Polls measure anything, everything
Ann McFeatters
WASHINGTON - America is in danger of being polled to pointlessness. From political polls to market research to attitudes on alien life forms, there is a poll to ``prove'' every opinion, whether it's vital, vapid or vindictive. Everybody wants to poll, whether or not they know ``the science'' of polling. The future holds more, not fewer, polls. But the bottom line is that polls can be used to ``prove'' anything. There are polls that show women no longer want to get married and polls that show they do. There are polls that show parents would not have their children if they had it to do all over again. There are polls that Americans would do anything for their children. There are polls that more Americans believe in conspiracies. There are polls that show Americans are more skeptical. There are polls that more Americans are flossing and other polls that we're eating more sugar and not brushing. There are polls that Americans wanted to go to war in the Balkans and polls that they did not. Whether the issue is gun control or tobacco taxes or censorship of violent movies or tax cuts or saving Social Security, polls support both sides of the debate. For many people, polling touched raw nerves this past year as President Clinton's job approval rating continued to hold high and steady despite impeachment. Many people offended by Clinton's behavior were convinced the polls were fixed. But the experts who have studied this phenomenon said the polling was correct because the public was persistent and steadfast, across demographic lines. Even though Clinton's personal negatives with the public soared, the majority of Americans did not waver: They wanted him to stay in office and said he was doing a relatively good job. A new study of a new poll (yes, really) by professors at the University of Michigan, the University of Pennsylvania and Ohio State University argues that Americans like polls because they like to know what their fellow citizens believe about the issues and how their own opinions fit in. On the other hand, while everyone pays lip service to making public opinion count in a democracy, the study says that the more educated and politically knowledgeable those surveyed are, the less they want government to act based on what polls show. And it seems to be almost universal that if a poll agrees with what you believe, you are more likely to think its findings are accurate. If you are asked if you are influenced by polls, you are almost certain to say that you are not but that you are worried that your fellow citizens are far too influenced by polls. However, the study says, those who are influenced by political polls tend to have lower levels of knowledge about the issues or the candidates. The real issue in any poll is how a question is asked. For example, polls show that Americans are not clamoring for tax cuts if they are asked if they'd prefer tax cuts or prescription drug benefits through Medicare and reassurance that Social Security benefits will be there in 30 years. But if they are asked if they would rather see more government spending or tax cuts, they say they would prefer tax cuts. When asked if they would prefer better health care, a cleaner environment, paying down the national debt and a strong defense or a tax cut, they do not choose a tax cut. When asked if they think the projected federal surplus (if any) is their money and if they want it back, they will answer that it is and that they do. Another crucial point with polls is timing. There are almost daily polls on whether likely voters prefer Texas Gov. George W. Bush or Vice President Al Gore as their next president. Invariably, Bush would win if the election were tomorrow. But the election is not for 15 months. Anything can happen. Voters won't start focusing on all the candidates for weeks. At this point, they know little about Bush and haven't paid much attention to Gore apart from Clinton. Another issue is analysis. If incumbent A gets a 45 percent favorability rating and his or her opponent gets 20 percent, this is not, as it might seem, good news for candidate A. It means more than half of those surveyed aren't pleased with their office-holder. In preparation for the upcoming season of polling frenzy, the experts have some advice when evaluating a poll: A national sample should be at least 1,000 people randomly selected. Who is paying for the poll? What's the purpose? Are the results exactly what you would expect the sponsor would want? Who conducted the poll? Is the polling firm reputable? A poll, only a snapshot of the situation at that moment, must be timely. The methods, response rate and margin of error should be fully disclosed. If there are previous polls, the new results should be put in context. Exact questions must be disclosed and be clear and unbiased. Are the polls interpreted correctly? Are apples compared with apples? If the results defy common sense, maybe the polling methods were goofy. Or maybe those surveyed just didn't feel like being polled that day.
Ann McFeatters covers the White House for Block News Alliance.