JOAN BARRON

CHEYENNE - Erratic poll results that popped up this year in the presidential election have been blamed in part by the difficulty in surveying people who use only cell phones.

The cell phone-only users are likely to be young people under 30. This group also is likely to heavily favor Democratic presidential candidate Barack Obama. Democratic Party people speculate that the problem in reaching this young crowd is why polls show Obama with only a single-digit lead over Republican John McCain.

A Pew Research Center report said three major surveys conducted since the presidential primary elections, using both cell and land line phones, show only a slight difference. The cell phone users were slightly more in favor of Obama for a difference of 2 to 3 points in the margin, the report said.

The polling industry has adopted some strategies to adjust for the shift in new age communication. "It is a major interest of the entire polling industry," said Burke Grandjean, director of Wyoming Survey and Analysis Center. "For the past several years it has been a central topic of discussion at the annual meeting."

The center is a unit at the University of Wyoming that provides independent and objective data for the state and the region. The center does no polling for political candidates.

Every two years the center does a statewide election survey for the political science department at UW. The department faculty decides on the set of questions, including preference for candidates for statewide offices, and on general policy issues.

In a week or so Wyoming residents will begin getting phone calls from the center's pollsters. The poll results, however, will not be released until after the November general election.

Major national pollsters no longer rely on random calls to land line households and, instead, incorporate cell phones into the sampling, Grandjean said. The main strategy used to count the cell phone-only users only is sampling from exchanges that are known to be cell phone exchanges, Grandjean said.

"It's difficult because an area code doesn't tie a cell phone to a particular geographic area," Grandjean said.

Although the mathematical work is progressing to incorporate cell phone-only users into the sampling, the results are not yet perfect.

Another tactic is to compare the demographic characteristics of the combined land line and cell phone samples with information from the census bureau's annual house-to-house surveys. This work gives pollsters pretty good data on what the distribution of age, gender and income ought to be in a geographic area. If the sample departs from the census data, the response can be differentially weighted.

For example, if there is a shortage of young men in the combined sample, the responses of those who responded get a little higher weight in tabulating the results, Grandjean said.

"This is not a perfect solution either," he said.

Many other adjustments are made as well.

"Some are soundly based in statistical theory, others of them are 'seat of the pants - it worked in the past and we'll try it again,'" he said.

Another difficulty is the center's introduction to potential voters. After announcing the university connection, the pollster's first question is whether the respondent is on a cell phone. If the person is driving, the caller must get off the phone immediately to avoid putting anyone at risk.

"To start every interview with that question puts some people off," Grandjean said. "They think we are telemarketers rather than legitimate pollsters.

"It has made our life difficult at a very practical level in addition to the more arcane statistical problems."

Contact Joan Barron at joan.barron@trib.com or by phone at 307-632-1244.