Is online polling.

The way political pollster John Zogby sees it, online polling is a ship that has already sailed and is proving itself as the vessel of choice for quickly, cheaply and accurately handicapping campaigns. Anyone in his business who has not caught the Internet wave is missing the boat, like the naysayers who were slow to embrace telephone polling 35 years ago.

[ILLUSTRATION OMITTED]

"I think there are some who would have advised Columbus to not sail west because he would fall off the Earth," Zogby said.

Critics of online political polling, however, are legion. They wonder how any pollster could stake a reputation on surveys that under-represent blacks, Hispanics and senior citizens--groups much less likely to be wired than whites. The lower minority penetration on the Internet has given rise to the term "World White Web."

Nearly three-quarters of Americans use the Internet, according to the latest survey by the Pew Internet & American Life Project. And while Internet access is improving among blacks and seniors, those groups still lag behind whites. Studies also show that younger, more affluent Americans tend to use the Internet more.

"There are grave sampling limitations trying to do a horse race survey with an online sample. It's a huge problem," said Nancy Belden of Washington-based Belden Russonello & Stewart and past president of the American Association of Public Opinion Researchers. "They're ignoring the people who are totally not online at all. Just because you get it right once or twice, or some of the time because you correct your sample through weighting doesn't mean you're not missing something that's out there."

Weighting is a statistical technique used to ensure representation of certain groups in the sample. Data for under represented groups are weighted to compensate for their small numbers, making the sample a better representation of the underlying population, according to the U.S. Bureau of Justice Assistance.

"The allure of Internet polling has several parts to it," she said in response to Zogby's praise for the technology. "It's much cheaper. It's a lot faster. What's the other argument for it? There's nothing better about administering it that way. You wouldn't send Columbus sailing on a cheaper, faster boat without the latest navigational equipment onboard. He could get lost or fall off the face of the Earth."

More and more Americans are relying on cell phones, which, so far, has put them out of the reach of political pollsters who call landlines. But the telephone will remain the gold standard of political polling for several more years, some experts insist.

"Telephone polling is still going to be far more accurate," said Jef Pollock, president of Global Strategy Group in New York. "Not everybody is online. There are all sorts of problems with online sampling. If you call 10 campaigns and talk to 10 candidates, not a one will accept an online poll for their candidate because they believe they would be skewed. None I know of would use it.

"Over time, the phones are going to become more problematic for us as researchers. We all know that," he said. "But for now, they're going to be a much better bet. It's a way of getting the most accurate information from governor, Senate, House and state races and lower."

GOP pollster Ed Goeas of the Tarrance Group agreed with Pollock. "We don't buy into that quite yet," he explained. "We still have to kind of question it, particularly the most important voter group: senior voters. Are you getting the saturation into that group that you would expect?"

Zogby scoffs at the critics noting how his firm fared in the 2004 presidential election. Zogby Interactive called 17 of 20 states correctly (it missed the tight races in Ohio, New Mexico and Iowa). Nationwide, Zogby Interactive's poll had George Bush defeating John Kerry 50 percent to 49 percent; the final was 51 percent to 48 percent.

Last year in the Virginia gubernatorial race, Zogby Interactive's Internet poll picked Democratic Lt. Gov. Tim Kaine to defeat Republican Attorney General Jerry Kilgore by 48 percent to 46 percent, and the final tally was 52 to 46. In New Jersey, Zogby Interactive projected Democrat Jon Corzine to win over Republican Douglas Forrester by 51 to 44, very close to the actual results of 53 to 44.

Zogby conducted six years of research and development before unveiling the online poll in the 2004 presidential election. Those results, combined with 2005, have validated the company's technology and methodology, he said.

Here's how Zogby Interactive works: The Utica, N.Y.-based firm has compiled a database of several hundred thousand voters. These were people solicited since 1998 on the company's Web site to participate in online polls. The database also includes participants of Zogby International's random telephone surveys who then were asked to submit their e-mail addresses. Zogby Interactive calls about 2 percent of the online respondents to validate their personal data, such as home state, age and political party. The database includes liberals, conservatives and moderates, voting diehards and the apathetic, the young and the old.

