Regime switching VAR-GARCH-BEKK model

[thulz]

does anyone have codes for estimation of a bivariable regime switching VAR-BEKK model or regime switching VAR-GARCH-in-mean model? any assistance would be highly appreciated.

[TomDoan]

Do you have a reference? I can't imagine a model like that would actually work.

[thulz]

not really. my professor suggested i try the model. my paper is on estimating the effects of commodity price volatility regimes on stock returns.Or could a normal VAR GARCH-BEKK be used instead?

[TomDoan]

I would certainly start with a single regime model. GARCH models, after all, effectively generate "regimes" on their own, since there will be high volatility and low volatility periods because of the persistence of the GARCH process. Threshold models (as opposed to hidden Markov switching models) can work for GARCH, so that would be the direction to go if you're trying to generalize. But start with a basic VAR-GARCH and see if there is any sign that it's inadequate.

[thulz]

my reference paper is "Capital Inflows and Exchange Rate Volatility in Korea" by Choi, Chung and Kim. the paper uses elder's 2004 model as well a hamilton markov model. but i dont know how to incorporate the regime switching part. Any assistance would be appreciated.

[TomDoan]

I think you're a bit confused about what they're doing. They're doing a Markov switching model (univariate model on the exchange rate), and also doing a separate VAR-GARCH model. They aren't combining the two.

[thulz]

i think i am confused. how were they able to regress the low and high volatility regimes on capital flows?

[TomDoan]

It looks like the description in the text of the article has the regression described backwards. Table 3 shows capital flows regressed on the "HighRegime" series (which I assume is the smoothed probability of the high volatility regime) not regime regressed on capital flows.

[thulz]

ok. that makes sense. thanks for the clarification. Is there a code or procedure i can use to regress the smoothed probability on capital inflows?

[TomDoan]

Prof. Choi graciously provided the data used for the MS part of the analysis. See https://estima.com/forum/viewtopic.php?f=8&t=2343. Again, you are describing that backwards---the regression is of capital flows on the regime, not regime on capital flows. Since the paper isn't about regime switching VAR-GARCH-BEKK models, please use that other thread if you have questions about it.

One thing to note is that the reason they went to the VAR-GARCH model is that the capital flows on volatility regime is a two-step procedure which uses a generated regressor, while the VAR-GARCH model does a joint estimate.

[thulz]

I think I understand. thank you for all the assistance. the link you provided was very helpful.

[irfanawan]

Looking for code of Markov regime switching GARCH model proposed by Alizadeh, A.H., Nomikos, N.K., Pouliasis, P.K., 2008. A Markov regime switching approach
to hedging energy commodities. J. Bank. Financ. 32, 1970–1983

Regards,
Irfan