Rasmussen (re-2008)

[A18]

That's probably true, but those are the people that will vote for their respective parties anyway. What matters is how the swing voters view things.

The same sample was used across the four scenarios but doesn't allow for any candidates specific-appeal and both Democrats and Republicans broke heavily along party lines (whatever the face-off), it was unaffiliated/Independent voters who accounted for most of the differentials across the four potential match-ups

What does that have to do with tyhe obvious truism I posted?

[Storebought]

Also according to Rasmussen, the GOP tends to win back-to-back elections by small majorities; the Democrats win only when the GOP is fractured.

If the national mood is such that the Democrats can win only by nominating a Bayh type, then, more likely than not, the GOP will win its third consecutive election.

Now, if Hillary! is nominated, then the US is already seriously f*ed, in which case the GOP would not stand a chance.

[WMS]

Not bad guesses at all. In case 3, I'd be parsing the hell out of the policy differences to see what issue would break me toward one candidate or the other. In case 1, I'd flip a coin.

And from your other post, yep, it's all us unaffiliated/Independents (who have a range of opinions of our own of course) who determine these races.

I've been cautious because I think that only under the most favourable, or adverse, of circumstances can I see support for the winning or losing party rising much over 55% or below 45%, respectively - but moderates do have seem to have an early advantage over ideologues, which, personally, I hope holds come 2008

Dave

I agree - it's not likely the election will go past 55-45, unless something really bizarre happens.

[muon2]

Look at the undecided vote:
1) 21% (this would include me, BTW, and would be 2004 all over again)
2) 19% (not likely to include me)
3) 20% (probably including me again, for the reverse of the reasons I'd be undecided in case 1 since I'd actually like the candidates this time )
4) 20% (not likely to include me either)

Up in the air, really...

I've been doing a bit of guess work to see how the 'undecideds' would break in each scenario:

1) Liberal Democrat 40% vs Conservative Republican 39%

I think the undecideds would break 3/2 in the conservatives favour, so it would be Conservative Republican 53%, Liberal Democrat 47%

2) Moderate Democrat 46% vs Conservative Republican 35%

I think the undecideds would break 55/45 in the moderates favour, so it would be Moderate Democrat 56%, Conservative Republican 44%

3) Moderate Democrat 42% vs Moderate Republican 38%

I think the undecideds would break 50/50, so it would be Moderate Democrat 52%, Moderate Republican 48%

4) Liberal Democrat 36% vs Moderate Republican 44%

I think the undecideds would break 60/40 the moderate Republican's favour, so it would be Moderate Republican 56%, Liberal Democrat 44%

This is pure guess work on my part, and doesn't allow for any major third party candidates

As you say, it's up in the air but, nevertheless, still pretty encouraging for Democrats, especially moderates

Dave

The numbers are in the general range, but some details could be tuned a little better. For instance, if a cons. Rep. splits undecideds 3/2 (= 60/40)  with a lib. Dem, the overall poll results suggest that a mod. Rep. should do better splitting the vote with a lib. Dem. Why not use 70/30, or at least 65/35, instead of repeating 60/40?

If the candidates are equally distant from the center, and their are no other factors, the conservative generally gets the edge among undecideds. Why would a mod. vs mod. be different than a con. vs. lib.? If we use the same 60/40 factor for an undecided split in case 3, the result becomes a 50-50 tie.

Of course, as others have posted, the split depends on real candidates with issues that can affect the undecided in different ways.

[TheWildCard]

You can't use this system. They don't mention what a moderate Republican or Democrat is. It also depends highly on the person who is running

For all I know a Moderate Republican is a social conservative and a fiscal liberal which would mean I probably would not support him.

