Did Facebook Predict the Election?

[Pangloss]

I'd love to see some statistical analysis on this. But here's what the article says:

 

http://abcnews.go.com/Politics/facebook-twitter-predict-2010-midterm-election-results/story?id=12227898

 

In November's elections, the candidate who more people "liked" on Facebook won in 71 percent of Senate elections. Twitter was even more accurate, with the candidates with more followers winning in 74 percent of elections.

 

Furthermore, candidates with twice as many fans as their opponent won by at least 3.9 percent.

 

Social media, such as Facebook and Twitter, were often better predictors of election results than how much money a candidate raised and spent, according to Facebook. In 42 of the races Facebook analyzed, the winner had more "likes" but less money. /quote]

 

But popular results are what elections are all about, so what I'd really like to see is how well this correlates when taking a standard deviation (on what I don't know) into account. 71% may not be all that impressive given that most Facebook users have only one account and can't spam this the way they might attack an American Idol automated telephone poll -- I would have thought the correlation would have been higher. OTOH, a "like" is not the same thing as a vote.

 

Kinda raises an interesting question, though: What if Facebook had done polls on elected officials? Could they have been as accurate as scientific polls, or exit polls?

 

What do you all think?