There seems to be a spate of recently released flawed surveys gaining media attention. Three methodologically deficient polls involve the Republican race for president, while another covers the Virginia US Senate campaign. All should be looked at with a skeptical eye.
Last week we covered the Wall Street Journal/NBC national poll that showed real estate magnate Donald Trump tied for second place with Mike Huckabee and only one point behind leader Mitt Romney. But this study is seriously inadequate. The survey sample included 1,000 adults without even screening for registered voters, and the self-identified number of Republicans answering the presidential preference question totaled only 238. This number is seriously below a proper national sample cell size, which should exceed 1,000 respondents.
Yesterday, CNN/Opinion Research released a new survey that places Trump and Huckabee tied for the lead with 19 percent apiece, followed by Sarah Palin at 12 percent, and Newt Gingrich and Romney closely trailing with an 11 percent tally. Though the sample size is 864 likely voters, the number of self-identified Republicans responding to the presidential ballot test that produced the aforementioned results was only 385 people. The methodology of this poll is better than the WSJ/NBC effort but is still not in the range of reliability for a national survey of likely Republican primary voters.
Also last week, Fox News released their national poll of 914 registered voters yielding similar results, though Trump and Gingrich did not show nearly as well. According to Fox, in data produced jointly by the Democratic firm of Anderson/Robbins Research and the Republicans’ Shaw & Company Research, Huckabee (15 percent) and Romney (14 percent) are virtually tied, with Palin following at 12 percent and Trump scoring 11 percent. Gingrich falls all the way to 7 percent. But here, too, the number of self-identified Republicans actually answering the questions is only 344.
The fact that all polling shows the race to be close with no clear leader and is verified by both methodologically sound and unsound studies does provide sufficient support for the supposition that the race is already extremely close, and that no candidate has a particular advantage. But, even if these conclusions prove to be spot-on, they are of little consequence. National polls for a political campaign decided by individual state contests are not very useful. The data helps to paint a simple story for the media to tell, but the findings are largely irrelevant in relation to the actual presidential horse race.
Roanoke College just released another questionable poll, but this one pertains to the Virginia Senate race. It shows former senator and governor George Allen (R) jumping out to a large lead over ex-Democratic National Committee chair and governor Tim Kaine (47-32 percent). This poll conflicts with every other piece of data showing the race to be in toss-up range. The results are even more curious when considering that Kaine just officially announced his candidacy, which normally provides a polling bump, not a decline.
The flaw in this poll concerns not so much the sample size (though 437 registered voters is low for a state the size of Virginia), but rather the length of the interview period, not screening for registered voters, and excluding respondents using cell phones.
The questions were asked from March 17-30, a 14-day period, when three days is usually considered the norm. Using only residents of Virginia who maintain land lines and not asking if they are registered voters badly under-represents the actual universe of people who will be casting votes in November of 2012.
Considering the aforementioned factors, the Roanoke College poll provides conclusions about the upcoming Senate campaign that are highly questionable and should not be considered as reliable information.
___________________________________________________
For further detailed insights, to sign up for my daily email updates, or to sign up to track specific issues or industries, please contact me at PR***@pe*******************.com.
