نقض ناگهانی در برآوردگر نوسانات رانش مستقل بر اساس دوره های متعدد باز، بالا، پایین، و قیمت نزدیک   Sudden breaks in drift-independent volatility estimator based on multiple periods open, high, low, and close  prices

Introduction This paper compares the performance of the Yang and Zhang (2000) (YZ) estimator and demeaned squared returns when applied with Inclan and Tiao’s (1994) iterated cumulative sum of squares (IT-ICSS) algorithm to detect the sudden changes in the volatility of a random process. Analysis of market risk plays a crucial role in the financial markets literature. Volatility is known to be a popular measure in evaluating financial risks, leverage effects and to examine the impact of asymmetric shocks on markets. Volatility plays an important role in financial markets due to its application in designing investment decisions and in portfolio rebalancing and management (Aizenman & Marion, 1999), in pricing derivatives securities (Hull & White, 1987), in quantifying risk (based on value at risk and expected shortfall) (Granger, 2002) and in implementing trading strategies (Poon & Granger, 2003). It is well known in the literature that the unconditional volatility of tradable securities and portfolios may be signifi- cantly affected by infrequent structural breaks or regime shifts, which may arise due to various domestic or global macroeconomic and political events (see Aggarwal, Inclan, & Leal, 1999; Kumar & Maheswaran, 2012) including terrorist attacks, wars, sudden hike in interest rates, changes in investors’ perception, crashes and crises in a market, or recession in an economy. Hence, it is important to consider the impact of sudden changes in volatility in the model for generating more accurate forecasts of volatility. This can be helpful for fund managers and investors to design investment strategies, to rebalance their portfolios and to hedge their positions based on an anticipation of future movements of the market. Regulators, policy makers and central banks also have an interest in volatility analysis to implement policymeasures effectively based on changes taking place in markets. This would enable them to maintain stability in financial markets and to assess the effectiveness of these policies depending on the required goals (Poon & Granger, 2003). There exist different methods to estimate daily unconditional volatility. The most popular method involves the use of square of close to close returns. The returns based volatility measures are well established in literature and also act as inputs to generalised autoregressive conditional heteroskedasticity (GARCH) class of models. However, the squared daily return is a noisy estimate of volatility and informational inefficiency (Alizadeh, Brandt, & Diebold, 2002). Another method involves the use of high frequency intraday data. This measure of volatility is also known as realised volatility which involves summing the squares of returns sampled at shorter intervals (for example, 5 minutes or 10 minutes) for a given day. However, the high frequency data exhibit non negligible microstructure issues, which may prevent the researchers in analysing its informational contents. On the other hand, high frequency data for many assets may not be available or may be available for a shorter duration. In addition, high frequency data are generally expensive and require substantial computational resources. The literature that started with Parkinson (1980) and Garman and Klass (1980), and extended by Rogers and Satchell (1991) and Yang and Zhang (2000), has highlighted the importance of using opening, high, low and closing prices of an asset for the efficient estimation of volatility. Alizadeh et al. (2002) highlighted that range based volatility estimates are highly efficient and are robust in terms of the non negligible market microstructure issues. Among all these range based volatility estimators, the YZ estimator proposed by Yang and Zhang (2000) is based on multi-period open, high, low, and close prices, is unbiased in the continuous limit, independent of any non-zero drift, and incorporates the impact of opening price jumps. However, the RS estimator proposed by Rogers and Satchell (1991) is also unbiased regardless of non-zero drift. The YZ estimator also makes use the of RS estimator for volatility estimation (see equation (5) in the section on Methodology). The other range based volatility estimators are biased in some way if the mean return (drift) is non-zero. The open, high, low, and close prices are also available for most of the traded assets, indices and commodities, and contain more information for efficient estimation of volatility