Skip to content
All library documents

BEKK-GARCH Availability in Stata and Multivariate Alternatives

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

Summary

This exchange discusses whether Stata can estimate a BEKK-GARCH model and clarifies when that model is appropriate. One answer says it is not aware of a Stata implementation and points to R software that can fit a BEKK(1,1) model, while noting that estimation may take time. It suggests constant-correlation GARCH or dynamic conditional correlation GARCH as alternatives available in both Stata and R, with Stata's multivariate GARCH commands mentioned as options.

A second answer emphasizes that BEKK is intended for multivariate volatility modeling, so it is not a natural choice for a single time series. It describes the alternatives as ways to represent changing conditional correlations. The exchange is brief and provides no implementation details, comparative results, or guidance on choosing among specifications. Its software availability claims reflect the answers in the discussion and may not capture later package or Stata updates.

Key ideas

  • BEKK-GARCH is designed for multivariate conditional volatility modeling rather than a single series.
  • The exchange reports that Stata does not directly fit BEKK, while R software can estimate a BEKK(1,1) model.
  • Constant-correlation and dynamic conditional correlation GARCH are presented as alternative multivariate approaches.
  • Dynamic conditional correlation models allow conditional correlations to change over time.

Tags

StatisticsVolatilityBacktesting
Full text
# BEKK - GARCH model in Stata


# BEKK - GARCH model in Stata












Is it possible to run BEKK-GARCH in Stata? mgarch is of a different model type and google provide me with no good hints.

## Answer by Alejandro Andrade (score 1)

https://quant.stackexchange.com/a/22138

I don't think that it is posible to fit a BEKK in Stata, i'm not sure if you are familiar with R but with the package `MTS` you can fit a BEKK(1,1), it takes some time to run, i would suggest that you consider a CC-GARCH wich can be estimated in Stata and in R with `ccgarch` or a DCC that is also posible in both programs

## Answer by Archie C (score 0)

https://quant.stackexchange.com/a/75036

It doesn't make much sense to me to use a BEKK if you were only looking at one time series. From my understanding you would use BEKK's for Multivariate (G)ARCH models. In this case you can use `mgarch dvech` or `mgarch dcc`. `dcc` stands for dynamic conditional correlation. Both of these modeling methods allow for changing conditional correlations.

see `help mgarch`

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.