[Erc]

In other Rasmussen news:

In yet another Rasmussen Virginia Poll:

In a matchup in 2008 between Mark Warner and George Allen, Warner beats Allen by 5 points:

46-41-5

While Allen beats Hillary by 9:

50-41-4


http://www.rasmussenreports.com/2005/Virginia%20President%202008.htm

[CARLHAYDEN]

It's early days but Rasmussen did a survey over 8-9 June of 1000 adults and found the following results in four possible generic match-ups:

1) Liberal Democrat 40% vs Conservative Republican 39%

2) Moderate Democrat 46% vs Conservative Republican 35%

3) Moderate Democrat 42% vs Moderate Republican 38%

4) Liberal Democrat 36% vs Moderate Republican 44%

www.rasmussenreports.com/2005/Election%202008%20Generic.htm

Finding 1) seems to blow my argument that a Liberal Democrat cannot beat a Conservative Republican out of the water, although a 1% lead is negligible

However, finding 2) does add weight to my argument that a Moderate Democrat could win 'big' over a Conservative Republican, while a Moderate Democrat would also will narrowly over a Moderate Republican. However, should it be a Liberal Democrat versus a Moderate Republican, then the Democrat is toast

It's early days, but at this stage adults seem to be warming to a moderate Democrats and I'll continue to argue that they'll have a better chance than liberals in winning the Oval Office

However, this is a generic poll and doesn't really allow for any particular candidates strengths and weaknesses and most of the changes took place among voters not registered with any party

Dave

Scotty has more Democrats than Republicans in his polls.

He made this change after 2000.

[Democratic Hawk]

That's probably true, but those are the people that will vote for their respective parties anyway. What matters is how the swing voters view things.

The same sample was used across the four scenarios but doesn't allow for any candidates specific-appeal and both Democrats and Republicans broke heavily along party lines (whatever the face-off), it was unaffiliated/Independent voters who accounted for most of the differentials across the four potential match-ups

What does that have to do with tyhe obvious truism I posted?

The point is I actually agree with your truism. Since Democrats and Republicans, invariably, break for their party's candidate - the battle is for the "swing" voters, who are usually Independents/unaffiliateds and it would seem, from this poll, that they would break for moderates over ideologues

So, yes, you are right the result matters on how the swing voters break in the actual election

Dave

[AuH2O]

The idea of "swing" voters is overrated. Most people that have that much trouble picking one side or another just don't vote. There are a number of weak partisans, and they are indeed relevant.

[Democratic Hawk]

It's early days but Rasmussen did a survey over 8-9 June of 1000 adults and found the following results in four possible generic match-ups:

1) Liberal Democrat 40% vs Conservative Republican 39%

2) Moderate Democrat 46% vs Conservative Republican 35%

3) Moderate Democrat 42% vs Moderate Republican 38%

4) Liberal Democrat 36% vs Moderate Republican 44%

www.rasmussenreports.com/2005/Election%202008%20Generic.htm

Finding 1) seems to blow my argument that a Liberal Democrat cannot beat a Conservative Republican out of the water, although a 1% lead is negligible

However, finding 2) does add weight to my argument that a Moderate Democrat could win 'big' over a Conservative Republican, while a Moderate Democrat would also will narrowly over a Moderate Republican. However, should it be a Liberal Democrat versus a Moderate Republican, then the Democrat is toast

It's early days, but at this stage adults seem to be warming to a moderate Democrats and I'll continue to argue that they'll have a better chance than liberals in winning the Oval Office

However, this is a generic poll and doesn't really allow for any particular candidates strengths and weaknesses and most of the changes took place among voters not registered with any party

Dave

Scotty has more Democrats than Republicans in his polls.

He made this change after 2000.

Do you think it would be better if pollsters were to 'normalise' to a 37% Democrat, 37% Republican and 26% Independent sample, unless this national exit poll data was somehow inaccurate? It might avoid any sample-bias

IT would seem that neither party has much of a numerical advantage in terms of party ID, so why do pollsters include significantly more Democrats than Republicans or vice versa?

Dave

[Democratic Hawk]

The idea of "swing" voters is overrated. Most people that have that much trouble picking one side or another just don't vote. There are a number of weak partisans, and they are indeed relevant.

