Testing for the presence of jump components in jump diffusion models

In this paper we propose a nonparametric test to determine whether an underlying jump diffusion process indeed contains jump component, or equivalently, is indeed a diffusion. Our test is based upon a robust threshold estimation of diffusive volatility and the kernel estimation of the conditional moment function of the squared instantaneous increments of the underlying process. We show that our test statistic has asymptotic standard normal distribution under the null hypothesis of no jumps, is consistent against fixed alternatives, and may detect local alternatives that shrink to diffusions at certain convergence rates, when sampling interval shrinks to zero and time span is either fixed or expands. We only assume that the jump diffusion process is recurrent, thus allowing for both stationary and nonstationary cases. In addition, we provide a regression bootstrap test and establish its validity. A Monte Carlo simulation is conducted to examine the finite sample performances of our test, and an empirical illustration is also provided.