Colombian voters’ rejection on Sunday of a peace agreement to end the country’s 52-year conflict with the Revolutionary Armed Forces of Colombia, or FARC, was a backbreaking defeat for the Bogota government. It was also an embarrassment for Colombian pollsters, who were so confident the accord would be approved that they muffed their country's most consequential election in a generation.
Four leading Colombian pollsters projected the “Sí” (Yes) side would win the referendum with support above 60 percent. Even two less optimistic polls showed the agreement being backed by well over 50 percent of voters, with a double-digit victory margin. The actual result? A 50.2 to 49.8 percent victory for the “No” side, though both the Colombian government and the FARC have promised to maintain a ceasefire while they work to save the deal.
Colombia’s referendum polls were much further off base than widely criticized polls taken in the United Kingdom before its vote to leave the European Union in June. One of those errant polls had the "Remain" in the EU option winning by 10 points, though most of the polls that erred did so by more modest margins. (The final tally saw a nearly 4-point victory by the “Leave” camp.)
It’s been a rough couple of years for pollsters around the globe. Polls failed to foresee the Conservative victory in the UK general election and forecast a difficult reelection bid for Israeli Prime Minister Benjamin Netanyahu in a race he won comfortably. Polls also anticipated a tight contest in Scotland’s independence referendum when the option for staying in the UK prevailed by more than 10 points.
"I wouldn’t say polling is getting less accurate, but it’s getting harder,” says Joshua Dyck, a public opinion specialist at the University of Massachusetts Lowell. “It’s more difficult to do a high-quality poll and get a representative sample.” That, he continues, is because the switch from land lines to cellphones has made it much more difficult for pollsters to obtain random samples that are truly representative of all of a country’s voters.
In an interview Wednesday, Dyck noted that referendums like the Colombian peace vote or Brexit "are among the hardest things to poll accurately," since voters tend be more unpredictable on issue elections than on votes involving candidates or parties. It can also be extremely difficult to accurately predict turnout for referendums, unless there have been past such votes in individual countries that can provide a clear track record.
Colombian pollsters openly concede to badly bungling their pre-referendum surveys, though they offer a variety of explanations. “It was a huge surprise for all of the pollsters,” adds Javier Restrepo, director of opinion research for Colombia’s IPSOS polling agency, which had Sí winning 66 percent to 34 percent. “No one thought it would be very close.”
Carlos Julio Lemoine, president of the Centro Nacional de Consultoría polling firm, which had projected a 65 percent to 29 percent win for Sí, attributed the polling failure to “the spiral of silence.” That’s a theory by the late German political scientist Elisabeth Noelle-Neumann, which posits that people prefer to keep quiet rather than publicly express views that they fear aren’t held by the majority.
Restrepo largely agreed. “In the weeks before the referendum, people who favored ‘no’ were called enemies of peace,” he said. “If someone expressed a ‘no’ sentiment on social media, they were immediately attacked. In the mass media, there were also attacks against opponents of the agreement.” In retrospect, he says, it’s clear that many opponents of the peace deal kept their opinions to themselves until voting day.
Opponents' reticence might have been especially problematic for IPSOS because it relies heavily on face-to-face interviews to reach working-class and rural Colombians, who can be tough to track down by phone. Restrepo says that in hindsight it might have been better to have conducted a simulated vote in which subjects deposited mock ballots into containers.
The act of voting, even if simulated, might have given interview subjects a greater sense of privacy and cut through the social pressure to say they intended to vote in favor of the pact. Restrepo says IPSOS has employed that technique with good results in neighboring Peru, a country where voters tend to be wary of pollsters due to a history of government spying on individuals.
Colombian pollsters say the other factor that might have tripped them up was the triumphant signing ceremony held by the government on September 26, the Monday before the referendum was held. The elaborate event, featuring more than a dozen foreign heads of state, was meant to trigger a surge of momentum for the Sí forces down the home stretch. But it seems to have had the opposite effect.
“People wanted to know why are they were holding the victory fiesta before they held the vote,” says Andrés Pérez, an executive at the EcoAnalítica firm, which was less wrong than other pollsters, projecting a 56 percent to 42 percent victory for Sí. He says the ceremony had a negative impact on the peace agreement that seems to have cut two ways: “Opponents felt angry that their views were being slighted, and supporters felt they didn’t need to vote because the agreement had already been signed.”
Some Colombians also seemed to have been galled by the prominent role that the FARC leader, Rodrigo Londoño Echeverri, known as Timochenko, was given in that signing event. “I think it bothered people to see the leader of the FARC treated like a great personality shaking hands with Colombia’s president and mixing with leaders from around the world,” Lemoine says. The conflict has claimed some 220,000 lives and displaced 5 million more, meaning that few Colombian families have been untouched.
The government prohibited polls from being published after September 27, the day after the signing ceremony, to avoid influencing the outcome of the vote. So polls didn’t capture impact from the event. “The last week was absolutely critical, and we don’t have data,” Restrepo says.
For followers of elections, whether in Colombia, the UK, or the upcoming US presidential vote, the lesson from the failed peace referendum is painfully simple: “It’s never safe to make assumptions based only on polls,” says Sheyla Dallmeier, a Bogota-based political consultant. “There’s a reason they hold the elections.”
Matt Moffett worked for three decades as a Wall Street Journal correspondent in Latin America and Spain.