I agree, weak partisans are relevant. Had 11% of self-identified Democrats not cast votes for Bush, then you might have just had a different president today? Kerry, in return, only received the support of 6% of Republicans

I'd guess that weak Democrats tended to be somewhat conservative and weak Republicans, somewhat liberal? But perhaps that's too simple an explanation. Significant numbers of liberals may have voted for Bush on the national security issue

The extent to which a party can retain the support of it's adherents can, indeed, be crucial to the outcome. Imagine if 11% of Republicans voted for Kerry and 6% of Democrats for Bush. If the Democrat candidate had been a southern or midwestern moderate, the outcome could have been markedly different

In 2004, Bush clearly appealed to Democrats more than Kerry appealed to Republicans

Dave

[© tweed]

Those poll results are very encouraging for Democrats.

It is May 2005.  Nothing matters until after the midterms.

[The Vorlon]

Do you think it would be better if pollsters were to 'normalise' to a 37% Democrat, 37% Republican and 26% Independent sample, unless this national exit poll data was somehow inaccurate? It might avoid any sample-bias

IT would seem that neither party has much of a numerical advantage in terms of party ID, so why do pollsters include significantly more Democrats than Republicans or vice versa?

Dave

Normalization to a specific party ID is a controversial idea, not entirely without merit, but still risky.

I am going from memory here, so I may be a % or two out in any given year, but as I recall the breakout of party ID over the last 8 elections has been as follows:

1992 Dems +3
1994 - Gop +3
1992 Dems +4
1998 - Dems +1
2000 Dems +4
2002 - GOP +2
2004 Even

So what is the "right" party ID to use?

The way campaign pollsters deal with the issue is to weight any given night's sample to the party ID in a larger, long term rolling sample. (TIPP Infometrica, which BTW was a very accurate  presidential poll in 2004 (if you go by projected margin of victory anyway) - they had Bush +2.1% versus +2.45% actual) uses this method.

How this works is fairly simple.

You would keep a rolling average of, for example, your last 14 nights of polling, and if the average of those 14 nights was say a 37/37/26 split you would weight the current sample to that value.

That way if you do get a blip sample, you can compensate by adjusting party ID, while still taking into account that over time party ID does indeed ship - not hugely - but it does shift. - if 2 years from now the average of your last 14 polls is Dems +3, you weight to Dems +3.

Party ID can and indeed does change, but it changes pretty slowly. 

In the last 8 actual elections it has been between Dems +4 and GOP +2, with the dems having a bit of an edge, at least untill recently.

Trying to guess what it will be in 2006 or 2008 is, well, simply a guess.

As a general rule, party ID should be something fairly close parity, by which I mean a few % either way. -

Another factor to consider is how had leaners are "pushed" for a party ID.

"About" a quarter of the population will on a first pass not declare a party ID, but if you slam 'em hard about 2/3rds will offer a weak party ID.


Party ID is far, far, far from a simple cut and dried matter in polling.

[CARLHAYDEN]

It's early days but Rasmussen did a survey over 8-9 June of 1000 adults and found the following results in four possible generic match-ups:

1) Liberal Democrat 40% vs Conservative Republican 39%

2) Moderate Democrat 46% vs Conservative Republican 35%

3) Moderate Democrat 42% vs Moderate Republican 38%

4) Liberal Democrat 36% vs Moderate Republican 44%

www.rasmussenreports.com/2005/Election%202008%20Generic.htm

Finding 1) seems to blow my argument that a Liberal Democrat cannot beat a Conservative Republican out of the water, although a 1% lead is negligible

However, finding 2) does add weight to my argument that a Moderate Democrat could win 'big' over a Conservative Republican, while a Moderate Democrat would also will narrowly over a Moderate Republican. However, should it be a Liberal Democrat versus a Moderate Republican, then the Democrat is toast

It's early days, but at this stage adults seem to be warming to a moderate Democrats and I'll continue to argue that they'll have a better chance than liberals in winning the Oval Office

However, this is a generic poll and doesn't really allow for any particular candidates strengths and weaknesses and most of the changes took place among voters not registered with any party

Dave

Scotty has more Democrats than Republicans in his polls.

