High-frequency analysis requires techniques that separate continuous diffusion from discrete jumps in equity returns. Researchers focus on measures that are robust to market microstructure noise and that can reliably detect both the occurrence and magnitude of jumps. Evidence from leading econometricians clarifies which quantitative tools are most effective.
Statistical techniques
Realized variance versus bipower variation forms a foundation for jump detection. Barndorff-Nielsen Aarhus University and Shephard University of Oxford developed bipower variation to estimate continuous variation while filtering out jumps; comparing realized variance to bipower variation highlights discontinuities as excess variation. Power variation and its ratios underpin formal tests that are relatively simple to compute from high-frequency returns and remain central in practice.
Building on that foundation, high-frequency asymptotic tests provide statistical significance for jumps. Aït-Sahalia Princeton University and Jacod Université Paris developed tests based on asymptotic properties of power variations that control size and power as sampling intervals shrink. Lee and Mykland proposed local test statistics that flag individual jump events using extreme return behavior, and Mykland Northwestern University contributed to refinement of such tests in the presence of noise. For estimation of jump sizes and dynamics within parametric frameworks, researchers use stochastic volatility models with jumps, estimating them by likelihood methods or particle filtering to capture jump intensities and amplitudes.
Causes, consequences, and nuances
Jump risk often stems from macroeconomic announcements, corporate news, liquidity shortages, or abrupt order flow—drivers that differ across markets. Emerging markets tend to exhibit more pronounced jumps because thinner liquidity and less timely disclosure amplify discrete moves, while developed markets show frequent small jumps tied to algorithmic trading and news fragmentation. Microstructure noise, including bid-ask bounce and asynchronous trading, can mimic jumps; practitioners mitigate this with pre-averaging, subsampling, or robust variance estimators.
Accurate measurement of jumps matters for pricing and risk management. Option pricing models that omit jumps understate short-term tail risk and misprice wings of the implied volatility surface, prompting systematic hedging errors. For portfolio managers and regulators, underestimating jump frequency leads to understated Value-at-Risk and stress-test failures. Combining robust nonparametric diagnostics such as bipower variation with parametric jump-diffusion estimation yields the most informative picture of jump risk across cultural and territorial market structures, enabling better-informed decisions under sudden market shifts.