The variance of some common estimators and its components under nonresponse

In most surveys, the risk of nonresponse is a factor taken into account at the planning stage. Commonly, resources are set aside for a follow-up procedure which aims at reducing the nonresponse rate. However, we should pay attention to the effect of nonresponse, rather than the nonresponse rate itself. When considering nonresponse error, i.e. bias and variance, it is not obvious that the resources spent on nonresponse rate reduction efforts are time and money well spent. In this paper we address this issue, continuing the work begun in Tångdahl (2004), now focusing on the effect of follow-ups on estimator variance. The components of the variance for some common estimators are derived under a setup that allows us to take into account the data collection process, and follow-up efforts in particular.