1996 Poll + Methodology

Poll results and Methodology for the 1996 Presidential election.

1996 Poll and Methodology (download pdf)
1996 Poll and Methodology (Scribd)

Results of the Medill News Service/WTTW survey are based on telephone interviews conducted July 8 through July 21 with 3,323 adults, 18 years of age and older, living in the continental United States.

The sample of telephone exchanges called was selected by a computer from a complete list of working exchanges in the country. The exchanges were chosen so as to insure that each region would be represented in proportion to its population. The last four digits in each telephone number were randomly generated by a computer and screened to limit calls to residences. This procedure provided access to both listed and unlisted residential numbers.

The sample for each region of the country was released in replicates to ensure that the established calling procedures were followed for the entire sample. This procedure also helped ensure that the appropriate number of interviews would be obtained in each region.

At least four attempts were made to complete interviews at every sampled telephone number. The calls were placed on different days and at different times of the day to maximize the chances of reaching a respondent. In each contacted household, interviewers first asked to speak with the “youngest male 18 years of age or older who is at home now.” If no eligible male was at home, interviewers asked to speak with the “oldest female 18 years of age or older who is at home.” This systematic respondent selection process has been shown to produce samples that closely mirror the population in terms of age and gender.

Each eligible respondent was asked whether they are registered to vote at their current address, whether they voted in the 1992 presidential election, what kept them from voting in 1992 if they did not do so, and whether they intend to vote in the 1996 presidential elections. Respondents who said they are currently registered, voted in 1992, and would “definitely vote” or “probably vote” this November were included in the pool of likely voters, as were those who said they are registered, didn’t vote in 1992 because they were not old enough, and would “definitely” or “probably” vote in November. Once a likely voter was identified, the interviewer collected information on the respondent’s age, race, education level, and household income before terminating the interview. A total of 2,322 of these short interviews were conducted.

Respondents who said they are not registered to vote at their current address, or who cited some reason other than age for not voting in 1992, or who said they would “probably not vote” or “definitely not vote” this November were classified as likely non-voters. These 1,001 respondents were then asked a battery of sixty-four questions to determine the levels and patterns of their news and information consumption; the extent of their alienation from or affinity for governmental institutions, political parties and politicians; the extent of their participation in other forms of political behavior (such as attending political meetings or contacting their federal, state or local representatives); and their attitudes on selected social and public policy issues. Likely non-voters were also asked eight demographic questions.

To facilitate our exploration of both the similarities and differences among non-voters, a typology was constructed by using cluster analysis, a statistical technique that classified respondents into the most homogeneous and meaningful groups possible based on their news and information consumption, the extent of their alienation from government and the political process, their feelings of political self-efficacy, the extent of their participation in other forms of quasi-political behavior, and basic demographic characteristics, including length of time at their current address.

The results of the survey have been weighted to adjust for variations in the sample relating to race, gender, age, and education. For results based on the sample of 1,001 likely non-voters, one can say with 95 percent confidence that the error attributable to sampling and other random effects is plus or minus 3 percentage points. However, for results based on interviews with subgroups of respondents, the margin of error is larger. For example, the responses of the 288 non-voters classified as “Doers” have a margin of error of plus or minus 6 percentage points.
In addition to sampling error, it should be noted that question wording, question order and the practical difficulties of conducting any survey of public opinion can introduce error or bias into the findings.

Comments are closed.