Sitting at his laptop, Jon Black scrolled over "Arkansas" to see a burst of bar graphs, each representing a week's worth of data on an important slice of the electorate: Republican-leaning Arkansans who often don't vote. Black winced.
A crooked blue line skirted just above the top of seven consecutive bars, telling Black, director of voter turnout at the Republican National Committee, that his Arkansas colleagues were short of their goals for this particular cluster of voters labeled "High Value GOP." He called his colleagues in Arkansas.
Almost instantly, Tom Cotton's Senate campaign started visiting, calling, and aiming digital ads at every young white male in the cluster. The data showed them to be the most persuadable of the "High Value GOP" voters. For each young white man – thousands of them – Republican Party operatives had a name, an address, a voting history, and most important, a rich attitudinal profile based largely on their online activities. Within a week, Cotton was back on target.
Welcome to 21st century micro-targeting – the ability to collect and crunch vast amounts of data on virtually every American and use that data to shape how people shop, eat, work, worship, play, and, of course, vote. What data-miners know about you is mind-boggling – if not at least a little bit frightening.
In politics, over the last decade, the Republican Party virtually ceded a technological monopoly to Democrats, foolishly blinding the GOP to tens of millions of potential voters. Until now.
A review of the RNC's targeting operation (including a preelection sample of specific projections) suggests to me that the GOP has made significant advances on targeting and mobilizing voters. While the Democratic Party may still own the best ground game, GOP Chairman Reince Priebus has narrowed, if not closed, the tech gap.
A few Democrats saw this coming. "Our side has underestimated the GOP ground game," Democratic pollster Celinda Lake told me Tuesday morning. "Their electorate doesn't look like ours, so we don't recognize or respect what they're doing."
First, some background. Ten years ago, it was the Republican Party that introduced Washington to the micro-targeting tactics long used by U.S. corporations and mega-churches. Facing a tough reelection campaign, President Bush's team expanded the pool of GOP voters and revolutionized the science of politics. (In 2006, I co-authored a book with Bush's targeting guru, Matthew Dowd, describing the targeting operation.)
The Bush squad started with government-run lists of registered voters: their names, ages, addresses, and voting histories. Over time, those lists had been enhanced with the purchase of additional political information, such as the membership rosters of advocacy groups.
Bush's team sent the RNC list to a data-mining firm called Acxiom, which had purchased consumer information from credit card companies, cruise lines, airlines, retails stores, and scores of other places where people did business. Acxiom cross-referenced the Bush team's voter lists against its own list of consumers.
The Bush team never had access to raw consumer data, but it had all it needed: a mega-list from Acxiom showing the stage in life (age, marital status, number of children, etc.) and lifestyles (hunter, biker, home renter, level of religious interest, etc.) of each voter, drawn from a menu of more than 400 separate categories.
Next, the Bush team called 5,000 people from that Acxiom list and asked them a series of questions to determine their political behavior and attitudes, as well as specific issues that made them angry.
Finally, those results were run through a computer program that grouped the 5,000 voters into 30 or so segments, each consisting of people who shared political and lifestyle traits; think of them as a virtual community.
Then every other voter – tens of millions of them – was assigned to one of the segments that fit based on the lifestyle and political habits he or she shared with those surveyed and already placed into groups.
It worked. But the GOP let its operation ossify while Democrats leapfrogged the Republicans in 2008, when Barack Obama's team harnessed rapidly advancing technologies to amplify his message. Those gains were multiplied in Obama's 2012 reelection, after which Priebus vowed to get the RNC back in the game.
Fast-forward to a week before Tuesday's election, when Black toggled through his models and showed me the RNC's projections for the Senate race in Arkansas. The most likely scenario:
Total Turnout: 808,337
GOP turnout: 412,252.
In other words, a narrow victory. A week later, on the morning of the election, fresh polling and other data pointed to a GOP surge – at least a 7-point margin of victory for Cotton. One best-case model assumed that unaffiliated voters would swarm to Cotton and swell overall turnout. That projection, sent to Priebus on Monday night, showed Cotton routing Sen. Mark Pryor, 54-38.
The actual margin as of Wednesday morning was 18 points. Actual turnout: 837,026.
Democratic consultant Doug Sosnik, who co-authored "Applebee's America" with Dowd and me, declared Tuesday that the GOP has clawed back into the ground wars. Comparing Republicans to other fast-adapting institutions, Sosnik said, "They don't invent anything, but they can take an invention and make it better, faster, and for less money. That's how technology works."
Credit goes to Priebus and a data team headed by Jesse Kamzol. Last week, Kamzol described how far the world of micro-targeting has progressed. The advances benefit both parties, assuming they're both willing to take advantage.
In 2004, voter registration lists were sparse. Now, they're updated routinely by state governments that often pay for outside data to supplement their lists.
In 2004, the Bush team could track voters based on a couple hundred consumer variables – brands of beer, cars, etc. Now, targeting gurus can choose from 3,600.
In 2004, Acxiom cross-referenced the voter list once for the entirety of the campaign. Now, it's updated weekly.
Where the Bush campaign conducted a one-time poll of 10,000 people nationwide to produce modeling for the duration of the campaign, the RNC this year polled 26,000 people each week –1,000 in each of the key states.
Finally, there is just so much more digital data on all of us than there was a decade ago – collected and analyzed in orders of unfathomable magnitude. The RNC had enough data this year to rate each individual registered voter on a scale of 1 to 100 on their likelihood to vote strictly along party lines; to vote for a GOP candidate or a Democratic candidate; to participate in elections; and to cast an early ballot. They also rated each voter's view on Obamacare.
Some pundits, including David Brooks of The New York Times, worry that big data is sucking the life out of politics. I don't see it that way. Actually, micro-targeting could return politics to its roots – to the time before mass media when people joined a party because of a personal connection. Consider the 19th century immigrant who arrived to the United States by boat and was greeted at the docks by a Democratic official who helped him find a home, a job, and a good school for his kids – and then six months later, showed him how (and for whom) to vote.
It was a highly personalized, transactional relationship built upon the immediate needs of the voter. In that sense, technology can take us back to the future: The maturation of digital-data politics will place a higher value on people on the ground than negative ads on the air, which may not be such a bad thing.
The Democratic Party's culture and history are easily suited to these ground games, which is why it's notable that Priebus deployed 2,000 operatives to competitive states this year and will leave 300 people in the field full time, the largest permanent commitment of GOP ground troops. (Democrats can still muster more through unions alone.)
Out of curiosity, I asked the RNC team Tuesday morning what they knew about my political tendencies – based only on what their computers told them about my history as a voter and consumer. A few keystrokes later: "You lean Democrat," said communications director Kirsten Kukowski. Specifically, for every race Tuesday, there was a 50 percent chance I'd vote for a Democratic candidate, the RNC predicted, and a 26 percent chance I'd back a Republican. Also, the models reported a 70 percent chance I would vote in the midterms. As a persuadable likely voter, both parties had me on their target lists.
I had just voted near my home in suburban Virginia when Kukowski called with my electronic profile. The RNC pegged me pretty well: I had split my ticket between Democrats and Republicans.