Roughly 100,000 respondents are picked randomly for each online poll. Interactive polls are supplemented by telephone surveys to "ensure proper demographic representation, especially among hard-to-reach groups," the Zogby International Web site states. Those telephone follow-ups allow Zogby to weight his polls against under-representation of blacks and Hispanics.

"We really are near universality in terms of Internet access," Zogby said. "Seventy-four percent of all adults and 91 percent of likely voters have access to the Internet at home. It really is the wave of the present and even more so the wave of the future."

Zogby's numbers are overstated, according to the Pew Internet Project. The study found that 73 percent of adults have access to the Internet, and 90 percent of those have it in their homes. Fifty-seven percent of blacks use the Internet, while about 70 percent of whites and English-speaking Hispanics are online.

The Internet usage percentage rate for young adults is in the upper 80s and begins a gradual slope down as people grow older, said John Horrigan, associate director of the Pew project. "When you get over the age of 70, Internet access really drops off," he added. And people 70 years old and above are among the stalwarts at the polls, even in off years, because their generation sees voting as more of a civic obligation.

Horrigan has no current data on Internet usage by registered voters, but doubts Zogby's 91 percent figure. After the last presidential election, for example, 66 percent of the respondents Pew researchers interviewed claimed to have voted and to have access to the Internet; at the time, Internet penetration was 59 percent versus today's 73 percent.

"Everyone isn't online," said Maurice Carroll, director of the Quinnipiac University Polling Institute in Hamden, Ct. "But one thing is clear: Things are going to change. Every pollster is experimenting or at least thinking of experimenting to make sure that the technology doesn't invalidate what we're doing."

[ILLUSTRATION OMITTED]

Bob Roberts, a political science professor at James Madison University in Harrisonburg, Va., said online polls are no more accurate than those conducted by Fox News, which would skew toward white, conservative, upper-income voters.

"That doesn't tell us how the entire population is perceiving things," said the professor, who compares online polling to telephone voting for "American Idol." "Those polls theoretically could be manipulated.... It's not going to be scientific."

Zogby, though, touts the response rate to Internet surveys.

"The telephone is pushing us away," he said. "When I started in 1984, average response rates were 65 percent nationally. Today, we're at about 28 to 29 percent. If you get into select metropolitan areas like New York City and Southern California, it's 8 to 9 percent. People mostly don't want to be bothered."

No public opinion pollster doubts the promise of the Internet and the value of select uses. There's no better way to shop political ads than on someone's home computer screen. The feedback is virtually instantaneous, cheap and avoids the collective thinking that can afflict focus groups. Internet polling also is an effective tool at surveying consumers about toothpaste and other goods. You can reach consumers day and night, while telephone polls target voters around the dinner hour and up to 9 p.m., when they don't want to be bothered.

The current method of testing advertisements is to convene a focus group or corner people at a shopping mall. It's time consuming, expensive and narrows the list of respondents, resulting in tiny samples. Political consultants now screen their ads for Internet users in the privacy of their own homes.

The Denver firm iModerate has taken the concept to the extreme. It uses national panel companies such as Greenfield Online, which has a database loaded with millions of people who clicked "yes" on pop-up ads that asked if they wanted to take a survey.

iModerate then marries the panel company's list to the database kept by Voter Contact Services, a private firm that has been compiling computerized data for 30 years and includes the respondent's party affiliation and vote history. "We have a different sample methodology," said iModerate co-founder Carl Rossow. "When we draw a sample to send out invitations for a survey, it is a representative sample. There are definitely gaps, but I think those are getting smaller and smaller every day."

What iModerate does is inject online chats into its clients' surveys. A trained moderator is able to send an instant message to the respondent for a live chat, allowing iModerate to probe deeper into the survey answers.

"Online surveys are very static," said Adam Rossow, director of marketing and communications (and Carl's brother). "There's no interactivity. This is providing an outlet so the people's real thoughts and emotions and feelings can be heard.... The Internet is becoming so popular for polling. It creates anonymity. When you're typing online, you'll say anything without risking embarrassment. We're kind of getting at the true voice of the consumer."

