A word on the limitations of polling

May 20, 1988

THE explosive growth of polling in and on American politics - so evident this election year - is yielding a wealth of information on the public's views and values. Unfortunately, though, it is also producing an abundance of misuse and confusion. Some of the problems involve gross misinterpretation of perfectly good data. For example, a recent Associated Press story told readers that ``Americans view Soviet leader Mikhail S. Gorbachev nearly as favorably as they do President Reagan, and Gorbachev outpolls Reagan among college-educated Americans....'' The survey, by the Associated Press and Media General, found 59 percent of all those polled and 62 percent of the college educated expressing favorable opinions of the President, 54 percent and 71 percent, respectively, of the general secretary. But, of course, Mr. Reagan was being assessed as an American president, and Gorbachev as a Russian leader. People were saying that they think quite highly of Mr. Gorbachev compared with his predecessors, not compared with Reagan.

Another common set of problems arises from trying to measure opinion that has not formed and then reporting the results as though they represent something substantial. Early presidential trial heats - at present, asking people whether they favor George Bush or Michael Dukakis - are an instance of this. It's not just that people may change their minds about how they will vote between now and the November balloting; it's that many of them haven't made up their minds at all. When asked whom they prefer, they will usually give an answer - but often the judgment is highly insubstantial. It really hasn't been made.

Presidential trial-heat findings six to 10 months before an election are confusing only to the extent that they are taken seriously as predictors of the November vote. In January 1980, the Gallup poll showed Ronald Reagan the weakest of the three Republicans being prominently discussed as presidential nominees (the others being Mr. Bush and Gerald Ford). Jimmy Carter ``led'' Reagan, 62 percent to 33 percent. In February the poll had Mr. Carter ahead of Reagan 60 percent to 31 percent; at the beginning of March the margin was 58 percent to 33 percent; in mid-May Carter ``won'' 49 percent to 39 percent.

Not surprisingly, when an opinion really hasn't formed, various poll attempts to measure it are likely to yield highly disparate results. Thus CBS News/New York Times surveys in April and June found Reagan ahead, while Gallup surveys of the same date had Carter in front. This spring, Bush-Dukakis trial heats have frequently yielded dissimilar results.

The reach of polls has exceeded their grasp even more notably when they have tried this spring to determine whether the addition to the ticket of a particular vice-presidential candidate helps or hurts electorally. There is nothing wrong, of course, in wanting to know whether having the Rev. Jesse Jackson as his running mate would strengthen Michael Dukakis's bid or weaken it. But questions posing this issue directly just don't work, because a very large proportion of the public has not thought the matter through at all.

I am often asked about how reports of campaign polls influence voters' decisions. For example, are voters led to the candidate who is shown to be way ahead in a ``bandwagon'' effect, or are they perhaps turned the other way, in a ``sympathy-for-the-underdog'' reaction? I reply that the public at large doesn't seem to be much affected - in large part because it doesn't pay much attention to polls. It is the political community that we need to worry about. Many in this community of officeholders, campaign technicians, financial contributors, commentators, and activists seem almost obsessed with polls and are sent careening this way or that by even the most insubstantial poll findings.

The political community's polling fascination has given polls a dominant role in how elections are thought about and covered. This means that judgments about candidates are increasingly poll-driven. Worse of all, perhaps, our compulsive use of polls has contributed to an inclination to portray elections - including those for the nation's highest office - as mere exercises in political manipulation.

The proper response to these problems is not, I think, to be found in government regulation of polling. Rather, those who consume poll information need to cultivate a more vivid sense of its limits: the limits set by the nature of public opinion on what can be reliably measured; and the limits on the part a polling picture of elections should play in these great celebrations of democracy.

Everett Carll Ladd is executive director of the Roper Center for Public Opinion Research at the University of Connecticut.