IN-Mason-Dixon: Good news for McCain

[Tender Branson]

McCain - 48
Obama - 43

http://firstread.msnbc.msn.com/archive/2008/10/24/1588502.aspx

[Josh/Devilman88]

Today is a very good day for McCain.

[Tender Branson]

Today is a very good day for McCain.

Except Pencilvania (Muhlenberg Tracking).

[Rococo4]

ive never concerned myself with indiana.  my thing is that if we lose IN than nothing else even mattered really.

[Eraserhead]

I really think MD has been too pro-GOP this cycle.

Also, lol at the fact that we have to consider a 5 point lead for a Republican in Indiana "good news" for him.

[Sam Spade]

I really think MD has been too pro-GOP this cycle.

Also, lol at the fact that we have to consider a 5 point lead for a Republican in Indiana "good news".

A couple of days ago, I did my averages for SurveyUSA, Rasmussen and Mason-Dixon in 2006, utilizing the 16 Senate races and the 21 Governor races in which at least two of the companies did one poll at minimum one month before election day.

I will additionally do a list of all polls done by the companies within one month of election day, regardless of whether other polling companies polled.  That is for later.

As I have said before, the dynamics of 2006 were quite obvious afterwards.  Basically, on Election Day, Republicans and Democrats showed similar enthusiasm in the last day polls.  The fact that there was a 3% difference in party ID is probably due to slight ID shifts that occurred and the fact the GOP base wasn't as motivated as in 2004 (obvious).  The key point being is that there was no *surge* in enthusiasm or *drop* in enthusiasm in one side's supporters that was not detected in the pre-election polls.

However, undecided independents, as the numbers clearly showed, broke about 2-1 or 3-1 Dem on election day.  That skewed the polls to being slightly Republican-leaning overall.  My best guess is that this skewed the polling to being about 2% too Republican.  My figures below list the firms and their performance *without the skew* and *with the skew*.

The question, then, fundamentally, is whether any similar *skews* will exist in regards to the two dynamics set forth above.  TBD...

List
Without *skew*:  Mason-Dixon (28 races): R+2.51% (2 wrong winners, 8 outside MOE (MOE = 5%)).
With *skew*:  Mason-Dixon (28 races): R+0.51% (1 wrong winner, 4 outside MOE)
Without *skew*:  Rasmussen (35 races): R+1.65% (2 wrong winners, 8 outside MOE)
With *skew*:  Rasmussen (35 races): D+0.35% (1 wrong winner, 6 outside MOE)
Without *skew*:  Survey USA (30 races): R+2.46% (0 wrong winners, 13 outside MOE)
With *skew*:  Survey USA (30 races): R+0.46% (1 wrong winners, 8 outside MOE)

[classical liberal]

I really think MD has been too pro-GOP this cycle.

Also, lol at the fact that we have to consider a 5 point lead for a Republican in Indiana "good news".

A couple of days ago, I did my averages for SurveyUSA, Rasmussen and Mason-Dixon in 2006, utilizing the 16 Senate races and the 21 Governor races in which at least two of the companies did one poll at minimum one month before election day.

I will additionally do a list of all polls done by the companies within one month of election day, regardless of whether other polling companies polled.  That is for later.

As I have said before, the dynamics of 2006 were quite obvious afterwards.  Basically, on Election Day, Republicans and Democrats showed similar enthusiasm in the last day polls.  The fact that there was a 3% difference in party ID is probably due to slight ID shifts that occurred and the fact the GOP base wasn't as motivated as in 2004 (obvious).  The key point being is that there was no *surge* in enthusiasm or *drop* in enthusiasm in one side's supporters that was not detected in the pre-election polls.

However, undecided independents, as the numbers clearly showed, broke about 2-1 or 3-1 Dem on election day.  That skewed the polls to being slightly Republican-leaning overall.  My best guess is that this skewed the polling to being about 2% too Republican.  My figures below list the firms and their performance *without the skew* and *with the skew*.

The question, then, fundamentally, is whether any similar *skews* will exist in regards to the two dynamics set forth above.  TBD...