While the technology improves all the time, the political opinion researchers' association cautions against putting too much stock in online political polling. In 2000, for example, online polls showed that Alan Keyes, Orrin Hatch and Steve Forbes were favored Republican presidential candidates, something no scientific poll ever found. Earlier in 2000, an online poll found that a majority of Americans opposed the government's action to send Elian Gonzalez back to Cuba away from his relatives in Miami. But a random national poll found 57 percent of Americans approved of the federal government's actions.

The AAPOR has a lengthy manual for surveying public opinion. Among the chief tenets is to use random or probability sampling to obtain a composite profile of the population. "In a bona fide survey, the sample is not selected haphazardly or only from persons who volunteer to participate," according to the group's ethics code listed on its Web site. "It is scientifically chosen so that each person in the population will have a measurable chance of selection. This way, the results can be reliably projected from the sample to the larger population with known levels of certainty/precision."

Large sample sizes are recommended, but so are solicitations of non-respondents and the use of well-trained interviewers. AAPOR encourages pollsters to disclose all methods of their surveys so they can be evaluated and replicated.

"If you want to model a likely electorate, it's virtually impossible to do that very well by what's come to be commonly known as online polling," said Robert Daves, president of AAPOR.

Mark Mellman, of the Washington-based Mellman Group, said, "There's a lot of things you can do online. Getting a pinpoint-accurate survey result is not one of them at the moment.... Online polling has generated a lot of controversy and a lot of heat, but so far there is certainly no consensus that online polling is an efficacious method of getting accurate results."

Public Opinion Strategies in Alexandria, Va., is the largest Republican polling firm in the country, with a list of clients that includes 54 members of the U.S. House of Representatives, 13 governors and 13 U.S. senators. For the 2006 midterm election, senior citizens, as they always do, will vote in far greater numbers than young adults, said Rob Autry, vice president of political polling.

The firm embraces online polls for corporate and public affairs work, but not for political campaigns. "Of all the political campaigns we're doing, we'll not do a single one of the polls online," Autry said. "What scares most political pollsters away from doing surveys online is the generational gap. No one uses them for political work, certainly not campaigns. It's a neat idea, but not very practical."

The debate will continue to rage over whether Internet polling is scientific or not. The answer really depends on the methodology. Are samples truly random, or arc survey respondents predominantly people who have volunteered? People who take surveys tend to be more educated, driven and politically active, experts say, and are not representative of the population as a whole.

At Knowledge Networks in Menlo Park, Calif., the firm has developed an online panel of respondents found by dialing random telephone numbers; the sampling is more representative of the country, said Dave Stanton, vice president of marketing and communications. If a respondent does not have a computer for online access, then Knowledge Networks can connect them online through their television sets. That means they are creating Web surfers, not just channeling existing online users.

"Our Internet methodology is just as accurate as the telephone," added Mike Dennis, the company's managing director for government academic research. "I think there's a tremendous variety in the way online polling is done. It's difficult to have an opinion in general on the genre. There is a concern that a small fraction of the online population is taking those surveys, and there could be a bias we're not aware of. There is a significant attempt to overcome the fact that four out of 10 households don't have the Internet and to eliminate self-selection bias"--in which respondents volunteer to take polls and usually are more politically active and educated, unlike a sizeable slice of voters.

At Polimetrix in Palo Alto, Calif., founder and Stanford University professor Douglas Rivers is seeking a patent for a system called sample matching. His database of more than one million names is culled from voter-registration lists and consumer databases. Polimetrix's respondents have provided specific biographical information, such as party identification, ideology and demographics (gender, age, race, religion, education and marital status).

"Our samples look exactly like the registered voter population in terms of the things that we can measure," Rivers said.

"The phone is going to go away," he continued. "The ability to find phone numbers for people is waning, the willingness of people to take polls on the telephone is decreasing and the quality of your samples is declining. The Internet is the obvious replacement. What's left unclear is exactly what Internet methodology will work."

J. Todd Foster is a freelance journalist based in the Shenandoah Valley of Virginia and is also the managing editor of the News Virginian in Waynesboro.