© 1991 World Association for Public Opinion Research
PRE-ELECTION POLITICAL POLLING AND THE NON-RESPONSE BIAS ISSUE
Abstract
Non-response bias is of great concern in pre-election political telephone polls, where the number of callbacks is far less than in most academic and government studies. The fewer number of callbacks raises the issue of whether there might be a greater non-response bias problem in such surveys. Given the public importance of political polls, there is a need to know something about this. The general problem faced in trying to assess this issue is that it is very difficult, if not impossible, to obtain information on non-respondents. There is an additional problem with political polling in that there are few public studies about non-response in such polling. This study assesses non-response bias for five pre-election political polls which used four callbacks each. The study is unique in that information is available for all respondents and non-respondents. The initial list of individuals to be called was taken from a computerized voter registration list which contains the age and sex of each registrant. The non-response bias associated with not-at-homes, bad numbers, and refusals is assessed. The results indicate that bad numbers and not-at-homes have different bias effects from refusals, but they offset each other to some degree in a situation of limited callbacks. More callbacks may reduce bias associated with not-at-homes, but not the bias associated with bad numbers and refusals. We usually do not know how biased nonresponse is, but it is seldom a good assumption that nonresponse is unbiased (Flower, 1984, p. 51–2