Matthew Blackwell

With the advent of the Internet and super computing, political scientists are discovering novel ways to monitor the mood of the public through blog texts, Facebook postings, Twitter messages, and even the expressions on candidate photos posted to the web.

 

Super Computing Takes the Pulse of Politics

Matthew Blackwell

With the advent of the Internet and super computing, political scientists are discovering novel ways to monitor the mood of the public through blog texts, Facebook postings, Twitter messages, and even the expressions on candidate photos posted to the web.

Add to that list campaign contributions. Matthew Blackwell is developing methods to use contribution data, now available in online databases, to follow the pulse of campaigns. In particular, he has developed a novel way to identify the critical moments when a political campaign either takes off or falls flat.

“Campaigns rarely have smooth trajectories,” says the assistant professor of political science. “Instead, they tend to go through discrete phases punctuated by decisive turns, up or down.” For example, a convincing debate performance by a candidate can give the campaign a surge of support, while a gaffe may totally break it.

Traditionally, political observers have tracked such change points by relying on polls. The problem is that in smaller campaigns, like state legislative contests or less contentious congressional races, polls can be scarce. With campaign contributions, “we can take all of the data that the federal government collects on campaign contributions to get a sense of a race on a day-by-day level,” says Blackwell.

Using statistical analysis, including a Bayesian change point model and a Markov Chain Monte Carlo estimator, Blackwell is able to tease out change points amidst the normal ebb and flow of campaign contributions. Recently, he applied this approach to the campaign of Herman Cain during the 2012 Republican presidential primaries. He found that the change points predicted by his model corresponded almost to the day with major events in Cain’s campaign.

“You find that change points tend to happen when there’s a lot of news about a campaign,” Blackwell explains. “More people are paying attention to the candidate, and so the campaign gets more money. By looking through this data, we can get a sense for what causes these things.”

Blackwell is optimistic that rigorous analysis of data can clarify certain ideas in political science. Already he’s discovered something surprising: “There’s a feeling in political science that campaigns don’t matter a lot. People are supposed to base their decisions on things that no one can control, like the state of the economy. But the data reveals that voters very much do pay attention to the horse race.”

He predicts that political commentators will increasingly embrace data-driven methods. The 2012 presidential election, for example, proved the predictive power of data aggregators over traditional polls alone or pundit forecasts, he says. “There’s a feeling among a lot of people that politics can’t be predicted,” says Blackwell. “But I think that this is just a change in technology, and you’ll see a slow acceptance of these kinds of models.”

Technology change is redefining the practice of political science as well. For decades, researchers have relied on large, canonical databases that were compiled by numerous researchers and an army of assistants. “It’s become much easier for researchers to pull down completely original sets of data in a relatively short amount of time very cheaply and then analyze that on their own very quickly,” says Blackwell. “There’s been a very real increase of neat projects with cool new data.”

“What’s exciting about Big Data is that it’s leading people in a number of fields to realize they have similar problems,” he says. “The more data researchers collect, the more people look to different fields to see how others have solved similar problems. And the more this kind of iteration goes on, the more people start to come together.”

Blackwell is excited about the collaboration. “Even though disciplines use different terminology, researchers realize that there are fundamental issues that are similar across fields. Maybe there are creative solutions that we can find for problems in our own discipline from the approach others have developed.”