Audio By Carbonatix
Sergio Bendixen and the polling creature that ate your telephone
Surveys have become an intrusive fact of modern life. Every day, it seems, the news carries word of some new poll that shows Americans favor this candidate or that policy, are eating more or reading less. And those are just the polls released to the public. Many more are conducted for private consumption by businesses or other interests. As you read this, phones are ringing in homes across America, a pollster on the line hoping to reel in a respondent with these reassuring words: “I can assure you that we are not selling anything and that the interview will only take a few minutes.” We’re saturated by polls.
That, however, didn’t stop us from jumping at the opportunity to produce our very own New Times poll. And why not? Polls, if they’re done right, can provide a fascinating look at what people are thinking, a statistical snapshot that can capture the mood of a community. Often they affirm conventional wisdom; occasionally they give voice to people who otherwise would not be heard. A survey can be a valuable tool for many, from medical researchers to product designers to journalists. “It gives you a chance to speak for groups of people,” says Sergio Bendixen, the prominent Miami-based pollster. “Sometimes it allows you to speak for the electorate, for all Hispanics, for the people of Miami-Dade County — and that has impact on the way our society runs, the way that our world is, the decisions that are made.”
Bendixen made this New Times poll possible. In fact it was his idea. (More about that in “Behind the Scenes”) We liked the idea of fashioning a poll that would indulge our journalistic interests, and we liked the fact that he would be administering it. In Bendixen we had the quintessential Miami-Dade resident (born in Peru, educated in the U.S.), whose skills have made him uniquely qualified to navigate the cultural mélange that is Miami. He pioneered Spanish-language polling in this hemisphere. For years he monitored elections for the fledgling Spanish International Network, which has since become Univision. The job meant dodging bullets in places like El Salvador, Guatemala, Peru, and Colombia. As the number of Hispanics in the U.S. began to grow dramatically, Bendixen brought his talents back home and, after several more years on Spanish television with Univision and later Telemundo, he started his own firm, Bendixen & Associates. Today he’s considered one of the best in the business, a meticulous planner and superior statistician. “I’ll be doing math on a calculator and he’ll be doing it in his head,” says Susan Vodicka, who has worked with Bendixen for years and has nestled her consulting group in his office.
Mathematics may be the foundation on which polling is built, but it is the analysis of the resulting numbers that often reveals meaningful information. Bendixen has made a name for himself in that regard, especially when it comes to Hispanics, the nation’s largest minority group. His insight and experience have caught the attention of the upper echelons of the Democratic Party (he’s a partisan political animal) as well as the nation’s top think tanks. “Sergio’s a brilliant pollster and an even better strategist,” says Rob Schroth, Bendixen’s former partner and now his chief competitor in the South Florida market. “If someone wants to get to the bottom of an issue, they’d be very smart to hire him.”
For generations straw polls once measured the opinions of people on street corners, at county fairs, and in town plazas. While they yielded interesting results, the limited sample gave no indication of the overall population’s opinion. Only in the Thirties, when market researchers began using probability samples, were the views of the masses reliably approximated.
As opposed to straw polls, probability samples allowed researchers to target a specific population and measure the likelihood of any single person’s participation in that poll. They could also extrapolate to the general population from the results and quantify the margin of error.
George Gallup and Elmo Roper saw entrepreneurial possibilities in the new techniques, and soon attracted the attention of President Franklin Delano Roosevelt, who used polls to measure support for his New Deal programs. Other presidents followed his lead. By the Sixties John F. Kennedy had a full-time pollster on his staff, and polling had become a big business on which political leaders relied. Bill Clinton was polling everything from healthcare to where he should go on vacation.
Many people believe incessant polling shapes policy. Others say polls shape public opinion. In either case, politicos are paying attention. Even the Bush administration, which is publicly dismissive of polls, surveys the public on virtually every important subject, from Iraq to Social Security. “Leaders ignore public opinion at their own peril,” says Scott Keeter, associate director of the Washington-based Pew Research Center for the People and the Press. “They’ve got to keep their eye on it.”
