At 1 a.m. on Monday morning, legislator Epsy Campbell was a fraction of a percentage point away from the vice-presidency, but she had another target in her sights: the pollsters who were so far off in their predictions that, she feels, they should “apologize to this country.”
“They didn’t do systematic work, and they didn’t observe the real tendencies,” she told The Tico Times, visibly weary but still with plenty to say outside the Citizen Action Party (PAC) headquarters in the eastern suburb of San Pedro. “If they hadn’t manipulated so many polls, we would probably be at 43-44%, because many people lost their incentive with that story that we were at half (of Liberation’s Oscar Arias)… They’ll have to make substantive changes.”
According to analysts, this is an accurate assessment in an election where most polls showed Arias with a 20% lead over PAC candidate Ottón Solís throughout the campaign, approximately 10% in the days before the election, and 7% in exit polls – allowing the too-close-to-call outcome to take the nation by surprise.
Why were the polls such a poor indicator of the results? Unusual levels of uncertainty and dishonesty on the part of interviewees, and the failure of polls to adapt to the collapse of the country’s bipartisan system, are the factors on the lips of those who interpret and design the polls. Also to blame, they say, are politicians and the media for inflating the significance of poll results.
“Yes, we were wrong.We failed,” Gustavo Araya, poll analyst for CID-Gallup, said Monday. “We couldn’t correctly read what the Costa Rican people were trying to tell us.”
New Era, New Polls?
According to José Carlos Chinchilla, a sociology professor at the Universidad Nacional (UNA) in Heredia, north of San José, the problem is that Costa Rica’s political scene changed drastically since the last elections, but the polls haven’t kept up. Namely, the drastic decline of the Social Christian Unity Party (PUSC), once a pillar of the government along with National Liberation, coincided with the growth of relatively new parties such as PAC and the Libertarian Movement Party.
Because of this, most voters don’t decide how to vote based on lifelong affiliations or family history as in the past, Chinchilla told The Tico Times. Polls measured “the voting intentions of people who’d decided, but that was only 15-17% of people” – and they were almost all lifelong Liberation supporters, and therefore votes for Arias, he said.
“When people began to change, (the polls) didn’t have the technical capacity to get that information from people,” Chinchilla added. “The polls weren’t precise enough, profound enough… They used a technique that was valid during the bipartisan era.” Ana Jimena Vargas, project director for polling company Unimer, agreed.
“Arias had a greater advantage among people who had decided,” she said Monday.
“Among less firmly decided people, Ottón had a greater advantage.”
Both Vargas and CID-Gallup’s Araya insist that Solís’ surprise upward surge was therefore not a polling fluke, but the result of last-minute decisions by a large part of the population – a new trend in Costa Rican politics.
Unimer found 47% of survey respondents had some degree of uncertainty, which Araya called an “atypical” situation.
“The spirit of these elections…was very distinct,” she added. “The issue isn’t whether the polls did their work well. It’s that Costa Ricans are sending a message.”
Another possibility: interviewees may have been lying.
“People expressed a position that didn’t necessarily match their vote,” said sociologist Carlos Carranza. “Perhaps they wanted to hide their vote… or could perceive the ends that (pollsters) were seeking.”
Chinchilla, however, discarded this notion.
“People don’t lie,” he said. “The questions were very closed-off. (Pollsters) didn’t explore other elements… This looks to me like an excuse for private polling companies.”
Another possible cause of the failure of polls to capture public opinion is the vortex of strong feeling surrounding Arias himself – and the reaction of the candidate to his soaring numbers.
“Oscar Arias is such a strong figure that perhaps he created polarization around himself,” said CID-Gallup’s Araya.“People started to migrate… toward the party of Ottón.”
Arias brought some of this upon himself, according to analyst Cerdas, who said the leading candidate was “deceiving himself” because of the numbers. The polls led Arias to devote his campaign to getting Liberation Party legislators elected, neglecting a presidential race that was closer than he realized, Cerdas added.
Often, politicians are most to blame for over-reliance on polls, according to Araya.
“What modern politics shows us is that it changes, it varies,” he said. “But politicians tend to say, ‘Look, I’m winning!’ When the polls are favorable, they support them.”
Seeking a Better View
To improve the accuracy of polls, analysts recommend changes such as including more general questions designed to reveal subjects’ overall political tendencies before asking about how they’ll vote; increasing sample size, in some cases, and frequency of polls to keep up with changing opinion; and, for both pollsters and the media, relying on informal signs as well as hard data to help inform reporting.
“What people are putting on the walls, coded messages… all this expresses a search for an (alternative) position,” Carranza said.
“E-mails, rumors, in the good sense of the word: the language of popular politics.”
For Araya and Vargas, the biggest problem isn’t the polls themselves, but rather people’s over-reliance on the results they show.
In fact,Araya, whose company’s poll published Feb. 1 showed Arias in the lead by 19%, said he’s grateful that Sunday’s results, or lack thereof, may encourage a more realistic approach to polls.
People have attributed “a supernatural power to us,” he said. “Thank God, with this election, such an (attitude) falls away… The poll is just a photograph of a population at a determined moment, not before, not afterward.”
Vargas, whose company last showed Arias with an 11.1% edge over Solís on Feb. 2, agreed.
“I’d love it if the lesson here would be to learn that polls aren’t crystal balls, but they shouldn’t be buried as a tool,” she said, adding that focusing solely on whether polls were flawed would be “an injustice to the political reality of this country.”
This isn’t the first Election Day that served up a nasty surprise for Costa Rican pollsters. In 1998, exit polls showed presidential candidate Miguel Angel Rodríguez of the Social Christian Unity Party (PUSC) with a 10% lead, causing his opponent, José Miguel Corrales, to concede before all the ballosts were counted. The daily Al Día published a special edition with the headline, “Rodríguez Sweeps.” However, PUSC received an unpleasant surprise when the official results proved to be far different from the polls’ projections.
In the end, Rodríguez squeaked by with a 2% victory (TT, Feb. 6, 1998).