Jewish World Review Sept. 21, 2000 / 20 Elul, 5760
John H. Fund
All this makes you wonder why we still have campaigns to capture the attention of inattentive voters and change their minds. The polls vary in their methodology, and too often the poll-obsessed nature of campaign coverage freezes public perceptions of the candidates and contributes to the low voter turnout everyone in the media claims to be against.
The Newsweek poll turns out to be of only 580 likely voters over two nights. One of them was a Friday, when pollsters say more Democrats are at home. In 1996, Newsweek's late-October survey had Bob Dole losing to Bill Clinton by a whopping 23 points. Mr. Dole lost by eight. (Newsweek's pollster says it did a final but unpublished survey that showed an 11-point margin.)
Other polls with a better track record show different results. The bipartisan Hotline poll shows the race tied. The Battleground poll, a joint project of Democratic pollster Celinda Lake and GOP pollster Ed Goeas, got the 1996 Clinton victory margin right within 0.5% of the vote. As of this Tuesday, their nightly tracking poll has Mr. Bush leading Mr. Gore by four points.
Similarly, the Rasmussen Research nightly tracking poll now shows Mr. Bush with a 44% to 41% lead among likely voters. Mr. Rasmussen uses a telephone prompting system rather than live interviewers to survey respondents, a controversial technique. But an analysis by the Progressive Review found that his polls -- along with those of Gallup and Zogby -- proved to be the most accurate during this year's primaries, and his lower costs allow him to survey more people and reduce his margin of error to 2%.
The third nightly tracking poll is Gallup, which has a good record even if its final 1996 survey overstated Mr. Clinton's winning margin by three percentage points. It currently shows Mr. Gore holding a four-point lead over Gov. Bush. Averaged together, the three nightly tracking polls show a one-point Bush lead.
In recent years, election polls have often overestimated the Democrats' share of the vote. People like to tell pollsters they plan to fulfill their civic duty and vote. In reality, about one-third of people who say that won't show up. The best polls figure out if people are engaged enough to be likely voters. Even with those efforts, polls often oversample groups such as blue-collar women that end up not voting in high numbers. Turnout, which in 1996 fell below half of all adults, does matter. The latest Wall Street Journal/NBC poll has Mr. Gore leading among registered voters by 45% to 42%, but Mr. Bush leading by 45% to 43% among likely voters.
There are other reasons polls may tilt slightly toward Democrats. Warren Mitofsky, who developed exit polling for CBS News in the 1960s, believes Democrats are more likely to respond to media polls than are Republicans, who may distrust the "liberal" news media. More than 60% of those pollsters try to contact routinely hide behind answering machines or otherwise refuse to answer. "This makes survey results more uncertain, and should cause concern, caution and above all humility in reporting the results," says Leo Bogart, a former president of the American Association for Public Opinion Research.
That humility was not found among the media in 1996, when almost every major poll overstated President Clinton's victory margin. Purdue University professor Gerald S. Wasserman says his mathematical analysis of the 1996 polls showed them to be "a collective failure," all erring in Mr. Clinton's favor. He found that only once in 4,900 elections would chance alone produce that. "To continue to use current polling technology without calling for a change would be like having a gambler play after the roulette wheel comes up red 12 times in a row," he says. "In both cases, it might just be chance, but any sensible person would stop and check the apparatus before going on."
Unfortunately, polls help dictate media coverage. Reporters overemphasize certain elements of a candidate's personality -- Mr. Gore's penchant for exaggeration or Mr. Bush's verbal flubs -- if the polls turn against one or the other. Then until the polls turn again, coverage conforms to the stereotype. ABC's George Stephanopoulos, a key strategist in Bill Clinton's 1992 campaign, said last week that political scientists "talk about the bandwagon effect, that once a candidate gets in the zone, all of the coverage is good, almost no matter what happens, and when you're out of the zone, even when do you do things right, it goes against you."
The polls are a prime mover in this bandwagon effect. But before they
drive a candidate's supporters to despair or allow the media to beat a
stereotype to death, some basic consumer reporting would help. Reporters
should tell us which polls have a good track record, which have been
clunkers, how much pollsters "push" the undecided to a candidate, and how
much less reliable some individual state surveys are. To do anything less
risks heightening the cynicism of an electorate that increasingly feels
it's being manipulated by both parties and not told the whole story by the