I have done the Bayesian estimation of a medium scale DSGE model through Dynare and the results with the respective bayesian impulse response functions are attached. Please guide me if these make some economic sense. The thing which I wanted to see was the impact of fiscal shocks on output Y and debt/gdp ration b. I will be grateful for the valuable insight on the results.estimation outcome.pdf (1023.5 KB)
There is nothing immediately suspicious, but you did not provide anything regarding the properties of the MCMC and the mode-finding like mode_check-plots, trace_plots and stuff like that.
Thank you so much for the guidance, I have attached the mode_check-plots, trace_plot of the parameter in which we are interested and variance decomposition of the variable (debt/gdp ratio) which is the point of focus in the study. please find the attachment and guide about the areas which need adjustments. I will be grateful for the guidance.model outcome.pdf (662.3 KB)
- Some parameters seem not to be identified, i.e. there is a horizontal likelihood.
- You seem to be using too few draws
- It seems your observation equations are wrong. Your smoothed variable features a constant that should not be there.
Thank you so much for the guidance. I think the problem is with these equations. I will be grateful for help in this regard.
nr = kp(-1) + u + rk -w;
// (11) Regular labour demand;
ns = kp(-1) + u + rk - w0;
// (12) irregular labour demand;
n = etanrnr0_n0 + (1-eta)nsns0_n0; % here eta=0.80 share of regular labour force in economy
// (13) labour demand;
nr0 = 0.3335;
ns0 = 0.0700; % computed on the sample mean of labour force and multiplying it with the share of this type of force in the economy.
k0 = 0.32kp0 + 0.10kg0; % capital at steady state to get the steady state value of labour share in the economy kp=private capital, kg=public capital both have same depriciation rate.
Without context, it is impossible to know where these equations appear and what they mean. But generally, having fixed numbers appear is unusual.
Thank you so much for the direction, these equation are in the production/firms segment of the DSGE model and the steady state values are sample means and they are in the steady state block of the same model. I have attached my data file and mod file for the reference. please find the attachment and I am grateful for guidance, help and your time.bayest1.zip (85.1 KB)
- I still think you are not handling parameter dependence correctly. You should use model-local variables.
- Looking at the data series, I think I spot a seasonal pattern. You need to make sure your data is seasonally adjusted.
Thank you so much for the guidance, I will adjust the seasonal effect in the data series. And I will be grateful if you can share some example for specification of model-local variable in the DSGE models.
Thank you so much for your kind guidance, time and help. I will study the suggested information package and try to improve these aspects of the model. Have a great time.