On impulse response analysis

[Voyant]

Hello comrades! Nice to meet you!

I run a a VECM model with 4 endogenous variables and 3 cointegration equations from 2003 to 2013 in quarterly frequency.
I use Cholesky dof in 10 periods in impulse response analysis.

Here are my questions:

1) At the end of the determined period window (mine is 10) the shocked variables must return to zero or it depends on their actual evolution?

2) Why Cholesky dof is recommended while many papers use generalized impulse responses?

3) Which are the actual restrictions of default VECM model (are only those which concern the intercept?) and how they affect the outcome of the impulse response analysis? Namely, how they affect the presuppositions of the interaction of the variables?

Thank in advance guys for any reliable answer!

[londonphd]

Hello comrades! Nice to meet you!

I run a a VECM model with 4 endogenous variables and 3 cointegration equations from 2003 to 2013 in quarterly frequency.
I use Cholesky dof in 10 periods in impulse response analysis.

Here are my questions:

1) At the end of the determined period window (mine is 10) the shocked variables must return to zero or it depends on their actual evolution?

2) Why Cholesky dof is recommended while many papers use generalized impulse responses?

3) Which are the actual restrictions of default VECM model (are only those which concern the intercept?) and how they affect the outcome of the impulse response analysis? Namely, how they affect the presuppositions of the interaction of the variables?

Thank in advance guys for any reliable answer!

1. No, it depends on your data. The response function may not return to its steady state i.e. 0 let's say. It could stay away from it for more than 10 periods

2. Identification using cholesky decomposition or any other method is arbitrary and really depends on what you want to achieve or looking into.

3. The deterministic terms in VECM are there to describe the data generating process. If you do not include the right deterministic terms, you may not obtain the correct cointegrating vectors and hence your impulse responses analysis would be affected.

For more technical discussion, you should read books on applied econometrics