If the election is close, it's nerd time
Armies of nerds and their algorithms saved President Obama’s reelection in 2012, but Donald Trump has shown you can also run for president just by using your celebrity and a Twitter account.
Geeks and their campaign technology are not saviors. Trump’s disregard for data and digital expertise has reminded us that candidates shape elections more than anything. But if a race gets to the end and it’s still close, that’s when candidates need their nerds.
The data and digital teams, in other words, are the field-goal team. They don’t move the ball up and down the field, but they can win the game for you if it comes down to the final seconds.
In the 2016 presidential election, Democratic nominee Hillary Clinton has an all-pro data and digital team. Trump, meanwhile, is relying almost entirely on the Republican National Committee for his tools and their execution.
At the moment, with the Trump campaign in full meltdown with just under a month to go, it doesn’t appear that either campaign’s data technology operation will play a decisive role in electing the next president.
But if the Clinton vs. Trump showdown were to come down to the wire, it’s almost certain Clinton would have an edge. And regardless of whether it’s close, she’ll be better able to know which voters in which states she need to get to the polls to make a difference, and she will have more capacity to get them there. The size of her advantage is hard to know without seeing inside each campaign’s data and analytics.
But here’s how each campaign is seeking to use technology, and what we know so far about how they stack up.
It all begins with a voter file. There are many variations of voter files, with varying degrees of quality and breadth. There is no one single voter file. Each state has one, and it’s up to the political parties or private companies that compile a national list from all the state data to continue refreshing, improving and expanding their lists year in and year out. It’s one of the most important roles that state parties and the national party committees can serve.
But a campaign will start with a file built over the course of many years by their political party, which has the names of all 190 or so million registered voters in the country. They’ll build a separate list of unregistered voters, which in the U.S. is about 58 million people.
These lists include voting history, party affiliation and as much publicly available information as can possibly be assembled about a person, culled from consumer databases, donation histories, past contacts by phone, email or door-knocking, and some social media activity. Companies like Cambridge Analytica — which was hired by Sen. Ted Cruz’s presidential campaign and is now doing work for Trump — have made big claims about their ability to gather intelligence about individual voters from social media platforms. But many remain skeptical that social media can actually be mined as effectively as Cambridge claims to be doing so.
Regardless, a campaign uses a voter file to build a unique voter universe for their contest. They do this by making thousands of phone calls to people to gauge sentiment about their interest in the coming election and their candidate. Then they combine that with past behavior to award every voter a score between 1 and 100 in two categories: how likely they are to vote and how likely they are to support the campaign’s candidate.
This data set allows campaigns to predict how many people will vote in each state, and how many of those people they think will vote for their candidate. At that point, a campaign knows what their task is state by state.
In one state, their models will tell them they can win as long as they turn out voters who usually vote for their party. They will put the most effort into making sure reliable supporters who are unreliable voters show up to the polls. And their messaging — on TV, through targeted digital advertising, through radio and through the scripts that volunteers follow when making calls and knocking on doors — will be aimed at motivating people already in their camp, which is often a different kind of message than one intended to persuade a more moderate, middle-of-the-road voter.
In another state, a campaign’s models will tell them that the undecided voters who are open to persuasion hold the key to victory. That will require a very different kind of messaging campaign to those voters. Different ads will play on the TV stations in that state. Different types of mail pieces will be sent, emphasizing different issues. And the scripts for voter contact will follow suit.
And in a third state, a campaign will know that unless they find a certain number of unregistered voters who are likely to support them, get them registered and turn them out to vote, they will lose that state.
These state-by-state models will also dictate what positions a candidate takes on issues specific to each state — such as ethanol subsidies in Iowa or the Yucca Mountain nuclear waste site in Nevada — and where the candidate and their surrogates travel in the state. Trump was asked about Yucca recently and did not have an answer. That does not necessarily mean his campaign had not studied the impact of that issue on their voter universe, though it’s possible. Trump’s well-established lack of interest in policy details could also have been to blame.
Every action the campaign takes flows from those universes and voter scores. But they will also continue to update their universes and their models week by week and even day by day, often doing nightly rounds of calls to voters in states that are the closest contests. The information they get from the hundreds or thousands of responses they get every night is extrapolated to draw conclusions about larger groups of voters.
