(Demography, Vol. 30, no. 2, 1993) Riley, Ann P.; Hermalin, Albert I.; Rosero Bixby, Luis
High levels of nonresponse or inappropriate response to items are a persistent concern in survey research because those who do not answer may not be representative of the study population. Thus nonresponse introduces potential bias in the point estimates as well as in multivariate analyses, which use the responses in question as either an independent or a dependent variable. Researchers often have little recourse but to form a "don't know” or "not available" category from the failure to respond (or to be responsive). They must either omit these cases, treat them as a separate category, or impute a value on the basis of other characteristics of the respondent (Croft 1991; Kalton and Kasprzyk 1986).
