Puzzles about replicating Gertler-Kiyotaki-Queralto 2012 using stochastic steady state

Dear Dynare team,

I recently tried to replicate the Gertler-Kiyotaki-Queralto (2012) article using Basu/Bundick 2017 method recommended by Prof. Pfeifer at forum given. In the case of a macroprudential policy, I got an impulse response similar to the original paper, but in the absence of a macroprudential policy (the only difference is the value of parameter tau0=0), the results of the “x” (equity ratio) and “mus” (excess value of bank assets) variables are opposite to the original. How can I fix the problem?

Attached are my mod files, and the paper.


GKQ2012_SSS.mod (11.4 KB)
nopolicy_GKQ2012.mod (11.4 KB)
Gertler-Kiyotaki-Queralto 2012.pdf (402.2 KB)

1 Like

I have tried my best to debug the program from “Timing” and “Expectation of function,vs, function of expectations” but still can’t find the exact causes. I was wondering if anyone can help me debug the program.

I quickly skimmed the original article. Their method is different from the Basu/Bundick one. This will be very challenging to replicate. Have you asked the authors for codes?

Thank you Prof. Pfeifer for your reply. I tried to ask the author for the code but didn’t get a reply.

  1. Professor Pfeifer, you are right, the two methods are different. GKQ(2012) using iteration method, but the differences of the concept “the risky steady state in GKQ(2012)” and “stochastic steady state in Basu/Bundick 2017” are really confused me. According to Groot (2013) and Groot et al.(2018), these two concepts are similar to me. How should I understand this problem?

  2. When I use BB2017 method, the irf of variable “mus” is opposite with the first order approximation around deterministic steady state methods, which also makes me confused. Am I having a problem with the timing understanding in this model?

Thank you very much!

nopolicy_GKQ2012_firstorder.mod (4.7 KB)

In principle they should yield similar results, but this is not assured. Also, pruning can make a difference here. Without digging very deeply into this, it is hard to tell what is going on. So I am afraid, I cannot help you here.

Thank you prof.Pfeifer.

Hi @June, thanks for sharing your experience on replicating the paper. Just to follow up on the thread, have you figured out the issue?

Indeed, in the IRFs under the 1st order approximation, I can’t replicate the sign for “x”. It should be negative according to the paper, but the result is positive (very close to zero though).

Anyone figured out this issue? Is this something I can only see in the 2nd order approximation? I didn’t find any error (e.g. timing) in @June’s code above nopolicy_GKQ2012_firstorder.mod.

Here I am attaching his code with a slight modification: stoch_simul to the first order, shocks to be iid and variables in the same order as in the paper.

nopolicy_GKQ2012_firstorder_2.mod (4.8 KB)