Bootstrap Methods for the Realized Bipower Variation and for Jump Testing

This paper proposes bootstrap methods for the realized bipower variation and the Barndorff-Nielsen and Shephard (2006a) test for jumps. These results enable inference for the realized bipower variation in the presence of jumps in prices. Both the i.i.d and the WILD bootstrap are shown to outperform results obtained through the asymptotic theory. To detect jumps in the presence of microstructure noise, we propose a procedure that averages test results across multiple sampling frequencies. This method considerably improves jump detection, by generating a higher level of power than the asymptotic test, unaccompanied by a simultaneous increase in size.