Skip to content
All library documents

When Hodges–Tompkins Volatility Debiasing Applies to Estimators

Article Quant Q&A · Author: earthlink
Read the original

Summary

The document raises a methodological question about the Hodges–Tompkins adjustment, a correction associated with volatility estimates from overlapping observations. It asks whether the adjustment can remove bias for any estimator constructed from overlapping data, or whether its applicability depends on the estimator’s assumptions. The question specifically contrasts common close-to-close estimation with range-based estimators such as Garman–Klass and Parkinson.

No answer, derivation, or empirical comparison is included, so the document does not establish that the correction transfers across estimators. It points to constant volatility as an assumption used by many implementations, making model assumptions and estimator construction central to assessing validity. A researcher would need to examine the adjustment’s derivation under the relevant sampling scheme and volatility model before applying it to overlapping range-based estimates. The excerpt is a focused research question rather than a complete guide or evidence-backed recommendation.

Key ideas

  • The Hodges–Tompkins adjustment is considered for volatility estimates based on overlapping observations.
  • The document asks whether applicability extends beyond a particular estimator family.
  • Close-to-close, Garman–Klass, and Parkinson estimators are raised as examples.
  • Constant volatility is identified as an assumption in many implementations.
  • No derivation or result is provided to settle the question for these estimators.

Tags

VolatilityStatistics
Full text
# Is Hodges-Tompkins adjustment applicable for all volatility estimators?


# Is Hodges-Tompkins adjustment applicable for all volatility estimators?












Can the Hodges-Tompkins adjustment (S. Hodges, R. Tompkins, The Sampling Properties of Volatility Cones 2000,2002 JoD) be used to de-bias any estimator computed from overlapping observations? It appears that most implementations apply it under the assumption of constant volatility. Is this adjustment applicable to estimates obtained from close-close, Garman-Klass, Parkinson, etc. on overlapping observations?

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.