Election Polling Made Simple – Look Deeper

Like an alley cat in heat, those Democrats and MSM polls are crying "Doom!' well past the point of credibility. But, since it's out there, we need to look at those numbers. Sharpen up those knives, it's dissectin' time!

OK, let's start by clearing out the trash. There are 33 seats in the Senate being decided, along with all 435 of the House Representatives. On the face of it, that makes it seem reasonable that control could change. But historically, better than 90% of incumbents get re-elected. It would take a truly significant set of conditions for the public as a whole to throw over control of Congress in the way that Democrats are predicting, conditions which the facts do not support. And the polls, if you look closely, are no great friend to the challengers.

I went through the Real Clear Politics' list of recent polls for the House and Senate, and here's what I saw:

For 33 Senate races, RCP has poll numbers for only 19 races. For 435 House races, RCP only shows poll numbers for 26 races. Now, I agree that it's possible that RCP is simply posting only the newest data, but I found it hard to find hard specific data on many of the predicted GOP losses. They are just assuming the results on no evidence.

Polling Report, for example, just focused on the generic poll. OK, I grant that there is some effect to a national preference, assuming for now that the polls are not skewed in their demographics or biased in their methodology. But even if we grant that, the notion that a specific Congressional District or State Senate race would ignore its own conditions and requisite politics in order to comply with the opinion from outside their borders is specious at best.

Also, one wonders about the sort of assumptions demanded of a reasonable observer, to accept a claim without something beneath the claim itself for support. Take the 22nd Congressional District in Texas, for instance. Tom Delay's resignation and subsequent dirty tricks by the Democrats certainly made it more difficult for a Republican to win against Nick Lampson, but the notion that voters who gave President Bush 64% of their vote in 2004 and whose Republican position has been unquestioned for a very long time would just hand over the district to the Democrats is laughable. I looked and looked for a poll on Lampson v. Sekula-Gibbs, but only found an old one from the start of September. But that one had Sekula-Gibbs ahead in double digits. Certainly, the need to write-in Sekula-Gibbs will make it tougher, but those people claiming that TX-22 is a sure thing for the Democrats, or even likely at this point, either don't know what they are talking about, they're liars, or both.

My point is, as a starting point for any projection we need to begin with the historical base, then advance conclusions only when there is substance to back them. And the baseline for this election, is that the Republicans hold the majority. It takes substance, not rumor, to change that condition.

Looking at those polls I could find in RCP, in the Senate only four races show polls with the incumbent trailing, and one of those is the Democrat Menendez in New Jersey. You can talk all you want about 'close' here and 'if' there, the fact right now is that the incumbent or incumbent party is leading in 15 of the 19 states where polls were posted. Sure, Chafee could lose Rhode Island, but where's the proof? In fact, from what I could find of the older stuff from September, Chafee was leading and if there is no evidence to the contrary, that's your scoreboard for now.*

In the House, RCP noted 26 races with polls, and the incumbent is leading in 10 of them, with another incumbent behind in one poll and ahead in another. That shows sixteen incumbents with polls showing them trailing, but only nine outside the statistical margin of error and only six where the challenger's lead is greater than the number of undecideds.. Someone tell me please how six seats equals Speaker Pelosi, when they need 15 or more? Yes, it could happen and I will be the first to scream that we all have to get out and vote, but where's the reason for the panic?

I also want to point out something which never seems to get mentioned. Intelligent people will remember that when a margin between two candidates is less than the margin of error, that's a statistical tie. But also, never forget that many of these polls, even as October winds down, still show a large number of undecideds. Dewine, according to the less-than-honest John Zogby, trails challenger Brown, 49% to 45%. Besides being darn close to the margin of error for a poll which queried only 750 people, the number of undecideds - six percent - means that this race is not at all clear. And yes, CBS News did broadcast a much larger margin - 14 points - but that one still only gave 49% to Brown, claiming only 35% for Dewine in a poll which allowed for a sixteen point undecided stake. And the difference between those two unfriendly-to-GOP polls also indicate that there is a lot more doubt in that race than you will hear on the air.

Nothing is over yet, unless you give up.


UPDATE: I went to look up some specific races which, for some reason, were not cited in the link from RCP I originally noted. Here's the detail when you don't just accept the headline:

Chafee: Trails by 5.6 in the RCP average, which accepts as valid polls from Zogby and Rhode Island College, but more to the point cites 13.4% "undecided".


Talent: Within statistical margin of error, with 8.4% "undecided"

Menendez: The link I originally cited had him down by 2, but in any case the average is well within the margin of error, and again cites 12.0% "undecideds"

Burns: None of the polls cited is even close to nominal threshholds for validity, plus again we see 9.3% "undecideds"

Anytime the "undecided" portion exceeds twice the margin of error cited for the poll, that poll is garbage. And traditionally, in Senate and Congressional elections, the undecideds break for the incumbent. Heavily so. I stand by my call.