Repoll man; one politician's solution to the preelection polls that almost sunk him: one more poll.

Readers of the May 16, 1992, Oregonian couldn't help but draw a conclusion about my bid to win the Democratic nomination for secretary of state, Oregon's second ranking state office: I was toast.

With just three days until the primary, I was 11 points behind my chief rival, the state's four-term labor commissioner, Mary Wendy Roberts. True, I'd made some progress. Six weeks earlier, I'd been 25 points behind, according to an earlier Oregonian poll. But this Il-point gap, so close to the election, was bleak news. I began working on my concession speech.

Three days later I scored an "upset" victory by a convincing seven-point margin. So what happened? What combination of my brilliant strategy and/or my opponent% strategic blunders had tipped the balance? Did all the news-. paper endorsements turn the tide? My record of redistricting the legislature and conducting aggressive performance audits of government programs? Or was it the stirring Aaron Copeland music that accompanied my television commercials?

Danreed if I know. And neither the Oregonian nor anyone else seemed interested in finding out. That's because, beyond the requisite exit polls--whose sole object is to establish the winner as early as possible--polling almost always stops on election day, depriving candidates and voters of answers to a question more important than "who?" Why?

The why of elections could easily be answered with another poll, a day or two after the fact, asking voters what factors influenced theft decisions. This sort of postmortem would shed light on how to fine-tune future preelection polls. Postelection polls might even wean the press and deep-pocket contributors--who often affect races by making more than they should of these surveys--from their current poll addiction. But perhaps their most important function would be served well after the election is over. Postelection polls might give all of us a better sense of what is important to American voters.

Poll cats

The absence of postelection surveys is puzzling, given the increasing prevalence of media-financed polling, and it's not just the CBSs and Newsweeks of the world that devote tens, or hundreds, of thousands of dollars to preelection surveys. Today, a growing number of daily newspapers and local TV stations are rushing into the act. More and more, their polls are going beyond the "top of the ticket" presidential and senatorial races to include state, mayoral, and even city council contests.

Media-sponsored polls have considerable appeal for candidates (we save money) and also a broader logic. When major candidates conduct their own polls, they inevitably try to spin the results to gullible journalists. Thus, there's arguably civic value in surveys that can be characterized as "objective." However, as my race illustrates, there are real dangers in this trend. For one, polls can mislead by testing name 'familiarity rather than true preference. They can also be downright inaccurate if the questions are badly phrased or a sample improperly drawn. For example, a poll that samples from all registered voters--rather than from the much smaller pool of voters likely to cast ballots in a particular election--will often give a much different (and less accurate) picture.

Yet perhaps most damaging aren't the numbers themselves, but how they play out in a campaign-- creating a phenomenon similar to the Heisenberg uncertainty principle known to physicists. Scientists have learned that the mere act of looking at certain subatomic particles in itself disrupts the reality of what's there. (In a nutshell, the light waves necessary for "seeing" the particles change their location.) The political version of this phenomenon occurs when polls, especially inaccurate ones, create their own political reality. A candidate at the short end of an early poll--one who seems to be lacking "big Mo"---can suddenly find fundraising calls unreturned and volunteers vanishing. Candidates don't have the resources to get their message out; reporters write about a "struggling effort"; and so on, in a familiar downward spiral. In effect, the media's quest for news creates news itself. Unfortunately, that news isn't always based on terra fnma.

Inaccurate polls occasionally afflict higher-profile races, too. Two days before the 1990 Senate race in New Jersey, the Newark Star Ledger reported the resuits of a poll that showed Bill Bradley burying Republican challenger Christine Todd Whitman by 17 points. On election night, Bradley escaped with a three-point victory. In the Michigan gubernatorial race that same year, the Detroit Free Press published a poll that forecast a 14-point drubbing for Republican challenger John Englar. Governor Englar, as he is now known, won by. 1 percent of the vote.

Of course, the bigger the candidates, the more they can protect themselves from bad polling. Misleading or inaccurate polls in presidential or U.S. Senate races are produced against a backdrop where issues are more prominent and voters have at least heard candidates' names. Incumbents and hopefuls alike have more options for fighting back and engaging voter attention. But in more obscure races like mine, sexy issues can be hard to come by, and finding money for a mailing or two--much less for constant polling--can be difficult. Getting the public to even notice your race is hard enough. Indeed, beyond a handful of highly visible contests, voter interest in the lower end of the ballot plunges precipitously. It's here that polling can do the most damage.

