Price Jumps Identification Using Integrated Variance Estimators

Technological advances and increasing availability of high-frequency data observed at very short time intervals, e.g. every minute or second, have enabled to use more complete information for nonparametric estimation of the continuous stochastic price process. Therewith, the jump component which is commonly described by the Poisson stochastic process, additively extends the Itô process components with time-varying parameters. Adding the price jumps to the stochastic price process significantly changes traditional understanding of the financial asset pricing models and has serious implications on financial risk management. Therefore, the objectives of this paper are theoretical explanation of the causes and the consequences of the price jumps, empirical identification of the price jumps and determination of their contribution to the total variance of returns at Croatian stock market. Since the realized variance is actually an estimate of the integrated variance (stochastic integral), statistical tests were developed that compare the realized variance of returns which is robust and the one which is not robust to the price jumps. Against this background, thet Barndorff-Nielsen and Shephard test indicates that the intensity of jumps is 76.19% in the observed period, and their contribution to the total variance of CROBEX returns ranges between 7.36% and 83.73%. Findings also confirm that price jumps are more induced with the well-known shocks rather than disagreements among investors.