List
Without *skew*:  Mason-Dixon (28 races): R+2.51% (2 wrong winners, 8 outside MOE (MOE = 5%)).
With *skew*:  Mason-Dixon (28 races): R+0.51% (1 wrong winner, 4 outside MOE)
Without *skew*:  Rasmussen (35 races): R+1.65% (2 wrong winners, 8 outside MOE)
With *skew*:  Rasmussen (35 races): D+0.35% (1 wrong winner, 6 outside MOE)
Without *skew*:  Survey USA (30 races): R+2.46% (0 wrong winners, 13 outside MOE)
With *skew*:  Survey USA (30 races): R+0.46% (1 wrong winners, 8 outside MOE)

That is the most important point IMHO.  We always assume that undecideds will split evenly, but they never do.  I think that in midterm years there is generally a 3-1 anti-incumbent split, while in presidential years there is generally comparable pro-incumbent split.  Though that, of course, is a generalization.

[Eraserhead]

I really think MD has been too pro-GOP this cycle.

Also, lol at the fact that we have to consider a 5 point lead for a Republican in Indiana "good news".

A couple of days ago, I did my averages for SurveyUSA, Rasmussen and Mason-Dixon in 2006, utilizing the 16 Senate races and the 21 Governor races in which at least two of the companies did one poll at minimum one month before election day.

I will additionally do a list of all polls done by the companies within one month of election day, regardless of whether other polling companies polled.  That is for later.

As I have said before, the dynamics of 2006 were quite obvious afterwards.  Basically, on Election Day, Republicans and Democrats showed similar enthusiasm in the last day polls.  The fact that there was a 3% difference in party ID is probably due to slight ID shifts that occurred and the fact the GOP base wasn't as motivated as in 2004 (obvious).  The key point being is that there was no *surge* in enthusiasm or *drop* in enthusiasm in one side's supporters that was not detected in the pre-election polls.

However, undecided independents, as the numbers clearly showed, broke about 2-1 or 3-1 Dem on election day.  That skewed the polls to being slightly Republican-leaning overall.  My best guess is that this skewed the polling to being about 2% too Republican.  My figures below list the firms and their performance *without the skew* and *with the skew*.

The question, then, fundamentally, is whether any similar *skews* will exist in regards to the two dynamics set forth above.  TBD...

List
Without *skew*:  Mason-Dixon (28 races): R+2.51% (2 wrong winners, 8 outside MOE (MOE = 5%)).
With *skew*:  Mason-Dixon (28 races): R+0.51% (1 wrong winner, 4 outside MOE)
Without *skew*:  Rasmussen (35 races): R+1.65% (2 wrong winners, 8 outside MOE)
With *skew*:  Rasmussen (35 races): D+0.35% (1 wrong winner, 6 outside MOE)
Without *skew*:  Survey USA (30 races): R+2.46% (0 wrong winners, 13 outside MOE)
With *skew*:  Survey USA (30 races): R+0.46% (1 wrong winners, 8 outside MOE)

Interesting.

I do seem to recall MD screwing up the Corker-Ford race pretty badly.

[Sam Spade]

Interesting.

I do seem to recall MD screwing up the Corker-Ford race pretty badly.

Thought you might like it.  Every one of the three major state polling firms (Strategic Vision and Research 2000 didn't poll enough to make it) had screwups.  M-D and Ras were more stable, with two wrong winners, though overall Ras was better in the end result without the skew.  SUSA had no wrong winners, but its results were highly erratic.

I didn't do Zogby you'll be happy to note.

[classical liberal]

Interesting.

I do seem to recall MD screwing up the Corker-Ford race pretty badly.

Thought you might like it.  Every one of the three major state polling firms (Strategic Vision and Research 2000 didn't poll enough to make it) had screwups.  M-D and Ras were more stable, with two wrong winners, though overall Ras was better in the end result without the skew.  SUSA had no wrong winners, but its results were highly erratic.

I didn't do Zogby you'll be happy to note.

Assuming a normal distribution:

Since M-D was outside the reported MoE 1/7 of the time, their actual confidence interval would probably be better assessed as 6/7, not 19/20.  To get their MoE to the 19/20 level I would expect that we should multiply their stated MoE by 1.33 to get a more accurate read on their error.

With Rassmussen, the same process gives a factor of 1.43, and with SUSA it gives 1.765.

Applying those penalties to the different firms and then taking least squares averages from each is a good methodology for compiling an expected baseline level of support for each candidate.

[TheresNoMoney]

I don't see Obama winning Indiana, I think he has a better chance to win Geogia.