Why not subscribe?

Wednesday, October 24, 2007

Who would you never commission a poll from? Zogby!

Here's the poll story, followed by a comment. This was sent to me by the Richardson campaign.
=================================================================
Zogby Poll: Half Say They Would Never Vote for Hillary Clinton for President

Other top tier candidates in both parties win more acceptance;
Richardson & Huckabee favored most


While she is winning wide support in nationwide samples among Democrats in the race for their party's presidential nomination, half of likely voters nationwide said they would never vote for New York Sen. Hillary Clinton, a new Zogby Interactive poll shows.

The online survey of 9,718 likely voters nationwide showed that 50% said Clinton would never get their presidential vote. This is up from 46% who said they could never vote for Clinton in a Zogby International telephone survey conducted in early March. Older voters are most resistant to Clinton -- 59% of those age 65 and older said they would never vote for the New York senator, but she is much more acceptable to younger voters: 42% of those age 18-29 said they would never vote for Clinton for President.

Whom would you NEVER vote for for President of the U.S.? %
Clinton (D) 50%
Kucinich (D) 49%
Gravel (D) 47%
Paul (D) 47%
Brownback (R) 47%
Tancredo (R) 46%
McCain (R) 45%
Hunter (R) 44%
Giuliani (R) 43%
Romney (R) 42%
Edwards (D) 42%
Thompson (R) 41%
Dodd (D) 41%
Biden (D) 40%
Obama (D) 37%
Huckabee (R) 35%
Richardson (D) 34%
Not sure 4%

At the other end of the scale, Republican Mike Huckabee and Democrats Bill Richardson and Barack Obama faired best, as they were least objectionable to likely voters. Richardson was forever objectionable as President to 34%, while 35% said they could never vote for Huckabee and 37% said they would never cast a presidential ballot for Obama, the survey showed.

The Zogby Interactive poll, conducted Oct. 11-15, 2007, included 9,718 likely voters nationwide and carries a margin of error of /- 1.0 percentage point.

======================================================================

There's some quirk here. First of all, some of these candidates don't even have NAME recognition this high. Is Brownback male or female? Who's Hunter? Is Trancredo the guy who owes me money from high school? Is that Tommy Thompson, Fred Thompson, Jim Thompson (all Republicans of some prominence)? This name recognition factor may very well include Richardson.

Second, this follows the industry practice of stating "carries a margin of error of /- 1.0 percentage point". This, basically, is a crock and no person knowledgable in survey research believes such a statement. There's a margin of "sampling" error (probably assuming away nonresponse, maybe not) of 1%, but the nonsampling errors around a question like this are likely huge, so any notion that 50% would actually never vote for Hillary should be thrown out the window.

Third, there's the Richardson camp spin: "Richardson & Huckabee favored most". Since when does "least hated" translate into "favored most"? Are there no English majors volunteering for the Richardson campaign?

Fourth, there's no such thing as full disclosure. The above press release is on the Zogby site at http://www.zogby.com/news/readnews.dbm?ID=1376 . The end of the release notes "For a complete methodological statement on this survey, please visit:
http://www.zogby.com/methodology/readmeth.dbm?ID=1223 "

So what's the "complete methodological statement"?

"Zogby International commissioned an online survey of [9976 likely voters].

"A sampling of Zogby International's online panel, which is representative of the adult population of the US, was invited to participate. Slight weights were added [region, party, age, race, religion, gender] to more accurately reflect the population. The margin of error is +/- 1.0 percentage points. Margins of error are higher in sub-groups.

"Zogby International’s sampling and weighting procedures also have been validated through its political polling: more than 95% of the firm’s polls have come within 1% of actual election-day outcomes. (10/20/2007)"

This leaves you wondering what an "incomplete methodological statement" would look like.