On November 3, voters at polling stations across the country cast their ballots in state and local contests. In addition to elections for offices such as school boards and city councilor, state and local ballot measures will also be decided. Texans, for instance, are voting on seven constitutional amendments, including a proposal that would repeal the requirement that elected state officials reside in Austin, the state capital. People living in Portland, Maine, are deciding whether or not to raise the city’s minimum wage to $15 per hour for private employees.

To persuade voters, some campaigns have spent millions of dollars, particularly on controversial issues. One hot button in this election is Issue 3 in Ohio, which would legalize limited sale and use of marijuana, if it is approved. The backers of the yes campaign, named ResponsibleOhio, have reportedly committed more than $20 million. But campaign spending is not the only factor that can sway voters. Research suggests that even the position on the ballot where a certain contest appears will determine how some people vote. If the proposal appears earlier on the ballot, it is more likely to pass. And the difference between a “good” and a “bad” position on the ballot is not small: It can result in a swing of several percentage points.

Why would something so seemingly trivial as the ballot position impact how people vote? Psychologists call the effect “decision fatigue,” which suggests that as people make several consecutive choices, the quality of their decisions deteriorates. This phenomenon shows up in commonplace situations. If you go out on a shopping spree, for example, you are more likely to demonstrate poor judgment towards the end of the run and make purchases that you later regret. But decision fatigue can also be a real issue in professional settings. One study found that judges were more likely to reject parole requests towards the end of the day, after they had already made many sequential decisions. The judges, apparently tired of making choices, may have become more willing to simply preserve the status quo and to not grant parole.

Considering how many races and proposals Americans vote on when they head to the polls every November, it is not surprising that some people give less consideration to decisions towards the end of the ballot. In 2010, the average state ballot across America consisted of 17 elections for office and five ballot questions. And some were much longer: Residents of Cook County, Illinois, voted on no fewer than 93 candidate races and four ballot measures.

But while it may seem intuitive that decision fatigue could have an impact on election outcomes, it is harder to study and quantify the effect. Contests are typically categorized and presented in a certain order on the ballots, for example by listing congressional elections earlier than state propositions. Comparing how people vote on the two will not reveal much; an election for congressional office is fundamentally different from an amendment to the state constitution. But in a forthcoming paper, Ned Augenblick of the University of California at Berkeley and Scott Nicholson, the chief data scientist at Poynt, Inc., present a clever way to overcome this methodological challenge.

“This is a topic that I have been thinking about a lot,” says Augenblick. “I know that when I spend a day grading papers, I evaluate those that are closer to the bottom of the stack differently simply because I get tired of making decisions. I was interested in seeing whether we behave in a similar way in the voting booth.”

Augenblick and Nicholson looked at every federal, state, and local election in all of San Diego’s precincts between 1992 and 2002. In California, local issues like school board elections automatically appear earlier on the ballot than state propositions, such as Proposition 8 on same-sex marriage in 2008. This means that the ballots look different across precincts. In some precincts, the same state proposition will appear much later on the ballot than in other precincts, because more positions on school boards and in city government have to be voted on first.

So while some voters may decide on a given proposition as, say, the 12th contest on the ballot, residents of other precincts may not choose to approve or reject the same proposition until they first have voted on, for example, a total of 21 contests. Since the number of local contests presumably is unrelated to how voters in a precinct feel about a certain proposition, Augenblick and Nicholson were able to analyze this as a so-called natural experiment.

They found that when a certain state proposition appeared later on the ballot, voters were more likely to vote no (or to abstain from voting altogether). “When we get tired of choosing, we are more likely to want to preserve the status quo, which for state and local propositions means voting no,” says Augenblick. He and Nicholson estimate that the proportion of no votes on the propositions that they studied would have been, on average, 3.2 percentage points lower if the propositions had appeared at the top of the ballots instead of their actual positions. They reckon that this would have been enough to swing the outcome of 24 contests (or 6 percent of all contests), which means that they would have been approved rather than rejected. In other words, decision fatigue is a blessing for the “no” campaigns.

Even though the study only looks at contests in San Diego, it seems reasonable to assume that voters elsewhere in America also are affected by the sheer number of contests that appear on the ballot. The results extend beyond state and local propositions, too: The authors reckon that a late position on the ballot significantly increases the number of votes for the first-listed candidate in a race, as people are more likely to want to make an easy and quick selection.

So there is a cost to multiplying the number of decisions that are presented to voters. The benefit of holding many electoral contests is that people can get more involved in public policy and have a say on their representatives and on issues that could have a large impact on their lives. But the downside is that voters may end up making poor choices that they would not have made if their minds were fresher. Unfortunately, there is no simple solution to the problem. In some countries, such as Canada, national and local elections are spread out across multiple days, which make individual ballots shorter. But more elections may increase costs and can also lower turnout. Another option is to encourage absentee ballots to give voters more time to digest each choice they have to make. In Oregon, for example, all elections are conducted by mail. However, if people wait to fill out the ballots until right before they submit them, this may not result in any more careful deliberation than walking over to a traditional polling place.

Unfortunately, leaving the system unchanged may encourage the groups that introduce ballot initiatives to try to get their proposals as high up on the ballot as possible. Indeed, during the 2012 election season in California, supporters of a tax-increase proposition were accused of convincing the state legislature to pass a bill that put their proposal at the top of the propositions on the ballot. One way to resolve this would be to randomly determine the order in which contests of a certain type—such as local propositions—appear on the ballot. This could at least potentially help eliminate the incentive for powerful groups to game the system in order to place their propositions at the top.

Of course, one may ask whether it is even desirable in the first place that people vote on up to a hundred electoral races and propositions in a single cycle. In some states and counties on the extreme end of the scale it would probably be wise to consider reducing the total number of contests so that voters make fewer but possibly better-informed decisions.

“Democracy cannot succeed,” President Franklin D. Roosevelt said, “unless those who express their choice are prepared to choose wisely.” That is true—and it appears to be less likely the longer the ballot.