Oct 30 2006

Poll Bias In Turnout Modelling

Published by at 9:01 am under 2006 Elections,All General Discussions

I am surprised Michael Barone was not more forceful in his article today where he notes the polls we see are all way off from the make up of the country. Pew Research has been measuring the political affiliations of the country and it has seen about a 33-33-33 split (D-R-I) for the Bush years. In 2004 we saw the continued trend that was established post 9-11 with a population that was 31% Dem and 30% Rep. In 2005 there seemed to be a slight shift where the Dems were 34% and Reps 31%. So when Barone notes what is in reality a few point gap in affiliation showing up in polls as 10+ points, one has to wonder at the accuracy of the poll to measure anything but liberal wishful thinking:

In 2004, the electorate that went to the polls or voted absentee was, according to the adjusted NEP exit poll, 37 percent Democratic and 37 percent Republican. In party identification, it was the most Republican electorate since George Gallup conducted his first random sample poll in October 1935.

But most recent national polls show Democrats with an advantage in party identification in the vicinity of 5 percent to 12 percent. Party identification usually changes slowly. Historically, voters have switched from candidates of one party to candidates of the other more readily than they have changed their party identification.

Over time, big changes in party ID can and do occur. When I started in the polling business, in 1974, national party identification was almost 50 percent Democratic and not much more than 25 percent Republican.

Basically the polls are assuming a larger democrat turnout to support their giddy predictions. But the GOP has the better turn out machine in their Get Out The Vote (GOTV) process. Personally, I would report a poll in three numbers so as to be accurately telling the possible outcomes. But that would show the Dems and Reps that if the Reps simply go to the voting booths the Democrats would lose. If there is a 10% skew in a generic congress poll that shows a 12% lead for the dems, then one of the realistic variations would be to normalize that poll to show what would happen if the Reps came out in equal numbers. That lead would probably shrink to maybe 4% or less. And what would happen if the Reps really surged like they can and got more people to the polls? What would the result be then?

Pollsters should be honest and tell the people the viable range of results based on different voter turn out assumptions. I feel this is more important to the left who have had their hopes dashed numerous times by having pollsters pretend the nation is still a liberal nation (which it is not).

The conservative tide is just beginning as Barone notes in his piece. In 1974 the country was 50% democrat and 25% Republican. If the pendulum is indeed swinging then this 33%-33% equilibrium point is just the midway point towards a period of conservative dominance in politics. There is no reason not to believe, given the momentum since 1974 away from liberalism, that the pendulum is going to reverse course right now. Events can slow things down a bit (take Nixon for example), but stopping and reversing the trend is going to take a lot more than even a Nixonian scandal. This too must be factored into any assessment of turnout.

And Bush doesn’t have one of those Nixonian scandals going. Iraq is a hard thing to do, but it is not the scandal the Dems salivate over. Those of us who signed up to the effort knew that in advance. And those of us old enough (and honest enough with themselves) know this war has had the lowest casualty rate of any major military effort we have taken on – with the exception of the first Gulf War.

So if there is one thing the American people should demand it is a full reporting on the range of probable outcomes a poll tells us is possible. Because that is ‘informative’ instead of speculative or manipulative or biased. Those of us who use statistical modelling in a predictive manner know that all we are doing is reducing the range of possible outcomes. It is time the pollsters where honest about what they do too. And that is pretend they have some magical powers to determine voter turn out. The truth is, if the Reps go to the polls en masse and the Dems see the same lackluster turnout they saw in their primaries, the Dems will be crushed. Difficult? Yes. A scandal? No. Need to surrender? Not ever.

The most publicized race in the country was the CT primary. Massive numbers of voters came out and turned every poll on its head when Lamont and Leiberman split 51-48 in the Democrat primary. There is still no better indicator than the CT primary that the so called anti-Iraq War and anti-incumbent wave is just not big enough. If it was, then Lieberman would be on his way out, not on his way back in.

So what is the right model? No one knows so they should report the probable outcomes based on turnout that is equal, and turnout that favors each side. With these three numbers we the people can decide what is happening and what we will do about it.

6 responses so far

6 Responses to “Poll Bias In Turnout Modelling”

  1. archtop says:

    I think technology is also making the polls more unreliable, especially this year. Most organizations (as far as I know) still employ phone polls. But a large majority of people in the US have either caller ID on their land lines or have an ID system on their cell phones (I have both). If the caller is the infamous “unknown caller” or “out of area” I just don’t pick up, especially this time of year when everyone and their brother who’s running for office is trying to contact you. So, in my opinion, it is very difficult to get true statistically “random” sample upon which the poll’s accuracy lies. I chuckle to myself when read that a poll supposedly has a margin or error of +/- x% – How can they prove that about any particular poll? The percentage error relies almost entirely on obtaining a statistically significant “random” sample – and I think their “sampling” techniques leave a ** lot ** to be desired…

  2. Limerick says:


    Is there any ‘review’ of the polling firms as to their competence that we could access. With all the different polls the MSM quote all the time it would be nice to have something like a ‘handicap’ list of who’s who in the business and their track record. Any ideas of where you could point a poor uneducated red-neck pick-up driving Texan for a run down?

  3. First Cup 10.30.06…

    A mathematician is a machine for turning coffee into theorems. ~ Alfred Renyi


  4. az redneck says:

    Limerick: So many things have changed so rapidly, I don’t think that any comparison would be that predictive today.
    That said, the Rassmussen polls have been close for the past 2 elections. RealClearPolitics has a running update that is an average of all recent polls. When RCP prints articles by either Barone or Jay Cost, they are usually worth while.
    Electionprediction.com and CQPolitics.com are two other sites that use a variety of polls. Strategicvision.biz gives you the questions asked, which are always useful, though their coverage is not as complete.
    If any were spot on, we wouldn’t need elections!

  5. Race For Congress Tightens Up In Last Days…

    It looks like it will be a nail biter for both sides once again. While some pundits have already written post-mortems for the GOP and many Conservatives have all but given up on retaining control of the House, it looks to be a close race as they appr…

  6. Snapple says:

    Well, we can speculate, but the only poll that counts is the election.