Consulting the Oracle
Everyone loves polls. But can you trust them?
When the Republicans were deciding what to put in the budget that they now must negotiate with President Clinton, they consulted the polls, which told them that promising to "put the government on a diet" is popular. When the president was trying to decide how hard to fight the Republicans' budget plans during the recent government shutdown, he consulted the polls, which said that accusing the Republicans of trying to cut Medicare scores points. And when the media wanted to know what to make of it all, they consulted the polls, too. Nearly every day of the budget stalemate, at least one major news organization released a new poll showing whom the public "blamed."
Polls have become for modern politicians and pundits what the oracle at Delphi was to the ancient Greeks and Merlin was to King Arthur: a mysterious and almost divine source of wisdom. There are polls to help politicians decide which euphemisms to use for orphanages, if they're in favor of them, and what epithets to use for government employees, if they're against them (call them "foster homes" and "bureaucrats"). There are polls to tell them what kind of shirt to wear (the checkered, lumberjack variety). There are polls to tell them everything except how to behave with integrity, judgment and honesty.
Pollmania. The sheer number of polls being conducted is staggering. President Kennedy commissioned 16 polls in the three years he was in the White House. Richard Nixon did 233 during his six years in office. During President Clinton's first 2 1/2 years in office, the Democratic National Committee paid pollster Stanley Greenberg more than $4.5 million for White House polling, which is enough to buy about 100 to 150 polls.
Politicians have now become so conditioned to polling first and deciding later, writes Brian Tringali of the polling firm the Tarrance Group, that some are virtually incapable of "making a decision without it." Pollsters now can--and regularly do--"go in the field at a moment's notice and have results to the candidate often within a few hours," Tringali notes.
That polls have replaced judgment in the conduct and coverage of American politics is scary enough. Scarier is what a shoddy substitute they are. On a narrow range of clear, well-defined issues that people have thought about carefully and hold strong views on (such as how they will vote in tomorrow's presidential election), a well-designed poll can accurately reflect the views of the electorate. (Even here, however, spectacular bloopers have occurred, such as the polls' failure to predict John Major's victory in the 1992 British election or the Republicans' sweep last fall.) But on nearly everything else, polls are inaccurate, speculative, volatile and often utterly devoid of the meaning that politicians, pollsters and the press all eagerly impute to them.
1. The "margin of error" is the least of it. A survey of 1,000 Americans will typically be reported to have a sampling error of plus or minus 3 percent. But anyone who takes that figure as a literal statement of a poll's accuracy is making a big mistake.