Political pollsters are supposed to feast during a presidential election year. So why are so many of them worried that famine waits over the horizon?

		More on Campaign 2004 Edwards' wife aims at Bush on terror in visit here Hoeffel ad: Specter, Bush are birds of a feather Specter ad is fast and loose with the truth

The reason is that people are getting harder to reach and survey response rates are getting lower. And they say it's only going to get worse.

Potential respondents screen calls with caller ID or answering machines and either don't pick up or don't return messages. Millions rely solely on cell phones and legally cannot be called for surveys. And millions of new voters have registered who may or may not be found by pollsters.

The problems are magnified because of the role polls now play in shaping the electorate's perception of a race, as well as in influencing a candidate's strategy and ability to raise money.

Nationally, about 6 percent of wireless phone users go without traditional land-line service, according to the Yankee Group, a communications research firm from Boston. The percentage approaches 12 percent among all people under 35 and rises to 14 percent among 18- to 24-year-olds. Cell numbers aren't listed in phone directories, and even if they were, pollsters are not allowed to call them.

So if these young, high-tech voters can't be reached, does that mean polls might be off by a point or two or three?

Maybe. Maybe not.

"The issue of the cell phones is a problem, but it's not a catastrophe," said Shawnta Wolcott, communications director for Zogby International, a top polling firm. "To compensate, [it] takes us a little longer to reach a balanced sample, so we just stay on the phones a little longer."

Eventually, she said, Zogby reaches enough younger voters to meet the poll's demographic criteria. The only way poll results could be tainted is if cell-phone-only professionals have different voting tendencies than landline users in the same age bracket, she said, and there's no evidence that this is the case.

For pollsters, it's a happy circumstance that 18- to 34-year-olds on the whole are less likely to vote than older Americans and therefore are less vital to the polling process.

Still, as the number of people who drop landline phones grows -- and assuming it remains illegal for pollsters to call mobile phones -- it will become increasingly difficult to conduct surveys.

Pollsters could face huge problems by the next presidential election in 2008, and by 2012 or 2016, the landline poll will have gone the way of the dodo.

Polling firms are constantly exploring the need to find new ways to reach the right people, and e-mail is gaining the favor with some in the industry. Zogby already uses "interactive polling" -- e-mails that direct poll participants to a Web site to answer questions -- but other firms, such as Gallup, aren't so sure about the reliability of Web polling.

The rise of random sampling

The American political poll was born in Pennsylvania, in 1824, when a Harrisburg newspaper tried to predict the presidential election, and since the beginning, there have been difficulties.

Straw ballots, which newspaper readers were asked to fill out and mail in, were unreliable. Canvass polls, where workers were sent to polling precincts, were marginally better.

It wasn't until the 1930s that scientific "sampling" -- the process of finding groups of random individuals who mirror the demographic characteristics of a given population -- became the method of choice.

Since then, the pollster's top challenge has been finding an accurate sample population, then getting in touch with those most likely to vote. Since World War II, the telephone has been the primary tool for conducting scientific surveys, but as the Chicago Tribune discovered, when it relied on poll results to craft its 1948 "Dewey Beats Truman" story, calling people on the phone has its limitations.

It turned out that people without telephones who couldn't be contacted had different voting patterns than people who had phones. Could the same eventually prove true of the cell-phone-only crowd?

"I don't think any of us know the answer to that," said Carroll Doherty, editor of the The Pew Research Center for the People and the Press.

Other obstacles to modern polling include screened calls, voice mail and caller ID. Women, blacks and people younger than 30 are more likely to block or not answer a call than someone older. The national and state Do Not Call lists lead to a lot of hang-ups on pollsters, who are allowed to call landline phones but are mistaken for telemarketers.

"People are dubious of pollsters to begin with because of privacy issues," Doherty said.

Response rates -- people who agree to answer questions -- have dropped to around 25 percent from nearly 40 percent a couple of decades ago, said Larry Jacobs, director of the 2004 Elections Project at the University of Minnesota.

