Careful with those poll numbers … Rasmussen on O’Malley and Ehrlich
I’ll be talking about this at 1p on WYPR, but here’s a brief rundown on the new Rasmussen poll showing Bob Ehrlich and Martin O’Malley neck-in-neck (linked here):
- The poll shows O’Malley beating Ehrlich again, if the election were held today, 47-44.
- It shows strong approval for Ehrlich among men and for O’Malley among women.
- It also shows O’Malley with higher unfavorables than Ehrlich (no doubt a function of O’M being the incumbent.)
Some general critiques:
- Rasmussen only polled 500 likely voters. An excellent sample would be 1,000 likely voters and Gonzales Research, which has a long history in Maryland, typically does around 800 people.
So the margins are going to be bigger, making this either a lot tighter of race than the poll would have it appear, or a lot less close.
- Rasmussen’s robocall method of polling has drawn criticism because, among other things, it’s harder to judge the validity of your respondents than if you had a person asking the questions.
- All of that said, one national analyst, Nate Silver, has lauded Rasmussen’s national polling work.
- UPDATE: A reader took issue with the assessment that Nate Silver “lauded” Rasmussen and provided a link to this post. In the post, Silver lays out why Rasmussen’s polling methodology tends to favor Republicans. It’s a good read, tad esoteric, but worth the time as we are entering heavy-speculation and prognostication season.
I have to disagree with you on your characterization that Nate Silver “lauded” Rasmussen’s polling work. He has actually taken Ras to task for its methodology and its lack of transparency in its screening process.
http://www.fivethirtyeight.com/2010/04/use-of-likely-voter-model-does-not.html
Thee is a decided “house effect” in the Rasmussen polls. If we account for that, we can safely assume O’Malley is leading by much more.
IPFrehley
April 23, 2010 at 1:25 pm
Mr./Ms. Frehley, thanks for the reply and link. I’m updating the post shortly to include that analysis. Do you have a non-prank name, so I can credit you with the find in the updated post?
tomlobianco
April 26, 2010 at 10:11 am
Tom, PPP’s poll showed a 3 point lead for O’Malley with a robocall which is the same as Gonzalez’s poll out today. The issue is not so much if a call is live or automated, it it the sample they take from. 500 versus 1,000 is irrelevant in statistical purposes. But if you assume 30% African-American turnout over 25% over 20% that may greatly impact your results (22%=2006, 25%=2008). Or if you assume a great Dem then Rep or great women than male, etc. And that part of polling is a guessing game.
Thomas the Tank Engine
July 27, 2010 at 7:27 am