Peter Lundquist: Balanced response set

Large nonresponse is typical of many sample surveys today. This can be a serious detriment to survey quality: Nonresponse causes systematic error in the survey estimates. In recent literature on survey nonresponse, new indicators of the quality of the data collection have therefore been proposed. These include indicators of balance and representativity (of the set of respondents) and distance (between respondents and nonrespondents), computed on available auxiliary variables.

At the seminar I will give a theoretical overview and demonstrate how to use the developed indicators in the statistical production. The indicators could be used during the field work to intervene and guide the data collector with a better balance in the response set as a goal. Other areas of use are for example: To compare response between different surveys, to do a longitudinally comparison for a survey and to evaluate the effect of different data collection modes in a survey. In the presentation I will use examples from Statistics Sweden.

My presentation is based on a joint work with Carl-Erik Särndal.

Reference: Lundquist, P. and Särndal, C.E. (2013). Aspects of responsive design - With applications to the Swedish Living Conditions Survey, Journal of Official Statistics, 29, 557-582.