Testing for Jumps in Near Non-Stationary Diffusion Processes
26 Pages Posted: 13 Oct 2017
Date Written: October 12, 2017
In this paper, we show that despite the fact that Ornstein-Uhlenbeck (OU) processes fall within the general specification of asset price dynamics studied by Lee and Mykland (2008), the finite sample performance of their two tests for additive jumps is far from being satisfactory when the process deviates from the random walk, resulting in a strong size distortion and a power loss. Therefore, we propose a modification to their test that improves the finite sample performance for local-to-unity processes (in both explosive or stationary directions). We apply the tests on 21 years of 5-minute log returns of the Nasdaq stock price index and find that, unlike the other two tests, our test allows to detect jumps when log-prices exhibit clear upward or downward trend movements.
Keywords: Jumps, Ornstein-Uhlenbeck processes, random walk, local-to-unity, intraday data
JEL Classification: C12, C14
Suggested Citation: Suggested Citation