Portfolio allocation: an analysis of stability and non-gaussianity

This thesis is the result of my internships experience at Fondaco S.G.R. The work is the answer to two of the many possible problems that can be found in the modern financial industry. In the first place, the thesis utilizes Principal Component Analysis and Monte Carlo simulations to perform an analysis on the stability of the asset allocation model output. Then looks for the sources of instability by extracting the factors featuring the highest levels of correlation with a measure of distance from values of reference. In a second phase, the thesis aims to solve the problem of mean-variance-skewness optimization in a framework characterized by factor models with scenarios. The solution of this problem is structured in a three steps process. First deriving the optimization criterion as an approximation of the expected utility, secondly proposing an analytical formulation for the skewness-coskewness matrix adapted for factor models with scenarios and lastly evolving the portfolio selection criterion to a theoretical framework in which risk is not entirely defined by volatility.