COLUMBIA, S.C. - The latest Winthrop/ETV Poll survey found that among likely voters in North Carolina and Virginia, there is a statistical dead heat between presidential candidates Barack Obama and John McCain. In South Carolina, McCain is expected to carry the Palmetto state by 20 points.
Scott Huffmon, associate professor of political science and poll director, said callers talked to 2,026 likely voters between Sept. 28 and Oct. 19. The Social & Behavioral Lab called only on weekday evenings and during the day and evenings on weekends and talked to registered voters in the three states who are likely to vote.
Those surveyed include 617 respondents from South Carolina (margin of error of +/- 3.94% at the 95% confidence level), 744 respondents from North Carolina (margin of error of +/- 3.59% at the 95% confidence level), and 665 respondents from Virginia (margin of error of +/- 3.8% at the 95% confidence level). For survey data, go to: http://www.winthrop.edu/winthroppoll
Huffmon found that by an overwhelming number in Virginia and North Carolina say that Obama, the Democratic nominee Senator from Illinois, understands their lives. Other highlights of the poll were:
• Almost 40 percent of white women in Virginia were less likely to vote for McCain, the Republican presidential nominee from Arizona, because of his selection of Alaska Gov. Sarah Palin as his vice president.
• More than 50 percent of voters give a thumbs-down to public funds saving private companies.
• Six in 10 likely voters say history will judge President George Bush’s presidency a failure.
Questions also covered the war in Iraq, health care, personal finances, gas prices, illegal immigration and taxes.
The results break out two groups of voters, white women and working class white voters, both of whom are thought to be pivotal in the race. For the purpose of these results, "working class" is defined as a respondent who reported $50,000 or less total annual household income.
“The white working class is economically vulnerable, however they are strong McCain supporters and they believe McCain on virtually every issue,” Huffmon said. “White women are a tipping point constituency – they represent a segment that could be pulled in either direction. They could change from a Bush security moms back to Clinton soccer moms based on the economy.”
The results were released Oct. 23 on South Carolina ETV’s “The Big Picture” program and in the Oct. 26 edition of The State newspaper.
Questions Regarding the Length of Time in Field
While this is a long time for a political survey to be in the field, it does two things: 1) Increases the “representativeness” of the sample without weighting and 2) Smoothes the erratic jumps and dips seen when overnight polls are aggregated. First, most polls done in a few days call during the daytime and must use a complicated weighting scheme for the data. For example, retirees and stay-at-home moms are more likely to be reached during the day and this over-sampling must be accounted for in the final results.
The Social & Behavioral Research Lab at Winthrop University calls only on weekday evenings and during the day and evenings on weekends. This allows respondents to be randomly sampled in their proper proportions without additional weighting. While the trade off between speed of data gathering and un-weighted accuracy of the sample is unacceptable to some, un-weighted sample quality holds a supreme position for us since this data will be used for academic research after the conclusion of the election. Second, a longer time in the field has a "smoothing effect" on the data.
Trends are more fluid compared to the choppy ups and downs seen in many quick snapshot polls. However, spending too long in the field can mask significant shifts. The consumer of this data should use these results to paint a fuller picture of trends seen from the multitude of quick snapshot polls; she or he should not assume these results supplant any other methodologically sound data.
Questions Regarding the definition of "Likely Voter"
Respondents were randomly selected from lists of registered voters in each state. The interviewer asked for the selected registered voter by name and then verbally confirmed their name and current registration status. After determining that we had the correct individual and that they were, in fact, still registered in the state where they lived (and were called), we screened the respondents using an additional question to determine likelihood of voting.
Only those currently registered voters who responded that they "definitely plan to vote in the November presidential election" are included in these results. Some polls use Random Digit Dialing and then apply a series of questions to determine likelihood of voting. These questions may include whether the individual is registered (we already know our respondents are registered because they are pulled from lists of registered voters), interest in the outcome of the election, how closely they have followed the election, how often they have voted in the past, and others.
Huffmon said he is concerned that any method that screens for, or more heavily weights, the responses of those who have previously voted will under-represent newly registered voters. While in most elections, newly registered voters make up so few of the electorate as to make this concern unwarranted, we believe that newly registered voters may play a pivotal role in the presidential election of 2008 and we chose a methodology that would ensure their inclusion.