|
The presidential race may still be too close to call the winner -- it's "a statistical dead heat," as the new parlance goes -- but it's not too soon to declare a clear loser: the polling industry. All the returns, in the form of polls, aren't in yet, but the magic number to allow us to call it has been reached: enough.
In a move emblematic of the ludicrous results that have characterized this campaign season's polling headlines -- "HE'S DOUBLE DIGITS UP!" "NO, NOW THE OTHER GUY IS UP!" -- CNN released two separate polls with two very different findings on the same day last week. At 6:23 p.m. on Friday, CNN/USA Today/Gallup proclaimed that George W. Bush held a 13-point lead over Al Gore, 52 to 39 percent. This would have been ridiculous in itself, but the absurdity was compounded by the release -- a mere two hours later at 8:36 p.m. -- of a CNN/Time poll that had Bush's lead at 6 points. And, no, W. had not been caught buying an eight-ball of blow at 7:02 p.m.
As if this were not enough, the compounding was compounded by CNN polling director Keating Holland's explanation that the wildly divergent polls were "statistically in agreement ... given the polls' margin of sampling error." Bewildered by Mr. Holland's assertion that a 13-point lead is somehow "in agreement" with a 6-point lead, I called him to clarify.
Instead, I got a lesson in numerical alchemy. "Take Bush's number in our first poll," he explained patiently, "52 percent. With a margin of error of 3.5 percent, that means it could have been as high as 55.5 percent or as low as 48.5. Same with Gore's numbers: He could have been as high as 42.5 or as low as 35.5." He then walked me through the same margin of error calculations with the second poll, reaching the conclusion that anything from Bush being ahead by 20 points to Bush and Gore being tied would fall within the range of results "statistically in agreement." Or as he put it: "They agree as close as you can demand two polls to agree." In other words, two polls -- one showing a landslide, the other a dead heat -- do not contradict each other in the through-the-looking-glass world of polling.
I was stunned. Call me naive, but I have gone through 25 years as a political junkie under the impression that the familiar "margin of error" referred to the spread between the candidates and not to each candidate's percentage. Not so. It turns out that the margin of error should be doubled when it refers to the spread.
Frank Newport, president of Gallup, confirmed that the way the margin of error is reported is misleading. "You'll be interested to know," he told me, "that we have discussed in our meetings the implications of journalists focusing on the gap between the candidates rather than on each candidate's percentage." It would have been nice if they had let the public in on these concerns. After all, it's focusing on the gap that gives the polls some predictive value. The alternative, applying the margin of error correctly, offers such dramatically varied possibilities that the only legitimate response has to be: Why bother?
On top of pollsters admitting that their response rates have plummeted to 35 percent and that variables such as the order of questions, weighing for party affiliation, the size of the polling sample or the time the poll was taken all influence the result -- we now have the revelation that the infamous margin of error number should actually be doubled.
But polls have become established as the lingua franca of political punditry and basic cable. Without them, the entire CNBC prime-time lineup would be reduced to 60-second interstitial blurbs in between the latest market reports. Practically every political show leads with the latest polling results and then turns to its correspondents to explain what they "mean." The truthful response -- nothing whatsoever -- is seldom the answer.
When the public hears that a race is a statistical dead heat, it assumes that the candidates are tied. In fact, there is no such clarity in a result that we now know has a plus or minus 6-to-8-point margin of error. Maybe reporters should be required to spell this out, but they'd probably just race through it like those pro forma legal disclaimers tacked on to the end of car commercials. You can just picture Peter Jennings closing out a poll-based story with "Candidate totals are merely Pollster Suggested Figures. Does not include diminished response rate, increased margin of error or Dealer's prep. The electoral results you get may vary. Polling data not good in all 50 states and Puerto Rico. If ingested into the brain, induce vomiting immediately and disconnect your cable."
Will this latest news lessen the media's habit of chewing over polling results as if some collective wisdom will be produced out of all this mastication? What gives me optimism is that for the first time a critical mass of prominent voices is being raised against the dominance of polling. From William Safire and Alison Mitchell at The New York Times to John Fund at The Wall Street Journal and Richard Benedetto at USA Today, the disgust with this journalistic junk food is rising.
And this conclusion comes with a margin of error of plus or minus zero.
[ Printer-friendly version ]
|
