Testing VaR Violation Independence and Clustering

Summary

This discussion asks how practitioners test whether Value at Risk (VaR) violations are independent over time, with particular interest in detecting clustering and deciding when a model needs review. It points to violation-clustering tests in a thesis on VaR backtesting and mentions the runs test, including its availability in an R time-series package. It also identifies the Kupiec test as a commonly used VaR backtest and directs readers to Federal Reserve research for further reference.

The material offers leads to methods and references rather than a comparison of their power or practical acceptance criteria. It does not explain test assumptions, provide empirical evidence, or specify thresholds for triggering model reassessment. Readers would need to consult the cited research and broader backtesting literature to answer the original practitioner-focused questions.

Key ideas

• VaR backtesting can examine whether violations cluster over time.
• The runs test is one method mentioned for assessing violation clustering.
• The discussion identifies the Kupiec test as a frequently used VaR backtest.
• The document provides references but does not compare test power or define reassessment thresholds.

Tags

# Backtesting VaR model violation independence


# Backtesting VaR model violation independence












I am interested in hearing about the practitioner state of the art for testing the time independence of a VaR model (i.e. that VaR violations are independent in time). There are a number of tests in the literature. What do people actually use? Which are viewed as the most powerful and useful in the real world? I'm interested in hearing about what kinds of specific quantitative criteria people might use to cause them to re-evaluate their VaR model's current ability to handle violation clustering.

## Answer by Richi Wa (score 4)

https://quant.stackexchange.com/a/4046

The following thesis deals with VaR back testing procedures in the Basel framework link. In chapter 7 tests for violation clustering are presented. An R implementation of the runs test is e.g. given in the tseries package.

## Answer by pyCthon (score 2)

https://quant.stackexchange.com/a/4055

The Kupiec test is frequently used to back test VaR values. I would refer you to read this article from the federal reserve

http://www.federalreserve.gov/pubs/feds/2005/200521/200521pap.pdf

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.