Polling can make or break presidents. Witness the Monica Lewinsky scandal. Keeter relates that pundits thought Clinton was “toast” when allegations emerged of sexual impropriety in the White House. Then pollsters went to work and found that people just wanted Clinton to face up to his wrongdoing. Clinton stopped lying and public opinion shifted in his favor. “I would argue that Clinton survived in office because of the polls,” says Keeter.
Advances in information technology have been both a blessing and a curse for those in the polling business. Sophisticated computer software has made sampling more precise and surveying much more efficient and cost effective. The result has been the rapid proliferation of polling. News organizations, political pollsters, special-interest groups, and businesses are swamping the telephone lines and, more and more, the Internet. A backlash was inevitable. “The biggest problem today for polling is access,” says Columbia University political scientist Robert Y. Shapiro.
Shapiro reports that pollsters used to get a 50 to 60 percent response rate, but these days are lucky if they get 30 percent. Voice mail, caller ID, and other filtering devices have combined with growing public hostility to create serious problems for people whose businesses rely on the telephone. University of Virginia political scientist Larry Sabato says this makes it very difficult to draw scientifically acceptable samples. “How do you draw a sample today to really reflect the population given the massive refusal rate?” he asks.
Complications aren’t limited to busy signals and angry hangups. Sabato believes respondents will say anything to please an interviewer, even if they don’t know the answer. “I’ve seen detailed polls about NAFTA [the North American Free Trade Agreement]. Get real! Ninety percent of Americans don’t know what NAFTA is,” Sabato says. “How do people react to those sorts of things? The answer is that they generally make up an opinion. They listen to the verbal and nonverbal clues [from interviewers] and then give the answer they think you want to hear. Nobody wants to look stupid.”
Bendixen ran into this very problem recently. In a poll of Hispanic voters for the New Democrat Network, a center-right Democratic organization, many respondents confused actor Erik Estrada with Miguel Estrada. Erik played Ponch on the popular television show CHiPS. Miguel had been nominated by the Bush administration for a seat on the U.S. Circuit Court of Appeals in Washington, D.C. (He has since withdrawn his name.) Asked whether he supported Miguel’s nomination, which was being held up by a Democratic filibuster, one of the respondents said he was a good cop and so he’d make a good judge. Another said he didn’t realize the actor was also a lawyer. The majority of respondents said they didn’t have enough information to answer the question.
Language is also increasingly an issue, especially in critically important electoral states like New York, California, and of course Florida, where significant portions of the population speak little or no English. “We have to do huge oversamplings of Latinos to try and get likely voters,” says Susan Pinkus, the Los Angeles Times‘s in-house pollster. “And not every media or polling group takes the time and money to do it.” The same is true in Miami-Dade County. “Any pollster who doesn’t use fully bilingual interviewers is destined to fail miserably,” says Rob Schroth, who conducts polls for the Miami Herald, among other news organizations.
Bendixen agrees. “[Miami is] like polling two or sometimes three different worlds,” he says. “And we’re moving in the direction of the number of worlds expanding. You poll the Hispanics in Spanish and the Hispanics in English and they give you two different answers to the same question.” A good pollster, he adds, strives for simplicity and avoids jargon in his questions, which helps when translating the poll into different languages. A competent pollster also tries to avoid sending messages to the respondent. “The objective of good polling is to make the questions so balanced that no one can figure out what my agenda is, if I have one, or who the client is or who paid for the poll. But also to not give either side of the argument any advantage.”
Bendixen has been honing his skills for more than 30 years. In 1972 he volunteered to work for Democratic presidential nominee George McGovern. Bendixen had been devouring newspapers since he was a teenager but had no formal experience in politics. In fact just a few months earlier he was about as far away from the U.S. presidential campaign as you could get: He’d earned a chemical engineering degree at Notre Dame and was traveling through South America with his cousin. The two were on a skimpy budget so his cousin made papier-mâché dolls they sold on city streets and at village fairs. Sometimes the money got them to the next town, sometimes it didn’t. “There were times when I thought I was going to starve to death or freeze to death,” Bendixen says from the comfort of his Coral Gables office. “It made me a lot stronger as a person. It made me feel like I could withstand strong challenges.”
