Handling nonresponse in business surveys

Abstract

"Business surveys are a valuable indication of the current and the future economic situation. However refusals are very common in this context and may induce bias in the estimates of interest. In this paper the problem of adjusting for nonresponse estimators of relevant economic outcomes is considered. Using a large simulation study, the authors investigate how weighting adjusted procedures are influenced by the specification of the response model utilized in calculation of the weights. They consider, in particular, how trimming the weights can have an impact on the final estimates and they propose a bootstrap-based procedure to determine an optimal trimming threshold. The authors illustrate the procedure using a 2009 survey of a large population of Italian firms." (author's abstract)