Work detail

Portfolio selection based on hierarchical structure of its components

Author: Bc. Robert Ševinský
Year: 2015 - summer
Leaders: prof. PhDr. Ladislav Krištoufek Ph.D.
Work type: Bachelors
Language: English
Pages: 62
Awards and prizes:
Abstract: This thesis investigate empirical performance of three portfolio selection and
covariance matrix models. The goal is to find a strategy that outperform
equally weighted portfolio in the long run and survives even in times of financial
distress. Two models based on Markowitz approach absolutely failed in
this context, however the last approach based on network analysis indeed outperform
the market even after risk adjustment of returns. Moreover this model
have sparse transaction matrix throughout time, therefore exhibit excellent
properties even in the presence of transaction costs. Results for network based
portfolio were obtained from running a back test on 160 member companies of
S&P 500 index for 6’000 trading days.




Patria Finance