An article in the most recent issue of Political Research Quarterly by Juliet Carlisle and Robert Patton of Idaho State University looks at political activity on Facebook during the 2008 U.S. presidential election -- the "first Facebook election." Looking at the political discussions and politically themed status updates of over 1,000 University of California undergrads, they found that "Facebook during the 2008 presidential election was not the hotbed of political activity that popular accounts may have us believe. During both the primary and general elections, users were more likely to be nonactive politically than they were to be politically active."
I suspect this dataset is pretty outdated at this point. Social media campaigning was still in its relative infancy in 2008, compared to the 2012 cycle, when campaigns, media counsultants, and pressure groups discovered the awesome power of gifs, viral vidoes, and image macros. (The Biden memes alone probably eclipsed all campaigning from the previous election.)
I did find it interesting that the authors used Facebook data to test whether Robert Putnam's famous Bowling Alone thesis -- that the breakdown of American social structures and peer networks are causing a decline in civic engagement -- applies to Facebook friends. Apparently, on Facebook it's actually the lone bowlers who are more engaged:
In considering Putnam's model, we predict a positive and significant relationship between network size (num-ber of friends and group membership) and political par-ticipation. We find that the number of Facebook "friends" a user has is not significantly related to political participa-tion, contradicting our expectation on the effect of net-work size. Moreover, our results demonstrate the direction of the relationship to be opposite of our expectations so that those with more friends are less likely to be politi-cally active. This suggests that collecting friends or build-ing one's social network in Facebook is an independent activity undertaken by users who are less inclined to be politically engaged with that network.
They later note that this doesn't necessarily disprove Putnam, "there may be reason to believe there are both practical and conceptual differences between “Facebook” friends and traditional friends." I'd say that's a pretty safe assumption.