환율의 연속적 파워변동성 분포와 이산적 점프 및 미시적 시장교란 효과의 비모수 추정

In recent decade, the volatility has been one of the most important topics in finance and time series econometrics. Most studies adopted the parametric approaches such as ARCH class of the modes and stochastic volatility models. The parametric approaches, however, rely on explicit functional forms which can be misspecified. In recent years a few studies adopt nonparametric approaches which are free from specific functional forms. This paper uses a new nonparametric realized volatility model by summing intraday squared returns to explain the discrete jumps as well as continuous volatility so that realized volatility, bipower volatility, relative jump models are introduced. 2)Thus this paper analyzes the recent realized continuous volatility and discrete jump volatility of US dollar returns against the Euro using the ultra-high frequency five minute returns spanning the period from January 2001 through January 2010 during which volatility was appreciably large and the financial crisis appeared in the U.S. and European countries. As results of basic statistics and the density functions, the realized rate of returns, realized variation and realized bipower variation appear the characteristics of nonnormal distributions as usual in the financial time series which have the fat tails. The realized rate of returns, realized variation and realized bipower variation also have the volatility clustering effects and the significant volatility appeared particularly in years 2004 and 2008-2009. The realized jumps of US dollar returns against the Euro appeared to be distributed around zero according to its density function. This paper also uses several tests such as ARCH test, the variance-ratio tests,the long memory tests and Runs tests to analyze the characteristics of realized rate of returns, realized variation and realized bipower variation of US dollar returns against the Euro. Furthermore, this paper introduces the realized jumps using realized bipower variation and then use several jump statistics to identify whether the observed jumps are significant or not. To calculate the jump statistics,this paper uses the tripower quarticity and quadpower quarticity and then estimates the daily several jump statistics at several different crtical values. The empirical results show that many large jumps appear associated with the news announcements after the financial crisis. Especially before and after years 2004 and 2009, the realized variations had the considerably large and discrete jumps when I obtain the jump statistics using tripower and quadpower quarticity. Thus the jump component is important in explaining US dollar exchange rate versus Euro during this the period from January 2001 through January 2010. In the application to the US dollar exchange rate versus Euro, this paper adopts the jump statistics and jump probabilities using tripower and quadpower quarticity, respectively and then estimates the several jump probabilities at different critical values. The empirical estimates shows that the realized US dollar five minute returns against the Euro have the jump probability that there is at least one significant jump per five or seven days during January 2001 through January 2010 at common critical levels.