The link between tweets and votes in political elections could be stronger than once thought, a new study has found.
An Indiana University study found that the percentage of votes for Republican and Democratic candidates in 2010 and 2012 races for the US House of Representatives could be predicted by the percentage of tweets that mentioned those candidates - and it didn't matter whether the tweets were positive or negative.
"Think of this as a measurement of buzz," said Fabio Rojas, an associate professor of sociology in the College of Arts and Sciences at the university.
"We call this the 'all publicity is good publicity' finding. Even if you don't like somebody, you would only talk about them if they're important," said Rojas.
Rojas and colleagues analysed a sample of 537 million tweets to examine whether on-line social media behaviour could be used to assess real world political behaviour.
Study lead author Joseph DiGrazia will present the findings at the Annual Meeting of the American Sociological Association.
Until now, DiGrazia said, polls and surveys were the primary way to gauge public attitudes.
"Our findings show there is massive, untapped, reliable data out there that can give insights into public opinion," he said.
Unlike other studies that have looked at the influence of social media on election outcomes, the study took into account such variables as incumbency, partisanship, media coverage, and socio-demographic makeup of the electorate.
They say Twitter provides another tool for analysing races, particularly when other data is in short supply.
Researchers said the Indiana Twitter database contains the largest sample of tweets in the world that is available to academics.