An Examination of the Random versus Nonrandom Nature of Item Omissions

Item omission in mail surveys has received little attention other than as a variable measuring response quality in questionnaire design studies. Few researchers have attempted to examine item omission as a nonrandom process. The authors re-evaluate the nature of item nonresponse and analyze item omissions within topical domains. Two independent surveys and a replication study are reported. The tendency to omit survey items is tested for generalizability across domains and between similar studies. Sociodemographic correlates are investigated and a longitudinal analysis by item is performed. The results indicate that item omission is apparently a nonrandom process at an aggregate level, but the magnitude of the relationship is weak. When the data are disaggregated by topical domain, significant systematic relationships become less apparent.