Financial frictions and Steady-State problem

[ohouari]

Dear all,

I would really appreciate any help or indication.

As part of my Master thesis I am trying to achieve a DSGE model to analyze the behaviour of entrepreneurs under different uncertainties (micro/macro) by developing an optimal contract (in the spirit of financial acceleration à la BGG 1999) with risk averse agents (Both households and entrepreneurs like in Mikhail Dmitriev 2015: Risk Aversion and the Financial Accelerator) and imperfect protection of investors (banks).

To determine the impact of investors imperfect protection parameter (eta) in the financial acceleration equation (equation 102 in the .mod file) I must calculate both elasticities nup and nusigma (values used in the .mod code are issued from Dmitriev 2015 just to run the model) which is what I have tried to do in the two .m files. The results (of nu_p and nu_sigma) depends on the steady state values of variables x_low, x_up, R, s. These variables determine the repayment and default rule of entrepreneurs.

I understand that I have to specify a separate steady_state.m file. The problem I have is that these variables are not present in my log-linearized model which means that they will not be reported and therefore not calculated.

I dont understand how to solve this problem or what I'm missing.

Again I would appreciate any help.

Best regards,
Houari

[jpfeifer]

First off, I don't think it is a good idea to work with this type of complicated model at the master's level. It hardly ever works out in the end. You may be the exception, but most people should start with simpler models (fewer than 50 equations).

You need to understand the calibration strategy of the model. If the objects you refer to do not appear in the model (and this was calculated correctly), then they are only used during calibration to determine the steady state levels of some endogenous objects in the model. Note that in linearized models the steady state of the original variables is required to compute parameters showing up. This might be the case here. It could be that you need to solve the nonlinear model in your steady state file to get the steady state values you need to set your elasticities.

[ohouari]

Dear professor Johannes Pfeifer,

First, I would thank you very much for your response and explanation of the calibration strategy and of course for all the great work you do here (and all the dynare team).

Second, my complete non linear model have just 25 equations and the log-linearized one only 18 equations as you can see in .pdf file I attached here (the number 102 in the last post and in the code was reffering to my demonstration file).

Moreover, when you say:" It could be that you need to solve the nonlinear model in your steady state file to get the steady state values you need to set your elasticities."I think that is axactly what I shoud do and I was thinking about it this last days. I try to follow the NK_baseline.m method and Fernandez-Villaverdé (2006) argumentation but since I have a lot of non-linearities (specialy integrals in g anf f functions) I dont know how to write this . I'm trying to define integrals like functions and use fsolve and this seems not working. Do you think this is the correct method to use? And can I introduce in this steady_state.m file my two elasticities nup and nusigma calculated in the previous .m files?

Thank you for your response.

Best regards,
HOUARI

[jpfeifer]

You might want to have a look at Christiano/Motto/Rostagno (2014): Risk Shocks. In their replication files at the AER they use a steady state file to deal with the financial accelerator.