The Superpollsters: How they Measure and Manipulate Public Opinion in America. - book review

Public Interest, Wntr, 1993 by Joshua Abramowitz

DESPITE then-Governor Clinton's solid lead in the weeks leading up to the 1992 election, it seemed unwise to have too much confidence in the polls. There was the matter of H. Ross Perot's rising, falling, and again-rising fortunes: One never knew from which of his adversaries he would draw more support. Meanwhile, Clinton's lead, though substantial, was shaky: Voters were often more anxious to cast an anti-Bush vote than to endorse the Democratic challenger. Pundits' predictions spread out across the spectrum. Some siad that the polls were underestimating a Clinton landslide. Still others awaited a "Major miracle," whereby the polls underestimate the strength of a conservative candidate, a la Britain's election of Prime Minister John Major last year.

There would have been something pleasant about a surprise result, be it in one direction or another. An alliance of high-tech computers, telephone banks, and Ph.D.s poring over the 200 polls taken between the conventions and the election would have failed, and we would have all learned how the sentiments of a vast nation can be neither pinned down nor reduced to a set of facts and figures. Alas, it did not happen. The polls wee right. Public opinion was both read and interpreted with a high degree of precision.

The significance of polls in the 1992 election, however, is not just that they successfully predicted the outcome. It is also that the polls themselves were the focus of so much of the nation's attention. From anchormen to armchair pundits, discussion of the race often focused on, well, the race. News stories frequently relied on, or were at least put in the context of, "the latest polls and what they mean." And individual voters had to determine if they would vote for their true favorite or, after reviewing the latest poll numbers, cast a strategic bid for their second choice. This is not to say that the election was so entirely poll-driven as to be without substance. It does mean that in some crucial way following presidential politics resembled following horses at the track.

In Thomas E. Mann and Gary R. Orren's Media Polls in American Politics, a number of prominent pollsters and academics have written essays on some of the issues surrounding the polls conducted by major media organizations. David W. Moore's The Superpollsters: How They Measure and Manipulate Public Opinion in America takes a broader, more historical, and more anecdotal look not just at media polls, but also at those conducted by campaigns and major polling organizations. They are uncontroversial books--explaining, rather than arguing--but each offers valuable information on how and under what circumstances polls do and do not work, as well as a more limited look at their effect on the American social and political scene.

WHEN DOUGLAS WILDER ran against Marshall Coleman in the 1989 Virginia gubernatorial race, polls put Wilder ahead, on the average, by 7 points. Yet he won by less than a point and a half. For David Dinkins in the 1989 New York mayoral election, a 14-point lead in the polls gave way to an ultimate victory of just 2 points. And although George Deukmejian won the 1982 California gubernatorial contest, each and every one of the nineteen polls that were conducted put Tom Bradley ahead. In all the above elections, the polls misgauged support for a black candidate, predicting that he would attract a substantially greater number of votes than turned out to be the case. Now, as the essay by Henry E. Brady and Gary R. Orren in Media Polls in American Politics points out, the evidence here does not make for conclusive proof of racism. Normal campaign instability, unusual turnout patterns, and undecided voters making last-minute shifts all might sway an election. And the increased media focus on a mixed-race contest could also affect voters. But it could also be true that voters are lying to pollsters (or hiding out in the "undecided" column) to hide their support for the white candidate. There are two possible motivations for this: Voters might be nonracist but concerned about being perceived otherwise, or they might want to hide genuine racism.

As a social phenomenon, this latter possibility is highly disturbing if true. But the specific technical difficulty here--determining if people are lying--is far from the only source of polling inaccuracy. Brady and Orren discuss in some detail the various "sampling" errors, which result from the failure to reach representative members of the group being polled. In attempting to steer around such problems, the pollster must account, for example, for the 5 percent of people who do not have telephones, for the fact that certain types of people are more likely to answer the phone in a household, and for the often substantial percentage that refuse to participate (although pollsters employ "refusal converters" to attempt to sway those reluctant to take part).

Meanwhile there are a variety of problems with the questions themselves: question wording (during the Cold War, for example, polls showed that a majority of Americans simultaneously favored "detente" and "getting tough" with the Soviet Union); question placement (support for a president can be lowered if an approval-rating question is preceded by questions that put the president in a negative light); and familiarity with the subject (Michael Dukakis' favorable poll ratings early in the 1988 campaign were based on a vague, and hence easily changed, impression of the candidate).