1. Archive

The Pollster's Plight

The oracles of politics, like those of baseball, have been stumbling. Consider these analytical foul balls from recent elections: Two days before last month's Democratic primary for Massachusetts governor, a KRC Communications poll showed Frank Bellotti leading John Silber by 9 percentage points. On Election Day, Silber romped to a 10-point victory.

Four days before the Louisiana U.S. Senate primary, a Mason-Dixon poll showed ex-Klansman David Duke with 26 percent of the vote to 53 percent for incumbent Democrat Bennett Johnston. Duke drew 44 percent of the vote on Election Day.

Two days before Florida's Democratic primary for governor, the New York Times Florida Newspapers published a poll showing Bill Nelson trailing Lawton Chiles by just 12 percentage points. On Election Day, Chiles won by 39 percent.

Less than a week before the Virginia gubernatorial election last November, a Washington Post poll showed Democrat Douglas Wilder leading Republican Marshall Coleman by 11 points. On Election Day, Wilder edged Coleman by less than 1 percent.

These Election Day surprises could be mere aberrations. But they at least raise some intriguing questions in a country grown addicted to polls not simply for politics but for all manner of commerce as well. Could the high-tech people meters that guide elected leaders and corporate executives be teetering out of alignment? Are current political developments so swift and volatile that surveyors can't keep up? Or are Americans starting to heed columnist Mike Royko's advice: to lie to the pollsters?

Some would consider these welcome developments in an age when we've come to expect television networks to call the outcome of an election hours before the polls close. "People don't like to be told what they're going to do before they do it," observes Mark DiCamillo, managing editor of the independent California Poll. "People love to prove the so-called experts wrong."

Americans probably haven't started lying to make

that happen. But there are other reasons why election results are making more and more polls look bad.

One is the proliferation of bad pollsters. "Don't assume a conspiracy until you've ruled out incompetence," quipped Harrison Hickman, a Washington-based pollster for Democratic candidates. Hickman and others point to surveys conducted by news organizations that skimp on costs by conducting too few interviews or failing to identify a representative sample of the public.

The elusive random sample

Pollsters also complain that news organizations sometimes misuse their results. Brad Coker, president of Mason-Dixon, says his Louisiana poll showing Duke with 26 percent cautioned that a "hidden" vote could bring the ex-Klansman up to 40 percent; reporters mostly ignored it. Gary Orren, a political scientist at Harvard's Kennedy School of Government, notes "a mismatch between the culture and ethos of journalism" _ which places a premium on speed, drama and novelty _ and that of conventional survey research, which values painstaking and cautious analysis.

Yet even the heavyweights face increasing difficulty finding representative samples of the population to analyze. In growing numbers, time-pressed Americans are refusing to respond to pollsters. "Random samples are becoming impossible to obtain," Don Schultz, a journalism professor at Northwestern University, told the New York Times recently. "You may have large numbers of people who hold violently different views from those who are willing to talk."

Further complicating the pollster's task is the long-running decline in voters' affiliation with political parties. Party identification remains the most reliable single indicator of someone's political leanings. But it is much less reliable than it used to be, especially among the young, who are apt to disregard party loyalty in response to campaign developments.

For all the technological progress and accumulated experience of recent decades, the lack of stable party ties is the major reason why "on the whole, polls are further off today than they were in the 1940s and 1950s," according to Everett Carll Ladd, executive director of the Roper Center for Public Opinion Research at the University of Connecticut. Primary elections like the Chiles-Nelson contest, in which party label plays no role whatever, are easiest for pollsters to trip up on.

The opening of the political process over the last quarter-century has also produced increasing numbers of unconventional, unpredictable candidacies. Though women candidates have become commonplace, some consultants believe survey results overstate their prospects by a couple of percentage points. Their explanation: Some voters remain uneasy with the idea of female office-holders but are loath to say so to an interviewer.

The problem of race

By some accounts, gauging candidacies with racial implications is even trickier. Larry Hugick of the Gallup Organization examined eight campaigns pitting black Democrats against white Republicans, including Wilder's bid for Virginia governor against Marshall Coleman. In each case, the black's Election Day showing trailed his final poll result by margins ranging from 5 to 12 percentage points.

"If the black candidate is ahead (in a poll) by less than eight points, chances are he's not ahead at all," Hugick says.

Pollsters disagree among themselves about whether this is so and why. Some believe that respondents, on the touchiest of questions, will give what they consider the socially acceptable answer whether or not it reflects their actual preference. Before the recent Nicaraguan election, a University of Michigan researcher found that face-to-face polls varied widely depending on whether the interviewer's pen was in the colors of the Sandinista party or the UNO party, which won. In racially-sensitive contests here, such as the Duke and Wilder campaigns, some pollsters suspect responses to black and white interviewers vary. But a review by the Washington Post of its last Wilder-Coleman poll found that black and white interviewers reported very similar numbers. The California Poll reached the same conclusion after reviewing its survey in the 1982 governor's race between black candidate Tom Bradley and his white opponent George Deukmejian. DiCamillo said an unexpectedly high turnout of gun enthusiasts, not race-related factors, led to the 8 percent discrepancy between the final poll result and Bradley's narrow defeat.

And indeed, voter turnout is harder to anticipate than broad public preferences. "We've got to factor in intensity (of support) as well as direction," notes Ted Arrington, a political scientist at the University of North Carolina-Charlotte. "We don't know how to do that."

The problem is exacerbated in the low-turnout, mid-term elections _ especially primaries _ that take place between presidential contests. Burned by the 1989 Virginia governor's race, the Washington Post decided a "predictive" poll immediately before the recent D.C. mayoral primaries wasn't worth the effort.

One can only fantasize about the ramifications if more American news organizations, politicians and business leaders decided to show similar restraint. Movie producers, for example, might have to rely on their directors instead of focus groups to determine the right ending for film dramas. Reporters might devote more attention to substantive differences between candidates than to their poll standings. And candidates, instead of consulting their pollsters on tough policy questions, might once again have to consult their guts.

The turmoil of changing times

Which leads to one more possible source of scattershot poll results _ the roiling gut of the American body politic itself. Paul Tully, political director of the Democratic National Committee, ticks off the unusual number of big-ticket issues the public had to digest in recent months: the liberation of Eastern Europe from Soviet domination, the possibility of war in the Persian Gulf, the savings and loan bailout, the federal tax and budget debacle, growing fears of recession. As voters take it all in, currents of opinion can shift quickly.

"It's like asking somebody, "What do you think of this baseball?' _ and then you throw it 100 miles an hour at their head," says Tully. "It'll take them a while to get on their feet and find the ball."