Some months ago, I started to look at polls with and without Gallup and Rasmussen (see Why I remove Gallup and Rasmussen from the Abbreviated Pundit Round-up charts). This is a feature built into Huffington Post's pollster.com, intended exactly for that purpose.
Without taking sides as to which is correct, it is instructive to see the difference between those two pollsters and everyone else (in addition to the references in my post above, see The Rasmussen Difference from Alan Abramowitz for further insights).
In any case, check this out.
This is everyone together, no exceptions:
This, however, is Rasmussen and Gallup alone:
And everyone else, without those two?
Now, I am not advocating pulling them out at this late stage of the game. Better to include them, and let the mix guide you. But knowing what's going on with and without helps put things in perspective, particularly when trying to match the tie national race with the predictions for an Obama electoral win, or talk of a popular vote/electoral vote split.
Simon Jackman has an elegant discussion of "house effect" here. Whether you use the mean of all polls as your compass pointing to "true reality", or performance in previous elections, even name brand pollsters can be wrong, and that's especially true when the race is as tight as this (polling is a blunt instrument to resolve a tie.)
The advice is the same as always... look for trends, use the aggregates, and recognize that a close race is a close race, no matter which aggregate site you use.
But for Pete's sake, don't declare Romney ahead because of momentum before you take a look at the above graphs.
Update: Looks like Andrew Sullivan had the same thought (see my first link for similar exercise last August). Love these chart tools! How could you not use them?
No comments:
Post a Comment