The Inclusion of Liquidity Risk Into a Parametric Value-at-Risk Framework

V magistrskem delu vpeljujemo likvidnost kot dejavnik tveganja v standardno analizo tvegane vrednosti. Osnovne parametrične VaR modele nadgradimo z informacijami o cenovnem razponu med ponujeno in povpraševano ceno določene naložbe. Pri tem nestanovitnost donosov izračunavamo na dva načina, in sicer z uporabo ne–tehtane metodologije in z uporabo GARCH metodologije. Rezultate nadgrajenih modelov testiramo na podlagi štirih naborov delnic: treh tujih in slovenskega. Ugotavljamo, da likvidnostni VaR modeli, ob upoštevanju predpostavk raziskave, primerno ocenjujejo tržna tveganja in da se v primerjavi s klasičnimi VaR modeli izkažejo za bolj ustrezne. Do takšnih ugotovitev pridemo pri uporabi obeh metodologij za izračun nestanovitnosti donosov. Ob primerjavi rezultatov modelov po obeh metodologijah med sabo pa se kot primernejši izkažejo likvidnostni VaR modeli, izračunani na podlagi GARCH metodologije. Metodološko na eni strani Likvidnostni VaR modeli predstavljajo napredek v okviru obravnave tržnih tveganj, vendar na drugi strani rezultati testiranj modelov kažejo, da slednji še niso dovolj robustni, da bi zadovoljili vsem statističnim testom. Razlog za to je v večji podatkovni zahtevnosti likvidnostnih VaR modelov. Glede primerjave rezultatov po naborih delnic pa ugotavljamo, da so rezultati za slovenski nabor, kljub manjši globini trga, primerljivi s tistimi iz tujine.In this masters thesis we implement liquidity as a risk factor in the standard value-at-risk framework. We incorporate bid-ask spread information into the basic VaR models. In the models we calculate the volatility of returns in two ways: with the use of non–weighted methodology and with the use of GARCH methodology. We then test these models on four different markets: on three foreign and on a domestic one. We conclude that based on assumptions of our research, liquidity VaR models adequately measure market risk and even prove themselves superior to ordinary VaR models. We conclude this for all the models, based on both types of volatility methodologies. In comparison between the models based on different volatility methodologies we conclude that those using the GARCH methodology prove themselves superior to those using non–weighted methodology. On one hand the liquidity VaR models represent a step in the right direction in market risk analysis but on the other hand those models are not yet robust enough to pass all backtests. Reason for this is that liquidity VaR models have a higher demand for input data quality and quantity wise. About the comparison of results between markets we conclude that the results for the domestic market are comparable to those of foreign ones despite the domestic one being smaller