CMM(2011) - generalized Schur (QZ) failing

[mfarkas]

Dear all,

we are trying to replicate the medium scale NK model of Christensen, Meh and Moran (2011). Please if interested, find it here: https://ideas.repec.org/p/bca/bocawp/11-32.html

We departed from the NK_MM2010 replication file from the MMB 2.0 (BankCapital.mod) that is a simplified version of the model. After realizing we needed to target values from the original paper to reduce the dimensionality, we could finally solve for the the steady state.

Everything seems to be , as it should be, yet we still get the following error message:

Error using print_info (line 36)
The generalized Schur (QZ) decomposition failed. For more information, see the
documentation for Lapack function dgges: info=23, n=23

We have consulted the previous entries on the error, where Dr. Johannes Pfeifer recommended to run the model_diagnostics(M_,options_,oo_).
For that we get the following error:

Error using svd
Input to SVD must not contain NaN or Inf.

Searching for this error message provided us with possible sources, like some variables being not initialized before estimation.
Therefore we have simply took all endogenous variables and initialized them at their steady state values.


What other variables could be not initialized or left to be equal zero?
If we could get the model_diagnostics running, that would help us great time to solve the issue, and estimate the model.

Could you please provide us comments, what could be going wrong and what is needed to circumvent it? Thank you very much!

Sincerely,
Mátyás Farkas

[jpfeifer]

It comes from

Code: Select all

alpha = 1 - (1-alpha_ss) - ((I-bigN)/Y - (steady_state(I)-steady_state(bigN))/steady_state(Y)) ^ varsigma ; %Endogenous riskiness of the banking sector


The bracket raised to the power of varsigma evaluates to 0, resulting in division by 0.