Read more about methods of electoral research

The major US news networks that make up the National Election Pool have enjoyed a monopoly over the collection of exit polling data for almost three decades. However, the 2016 presidential election saw the emergence of some new, digitally minded individuals involved in exit polling.

Exit polling data analyzed by the National Election Pool during the 2016 presidential election was based solely on surveys carried out by Edison Research. In a previous blog post, we saw that these surveys were conducted both in person and over the phone. However, some of the younger, more digitally focused news outlets looked to technology to help them collect data.

VoteCastr’s ‘Exit Polling’

Digital media outlets Slate and Vice News teamed up with VoteCastr to collect and publish live election data. However, instead of using traditional exit poll-type surveys like larger news networks, the focus was on tracking voter turnout on Election Day. In order to do this, VoteCastr employed hundreds of voter turnout ‘trackers’ to polling stations in key battleground states. These trackers kept an eye on how many voters were showing up at polling stations throughout the day, updating the data regularly through a smartphone app.

But how does collecting such information help these news outlets report on an election? The answer lies in combining it with pre-election opinion polls. As real-time voter turnout data was updated throughout the day, VoteCastr’s statistical models predicted who was winning and losing each state in real-time. Before the election, the creators of the model claimed this methodology to be highly accurate in predicting election outcomes. However, the final predictions were well and truly off the mark. For example, VoteCastr predicted a sizeable 8-point lead for Hillary Clinton in Wisconsin. Donald Trump won the state by a single point.

So what went wrong? The problem appears to be the inaccuracy of the pre-election opinion polls that heavily favored Hillary Clinton. However, it could also have been a failure to accurately assess who was actually turning out to vote.

Depressing turnout or empowering voters?

We’ve seen in previous posts how seriously the issue of publishing exit polling data early is taken. There are legitimate fears that polls will have a negative effect on voter turnout. Networks making up the National Election Pool have accordingly, agreed not to publish sensitive data before polls close.

The decision by Slate and Vice News to provide live information to voters on election day goes against this logic. The justification for doing so is that it enables voters to access the same information held by campaign insiders. By doing so, voters will be empowered to make more informed decisions.