A few questions about dynare code?

General DSGE Modeling
1

Dear Professor Pfeifer，

I have been learning Dynare for some time and I encountered a few questions when writing the dynare code. Would you please help me out ? Any reply will be appreciated, thank you !

(1) It’s a question about the form of shock equation. In a non-linear model, when the code is written like “exp(var)” form, the shock equation is written as:
“exp(A_e) = 1 - rho_A_e * 1 + rho_A_e * exp(A_e(-1)) + e_A_e”,
but if the code is written like “var” form(without “exp” ), Does the shock equation should be written as:
“A_e = 1 - rho_A_e * 1 + rho_A_e * A_e(-1) + e_A_e” ?

(2) It’s also a question about shock equation. In a non-linear model, when the code is written like “var” form, Is there any chance that the shock equation is written as:
“log(A_e) = rho_A_e * log (A_e(-1)) + e_A_e” ?

(3) It’s a question about Bayesian estimation. Is there a normal range of the " initial value of the log posterior(or likelihood)" so that I could tell whether the code is right or not from the beginning of the estimation ?

Any reply will be appreciated, thanks again !

2
  1. Yes, the two specifications are consistent and can be used.
  2. That depends on how A_e enters the other model equations. It is common to have
Y=A*K^alpha*L^(1-alpha)
log(A)=rho_A*log (A(-1)) + e_A
  1. That depends very much on the data you use, but anything too big in absolute value, say bigger than 1e7 is highly suspicious.
3

Dear Professor Pfeifer，

Happy New Year and thank you for your reply, that helps a lot.

But as you said that it is common to have

‘‘Y=AK^alphaL^(1-alpha)
log(A)=rho_A*log (A(-1)) + e_A’’

Could the shock equation be written as:
“A = 1 - rho_A * 1 + rho_A * A(-1) + e_A”
in this situation ?

4

At order=1 the two specifications are identical.