Email targeting in 2012 political campaigns | MIT Center for Civic Media
I am a senior at MIT in computer science with a concentration in science writing. I am currently the online media editor at The Tech and a collegiate correspondent for USA TODAY. I am interested in creating useful tools and platforms for journalists and news developers, interactives that tell a story, and discovering a way to disrupt the news industry.
Places you can find me:
Doing silly things to get myself on late night talk shows
Email targeting in 2012 political campaigns
An article from The Atlantic about the bearded geeks behind Obama's campaign recently caught the attention of many — as of today, more than 12,000 people have liked it on Facebook and over 3,000 people have tweeted about it. Part of the story's intrigue was the campaign's ability to collect massive amounts of data on voters and volunteers and use that data to target their messages.
This week's blog post will focus on ProPublica's Message Machine and its conclusions about email targeting in the 2012 election campaigns. This blog post is the second post in a series leading up to a final project about mass communication from 2012 election campaigns.
Message Machine is a project that aims to reverse engineer email targeting by political campaigns. It utilizes crowdsourcing to gather a panoply of campaign emails across a variety of demographics. Created by Jeff Larson and Al Shaw, news applications developers at ProPublica, Message Machine gathered more than 30,000 emails (nearly 3000 emails are distinct and available on the website) from over 600 recipients from the 2012 election cycle.
Message Machine tracked emails from Obama for America, Romney for President, Democratic Congressional Campaign Committee, Democratic National Committee, and Democratic Senatorial Campaign Committee. It also collected personal data from people submitting emails, such as age, salary, and education level. Using NLP and machine learning, ProPublica modeled the data and presented their results in an article posted on October 18.
Through the emails they received, they discovered the following (short summary of conclusions from ProPublica's results article):
1. The most common difference in single blast emails was the suggested amount to donate, which was based on the recipient's donation history
2. Voters in non-swing states were encouraged to travel to swing states to participate in get-out-the-vote operations.
3. Recipients between 18-29 tended to get more emails about contests.
Have any questions or suggestions for my final project? Let me know in the comments or email me at jskao at mit dot edu!