New developments in frontier models for objective assessments

This dissertation is the result of some innovative proposals, in the wide framework of production efficiency frontier models, that have the common goal of reducing subjective choices of the researcher by using, as far as possible, objective methods. In particular, the first proposal links the economic efficiency theory to the spatial econometrics with the aim of taking into account - in the efficiency evaluation of a productive unit - the neighborhood effects in a global way avoiding the subjective selection of a set of variables identifying territorial effects. The method called Spatial Stochastic Frontier Analysis (SSFA) has been published in Fusco and Vidoli (2013) for the production efficiency analysis and generalized in this thesis to be able to also analyze the cost efficiency. The second proposal, instead aims to introduce enhancements in the methods using frontier techniques to aggregate simple indicators in a composite indicator. Subjectivity is avoided in the identification of the set of aggregation weights necessary for constructing the composite indicator, in the definition of a preference structure among simple indicators and in the extreme values and outliers influence removal. The two methods proposed, called respectively Directional Benefit of the Doubt (D-BoD) and Robust Directional Benefit of the Doubt (RD-BoD), have been published in Fusco (2015) and Vidoli, Fusco and Mazziotta (2015). The dissertation consists of four parts: the first one introduces the foundations of the economic efficiency analysis and gives key economic concepts and definitions needed for a proper understanding of the following parts, focusing both on parametric and on nonparametric methods for cross-sectional and panel data and for mono-output and multi-output production processes; the second one discusses the fundamentals of the spatial econometrics, on the main connection proposals with the efficiency theory and shows in detail the SSFA method and the related R package called SSFA implemented to allow other researchers to use it; in the third part the concept of composite indicator and the required steps for its construction are discussed and D-BoD and RD-BoD are shown, moreover the related R package Compind is presented; all proposed methods have been tested both on simulated data and on real data and the results are shown in the fourth part. In the last part, two innovative applications, respectively on the estimation of non performing loans of commercial banks (Fusco and Maggi, 2016) and on the estimation of the local governments’ expenditure needs (Vidoli and Fusco, 2017) by using the efficiency and spatial theories, are also included.