This is an interesting article on using Facebook to try to predict elections. Apparently in 2012 higher Facebook fan engagement was correlated with winning close elections, so political scientists are looking what this might mean for 2014. From Politico:
So far, we have trained our forecasting model on four key Senate elections in North Carolina, Alaska, Kentucky and Michigan. In those races, our two-party contest Facebook model shows Republicans potentially picking up one seat.
According to our Facebook forecast, in North Carolina, incumbent Sen. Kay Hagan, a Democrat, leads State House Speaker Thom Tillis, the leading Republican contender, by 52.7 to 47.3. In the open Michigan Senate race, Democratic Rep. Gary Peters leads the announced Republican candidate, former Secretary of State Terri Lynn Land, 55.6 to 44.4. And, assuming he survives a primary challenge, Kentucky Sen. Mitch McConnell holds a 53.9 to 46.1 lead over Alison Lundergan Grimes, the Democratic secretary of state. In Alaska, Democratic Sen. Mark Begich is in a dead heat whether he faces Joe Miller, the 2010 Republican Senate nominee, or Lt. Gov. Mead Treadwell.
It will be interesting to see if the results hold up but there are two points of caution. First their model is basically only built with data from 2012, so this correlation could just be coincidence. Second, how people are using the internet is changing so dramatically any forecasting model like this might be out of date very quickly. It is possible 2012 was only a brief period in time when Facebook engagement was an important indicator but the shift in social media demographic and use has changed. Just think how quickly MySpace dropped in relevance.