Realized normal volatility and maximum outlying jumps in high frequency returns for Korean won–US Dollar

Abstract

This paper analyzes the realized volatility and jumps of five-minute returns for the Korean won–US dollar exchange rate from June 2010 to April 2021. If standard normal distributed jump statistics are used, the jump probabilities of Korean won–US dollar are lower when jump occurrences are frequent, and the jump returns for Korean won–US dollar will be considerably underestimated. However, if we utilize maximum outlying jump statistics, the jump probabilities of Korean won–US dollar are much higher. Furthermore, if we include periodicity filters such as median absolute deviation (MAD), shortest half scale (ShortH), and weighted standard deviation (WSD) in the maximum outlying jump statistics, the five-minute returns of Korean won–US dollar have considerably lower jump probabilities. Thus, the jump probabilities of Korean won–US dollar returns depend on whether standard normal distributed jump statistics or outlying jump statistics are used with periodicity filters such as MAD, ShortH, and WSD, in jump periods. Therefore, outlying jump statistics with periodicity filters such as ShortH and WSD ought to be used to obtain more consistent jumps probabilities for volatility and frequent jumps in the foreign exchange rate returns of 2010–2021.