Building composite indicators for the territorial quality of life assessment in European regions: combining data reduction and alternative weighting techniques

Development of composite indicators is a challenging task given that sustainability indices are strongly dependent on how the sub-indicators are weighted. This is because relative indicator weights may significantly differ based on the chosen weighting methods used in the analysis. There is hardly any study that has paid attention to this issue so far. Therefore, this paper aims to fill this gap in the literature by searching the robustness of selected weighting methods, i.e. entropy-weight (EW), principal component analysis (PCA), machine learning approaches (random forest-RF), regression analysis (RA) and benefit-of-the-doubt (BOD) when constructing a composite indicator. To research the current sustainability performance of European regions, the present study focuses on the Territorial Quality of Life Index—initially proposed by the ESPON Programme—that are aligned with the specific targets of the Sustainable Development Goals of the 2030 Agenda. The methods to construct composite indicators include stages of data preparation (including the estimation of missing values with random forest method), normalization, statistical transformation of raw data, reduction of indicators in order to ease public communication (using the PCA method) and data interpretation, weighting of the sub-indicators using EW, PCA, RF, RA and BOD methods and their linear weighted aggregation, and checking for robustness and sensitivity. The results suggest that there are significant differences in the rank and spatial distribution of composite indicators based on the use of different weighting methods considered in the analysis. The results from sensitivity analysis support the robustness of entropy-weight method among others. The methodology used in the current analysis can be adapted to other study areas and regions internationally. The findings showed that Eastern European countries and some Mediterranean countries have relatively lower index values compared to other European regions; therefore, policy and planning actions are needed covering these regions specifically.