Sigh...same tired old question about observation equations and models

[ChrisL]

I have a complicated NK model implemented into dynare in levels (no growth trend, no exp() function; it is as straughtforward and “vanilla” as possible). Steady-state (using initval) comes out perfect; hits all targets. I use stoch_simul command and this implies that IRFs and model output is in absolute deviation from steady-state.

1. If I use stoch_simul(…, loglinear, …) I’m assuming that this implies the model output and IRFs are now going to be in percentage deviation from steady-state?
2. Follow-up to number 1: If “YES” to number 1, that applies to all the variables in the model, including net (not gross) interest rates?
3. If this is the case and now I’d like to estimate the model using Bayesian methods, AND my data is defined in percentage deviation from steady-state, I need to leave the “loglinear” subcommand in the estimation command?

Sorry about this question, I think I’ve read every thread on this and Johannes’ guide, but I’m still paranoid I am doing something stupid. Or I am just stupid.

[jpfeifer]

1. Yes, the loglinear option divides all variables by their steady state value, expressing everything in percent from the base level.
2. Yes, that will apply to interest rates as well. If the model has a net interest rate, you will have percentage deviations from the net level, which is usually undesirable.
3. Yes, that is usually true. But generally, I would recommend appending logged auxiliary variables for the variables of interest instead of a blanket loglinear option.

[ChrisL]

Thanks, Johannes.

Is there any way to remedy #2 in the models equilibrium equations? Can I just multiply the IRF value for the interest rates by their steady state values? If so, is it possible to perform a similar conversion to the models simulation output somewhere so that the estimation makes sense, or is the model doomed to fail?

Also, can you expand a little bit on your explanation #3?

Thank you and sorry for the pestering.

[jpfeifer]

Have a look at

Question about understanding irfs in dynare

To be more precise: it depends on how the variables in your model are measured. As @stepan-a says, Dynare conducts a linearization. If you for example have a nonlinear model and your model variable is simply the level of output, the IRF will measure the difference between the variable and its steady state (or its mean at higher order). This is the case with https://github.com/DynareTeam/dynare/blob/master/examples/example1.mod In contrast, if your model variable already measures percentage…

[ChrisL]

I went through the model and changed the net interest rates into gross interest rates where it was obvious it could be done. Some parts of the model required a little more work to make it happen. It runs the exact same, and the steady-state target unaltered. All of the IRFs are as they were before, except for the interest rates. The values make a lot more sense now.

I’ll have to double-check the simulated moments to see if something similar happened, but I’m guessing given how identical the output looks.