As another election season heats up, researchers are engaged in a battle of their own over voter behavior: Can it really be swayed by whether the local college football team just won or lost?
Five years ago, researchers drew attention with data suggesting that a win within 10 days before Election Day gave a small ballot-box boost to the party of the incumbent senator, governor or president.
The idea was that the victory makes local people happier about the way things are in general, which nudges them to vote for incumbents.
Now a new analysis concludes that the statistical pattern found in that study was just a coincidence, not really the result of success on the gridiron.
The finding was "bad luck for researchers rather than a shortcoming of American voters," as the authors put it in a paper released Monday by the Proceedings of the National Academy of Sciences.
For the new work, political scientist Anthony Fowler of the University of Chicago and B. Pablo Montagnes, now at Emory University in Atlanta, analyzed records from 1960 to 2012. They reasoned that if the football effect is real, it should also produce certain other outcomes. But it didn't.
For example, no influence from football wins appeared in years when a county's voters faced multiple election races that had different incumbent parties. Nor was it especially strong in counties where citizens care the most about college football, as measured by Facebook "likes." Nor was it stronger when the incumbent was up for re-election than when the seat was open.
And the researchers found no local effect of wins by NFL teams.
But Stanford University political scientist Neil Malhotra, an author of the 2010 study, said the new analyses weren't appropriate for testing whether the original results were a fluke.
They simply show that the football effect "is not something that happens all the time, in every circumstance," he said.
Explore further: Can black Republicans win black votes? Not likely, study finds
College football, elections, and false-positive results in observational research , PNAS, www.pnas.org/cgi/doi/10.1073/pnas.1502615112