Yesterday, PSCI 1150: U.S. Elections had the honor of hearing from Dr. John Lapinski, the Director of the Elections Unit at NBC News, a renowned professor at the University of Pennsylvania, and our very own Professor Clinton’s dear friend. Dr. Lapinski has been at NBC since he was a graduate student studying political science in the year 2000. Despite coming into this role largely by chance, Lapinski successfully brought data science to the Decision Desk, and three election cycles later, he became the Director of Elections at NBC.
NBC News Decision Desk, established in 1964, plays a critical role on and after election night by gathering vote counts, projecting races, and conducting exit polls. Due to the decentralized nature of the American electorate and the slight differences in methods used by each state to count their votes, gathering vote counts and projecting races is of the utmost importance. However, in addition to telling you who won on election night, the Decision Desk also tells you why the results are the way they are; they do this by interviewing voters at physical polling locations and over the phone to capture their perspectives on certain candidates and salient issues at hand. It is crucial that exit polls account for a wide range of voting methods such as in-person voting, early voting, and mail-in voting, because the way people vote today has become very partisan. In 2020, we saw that a majority of early/mail-in voters were Democrats while a majority of in-person voters were Republicans. Before 2020, this distribution was flipped, with a majority of mail-in votes being Republican. So for exit polls to be accurate, they must capture all of these modes of voting.
Today, NBC does not use exit polls to project most races. But this was not the case in 1980, when a few states were called wrong and the winner of the presidential election was announced based on exit polls even before the polls closed on the west coast. In this election, the media was accused of “putting their finger on the scale” and impacting the actual results of “down-ballot” elections by discouraging people to show up to the voting booth since Ronald Regan had already been declared our next president. However, times have changed. Today, instead of using exit polls to call races, NBC News uses them to tell stories of the American electorate and provide qualitative support for the results they derived solely from quantitative data. Accuracy is paramount for Dr. Lapinski’s team, who will not call a race until they are at least 99.5% confident in its result. “The reality of our current political environment is that I cannot, under any circumstances, call the race wrong,” Dr. Lapinski begins. “There is already so much mistrust in the media that I don’t think that our democracy would survive it. So we have to be very very careful.” For this reason, Dr. Lapinski and his team use hundreds of statistical models to analyze data and do not call the results until they meet the “minimum threshold” of 99.5% confidence in the results.
To further stress the importance of accuracy, Dr. Lapinski reflects on the 2020 election when Fox News and the Associated Press (AP) called Arizona for Biden extremely early on election night (at 11:20 pm and 2:51 am, respectively). At the time these networks called the results, only 75% of the vote had been counted. In their defense, Biden was up 8 points and it was looking unlikely that Trump would pull ahead. But given that NBC’s models showed only 78% confidence in a Biden win at that time, not to mention the grave consequences of calling a race incorrectly, Lapinski argued that the call was made too early. “Had Fox News and the AP called the race wrong, all of America would suffer. And if something like that were to happen today, it would be even more damaging to our democracy,” says Lapinski.
Professor Hemmer kicks off the Q&A section of the class by acknowledging that Dr. Lapinski brought data science to the Decision Desk. “How has this changed the process of calling elections?” she asked. Lapinski explains that in the 2000 election, there was a lot that people didn’t know, and even the experts relied on their gut feelings when calling races. In fact, the media prematurely called the result twice on election night, only to later retract those calls. It was then that the people calling the races realized that it was necessary to improve their methodology in order to accurately predict future elections.
Next, Professor Geer brings up a general concern that there's a partisan bias on calling elections. “How do you set up the Decision Desk at NBC to ensure that you are making the right call as opposed to a partisan call?” he asks. “If I were to call an election wrong, I would be sent straight to Washington to testify in front of Congress,” Lapinski responds. He explains that when you are at the Decision Desk, you lose your partisan motivation and focus strictly on the numbers. “The data is the data, there is no fudging it.” Lapinski states. The consequences of doing so would destroy the network and further erode people’s trust in the electoral process.
Lastly, Geer asked Lapinski to distinguish between prediction and projection. Lapinski explains that while some people, like Nate Silver, are in the business of predicting who is going to win the election, he himself is much more in the business of projection. According to Lapinski, NBC does make predictions, but not on election night. And those predictions aren’t even released to the public. Predictions are pretty volatile, and that volatility isn’t usually seen in American elections. What NBC does on election night is projection, which is essentially figuring out when it is mathematically impossible for the winning candidate to lose and the losing candidate to win.
When you are sitting in front of your TV on November 5th, anxiously watching the election unfold, trust the process and remember that Dr. John Lapinski and our very own Josh Clinton are behind the scenes, making the calls. They won’t let us down.
Comments