Statistical Processes Under Change: Enhancing Data Quality with Pretests

Abstract

Statistical offices in Europe, in particular the Federal Statistical Office in Germany, are meeting users’ ever more demanding requirements with innovative and appropriate responses, such as the multiple sources mixed-mode design model. This combines various objectives: reducing survey costs and the burden on interviewees, and maximising data quality. The same improvements are also being sought by way of the systematic use of pretests to optimise survey documents. This paper provides a first impression of the many procedures available. An ideal pretest combines both quantitative and qualitative test methods. Quantitative test procedures can be used to determine how often particular input errors arise. The questionnaire is tested in the field in the corresponding survey mode. Qualitative test procedures can find the reasons for input errors. Potential interviewees are included in the questionnaire tests, and their feedback on the survey documentation is systematically analysed and used to upgrade the questionnaire. This was illustrated in our paper by an example from business statistics (“Umstellung auf die Wirtschaftszweigklassifikation 2008” – Change-over to the 2008 economic sector classification). This pretest not only gave important clues about how to improve the contents, but also helped to realistically estimate the organisational cost of the main survey.