Tuesday, July 6, 2010

Daily Kos charges pollster fraud
An abject lesson in the realities of polling data

DAILY KOS, INTERNET—Liberal blogger and political commentator Markos Moulitsas Zúñiga is publicly calling out Research 2000, a polling firm that he contracted, claiming he was the victim of polling fraud. Moulitsas states that he will file a lawsuit against R2K—this will be a difficult legal standard to meet and given that R2K claims its now having to work out of a Kinko’s location, Moulitsas is unlikely to see any money.

Furthermore, John Cook of Yahoo News reports, "Del Ali, [R2K’s] president, has been sued numerous times in his home state of Maryland for nonpayment of debt and has been hit with several tax liens, according to court records."

Now on to the abject lesson—in layman’s terms, this is how a poll is generally conducted: the poll questions are engineered to be non-biased, and respondents are supposed to be equally weighted. In other words, equal or close to equal number of party affiliation and/or self-described political philosophy. For the most accurate results, respondents in electoral race polls should be LV’s or Likely Voters, meaning they have a party affiliation and vote in nearly every election. With more general polling such as direction of the country, A’s or RV’s are sampled (Adults and Registered Voters). General polls are also conducted over a large geographic area to ensure a solid cross-section.

If you wanted a poll to show a predetermined result, you simply skew the weighting (more republicans or more democrats) or confine it to an area that has a proven voting block for one particular philosophy and party. That’s what happened in the 2004 Bush/Kerry election with exit polls. They were conducted in heavy democrat districts and the results showed Kerry winning—but he didn’t and that’s the abject lesson in polling.

--Killswitch Politick

1 comment:

  1. Many technical errors above. The weighting is not supposed to be "equal" but rather to match the population percentages, as best as they are known. Equal party weighting in the US would slant polls to the right, equal "political philosophy" weighting would slant them to the left."Party affiliation" is not a required element to be an LV, since there are huge numbers of non-affiliated voters.