CAN “SAFE” STATE TV DATA DETERMINE “SWING” STATE WINNERS?
You don’t have to be a political scientist to understand the importance of “swing” states in national elections. As any newscaster or penny-ante pundit can tell you, they pick our presidents. But can data from “safe” states – or states in which electoral outcomes are assumed – be used to model which candidate swing state voters choose?
Yes – according to a new study that could impact how billions of dollars in advertising are spent during the 2016 election.
The study – entitled “Does Television Viewership Predict Presidential Election Outcomes?” – was conducted by Arash Barfar and Balaji Padmanabhan of the University of South Florida, Tampa. It was published in Big Data, a peer-reviewed journal from Mary Ann Liebert, Inc.
Using television viewership data in “safe” states in the four weeks leading up to the 2012 election, Barfar and Padmanabhan found that “models may be trained with the television viewership data in the ‘safe’ states … to potentially forecast the outcomes in the swing states.”
In other words by tracking which television programs viewers in safe states were watching, the winners (and losers) of swing states could be identified ahead of time.
“In addition to their potential to forecast, these models could also help campaigns target programs for advertisements,” Barfar and Padmanabhan added. “Nearly two billion dollars were spent on television advertising in the 2012 presidential race, suggesting potential for big data–driven optimization of campaign spending.”
And while Barfar and Padmanabhan were quick to point out their research “did not imply causality,” the strong correlations they uncovered have the potential to reshape future elections.
“The models are robust with respect to the battleground states and the counties with narrow victory margins,” they wrote.
Such data is obviously a potential gold mine for political campaigns – which can use the information to make more efficient use of their advertising budgets as well as better informed messaging decisions.
“Campaigns can use such programs for messaging, sometimes at potentially lower costs since these programs may be relatively less known for their political signal correlation,” Barfar and Padmanabhan wrote.
The findings could also dramatically impact the way presidential elections are covered. Using these predictive analytics, “it may even be possible to forecast outcomes in real time.”
“This very interesting research demonstrates the prediction of election outcomes at the state and county levels based on an analysis of television viewership across the country,” said Big Data Editor-in-Chief Vasant Dhar, professor at the Stern School of Business, New York University. “The results from the predictive model provide useful insights into some of the major drivers that drove 2012 election results. It will be very interesting to see the model applied to the 2016 elections.”
Indeed. It will also be interesting to see how the model might be expanded to incorporate internet data (site visits and searches) to help solidify the predictive strength of the television viewership data.