Replication for Jermann and Quadrini (2012)

General DSGE Modeling Replication Attempts
1

Dear All,

I am now replicating the NK model in Jermann and Quadrini (2012), using Johannes Pfeifer’s dynare code as reference. When I run my code, it keeps saying that “line 212, cols 1-0: syntax error, unexpected $end”. Could anyone explain to me where I went wrong? Thanks so much!
JQ2012.mod (6.81 KB)

2

Your line 120 reads

exp(x) = (exp(b(-1))/((1+exp(r(-1))))-exp(b)/(1+exp(r)))/*exp(y);
You cannot have

/*

at the end. It is not a valid operation

3

Thanks so much for your help! Since I could not find the steady state in NK model, I decided to start replicating the basic RBC model. The system says that “The steadystate file did not compute the steady state”, is that mean that I have problems in the initial state I put? The model part seems ok to me. Thanks so much for your concern.
rbc.mod (2.29 KB)

4

What are you doing? I provide full replication files at github.com/JohannesPfeifer/DSGE_mod/tree/master/Jermann_Quadrini_2012, including the _steady_state.m files that compute the steady state

5

Thanks so much! I have successfully code the RBC model! However I have problem coding NK model, the steady state occurs, but it says '“There are 15 eigenvalue(s) larger than 1 in modulus
for 14 forward-looking variable(s). The rank condition ISN’T verified!”

I used the same equation as you did in the tutorial file, so could you please notify me where I went wrong? Or is that possible to do the estimation without solving the trouble above? I am really new to Dynare and trying to practice as much as I could, please forgive my stupid questions. I am really appreciate for your help. The new code is att. Thanks!
jj111.mod (7.38 KB)

6

Dear Jpfeifer,

I figured out where I did wrong, and thanks for your code. I have one more question, since I use exp() form in my model, for those variables have steady state values equal zero and do not have percentage deviations, should I just leave it without putting exp()? Thanks for your concerns.

7

Yes, variables with mean 0 are usually linearized, not log-linearized (particularly as they often are already in percent)

8

Thanks so much! I am really appreciate for your help!

9

Dear Professor,
I looked at your replication code, Could you tell me why the value of labor is normalized to 0.3? I do not understand the reason why in this kind of DSGE model the value of output can be normalized.

10

Both labor and output are measured in arbitrary units. That allows you to conduct a normalization. After normalizing the non-sleeping hours to 1, you can calibrate the model to have agents work 30% of their time. Similarly, you can normalize TFP to get an output of 1 (unit).

11

Thank you very much for your answer, however, I do not understand the reason why it is possible to normalize both labor and prices. This double normalization should imply that there are two redundant optimality conditions, right?

12

No, because there are two parameters pinned down by this. Labor disutility and steady state TFP.