He returned to his middle-class existence in Florida confident enough to pursue a passion he thought he would only be able to observe from the sidelines: politics. His timing was perfect. The McGovern ticket couldn’t find any volunteers for its Miami office. “Absolutely no one wanted anything to do with it, let alone run it,” Bendixen says of the McGovern campaign. “So they ended up giving it to a 23-year-old kid who had never really done much but was willing to work for $50 a week and put in a lot of energy. And what we built during that campaign still stands.”
Indeed the McGovern ticket brought together Bendixen, his business partner Susan Vodicka, and Mike Abrams, who would go on to become a state legislator. The three of them helped push Jimmy Carter to the top of the Democratic Party in Florida and then to the presidency. Bendixen later worked as press secretary for Rep. William Lehman and led California Sen. Alan Cranston’s spunky bid for the Democratic presidential nomination in 1984, during which Newsweek said he’d single-handedly “outflanked and outhustled” Walter Mondale’s campaign. Bendixen’s strategy during the campaign: publicize straw polls that showed Cranston had strong support among Democrats.
But not everything has been roses. In Nicaragua in 1989, Bendixen & Schroth, working for Univision, wrongly predicted a Sandinista landslide in the presidential elections. The Sandinista candidate lost by fifteen percent. It wasn’t just Bendixen who was chagrined; many major U.S. newspapers followed his lead in anticipating a Sandinista victory.
Bendixen has also been accused of drawing too much from too little information, a common gripe against pollsters. Still he has a reputation for being fair. “My job is to figure out how to nonmanipulate, how to be objective,” he says. “So I don’t stay up at night trying to think of the best way to get an answer that’s going to be meaningless. You don’t even make your client happy.”
He has long since recovered from the Nicaragua debacle. In the last five years he’s built a two-million-dollar business with a client list that includes the Cuba Study Group, the Pew Hispanic Trust, and the New America Alliance. Two months ago Bendixen met with twenty Democratic senators to show them how to win the Hispanic vote against President Bush. Tom Daschle, John Kerry, and Hillary Clinton sent personal notes thanking him. Last year Bendixen’s polls revealing a significant moderation of views toward Cuba among exiles sent tremors through el exilio. Although many were angry at the results, few faulted the methodology.
Bendixen’s hands-on approach to the business is all-consuming. He spends at least half his time traveling to see clients. He’s not married. “His work is his life,” one colleague says. “That’s all he has time for.” In the last month alone he’s been to Guatemala, Ecuador, Mexico, Puerto Rico, Los Angeles, Jacksonville, Tampa, Baltimore — and 2800 Biscayne Blvd.
Behind the scenes:
It’s not quite as bad as sausage-making but …
This project began with a phone call about a television cop show. A recent poll garnered national attention when it revealed that many Hispanics had confused Erik Estrada, the television actor, with Miguel Estrada, the lawyer who was a Bush administration nominee for the federal appeals court. The man who had conducted the poll was Sergio Bendixen, whose company, Bendixen & Associates, is located in Coral Gables.
An inquiry about that case of mistaken identity led to more involved conversations about polling generally. Eventually we asked Bendixen if we could follow one of his polls from start to finish, then write about it. His response was unexpected. He suggested we do a poll together. New Times and Bendixen & Associates could formulate the questions, we would publish the results, and he would pick up the tab — a hefty $25,000.
Sounded good to us, but of course there had to be a catch. What was in this for Bendixen? In part, he explained, he welcomed the opportunity to open the polling process for scrutiny because he gets so many questions — and too many complaints. “I obviously have pretty good confidence in my business,” he said. “I just think of what we have accomplished in Miami in setting up a professional business…. To me it’s a great accomplishment, and I’d like it to get publicized. I thought if we do a public-opinion survey, it serves our purposes and it serves your [New Times‘s] purposes.”
With motives fully disclosed — access to the polling process and a countywide survey for us, publicity for Bendixen — we moved on to developing questions, and immediately ran into problems. The Miami-Dade mayor’s race was an obvious area of interest for both of us. But Bendixen is a close friend of mayoral candidate José Cancela (who was also his boss at Telemundo) and is considering working for his campaign. If our poll touched on the mayor’s race, he felt he would have a conflict of interest. We agreed.
