Error message: The Jacobian contains Inf or NaN

Stochastic simulations
1

Dear all,

I modified Gali and Monacelli (2005)’ small open economy model to do the policy evaluation. Thus I write the model in levels. However, I got the error message"The Jacobian contains Inf or NaN". The code is attached. Anyone can help?

Thanks in advance.code.mod (2.5 KB)

2

Dear Grant,

Just remove the check command before the model_diagnostics command and you will obtain more informations:

MODEL_DIAGNOSTICS: The Jacobian of the static model contains Inf or NaN. The problem arises from: 


Derivative of Equation 3 with respect to Variable s  (initial value of s: 1) 
Derivative of Equation 8 with respect to Variable s  (initial value of s: 1) 

MODEL_DIAGNOSTICS:   The problem most often occurs, because a variable with
MODEL_DIAGNOSTICS:   exponent smaller than 1 has been initialized to 0. Taking the derivative
MODEL_DIAGNOSTICS:   and evaluating it at the steady state then results in a division by 0.
MODEL_DIAGNOSTICS:   If you are using model-local variables (# operator), check their values as well.

MODEL_DIAGNOSTICS: The Jacobian of the dynamic model contains Inf or NaN. The problem arises from: 


Derivative of Equation 3 with respect to  Variable s  (initial value of s: 1) 
Derivative of Equation 8 with respect to  Variable s  (initial value of s: 1) 

MODEL_DIAGNOSTICS:   The problem most often occurs, because a variable with
MODEL_DIAGNOSTICS:   exponent smaller than 1 has been initialized to 0. Taking the derivative
MODEL_DIAGNOSTICS:   and evaluating it at the steady state then results in a division by 0.
MODEL_DIAGNOSTICS:   If you are using model-local variables (# operator), check their values as well.

An obvious problem seems to be the value of eta (which should not be equal to one)… But they are other issues after.

Best,
Stéphane.

3

Dear Stéphane,

Thanks for your reply. It seems that the problem comes from the calibrated values. Am I right?

4

In a sense yes. You are trying to run the model in a limit case with eta=1. Mathematically, that is not a problem, but numerically, the computer does not know how to handle the limit where you have a variable with (s^0)^\infty

1 Like
5

At least the first problem… It is not possible to have eta equal to 1 because you have terms like 1/(1-eta). If you change this, you will fix the NaNs issue, but you will see that you have other issues with the model.

Best,
Stéphane.

6

If you really need to consider the limit case, you’ll have to do it analytically and change the equations accordingly (i.e. change a CES like production function for a Cobb Douglas). Unfortunately Dynare cannot do this for you…

Best,
Stéphane.

7

Hello everyone

I am having similar problem. In my case dynare is able to solve the model and generate irf, but when I am using model_diagnostics command then it is saying that:

MODEL_DIAGNOSTICS: The Jacobian of the static model contains Inf or NaN. The problem arises from: (%%%%)
MODEL_DIAGNOSTICS: The problem most often occurs, because a variable with
MODEL_DIAGNOSTICS: exponent smaller than 1 has been initialized to 0. Taking the derivative
MODEL_DIAGNOSTICS: and evaluating it at the steady state then results in a division by 0.
MODEL_DIAGNOSTICS: If you are using model-local variables (# operator), check their values as well.

I read that such problem may come due to small bug in dynare ( I am suing 4.5.7).
I am attaching the code. I am not sure if I am doing some mistake or it is due to some small bug.

I will be thankful if someone can help me in this regard.

model2.mod (9.1 KB) model2_ss.m (3.3 KB) model2_steadystate.m (308 Bytes) param.mat (2.8 KB) param2_ss.m (2.8 KB)

Thanks
Ankit

8

Please upgrade to 4.6.1. There the message is gone. Here is the steady state file you will need.
model2_steadystate.m (318 Bytes)

1 Like
9

Dear Prof. Johannes

Thank you for your reply. Code runs without any error now.
Ankit