IRF from VECM using @BERNANKE

[kb1]

Hello,

Is there a procedure or suggestion for coding up IRFs (including confidence intervals) from a VECM, using the @BERNANKE procedure for orthogonalized innovations? Apologies if the answer is obvious, I have not been able to find much.

Thanks.

[TomDoan]

You won't find anything using the BERNANKE procedure since that's almost 20 years old now and has long been deprecated in favor of the CVMODEL instruction. The procedure for MC integration with a VECM is effectively the same as for a VAR if the cointegrating vector is treated as known. (If it's not treated as known, it's an incredibly difficult problem). If you're doing a parametric structural VAR (the type you would do with CVMODEL), the MC integration for those parameters is doable but requires some "art" and is generally not recommended as a first foray into modern simulation techniques. The MONTESVAR.RPF program included with RATS shows one of several possible techniques. The 2nd Edition of the VAR e-course has almost an entire chapter devoted to the three main methods (importance sampling, random walk Metropolis and the Waggoner-Zha sampler).