Question on Cushman and Zha (1997) and Zha (1999)

[jacarandas]

Hi,

I have been reading the following two papers that works on block exogeneity in a VAR.

1. Identifying monetary policy in a small open eocnomy under flexible exchange rates - Cushman and Zha (1997)
2. Block recursion and structural vector autoregressions - Zha (1999)

I have a few really basic questions:

1. From my understanding, Cushman and Zha (1997) uses a frequentist approach, while Zha (1999) develops a Bayesian approach that allows for more complicated restrictions to be imposed on the contemporaneous and lag matrices. Is that right?
2. Is there any reason why the domestic block is ordered before the external block in both papers? I know it doesn't really matter since my understanding of block exogeneity is that the blocks are estimated separately. However, it just seems intuitive for the external block to come before the domestic block.

Thanks

[TomDoan]

Hi,

I have been reading the following two papers that works on block exogeneity in a VAR.

1. Identifying monetary policy in a small open eocnomy under flexible exchange rates - Cushman and Zha (1997)
2. Block recursion and structural vector autoregressions - Zha (1999)

I have a few really basic questions:

1. From my understanding, Cushman and Zha (1997) uses a frequentist approach, while Zha (1999) develops a Bayesian approach that allows for more complicated restrictions to be imposed on the contemporaneous and lag matrices. Is that right?

That's basically correct. The Monte Carlo integration described in the 1997 paper isn't technically correct---the point estimates done with ML are fine, but you can't do importance sampling with the missing block of variables. http://www.estima.com/forum/viewtopic.php?f=8&t=1857 uses an efficient sampling method not covered in either of the papers.

2. Is there any reason why the domestic block is ordered before the external block in both papers? I know it doesn't really matter since my understanding of block exogeneity is that the blocks are estimated separately. However, it just seems intuitive for the external block to come before the domestic block.

I think the only reason would be that the interesting contemporaneous structural restrictions are on the domestic block; the foreign block is just a simple Cholesky factor "model".

[jacarandas]

Thanks Tom. I am interested to estimate a simple model with cholesky ordering in each of the block using Bayesian methods. I suppose Zha (1999) would be a good starting point for me?