15 August 2002
Source:
http://usinfo.state.gov/cgi-bin/washfile/display.pl?p=/products/washfile/latest&f=02081402.nlt&t=/products/washfile/newsitem.shtml
US Department of State
International Information Programs
Washington File
_________________________________
14 August 2002
(Modern techniques allow for accurate prediction, greater involvement) (1350) By Benjamin Gross Washington File Staff Writer Washington--Their results appear everyday in newspapers and television programs around the world, influencing the decisions of political leaders and ordinary citizens alike. Still, despite their widespread acceptance, the mechanisms behind public opinion polls remain a mystery to the general public. They wonder how a survey of only 1,000 people can accurately reflect the opinions of an entire nation. John Zogby, an internationally respected pollster, suggests that people should be aware of how polling works, because it is first and foremost a participatory process. "It offers people a sense of connectedness whether they are isolated by modern life: the cubicle, the long commute, the gated community; or isolated by traditional life: the farm, the desert, or whatever. People with views are connected to each others via the polls." All public opinion polling aims to record opinion in a sample that accurately represents a wider population. Pollsters seek a truly random sample -- one in which every member of a population has an equal chance of being selected. Assuming a large enough group is selected, the views of those in such a sample will truly reflect those of the broader population. The challenge however, lies in identifying the target population and designing a means of sampling that population randomly to avoid bias. For target population, a typical Zogby political poll limits the scope of its survey to likely voters, rather than the entire adult population. "There's enough of a different demographic between the adult population and the likely voting population so as to be significant," Zogby explained. "[We screen out] those who have not voted and don't intend to vote in a similar election, those who aren't registered to vote, and those who refuse to give us a party identification." In the past, polling was conducted door to door. But when telephones became common in U.S. households, random telephone dialing became an effective sampling method. For his surveys, Zogby carries the selection process one step further, stratifying the samples to ensure regional, as well as demographic, diversity. "That means that when the telephone callers start calling, I have a much better chance of ensuring that my survey is both regionally and state-by-state represented," he noted. "And if I do that, I've got a very good chance of ensuring that my sample is also representative demographically." Once the survey mechanism has been designed, the prospective pollster must decide how large to make the sample. In general, increasing the sample size decreases the survey's margin of error. The margin of error is an indication of the accuracy of a poll's results. If a poll were conducted 100 times, then 95 of those polls would yield the same result, plus or minus the margin of error. A sample of 400 people, for example, might result in a margin of error of +/- 5 percent. As sample sizes increase, however, the corresponding decrease in the margin of error would become smaller. Pollsters must strike a balance between accuracy of results and time and cost of extra interviews. For his polls, Zogby typically aims for 1,000 respondents in the sample. "It's a bottom line. It's a good margin of sampling error, [plus or minus] 3.2 percent, and it's also just large enough to give me some statistically significant subgroups: approximately 500 men and women each, among 4 different age groups," he said. Once they devise a means of selecting a random sample and finalize its size, pollsters must consider what questions to ask and the proper order in which to present them. Even the most well designed selection mechanism can be undermined if the poll questions are misleading or biased. Clients work closely with the pollster through several drafts before deciding a final order. Still, Zogby says, a pollster must make great efforts to maintain a high level of objectivity in the work. "We insist on the final wording of the questions," he said. "It has a lot to do with standards. We don't feel that we're serving the client if we're loading the questions, because they're getting inaccurate information. The survey has to be designed in accordance with our ethics, with the ethics of the business, and also with the sense that the more ethical you are, the better your survey will work." The exact types of questions used vary depending on the poll, but tend to become increasingly specific over time to keep people from perceiving for whom the survey is being conducted and thereby skewing the results. A typical Zogby political poll might begin by asking a person about his initial impression of the candidates and whom he planned to vote for before moving on to specific issue-oriented questions. It might then ask for responses to open-ended questions, asking how people would react to new information about each candidate. Finally, it concludes by asking the person whom he plans to vote for, given all the new information presented in the survey. "Since the average voter out there doesn't pay attention to more than 10 to 15 minutes worth of political news in the course of a campaign, if their mind is changed in a 10 minute interval then I know what kind of image proves more effective," Zogby said. Polling results are simultaneously tabulated during the interviews using computers, and are often published within hours of completion. Public opinion polling allows candidates to better understand the electorate and determine what issues and trends will prove relevant during the course of a campaign. The media use survey results to determine public reaction to government policies and predict the outcome of elections. Zogby is quick to point out that these snapshots of public opinion do not determine the outcome of elections and should not be used to dictate policy. "Polling doesn't dictate policy, but it always has to be an element in policy. Whether it's a democracy or an autocracy, no ruler can afford to be contemptuous of public opinion. It has to be part of the mix," he said. Given its long and successful history in the United States, it is not surprising that polling is beginning to be used more frequently in other parts of the world. Zogby's company has been hired for elections in countries like Iran and Israel. Although logistics can prove challenging in regions like the Middle East, Zogby feels that given time and greater awareness of how the process works, public opinion polling can help spread democratic values throughout the world. "Polling in itself isn't just information collection, it's participatory in itself. It's vox populi [the voice of the people]. And so the more Middle Easterners become accustomed to sharing their voice in a more public and legitimate participatory setting, the greater will be both the expectations for democratization and also the skills development for democratization," he said. (The Washington File is a product of the Office of International Information Programs, U.S. Department of State. Web site: http://usinfo.state.gov)