BOOK REVIEW / It's the real swing: The Superpollsters - David W Moore: Four Walls, Eight Windows, pounds 13.95

ABOUT the only thing American pollsters agree on these days is that most Americans, given the opportunity, prefer voting for politicians they have never heard of rather than ones they have.

The problem with American presidential elections, however, is that they last so long that no politician can remain unknown long enough to remain popular. Witness Ross Perot, the talk-show phenomenon already chased back into the primordial media-muck from which he crawled. Since what Americans know of their politicians is pretty awful, and since they cannot vote for politicians who they don't know because, of course, they don't know who they are, nearly half of them do not vote at all. And those who do, try not to think too hard or expect too much. They vote for the green balloons or the blue ones.

It's called the 'Mr Smith' effect, and was first identified by Pat Caddell, an influential Democratic pollster in the Seventies and Eighties. Caddell helped to engineer George McGovern's surprising dark-horse victory in the Democratic presidential nomination in 1972; he went on to mastermind Jimmy Carter's victorious campaign in 1976, and Gary Hart's failed one in 1984. Caddell did not want to identify public opinion; he wanted to shape it. Eventually, when he recognised his methods were becoming 'indefensible', he quit what he called 'the heroin of politics' in 1988. 'This country isn't apathetic,' he recently declared, 'it's angry as hell.'

Polling began as a subscription gimmick for early 20th-century magazines and developed into a substitute both for politics and political journalism. George H Gallup pioneered professional polling by predicting Roosevelt's startling 1936 landslide victory over Alf Landon; he was a populist who believed that by making the public's voice heard, he might teach politicians to represent their constituencies better. By 1946 the newspapers treated polling results so seriously they made fools of themselves; it was the year many front- page headlines prematurely announced Dewey's victory over Truman - which, of course, was totally back-asswards. Journalists and politicians had begun listening to polls rather than people. In America (and in Britain, too, as the last election showed) the problem has since grown worse.

Polls have become an intractable part of the American news media. By reporting on the statistical 'mood' of the nation, journalists can avoid the onerous work of educating that nation, or providing people with the crucial, hard-won information they need. As a result, public opinion has become a volatile phenomenon, obliged to base its decisions on the insubstantial issues debated in bad newspapers and television programmes - issues focusing on the 'character' or 'integrity' of political candidates rather than on the hard economic or political programmes they endorse.

In other words, people are smart enough to know when there is a lot they still do not know. As a result, support for presidential candidates can shift as much as 20 or 30 percentage points in a matter of weeks.

The way public opinion is reported and manipulated by media is a fascinating subject, and it is a shame David W Moore has not done it justice. The Superpollsters, I'm afraid, is dead, dry, interminable and vapid. Like the polls themselves, it gives us all the superficial facts and figures and none of the ontological stuff about how public opinion is shaped, guided or misrepresented by polling. It is a book filled with angular, unengaging and mechanical sentences such as the following: 'Probably no single event in the primary campaign is more important than the acceptance speech of the nominee, where he will address the nation and establish his vision of the country. It is a unique opportunity for the candidate to define his campaign on its own terms, and to persuade the American people to give him their support.' Might as well say: Good presidents do good things, bad presidents do bad ones. Nobody likes the bad ones, because they don't do a good job.

Random sampling indicates that books about polling are not necessarily dull, but this one is.