An Interview with Don Kotecki, Director of the St. Norbert’s College Survey, and Wendy Scattergood, Assistant Professor of Political Science and St. Norbert College Survey Analyst.
What is the survey mode?
The survey was conducted by telephone.
Can you tell me about the sampling technique?
The methodology employed is random digit dial, or RDD.
What was the type of in-household selection, if any?
There is a computer- assisted selection process. Once the phone is answered, we randomly select someone in the household to talk with. We select among the adults who currently reside in the house. Basically, if I call you, I’ll ask how many adults reside in your house, then I may ask how many females and how many males. You will response in turn. Then the matrix is set up to select an interviewee based on age and gender. For example, the computer may ask for the oldest male. If a woman answers, and the oldest male is not there, we will call back to attempt to complete the interview with the selected adult. The name for this process is: the Tordahl-Carter selection method.
Who were your interviewers?
We use a combination for the interviewing staff. Students made up about half the staff; the other half is composed of community workers, for the most part elderly or those looking to augment their income.
If you weighted the data, how did you do so? What parameters did you use? Do you weight by party identification, for example?
In this particular survey, we did not -- and typically do not -- weight the data.
We do not try to predict who’s going to show up at the polls. It’s a philosophical issue, whether we weight the data based on past voting demographics -- or whether we randomly sample and take the results. In the last governor’s election, for example, without weighted data, the St. Norbert-WPR Survey was the closest to predicting the outcome of the election of any national or state poll.
Can you give me a general description of likely voter modeling, if there is any?
We just ask how likely are you to vote in a given election. If we are screening for that, at the beginning of the survey we continue interviewing the respondent if they answer “somewhat likely” and “very likely.” We discontinue the call if the reply is “somewhat unlikely” and “very unlikely.” The question really is: How do you screen for likely voters, and different states have different laws. We don’t use “Have you voted in the last election?” or “Are you a registered voter?” because in Wisconsin one can register at the polling place. We prefer not to pigeonhole those who did not vote in the past.
What is the typical sample size?
We strive to acquire 400 competed cases, which based on an infinite population will give us results with a 95% confidence interval, plus or minus a 5% margin of error.
In terms of sponsorship, who pays for the poll?
It’s a combination of sponsors, St. Norbert College and WPR.
Next, in terms of branding, what are the name of poll and polling partners?
When the call is placed, St Norbert College is mentioned. If people inquire further about sponsorship, we also mention WPR.
Lastly, if your methods vary by communities or cities, can you detail the differences?
There is no variation across the state.