He made this change after 2000.

Do you think it would be better if pollsters were to 'normalise' to a 37% Democrat, 37% Republican and 26% Independent sample, unless this national exit poll data was somehow inaccurate? It might avoid any sample-bias

IT would seem that neither party has much of a numerical advantage in terms of party ID, so why do pollsters include significantly more Democrats than Republicans or vice versa?

Dave

Well, from the 19cos to the 1990s, Democrats consistently outnumbered Republicans in the electorate.

As the Democrat party came increasingly under the sway of liberals, more and more conservative Democrats left the party which abandoned them.

Some pollsters haven't recognized this phenomena.

[CARLHAYDEN]

Do you think it would be better if pollsters were to 'normalise' to a 37% Democrat, 37% Republican and 26% Independent sample, unless this national exit poll data was somehow inaccurate? It might avoid any sample-bias

IT would seem that neither party has much of a numerical advantage in terms of party ID, so why do pollsters include significantly more Democrats than Republicans or vice versa?

Dave

Normalization to a specific party ID is a controversial idea, not entirely without merit, but still risky.

I am going from memory here, so I may be a % or two out in any given year, but as I recall the breakout of party ID over the last 8 elections has been as follows:

1992 Dems +3
1994 - Gop +3
1992 Dems +4
1998 - Dems +1
2000 Dems +4
2002 - GOP +2
2004 Even

So what is the "right" party ID to use?

The way campaign pollsters deal with the issue is to weight any given night's sample to the party ID in a larger, long term rolling sample. (TIPP Infometrica, which BTW was a very accurate  presidential poll in 2004 (if you go by projected margin of victory anyway) - they had Bush +2.1% versus +2.45% actual) uses this method.

How this works is fairly simple.

You would keep a rolling average of, for example, your last 14 nights of polling, and if the average of those 14 nights was say a 37/37/26 split you would weight the current sample to that value.

That way if you do get a blip sample, you can compensate by adjusting party ID, while still taking into account that over time party ID does indeed ship - not hugely - but it does shift. - if 2 years from now the average of your last 14 polls is Dems +3, you weight to Dems +3.

Party ID can and indeed does change, but it changes pretty slowly. 

In the last 8 actual elections it has been between Dems +4 and GOP +2, with the dems having a bit of an edge, at least untill recently.

Trying to guess what it will be in 2006 or 2008 is, well, simply a guess.

As a general rule, party ID should be something fairly close parity, by which I mean a few % either way. -

Another factor to consider is how had leaners are "pushed" for a party ID.

"About" a quarter of the population will on a first pass not declare a party ID, but if you slam 'em hard about 2/3rds will offer a weak party ID.


Party ID is far, far, far from a simple cut and dried matter in polling.

[CARLHAYDEN]

Do you think it would be better if pollsters were to 'normalise' to a 37% Democrat, 37% Republican and 26% Independent sample, unless this national exit poll data was somehow inaccurate? It might avoid any sample-bias

IT would seem that neither party has much of a numerical advantage in terms of party ID, so why do pollsters include significantly more Democrats than Republicans or vice versa?

Dave

Normalization to a specific party ID is a controversial idea, not entirely without merit, but still risky.

I am going from memory here, so I may be a % or two out in any given year, but as I recall the breakout of party ID over the last 8 elections has been as follows:

1992 Dems +3
1994 - Gop +3
1992 Dems +4
1998 - Dems +1
2000 Dems +4
2002 - GOP +2
2004 Even

So what is the "right" party ID to use?

