When Campaigns Backfire

Campaigns and advocacy groups spend millions of dollars on these direct-mail efforts every election cycle, believing that, with the right messaging, they will win over more recipients than they will lose. But new research shows that these campaigns can backfire, winning the support of the target audience but losing the support of all other voters.

“The problem is, nobody really understands how a given leaflet or ad might motivate people to talk to others and spread information across social networks,” says Georgy Egorov, a professor of managerial economics and decision sciences at Kellogg.

Seeking to explore how messages ripple through social networks and to better understand the full scope of a campaign’s impact, Egorov collaborated with Sergei Guriev of London Business School, Maxim Mironov of IE Business School, and Ekaterina Zhuravskaya of Paris School of Economics.

The economists developed a novel way to simultaneously measure the direct and indirect effects of a political campaign and applied it to study the 2023 Argentine presidential election, which was won by the outsider, antiestablishment, and right-wing candidate, Javier Milei. They focused their efforts on a leaflet campaign that provided expert information about the likely adverse consequences of some of Milei’s extravagant proposals.

Overall, people who received one of the leaflets were less likely to vote for Milei. So at least on the surface, the campaign went as expected.

But the campaign also had an indirect effect on voters in the same localities, which completely upended expectations. Among those who did not receive a leaflet, support for Milei not only increased; it went up so much that it overshadowed the direct effect.

In other words, this particular campaign backfired, ultimately hurting the cause it was supposed to help.

“That the indirect effect was bigger than the direct effect—and also in the opposite direction—is not something that we anticipated,” Egorov says. “It really challenges what we know about the spread of ideology over social networks.”

An ideal setting

Unlike in many other countries, Argentina’s election results are reported at the sub-precinct level, with voters assigned to a specific “mesa” based on the first letter of their last name. Political parties in Argentina are also legally required to make public a list of their members’ names and addresses.

This setup allowed the NGO that the team was working with to know not only which residents voting in a given precinct would receive leaflets and which would not, but also how the groups that received the leaflets and those that did not ended up voting.

“This allowed us to measure how people living in the same area, and differing only by whether the campaign targeted them directly or not, voted,” Egorov says. “Short of observing individual votes, this is an ideal situation to identify the direct and indirect effects of a campaign.”

Together with the NGO, Egorov and his team designed two different leaflets that fact-checked Javier Milei’s policy proposals.

The first design explained that Milei’s plan to abolish the Central Bank and dollarize the economy would cause significant devaluation and inflation. The second one described how Milei’s plan to replace public funding for secondary schools with vouchers would harm the country’s educational system.

A highly unusual outcome

Before the start of the general election, the team mailed 5,000 of each type of leaflet to a random selection of individual residents voting in particular sub-precincts of Salta Province (excluding the capital city), where Milei had received his highest share of votes in the recently concluded primary election.

Overall, the first leaflet about inflation had no impact on voting outcomes, likely because the topic was highly familiar to the voters and discussed extensively in the media.

In contrast, the leaflet about education decreased voter support for Milei by about 20 votes per 100 leaflets mailed—but only among residents who received a leaflet. Among residents in the same precincts who did not receive a leaflet, voter support for Milei increased by about 30 votes per 100 letters mailed.

The net result was an increase of 10 votes in favor of Milei per 100 leaflets mailed.

“When we got the results, we thought, ‘Something highly unusual is happening here,’” Egorov says. “We questioned to what extent this was a statistical fluke.”

When Milei emerged from the first round of the general election as one of the two finalists, the researchers realized they had an opportunity to corroborate their initial findings in a follow-up test. They sent out 5,000 of their “education” leaflets to a random selection of residents—who lived in the same province but did not receive a leaflet in the first phase—ahead of the runoff election between Milei and the frontrunner, Sergio Massa.

Just as in the first phase, the campaign reduced voter support among residents who received a leaflet, though not by as much this time around—perhaps because more voters had made up their minds about Milei by the time of the runoff. In contrast, the indirect effect stayed positive and just as strong: even if leaflets were less likely to change people’s minds, their desire to talk about the election to friends and neighbors seemed to remain just as high.

The team also looked at the effects of the first phase on the runoff vote to see if the effects were persistent or short-lived. “Surprisingly, we didn’t see decay in the runoff,” Egorov says. “If anything, the indirect effect grew bigger. It’s almost as if people had more time to talk to their friends and neighbors about the information they had learned.”

Motivated to talk

Why did the “education” campaign backfire so spectacularly?

One likely explanation, according to the researchers, is that the voters who received a leaflet and disagreed with it were much more vocal about their opinion and more likely to convince their neighbors to vote for Milei.

“Clearly, some voters found the arguments persuasive. But some apparently thought it was an unfair attack against a candidate that they really liked—well, it turns out these voters were more motivated to talk to their friends,” Egorov says. “And they happened to be much more effective than any sort of political campaign might hope to be.”

The team substantiated this theory in a survey of voters long after the presidential election concluded. The survey asked people to imagine either believing or not believing information in a leaflet about a hypothetical presidential candidate and then report how likely they would be to talk to their friends and neighbors about it. Survey respondents who were asked to disbelieve the information were not only ready to talk to more people about it but also more motivated to convince others to vote for the candidate whom the leaflet criticized.

“Apparently, upsetting and outraging people from the other side is easier than energizing your supporters,” Egorov says.

Putting assumptions aside

The study’s findings offer valuable lessons for campaign and outreach efforts, even beyond the political realm. For one, “you cannot always take things that work in the lab into the field,” Egorov says.

The researchers’ “education” leaflet, for example, would have aced the typical focus group or lab-based test that campaigns usually rely on to predict how people will respond to their messaging. But as the study’s findings revealed, such test results would likely have been very misleading when implemented in the real world.

Just as important for campaign and outreach efforts is to consider how their messaging might affect people beyond their target audience. For instance, political campaigns all too often fixate on engaging swing voters and overlook large swaths of the population. For managers and marketers, the lesson is similar: it’s not enough to understand how the target audience reacts; you also need to anticipate who is more likely to talk about your message, and how.

Finally, the study’s findings highlight the importance of putting assumptions aside when trying to understand how people might respond to new information and interact with others.

“When we accepted that our assumptions—and those of our profession—about how things spread over social networks were too simplistic,” Egorov says, “that’s when we learned something new.”