeg1011 @ Wed Sep 17, 2008 at 13:50:15 PM CDT

 

But are state polls' methods any different (0.00 / 0)

In other words, do state polls lag simply because they are not undertaken as often, or do the results of a state poll somehow “lag” behind a simultaneous national poll?

---

by: ThomasPaine @ Wed Sep 17, 2008 at 13:55:39 PM CDT

[ Parent ]  

can't think of any reason they would (0.00 / 0)

but it’s possible, I guess, that there are regional differences. Some states may be slower to “change” than others…

Mostly it’s the fact that they are less frequent and most of the projections are done the basis of averaging the most recent one’s done in a state.

John McCain or Barack Obama?

---

by: tremayne @ Wed Sep 17, 2008 at 14:04:07 PM CDT

[ Parent ]  

I always figured it's a matter of cost. (0.00 / 0)

Getting a statistically valid sample of one state takes a lot more than one-fiftieth of the respondents for the whole country. So pollsters would be inclined to do more national polls than one-state ones.

---

by: DaveW @ Wed Sep 17, 2008 at 14:09:42 PM CDT

[ Parent ]  

yes (0.00 / 0)

towards the end, though, they’ll put more resources in the key states.

John McCain or Barack Obama?

---

by: tremayne @ Wed Sep 17, 2008 at 14:12:23 PM CDT

[ Parent ]  

Question About Daily Tracking Polls (0.00 / 0)

Why did the Daily Kos poll pick up the trend before the other daily polls.  Could it be that some are manipulating their data or is it just the newness of the Kos?

---

by: steve katz @ Wed Sep 17, 2008 at 13:53:08 PM CDT

 

trend? (0.00 / 0)

that poll (and Diageo I think) have calculated there are more Dems than the other polls (Rasm. and Gallup maybe) think. They weight their data accordingly. They would make them consistently lean a bit to Obama rather than trending in that direction recently.

John McCain or Barack Obama?

---

by: tremayne @ Wed Sep 17, 2008 at 14:06:48 PM CDT

[ Parent ]  

Why that, though? (0.00 / 0)

I mean, why do Dailykos and Diageo calculate more dems than the other polls? Is there some decision they made in structuring the DailyKos poll that makes it that way? Or is it just a coincidence that the tracking poll commissioned by a Democratic group happens to predict a larger number of Democrats?

I dunno. This whole practice of normalizing electoral poll percentages to partisan identification percentages, at all, seems bizarrely circular to me.

---

by: mcc @ Wed Sep 17, 2008 at 14:55:52 PM CDT

[ Parent ]  

R2k report what they find from what I understand (4.00 / 1)

So does Hotline who certainly aren’t a Democratic group. Gallup we don’t know. Rasmussen weights by averaging over a certain period. The to weight or not to weight debate is as old as polling and both sides have decent arguments.

---

by: conspiracy @ Wed Sep 17, 2008 at 15:03:51 PM CDT

[ Parent ]  

"report what they find" (0.00 / 0)

You’re saying R2K is not weighting responses to partisan identification?

Interesting.

---

by: mcc @ Wed Sep 17, 2008 at 15:33:30 PM CDT

[ Parent ]  

What Is The Trend (0.00 / 0)

Are there more Dems?  By what %?  By what you are saying, using the number of Reps to Dems to Indies you can skew a poll anyway you want.  

---

by: steve katz @ Wed Sep 17, 2008 at 16:45:41 PM CDT

[ Parent ]  

What happens (0.00 / 0)

to the polls’ margin of error when you calculate the average? Is there anything in today’s result that is statistically significant? Is there a way to quantify that?

---

by: DaveW @ Wed Sep 17, 2008 at 14:02:04 PM CDT

 

using an average of the (0.00 / 0)

4 polls (which are each averages of 3 days worth of polling) takes a lot of choppiness out of the day-to-day movements. It also means we are looking at a much larger daily sample than the individual polls have. Yes, this reduces the likelihood of error.

Another thing that increases the odds of movement being “real” is a consistent pattern over time. Here we have 4 polls that all show a slipping McCain. Very unlikely that occurs by random chance.

John McCain or Barack Obama?

---

by: tremayne @ Wed Sep 17, 2008 at 14:10:45 PM CDT

[ Parent ]  

The seat of your pants high school statistics calculation (0.00 / 0)

tells you (for the case of equal margins) to reduce the margin of error by one over the square root of the number of polls you include.  For unequal MoE’s, there is a (slightly longer) procedure.  A real calculation would take a bunch of other factors into account since you are measuring a ratio, weighting to various groups (each of which has a size with a margin of error),  blah blah blah.  

Of course, this is all assuming that systematic errors in the data collection are absent.

But the overall pattern is to reduce it by one over the square root.  So, roughly, if you take four polls, each with a 5% MoE, and average them, you get an average with a 2.5% MoE.  

---

by: Valatan @ Wed Sep 17, 2008 at 15:45:36 PM CDT

[ Parent ]  

and 47-45 is same as Reuters/Zogby national poll (0.00 / 0)

Released today.  

New Jersey politics at Blue Jersey.

---

by: Hopeful in NJ @ Wed Sep 17, 2008 at 14:44:19 PM CDT

 

Whatever happened to what's-her-name?… (0.00 / 0)

… Nobody seems to be bothered about Sarah Palin on Hannity tonight.

I think that particular caravan has passed by ….

---

by: tobyj @ Wed Sep 17, 2008 at 14:55:52 PM CDT

 

The key… (4.00 / 1)

 …was to contain Sarah Palin’s appeal to the wingnut world, and ONLY the wingnut world.

 It appears that we’re getting there.  

“We judge ourselves by our ideals; others by their actions. It is a great convenience.” — Howard Zinn

---

by: Master Jack @ Wed Sep 17, 2008 at 15:27:16 PM CDT

[ Parent ]  

Cramer is a Democrat (4.00 / 2)

Has been for years, as is his wife, based on 10 years of reading him.  For a time after 9/11 he was enamored of Bush and talked very tough about terrorism (he lost many friends and neighbors that day).  But he began to realize that the Bushies knew nothing about economics or finance and were just ideologues.  He began to rail against them more and more, becoming very outspoken over the past year or more, and pretty heated over the past 6 months.

John McCain–He’s not who you think he is.

---

by: Mimikatz @ Wed Sep 17, 2008 at 14:58:41 PM CDT

 

Obama's ahead again on pollster.com now, too (0.00 / 0)

That wiggly line there sure is wiggly.

---

by: mcc @ Wed Sep 17, 2008 at 15:37:45 PM CDT