The paper investigates the problem of the definition and the construction of composite indicators. The indicators are increasingly a valuable tool to assist people for the definition of appropriate policy that is based on effective analysis of real-world.
Methods and Results
Principal component analysis is often used to define composite indicators. Unfortunately, when dealing with spatial units, this technique is not appropriate, since it does not consider the spatial effects, namely spatial heterogeneity and dependence that are inherent characteristics in spatial data. To overcome this problem, the authors use a modified version of principal component analysis that has recently been introduced in literature and that explicitly considers the spatial heterogeneity effect. This method is denoted as geographically weighted principal component analysis. The method is applied for the definition of well-being composite indicators at local level for 110 Italian provinces. The analysis is performed for 2011.
The empirical evidence shows that the multidimensional concept of well-being is differentiated at local level, supporting the extent of spatial heterogeneity. Therefore, the obtained results support the use of the geographically weighted principal component analysis as more suitable method for handling spatial data.
Keywords: Geographically weighted regression, composite indicators, well-being indicators, spatial econometrics, kernel function
Jel Code: C01, C21, C43, C54.