It’s one thing to get a good grade on a final project. It’s quite another to see your final project mentioned in The New Yorker.
But that’s exactly what happened with Robynn Singer-Baefsky. Her end-of-year project for Senior Scholar-in-Residence Spencer Kimball’s polling class. Singer-Baefsky ‘18, co-president of the Emerson College Polling Society (ECPS), a student-run organization advised by Kimball, polled likely New York state voters in the primary, and accurately predicted Democrat Hillary Clinton and Republican Donald Trump winning by 15 and 34 points, respectively.
“That was incredible, because it was just an exciting moment,” Singer-Baefsky, a Political Communication major, said. “I plan my life around election cycles, so knowing polling and different methodologies, it’s really opened my eyes to a lot of things in the political sphere.”
Just four years after coming into existence, ECPS polls have garnered a lot of attention during the 2016 election cycle. Politicos from across the spectrum, including Nate Silver of FiveThirtyEight, MSNBC’s Chris Matthews, and FOX News’ Bill O’Reilly have all praised Emerson’s polls. In June, Bloomberg ranked Emerson number one in accuracy among collegiate pollsters during the primaries.
But for all the pundits’ plaudits, ECPS really has one overarching mission, according to co-president and fellow Political Communication major Christine Kane ‘17.
“We consider ourselves to be a learning organization, and we…follow the school’s philosophy of learning through immersive experiences,” she said. “The purpose of the organization is to teach students about this particular aspect of the political world.”
How immersive is it?
When the data comes in from a particular poll, it needs to be cleaned up and anonymized, so there is no individual identifying information, Kane said. It’s then analyzed using statistical software, and interpreted by the polling society members, who glean what the data says about voters' preferences and intentions. Finally, the team writes up a press release and shares it with media.
“Typically, that takes place in about an 18-hour span, so it’s pretty high pressure,” Kane said.
The Polling Society did 30-40 polls during the primaries, and as of this writing, have polled 14 states in the general election cycle, many of which included Senate races.
Students aren’t just analyzing data from the polls, they’re also analyzing the polling process itself – specifically the strengths and weaknesses of the methodology Emerson uses, said Kimball, the advisor. ECPS conducts Interactive Voice Response (IVR) polls (commonly known as “robo-polling”).
“The low cost of so-called robo-polling has led to its proliferation in politics, and in 2012, nearly two-thirds of all state-level polls used an automated system,” Kimball said in an email. “Independent studies done after the 2008 and 2012 elections found the automated methods to be as reliable as live-operator telephone data collection. However, academic research on IVR’s validity is sparse, and we are currently implementing the best practices to test this method of research.”
The goal for the general election, Kane said, is to focus on 12-15 battleground states, which Singer-Baefsky is looking forward to.
“I’ve been wanting to poll Florida for months now…because it’s [such a] swing state,” she said.
This election cycle has been fascinating, Singer-Baefsky said, partly because party loyalty is far shakier than normal. Their polling has suggested that people who had supported John Kasich, Ted Cruz or Marco Rubio in the primaries are not migrating over to the Republican nominee in the numbers they normally would, with a larger share than is typical saying they’ll support Clinton or Libertarian candidate Gary Johnson.
On the Democratic side, while a wide majority of Bernie Sanders supporters say they’ll vote for Clinton, an unusual number say they’ll support a third-party candidate, or even Trump. In bluer-than-blue Rhode Island, Singer-Baefsky said, 24 percent of Sanders supporters polled say they’ll vote for the billionaire developer.
Third-party candidates are getting a lot more support this cycle compared to other years, which kind of throws a “wild card” into the mix, Kane said.
“There’s a lot of talk that the electoral college is very favorable to Clinton,” Kane said. “So even if national polls are tight, and even if state polls are tight, there’s something of a consensus that she has a big advantage in the electoral college. But again, you can’t take anything for granted.”
The bottom line, Kane said, is this election is very, very difficult to predict. And unlike the primaries, which provide almost immediate feedback, pollsters won’t know how accurate they are in the general elections until November 9.
“This year is such a complete anomaly in so many ways that almost nothing surprises me,” Kane said. “Seriously.”