Three Nevada professors compare accuracy of polls

Friday, Oct. 22, 2004 | 8:13 a.m.

WEEKEND EDITION

October 23 - 24, 2004

Three university professors say that the statewide Las Vegas Sun/Channel 8 Eyewitness News/KNPR Nevada Public Radio poll that shows President Bush leading Democratic challenger John Kerry by 4 percentage points is more accurate than another Nevada poll released last week.

In a poll released Thursday by the Las Vegas Review-Journal, Washington-based Mason-Dixon Polling & Research Inc. had Bush leading Kerry 52-42 with Ralph Nader getting 1 percent and 4 percent undecided. Both polls had margins of error of 4 percentage points.

"Your poll is more accurate," UNR political science professor Eric Herzik said of the Sun/Channel 8/KNPR poll. "Mason-Dixon has done a very credible job over the years but every once in awhile a sample can be a bit exaggerated and that's what I think they did here. They may have oversampled the rurals a bit."

The Mason-Dixon poll of 625 likely voters included 61.6 percent from Clark County, 20.8 percent from Washoe County and 17.6 percent from rural counties. The Sun/Channel 8/KNPR poll by Washington-based Belden Russonello & Stewart queried 600 very likely voters, including 65.2 percent from Clark County, 18.6 percent from Washoe County and 16.2 percent from rural counties.

UNLV history department chairman Hal Rothman said the Sun/Channel 8/KNPR poll is "a much more realistic look than the Mason-Dixon poll because it reflects what people are saying and thinking." Rothman said Mason-Dixon polls done elsewhere have tended to favor Bush more so than other polls in those states.

"The fact that Mason-Dixon has a 10-point margin puts Nevada at odds with the rest of the nation," Rothman said.

Based on conversations with campaign insiders, UNLV political science professor David Damore also said he thought the Sun/Channel 8/KNPR poll was more accurate. Damore said he believes the race between Bush and Kerry in Nevada is still "within the margin of error."

"The R-J numbers for Bush are a little high," Damore said.

The reason that two polls can reach different conclusions within a short period of time is that polling organizations often rely on different methodologies, he said.

"Different organizations use different methods to root out voters from non-voters," Damore said.

"Polling organizations also want to keep their methodology a secret. Their reputation is based on their accuracy so they want some of that proprietary information kept secret."

Larry Hugick, a researcher at Princeton Survey Research, an independent polling firm in Princeton, N.J., said it is not unusual for two polls released in the same state within a short time of each other to have different results.

In the case of the presidential race, Hugick said many polls that were taken immediately after the final debate on Oct. 13 gave Bush a bump. The Review-Journal poll was taken Oct. 14 through Oct. 16, immediately after the debate. The Sun/Channel 8/ KNPR poll began Oct. 16.

Other factors that can make poll results differ include the order in which questions are asked and the definitions of likely voters, Hugick said.

"What determines a likely voter differs from poll to poll because some polls put more weight on whether they have voted in comparable elections in the past," Hugick said.