Diagonal BEKK model

[nifa]

Hello Everybody. I am trying to run this model in eviews in order to calculate the dynamic hedging ratio by calculating the conditional covariance and variance. I run it with two different ways:

1) Solve the system of the following equations:
x = c(1) + c(2)*x(-1) + c(3)*y(-1)

y = c(4) + c(5)*y(-1) + c(6)* x(-1)

where x is the spot and y the futures, and the take the matrixes with variances and covariance.

2) Click on Spot and Futures time series -> open as a system -> estimate -> Diagonal BEKK model


Unfortunately the results deviate between 1 and 2 method. Which one is the most accurate and is there any alternative way to run it?

[trubador]

Multivariate GARCH models are more appropriate in terms of building dynamic hedging strategies. Therefore, the second approach would serve the purpose better.

[nifa]

Both are subjected to the same model. With the first approach i try to constitute the BEKK model by estimating the system which was quoted. So it's more prudent to follow the second method?

[trubador]

I do not follow the difference between your specifications. It seems to me that second approach you mention only includes the constant terms.

The system equations you wrote correspond to "mean" part of your model. BEKK handles the ARCH dynamics in the "variance" part of your model, which operates on the stationary residuals. Since you work on price levels not returns, building a VAR(1) model as you did would be the better approach. You can also try modeling the log returns and see if it fits better.

[hasseb1]

Unfortunately the results deviate between 70-486 1 and 2 method. Which one is the most accurate and is there any alternative way to run it?