Data that is missing due to non-response impose a serious threat to the quality of statistics that are based on both surveys and registers. In most cases non-response relates to demographic and socio-economic characteristics of the selected persons or enterprises and importantly also on the data collection process. In recent years a growing focus can be observed in survey research on differentiated data collection protocols and adaptive designs.
The theory of Deming (1986) about improving quality and productivity in industry is well-known. Many of his famous 14 points for management also apply to the production of statistical information. Quality must be built in at the design stage. Deming’s points particularly apply to the data collection process.
The response rate is often used as an indicator of survey quality. However, literature gives various examples where increased data collection efforts led to a higher response rate but also to a larger or comparable non-response bias.
Therefore, to assess the effects of non-response, other quality indicators are needed.
These indicators should measure the degree to which the respondents of a survey or register still resembles the population. First examples of such representativity indicators have emerged recently. The main objectives of this project are to elaborate and develop these representativity indicators, to explore their characteristics and to show how to implement and use them in a practical data collection environment. It will be demonstrated that representativity indicators can be used in several stages of the data collection process to improve the quality of the resulting statistics.
The project facilitates the efficient allocation of data collection resources and a sophisticated trade-off between quality and costs. Furthermore, in order to enable a sensible incorporation of register data in the production of statistics, quality indicators are important tools in the reduction of respondent burden