It would probably be good for American democracy if we could learn to pay a little less attention to pre-election polling. After all, we are going to find out who wins on election day anyway. Until then, the really compelling question for interested citizens isn't "Who will win?" but "Who should win?"
It's evident, though, that the political community's and journalists' appetites for election polls are growing, and elections are increasingly framed in terms of trial-heat results.
Pre-election polling was begun non-systematically by Literary Digest magazine in the late 19th century - and then systematically by Gallup and Roper in the 1930s. Still, polling was but a tiny part of American electioneering through the 1960s. In 1968, the Roper Center's exhaustive catalog of presidential trial heats (done nationally) contained only 10 askings of "How would you vote if the election were being held today?" from Sept. 1 through election day. This year, in contrast, we have located more than 50 for the month of September alone and project more than 110 from Sept. 1 through Nov. 5.
As the number of polls climbs, many leading research organizations labor to improve their methods. They try, for one thing, to get better estimates of "likely voters," meaning those in the entire eligible population who are likely to cast ballots. For all these efforts, the polling environment is more difficult now than ever before. One big reason why it's harder now to get reliable pre-election vote estimates is because a growing segment of the electorate is no longer anchored by party loyalties. The proportion of Americans who think of themselves as independents rather than Democrats or Republicans and who, even more important, vote the "best person" rather than the party, is the largest ever. Because today's electorate is less anchored, it typically takes less to move it.
Challenge for pollsters
This weakening of party ties may not pose problems for the democratic process - but from a polling standpoint, the unanchored electorate is a harder read. It has the potential of moving from one presidential candidate to another on relatively little "provocation." Especially for that large segment of Americans that doesn't think much about its choice of candidates until the approach of election day makes it necessary to do so, the absence of firm party loyalties creates a situation in which answers to early trial heats often have very little substance.
The potential for electoral volatility in today's unanchored electorate is heightened in an election like this year's, where many voters find themselves tugged in opposite directions. Both Bill Clinton and Bob Dole get high marks from voters on some leadership dimensions but low marks from many of the same voters on others. Many say that they consider Dole of high integrity and properly credentialed for the presidency, but perhaps too old and too cold. In turn, many voters like Clinton and credit him with being an able and hard-working politician, but also say they don't trust him. In early September, when the president enjoyed a lead of 20 points or so in the trial heats, an ABC News/Washington Post survey asked whether "he has high personal moral and ethical standards." Just 39 percent said yes; 56 percent that he did not.(In contrast, 70 percent credited Dole with high moral standards.) An incumbent president may win reelection despite such judgments, because people make their final decision on a variety of considerations. But now, as four years ago, Mr. Clinton gets decidedly mixed grades on key elements of personal leadership.
The parties get mixed verdicts, too. Many people at once give the GOP the edge on such important matters as controlling the growth of government spending and managing foreign policy, while crediting the Democrats with "caring more about needs and problems of women," and as better on environmental issues. Overall, the two parties stand in essential parity in public ratings.
Clinton is running for reelection at a time when the economy is strong - and more important, when its strength is widely recognized. At the same time, substantial pluralities say the country's most serious problems are moral, not economic. A large body of polling emphatically rejects the suggestion that the public is "feeling good" about the state of the nation as it nears the November vote.
What's more, while much of the public is satisfied with the past four years' economic developments, which favors the incumbent, it remains broadly conservative on a wide range of social issues and role-of-government questions. A recent Roper Center review of the data shows Americans as disinclined now to endorse more government for national problem-solving as they were in the fall of 1994 and early 1995.
Many Americans' assessments of the candidates, parties, issues, and state of the nation thus leave them of mixed minds on how to vote. Such an electorate isn't likely to be "locked up" for any candidate five weeks before the balloting. It's surprising, then, that so many observers have given so much weight to the trial heats to date - and seem so ready on that basis to award Clinton the election.
Poll soundings have bounced around a good bit. The daily tracking polls clearly reflect not only actual changes in sentiment, but who happens to be reachable by phone on particular days. Republicans generally seem to be harder to reach than Democrats on weekends. Still, Clinton is certainly ahead as the campaign enters its home stretch.The critical question is, "How firm are the numbers?" Is Clinton's lead so well grounded that it's unlikely to change, or is it wide but shallow? The underlying structure of this election points to a fairly close final outcome, not a landslide, and suggests that the preferences trial heats have recorded are weakly held.
In 1948, Gallup showed Harry Truman trailing in late September, but he won in the November balloting. Nonetheless, in the 60-year span for which we have survey data, the candidate leading in September has usually triumphed. But there's often been a lot of movement from poll findings a month or so before election day to the actual results. For example, a Gallup tracking poll of Sept. 28-30, 1992, found Clinton leading Bush by 16 points. The Democrat won, of course, but by a 5.6 percent popular-vote margin. In 1968, a Gallup survey of Sept. 26 through Oct. 1 had Richard Nixon ahead of Hubert Humphrey by 15 points, with George Wallace a strong third. Humphrey came on strong, though, and Nixon won by just 7/10ths of a percentage point in the popular vote.
There was a good basis for predicting the race would tighten in years like 1968 and 1992. In the Nixon-Humphrey contest, the Democratic Party found itself bitterly divided over Vietnam, but at the same time it enjoyed great advantages - including a huge margin over the Republicans in party identification. It was unlikely that the overall mix of factors - some favoring the Democrats, others the GOP - would in the end produce a lopsided result. In contrast, the 1984 race between Ronald Reagan and Walter Mondale looked like a landslide from the beginning. Not only did the polls point to it: All of the underlying elements, from Reagan's personal standing to a resurgent national mood and a strong economy, were going the Republicans' way.
Five weeks from presidential balloting, analysts should focus on the strength of the attachments the trial heats capture. This year's contest has a great potential for late change. The electorate is remarkably unanchored, and brings sharply contradictory assessments to its presidential choice.
*Everett Carll Ladd is professor of political science at the University of Connecticut and president of the Roper Center for Public Opinion Research.