Skip to main content
Expand AU Menu

News

Research

The Numbers Game: Math and Stats Department Tackles the 2016 Election

By Gregg Sangillo

Colorful silhouettes of hands grasping ballots, positioned to place them in a collection box.

The Mathematics and Statistics Department can see through the fog and dissect complex electoral data. Credit: iStock.

Despite all the angry barbs and hot takes, the 2016 election will be decided by numbers. In the critical battleground states, will Hillary Clinton or Donald Trump turnout more voters?

Polling has never been more obsessed over and analyzed. But though we all may think we’re Nate Silver, experts are needed to dissect complex electoral data. That’s where the American University Mathematics and Statistics Department can help us sift through the madness.

This election season, AU professors and students are doing invaluable research. For AU statistics professor Mary Gray, this underlines the significance of their field of study.

“Students need to know why statistics are useful,” Gray says. “And there’s civic engagement involved. Having students see how they can do something, and how statistics are meaningful, is also very important.”

Voter ID

Voter ID laws have been extremely controversial, particularly in states like North Carolina and Texas. Many Republicans have favored these measures to combat voter fraud, but Democrats and some public interest groups counter that voter fraud is quite rare. Critics charge that voter ID laws are a thinly veiled GOP attempt to suppress votes in minority—and mostly Democratic—communities.

Gray will help lead a voter ID project with many AU students—in both the math and stats department and beyond. On election day, AU students will disperse to Northern Virginia polling stations in Arlington, Fairfax, and Loudoun counties to assess the number of people unable to vote because they lack voter identification (such as a driver’s license or passport).

Gray says they’re hoping to recruit up to 100 students, with the goal of covering about 50 precincts. Through systematic sampling, they’ll interview one in five people, or one in 10 (depending on crowd size), as they leave polling stations. Mostly math and stats people analyze the data afterwards, but politically-inclined AU students from all departments are invited to participate as election day questioners.

In Virginia, AU collected similar voter ID data in the gubernatorial year of 2013 and for congressional midterms in 2014. For those two elections, they found that roughly 2 percent of people interviewed were prevented from voting because of insufficient voter ID. People turned away were “minorities, and they were all in low-income districts,” says Gray.

In many instances, people don’t even know about the voter ID laws until they show up at the polls. In 2015, AU conducted surveys primarily in pharmacies and grocery stores, and discovered that while most people had IDs, fewer than 50 percent of interviewees were aware of voter ID laws.

Though 2 percent of would-be voters might not sound like much, it could make a difference in a close election. More importantly, Gray says, it disenfranchises those individual voters.

“The right to vote is fundamental. All of the other rights fall apart if you don’t have the right to vote,” she says.

Strategic Voting and How to Win

Kenneth Ward, a professorial lecturer in the Mathematics and Statistics Department, is also working on the voter ID project. In addition, he’s utilized fascinating strategic voting models to better understand candidate and voter behavior.

Though formulated in 2003, these models sound strikingly prescient in 2016.

A synopsis of one model reads: “Opportunistic candidates can strategize by revealing little information.” Many media outlets have ridiculed Trump for being light on policy details.

As a result, Ward says, it can benefit the other candidate (e.g., Hillary Clinton) to attack the opportunistic candidate. “It’s in the best interest of the other candidate to give a significant amount of negative information about the candidate who withholds information about his or her policies,” he says.

With the election less than a month away, Ward says Clinton’s most effective strategy would be to highlight “provably negative information” about Trump.

“One candidate might only say, ‘Well, my opponent is corrupt in general,’ but not really point towards anything. But the other candidate can give definite stories where people say, ‘I lost everything because of my business dealings with him,’” he says. “The models predict that if she does that, there’s a high probability that she will win.”

Another synopsis states: “Newer models say that voters’ knowledge of each other can matter more than knowledge of the candidates.”

This is the essence of Ward’s research findings, as voters behave quite strategically. They make their candidate selections based on their expectations of what other voters will do.

“Any strategic voter who does vote must believe, in some sense, that this decision should be conditioned upon the possibility that his or her vote is the deciding vote in the election,” Ward explains.

Part of this is derived from game theory. Ward was friends with the late John Nash, a game theory pioneer and subject of the Oscar-winning biopic A Beautiful Mind. Based on what’s known as “Nash equilibrium,” people behave in their own best interests in a system that is stable. The way to make it stable for voters, Ward adds, is by providing them with information.

If the Clinton and Trump campaigns have internal polling numbers, Ward says they have reasons for keeping their findings private. Clinton’s internal polling numbers may be really strong, but her campaign might not want to divulge that over fears that her supporters will stay home. Conversely, even if Trump’s internals are weak, he’d want his voters to feel like their views are consistent with a sizable portion of the population.

“The candidate might want voters to have misconceptions about what others believe,” he explains.

Polling, Accuracy, and Seeing Through the Fog

Having said that, Ward notes the potential pitfalls of an abundance of data. A voter might not know what to believe. “There’s not a lot of modeling of, ‘What happens when voters receive too much information?’ And that’s a big issue,” he says. “Part of the goal here, on both sides, is to give individuals information that they remember. Now will they factor in 1,000 polls? How do they choose between all of those?”

Gray is skeptical about some political polls. Various polling outlets use panels, essentially a database of people who answer surveys online. Gray notes that those kinds of surveys—which only measure a certain type of voter—are what failed to predict the United Kingdom Brexit vote to leave the European Union.

Businesses need those online databases for consumer and marketing research, but politics is a whole different animal. “It’s fine if you’re trying to sell toothpaste. But to put that forward as an accurate reflection of public opinion on important issues is harmful,” she says.

And Gray won’t venture to predict what will happen this election. “I try to keep politics out of this, but this election is so crazy, all bets are off.”