Caught Kord exception

[nastforest]

Dear Johannes,
Could you please help?
The error that I get after trying to run stochastic simulation (up to the 3rd order) of my model is:

"There are 10 eigenvalue(s) larger than 1 in modulus
for 9 forward-looking variable(s)

The rank condition ISN'T verified!

dynare:k_order_perturbation: Caught Kord exception: NaN or Inf asserted in first order derivatives in FirstOrder::solve
Error using print_info (line 68)
k_order_pert was unable to compute the solution

Error in stoch_simul (line 98)
print_info(info, options_.noprint, options_);

Error in BGG_mybank_timingRA (line 478)
info = stoch_simul(var_list_);

Error in dynare (line 180)
evalin('base',fname) ;"

Could you please tell me what might be a possible source of the problem? What could I look at to solve it? (Model_diagnostics command does not return any problem.)
I read elsewhere in the forum that this could be due to singularity of the first order approximation. What does singularity mean in this case? Is it that something is wrong in the model equations? How can the singularity be detected?
I know that the problem is most possibly with the Euler equation. Interestingly, when I change the value of one parameter values (which is a part of the Euler equation), the problem disappears.
Thanks in advance!

[jpfeifer]

Before going to third order, please make sure your model runs at first order. Usually, there is a timing mistake involved (or a parameter problem).

[nastforest]

Indeed, am not able to run the model at first order:
"There are 10 eigenvalue(s) larger than 1 in modulus
for 9 forward-looking variable(s)

The rank condition ISN'T verified!

Error using print_info (line 42)
Blanchard Kahn conditions are not satisfied: no stable equilibrium"

Given that changing one parameter value allows to circumvent the problem, does it mean that it is not a timing issue (because the same timing allows to run the model with a different parameter value)?
And, when you say 'a parameter problem', what exactly do you mean? How could I check what values of parameters are compatible with my model and which are not?

Thanks a lot in advance!

[jpfeifer]

There are two (not necessarily exclusive) problems that can be the cause.
1. There is a timing problem and changing the parameter value allows you to override this problem by introducing a second error that cancels the effect of the first one on the BK conditions but still gives wrong results.
2. The timing is correct, but the parameter value chosen is not compatible with the BK conditions. In this case, altering the parameter values is the correct way to go.

According to my experience, it is 1) most of the time. A good indication might be checking whether the parameter values are common given literature. If you e.g. need to set beta>1 to get the BK conditions to hold, it is not the parameter values that cause the problem.

[nastforest]

The value of the parameter that I change in order be able to run the model is risk-aversion that is rather flexible (when it is positive) as far as I know. In particular, with kappa=1 I can run the model, while with kappa=2 I cannot. What would be the way to check that the value of risk-aversion=2 is incompatible with BK conditions (in case I have a parameter problem)?
I have been trying to find a timing problem in my equilibrium conditions, but wasn't able to find it so far.
Could you may be look at my .mod file to help me spot the errors?
Thank you a lot in advance!

[jpfeifer]

This is highly unusual, because the risk aversion should usually not affect the stability and uniqueness of the model (unless you account for growth explicitly where the discount factor might depend on it). Therefore, I guess there is still a problem with your Euler equations (which is the only place where risk aversion shows up)