Another subject we hoped to explore was the Cuban-exile community’s thoughts about recent developments in Cuba and evolving U.S. policy toward the island nation. Again Bendixen couldn’t oblige us. For the past three years he’s been working with the Cuba Study Group, a nonprofit organization of wealthy and influential Cuban Americans devoted to studying el exilio and promoting nonviolent change in Cuba. Because of that ongoing relationship, and the likelihood of more polling on the subject of Cuba, he wanted to avoid any potential conflicts.
“Cuba polling is very sensitive,” Bendixen noted. “You’d be a fool to work with different groups on the issue.” He feels the same about working for multiple political parties. He’s a staunch Democrat, but even if he weren’t, he would avoid polling for more than one party. “Everyone has his own code of ethics,” he said, “but if you’re known to be a pollster who works for both parties — and there are some — your clients are not going to trust you as much. You’re going to ruin a relationship between the pollster and client. They have to be able to speak to you with a lot of confidence and know that the information they’re giving you is not going to be used against them somewhere down the line.”
Had New Times commissioned a poll, such restrictions would have been unacceptable. But this project was a partnership, so compromise was appropriate. (New Times did maintain complete control over all the material published here.) So we decided to continue developing questions. Bendixen concentrated on the airport, transportation, and FCAT, the statewide testing procedure for students in public schools. They were buzz issues, he said, especially the airport: “The airport has become the symbol of everything that’s wrong with local government.” We were thinking about trust in government, quality of life, and Haitian and Cuban refugees.
After some back and forth fashioning the questions, Bendixen subcontracted the Atlanta-based firm SDR Sampling to obtain a list of possible respondents. SDR sells lists to clients who want to target specific geographic areas, demographic groups, or consumers. They can furnish names or phone numbers or both for a survey of people who have arthritis almost as easily as they can provide lists of people who recently purchased a new home. We were targeting the county’s general population, so SDR generated a “random-listed” sample from a computer program that gives every resident in Miami-Dade County with a telephone an equal opportunity to be part of the survey. The list included the name and phone number of each person, but excluded businesses and unlisted numbers (of which there are many in the county). “Polling works,” Bendixen pointed out, “because researchers have found that people with unlisted numbers aren’t any different from people with listed numbers.”
Bendixen purchased a list of 8000 telephone numbers from SDR, from which we hoped to obtain at least 400 respondents. Pollsters normally use a list about ten times the size of the target sample; since Miami-Dade has a lower response rate, Bendixen requested more. Most analysts consider 400 an acceptable sample size, although large media organizations usually poll 1000 people or more. The larger the number of respondents, the lower the margin of error and the more accurate the analyses of subgroups like African Americans, Spanish-only speakers, Anglo women with incomes over $50,000, and so on. The minimum number of respondents required for a subgroup to be considered valid is 50.
Next Bendixen contracted with Miami-based National Opinion Research Services (NORS) to call people from that list. NORS, one of a handful of research phone banks around the nation, likes to refer to its operators as “professional data collectors,” an effort on the industry’s part to distinguish itself from telemarketers. Increasingly telemarketers are employing a tactic known as “sugging,” selling under the guise of market research, to trick respondents. The resulting confusion and growing animosity is costing companies like NORS and their clients some serious money. When respondents hang up in anger, more calls are required to achieve the needed target figure, and that becomes expensive.
NORS conducts more than 100,000 telephone surveys per year, and several thousand more in-store, door-to-door, one-on-one, and mail polls. The company’s client list is secret, but Dan Clapp, the fit-looking baby boomer who owns NORS, said he works with all political parties and competing companies, “sometimes at the same time.” NORS maintains a staff of 170 people, all but a handful of whom are bilingual in Spanish and English. The company also employs a woman to conduct interviews in Haitian Kreyol, but Clapp said there isn’t enough work to keep her on staff full time. We reached just two Haitians for our poll, both of whom answered the phone in English. “Even if they’re struggling, they prefer to do it in English,” Clapp said of Kreyol speakers. He would know. He began his polling career going door-to-door in Little Haiti for FIU sociologist Alex Stepick. “It’s difficult to interview by phone with the Haitians because they always think there’s a slant against them,” he added.
Interviewers at NORS get one day of training that includes role-playing and pep talks about how, as Clapp put it, “they’re on the front line.” He emphasized that interviewers should be “plain vanilla” on the phone. “We teach them that there are no right answers,” he said. “They’re taught to read the script word for word without interruptions or corrections.” Still mistakes happen. Clapp has caught interviewers ordering respondents to be quiet, or telling them they were stupid.
