Who conducted the poll: The poll was conducted for The Atlanta Journal-Constitution by the School of Public and International Affairs Survey Research Center at the University of Georgia. Students made the calls, under the direction of M.V. (Trey) Hood III, the director of the center and a professor of political science.
Who selected the questions? The AJC created the survey, which included several questions we have asked in previous polls.
Whom did we talk to? The center contacted 940 registered voters from across the state. The numbers were randomly drawn from a voter registration list obtained through the sampling vendor Revily. The company maintains a database constructed from the state voter registration lists. Through commercial sources, phone numbers have been appended to the individual records (registrants) that make up these lists. Sixty-five percent of the phone numbers in the sample were cellphone numbers; 35 percent landline numbers.
What is weighting and how do you do it?
Some adjustments are made to the survey results to accurately reflect the demographics of the population of registered voters in the state. For this poll, adjustments were made to ensure the poll results looked like the population of registered voters in Georgia in terms of race, age and sex. For example, if the survey sample was found to contain 51 percent female registrants and the registrant population in the state consisted of 53 percent females, we would give female respondents more “weight.” (In this example the converse would be the case for male respondents. These respondents would be given less “weight” in order to ensure they are represented in proportion to their share in the overall population.)
What is the “margin of error” for the poll and what exactly does that mean?
No matter how carefully a poll is conducted, there will always be some measure of uncertainty when you survey a small portion of a larger population, such as the state of Georgia. The margin of error is the measure of the uncertainty in the sample. The margin of error that we report accounts for these sources of uncertainty. For example, with a margin of error of 4 percentage points, a candidate polling at 50 percent could have support of anywhere between 46 percent and 54 percent, with a 95 percent level of confidence. That means that if we drew 100 different samples using the same methodology, then no more than five times out of 100 should chance variations in the sample cause the results to vary by more than 4 percentage points from the answers that would be obtained if all Georgians were polled.