Number obsession
Ever since the Democratic primaries I've been getting more and more interested in polling numbers and statistics. To me, it's fascinating to follow this massive attempt to predict where the presidential race is heading. I've been getting familiar with terms such as margin of error, sample size, house effect, trend line adjustments, it's a world on it's own.
I've been following these numbers so closely that I have to keep reminding myself that there's only one poll that counts: the one on election day. Still, looking at numbers is suprisingly entertaining to me (and at times: nerve wrecking) .
There's so many pollsters out there at the moment that the best source for information on where the race stands are websites that weigh and average all the polls (poll of polls).
For raw information I suggest pollster.com:
pollster.comHere you'll find every single poll that comes out during the day. They even have an RSS feed for the hardcore number junkies.
For an accurate poll accumulation and analysis there's nothing better than fivethirtyeight:
fivethirtyeight.comHere they weigh polls based on their reliability in the past, adjust the data according to recent trends and what's most impressive to me: simulate the results 10,000 times based on the results of the projection to account for the uncertainty in the estimates. On top of that I also enjoy reading their background analysis.
One post I want to point out:
Tie Probability Continues to IncreaseThis has never crossed my mind until I saw it mentioned a couple of months ago: but a tie on election day is a very probable scenario. In case you don't know what happens should this occur: I have learned that members of the House of Representatives will choose the next president. Since it's almost certain that Democrats will have the majority of House delegations, this should be fun. On top of that, in 80% of the simulations where the result ends up in a tie the Republicans actually win the popular vote.
Can you imagine? The ultimate payback for 2000!