Judging the influence of opinion polls

Inbal Bracha might not personally know the people she bombards with probing questions, but she can size them up within seconds. The executive, for instance: "I can hear in his voice that he's a manager - it's a very formal voice. It's a very confident voice. He's very decisive." Hunched over the phone at the desk behind Bracha, Oded Blech has a list of the different personality types he encounters. There are the people with poor excuses (out shopping for the whole night). There are the people who lie (the father whispering loudly that he's not home). There are the poor disciplinarians (the kids pick up and hang up the phone throughout the conversation). And then there are the husbands who ask their wife every question before answering. "I'm thinking about creating a movie about this job," says the 18-year-old aspiring filmmaker. But Blech isn't on a movie set. He's executing one of the major components of election season: the political polls whose results fill news pages and TV screens. And they are supposed to be as far from fiction as possible. "It's kind of a side effect of the political stuff. It's supposed to give you a reflection of what really happened and not create a reality," says Sima Aharoni, manager of the quality department of Dialog, which conducts political polls on behalf of Channel 2. "They give information to people," says Tel Aviv University statistics professor Camil Fuchs of the Dialog political surveys he supervises. Mainly, he explains, they tell the public who's popular. That, in turn, can either cause people to lend support to a leading candidate based on the "bandwagon" effect, or to back a trailing figure out of "underdog" sympathies. But Fuchs describes the influence of polls as "marginal." He maintains that "political surveys don't have a large effect on how people vote. They reflect how people think." Still, Fuchs acknowledges, they don't always accurately gauge what will happen. He quips that ahead of the elections, he can only predict one thing for sure: the results will be different from what the polls indicate today. "Things change," he says. "Things change all the time." But one thing that doesn't change is the problem of "biases" that creep into the surveys, notes Fuchs. For one thing, he says, "we're forcing [the subject] to make up his mind and he may change it tomorrow." For another, some people refuse outright to respond to telephone polls, while busier people are often unwilling to participate in longer surveys. And certain populations can be harder to reach. To correct for the latter, callers such as Bracha and Blech also keep track of demographic information, which Dialog uses to weight the results. Fuchs gives the example of voter representation: if 5 percent of the population voted for the National Union in the last elections, but only 2% of the interviewees say they voted for the National Union, he multiplies their answers to boost their percentage among the surveyed population. For this process, he stresses, past voting rather than current allegiance is key, since the past is what is known. The ultra-Orthodox, Russian immigrant and Arab populations are key constituencies whose representation often has to be calibrated. Then there is the issue of the "desire to please," Fuchs continues. That's the instinct on the part of those being surveyed to answer the question in the way they perceive as more socially acceptable, for instance, rather than as they truly feel. He points to a telephone survey done by one newspaper following the last elections showing that 93% of the public had voted - when the turnout was only 67%. "That's a fantastic example of the desire to please," he says, as well as of a newspaper failing to contextualize responses. To combat such biases, the callers need to be as "neutral" as possible, according to Aharoni. "Everything is okay," she explains. "You're not supposed to react. You're not supposed to give any feedback to the person." And to overcome the refusal factor, she says, "You have to smile, to sound convincing, to really appreciate" their responding. Bracha, an energetic engineering student, does smile over the phone. And she has a strategy for getting people to answer - necessary as the small room she and a dozen or so other phone operators work from resounds with the clicking of phones being hung up rather than interviewers reading off questions. Bracha speaks politely and quickly. She tells them it will just take a minute to complete the survey. And, most importantly, she explains, "I give the person who is being interviewed the feeling that he is very important." In a regular survey, the latter might be a stretch, since Bracha can usually just move down to the next entry on her list of hundreds of names. But for the poll of Likud Central Committee members that Bracha worked on this week, each person mattered. Usually phone operators make 10 calls to reach one person, but the Likud Central Committee consists of only 3,000, and Dialog needed to reach 500 of them for to get a large enough sample size. That cranks up the pressure, which is already intense for political polls dictated by media deadlines. They are generally conducted on the same day they are ordered by the news outlet, between 4 p.m. to 9 p.m. Making this week's poll even more challenging, Aharoni notes, is that Likud Central Committee members are highly sought after by pollsters. Soon after making this comment, Bracha speaks to a member who says he has already been questioned by Ma'ariv and Yediot Aharonot. Another person asks her who the survey is being done by and only answers after hearing it's on behalf of Ha'aretz. Aharoni explains that he wants to be assured the poll isn't an internal survey being done of behalf of one of the candidates. Internal political polls often turn out results more favorable to the candidate conducting the survey, and Fuchs warns that they have to be taken "with a grain of salt." It's not that the results are falsified, he clarifies, but that the questions are phrased in ways that elicit certain results. He says queries have to validate both a yes and no response. One question in the original Likud questionnaire, for example asked whether the so-called party rebels should be "punished." Fuchs changed it to ask whether they should be "punished or encouraged." The sequencing of questions is also important, he adds. He points to a surprising Central Bureau of Statistics survey finding that 82% of Israelis reported being happy with their lives, despite the economic and security situation. In that case, he says, the CBS first asked about the respondent's health, employment and children's well-being. After answering positively to those questions, fewer people would be inclined to say they were unhappy than if just asked whether they were happy without the opening questions. Despite these flaws, Fuchs defends public opinion polls as being good indicators of trends over time rather than facts at certain points in time. When the same questions asked in the same manner over a period of time show an increase in the rating of a candidate, it's clear that "he's better now than he was two weeks ago." Though there has been criticism of polls for unfairly influencing elections, Fuchs maintains that political polls "bring news" and that "the people have a right to know, for better or worse." But even Blech doesn't think people have the right to know what he's thinking, which puts him in the category of those who screen their calls. "I don't answer," he says. "I don't like to be bothered."