Oct 29 2012

The Problem With Polls

Published by at 12:38 pm under All General Discussions

You know, this new poll math is pretty silly. Apparently 2+2 = +1 Obama (every time). As Ed Morrissey points out, if the turn out/sample models were aligned to some semblance of reality then it probably looks like a Romney win.

If Romney wins independents by 15, and the gender gap by 2, how can he only be leading by one overall? Simple — the Post/ABC sample has a declining number of Republicans in its sample.  Today’s rolling three-day average has a D/R/I of 35/28/34, which would put Republicans at seven points below their 2010 turnout and five points below their 2008 turnout.  Two days ago, the sample was D+4 at 34/30, but now it’s D+7.

Look at those numbers for a second and just marvel at the lack of reality. In an anti-Obama year, when the GOP has the opportunity to take the House, Senate and President – especially after 4 years of economic hell which typically turns out incumbents – the least likely to turn out are GOP voters (28%)? Independents and Democrats will be the most engaged and energized this year?

Not in this reality.

I would say left-of-center Democrats are disillusioned and either will sit home or defect to Romney. Independents who have awoken from their 2008 Kool Aid are either going to sit out or switch to Romney. Only the rabid far left (making up 20-25% of the electorate) is panicked enough to be energized to try and save this failed administration.

60% of the electorate feels the country is off track, which means they don’t want 4 more years. The reality is Obama has little chance to pull out of this tail spin. I am just curious how badly he will lose, by a little or a lot?

So let’s do some recalculations. Yesterday Rasmussen listed the candidate support by party ID:

  • Obama takes 86% of Democrats (leaving up to 14% to go to Romney)
  • Romney takes 90% of Republicans (leavinup to g 10% to go to Obama)
  • Romney led Independents by 11% (so we could assume a 54-43% split)

If we use these allocations, and assume no one goes to another candidate, we can compute the maximum number (the ceiling) each candidate can obtain under various turn out models.

If we assume a D/R/I of 33-33-33 (all equal with no edge to any group) the results would be 52% Romney and 45% Obama (which interestingly is very close to the result from the latest Battleground Poll of 52-47%).

If we use the ABC News D/R/I (35-38-34) and the lead with independents (55-40) that Morrissey was analyzing, the results is 49% Romney to 46% Obama! For Obama to get a point ahead of Romney (48-49) in this poll means some amazing statistical gymnastics. Note that I am computing independents going 55-40 Romney, but I did not change the 86% of Dems for Obama or 90% of GOP for Romney. Honesty, I could not get to an Obama 48-49 lead with the independent numbers quoted (55-40). No reasonable combination of party base win-loss gets me there, so whatever they did is pretty wild.

Update: Was being interrupted over and over while writing this post, so I somehow flipped who had the lead (it is Romney +1, not Obama). I still cannot get the numbers to come out right for the ABC News poll assuming a range of party base win-loss scenarios while holding to the 15% lead with independents. So I still maintain Romney is more likely to be way ahead than tied. Anyway, may try and clean up the rest of the post if I have time. – end update

Let’s do another experiment. Let’s use Rasmussen’s party base numbers, but calculate the result using Rasmussen’s party ID for September 2012 (36.8-34.2-29). When we compute the result Romney would win 51-46%. Again, very close to the latest Battleground Poll prediction.

And if we use Gallup’s D/R/I for 2012 (37-39-24) then the result is 51-49% Romney.

Honestly, there is no realistic computation that comes out with Obama ahead. And many come out showing a landslide for Romney (you don’t win by 5+% and have a close electoral college result). I am still very bullish on this election, bizarre polls aside. The Obama administration and its allies keep putting out propaganda that the race is closer than it probably is. Basically, a high Democrat turn out does not equate to Obama win, especially if the turn out is from disaffected center-left Dems voting for Romney.

I think this is the year the pollsters are going to be exposed as being much less reliable than everyone assumed.


5 responses so far

5 Responses to “The Problem With Polls”

  1. Polling is one thing, campaigning is another.

    One of the commentors over on the Polipundit blog just noted the following campaigning pattern, which is interesting in light of Michael Barone’s recent “White voter preference” column —

    “Biden goes to states where Obama has grown in unpopularity to the extent that Obama won’t go.

    Observing the pattern;

    – Biden campaigned in FL – polls shifted to Romney.
    – Biden campaigned in NC – polls shifted to Romney.
    – Biden campaigned in VA – polls shifted to Romney.
    – Biden campaigned in CO – polls shifted to Romney.

    Obama only goes where he can be assured of a non-embarrassing crowd such as a university or inner city park.

    And he doesn’t go where he is underwater. He sends Biden.”

    This commentor thinks based on the above —

    “PA is the next to go Romney. MI will soon follow.”

    We shall see how that turns out, but it does seem to fit the available campaigning annoucements, including a VP Biden campaign visit to Scranton, PA on Nov 01, 2012.

  2. oneal lane says:

    Currently, I just don’t believe much of anything I hear regarding this race, good or bad.

    There is too much wishful thinking going on on both sides and too much corruption to say “what is supposed to happen will happen.”

    Ths system is “out of order.”

  3. Stirner says:


    “Dave in Fla” over at http://difpolls.blogspot.com/ has been doing similar analyses to the polling during the last few months. He does a great job of teasing out the surreal internals of many of the published polls and he tweaks the D/R/I weighting to reflect several different election day outcomes. Definitely worth checking out. (One of his first posts lays out his assumptions and methodology.

    Bottom line: Romney/Ryan has this in the bag. The main question at this point is how far they can run up the score and flip Senate races and bring of a decimation of Democrats down-ticket along the lines of 2010.

  4. dhunter says:

    R 33 D33 I 33

    R x 90% D x 90% Indy R x 60 % D x 40%

    R 90% + 60% = 150%
    D 90% + 40% = 130%

    R 150 – D 130 = Obama + 2%

    Math is HARD
    The Presstitutes!