Sorry, you need to enable JavaScript to visit this website.

IRI's Tom Garrett and Dan Fisk Talk to the Miami Herald about Polling in Cuba

April 9, 2015

How secret poll in Cuba was done
The Miami Herald
By Glenn Garvin

Miami pollster Fernand Amandi was watching a TV newscast shortly after the historic thaw in relations between Washington and Havana was announced last December when he saw Cuban leader Raul Castro brushing off a suggestion that the island’s dissidents might now get a more respectful hearing from their government. Why should they? said Castro. They don’t really represent anybody. They don’t speak for Cubans.

Amandi had two quick thoughts. The first was: In a country that hasn’t held free elections in more than 50 years, who knows what Cubans really think? The second was: Why don’t we ask them?

Two months later, a squadron of interviewers working for his company, Bendixen & Amandi International, were secretly fanning out across Cuba conducting 1,200 interviews for what would be the country’s first nationwide survey conducted by an independent private polling firm since Fidel Castro came to power in 1959.

“We’re very, very proud of this work,” Amandi said of the poll, published Wednesday by his clients, the TV networks Univision and Fusion and the Washington Post. “We’ve done difficult polls in the past, but this one presented some very unusual obstacles.”

Three of B&A’s group of 18 Cuban surveyors quit in fear during their training. Three others had to evade police after they were denounced by Cuban citizens who were suspicious about their questions. In order to meet B&A’s goal of collecting data from all 15 of the island’s provinces as well as the Isle of Youth, some of the surveyors had travel as much as 10 hours on rickety buses just to do 10 interviews.

Opinion polling in any Third World country, where the primary tools of the pollster — telephones and solid census data on demographics — are usually unreliable is always difficult. But in Cuba, where polls are illegal and political repression makes answering questions from strangers chancy, it’s a truly formidable task.

“Cuba is a closed and closely monitored society,” said Tom Garrett, vice president of the International Republican Institute, an arm of the U.S. National Endowment for Democracy — and largely staffed by Republicans — that has conducted its own polls in Cuba. “It’s one of the toughest places we’ve worked.”

The difficulties of polling — in particular, the lack of good demographic data and the difficulty of hiring trustworthy interviewers — had aborted two other fledgling B&A plans to work in Cuba, in 2006. Amandi is reluctant to say exactly what changed for fear that disclosing the details of his operation would compromise the company’s plans to poll again in Cuba in the future.

“I’ll just tell you that we had demographic information we were confident in and we had a vetting process that we were also confident in, to ensure that the confidentiality of the people we interviewed would be protected,” he said.

Venezuelan pollster Joaquin Perez Rodriguez helped design the 79-question survey, which started with non-controversial questions like “what would you like to accomplish in the next five years?” and worked up to potentially troublesome ones like “do you have a positive or negative opinion of Fidel Castro?” Another polling company, Centro de Investigaciones Sociológicas in Mexico, actually ran operations on the island, hiring and training the Cubans who would ask the questions.

B&A officials wouldn’t reveal exactly what the interviewers were told about who they were conducting the survey for, or what they, in turn, told Cuban citizens who agreed to answer questions. But they did say the training was time-consuming.

“Given that we couldn’t take all the interviewers to a hotel without being exposed — all the hotels have government personnel who inform about every little thing — we had to rent rooms and houses in various places around the country,” said Fernando Civera, the project director for CIS-Mexico.

Even then, the training was done one on one rather than in a group class so that one if surveyor was caught, he couldn’t identify the rest. None of the surveyors knew who else was participating.

A similar poll in the United States or Europe, where something over 99 percent of households have telephones, would be conducted by phone using what pollsters call RDD, random digit dialing. But telephones are far less common in Cuba, especially in the provinces outside Havana, and the pollsters doubted anybody would be willing to talk politics honestly with a random phone caller.

Instead, they used a technique known as random route, in which certain addresses are established as “sampling points.” From there the surveyors fan out in all directions to visit, say, every third house for a face-to-face interview.

