conditional forecasts with shocks limited to some variables

[tclark]

Has anyone coded up conditional forecasting in which shocks to some variables are shut down? The condit.src procedure provides the general Doan, Litterman, Sims (1984, Econometric Reviews) approach to conditional forecasting, in which hitting the condition of interest can involve shocks to all variables. DLS also describe hitting the condition of interest by forcing the shocks to come from only some variables (e.g., one might want to hit an interest rate path conditional on only structural shocks to the interest rate). Doing so involves zeroing out the relevant columns of the R matrix in DLS. If anyone has ever coded up this up and would be willing to share code, I would really appreciate it.

[TomDoan]

You can accomplish the same thing by zeroing out columns of the factor matrix that's input to @CONDITION. For instance, the following F matrix will eliminate the 2nd and 4th Cholesky shocks:

compute f=%decomp(s)
ewise f(i,j)=%if(j==2.or.j==4,0.0,f(i,j))

[tclark]

That's great -- thanks very much

[TomDoan]

My bad. It's just zapping the columns, not the rows as well. I corrected the post above.