November 26, 2007

Ignore This Poll!

Hatched by Dafydd

Don't pay any attention to this poll showing Sen. Hillary Clinton (D-Carbetbag, 95%) running behind all of her major Republican opponents in the presidential race; as enjoyable as all the hype may be, the source -- Zogby Interactive -- is completely unreliable.

Sorry to burst the bubble, but John Zogby's "interactive" -- that is, online -- polling is execrable. It's untrustworthy when it goes against Republicans; and it's equally untrustworthy when it cuts in our favor.

Here is the key graf from the Breitbart story, and the only thing you need know about the poll:

The Zogby poll was conducted online among 9,150 likely voters across the United States between November 21 and 26, and carried a margin of error of plus or minus 1 percentage point.

So yeah... among Americans who spend significant time online and are willing to answer polling questions at a web site -- Hillary now runs behind. But besides looking at an incredibly volatile group, it's also very small and highly unrepresentative of the voting population as a whole.

The Zogby online poll is very much like the infamous 1948 Chicago Daily Tribune poll during President Harry Truman's reelection campaign; the Trib conducted the poll by telephone, and it ended with Truman, newly reelected, holding up the election-day plus one edition of the Tribune with the banner headline "Dewey Defeats Truman."

And as Isaac Asimov pointed out, the poll was perfectly accurate: If the election were limited to only those people who owned telephones in 1948, then ultra-liberal Republican Gov. Thomas Dewey of New York would have won.

In the Zogby case, we do actually allow those who don't camp online -- in fact, even those who don't have computers at all! -- to vote for president; the poll is wildly unrepresentative... no matter who an individual instance of it supports.

Look, I do believe that current polling has an inherent partisan Democrat bias: They poll over the weekend, which favors Democrats; they overpoll big cities, which favors Democrats; and just in general, Democrats tend to be more willing to sit still for a call from a political pollster... Republicans are much more likely to hang up. Pollsters could fix this bias by simply asking party affiliation, comparing the percentages to the percentages in the districts in which they poll, and then weighting the responses accordingly.

But they refuse. Pollsters claim that respondents answer the party registration question not according to how they are actually registered, but according to how they feel about their party's candidates that day instead. In other words, suppose the Gallup poll calls a person who is registered as a Republican; if he likes the Democratic candidate better that day, he'll sail under false colors and claim that he's a registered Democrat. Ergo, when they get a big imbalance in favor of Democrats, that just means lots of Republicans like the Democrats better, and they're lying about their own registration to jump on the Democratic bandwagon.

I find this argument risible. For one thing, the imbalance exists even when the Republican wins the race. Do pollsters really want us to believe that Republican respondents are so down on the GOP that they falsely claim to be registered Democrats -- and then go ahead and vote Republican?

The more plausible explanation is that the heuristic that pollsters use to select samples of respondents tends to skew Democratic; therefore, they should weight their samples to match the partisan breakdown of the actual population of voters, according to previous votes. Picture this silly example: If Gallup polled exclusively at singles events, would that be a reasonable sample base from which to project an electoral winner? Of course not: Singles are much more Democratic than married couples. Alternatively, if they polled only at churches, that would be equally skewed towards the right.

It's not an exaggeration to say that the sample "makes" the poll. When the sample differs significantly from the population -- such as when it consists exclusively of people who own computers, spend much time online, and are willing to take the time to answer an online poll -- it loses all predictive value.

I do believe that Rudy Giuliani, Mitt Romney, Mike Huckabee, and John McCain would actually be running ahead of Hillary Rodham Clinton Rodham if the poll samples were more accurate. Alas, nothing by John Zogby can be counted as evidence of such.

But keep watching the polls; I suspect they'll start to change as the primary process kicks off. And in particular, when the two nominees are known, that's when we'll really start seeing some movement.

Hatched by Dafydd on this day, November 26, 2007, at the time of 2:50 PM

Trackback Pings

TrackBack URL for this hissing:


The following hissed in response by: snochasr

I am reminded of the old idea about watching only the last two minutes of a football game because that's when all the action takes place. I wish we could conduct our presidential politics that way.

The above hissed in response by: snochasr [TypeKey Profile Page] at November 26, 2007 4:23 PM

The following hissed in response by: Terrye

Didn't this poll also have Edwards and Obama beating the Republicans? Oh well, I never trust a Zogby poll anyway.

The above hissed in response by: Terrye [TypeKey Profile Page] at November 26, 2007 4:55 PM

The following hissed in response by: Geoman

I think people lie about party affiliation all the time, Republicans more so than Democrats. Why? Republicans get demonized by the media, so when a stranger asks you what party you belong to, what should you say? The easiest answer is to say Independent, which, if you are a Republican, is probably reasonably accurate (i.e. you are intelligent, logical, open minded, results oriented, etc.). Democrats are much more dogmatic than most Republicans I have met, which is odd, because the media portrays just the opposite.

Another factor: Before the vote a citizen may not necessarily be very serious, making a flirtation with Democrats more likely. In the voting booth, faced with actually pulling the lever for the Democratic demagogue of choice, voters have second thoughts. This is like sitting in your living room and fantasizing about skydiving, but not actually being able to jump out of the plane.

Online polls are not worth the paper they are printed on.

The above hissed in response by: Geoman [TypeKey Profile Page] at November 27, 2007 11:07 AM

The following hissed in response by: hunter

I think the polls, no matter the bias, are unreliable and useless right now. It is too far in advance of the election.

The above hissed in response by: hunter [TypeKey Profile Page] at November 27, 2007 1:13 PM

Post a comment

Thanks for hissing in, . Now you can slither in with a comment, o wise. (sign out)

(If you haven't hissed a comment here before, you may need to be approved by the site owner before your comment will appear. Until then, it won't appear on the entry. Hang loose; don't shed your skin!)

Remember me unto the end of days?

© 2005-2009 by Dafydd ab Hugh - All Rights Reserved