After a week of testing in which 100 interviews were successfully conducted in three provinces, the interviewers went to work for real between March 17 and 27. As they filed their encrypted reports electronically from Cuba’s Internet cafes, the pollsters were delighted at the results, which showed the same general trends no matter which province or interviewer they came from.

“That was a good indication that the poll hadn’t been compromised in any way,” Civera said. Some of the interviews were taped, with the interviewees’ permission, so the pollsters could spot-check that questions were being asked correctly and without signs of bias, and those were reassuring, too.

“It was really something to listen to the tapes, to hear how matter-of-factly people gave their opinions,” said Fernando A. Amandi, another B&A associate who work on the poll. “For 55 years, the people have been totally silent. And suddenly here came their voices, out of the darkness.”

There were also indications that, in a wired world, Cuba remains largely disconnected. “One of the questions was, what celebrity would you like to visit Cuba?” said Fernando A. Amandi. “I almost fell out of my chair when one of the guys answered, ‘Michael Jackson.’”

Not all the news was good. “About three or four days into it, we got word that three of our interviewers had been detained,” Amandi said. “That was a really low point. It wasn’t just that we feared the poll was compromised, but the moral burden that something we had set up had resulted in people going to prison.”

But the pollsters quickly found that no one had actually been jailed. One of the interviewers had been involved in a fistfight with a citizen suspicious of the questions, another had run away, and a third got out of trouble by paying a quick bribe.

None of the incidents, apparently, were reported to the police. But they raise obvious questions about the reliability of a poll taken in a totalitarian country where expressing an opinion contrary to that of the government can result in anything from loss of a ration card to a jail sentence.

There is relatively little track record for polling in Communist countries, although there is some evidence that the Cuban government has occasionally taken its own surveys to gauge public opinion. One of the few attempts was made during the 1990 election in Nicaragua, when the Marxist Sandinistas were still in power, and it resulted in one of the greatest disasters in the history of polling.

Several American pollsters (including Sergio Bendixen, the president of B&A) forecast landslide victories by 15 percentage points or more for the Sandinistas. Instead, they lost by 14 percentage points. An polling experiment conducted by the University of Michigan’s Survey Research Center during the election pointed to the likely reason.

For a third of their 900 interviews, the Michigan pollsters carried pens with the red-and-black Sandinista party colors. A second group carried blue and white pens, the colors of the opposition. And a third had pens in neutral colors.

The result: The groups of pollsters carrying the Sandinista pens and the neutral pens got erroneous results showing a Sandinista landslide. Those who used the blue and white opposition pens got numbers showing an opposition victory by a margin very close to the real one.

“The inference was pretty clear that the pollsters themselves were not trusted,” Howard Schuman, the now-retired Survey Research Center chief who conducted the experiment, told the Herald last week. “The voters looked at their pens and then told them what they thought the pollsters wanted to hear. And the distrust was so strong that it even carried over to the pollsters with neutral-colored pens.

“I don’t know a lot about Cuba and I don’t know how freely people speak there. I’d guess a lot of the people who speak most freely have either left or are in prison. But I would expect a poll conducted there would face some of the same problems we encountered in Nicaragua.”

“It’s a very difficult problem in Cuba,” agreed Dan Fisk, operations director of the International Republican Institute, which has conducted eight polls on the island. “You’ve always got be to careful about it. The people you talk to are always calculating, ‘Who is this person and why is he asking me these questions and what should I say?’”

To help neutralize the suspicions, IRI interviewers don’t say they’re conducting a poll. Instead, they ask their questions during casual conversations struck up at bus stops and other public mingling places.

B&A’s interviewers encountered some evidence of mistrust. “The Cuban in general answers the polling questions amiably, but in the results you can also see that on some questions there was a fear of answering, and we got more ‘I don’t know’ or ‘no answer’ responses,” said Civera. And 75 percent of the people polled said they “have to be careful about what to say” when speaking in public.

Nonetheless, Armandi believes his interviewers for the most part got beyond any cautiosness on the part of the Cubans who were interviewed.

“Look, half the people said they disapproved of Raul Castro,” he noted. “Half said they disapproved of Fidel Castro. Obama got enormous positives. How afraid could they have been?”