In my own race, I had some ammunition to help fight the tyranny of polls. Before the first one showed me 25 points behind, we paid $12,000 for our own survey. We called 2,000 people and, by screening out unlikely voters, produced a sample size of 600. In this survey, we were just seven points behind. The Oregonian poll had not screened and used just 283 respondents. Thus, we had a plausible spin when potential supporters questioned our ability to win.

Still, the media's polls had a visible effect on my campaign--though not, interestingly, on the general public. In my encounters with voters, few even seemed to know about these polls, much less take an interest in them. Yet almost every time I went to a fundraiser or called a political colleague, the topic came up. And the dim news from the Oregonian directly affected our ability to raise money, especially from the political action committees and corporations that tend to gravitate toward likely winners.

Despite my incumbent status, I failed to receive endorsements from many of the major PACs interested in my race--such as organized labor and teachers. While my major opponent raised more than half of her funds from PACs and corporations, we ended up raising more than 80 percent of our money from individuals.

The sins of errant polls are multiplied when news organizations conduct no postelection surveys. Such surveys would likely tell us, in those instances in which preelection polls were wide of the mark, what exactly happened and why surveys failed to predict it. We might, for instance, discover that people are inclined to change their minds at the last minute-- which might make PACs more careful about exerting influence early in a race, when fledgling candidates are most vulnerable. Or we could learn that a given poll was poorly conceived-which would help us frame future polls so that they are more accurate and, therefore, more fair.

Currently, the closest the media come to postelection surveys are the ubiquitous exit polls conducted on election day. But even here, the primary motivation is to allow an early call-- "We beat CBS by 12 minutes on the Utah senate race !"--rather than to gain an in-depth understanding of the ebb and flow of voter decisionmaking. In the absence of postelection empirical data, journalists and political pundits are that much more inclined to speculate. Not surprisingly, their theories often look like efforts to rationalize their preelection analysis. Or they latch on to a clearly definable moment in a race, imbue it with great weight and meaning, and present it as a key turning point. They thereby cover an election much as a sportswriter covers a game, attributing a team's comeback to a key double play or an interception.

The conventional wisdom in Washington, D.C., for instance, attributed 1990's last-minute surge by mayoral candidate Sharon Pratt Kelly directly to an endorsement by The Washington Post--a winning three-pointer at the buzzer, as it were. But in contrast to sports contests, what happens in voting booths is beyond the public eye. And, especially in local campaigns, there's a general reluctance to look too hard behind the curtain, to try to ascertain (however inexactly) what really led voters in the end to prefer Candidate A to Candidate B.

Was it Kelly's promise to cut the lackluster city bureaucracy that put her over the top? Was it her tough-on-crime rhetoric? No one asked. The cynical might attribute that lack of curiosity to media selfprotection. After all, if voters' explanations of their behavior are wildly at odds with the preelection (and election night) handicapping, it raises questions about the acumen of the reporting staff.

Still, I think a likelier explanation is a simpler one: the continuing obsession among political insiders with the horse-race aspect of American politics. Staffers, lobbyists, reporters, contributors, consultants, and political junkies are more interested in putting a wreath around the winner, interviewing the loser, and preparing for the next trifecta than they are in a sober appraisal of what happened and why. On election night and afterward, the insiders don't much care. They are elated or discouraged-- but in either event it's time to move on.

Leaving an air of mystery around the final tally, not even trying to peek into the black box, allows bedrock assumptions to go unchallenged. It's all the easier to fall back on your favorite explanations, confident that, win or lose, you really do know the business.

But to understand what the public is missing, all you have to do is look at one race in which the media didn't retreat immediately after the returns: Harris Wofford's run for one of Pennsylvania's Senate seats last year. Largely because the race involved a national figure, former Attorney General Richard Thornburgh, it attracted extraordinary attention. And when David handily whipped Goliath, there was deep curiosity as to why. What postelection polls confumed --a fervent desire among the Pennsylvania middle class for health-care reform--is now tippling through races across the country.

Polling is clearly here to stay. If anything, future elections will use it even more frequently. (Look at presidential polls that have been issued on an almost daily basis since last spring.) And because we're going to have to live with them, it's important that we better understand them. We need to recognize the inherent limitations in the predictive power of surveys --and the dangers in giving them too much prominence. Even more important is taking full advantage of polls as an opportunity to better understand voters. It's time the media use their considerable resources to help explain to us--including those out there running, gabbing, and fundraising--what the voters are really trying to say.

Phil Keisling is a contributing editor of The Washington Monthly and Oregon's secretary of state.