Southern Focus Poll, Non-South Survey, 1999
SummarySoutherners tend to slip through the cracks between state surveys, which are unreliable for generalizing to the region, on the one hand, and national sample surveys, which usually contain too few Southerners to allow detailed examination, on the other. Moreover, few surveys routinely include questions specifically about the South.
To remedy this situation, the Odum Institute for Research in Social Science conducts a Southern regional survey, called the Southern Focus Poll. Respondents in both the South and non-South are asked questions about economic conditions in their communities, cultural issues (such as Southern accent and the Confederate flag), race relations, religious involvement, and characteristics of Southerners and Northerners.
All of the data sets from the Southern Focus Polls archived here are generously made available by the Odum Institute for Research in Social Science of the University of North Carolina at Chapel Hill (OIRSS).
The ARDA has added six additional variables to the original data set to enhance the users' experience on our site.
Data FileCases: 405
Weight Variable: WTVARN and TOTWT
The weighting variable WTVARN is for weighting the data by household size. The variable PEOPLE for each case is divided by the mean for the variable PEOPLE in the sample. The variable TOTWT was created to allow the combination of the Southern and non-Southern data sets into a representative national sample. NOTE: TOTWT is only designed for combining the general South and non-South samples, and NOT the oversamples. Use this variable when combining these two data sets.
Data CollectionFebruary 1 - April 12, 1999
Funded ByOdum Institute for Research in Social Science of the University of North Carolina at Chapel Hill
Collection ProceduresFrom five to ten student interviewers made calls each night of the survey. Each of the regular interviewers and alternates went through a training session before they began conducting telephone interviews. The training session familiarized interviewers with the purpose of the survey, its sponsor, the target population, and the specific interviewing procedures to be used for the survey.
The training also stressed familiarity and understanding of the opening script, call sheet, and questionnaire. The opening script is the model for establishing that a residential phone has been reached, explaining the purpose of the call, identifying the eligible respondent, obtaining consent and getting the eligible respondent to complete the interview. General and specific techniques to persuade reluctant respondents to cooperate were reviewed.
Once the survey began, a field manager was on duty at all times to monitor quality and handle any problems that arose. The manager compiled daily reports on each interviewer's number of completed interviews, refusals, efficiency, etc. and provided feedback as necessary.
At least eight call attempts were made to each telephone number, at least 24 hours apart. If no respondent was reached during the initial call, subsequent calls were scheduled at different times of the evening, within the specified calling times. Once callbacks were scheduled, eight attempts were made to complete the interview. Unless a day-time callback was scheduled by the eligible respondent, all interviews were done from 6:00 to 10:00 p.m. in the respondent's time zone, Sunday through Thursday, from February 1 to April 12, 1999. No calls were made March 5-15.
Refusals were coded harsh, firm, or mild, and attempts were made to convert mild and firm refusals into completed interviews.
Sampling ProceduresThe target population for the telephone survey was adults age 18 or older, residing in households with telephones in the United States. We oversampled the Southern states of Alabama, Arkansas, Florida, Georgia, Kentucky, Louisiana, Mississippi, North Carolina, Oklahoma, South Carolina, Tennessee, Texas, and Virginia. We also made two oversamples of African-Americans, one in the Southern states and the other in non-Southern states. During the period from February 1 through April 12, 1999, 1369 interviews were completed. Of these, 829 were general Southern, 405 general non-Southern, 26 Southern African-American oversample, and 109 non-Southern African-American oversample. In addition, when we reached non-African American respondents in either African-American oversample, we took only demographic information from the respondents. 36 non-African- American respondents in the Southern African-Amercian oversample provided us with demographic information. 70 non-African-American respondents in the non-Southern African-Amercian oversample provided us with demographic information.
IRSS purchased two random-digit dialing (RDD) samples from Genesys Sampling Systems, (GSS) of Fort Washington, Pennsylvania: a sample of Southern numbers, and a sample of non-Southern numbers. To give each residential telephone within each sample an approximately equal chance of being dialed, GSS first systematically stratified each sample to reflect each state's proportion of the appropriate region's population. For each state, GSS then estimated the
proportion of telephone numbers beginning with each three-digit prefix in use in that state. This estimate was based on the proportion of numbers with each prefix listed in the telephone directories for the state. Strings of four random digits, confined to working banks where known, were generated by computer to complete the phone numbers. Using this method, GSS produced sets of Southern and non-Southern numbers. The numbers were screened for digitally indicated non-working numbers, which were then eliminated.
IRSS also purchased two RDD oversamples of African-Americans, one Southern, the other non-Southern. To generate the targeted RDD oversamples, Genesys converts zip-code coded data to exchange-level data that assigns a density code to each telephone household reflecting the proportion of black households for the area code exchange in which the household resides. For this Southern Focus Poll, we chose to target exchanges with at least a 35 percent density
level. About 46 percent of all black households are covered by these exchanges, and Genesys estimates the probability of reaching a Black household using an RDD sample from the cumulative subset of exchanges to be about 51 percent.
IRSS staff printed the numbers onto calling sheets, which were divided into randomly selected sets of approximately 50 numbers (general Southern sample), 27 numbers (general non-Southern sample), 3 numbers (African-American oversample, Southern), and 13 numbers (African-American oversample, non-Southern). As calling proceeded, 56 general Southern, 50 general non-Southern, 65 Southern African-American oversample, and 53 non-Southern African-American oversample sets were used.
A sample number was retired when any of the following conditions were met: the number was determined to be non-working; the interview was refused or terminated; the number was of a business or government office; there was no eligible respondent at the number; or a callback could not be completed. After eight calling attempts resulting in any combination of no answer, a busy signal and/or answering machine the number was retired. Computer or fax line numbers were retired after one attempt.
The target respondent within each household was the person living in the household, age 18 or over, with the next birthday. The next-birthday method is a simple procedure used in order to reach a random respondent within each household. Also, in the two African-American oversamples, the survey was completed only by those who identified themselves as African- American or any of its equivalents in the demographic questions at the beginning of the poll.
Those who did not identify themselves as African-American were asked only demographic questions, and then skipped to end of the poll. This demographic information is separated out from the data sets for the oversamples. It was collected in case researchers are interested in the characteristics of non-African-Americans living in primarily African-American locales.