Field Substitution in Surveys

A possible approach to deal with nonresponse in surveys is replacing nonrespondents with other similar units during fieldwork. Field substitution is a general term encompassing many different practices. Both conceptual and operative aspects of substitution can influence the reliability of results, such as methods of selection of substitutes (random/nonrandom), mode of data collection (telephone/face‐to‐face interviews), fieldwork protocols, and controls (loose/tight). If substitution on a convenience basis is allowed or the decision to substitute is left to the interviewer's discretion, the sample representativeness could be undermined and a substitution bias may compound the effects of nonresponse. The risk is lower in telephone surveys, which make contact attempts less demanding and allow easier monitoring of operators' activities. The substitution bias can be evaluated testing the hypothesis that the subsamples of originally selected units and of substitutes belong to the same population. Field substitution can be advantageous, since it maintains the intended sample size and structure, ensuring the planned number of observations for each targeted subpopulation of the sample. It is considered especially suitable for surveys with an extensively stratified design. The choice to resort to field substitution depends on its anticipated usefulness, given the objectives, characteristics, and context of each survey.