It ain't over until the voters sing


October 30, 1992|By David W. Moore

ONCE again, pollsters and pundits, who should know better, have been caught short in predicting an election outcome.

For weeks, a Bill Clinton victory has been treated pretty much as a foregone conclusion.

Newsweek's Oct. 26 issue typifies the mind-set with its cover photo of the governor and the caption "President Clinton? How he would govern."

Frank Newport, the Gallup organization's president, said on TV last week that a comeback by President Bush would be unprecedented.

The ABC 50-state poll broadcast Friday showed such an overwhelming Clinton Electoral College lead that his victory was virtually assured.

Now the polls tell us the race has tightened significantly, and the public must again wonder about the validity of polls. This is not the first time the polls have produced an overwhelming consensus about the likely outcome, only to be proved wrong.

In 1948 every major commentator and newspaper accepted the pollsters' prediction that Thomas E. Dewey would defeat Harry S Truman.

The polls were not wrong; they correctly reflected Dewey's early lead. Rather, the pollsters were wrong in predicting that opinion could not change sufficiently to allow a Truman comeback.

Today, the situation is somewhat similar. Pundits used poll results showing Mr. Clinton with a double-digit lead to predict an election that was weeks away, expecting that too little time remained for Mr. Bush to make a comeback. Now those ` expectations have been upset.

The fault is largely with the way we pollsters ask the presidential-preference question. Our method inevitably suggests a more decisive electorate than is the case, thus masking the potential for significant shifts in opinion.

The standard question asked is: "If the election were held today, whom would you choose -- Ross Perot, the independent; Bill Clinton, the Democrat, or George Bush, the Republican?"

The order of the names may differ and many polls include the vice presidential candidates, but what we pollsters call a forced-choice question is standard.

jTC We do not offer the option of "unsure," and if voters indicate indecision, we press them for the candidate they lean to. The result: In recent weeks only about 10 percent of the voters have been classified undecided.

That these figures could not possibly have been true is illustrated by the volatility of the polls in August.

Early in the month, only 10 percent of the voters were classified as undecided, which suggests that 90 percent must have been decided.

Nobody believed that, and a couple of weeks later Ross Perot moved from first to last, Mr. Clinton took a major lead for the first time, Mr. Bush's support dropped and again the number of undecideds was reported at about 10 percent.

Clearly, something is amiss with our way of measuring the firmness of voters' intent. The small number of voters whom pollsters cite as undecided are determinedly so.

They defy our response categories, volunteering that they have not made up their minds and resisting our pressure to say they lean toward a candidate.

But there are many undecided voters who play our polling game -- who specify a candidate they might choose "if the election were held today" but who have no commitment to that candidate beyond the time of the interview.

If our question was framed to acknowledge that the election is in November and it included the option of "unsure," and if we excluded the so-called leaners from a candidate's vote total, the number of voters classified as undecided could be 20 to 25 points higher than it is with our forced-choice format.

Instead of seeing what appeared to be a lock on the election, the public would have known that many voters were mulling over their decision and were still subject to influence by ads and campaign appearances.

In 1948 the pollsters were humiliated by their prediction of a Dewey victory. They were embarrassed again just three elections ago.

In 1980 all the major media polls reported that with fewer than five days to go before the election, the Carter-Reagan race was a dead heat.

They all said the election was too close to call.

Jimmy Carter's pollster, Pat Caddell, showed the same results. But unlike the media pollsters, he tracked the vote over the weekend and saw the dead heat change into a 10-point victory margin for Ronald Reagan.

Mr. Reagan's pollster, Richard Wirthlin, also conducted daily tracking polls that recorded huge swings in the last several days, from a small Carter lead to a landslide Reagan victory.

These swings and the election results shocked the polling industry, which still couldn't believe that public opinion could change so dramatically in just a few days. Now once again major fluctuations in public opinion have surprised the country.

We pollsters must reconsider the validity of the forced-choice question.

We have clearly underestimated the number of undecideds and lulled everyone into forgetting that a sizable number of voters remain on the fence. Nobody knows where they'll jump.

The common assumption has been that they would vote the same way as those who had already decided.

But, as the latest polls show, that assumption does not always hold.

Every few elections the voters remind us that it is they, not we, who have the last say.

David W. Moore, author of "The Superpollsters: How They Measure and Manipulate Public Opinion in America," is director of the Survey Center at the University of New Hampshire.

Baltimore Sun Articles
Please note the green-lined linked article text has been applied commercially without any involvement from our newsroom editors, reporters or any other editorial staff.