Extreme value estimators and their performance across various beta levels

Volatility of an asset or a portfolio is one of the most important parameters that are estimated in asset allocation, risk management, options pricing etc. Various approaches have been developed to estimate and forecast the volatility of an asset. These include the conditional volatility estimators like ARCH and its various improvements and the unconditional volatility estimators like the extreme value estimators. This paper analyses empirically the efficiency and bias of the various extreme value estimators in estimating the volatility across different beta levels of the 50 stocks which make up the S&P CNX Nifty index. The Garman‐Klas estimator performs the best in estimating the volatility over one day and five day return periods for high and low beta stocks, while the other return periods and beta levels portray mixed results.