Composite indicator as a distance

Abstract

There is increasing interest in the construction of composite indicators to evaluate socio economic concepts. In general, the mathematical approaches on which the most commonly used techniques are based do not allow for reliable benchmarking. Moreover, the determination of the weighting scheme in composite indicators remains one of the most problematic issues. In this thesis, different methodologies are analyzed to extract their strengths and weaknesses. From this analysis it emerges that few of these tools allow comparison between observations. Using the vector space formed by all observations, a new method to construct composite indicators is proposed:a distance or metric that considers the concept of proximity between units. To do so, we take Pena Trapero’s P2 Distance method as a starting point. The proposed methodology eliminates the linear dependence of the model and looks for functional relationships that allow us to build themost efficient model. This approach reduces the subjectivity of the researcher by assigning the weighting scheme with unsupervised machine learning techniques. Monte Carlo simulations confirm that the proposed methodology is robust. As an application of this new approach, a vulnerability index is constructed for the European Union (EU) regions. The cohesion policy is analyzed for the period 2021-2027 focuses on five objectives to make the EU smarter, greener, more connected, more social and closer to citizens. A macroeconomic index is proposed as the predominant criterion for allocating Structural Funds among regions. It is hypothesized that it is possible to take into account new complementary criteria that better reflect the quality of life of citizens. To this end, a composite index of socio-economic vulnerability is constructed for the 233 regions studied. The results show that, following our multidimensional approach for the allocation of Structural Funds, there are notable differences in the maps of priority regions. Moreover, the COVID-19 pandemic represents a welfare threat. Multilevel models are estimated from which it follows that country characteristics interact with those of regions to alter vulnerability patterns. More specifically, increased public spending on education and improved political stability would reduce regional vulnerability or build resilience, while increased poverty would be associated with increased vulnerability. Likewise, the most vulnerable regions would be the most exposed to the negative socio economic effects of COVID-19.