Measuring and Monitoring Balance of Response Set

Today, non-response is a common issue in each sample survey. The resulting re-sponse set is usually out of balance with respect to the full sample, and causes biased estimators. Even special adjustment methods cannot remove the entire bias. In Särndal (2011) new indicators are proposed to measure balance of the response set. In this paper we use these balance indicators to monitor sample selection pro-cess. We argue that response rate alone is not enough to express quality of the sample. Moreover, increasing response rate by the efforts to get data from arbi-trary non-respondents may be even harmful to the balance of the response set. We describe methodology to identify those non-respondents who would bring highest increase towards balance. This knowledge could be used in the data collection pro-cess; directing more efforts for incorporating the most influential non-respondents. An example with three strategies is given