Poli-Sci Perspective is a weekly Wonkblog feature in which Georgetown University’s Dan Hopkins and George Washington University’s Danny Hayes and John Sides offer an empirical perspective on the issues dominating Washington. In this edition, Hopkins examines how an experiment in Wisconsin showed some surprising results when potential voters were canvassed.
In the aftermath of the 2012 election, one rare point of bipartisan agreement has been about campaign tactics: Obama had a significant edge in the ground game, many agree. One piece of evidence for that claim was the Obama campaign’s almost three-to-one advantage in its number of field offices.
Still, persuading a voter face-to-face doesn’t always work out as planned. In fact, as Michael Bailey, Todd Rogers and I show in a new working paper, large-scale efforts at interpersonal persuasion can backfire. Especially with voters who are less politically engaged.
Our working paper details a large-scale field experiment conducted in Wisconsin in October 2008. The effort first involved identifying 56,000 “persuadable” Wisconsin voters who were the only registered voter in their household. They were mostly voters thought to be in the middle of the partisanship scale. Starting with that target population, a pro-Obama organization then canvassed, called, and/or sent mailings to randomly selected subsets; 7,000 voters were slated to receive an in-person canvassing visit, a phone call and a series of mailings, while another 7,000 voters heard nothing from the organization at all. Still other groups received different combinations of the persuasive contacts.
In all cases, the persuasive messages focused on the economy. For instance, the voters who were called or visited in person heard a script that read, “John McCain says that our economy is ‘fundamentally strong,’ he just doesn't understand the problems our country faces.” It contrasted Sen. McCain with then-Sen. Obama, who would “cut taxes for the middle class and help working families achieve a decent standard of living.”
After the various campaign contacts were complete, a survey firm called the voters involved in the experiment. It managed to get a vote preference from just over 12,000 of them, an impressive response rate in this day and age.
By looking at who responded to the follow-up survey, we get our first hint that the campaigning—and the door-knocking in particular—was an unpleasant experience, at least for less politically engaged voters. Consider those voters who hadn’t voted in any of the nine previous federal elections, dating back to the 2000 primaries. If someone had knocked on their door, they actually became 4 percentage points less likely to respond to the subsequent political survey. The response rate for those who didn’t receive a visit was 21 percent, while for those who did it was just 17 percent. Being canvassed was a sufficiently negative experience that it led infrequent voters to avoid an ostensibly unrelated survey.
One of the most consistent findings from recent field experiments in political science is the capacity of an in-person visit to turn out voters on election day. But that didn’t prove true for this sample of Wisconsin voters. Overall, the canvassing effort did nothing to increase turnout. And among that same group of people who hadn’t voted recently, the impact of door-knocking on turnout was actually negative, at -1.3 percentage points. That result is marginally statistically significant —and it gives another indication that trying to persuade unlikely voters can backfire, leading them to disengage further from politics.
Most of the voters in the experiment didn’t take the follow-up phone survey, so we don’t know if they were leaning toward Obama or McCain in the aftermath of the experiment. To analyze the data, we borrowed tools from statistics, public health and other fields to estimate the effects of canvassing when making different assumptions about who failed to respond to the survey. And when we do so, we see a consistent and counter-intuitive pattern: the pro-Obama canvassing appears to have significantly decreased support for Obama in the follow-up phone survey, by one to two percentage points. Neither the persuasive phone calls nor the mailings seem to have left a discernible impression. On the whole, our best guess is that this large-scale effort on Obama’s behalf reduced the number of people willing to say that they backed him. And the negative effects of canvassing were strongest for—yup, you guessed it—those people who hadn’t voted recently.
Here in the United States, there are strong norms about the importance of voting and an in-person visit can increase voter turnout by reinforcing those norms. But those who seek to persuade voters to choose a particular candidate don’t have the same set of shared norms to build from and as a result, they run the risk of alienating voters, especially those who are already disengaged from politics. It’s plausible that the attack on McCain in the persuasive script increased that sense of alienation as well.
When making sense of the unexpected backlash this experiment uncovers, we need to keep the context in mind. As in 2012, Wisconsin was a swing state in 2008, one that saw heavy campaign activity from both sides. The economic message delivered in the experiment was likely to be familiar, and Wisconsin voters might well have passed their saturation point. Also, this experiment targeted voters living in homes with no other registered voters. Doing so made the post-experiment survey more efficient. But it also might have produced a population of voters who were less socially engaged, and perhaps less excited about a stranger paying a visit to talk politics.
In the past election cycle, the Wisconsin Democratic Party chair told journalist Sasha Issenberg that when you are persuading people, “[y]ou’re going to people who are undecided, who don’t want to hear from you, and are often sick of politics.” The backlash we observe in the Wisconsin 2008 experiment certainly bears out that observation.
In theory, campaigns should prefer persuasion to mobilization, since persuasion takes voters out of an opponents’ column at the same time that it adds votes to their column. Mobilization leaves opponents’ tallies unchanged. But as the experiment makes clear, face-to-face persuasion is an uncertain enterprise—significantly more so than mobilization.
Does all this mean that Obama’s allies would have been better off talking up McCain in 2008? Not exactly. Like the effects on survey response and voter turnout, the negative effect on Obama support appears concentrated among infrequent voters. So the people who were turned off were those least likely to turn out. But the experiment does reinforce just how difficult persuasion can be, and just how central targeting is to an effective ground game. And for campaigns and campaign watchers, it underscores that putting boots on the ground isn’t always enough—or even a net positive. If the volunteers in those boots are knocking on the wrong doors, or saying the wrong thing, they might not be helping their candidate.