I tried to replicate CMR Risk Shocks model. From what I’ve read, the replication has been troublesome for many.

While there are no other technical problems (errors) with my estimation, I have trouble with shock decompositions. I get the shock decomposition graph after estimation, but risk shock is missing.

In their .mod-file, they have a risk shock and news shocks, which are modeled differently than other shocks. My problem is that the risk shock and news shocks don’t appear at all in the shock decomposition and it is not clear to me how to fix this.

For the variance of risk shocks, they estimate

“stdsigma1_p, INV_GAMMA2_PDF, 0.00233333333333, 0.00329983164554;”

instead of something like

“stderr e_sigma , INV_GAMMA2_PDF, 0.00233333333333, 0.00329983164554;” .

Here is the original CMR .mod file where you can see how they have modeled risk shocks and news shocks (lines 268-359).

cmr_model.mod (18.3 KB)

So I’m using this exact structure for the estimation and I have run estimations with both “stdsigma1_p” and “stderr e_sigma” and only either of them being there. In every case, posteriors indicate that the variance is relatively big but there is still nothing in shock_decomposition.

It would be very helpful if you had ideas about how to deal with the issue that the risk shocks are not there in the shock_decomposition. Thank you!