Jump-robust volatility estimation using nearest neighbor truncation

Research output: Contribution to journal/Conference contribution in journal/Contribution to newspaperJournal articleResearchpeer-review

We propose two new jump-robust estimators of integrated variance that allow for an asymptotic limit theory in the presence of jumps. Specifically, our MedRV estimator has better efficiency properties than the tripower variation measure and displays better finite-sample robustness to jumps and small ("zero") returns. We stress the benefits of local volatility measures using short return blocks, as this greatly alleviates the downward biases stemming from rapid fluctuations in volatility, including diurnal (intraday) U-shape patterns. An empirical investigation of the Dow Jones 30 stocks and extensive simulations corroborate the robustness and efficiency properties of our nearest neighbor truncation estimators.
Original languageEnglish
JournalJournal of Econometrics
Volume169
Issue1
Pages (from-to)75-93
Number of pages19
ISSN0304-4076
DOIs
Publication statusPublished - 1 Jul 2012

See relations at Aarhus University Citationformats

ID: 69757482