The 2016 election tested a number of questions about American electioneering, among which was how much organization matters in the modern political environment. The Trump campaign had very little organization and no get-out-the-vote (GOTV) operation. The Clinton campaign went big on both. The election should have been as close to a laboratory test of the hypothesis as possible.
We have our answer: Donald Trump won the presidency with little more than a bunch of rallies and a Twitter account. (Plus Facebook and YouTube.) His victory seems to prove that organization is vastly overrated, a conclusion that strikes fear into the hearts of campaign consultants across America.
But the real answer may be more complicated. For one thing, it’s entirely possible that Hillary Clinton’s loss would have been worse had she forgone the framework. In Nevada, for instance, it was the machinery that put her over the top. “There were 24 people in Nevada doing GOTV for Joe Heck,” explained a consultant who worked on the Republican congressman’s unsuccessful bid for the Senate. “The Clinton campaign had 12 different organizations.” Clinton won the state by 26,000 votes. There was a similar dynamic in New Hampshire, where Clinton GOTV almost certainly minted her the 1,500-vote margin that allowed her to hold the state.
And at the end of the day, organization bumps up against the limits of the product. Chris Wilson, the data wizard who won the Iowa caucuses for Ted Cruz, puts it this way: “Trump had 6 or 12 data scientists. Clinton had a hundred. But the Clinton operation, you can’t be critical of it. They had the world’s greatest marketing and sales team trying to sell boxed wine to rich people.”
Yet tucked away in the results is a story that suggests organization can actually have quite large effects. Because the data operation built by Wilson and run by the Club for Growth in Wisconsin helped move 345,000 voters into Republican senator Ron Johnson’s column.
Johnson’s was supposed to be one of the two easiest GOP Senate seats for Democrats to capture. Of the 30 public polls taken during the campaign, Johnson trailed Russ Feingold in 29.
But the Club for Growth thought that Johnson’s situation wasn’t as bad as it looked. They had surveyed the race last year, and while their private poll agreed that the incumbent was down by 8 or 9 points, they learned that this was partly because he hadn’t solidified his conservative support. They also found that voters were open to the possibility that Feingold might be too liberal.
The advocacy group kept an eye on the race and, using pollster Jon Lerner, went back into the field in August of this year. While the public polls showed Feingold with an 11-point lead, Lerner’s detailed likely-voter survey suggested the race was much closer. (His poll also showed Trump at minus-11.) Instead of dumping all of their money into broad television ads, the club opted for a data-driven approach.
On September 28, they constructed a large-scale model of the Wisconsin electorate. They began with a 12,000-person survey, which combined robocalls and live interviews. The poll included detailed questions about voter preferences, issues, and demographic characteristics and yielded granular descriptions of two important groups of voters. The first were the “persuadables”—that is, voters who were either undecided or not firmly decided. The second were soft Johnson voters—that is, voters who naturally leaned Johnson but were less likely to turn out and vote.
Club president David McIntosh explains that they then ran a “data match,” comparing the results of their survey against voter lists and other pieces of individual information, such as magazine subscriptions. “We identified people with similar characteristics and found a total bucket of persuadables that was 850,000 people. And we also found a bucket of Johnson voters who just didn’t show up all the time and needed a different message to motivate them to the polls. There were 55,000 of them in this bucket.”
What the group now had was a targeted universe of almost a million individual voters—with names and addresses—who were gettable for Johnson. And in early October, they went to work on them.
Now they bought television ads, but instead of simply blanketing the airwaves, they hunted for their persuadable voters. They found that satellite TV providers gave them the ability to target specific houses, which was a boon. “We knew we had money to buy TV messaging in the two biggest markets, Milwaukee and Green Bay,” McIntosh explains. “And 70 percent of the persuadables were covered by those markets.” That left the 30 percent of the list who were either cord-cutters or lived outside of the two big markets. They went after those voters online, with digital ads.
This digital advertising was targeted specifically to the individual people on the persuadable list who were outside the reach of their TV ads. “We ran four different messages to these voters through our persuasion campaign,” says Club for Growth digital director Stacy French. The ads focused on tax increases, government spending, the problems of career politicians, and raiding Social Security.
This process, known as A-B testing, is a way to find which messages resonate best. The club watched the click-through rates in real time and after seeing that the “career politician” and “raiding Social Security” ads garnered the best response, shifted all of its online resources to them. They also tested the efficacy of different delivery platforms against one another—for example, pre-roll YouTube versus Facebook—so that they could maximize their per-dollar impact.
The results manifested almost immediately: Public polls began tightening the first week of October. After two weeks of the digital push, the club did an internal refresh poll. They found that 218,000 people from the persuadable bucket had moved to “likely Johnson.”
So they pressed on with the targeted data strategy for two more weeks before pivoting to turnout and early-vote messaging in the final week. Here, too, they A-B tested messages and found that pairing one of their core messages with “don’t forget to vote” was more effective than “don’t forget to vote” on its own.
By this point, Johnson had moved to within the margin of error in the public polling. The final internal poll showed that the Club for Growth’s data operation had moved 344,855 people out of the universe of persuadable voters and into Johnson’s column.
Johnson’s margin of victory over Feingold was just under 100,000 votes. It’s an amazing accomplishment.
And there’s one more wrinkle to appreciate: Johnson ran 70,000 votes ahead of Trump. Trump’s margin of victory in the state was 27,000 votes. It is entirely possible that the Club for Growth’s data strategy didn’t just win the election for Ron Johnson, but pulled Trump over the line in Wisconsin, too.
Jonathan V. Last is a senior writer at The Weekly Standard.