The way campaign pollsters deal with the issue is to weight any given night's sample to the party ID in a larger, long term rolling sample. (TIPP Infometrica, which BTW was a very accurate  presidential poll in 2004 (if you go by projected margin of victory anyway) - they had Bush +2.1% versus +2.45% actual) uses this method.

How this works is fairly simple.

You would keep a rolling average of, for example, your last 14 nights of polling, and if the average of those 14 nights was say a 37/37/26 split you would weight the current sample to that value.

That way if you do get a blip sample, you can compensate by adjusting party ID, while still taking into account that over time party ID does indeed ship - not hugely - but it does shift. - if 2 years from now the average of your last 14 polls is Dems +3, you weight to Dems +3.

Party ID can and indeed does change, but it changes pretty slowly. 

In the last 8 actual elections it has been between Dems +4 and GOP +2, with the dems having a bit of an edge, at least untill recently.

Trying to guess what it will be in 2006 or 2008 is, well, simply a guess.

As a general rule, party ID should be something fairly close parity, by which I mean a few % either way. -

Another factor to consider is how had leaners are "pushed" for a party ID.

"About" a quarter of the population will on a first pass not declare a party ID, but if you slam 'em hard about 2/3rds will offer a weak party ID.


Party ID is far, far, far from a simple cut and dried matter in polling.

I agree.

I really depends on whether the turnout is skewed.

In 1974 Republican voters stayed home, and they got shellacked at the polls.

[Democratic Hawk]

Do you think it would be better if pollsters were to 'normalise' to a 37% Democrat, 37% Republican and 26% Independent sample, unless this national exit poll data was somehow inaccurate? It might avoid any sample-bias

IT would seem that neither party has much of a numerical advantage in terms of party ID, so why do pollsters include significantly more Democrats than Republicans or vice versa?

Dave

Normalization to a specific party ID is a controversial idea, not entirely without merit, but still risky.

I am going from memory here, so I may be a % or two out in any given year, but as I recall the breakout of party ID over the last 8 elections has been as follows:

1992 Dems +3
1994 - Gop +3
1992 Dems +4
1998 - Dems +1
2000 Dems +4
2002 - GOP +2
2004 Even

So what is the "right" party ID to use?

The way campaign pollsters deal with the issue is to weight any given night's sample to the party ID in a larger, long term rolling sample. (TIPP Infometrica, which BTW was a very accurate  presidential poll in 2004 (if you go by projected margin of victory anyway) - they had Bush +2.1% versus +2.45% actual) uses this method.

How this works is fairly simple.

You would keep a rolling average of, for example, your last 14 nights of polling, and if the average of those 14 nights was say a 37/37/26 split you would weight the current sample to that value.

That way if you do get a blip sample, you can compensate by adjusting party ID, while still taking into account that over time party ID does indeed ship - not hugely - but it does shift. - if 2 years from now the average of your last 14 polls is Dems +3, you weight to Dems +3.

Party ID can and indeed does change, but it changes pretty slowly. 

In the last 8 actual elections it has been between Dems +4 and GOP +2, with the dems having a bit of an edge, at least untill recently.

Trying to guess what it will be in 2006 or 2008 is, well, simply a guess.

As a general rule, party ID should be something fairly close parity, by which I mean a few % either way. -

Another factor to consider is how had leaners are "pushed" for a party ID.

"About" a quarter of the population will on a first pass not declare a party ID, but if you slam 'em hard about 2/3rds will offer a weak party ID.


Party ID is far, far, far from a simple cut and dried matter in polling.

Thanks, for clarifying the matter Vorlon

Dave

[Starbucks Union Thug HokeyPuck]

The 2/3 of the population that is not the Christian Right is starting to realize exactly what today's Republican Party is.  I have little doubt in my mind that the Democrats are going to win the next presidential race provided they run a half-decent canidate.  I'd say Hilary has too much baggage and she could be painted as a commy, so she's out, but just about anyone else could win.  John Edwards or Mark Warner would be great.