I am doing research to study volatility spillover effects between several financial time series $\{x_{1,t}\}, \dots, \{x_{k,t}\}$ (in my case, $k=4$). What would be the best model to study the spillover effect of other series on $x_{1,t}$?

I was thinking of using VAR-GARCH or BEKK-GARCH Model in R. Any suggestion would be highly appreciated.

  • $\begingroup$ This is an edited copy of a thread that has been deleted. I repeatedly encounter similar questions, and the answer I gave in the deleted thread is relevant for many users. Therefore, I have copied the thread to preserve the answer. $\endgroup$ Commented Jun 18, 2022 at 7:10

1 Answer 1


VAR is a conditional mean model and as such has nothing to do with volatility spillovers – unless you apply it on realized variances instead of raw data. In that case it is a model that directly allows for volatility spillovers; see Diebold & Yilmaz "Better to give than to receive: Predictive directional measurement of volatility spillovers" (2012).

Unfortunately, full BEKK-GARCH appears to be a nonsense model; see McAleer "What They Did Not Tell You about Algebraic (Non-) Existence, Mathematical (IR-)Regularity and (Non-) Asymptotic Properties of the Full BEKK Dynamic Conditional Covariance Model" (2019). Diagonal BEKK-GARCH is fine, though (ibid.), but it does not seem to allow for volatility spillovers; see e.g. equations (6), (9) and (11) on p. 5 of Erten et al. "Volatility Spillovers in Emerging Markets During the Global Financial Crisis: Diagonal BEKK Approach" (2012) – even though people use it for the purpose (e.g. Zolfaghari et al. "Volatility spillovers for energy prices: A diagonal BEKK approach" (2020)).

Some people use DCC-GARCH for the purpose, but that model does not allow for volatility spillovers by design (just like the diagonal BEKK-GARCH), as I have explained elsewhere. Moreover, it is also full of problems: McAleer "What They Did Not Tell You about Algebraic (Non-) Existence, Mathematical (IR-)Regularity, and (Non-) Asymptotic Properties of the Dynamic Conditional Correlation (DCC) Model" (2019).

Other models (e.g. GO-GARCH)
There are other multivariate GARCH models out there, as reviewed in Bauwens et al. "Multivariate GARCH models: a survey" (2006) and Silvennoinen & Teräsvirta "Multivariate GARCH models" (2009). Perhaps some of them might work. E.g. consider GO-GARCH; there are at least a couple of packages for estimating the model in R, namely, gogarch and rmgarch. I am just afraid McAleer comes out soon and explains how GO-GARCH is full of problems, too :)
RIP Michael John McAleer. The world of econometrics has lost a great man.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.