Circular Block Bootstrap VaR and Artificial Extreme Returns

Summary

The document raises a practical concern about estimating Value-at-Risk with a circular block bootstrap. The author asks how to wrap sampled return blocks from the end of a series back to its beginning, particularly around the observations labeled P10 and P1, without creating an implausible return. They report that an unadjusted calculation produced a 99th-percentile VaR of -70% and say the implied observation was not real.

No solution, data description, or validation is provided. The example highlights that the way blocks are joined can create boundary transitions absent from the original data, potentially distorting tail-risk estimates. The reported figure is specific to the author’s calculation and cannot establish how common or severe this issue is. Readers would need the return series, block construction details, and sign convention for VaR to assess a remedy.

Key ideas

• Circular block bootstrap sampling can create joins between the end and start of a return series.
• The author reports that an unadjusted join produced a 99th-percentile VaR of -70%.
• The document asks how to avoid artificial boundary returns but does not give a method or resolution.

Tags

# Circular block bootstrap (CBB) Value-at-Risk (VаR)


# Circular block bootstrap (CBB) Value-at-Risk (VаR)












I am calculating Value-at-Risk (VAR) with the Circular block bootstrap (CBB) method.

How do I complete the circle and join $P_{10}$ and $P_1$ so that I do not get -70% yield? If I proceed with CBB VAR without any adjustments, my .99 percentile VAR becomes -70%! That observation is not even real.

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.