The Relation Between Unit Nonresponse and Item Nonresponse: A Response Continuum Perspective

Abstract

Nonresponse is a significant problem for survey research. The phenomena encompass nonresponse at both the unit and the item level. Unit nonresponse refers to the complete absence of an interview from a sampled household whereas item nonresponse refers to the absence of answers to specific questions in the interview after the sampled household agrees to participate in the survey. Traditionally, survey research has treated unit and item nonresponse as two separate problems with different impacts on data quality, different statistical treatments and adjustments, and different underlying causes (Beatty & Herrmann, 2002; Groves, 1989; Groves, Cialdini & Couper, 1992; Groves, Presser, & Dipko, 2004; Gorves, Singer, & Corning, 2000). Unit nonresponse is usually considered to pose a much greater threat to survey research than item nonresponse. Quantitatively, unit nonresponse is usually much larger than item nonresponse and there are a wide variety of statistical techniques that have been developed to address unit nonresponse. Item nonresponse poses additional threat to data quality because it reduces sample size if only completed cases are used in an analysis. Imputation techniques can be employed to impute for missing data to avoid the shrinkage of sample size. However, it is still problematic for surveys when item nonresponse produces nonignorable missing data. Nonignorable missing data happen when the missing data pattern is correlated with the values of the variable of interest (Little & Rubin, 1987). Of course, unit nonresponse can also create nonignorable nonresponse if the missing pattern in sample persons is related to the values of the variable of interest. The survey literature has recorded a persistent and well-documented rise over the past few decades in unit nonresponse rates for household surveys (Atrostic, Bates, Burt, & Silberstein, 2001; Curtin, Presser, & Singer, 2005; de Heer, 1999; de Leeuw