Researchers from universities including Stanford and Harvard have published new evidence that the order in which candidates are listed on ballots can influence how people vote — and the effect is so strong that they claim it likely would have changed the outcomes of presidential elections in 2000 and 2016.

Using data from New Hampshire, a state that rotates name order so that no candidate or party is listed first on the majority of ballots, the researchers found what they call a "ballot order effect" in several primary and general elections, including the 2016 presidential race between Donald Trump and Hillary Clinton. For ballots where Trump was listed first, he gained an extra 1.7% of the vote share; when Clinton was first, she gained 1.5%. 

The findings, which incorporated factors like party registration that affect vote share, were shared in a March paper for the journal PLOS One. 

While ballot order effects have been observed in other states for decades, this new paper marks the first comprehensive examination of the pattern in New Hampshire, which is especially important for U.S. presidential elections because it's the first state to vote in both the Democratic and Republican primaries, and because it has seesawed between both parties in general elections in recent decades. New Hampshire also has unusually high voter engagement and turnout. 

"The paper is a combination of the primary data as well as general election data, and we see the same prevalence there as we're used to seeing," said co-author and Stanford University professor Jon Krosnick, who has studied ballot order effects since the 1990s. 

Krosnick has also testified in several lawsuits seeking to require states to rotate ballot order, including in New Hampshire and Florida. He believes that ballot order effects likely swung the results of the 2016 and 2000 presidential elections.

In 2016, Trump barely defeated Clinton in several key states, taking Michigan by 0.22%, Wisconsin by 0.76% and Florida by 1.20%. Trump was listed first on every ballot in each of these states and therefore likely benefited from ballot order effects, according to Krosnick. 

"In the states that Donald Trump won by a very tiny margin, he was listed first on the ballot in all but one of those states," Krosnick said. 

Theoretically, if the three states had rotated ballots and eliminated a New Hampshire-sized ballot order effect from their results, Trump would have lost them and therefore lost the presidency. 

And in 2000, George W. Bush defeated Al Gore by a margin of just 537 votes — or 0.009% — in Florida, which determined the election, while being listed first on every single ballot in that state. So even if there was a ballot order effect of just 0.01% that year, it had the effect of giving the presidency to Bush. 

"Most likely, Donald Trump would have not been elected in 2016, George W. Bush would have not been elected in 2000 [if ballots had been rotated]," Krosnick said. 

Krosnick wrote the paper with Bo MacInnis of Stanford; Joanne Miller of the University of Delaware, Newark; Clifton Below, a local government official in Lebanon, New Hampshire; and Miriam Lindner of Harvard University.

Ballot order effects are the political manifestation of a broader psychological tendency called the "primacy effect," according to Krosnick. The inclination to choose the first candidate on the ballot is similar to the tendency to pick the first stall in a public restroom or to prefer the first beer out of four in a blind taste test, he said. 

"This is part of life. It's part of being human," he said. 

About 10 U.S. states, including Ohio and New Hampshire, currently rotate ballots in a manner Krosnick considers fair. The practice lets researchers better observe causality, he said, because assigning half of the electorate one kind of ballot, and the other half another kind, functions as a kind of "quasi-field experiment."

While rotating ballots can mean extra work for elections officials, he says more states should embrace the idea in order to increase the public's trust in elections. 

Krosnick also supports the idea of federal legislation requiring states to rotate ballots and allow outside audits to ensure compliance, which he said would help both Democrats and Republicans assure constituents that their representatives are serious about election security and fairness. However, he believes that national lawmakers may be wary of being accused of pushing states around. No such bills appear to be under consideration by federal lawmakers, and the issue has received only sporadic coverage in the political press. 

"You'd think the Democrats in Congress would have a lot of motivation to do something about this," Krosnick said. 

The paper, "Candidate name order effects in New Hampshire: Evidence from primaries and from general elections with party column ballots," published on March 16 in PLOS One, was authored by Bo MacInnis and Jon Krosnick, Stanford University; Joanne Miller, University of Delaware, Newark; Clifton Below, city councilor in Lebanon, New Hampshire; and Miriam Lindner, Harvard University.