"The steady state contains NaN or Inf "


#1

I have the following codes for the below system of equations which is hugely non-linear and I cannot do the linearization by hand. I did the codes in dynare in two versions. One using the exp(var) for the variables and the other one just by same equations.

However, i get always that “The steady state contains NaN or Inf” what would be the problem?
(The upload of the problem is below)

Best,

ex4.mod (1.9 KB)

ex5.mod (2.0 KB)


#2

The upload of the problem:

the problem:
project (2).pdf (135.2 KB)
There is an integral in the model I could not write it and calculate in the dynare. so just for making it compatible and have a solution at least I assumed it as a parameter in my codes. I would be glad if someone also help me for that.


#3

Please do not prematurely do an exp()-substitution. Use the model in levels and provide sensible initial values. You use 0 for most initval, but that is not allowed for e.g. capital.
Also, recheck all your equations. Steady state finding has a problem with b and k in particular. Check whether the equations containing those are correct.


#4

Thanks for response.
I recheck my equations, and also simplified in terms of notations, ( I also attached the equations as well)

I have no clue for my initial values. I tried several initval but none of them changed the final answer to be different from NaN. Do you have any suggestions.
exx.mod (1.4 KB)

project (8).pdf (141.8 KB)


#5

Initialize ALL variables. If you put

resid;

before steady, you will see that the starting values for various equations are still invalid.


#6

Could you elaborate on that? what does the command “resid” do?


#7

See the Dynare manual.


#8

Hi everyone,
It seem`s like I have the same problem:
I can’t count initial values by hand and ask Dynare to do this, and get an error:
Error using print_info (line 90)
The steady state contains NaN or Inf
What should I do with this?
thanks
HA1.mod (1.6 KB)


#9

Your initval block uses an implicit equation:

N=exp((1-N)/N*theta/(1+phi)-(1-theta)/(1-alpha)*log((1-theta)/theta*(1+gamma)/(1+gamma-beta)));

N appears both on the left and right of the equal sign. That cannot work. Either the equation is wrong or you need to call a solver.