Estimating the degree of activity of jumps in high freq外文翻译.pdfVIP

ESTIMATING THE DEGREE OF ACTIVITY OF JUMPS IN HIGH FREQUENCY FINANCIAL DATA By Yacine AÔt-Sahalia1 and Jean Jacod Princeton University and UPMC (UniversitÈ Paris-6) We propose estimators of the index of jump activity of a discretely sampled process, and derive their properties. These estimators are applicable despite the presence of Brownian volatility in the process, which makes it more challenging to learn about the small, inÖnite activity, jumps. When the method is applied to high frequency stock returns, we Önd evidence of inÖnitely active jumps in the data and estimate their index of activity. 1. Introduction. Using high frequency Önancial data, which are now widely available, we can hope to answer a number of questions regarding the characteris- tics of the process that drives asset returns. Let us model the log-price X of some asset as a 1-dimensional process, which we will observe over a Öxed time interval [0; T] at discrete times 0; n ; 2n ; with a time interval n between successive observations that is small: this is the essence of high frequency data. Let us further assume that this process is an ItÙ semimartingale, meaning that its characteristics are absolutely continuous with respect to Lebesgue measure: so it has a drift, a con- tinuous martingale part which is the integral of a possibly stochastic process with respect to a Brownian motion, and we will also let it have jumps, with a possibly stochastic LÈvy measure. For modeling purposes one would like to infer the characteristics of X; that is, its drift, its volatility and its LÈvy measure, from the observations. When the ti