In a new book describing changing race relations in the United States, public opinion analyst Howard Schuman and two of his academic colleagues home in on what they call "a fundamental transformation of social norms" over the past 40 or 50 years. Some of what they say is directly related to the polls The Washington Post conducted prior to the Virginia elections.

Some white people who harbor racist feelings today, the authors write, may succumb to social pressure and keep those feelings to themselves, even to the extent of giving racially liberal answers in the semiprivate setting of a public opinion poll interview.

"Antiblack speech and action that were once acceptable and common are now almost completely taboo," note Schuman and his associates, Charlotte Steeh and Lawrence Bobo, in "Racial Attitude in America, Trends and Interpretation."

That is the reverse of the situation 50 years ago, when, according to the authors, a survey now regarded as a classic "showed verbal responses to a questionnaire to be discriminatory, and actual behavior to be equalitarian. That our concern today is with exactly the opposite problem indicates how great has been the change from the 1930s to the 1980s."

My concern here, however, is exclusively with this question of social pressures on whites to give racially liberal responses to pollsters, and on the related problem of blacks not being exactly candid, either. Two polls I did for The Washington Post substantially overstated the lead of black candidate L. Douglas Wilder in his race for lieutenant governor and those of the other two candidates on the Democratic ticket. Wilder was victorious in his bid to become the first black in the South to hold high statewide elective office since Reconstruction.

If social pressure pushes some whites to give racially liberal answers, it also appears to push some blacks to say they have voted in the past and are certain to vote now, when, in fact, voting may be the last thing on their minds. A "likely voter" screen in both polls indicated that blacks would make up 20 percent of the electorate. My estimates on the outcome of the three elections were based on such a turnout.

Blacks are thought to have made up about 9 percent of the electorate in the 1979 Virginia elections, and about 14 percent in 1981. This time, with a black on a ticket, it seemed likely that the strong upward trend would continue and that my screen was correct.

But no such black vote materialized on Nov. 5, and the surveys were considerably off. They predicted the right winners, but by margins that were much too wide. It is now estimated that blacks made up about 12 percent of all voters. When the polls are adjusted to take that into account, the one I did the week before the election comes within 2 percentage points in the governor's race, 3 points for the attorney general's contest.

But in the Wilder race, even such an adjustment leaves the poll off the mark. He won by 52 percent to 48 percent; the final poll, weighted for the low black turnout, would have put the outcome at 59-41, an overstatement of 7 percentage points.

I think it is safe to attribute all or almost all that distortion to the kind of social pressure that Schuman and his colleagues write about.

The evidence exists in two forms. First, just over 1 percent of the people polled were blacks who said they had not made up their minds, and 13 percent were whites who claimed to be uncertain.

Some analysts, including those in Wilder's camp, were working with similar numbers and made the assumption that all the "uncertain" blacks would vote for Wilder and all the "uncertain" whites for his opponent. Mean-spirited as it sounds, they may well have been correct: Had I apportioned those votes that way, the Post's result would have been on the money, 52-48.

The other evidence is at least as strong. The survey included questions aimed at eliciting signs of racial prejudice. The most direct was to ask people whether they were more inclined or less inclined to vote for Wilder because of his race. Six percent of whites said "less inclined." Other questions asked whether the respondent would vote for a black for president (16 percent of the whites said "no"), and whether they felt racial integration was going too fast or too slow or about right (12 percent said "too fast").

Based on these responses and those to two other questions, I created a "bias scale" after the election. It showed that 76 percent of the whites interviewed exhibited no antiblack feeling in any of the questions, that 6 percent showed a possible bias and 18 percent perhaps a substantial bias.

People may argue whether those figures are too high or too low for a state such as Virginia; I won't maintain that this rough measure is a perfect gauge of bigotry. What is pertinent, however, is that among that 18 percent, fewer than half said they were voting for Wilder's opponent. Most of the rest said they were for Wilder; some said they were undecided.

Had I discounted these people's votes -- that is, moved them out of the Wilder column to that of his opponent -- that, too, would have resulted in a direct hit, a 52-48 score.

As more blacks seek high political office, more opinion analysts will have to learn this "fundamental transformation of social norms" or suffer Election Day embarrassment.