Averaging tests for jumps

We propose a new procedure to detect jumps in prices in the presence of microstructure noise. The procedure averages the results from existing tests across sampling frequencies. This approach overcomes the sub-sampling and low-power problems that plague existing tests for jumps. We use a modified version of Fisher’s method to combine p-values for the Barndorff-Nielsen and Shephard (2006) test applied at different sampling frequencies. We propose a double bootstrap procedure to obtain an empirical distribution for Fisher’s test statistic. As an additional contribution, we prove the bootstrap consistency of the Barndorff-Nielsen and Shephard (2006) test.