A new framework for assets selection based on dimensions reduction techniques

Abstract

We introduce a model called Asset Drivers Framework (ADF), which combines Dimensions Reduction Techniques (DRT) with a ranking procedure to find out assets to be inserted into a financial portfolio. The basic idea is that market securities can be described by a wider number of determinants, but only a few number of them can effectively characterize the assets to form well-balanced portfolios. The ADF manages this as a dimensions reduction problem, and extrapolates for each asset a reduced number of determinants as natural drivers of theirs. The procedure ends by assigning a score to the assets projected in such dimensionally reduced space, with a method of punishment/reward of the way the securities cluster into it. The beauty of the ADF scheme relies on a number of points: (i) it provides a platform to test various dimensions reduction techniques; (ii) looking at the performance, ADF makes possible to build portfolios whose returns are aligned to those of the traditional approach, but with lower variance, and hence lower risk. © 2011 Springer-Verlag.