Page 1 of 1

Impossible to find steady state/model diagnostics

PostPosted: Tue Apr 14, 2015 11:36 am
by mrk
Hello,

I'm trying to run a simple Iacoviello (2005) model with borrowers and savers and a collateral constraint with housing. Every time I run the model I get the following error message:

Error using print_info (line 74)
Impossible to find the steady state. Either the model doesn't have a steady state,
there are an infinity of steady states, or the guess values are too far from the
solution
Error in steady (line 92)
print_info(info,options_.noprint, options_);
Error in Iaco (line 217)
steady;
Error in dynare (line 180)
evalin('base',fname) ;


I know this is a very common problem for dynare rookies but I cannot find where my error is. I have tried the model diagnostics command but this returns the following message:

SOLVE: maxit has been reached
model diagnostic can't obtain the steady state


I am certain that there is a mistake in my model, is there a way of identifying where it is?

Any additional help would be much appreciated.

Thanks

Iaco.mod
(1.86 KiB) Downloaded 117 times

Re: Impossible to find steady state/model diagnostics

PostPosted: Tue Apr 14, 2015 2:07 pm
by jpfeifer
Increase
Code: Select all
maxit

in the steady command. The unstable version, where options_.maxit=50, finds
STEADY-STATE RESULTS:

cc -6.0298
cu 3.98492
wc -5.98856
wu 3.92978
lc 4.12366
lu -5.51468
r 0.0295536
dp 0.0197042
a 0
x 0.180025
y -2.04488
hc -10.0937
hu -4.13417e-05
q 6.28755
lam 1.42463
bc 2.7349
bu 2.7349

This does not look correct. Thus, your model equations seem to be incorrect. Looking at
Code: Select all
1/exp(cu)=BETA*exp(r-dp(+1)-cu(+1));

//Eq.(2) Labour supply savers log_linearised
wu=(ETA-1)*lu+cu ;

You seem to denote with cu both the log consumption level as well as the percentage deviation of consumption from its steady state. Those are two different things as the latter is cu-steady_state(cu).

Re: Impossible to find steady state/model diagnostics

PostPosted: Thu Apr 16, 2015 11:04 am
by mrk
Thank you for the reply.

After rewritting all equations in their most basic form (not log linearised) I discovered that the last equation of the model was completely wrong.

Thanks again for your help.