I’m going to delve into a bit of poll wonkery here, so if this kind of thing doesn’t trip your trigger, go check your email or something.
There are a lot of polls I’ll not bother to write about.
For example, the Zogby Interactive polls are pretty much bullshit because the samples don’t come close to approximating a random sample of eligible voters. Likewise, I usually ignore polls conducted on behalf of candidates or a party. In that case, the polls may use perfectly fine methods conducted to professional standards. The problem is that the results may be released strategically—that is, released if the findings are favorable to a candidate or party and kept private otherwise. In other words, the poll itself isn’t representative.
Yesterday I came across a new type of poll from SurveyAnalytics . Here are the sampling methods described in the top-lines:
SurveyAnalytics conducted an online survey of 2,001 voters in King County, including the City of Seattle. All of these voters were determined to be likely to vote in the November 2011 General Election. SurveyAnalytics recruited respondents from a voter list purchased from Labels & Lists, which included citizens who had voted in 2, 3 or 4 of the most recent General Elections. Respondents were contacted via e-mail following an e-mail matching process also conducted by Labels & Lists. Voters completed the survey online using SurveyAnalytics’ CityFeedback platform.
Okay…the methods sound interesting, although I don’t know how the “email matching process” works. Older folks are likely underrepresented because many still don’t have email addresses or computers. But traditional land-line polls under-represent young voters, who are more likely to have only a cell phone for a telephone.
SurveyAnalytics compares their sample to a SurveyUSA poll of King County taken in 2009, and they find very similar results for sex and ethnicity, as well as crudely categorized education and income variables. But the SurveyAnalytics sample is, as expected, slightly younger compared to SurveyUSA’s poll. An alternative interpretation is that the SurveyUSA sample—based on robocalling land-lines—was too old!
I cannot vouch for the representativeness of this poll. I do find the methods intriguing. With that…here are some results from their survey of 2,001 King County likely voters (MOE 2.24%) taken from October 29 to November 2:
- Obama 52%, Romney 24%, Neither 13%
- When asked for names of people running for Governor, McKenna’s name was given by 68%, Inslee’s by 57%
- A head-to-head match-up gives McKenna 36%, Inslee 35%
- Cantwell is at 54% to the sum of all other Republicans at 35%
- I-1183 (liquor privatization): 61% yes, 33% no
- I-1125 (Eyman/Freedman anti-tolls/anti-transit): 50% no, 38% yes
- I-1163 (background checks, training on long term care workers): 60% yes, 25% no
Again, keep in mind that these results are for King County only.
Out of curiosity, I’ve compared the SurveyAnalytics poll to the “Puget Sound” sample from the recent Washington Poll. I’m not sure what “Puget Sound” is defined as in the Washington Poll, but keep in mind that the samples in the two polls are not strictly comparable.
Obama does marginally well in King County by this poll at 52%. But the total of all Republican candidate percentages is only 24 36%, with Romney at 24%. The Washington Poll has an Obama—Romney match-up giving Obama 55% to Romney’s 37% in “Puget Sound”. The difference may largely be the number of undecideds, perhaps reflecting how the question was asked.
The most interesting finding is that McKenna and Inslee are practically tied in King County. This isn’t as bad as it looks for Inslee, as McKenna had an extra 10% of people who could say he was running. The Washington Poll found Inslee leading McKenna in “Puget Sound” 45% to 40%. As I mentioned earlier, Inslee probably does better relative to McKenna as more of the undecideds decide. And to win, he’ll have to do much better in King County in November 2012….
The I-1183 findings are more favorable at 62% Yes, 33% No compared to Washington Poll’s 52%, 42% split for “Puget Sound”.
Initiative 1125 loses by a whopping 50% to 38% in King County; the Washington Poll’s Puget Sound sample rejected the initiative by a more modest 43% to 42%.
It will be interesting to compare the actual vote in King County for the initiatives this election to these poll results, if only to assess whether this particular internet-based polling method is any good. And whether future polls of this type are worth our attention.
