Different result with different identification criteria

Identification/Sensitivity
1

Dear all,
I am estimating multisector NK model, and I did identification with ‘identification’ command before estimation.

I got different result with different identification criteria.
For criteria of “REDUCED-FORM”, “MINIMAL SYSTEM (KOMUNJER AND NG, 2011)”, and “MOMENTS (ISKREV, 2010)”, I got all parameters identified.
For “SPECTRUM (QU AND TKACHENKO, 2012)”, on the other hand, some of my parameters are not identified.
Whole messages from the identification are as below.

======== Identification Analysis ========
Testing prior mean
Evaluating simulated moment uncertainty … please wait
Doing 3870 replicas of length 300 periods.
Simulated moment uncertainty … done!
Note that differences in the criteria could be due to numerical settings,
numerical errors or the method used to find problematic parameter sets.
Settings:
Derivation mode for Jacobians: Analytic using sylvester equations
Method to find problematic parameters: Nullspace and multicorrelation coefficients
Normalize Jacobians: Yes
Tolerance level for rank computations: robust
Tolerance level for selecting nonzero columns: 1e-08
Tolerance level for selecting nonzero singular values: 1e-03
REDUCED-FORM:
All parameters are identified in the Jacobian of steady state and reduced-form solution matrices (rank(Tau) is full with tol = robust).
MINIMAL SYSTEM (KOMUNJER AND NG, 2011):
All parameters are identified in the Jacobian of steady state and minimal system (rank(Deltabar) is full with tol = robust).
SPECTRUM (QU AND TKACHENKO, 2012):
!!!WARNING!!!
The rank of Gbar (Jacobian of mean and spectrum) is deficient!
SE_vmu2 is not identified!
SE_vmu4 is not identified!
SE_vmu11 is not identified!
SE_vmu12 is not identified!
rhomu1 is not identified!
rhomu2 is not identified!
rhomu3 is not identified!
rhomu4 is not identified!
rhomu5 is not identified!
rhomu6 is not identified!
rhomu7 is not identified!
rhomu8 is not identified!
rhomu9 is not identified!
rhomu10 is not identified!
rhomu11 is not identified!
rhomu12 is not identified!
rhomu13 is not identified!
rhomu14 is not identified!
rhomu15 is not identified!
rhomu16 is not identified!
MOMENTS (ISKREV, 2010):
All parameters are identified in the Jacobian of first two moments (rank(J) is full with tol = robust).

In this case, can I trust whether my parameters are all identified?
I attached the result .txt log file and graphs for identification (and also .mod file).

Thank you for your help!

q_identifcation.zip (649.7 KB)
mod.zip (27.2 KB)

2

My experience is that the Iskrev ones are the most reliable (particularly given that the default options choices are geared towards it).

3

Professor, thank you for your response.
I got some clarifying questions.

By “Iskrev ones” you mean the Iskrev(2010) and reduced-form solution from Ratto and Iskrev(2011)?

In what aspects are they considered the most reliable?
Do you mean that the basic algorithm of identification is set up to be most suitable for these two?

4

Yes, I mean

  1. I mean that the default settings
Settings:
Derivation mode for Jacobians: Analytic using sylvester equations
Method to find problematic parameters: Nullspace and multicorrelation coefficients
Normalize Jacobians: Yes
Tolerance level for rank computations: robust
Tolerance level for selecting nonzero columns: 1e-08
Tolerance level for selecting nonzero singular values: 1e-03

are geared towards the moments check, not the Spectrum. Thus, the spectrum check tends to create false positives, i.e. treating too many small values as being singular.