Clapp charges Bendixen for the size of the call list and the amount of time he expects the average survey to take. The interviewers are strictly monitored to make sure they don’t let respondents drag out the conversation. “After twenty minutes you’re getting garbage,” Clapp explained; more important, they’re costing him money.
On August 25, the first day of data collection, we went to the NORS office, located in a gated office park on NW 107th Avenue near SR 836. The phone bank is set up in a long, narrow, carpeted room with six rows of fuzzy cubicles. That night it was nearly full of mostly young Latinos staring at computer screens and talking softly into headsets. Four supervisors paced behind them, eavesdropping on their style and delivery and coaching them through any problems. If someone paused too long, a supervisor would remind them of the time by tapping them on the shoulder.
At one end of the room is mission control — a large office with an eight-foot-long panel of windows. Inside are two computers, two phones, and head supervisor Luis Montenegro, in his midthirties and all serious business as he oversees the action. His computers show how long each caller is on the phone and who is on break, and track the results of the survey in real time. For example Montenegro can see how many respondents think Alex Penelas has been doing an excellent job, or how many Hispanics have been polled.
After a quick tour of the office, Montenegro allowed us to monitor some of the calls. (A representative from Bendixen & Associates routinely does this when NORS conducts a survey for the company.) First observation: It’s difficult to get people on the phone, which is why calling hours commonly are restricted to weekdays between 5:30 and 9:30 p.m., Saturday between 10:00 a.m. and 2:00 p.m., and Sundays between 5:00 and 9:00 p.m. Through two hours of calling on this Monday evening, Montenegro said they were averaging two and a half completed surveys per hour per caller. That meant 50 percent of the time was spent just making the phone calls. If that sounds like a lot, it’s quicker than it once was. These days the computer makes the calls, ensuring greater efficiency, accuracy, and randomness of the sample. Nevertheless it takes time. One typical spree of calls went like this:
First call: “I’m not interested,” a man said before hanging up.
Second call: Answering machine.
Third call: Busy.
Fourth call: Busy.
Fifth call: “Okay, okay,” a man said but insisted that the interviewer start “de una” — right away.
Respondents have ways of ducking even if they do answer the phone. One man said he didn’t speak English but still refused to participate when the interviewer offered to do the survey in Spanish. Another said no one over eighteen was home, even though it was obvious by her voice she was an adult.
Some respondents create a different problem — they want to engage in conversation with the caller. “They have nothing better to do,” Montenegro laughed. Other respondents become distracted or bored during the questions. One man yelled at his kid while the interviewer posed a question. Another man repeatedly answered “Good” to questions 19 through 23, seemingly without thinking.
Using our questions, Bendixen designed the poll to move from the simpler to more difficult questions within a subject area, and attempted to avoid jargon and loaded queries. For example, question number five — regarding the county commission and special interests — contains nearly the same number of words on the positive as the negative side of the query. “It’s as balanced as anything can be,” Bendixen said.
Yet there were problems. Respondents, in particular those speaking Spanish, often began with qualifiers such as “me parece que” (“it seems to me”), which inevitably was followed by a justification for a response that was being formulated as they spoke. Some long-winded questions, like number thirteen regarding the FCAT, forced interviewers to speed up and perhaps lose some of the respondents. “It shouldn’t have so much importance,” one woman said of the student test. “It should only say whether the person is weak.” So you’re not in agreement, the interviewer probed. “Yes,” she responded, but seemed confused. The interviewer, however, moved on to the next question.
Respondents also frequently added qualifiers that could never be registered in a survey like this. On question 27, about special treatment for Cubans, one woman said Cubans should get special treatment. Then she added, “But so should everyone else.” The same question often required the interviewer to interpret or change the answers. One man said, “What I think is that the wet-foot, dry-foot policy is the most absurd thing in the world.” But as the interviewer was inputting “No,” the man added, “They should just get rid of that law — it should be either wet foot or dry foot.”
Confused, the interviewer attempted to clarify. Finally the respondent said, “They should get special treatment.”
So, is that a yes?
