Advertisement

Real-time Election Day projections may upend news traditions

 
Lynn Suarez, left, and her husband, Raymond Suarez, stopped by the West Tampa Library on Union Street to participate in early voting on Aug. 15 for the primary. [OCTAVIO JONES   |   Times]
Lynn Suarez, left, and her husband, Raymond Suarez, stopped by the West Tampa Library on Union Street to participate in early voting on Aug. 15 for the primary. [OCTAVIO JONES | Times]
Published Sept. 12, 2016

PALO ALTO, Calif. — For decades, news organizations have refrained from releasing early results in presidential battleground states on Election Day, adhering to a strict, time-honored embargo until a majority of polls there have closed.

Now, a group of data scientists, journalists and Silicon Valley entrepreneurs is seeking to upend that reporting tradition, providing detailed projections of who is winning at any given time on Election Day in key swing states, and updating the information in real time from dawn to dusk.

The plan is likely to cause a stir among those involved in reporting election results and in political circles, who worry about both accuracy and an adverse effect on how people vote. Previous early calls in presidential races have prompted congressional inquiries.

The company spearheading the effort, VoteCastr, plans real-time projections of presidential and Senate races in Colorado, Florida, Nevada, New Hampshire, Ohio, Pennsylvania and Wisconsin. It plans to publish a map and tables of its projected results on Slate, an online newsmagazine.

The company will make its projections by looking at who is actually turning out to vote, and then processing that data through a method known as predictive turnout modeling. The process is similar to how presidential campaign war rooms operate on Election Day, when they track turnout by likely supporters so they can adjust get-out-the-vote efforts accordingly.

"It's what campaigns do," said Ken Smukler, the founder of VoteCastr. "We're flipping up the kimono and letting people see what campaigns do on Election Day."

Providing real-time updates will be a drastic departure from standard election reporting that television networks, national newspapers and the Associated Press have rigidly adhered to for decades. Many news organizations refrain from publishing exit poll data about the likely winners in a state until a majority of polls there have closed. (Exit polls are based on surveys of voters as they leave polling stations, and they are considered a polling gold standard because only people who voted are questioned.)

Though not legally bound to do so, news organizations have kept this information under wraps for fear of suppressing turnout and affecting down-ballot races if the presidential election were called before voting in most states had ended.

"Politicians from Western states have been very critical of any attempt to project election outcomes and report election outcomes before voters in their states have had a chance to cast their votes," said Joe Lenski, executive vice president of Edison Research, which conducts exit polling for leading news organizations on election nights.

There is no conclusive evidence that early reporting suppresses turnout. But news organizations have stuck to their decision to refrain from reporting any results early on Election Day.

"The media approach to covering Election Day broadly is to have a total media blackout on what voters are doing in real time," said Julia Turner, the editor-in-chief of Slate, who called the withholding of such information "ill conceived and anti-journalistic."

The group expects to make some bold predictions to kick off Election Day. At 6 a.m. Eastern time, before the first rays of sunlight kiss the Rockies or a single Election Day vote is cast, it expects to project who is winning the swing state of Colorado based on absentee ballots and an early vote count kept by the Colorado secretary of state.

Smukler of VoteCastr experimented with this style of reporting during the Philadelphia mayor's race in 2003, a rematch between Mayor John F. Street and Sam Katz, projecting results live throughout the day on the radio. The projections held true all day.

VoteCastr is basing its turnout models on key counties in the swing states, sampling from areas that favor Hillary Clinton, swing areas and areas that support Donald Trump. Those models can then be used to project results statewide.

On Election Day, hundreds of employees and volunteers will be dispatched to designated precincts in their key counties to report turnout. If turnout exceeds their modeled expectations in an area solidly for Clinton, for example, they can deduce that she is overperforming in the percentage of total statewide votes coming from similar areas and is therefore likely to outperform in total vote expectations statewide.

They then tabulate all of these real-time results to eventually report who is most likely winning the state.

VoteCastr argues that this method is far more effective than using exit polls, which it considers useful for explaining results after the election, but flawed for projecting turnout on Election Day.

"Exit polls are crude, inefficient and a bad way of predicting these outcomes," said Sasha Issenberg, a journalist and an author of the book The Victory Lab, who is a member of the VoteCastr team. "They have always been designed to tell us why certain types of people voted the way they did, not to predict the outcome."

The media has been skittish about reporting early results partly because of its own history.

In 1980, there were reports of voters leaving lines in California after television networks called the presidential race for Ronald Reagan before the polls there had closed. The networks made the call after President Jimmy Carter conceded, and voter data did not show a large drop-off in turnout in the West. Nonetheless, the presidents of ABC, NBC and CBS were brought before Congress for numerous hearings.

After the 2000 election, when some networks incorrectly called the Florida results before voting had ended in the Panhandle, which is in an earlier time zone, the network presidents were again summoned to Capitol Hill.

Television networks have taken their pledge not to post results too early so seriously that they even quarantine their election night forecasters and polling professors who work for them.

"The networks actually do this quarantine room because they do not want this info to leak out," said Ken Goldstein, a professor of politics at the University of San Francisco and a member of ABC's decision desk. "That's to their credit."

Both Slate and VoteCastr feel they are on solid legal ground reporting projections early.

"The legal constraints generally have been those on exit polling, and those were restraints about physical distance and approaching voters," said Karl J. Sandstrom, a lawyer for VoteCastr.

The group does not feel comfortable making outright calls of winners, but rather intends to show "where things stand" throughout the day. "The appeal is not to out-CNN CNN or declare the election over at 10 a.m. on Election Day," Turner said.

Either way, some say the group could be getting ahead of the results and threatening how elections should be covered.

"I'm profoundly uncomfortable with characterizing election results during Election Day," Goldstein of ABC said.

But the group, seeing its role as part informer, part disrupter and part pioneer in the world of election reporting, believes it will change the Election Day experience forever, even adopting the provocative terminology used by Sen. Ted Cruz of Texas during his presidential bid to describe the news industry.

Speaking about the networks' reluctance to share their information throughout the day, Smukler has a simple moniker for the election consortium: "the media cartel."