Garch hedge ratios using Dynamic Programming

[jman]

Hi all,

I am using Eviews6 and trying to implement the Haigh&Holt (2002) dynamic programming model. It is essentially a BEKK-BiVar-GARCH, but the hedge ratios are calculated slightly different. With a 4-week hedging horizon.

Beta(T-1) = Cov(Spot at T, Futures at T-1 | given information up to T-1) / Var(Futures at T-1| given information up to T-1)

But the hedge ratio for T-2 and T-3 is,

Beta(T-2) = Cov(Spot at T, Futures at T-1 | given information up to T-2) / Var(Futures at T-1| given information up to T-2)
Beta(T-3) = Cov(Spot at T, Futures at T-2 | given information up to T-2) / Var(Futures at T-2| given information up to T-3)
Beta(T-4) = Cov(Spot at T, Futures at T-4 | given information up to T-4) / Var(Futures at T-4| given information up to T-4)

As you can see the conditional covariances are slightly different in that it the lags change for the futures part but the spot stays at T). Even if you can't get the hedge ratios out, then the updating covariances/variances series with be a big help.

I have included the paper in this email. Also have included a sample work file with data. Any help would be greatly appreciated, really would be saving the day!!

The Paper: http://www.speedyshare.com/files/236634 ... lt2002.pdf
Data: http://www.speedyshare.com/files/23663350/data.wf1


Thank you
Jman

HaighHolt2002.pdf

(567.43 KiB) Downloaded 262 times

data.wf1

(42.26 KiB) Downloaded 223 times

[jman]

Dear All,

I noticed a mistake for betas. It should have read:

But the hedge ratio for T-2 and T-3 and T-4 is,

Beta(T-2) = Cov(Spot at T, Futures at T-1 | given information up to T-2) / Var(Futures at T-1| given information up to T-2)
Beta(T-3) = Cov(Spot at T, Futures at T-2 | given information up to T-3) / Var(Futures at T-2| given information up to T-3)
Beta(T-4) = Cov(Spot at T, Futures at T-3 | given information up to T-4) / Var(Futures at T-3| given information up to T-4)

If anyone could please help. Would be greatly appreciated.