steady state contains NAN or Inf

[shartjen]

Hi,

I am on my first project with dynare. I have modified the Smets and Wouters (2007) model (SW07) with endogenous growth. My code seemed to work fine but now I want to add a variable to measure persistent effects on output and end up getting the Error: "The steady state contains NaN or Inf." when trying to go for stoch_simul. I could narrow down the problem to two lines of code:

Code: Select all

dy=y-y(-1)+gamma(-1)+ctrend;
YL =  y-y(-1) + YL(-1); //+gamma(-1);


The dy is the measurement equation from SW07 for later estimation. The YL is the persistent effect I want to measure. Later I would add gamma(-1) to measure the cumulative effect in from gamma. But to make sure it is not the errors cause I commented it out and leave only the first part which should just equal YL = y in a complicated way. When I leave either line out the code works. If both lines are in I get the error.
I put in resid(1) to check and the dy-equation was not zero but equalled ctrend, which never was a problem before.
Why do these two lines interfere? Thanx for any help.

P.S.:
When making YL a trending variable with a deflator very close to one the problem disappears.

Code: Select all

trend_var (growth_factor = cgamma^0.001) TFP;
var (deflator=TFP) YL;


cgamma = 1.0043

[shartjen]

Someone else was already able to help me. Setting initival Values did the job, though I still dont really know why.

Code: Select all

initval;
dy=ctrend;
end;


As ctrend is trend-growth in this model and thus the intercept of the otherwise stationary variable dy.

[jpfeifer]

Unless you provide initial values in the initval block, Dynare takes them to be 0. But dy does not have a steady state of 0, but of the steady state growth rate ctrend. Hence, you should provide this value as the initial value.