Detecting sudden changes in volatility estimated from high, low and closing prices

In this paper, we assess the size and power properties of Inclan and Tiao's (1994) Iterated Cumulative Sum of Squares (IT ICSS) algorithm for detecting sudden changes in volatility. We make use of the variance estimator that utilizes high, low and closing prices proposed by Rogers and Satchell (1991) (RS) and compare it with the performance of the demeaned squared returns. We find that the IT ICSS algorithm exhibits more desirable size and power properties when applied with the RS estimator in comparison to the demeaned squared returns. On the empirical side, we apply the IT ICSS algorithm with the RS estimator and demeaned squared returns of the S&P 500, CAC 40, FTSE 100, IBOVESPA and SZSE Composite indices to detect sudden changes in volatility of both developed and emerging markets. We find that most of the structural breaks detected by the RS estimator can be related to major macroeconomic events while very few of the structural breaks detected by demeaned squared returns can be related to macroeconomic events and hence are probably spurious.

► This paper suggests the use of the RS estimator to detect sudden breaks in volatility. ► The RS estimator is an unbiased volatility estimator regardless of the drift parameter. ► The RS estimator exhibits desirable size and power properties in the IT ICSS algorithm. ► The breaks detected by the RS estimator can be related to major macroeconomic events. ► Hence, we propose the use of the IT ICSS algorithm based on the RS estimator.