Dear all,
I have just encountered the following issue that pertains to solving for a model’s steady state.
If I include my guesses and calculations for the steady state values in the initval block, dynare solves the model just fine. (See Bench_Old.mod)
If, however, I cut and paste the calculation of steady state guesses to _steadystate.m file, dynare responds with “The steadystate file did not compute the steady state.” Preceding this statement are the residuals of the static equations - I’m only providing a short excerpt:
Equation number 8 : 0 : mus2
Equation number 9 : 0.027437 : 9
Equation number 10 : -0.00047394 : 10
Equation number 11 : -0.070945 : y_out
Equation number 12 : -0.035545 : 12
Equation number 13 : 0.0091456 : 13
Equation number 14 : 0.0029947 : y_outs
Equation number 15 : 0 : yret
(See the pair Bench.mod and Bench_steadystate.m)
My initial guesses are not perfect: here’s an example of my guesses and/or calculations (that obtain from initval block and the _steadystate file) versus what dynare finally settles on as the true steady state values when I run Bench_Old.mod.
final answer initial guess
y_out 0.646149 0.5844
y_outs 3.80676 3.7866
gdp 0.60176 0.5501
gdps 3.80785 3.7879
expend 0.60176 0.6226
However, I thought the guesses would be good enough to get to the final answer using the _steadystate.m approach.
Obviously, for this particular model I can just work with the initval block version. However, I’m working on an extension which requires nonlinear equation solver to obtain some of the initial guesses and so I will need to rely on _steadystate.m route. With that in mind, why, theoretically, is dynare able to solve for the steady state one way and not the other? Is the former a superior approach? Is this indicative of multiple equilibria? Any insight would be greatly appreciated.
Thank you!
Bench_Old.mod (26.9 KB)
Bench.mod (21.9 KB)
Bench_steadystate.m (6.6 KB)