As a student of polling, I am always looking for interesting ways that poll data is being analyzed. Here are few of the places I look at when trying to make sense of the latest polls.
- Real Clear Politics
This site keeps an aggregation of national polls (and state polls). Its strength is that it is updated automatically and quickly. If Gallup releases a new national poll, it shows up on RCP withing minutes. Its weakness lies in that other than completely discounting some polls (they don't seem to include Zogby) RCP doesn't do anything with obvious anomalies. For example until recently every OH poll in the list was mildly McCain, but one poll had Obama +14. That one number skewed the entire average blue, but it was so far from the other poll results that you just had to question it.
- Hedgehog Report
The Hedgehog Report is maintained by David Wissing. Two elections ago, Wissing was an active blogger, commenting on many issues of that presidential campaign. Since then, he has stopped posting on much other than poll results. As a result, he has attracted a number of commenters that love to analyze poll results. As a warning, the commenters lean noticably to the right with a few obvious left "trolls" just trying to stir things up. But if you can get past the partisan bickering, there are often some insights to be gained about internals and trends.
FiveThirtyEight is a relatively new player, maintained by Nate Silver (it is named as such because there are 538 votes in the electoral college). Silver takes a hard statistical approach to analyzing poll results, to the point of performing Monte Carlo simulations. As part of my doctorate work involved Monte Carlo simulation, color me intrigued. He FAQ about his site is available here. Because of the depth of the analysis the site is only updated once a day. On the internet that is a lifetime, but it is the price you pay for having more than numbers from polling agencies just re-posted or aggregated for you.
If I have one concern about 538, it's that I don't know the political affiliation of Silver. His bio states he is a baseball fan (another plus!) and professionally has been involved with baseball statistics. But one has to be careful. Electoral-vote premiered on to the scene a few election cycles ago and made no obvious declaration of political bias. When the election was over, the creator of the site came forward and admitted he was a strong liberal and that defeating Bush was very important to him. I've doubted the accuracy of that site ever since. Such things have made me wary--I would like to know more about Silver's background and whether he is (unconciously or not) adding bias to his methodology.