SAN FRANCISCO—Assessing the voting public’s mood swings, likes and dislikes to win an election has been a top-line task for political campaigns in the United States for most of its 240 years of existence. But only in the last decade have the internet, social networks and real-time analysis of big data collected from all corners of the country played a central role in helping sway voters to determine the leader of the nation and of the Free World.
The John Kerry-John Edwards campaign in 2004 made extensive use of direct email to likely Democratic voters in its losing cause. However, President Barack Obama’s campaigns in 2008 and 2012 took political connectivity to a whole new level, making extensive deployment of multiple daily emails, targeted webvertising, social networks and conventional television and radio ads to attract likely voters.
In unprecedented fashion, the team then continued fundraising throughout Obama’s eight years in office, even after both campaigns had ended.
2012 Campaign Set High Bar for Performance
As history has noted, the last two campaigns were highly successful, setting a high bar of performance for all others to come.
In 2016, all four political organizations—Republican, Democratic, Libertarian and Green parties—are following the Obama template to varying degrees of success. Real-time connectivity for immediate access to the latest event, statement or insult thrown by one candidate at another is now not only available to the widest audience in history, it is now as commonly anticipated as water from a faucet.
At Oracle OpenWorld, which concluded Sept. 22, a session titled, “2016 U.S. Presidential Election Analysis,” and discussing the impact of new-gen tech on the campaign attracted a large turnout here at Moscone West. It featured Stephanie Cutter, Democratic political strategist and founding partner of Precision Strategies; Republican political strategist and “Meet the Press” regular Mike Murphy; and emcee Dr. Michael Krasny, professor at San Francisco State University and a longtime talk show host at San Francisco PBS affiliate KQED-FM.
Polling Numbers Have Been Inconsistent
As political devotees have observed, poll numbers have been inconsistent in this election cycle. One week it’s Hillary Clinton on top with a seemingly comfortable lead, the next week it’s Donald Trump closing in and making it a horse race. Some states have conflicting polls indicating both major candidates as being in the lead at the same time.
Some analysts contend the election will be a no-contest victory for Clinton on Nov. 8, while others insist it will be very close. Is it indeed possible that big data analytics, with all its power and reach, can offer some sort of indicator about what’s going to happen? Even IBM’s Watson and plenty of smart people at various think tanks have yet to come up with such a prediction.
Much more information is coming into storage arrays now than in years past, thanks not only to the growing population but also to the proliferation of mobile connected devices. One would think all this data could be easily harnessed to provide reliable predictions; after all, that’s what a growing number of businesses are doing now with batch analytics in Hadoop clusters and other systems.
Can Big Data Analytics Be Relied Upon?
How can big data analytics accurately predict an election, when data comes in from multiple sources, including smartphones, cellphones and the internet?
“Well, that’s a challenge. This is a common complaint from campaign hacks like ourselves,” Cutter said at OpenWorld. “There are so many flawed public polls with such wide shifts out there that it’s a challenge to not let them distract you. Some look at regular voters, some look at likely voters, some use only phones, some combine it with digital polling. Some of their base assumptions about who’s going to turn out for the election are flawed.
“The most recent example was in the (Mitt) Romney campaign, where they were thinking that the 2010 electorate (which voted heavily Republican in Congressional races) would show up for the election instead of the 2008 electorate. And once you make a flawed assumption from the start about who’s going to show up (to vote), everything else falls off.”
Both the 2008 and 2012 campaigns revolutionized how data analytics did predictive modeling of how people were likely to vote and help team leaders determine who within the neighborhoods were good potential persuaders of their neighbors, Cutter said.
“By 2012, the Republicans had just about caught up (with the Democrats’ systems),” Cutter said. “We didn’t rely on one piece of data, ever. We were doing 9,000 calls a night into the battleground states for a base track we were modeling and predicting. We had a database we kept updating and refreshing; not all campaigns can do this. By June (2012), we knew we had a 70 percent chance of winning.”
Why Telephone Polls Can’t Be Trusted Anymore
Murphy, who ran Jeb Bush’s 2016 campaign (Bush dropped out of the race in February), said that he simply doesn’t trust phone-call data anymore.
“We don’t trust pure telephone calling anymore because we have so much more data we can collect,” Murphy said. “We now capture everything; we have about 400 data points on the average American voter. We now have a legitimate computing model, thanks to Oracle and others.
“If you look at it objectively, a poll is one of the few stories where the media will create the story and then report on it.”
Cutter said that “there’s definitely a shift happening in how people predict how the elections will turn out. You can’t get somebody on the telephones anymore, particularly millennials who aren’t picking up the phone or who don’t have a home phone. There’s a reason why Hillary Clinton is up 10 (points) one day and the next day is down four.”
GOP Had Analytics Lab in Silicon Valley in 2012
Murphy, who has worked on both Bush campaigns and others for Meg Whitman, Arnold Schwarzenegger and even several international campaigns, told eWEEK after the 90-minute session ended that during the 2012 Romney campaign, he started up his own semi-secret research company in Redwood City, California, to do big data voter analysis for that election.
“We got a little office there near the seaport, and I hired about 10 developers from around here to just work on big data analytics for that campaign. We learned a lot from that experience.”