rolling VAR

[WALLE]

Hello,

I am trying to estimate a rolling bi-variate VAR as in Blanchard and Gali (2007): The Macroeconomic Effects Of Oil Price Shocks: Why Are The 2000s So Different From The 1970s? Basically its similar to the idea behind ROLLREG rolling regressions just that its a VAR with two variables. It's a more flexible approach to a normal VAR and its allows one to estimate rolling IRFs to shocks, based on a simple dynamic equation linking a variable of interest to its own lags and the current and lagged values of the other variable. It involves using a moving window of n observations which is 'rolled' over all observations to ascertain how the variables respond to shocks overtime in each other.

I would like to obtain the evolving Impulse response graphs of a shock in one variable to another, however I have had difficulty finding any codes similar to this. I will like to inquire if a procedure exists for this in RATS. If YES, I would like to be directed to it. If NO I would require assistance in coming up with a code to carry out the procedure if its possible on RATS. Thanks

[TomDoan]

Hello,

I am trying to estimate a rolling bi-variate VAR as in Blanchard and Gali (2007): The Macroeconomic Effects Of Oil Price Shocks: Why Are The 2000s So Different From The 1970s? Basically its similar to the idea behind ROLLREG rolling regressions just that its a VAR with two variables. It's a more flexible approach to a normal VAR and its allows one to estimate rolling IRFs to shocks, based on a simple dynamic equation linking a variable of interest to its own lags and the current and lagged values of the other variable. It involves using a moving window of n observations which is 'rolled' over all observations to ascertain how the variables respond to shocks overtime in each other.

I would like to obtain the evolving Impulse response graphs of a shock in one variable to another, however I have had difficulty finding any codes similar to this. I will like to inquire if a procedure exists for this in RATS. If YES, I would like to be directed to it. If NO I would require assistance in coming up with a code to carry out the procedure if its possible on RATS. Thanks

I've seen Jordi Gali's program for that, and there's really not much more to the basic analysis than putting a loop for the rolling estimation period around a standard estimation of a VAR (done using ESTIMATE, nothing fancy) and generation of IRF's by Monte Carlo integration. (The full program is much more complicated than that, but that's because they were experimenting with many different combinations of variables).