David W. Moore argues that polls are being misused. He argues that questions are formulated in such a way that the results make for interesting news stories while bearing very little resemblance to what the public actually believes. He argues that even good polls are presented by news organizations in dangerously misleading ways. He argues that polls, as they are practiced today, both misinform the public and elected officials, and actually can create public opinion and misguided policy decisions. Having been a senior editor at Gallup Polls for thirteen years, he has an insider's perspective on how they're run, and where they can go wrong. And by the end of the first chapter, I was sold.
By the end of the second chapter, I wanted him to stop trying to convince me of what he'd already established, and get on with developing his argument further. By the end of the third chapter, I was bored, impatient, and starting to suspect that he really didn't have anything new to add beyond what he'd said at the outset.
I turned out to be right. The book is filled with example after example after example, interspersed with reiterations of the same argument over and over and over again, scrupulously documented with endnotes. Almost any given chapter would make for a good essay; I'd pitch it to Harper's, or maybe Mother Jones. Possibly even Esquire. (Perhaps some chapters were essays originally; if so, the copyright page gives no indication of this.) But as a book, it never goes anywhere.
The fundamental problem, I think, is that David W. Moore believes in polls. He believes that, taken properly, they can yield valid data. He believes that, used properly, they can do tremendous good. I do not know why. I kept waiting for the chapter in which he would make the case for the utility of polls, but he never really does; the belief goes so deep that it's taken for granted. When he mentions, on the fourth-to-last page, that he wrote another book sixteen years ago about how nifty polls were, it comes as no surprise. (Perhaps he figured that having made that case sixteen years ago, he didn't need to recapitulate it here. If so, he was wrong.) I think this belief in polls may account for his driving his argument into the ground; he's trying to battle a resistance that he feels, and perhaps others in his field feel as well. He's trying and trying and trying to convince those who really don't want to believe that polls could be bad.
I have no such problem. After the first few dozen pages, I was ready to declare that all political polls sucked, that they had no utility whatsoever, and that the only good polls worth taking (or paying any attention to) were the actual elections themselves. And a counterargument for this never came. While he proposes some specific ways to fix some of the specific problems in some polls, he never makes a case for how—if all his suggestions were taken—the public would be better served than by having no polls at all.
(The closest he comes is in telling the history of George Gallup and his hopes that polls would give the public more of a voice in national policy, but the exposition is apparently enough. He appears not to see any need to establish that this could really work in practice, or would even be a good thing.)
Recommendation: read the first chapter. Or almost any one of the others; they're close to interchangeable.