Report on the impact of response and sampling on composite and multidimensional indicators. Explorative studies on the impact of data quality on composite indicators (Deliverable 13.63)

Composite indicators (CIs) are frequently employed to measure complex, multidimensional phenomena such as well-being, sustainability, or work quality. It is nowadays well studied, that the multiple decisions made in the construction process of building the CI strongly affect the result. These construction decisions range from the selection of a set of sub-indicators, to the standardization method, the weighting and the choice of the aggregation function. Sensitivity analysis is an established tool to analyze and quantify the impact of the choices made in this regard. Less attention has been paid to the impact of data quality on CIs: CIs are usally based on sub-indicators that are estimated from sample surveys and the resulting aggregated measure can only be as good as the underlying data. The uncertainty in the CI due to sampling and possible non-sampling errors is, however, frequently neglected. With this report we aim to fill this gap. To do so, we perform a sensitivity analysis for an example CI for working quality, that includes the selection of a sampling design and the sampling itself as possible sources of variability. Further, we propose a parametric bootstrap-based apporach to estimate the standard error of a CI and apply it to the illustrating example of a CI on environmental perfomance.