I Hear You Knocking But You Can't Come In

This article examines the effects of refusers and reluctant respondents to a general population survey on estimates of population distributions and multivariate relationships. Our data indicate that refusers do not constitute a simple random sample of respondents, and that efforts to correct for nonresponse by positing a “continuum of resistance” with refusers assumed to resemble reluctant respondents is not justified. Furthermore, we find that refusers and reluctant respondents can have serious effects on estimates of the relationship among variables. We do this by using analysis of covariance to assess the effect of callbacks on sample estimates and an imputation technique that can be used routinely in personal interview surveys to detect the magnitude and significance of refusal bias on final sample estimates.