I hope that this year’s fiasco persuades the news media to stop conducting and disseminating exit polls. Not because they are necessarily unreliable but because they have spoiled election night. Since at least 1980 the pattern has been consistent: Polls close in a bunch of states on the hour; with ten minutes the states are called based on exit poll numbers; airheads chat aimlessly for 50 minutes; repeat until ennui sets in. There is almost no old-fashioned discussion of how the vote count stands, where the counted votes come from or how the trends compare with earlier years. If a state’s exit poll isn’t conclusive, the talking heads don’t know what to talk about. It makes for a boring, uninformative show – and sometimes a misleading one. This year, for instance, it was clear from an early hour that President Bush would carry Florida, because he was running comfortably ahead of his 2000 margin in the northern and central parts of the state. The network commentary should have been emphasizing that fact and how difficult it would be for Senator Kerry to reverse the tide with returns from not-yet-reported Dade and Palm Beach Counties. Instead, only Republican operatives provided that information, and they were naturally discounted as partisan.
Modern computer databases make it simple to put a plethora of past election statistics at reporters’ command, and modern computer graphics allows the data to be presented in readily assimilable form. Just from a pure business point of view, wouldn’t that attract a bigger audience than long stretches of Dan Rather growing progressively more disconsolate?
One might think that, without exit polls, we would lose statistics on how various subsets of the electorate voted,
Although the aesthetic case against exit polling is overwhelming, I suppose that not everyone will consider it decisive. Millions of innocent words will be slaughtered in the debate over what went wrong with the 2004 surveys, if anything went wrong at all. The pollsters themselves, as I understand it, insist that they got everything right and have been misrepresented. Maybe so, but here are numbers, including virtually every state considered to be “in play”, that were being reported around the blogosphere shortly before the polls closed. I list Bush first and Kerry second, followed by the actual results and the difference between Bush’s predicted and real percentages.
Florida 49%-51% [52%-48% (-3%)]
Pennsylvania 46%-53% [49%-51% (-3%)]
Ohio 49%-51% [51%-49% (-2%)]
Michigan 47%-51% [49%-51% (-2%)]
North Carolina 52%-48% [56%-44% (-4%)]
Missouri 54%-46% [53%-46% (+1%)]
Arizona 55%-45% [55%-44% (0)]
Minnesota 44%-54% [48%-51% (-4%)]
Wisconsin 47%-52% [49%-50% (-2%)]
Colorado 51%-48% [52%-47% (-1%)]
Louisiana 56%-43% [57%-42% (-1%)]
Iowa 49%-49% [50%-49% (-1%)]
Arkansas 53%-47% [54%-45% (-1%)]
New Mexico 49%-50% [50%-49% (-1%)]
None of the deviations is huge. Individually they fall within the poll’s margin of error, but it is certainly striking that, in 12 of 14 cases, the Bush percentage is understated. Moreover, the larger states tend to have larger discrepancies. I am hesitant to offer an explanation, but innocent ones are not easy to formulate.
Anyway, I don’t want innocent explanations. When returns start arriving on November 7, 2006, I want to hear much more from guys like Michael Barone and Larry Sabato, who know what elections are about, and as little as possible from Dan Rather, Peter Jennings and Tom Brokaw. The lower the credibility of exit polling, the more likely we are to see real analysis.
Comments