Long Memory and Structural Breaks in Extreme Value Estimators

With numerous studies reporting long memory in financial volatilities, long memory became one of the stylized facts of volatility time series. Several researchers, however, including Granger and Hyung (2004) and Choi and Zivot (2007), argue that the long memory property of financial volatilities may be amplified by occasional structural breaks. This paper investigates the validity of the previous studies - whether long memory in extreme value estimators is overstated by structural breaks. I find an evidence that the degree of long memory in the extreme value estimators is inflated by structural breaks. I also find, however, that significant long memory is still discovered in the extreme value estimators even after the multiple breaks are controlled in the estimation.