The future of polling is hard to predict. Almost certainly, an alternative to the phone poll will be required. But will it be one that relies on the Internet? Face-to-face or door-to-door polling, which is extraordinarily expensive? Old-fashioned snail mail? Maybe a rewards-based system, in which prospective respondents are offered something of value in exchange for their time and opinions?

"Three elections from now, maybe we'll have one machine that will be the phone, Internet and TV," Doherty said. "Who knows?"

Even if pollsters sort out the technical challenges, they'll still face difficulty in figuring out which Americans are most likely to vote.

Polling is art and science

Polling is based on well-established statistical principles, but pollsters rely on art as well as science in trying to discern who will actually cast a vote.

Different polls use slightly different "screens,'' or series of questions, to determine who among their respondents are most likely to vote. The task is complicated by the fact that many people claim they will vote and then they don't.

Some sincerely plan to vote and simply don't get around to it. Others don't genuinely plan to vote, but will say "yes" because it seems the socially correct answer. As a result, questions designed to discern "likely voters" attempt to tease out a respondent's interest in the election, level of certainty about voting and voting history.

The respected Gallup Poll is among those that rely in part on past behavior. People are asked whether they know the location of their polling place and whether they voted in past elections.

One key question this year is whether the flood of newly registered voters -- more than 40,000 in Allegheny County -- will confound pollsters' expectations about the identities and tendencies of likely voters. Adam Clymer, political director of the National Annenberg Election Survey, said turnout among new voters is a potentially huge variable.

"I have a lot of sympathy for my colleagues who have to report the horse race number," Clymer said. "This represents a real challenge. We don't know how many people are going to vote, but most people think it's going to be substantially more.''

The classic example of voter turnout defying polls was Jesse Ventura's 1998 election as governor of Minnesota. A last-minute surge of support, much of it believed to have come from newly registered, nontraditional voters, brought him a stunning victory that no poll had predicted.

"In the case of Ventura, the reality on the ground changed overnight," said Jacobs, of the University of Minnesota. "The universe of voters destabilized in the last few days -- we saw a huge surge of voters that didn't vote in the past."

Brad Coker, managing director of Mason-Dixon Polling & Research, which conducts the Pennsylvania Poll for the Post-Gazette, said that while new, nontraditional voters were a factor in the Ventura race, their effect may have been exaggerated.

"There might have been some of that, some slackers who said, 'Hey, dude, let's go vote for Jesse,' " Coker said.

"But everyone forgets there was a widely carried debate the weekend before the election and Ventura just mopped the floor [with his opponents.] "That debate generated a lot of that enthusiasm for him. You could have been polling the right people [and] still missed a lot of that Ventura surge."

Jacobs said the Ventura race was "instructive but not a real useful guide" to current polling perils. But pointing to reports of widespread voter registration increases across the county, he said, "I do worry that some of the polls out there may not be capturing this higher level of voter interest."

Coker added a note of skepticism regarding how many of the highly touted new voters would actually show up on Nov. 2.

"Registering voters is one thing, getting them to turn out is another," Coker said. "Republican are working, the Democrats are working this real hard; we'll see if they can pull it off."

G. Terry Madonna, director of the Center for Opinion Research at Franklin and Marshall University, said he was not particularly concerned that a surge of new voters would elude his Keystone Poll or other reputable surveys when it comes to the presidential race.

"Presidential elections are easier to do in one sense because the turnout is so large that likely voters more closely mirror all voters," he said.

Jacobs said one problem with polling is the fact that many lay people simply don't understand the concept of margin of error and are confused by seemingly varied results. If two candidates are tied in a poll with a margin of error of plus or minus three percentage points, and another poll with a similar margin of error depicts one candidate with a six-point lead, it is possible the polls are reporting identical snapshots of opinion, with each candidate at the outside edges of the margins of error.

"The reality is that most of the polls [are] being done very well,'' Jacobs said.

Madonna noted that a variety of different polling organizations have been taking the pulse of the Pennsylvania presidential race this year, with relatively consistent assessments over time.

"They're all showing the same election,'' he said.

Which remains too close to call.