Several new statistical procedures for high-frequency financial data analysis have been developed to estimate risk quantities and test the presence of jumps in the underlying continuous-time financial processes. Although the role of micro-market noise is important in high-frequency financial data, there are some basic questions on the effects of presence of noise and jump in the underlying stochastic processes. When there can be jumps and (micro-market) noise at the same time, it is not obvious whether the existing statistical methods are reliable for applications in actual data analysis. We investigate the misspecification effects of jumps and noise on some basic statistics and the testing procedures for jumps proposed by Ait-Sahalia and Jacod (Ann Stat 37–1:184–222 2009; 38–5:3093–3123 2010) as an illustration. We find that their first test (testing the presence of jumps as a null-hypothesis) is asymptotically robust in the small-noise asymptotic sense against possible misspecifications while their second test (testing no-jumps as a null-hypothesis) is quite sensitive to the presence of noise.
- Asymptotic robustness of Jump-Test
- Continuous-time processes
- High-frequency financial data
- Micro-market noise
- Small-noise asymptotics