Timing Issue in nonlinear pricing using Kimball Aggregator (Replication of Linde and Trabandt, 2018 JAE, Fiscal Multipliers) [Solved]

I am trying to replicate the model used by Jesper Linde and Mathias Trabandt in their Journal of Applied Econometrics Paper “Should We Use Linearized Models To Calculate Fiscal Multipliers” with Dynare. I am particularly interested in the recursiver formulation of the Kimball Aggregator.
However, I do have a timing issue as Dynare tells me that the Blanchard and Khan Conditions are not satisfied (the rank condition is not verified). The steady state computations, model diagnostics are working fine.

Maybe someone could help me out or give me a hint about my timing issue?


PS: I have attached my two mod files. The “Orig” corresponds to the baseline model used in the paper and in the “Simplified” version I got rid of both the fiscal variables and the flex-price economy. The timing issue is the same.
PS2: I can get a verified rank if I change e.g. “a”, i.e. the third variable used in the recursive formulation of the nonlinear pricing (and which is zero in the Dixit Stiglitz case), into a predetermined variable, but that doesn’t make sense to me.
PS3: Here are the mod files.
LindeTrabantSimplified.mod (7.6 KB)
LindeTrabantOrig.mod (12.5 KB)

Mathias Trabandt just replied to me with the solution… the parameter for the net markup was negative and the Kimball parameter positive. Switching the signs solved my issue.


How to caclu. the fiscal multipliers?


@Wenddy You need to @wmutschl. Also, please elaborate on your question.

The replication codes of the paper are on Trabandt’s Research page. The paper on page 19 also defines the way they compute the (marginal) fiscal multiplier at the zero lower bound. I hope that answers your question?

Thanks Mutschl,
I konwn that paper, but I didnot get the same results!
You defined that g_o_y is measured relative to Y , so multiplier =1/g_o_y, right?@wmutschl

I guess you should look into their replication codes which are also on Trabandt’s website. They look at the multiplier a little bit different than just 1/g_o_y.

Thanks you very much!