Friday, October 12, 2012

Neutrality in Political Polls




A Facebook friend who lives in Florida recently posted about a polling call that his wife answered.  The interviewer asked if she planned to vote for Barack Obama in the upcoming election.  When she said no, the interviewer said they hoped she would change her mind and ended the call.

I commented on my friend’s post that this did not sound like a legitimate poll.  It was really a canvassing call.  A real poll would not ask a leading question.  Where I work, we even go so far as to rotate the names of the candidates in a race so that half of the respondents hear the name of the Democratic candidate first, and half of the respondents hear the name of the Republican candidate first.  That is one way that we work to reduce bias in polling.

Respondents often ask our interviewers who is sponsoring a survey.  When we conduct political polls we often do not tell the interviewers the name of the client.  We tell them to tell respondents that interviewers are intentionally not told who is sponsoring a survey, so as not to bias or influence anyone’s answers.  We have done several polls for the Columbus Dispatch.  At the end of the call we ask respondents if a reporter from the Dispatch may call them to discuss their answers.  When we do these polls, we tell interviewers to tell respondents who is sponsoring the survey after they have completed the survey.

When I brief interviewers on a new telephone survey project I like to remind them that they are free to have opinions about candidates and issues, but to keep those opinions to themselves when interviewing respondents.  If an interviewer asks:

“If the election were held today, who would you vote for President of the United States, Barack Obama, the Democrat, or Mitt Romney, the Republican?”

OR

“If the election were held today, who would you vote for President of the United States, Mitt Romney, the Republican, or Barack Obama, the Democrat?”

The interviewer should not emphasize either choice.  If the interviewer asks a respondent if they agree or disagree with a political position, the interviewer needs to read the question in such a way that the respondent cannot tell if the interviewer agrees or disagrees with the position.  We occasionally have respondents ask interviewers how they feel about issues.  They are permitted to discuss their opinions only after the respondent has answered all of the survey questions, and then to keep it brief.  We remind interviewers not to say okay or uh-huh after a respondent answers a question.  These utterances can be interpreted as approval or disapproval of answers to questions.

As far as I know, CBS News, Gallup, Quinnipiac University and Rasmussen train their pollsters in maintaining neutrality.  These organizations have an interest in making sure that the data they collect is accurate.  The larger the phone room, the more difficult it may be to make sure that each interviewer is asking questions in a neutral manner.  Errors in sampling may also skew polling results slightly.  That is why a margin of error is always reported with survey results.  For the most part, however, polling organizations have neither a Democratic nor a Republican agenda.

John C. Stevens
Saperstein Associates
(614) 261-0065

No comments:

Post a Comment