Browsing by Author "Eisenstein, Kaylah"

Now showing 1 - 1 of 1
  • Thumbnail Image
    Item
    Open Access
    Identifying jumps in financial time series: a comparative study of jump detection tests
    (2022) Eisenstein, Kaylah; Ouwehand, Peter
    There is consensus in the financial literature that traded asset prices may be subject to rare, but sudden movements, resulting in asset price discontinuities, known as jumps. It is therefore important to not only incorporate jumps into diffusion models but also to disentangle the diffusion component, which can be hedged, from the jump component, which typically cannot. Consequently, there is a need to identify jumps in financial time series. A number of non-parametric finite activity jump detection tests have been proposed by various scholars. In this dissertation, a comparative study amongst these jump detection tests is conducted. A Monte Carlo simulation is performed using a variety of data generating processes, model parameter values and sampling frequencies. The Matthews correlation coefficient and bookmaker informedness are used to compare the absolute and relative performances of the jump detection tests. In particular, the multi-power variation tests of BarndorffNielsen and Shepard (2004, 2006) and Andersen et al. (2004), the minimum and median variance tests of Andersen et al. (2009), the threshold multi-power variation test of Corsi et al. (2010), the instantaneous volatility test of Lee and Mykland (2008), the swap variance tests of Jiang and Oomen (2008) and combinations thereof are considered in the study. Generally, the absolute performances of the Lee and Mykland test are consistently strong. Consequently, it emerges as the most accurate jump detection test in most scenarios. However, when asset prices with stochastic volatility experience particularly high levels of volatility, the Lee and Mykland test experiences an inability to adequately disentangle the diffusion and jump components. The swap variance tests consistently emerge as the worst performing jump detection tests. Nevertheless, a combination of either the minimum variance and ratio swap variance tests, or the median variance and ratio swap variance tests perform notably well across the different scenarios, particularly when the volatility is high.