Sunday, May 04, 2008
Mark Halperin and bad polling
What I am going to talk about is reporting poll numbers from a bad pollster. What is a bad pollster? A pollster who is further off than the competition.... and often... is a bad pollster. The problem is that most bad pollsters are those of the news media. CBS/NYT comes to mind immediately... with Newsweek a close second. ABC News/Wash Post are decent pollsters.
The problem is that big "networks" can get their numbers out easily. More than that, political blogs will report them because "they are the big names". And of course, they are traditional and not IVR, so this is just another reason to report them.
So when CBS/NYT comes out with Obama over Clinton by 12 and both over McCain by double digits and Halperin posts it on the front page (http://thepage.time.com/), it worries me. In fact, CBS/NYT are the worst of the traditional pollsters... and only do Zogby Interactive (internet polling, wholly unreliable) and Columbus Dispatch (its some form of mail polling) do worse.
Take a look for yourself: http://farm4.static.flickr.com/3041/2453391209_6863309983_o.png
Not surprisingly, the national average have Obama up over McCain by a point and Clinton over McCain by 3. Further, Obama's lead over Clinton is about 3 points in the average.
Now, what does this mean? I'm personally very bothered by it. You got real people reading this blogs (especially Halperin who is one of the best in the business). People are given information that isn't the most accurate. And polling is already such a tricky business.
I'm not what you could a theoretical guy. I see what works... and use it.
CBS polling doesn't work. Don't use it. And if you do, don't put it up on your main page without any explanation.
Subscribe to Posts [Atom]