Skip to content
All library documents

Using EWMA Estimates as Inputs to VaR Calculations

Article Quant Q&A · Author: Bobby Digital
Read the original

Summary

The document asks how to incorporate more responsive volatility estimates, such as EWMA or GARCH, into VaR calculations with R's PerformanceAnalytics tools. One answer says the VaR function accepts externally estimated distribution inputs, including mean, covariance, co-skewness, and co-kurtosis, so estimates from another method can be supplied rather than relying only on default inputs.

A second answer outlines an EWMA route using a weighted covariance estimate, then converting the resulting volatility into VaR under a normality assumption. The discussion is brief and does not provide a complete implementation, establish that the suggestion applies to every VaR mode, or compare EWMA with GARCH in practice. The normal-distribution step also limits the approach's ability to represent non-normal tails unless additional distributional information is supplied.

Key ideas

  • VaR inputs can include externally estimated moments and covariance information.
  • EWMA or GARCH estimates may be calculated separately and passed into a VaR calculation.
  • An EWMA covariance estimate can provide volatility for a VaR calculation under a normality assumption.
  • The document gives suggestions rather than a full implementation or comparative evaluation.

Tags

Risk managementVolatilityStatistics
Full text
# Is there a way to adjust R PerformanceAnalytics function VaR with EWMA or GARCH method?


# Is there a way to adjust R PerformanceAnalytics function VaR with EWMA or GARCH method?












Is there a way to upgrade R PerformanceAnalytics function VaR with more risk sensitive approaches like EWMA or GARCH? Or is there another R package which can handle the issue?

## Answer by Kyle Balkissoon (score 1)

https://quant.stackexchange.com/a/15875

You can pass in the parameters are you estimating with EWMA or GARCH using the mu (mean), sigma (co/variance) m3 (co/skewness) and m4(co/kurtosis) arguments.

e.g. blahblah = EWMA(my_time_series)

VaR(my_time_series,mu=blahblah)

## Answer by m_099 (score 0)

https://quant.stackexchange.com/a/14759

If you want to take the EWMA approach have a look at cov.wt() from the stats package. It will give you an EWMA volatility, which you can then, given the normal distribution assumption, easily transform into VaR.

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.