Researchers examine correlation between political speeches, voting

Dec 08, 2009

Although politicians are often criticized for making empty promises, when it comes to their voting records, their words may carry more weight than previously thought, according to findings by two Penn State information technology scientists.

The researchers used a to compare voting records from the 110th Congress -- Jan. 3, 2007 to Jan. 3, 2009 -- to each senator's floor statements on the issues to determine whether the two matched up.

They did this by creating a computer-based regression model to scan the floor speech text and compare it to each senator's DW-Nominate score, a measure of how conservative he or she is based on voting record.

"This study validates what we had been thinking - that you can get a good idea about how people will vote based on what they said," said Prasenjit Mitra, assistant professor of information sciences and technology. "We were able to study political discourse and find biases quickly without having to look for specific clues in the text."

Findings showed that legislators do give floor speeches on topics that supported their ideologies, and the supporting materials chosen for those speeches also aligned with their ideologies. For example, conservatives cited sources like the Heritage Foundation, a well-known conservative think tank.

The word-mining technique used in this study is commonly used in analyzing online product reviews, but had not previously been applied to politics.

The researchers found that nouns used by the senators tend to signal the way they will vote -- a major difference from the online product reviews where typically adjectives are used to identify opinions.

For example, liberals tended to favor nouns like "disability" and "veteran" and adjectives like "homeless" and "ecological," while conservatives opted for nouns like "clone" and "missile" and adjectives like "homosexual" and "ballistic." Mitra said, however, that the voting analysis was based on the overall impact of the words, not each one's individual ideological leaning.

"We were able to gain a better understanding of how opinion is expressed and what clues to look for to identify ideological slants," said Mitra.

Mitra and Lei Zhu, doctoral student, information sciences and technology, presented their work at the Conference on Information and Knowledge Management in November in Hong Kong.

Future research in this area could include building an automated program that could compare real-time data and voting records. The method used for the U.S. political data could also be applied to other countries' political systems with the help of translation software.

Source: Pennsylvania State University (news : web)

Explore further: Study shows employees become angry when receiving after-hours email, texts

add to favorites email to friend print save as pdf

Related Stories

Media bias can net mistakes at the ballot box

Jun 25, 2008

The media slant political news to the left or right to increase ratings and profits, spinning up an information vacuum that can lead to mistakes at the ballot box, a new study by three University of Illinois economists says.

Recommended for you

Predicting human crowds with statistical physics

6 hours ago

For the first time researchers have directly measured a general law of how pedestrians interact in a crowd. This law can be used to create realistic crowds in virtual reality games and to make public spaces safer.

Broken windows thesis springs a leak

11 hours ago

The broken windows theory posits that minor misdemeanors, like littering or graffiti spraying, stimulate more serious anti-social behavior. LMU sociologists now argue that the idea is flawed and does not ...

User comments : 0

Please sign in to add a comment. Registration is free, and takes less than a minute. Read more

Click here to reset your password.
Sign in to get notified via email when new comments are made.