This paper proposes a newly-developed jump test based on the order statistics from intraday returns to identify the direction and magnitude of jumps. Our identiﬁcation strategy, as suggested in the recent high-frequency ﬁnance, hinges on the local Gaussianity of the intraday return distribution in absence of jumps. Our test allows for an operational threshold in examining and characterizing a potential spectrum of jump sizes with signs. Besides its better statistical size and power properties, our numerical results demonstrate its robustness to the threshold, market microstructure noises, and the underlying stochastic volatility; our empirical evidence delivers several interesting and meaningful points. First, the numbers of identiﬁed positive and negative jumps are signiﬁcantly different in their intensities and sizes, respectively. We ﬁnd that jumps come in clusters and such clustering patterns can be linked to forward looking variables such as VIX. Moreover, by varying the operational threshold, our jump test allows us to sketch a rough picture understanding various features of ﬁnite-activity jumps. Finally, we observe asymmetries in the intensities as well as the jump-contributed price variations between the positive and negative ones. Given these observations, we expect this jump test to be potentially useful in investments or risk management.
High Frequency, Finite Activity Jumps, Jump Diffusion, Realized Variance, Order Statistics