In 2012, the Obama campaign was doing 10,000 of these ID calls a night, which worked out to 1,000 responses in each of the 10 battleground states. These calls don’t go out to random voters. The IDs must get responses from a certain number of voters who are representative of a few different categories or voter score ranges, so that their responses can be used to update the larger categories.
It’s unclear how much of a challenge the Trump campaign’s shotgun marriage with the RNC has affected its ability to synchronize and tailor messaging state by state. Trump has had a data operation, but it has seen turnover, with one data director, Matt Braynard, leaving early last summer. He was replaced by Witold Chrabaszcz, a former RNC staffer.
If the RNC is trying to use data from Trump businesses or volunteers not using the national party’s data, that could make it harder for the joint RNC/Trump effort to make sure the campaign’s TV ads, digital content, voter scripts and surrogate messaging are all on the same page. The Clinton campaign tested this kind of coordination during the primary, and is fine-tuning that experience now. This kind of precision messaging also helps ensure that the Clinton campaign is not sending messages to voters that might have a negative impact on other elections in key states, like close battleground races for the U.S. Senate.
At one point over the summer, Trump’s travel to nonbattleground states flummoxed many observers and supporters, and was clear evidence that his campaign was not using any kind of data-backed analytics to guide internal decision making about where to send him. But RNC officials say that has changed, and there is now a weekly meeting to review fresh data and use it to shape upcoming travel.
Even with that, however, the Trump campaign does not have the kind of manpower directing resources that Mitt Romney’s campaign did in 2012. With any traditional candidate, “you’d have a 50 or 60 person operation at the campaigns directing strategy and guiding where resources need to be spent,” said Mark Stephenson, a Republican operative who oversaw the data and analytics operation for Wisconsin Gov. Scott Walker’s presidential campaign.
Early voting is another area where data and analytics plays a crucial role. The voter universes built by the Clinton campaign and the RNC will help them understand who is voting early and who is not. They’ll know, then, whether they are losing or winning the state well in advance of Election Day, and that will dictate how much money they spend on TV there, how often their candidate travels there, and who they try to contact to get them out to vote down the stretch.
One of Clinton’s big advantages appears to be in her precision in placing TV ads. Politico reported last month on an algorithm written by Clinton data analytics director Elan Kriegel to target TV ads in the primary, and Stephenson said his analysis of Kantar Media data shows that Clinton is getting far more bang for her buck on TV than Trump.
“Hillary is buying 50 or 60 channels while Trump is only buying 8 or 9. It points to a data-driven optimization,” Stephenson said. In other words, Clinton’s campaign knows their voter universe well enough that they can place a certain type of ad on a rerun show like “Wheel of Fortune” because they know it’s going to hit a certain target audience and they’ll pay a bargain rate. Clinton’s campaign has also maximized, like Obama did, advance purchasing of air time to lock in lower rates.
Clinton has already outspent Trump exponentially on the airwaves and through online advertising, and is projected to outpace him by a huge margin between now and Nov. 8.
The only caveat to Clinton’s advantage here is the effectiveness of TV ads in the modern media environment. Yet Clinton campaign officials said in private conversations that the election will be won over a “relatively small number of votes,” and they are maximizing every means of influencing those voters.
Digital content distribution and digital advertising is another huge category, as is building an email list, which is key to raising small dollar donations.
Trump’s digital director, Brad Parscale, has a team of RNC digital staffers headed up by Gary Coby embedded in his St. Antonio offices doing the bulk of the campaign’s digital advertising. The RNC also took over managing the entire 9 million person combined RNC and Trump campaign email list after Parscale hit a spam rate so high that there was risk of the campaign getting kicked off its email provider.
Kriegel, Clinton’s data director, said that 2016 has been “interesting from a pure data perspective” because of Trump’s unpredictability and the ways in which certain categories of voters have been tossed up in the air and rearranged.
For all the talk of Trump bringing new voters into politics to vote Republican, Kriegel said, he has seen statistical evidence that there are registered Democrats who have not been regular voters over the past few elections who have been roused in opposition to Trump and are more likely to vote now.
But despite the volatility of the 2016 election, Kriegel expressed the kind of confidence about the shape of the electorate that’s customary for the operatives who endlessly sift large data sets.
“I have a pretty good idea who is still deciding, who his core is, who our core is and who our new supporters are,” Kriegel said.