Robert Waldmann: The Gallup Anomaly

Robert's Stochastic Thoughts:

This year it has been widely noticed that the Gallup tracking poll looks much better for Romney than any other national poll except Rasmussen…. It's the talk of the web.  This makes 2012 just like 2008, 2004 and 2000. I posted on the return of the Gallup anomaly on May 11 2004….

[T]he Gallup likely voter filter… is irrelevant to the current Gallup anomaly, since Gallup is still reporting results for registered voters (and yet has a Romney house effect !).

The odd thing is that Gallup remains the most respected name in polling even after all these anomalies, while Rasmussen is considered a joke.

I think  that, this year, the explanation is pretty clear -- Gallup assumes low African American turnout based on 2010 when others base on 2008 or allow the sample to be the sample without weighting.  But I don't remember where I read that.  I just quote Marshall

Less clear is Gallup. They don’t have any obviously ideological bias. And they do use cell phones in their surveys. The issue seems to be giving more credence to a 2010 model of the electorate than a 2008 one.