Vol. 10, No 1, 2013 pp. 25 - 37
UDC 336.76
MEASURING VALUE AT RISK ON EMERGING STOCK MARKETS: EMPIRICAL EVIDENCE FROM SERBIAN STOCK EXCHANGE
Mirjana Miletić1, Siniša Miletić2
1National Bank of Serbia, Belgrade, Serbia,
2College for Business Economics and Entrepreneurship, Belgrade, Serbia
webadmin@nbs.rs
This paper evaluates the performance of a variety of symmetric and asymmetric GARCH type models based on normal and Student t distribution in estimating and forecasting market risk in Serbian stock market. Using the daily returns of the Serbian stock index BELEX 15 we tested the relative performance of a variety of symmetric and asymmetric GARCH type models based on normal and Student t distribution for period of October 2005 to October 2012, a sufficiently long period which includes tranquil as well as crisis years. For investors, in the current global financial crisis, it is particularly important to accurately measure and allocate risk and efficiently manage their portfolio. The impact of extreme events on changes in the financial markets in emerging countries is even more pronounced, as such markets are characterized by lower levels of liquidity and significantly smaller market capitalization. The possibility of application of VaR methodology, which is basically designed and developed for liquid and developed markets, should be tested on the emerging markets, which are characterized by volatility, illiquidity and shallowness of the market. This motivates us to implement methods that involve time varying volatility and heavy tails of the empirical distribution of returns. We test the hypothesis that using the assumption of heavy tailed distribution it is possible to forecast market risk more precisely, especially in times of crisis, than under assumption of normal distribution. Our empirical results indicate that the most adequate GARCH type model for estimating and forecasting volatility in the Serbian stock market is EGARCH model with assumption that the residuals follow the normal distribution and GARCH (1,1) model with assumption that the residuals follow the Student's t distribution. Our backtesting results for the last 200 observations based on the Kupiec POF test show that EGARCH model with normal distribution and GARCH(1,1) model with Student t distribution of residuals passed Kupiec test with 99% of confidence level, but not with 95% of confidence level, which imply that these models underestimate VaR at 95% confidence level.
Key Words: value at risk, BELEX 15 index, GARCH models, backtesting
MERENJE VREDNOSTI PRI RIZIKU KOD TRŽIŠTA U USPONU: EMPIRIJSKA ANALIZA TRŽIŠTA KAPITALA U SRBIJI
Osnovni cilj ovoga rada jeste istraživanje performansi različitih specifikacija simetričnih i asimetričnih GARCH modela uz pretpostavku normalne i Studentove t distribucije u cilju procene i predviđanja tržišnog rizika na tržištu kapitala u Srbiji. Koristeći dnevne stope prinosa srpskog indeksa akcija BELEX15 testirali smo različite asimetrične i simetrične GARCH modele uz pretpostavku normalne i Studentove t distribucije za period oktobar 2005-octobar 2012 što predstavlja značajan vremenski period koji sadrži periode pre i tokom finansijske krize.
U uslovima globalne finansijske krize za investitore je naročito važno da što preciznije izmere i alociraju rizik i efikasnije upravljaju svojim portfolijom. Uticaj ekstremnih događaja na kretanja na finansijskim tržištima zemalja u razvoju je još izraženiji, budući da je reč o tržištima koje karakteriše niži stepen likvidnosti i znatno manja tržišna kapitalizacija. Mogućnost primene VaR metodologije, koja je u osnovi kreirana i razvijena za likvidna i razvijena tržišta, potrebno je testirati na tržištima u razvoju koja karakterišu volatilnost, nelikvidnost i plitkost tržišta. To nas je motivisalo da primenimo metode koje uključuju vremensku promenljivu volatilnost i teže repove empirijske distribucije prinosa. Testirali smo hipotezu da li je moguće tačnije predvideti tržišni rizik, posebno u vremenima krize, korišćenjem pretpostavke o težim repovima distribucije nego pod pretpostavkom normalne distribucije. Empirijski rezultati ukazuju da model EGARCH uz pretpostavku da reziduali prate normalnu distribuciju i model GARCH (1,1) uz pretpostaku Studentove t distribucije imaju najbolje statističke karakteristike. Naši backtesting rezultati za poslednjih 200 opservacija prinosa zasnovani na Kupiec-ovom POF testu pokazuju da EGARCH model uz pretpostavku da reziduali slede normalnu distribuciju i GARCH tipa uz pretpostavku Student-ove t distribucije daju dobre procene vrednosti VaR-a uz interval poverenja od 99%, dok za 95% navedeni modeli nisu prošli Kupiec-ov test, što ukazuje da potcenjuju stavrne gubitke.
Ključne reči: vrednost pri riziku, BELEX 15 index, GARCH modeli, backtesting
