It is one of the great political conundrums of our times.
For two weeks, President Clinton has been at the center of lurid reports about marital infidelity and lying under oath - and his popularity rating has never been higher.
But don't ask the pollsters to explain, with any kind of clarity, why this is so. In some ways, the rapid-fire polltaking of the past two weeks - in which people are asked about their opinions but don't explain their answers - is shedding more heat than light on the latest Clinton flap.
Therein lies a central paradox of modern politics: In an era of instant polling and instant communication, surveys are being used to shape policy and perceptions more than ever, but, in many ways, paint an incomplete picture of what the public really thinks. "I'm beginning to wonder what it is that we're measuring when we poll so much," says Karlyn Bowman, a nonpartisan analyst of polls at the American Enterprise Institute here.
From CNN and CBS to The New York Times and the Corpus Christi Caller-Times, polling data on Mr. Clinton and the White House intern he allegedly had an affair with, Monica Lewinsky, are filling airwaves and newsprint. We are becoming a nation of mirror-gazers and arm-chair pundits.
"We've seen it several times in the last few years, and it's in part because of the proliferation of pollsters - they're all covering the same ground over and over again," says Ms. Bowman. "I don't know whether we're regurgitating what we hear in the polls. That sort of worries me."
One thing is certain: That the import of these polls is more than mere idle curiosity. Bill Clinton's presidency lives and dies by public opinion. If a new bombshell development were to come out today, his high job-approval rating - in some polls above 70 percent, unprecedented for him - could plummet. This would embolden his Republican opponents and perhaps revive talk of impeachment proceedings, which for now is dormant.
Former President Bush certainly will never forget how his approval ratings spiked above 90 percent during the Gulf War early in 1991 - only to lose the presidency to Clinton the following year.
For now, though, absent any new revelations, Clinton can take some comfort that his high approval numbers have improved the overall climate of thought about his presidency, which in turn could cause the press to back off.
"Public opinion does affect the media, but not necessarily in a conscious way," says Douglas Hodgkin, an expert on polling at Bates College in Lewiston, Maine. "The media are interested in attracting viewers and readers, and if the public is saying, 'enough already,' that may have an impact."
As for independent counsel Ken Starr, whose own "favorable" ratings in polls (at least among those who have any opinion at all) are dwarfed by Clinton's, Mr. Hodgkin doesn't think he'll be as influenced by the poll numbers as elected officials are.
In defense of polling
Understandably, the people who do the surveys defend their craft. Richard Morin, director of polling for The Washington Post, says it's been important to gauge public opinion from the beginning of the Clinton-Lewinsky story, to see how views have evolved.
"We have a presidency that may or may not be in peril, and ... we're dedicating reporting resources and polling resources to tell that story," says Mr. Morin. He says that "the numbers alone" tell only part of the story, which is why the paper supplements stories about polls with "voices" - interviews with people being surveyed - to flesh out the picture on public opinion.
Fact and speculation
But polls alone tell less than we may think. After Clinton delivered his State of the Union address Jan. 27, his approval ratings shot up - leading analysts to conclude the public responded positively to his speech. But there are no hard data to support that view.
"Did pollsters ask people if they watched the speech, and if so, if that explains their approval of the president's job performance?" asks Michael Traugott, a political scientist at the University of Michigan at Ann Arbor. "Have pollsters correlated people's approval of Clinton with how they voted in the last election? Or with their party affiliation?"
The bottom line, Mr. Traugott says, is that the data are incomplete. But this isn't surprising. Because many polls are commissioned by the news media - which, at heart, are focused mainly on one question: Is the president going to make it or not? - then the questions reflect their interests.
The techniques of pollsters also come under scrutiny at a time like this, when public opinion represents a key element in a news story. Most polls are pretty straightforward, says polling expert Karlyn Bowman, though there can be subtle differences, such as whether Clinton is referred to as "the president" or "President Clinton" or "Bill Clinton."
The positioning of questions can also make a difference. If the job-approval question is asked at the beginning of an interview (as most polls do), the president will get a higher rating than if the question comes at the end, after a series of scandal-related questions.
How the sample is selected can also affect the numbers. If a pollster arrives at a scientifically selected sample, and works hard to poll those particular individuals, he may end up with a different result than one who doesn't try hard to get to the people who are less interested in being surveyed.