The University of Michigan study, among the first to examine the Tea Party's social media strategies, also showed that analyzing Twitter activity can lead to good predictions of election winners.

Various social media tools have become a key part of campaign strategies in recent years. In 2010, nearly a quarter of online adults used social networks including Twitter to engage with the election.

In this study, researchers from the U-M School of Information and the College of Engineering looked at more than 460,000 tweets -- three years' worth from 687 candidates running for national House, Senate and gubernatorial seats.

"The conservative candidates -- Republicans and Tea Party members -- definitely used Twitter more visibly and showed a more coherent set of messages and topics," said Eytan Adar, assistant professor in the School of Information and the Department of Electrical Engineering and Computer Science. "They also followed each other much more closely. I think it's fair to say they were much more cohesive in a lot of ways and at the end of the day that makes for a stronger campaign."

Conservatives, who made major gains in the 2010 midterm elections, tweeted about similar topics and conveyed a coherent message with a particular attention to economic issues, the researchers found. The top terms in Republicans' posts were "spending," "bills," "budget," "WSJ" (Wall Street Journal), "Bush" and "deficit." Over the study period, Republicans tweeted an average of 723 times.

With an average of 551 tweets (text entries) during the study period, Democrats posted less frequently. Their tweets covered a wider range of topics. Top terms were "education," "jobs," "oil_spill," "clean_energy," "Afghanistan," and "reform."

The study zeroed in on the posts of self-identified Tea Party members. Despite its grassroots nature, the Tea Party appeared to be running an organized campaign. Not only did members tweet more often, averaging 901 tweets during the study period, they exhibited behaviors suggesting a stronger community than their counterparts.

Tea Party members retweeted one another more often, rebroadcasting a colleague's message an average of 82.6 times, compared with 52.3 retweets for Republicans and 40 for Democrats. They used hashtags (keywords used to categorize tweets) an average of 753 times, compared with Republicans' 404 times and Democrats' 196. The researchers suggest this may be because the Tea Party members joined forces on Twitter to attack key Democrats. Among the party's most popular terms were "Nancy Pelosi," "Barney Frank," and "Clinton."

The researchers found that overuse of Twitter might not correlate with better election performance, though.

"In fact over usage might even repel the targeted audience to some extent," said Avishay Livne, a doctoral student in the Department of Electrical Engineering and Computer Science. The study examined how Twitter behaviors could help predict election winners. By looking at the content of candidates' tweets, the number of followers they had, and whether the candidate was an incumbent, they were able to predict election outcomes with 88 percent accuracy.

"We found that candidates who are close to the middle of the network, and the middle of what is being discussed by everyone are more likely to be elected," said Lada Adamic, associate professor in the School of Information and the Department of Electrical Engineering and Computer Science.

Adamic says the work also sheds light on how a candidate's positions correspond to his or her likelihood of being elected.

"This has been attempted in the past by looking at, for example, a candidate's past voting record or their responses to standardized surveys," Adamic said. "However, this data was frequently incomplete. It is interesting to see how candidate's activity on Twitter is connected with election outcomes."

Livne presented the findings July 19 at the International Conference on Weblogs and Social Media in Barcelona. Matthew P. Simmons, a graduate student in the School of Information, is also a co-author. The study was funded in part by a grant from the National Science Foundation.

Note: If no author is given, the source is cited instead.

• Hacking
• Information Technology
• Encryption

• Political Science
• Educational Policy
• Privacy Issues

• Experimental economics
• Cyber-bullying
• Microeconomics
• Security engineering
• Game theory
• Local area network