By Richard Morin
By Richard Morin
Well, do most Americans believe that Monica and Bill did it, or not? Yes, said The Washington Post and ABC News. Our survey conducted during the weekend immediately after the scandal broke found that 57 percent of those interviewed said they were "inclined to believe" that President Clinton and former White House intern Monica Lewinsky did have an affair.
No, said CBS and The New York Times. Barely four in 10 42 percent said Clinton and Lewinsky probably had an affair, according to their poll conducted the same weekend as the Post-ABC survey.
That's wrong, countered CNN, USA Today and the Gallup Organization. Their survey, also conducted over the same weekend, found that 60 percent of those interviewed said Clinton and Lewinsky probably had sexual relations. That's virtually identical to the Post-ABC result.
No, that's wrong, reported The Wall Street Journal and NBC: Their weekend poll found that 38 percent of those interviewed said reports of the affair were probably true a result that would seem to confirm the CBS and New York Times finding.
Four polls, two answers. What exactly is going on here: How can polls be so consistent and seemingly so contradictory at the same time?
Actually, the explanation is simple, a classic example of how reputable survey organizations often make different and defensible decisions that seem to produce very different answers to the same essential question.
The Post-ABC News poll and the CNN-USA Today-Gallup percentages were the combined results of two questions: One that asked people directly whether they thought the president had an affair. The second question asked those who initially said they weren't sure which way they were leaning. CBS and The New York Times, as well as The Journal and NBC didn't "lean" respondents, and their results were based only on the first question.
My guess is that if the CBS-Times and the NBC-Journal surveys had leaned respondents, they would have produced identical results to The Post-ABC and CNN-USA Today-Gallup surveys. Actually, it's more than a guess, because I can break down the results of the Post-ABC poll and look at the results of the main question and the follow-up separately.
The main question reads, "As you may know, allegations are being reported that Clinton had an affair for a year and a half with a 21-year-old intern worker at the White House. Clinton has denied it. Are you inclined to think Clinton did or did not have an affair with an intern at the White House, or don't you know enough about it to say?"
The result: 44 percent said he did virtually identical to the 42 percent reported by CBS and The Times, and just slightly higher than the 38 percent in the NBC and Wall Street Journal poll.
That number rises to 57 percent in the Post-ABC News poll when this question was asked of those who said they didn't know enough to say: "Well if you had to guess, are you inclined to think Clinton did or did not have this affair?" Another 13 percent suspected that he had.
Which is the right answer? There obviously are two schools of thought among reputable polling organizations.
Some pollsters argue that leaning or "pushing" those who initially express no opinion produces a more accurate result. The theory is that people may be reluctant to express a view until they're invited to a second time or gently prodded, particularly on sensitive questions. Besides, these follow-up questions allow researchers to measure people's suspicions or inclinations. That's useful in assessing the current state of public opinion and where it may be heading as measurements of firmly held beliefs.
That's why everybody, including CBS, NBC, The New York Times and The Wall Street Journal, ask a follow-up "leaner" question when measuring voter preference in elections. (The drill is much the same: A pollster first asks which candidate the respondent would support if the election were held today. Those who don't express a preference between any of the candidates are asked a follow-up question like, "Well, which one are you leaning toward?" The results of the candidate preference question and the "leaner" question are then netted and reported as each candidate's share of the hypothetical vote.)
But there are pollsters who argue that leaning respondents on some kinds of questions may force some respondents to express views they donŐt really hold. The follow-up question may be perceived as a demand for a response, not merely an invitation. Thus the netted results are a curious stew of real opinions and non-opinions.
My personal view is that both sides are right. The critical issue is timing. Surveys done immediately after an event such as those done the night of Jan. 21, the day The Washington Post first reported details of the alleged affair risk capturing large numbers of true non-opinions if they push those who initially say they don't yet know enough to offer a view.
That's, in fact, exactly what our polling partners at ABC did. The night the scandal broke, ABC asked people whether they believed the allegations that Clinton had had an affair with Lewinsky, offered the "don't know enough to say" option, and did not ask the follow-up. Surveyors found that nearly half 46 percent had not yet enough information to say.
But as information builds, the risks of forcing non-opinions diminish.
In the case of the current Clinton scandal, information about essential details of the allegations spread quickly certainly quickly enough, in my view, to justify leaning respondents.
Richard Morin is director of polling for The Washington Post. "What Americans Think" appears Mondays in The Washington Post National Weekly Edition. Morin can be reached at email@example.com.
© Copyright 1998 The Washington Post Company