Nov 30, 2012

How data Mining helped Obama Campaign

Time published an article right after election on How Obama campaign used data mining to win the race.  In the article, the contributions of data mining to Obama campaign are listed as:
  1. Identify top donor profile ( i.e., single women between 40 and 49)  This leads to a fund raising dinner with George Clooney on the west coast, generating $15 million in one night.  
  3. Identify people who mostly like give online. This generates more targeted email campaign that goes out to these specific group of voters. 
  5. Identify voters who would respond to persuasion (Voter profiles include age, sex, race, neighborhood, voting record and consumer record) This allows the campaign to make more targeted calls.

The techniques used here seem like the classical data mining method: A machine learning-based classifier. More specifically, it is supervised learning. Given that Obama campaign accumulated all data from 2004 campaign, it is very easy to train a model from previous data that include who donated and who did not donate. The top donor profile can then be extracted.
Romney may not be so lucky as he did not run a presidential campaign 4 years ago. This probably applies to future 2008 presidential candidates. In case Hillary Clinton runs for president, she may have some advantage by using her 2004 campaign data (assuming she still keeps those data). Of course, she can use the data donated from Obama campaign, assuming the voter profile does not change.
As we can see, data mining and data collection become a big commodity. They bring more money for a political candidate and they generate turnout. Winning a big political battle is now intimately linked to technology.

(A side note, this article also mentioned, like every company that starts to work on data mining, consolidating databases and make data available for data mining is first step.)