Wednesday, November 7, 2012

Data Mining and Statistical Models Changing the Face of Campaigns

Recently a friend turned me to a New York Times blog, Five Thirty Eight, where Nick Silver has expanded his statistical model from 2008 that he used to correctly predict the electoral college in 49 of the 50 states.  The actual specifics of his method are beyond my statistical capacity but the fact that he was able to predict ALL 50 states in this election leads me to believe he must be doing something right. I have read several articles today that have not only discussed Nick Silver's success in depth, but that have gone on to say modeling of this type will be commonplace for the next election.

Barack Obama's campaign seems to have gotten this hint, as their data mining operation was five times larger than in 2008.  The Time Magazine article linked above details the complexity of the system and how it was not just about the political beliefs and tendencies of the electorate but more about who they need to target and what makes those people inclined to support.  I found it particularly interesting that they found Sarah Jessica Parker would be the best person to host a dinner event with the President and a contest winner based on the success of their event on the west coast with George Clooney.

The complexities of their data mining system goes beyond what I would have ever thought a campaign could coordinate, which is reflected in their impressive fundraising.  They were able to integrate the data from their campaign offices, grass roots organizers, polls and a number of other sources to target specific groups with specific language.  I remember getting annoyed with all the emails I was getting every day for a while, turns out they were testing language on me so they could use it for their fundraising efforts.

Their use of facebook was also quite surprising to me, allowing an app to access personal information and prompting individuals to ask their friends to take action really struck home with me.  It reminded me of the case study we read earlier in the semester with the online communities becoming offline actions and lead me to wonder if this really will be the future of elections.

I wonder if in 2016 instead of getting emails and phone calls to vote if I will instead receive tweets and personalized requests from my friends?  I think that it will be really interesting to see how these complex statistical prediction models and data mining efforts will reshape the landscape of elections in the near future and how it changes candidate interactions with the public.

No comments:

Post a Comment