9/20/2012

Polls and climate models

Via-Jer's Notes

I am 95% sure with a margin of error of (+/-3.25%) that the polling for this election is skewed badly. Whether this is due to some leftist institutional bias or just some structural defect that permeates the polling community I'll let you judge, but regardless the polls are, to put it bluntly, crap.

To understand how bad this defect is you must first understand that polls are not really scientific in much the same way that climate models are not scientific. Like the climate models that forecast increased temperatures as the result of increased CO2 in the atmosphere they are only as good as the suppositions fed into them by the modelers. If the suppositions fed into the formulas are wrong, for whatever reason, then what comes out can not possibly be correct. The old garbage in garbage out scenario.

Now I have no way of knowing where the garbage going into the various polls is coming from, though I do have my suspicions, but it is obvious that what is coming out is garbage.

The supposition that pollsters make which like climate modelers may be based upon some other scientific formulation or criteria is what they project the turn out on election day to be. In essence the pollster is projecting who is going to show up at the polls and then fitting the information of their polls into that projection. For example when you see or hear a poll referred to as a D+5, that simply means that the pollster used 5% more Democrat

Let's take the three battleground state polls released by FOX yesterday as examples Ohio, Virginia and Florida. The first thing you must know is that in both Ohio and Virginia the voter does not register by party affiliation so there is no way of knowing at any given time the make up of the electorate. In Florida we do register by part or no party affiliation so that information is readily available. However party registration is not really a precise indicator of turn out at elections if it were in Florida as an example all the state wide offices and both US Senators would be Democrat since there are just under 4% more registered Democrats than Republicans in Florida. Despite this advantage all the state offices, both houses of the legislature, the vast majority of US House Representatives (19-6)and one US Senator are all Republican.

In most elections in virtually all states regardless of the party registration, the Republican voters tend to back their candidate with more loyalty than do Democrats. Even with the extremely motivated Democratic base Barack Obama only received 89% of the Democratic vote in 2008. Lest you think that was the result of racism that is the same percentage of Democratic vote that Kerry received in 2004 and two points higher than Gore received in 2000. John McCain on the other hand in keeping with a typical Republican loyalty trend irrespective of the fact that it was a very uninspired Republican base received 93% of the Republican vote.

These historical details are not unimportant when looking at polls and it is not like pollsters don't know this, which makes the polling done this cycle all the more egregious which brings me back to the FOX battleground state polls. Here is the important figures from the polls

OHIO

Obama +5 (49-44)
Turn out model D+6 (42/37/20)
Party Vote
(D)92-3 (R)89-6 (I)48-51

In the 2010 mid term the parties turned out like this 36% (D) /37%/ (R) /28% (I). In 2008 the turn out was 39 (D)/ 31 (R)/ 30 (I)

Putting aside the oversampling of Democrats which reflect a turn out far closer to 2008 than 2010 the next thing you will notice is that according to this poll 92% of Democrats are going to vote for Obama and only 89% of Republicans are voting for Romney while 5% of both parties are either undecided or are voting for someone else. Really?. Does anyone believe that Romney has less support among Republicans than Obama has among Democrats? I guess that according to this pollster at least, the Tea Party is dead. So Obama not only is going to turn out 6% more Democrats than Romney he is going to win his base by 3% more than Romney while loosing the Independents, Right! Oh and by the way the Governor and both houses of the state legislature in Ohio are Republican, boy those Ohioans must sure be fickle.

VIRGINIA

Obama +7 (50-43)
Turn out modelD+5 (38/33/26)
Party Vote
(D)94-3 (R) 87-5 (I) 42-41

There were no exit polls in Virginia in 2010 since there were no state wide contest so we do not know the most recent party turn out results. In 2008 it was 39(D)/33(R)27(I). So according to this FOX poll the turn out in 2012 is going to be about the same as 2008, Do you believe that?

Here again we see that according to this poll , Obama's base is going to support him 7 points higher than Republicans are going to support Romney. Come on what universe are these pollsters living in. And despite all historical precedent to the contrary, like in Ohio, twice as many Republicans are going to vote for Obama than Democrats are going to vote for Romney. But even this does not capture the absolute absurdity of this poll. According to this poll not only is Obama slightly winning the Independents in Virginia, somehow, someway there are 17% of Independents not voting for either candidate. According to this poll 10% of Independents are undecided 3% are voting for Virgil Goode and 4% just sort of disappeared. 99% of Ohio's Independents are decided but 10% or is it 14%(?) of Virginia's are not? Not to mention the 5% of Republicans that can not seem to make up their mind for these pollsters. Can anyone say Shenanigans?

FLORIDA

Obama +5 (49-44)
Turn out model D+5 (42/37/20)
Party Vote
(D) 88-7 (R) 87-9 (I) 45-43

In 2010 the exit polls for Florida were (D) 36 (R) 36 (I) 29 in 2008 it was (D) 37 (R) 34 (I) 29. So not only is this pollster oversampling Democrats 5 points higher than the most recent election they are oversampling them 2 points higher than 2008! But that is not the worst of it, they are saying a higher percentage of Democrats will turn out than there are registered Democrats in Florida just under 40 % of the electorate.

Here too we see that Obama is going to do better among Republicans than Romney will do among Democrats, In 2008 12% of Republicans did vote for Obama in Florida but 12% of  Democrats voted for McCain. We also see that in this poll that a great many Independents (12%) are either undecided (9%) or voting for someone else (3%).

Historically the undecided vote breaks overwhelmingly for the challenger, there is no reason to believe that will not happen in this election particularly with the Republican undecided. According to this poll  of likely voters in Ohio 2% of Republicans would vote for someone else, 2% are undecided and 1% would not vote. In Virginia 3% vote for someone else and 5% are undecided. In Florida 2% will vote for someone else and 2% of Republicans are undecided.

These polls besides not reflecting the reality of the recent trends in the electorate are using turn out ratios that are simply trash, garbage in garbage out.

1 comment:

  1. oh come on...
    ...we all know that liberals don't poll

    they are scared to death of those radical Seventh-day Adventis and Presbyterians

    ReplyDelete