Use annual data to estimate a DSGE model with Bayesian technique

Hi everyone

Can someone help me a question that can we use annual data to estimate a DSGE model with Bayesian technique?

Thank you so much

You need to adjust the frequency of your model accordingly (e.g. annual instead of quarterly discount factor), but there is generally no problem with using annual data.

Thank you so much Prof. Pfeifer

So if my sample covers only for 34 years, let’say 34 obs, is this sample size enough to estimate model parameters with Bayesian Technique? and what about the shock process, I mean should I use the AR(1) or only White Noise process to model the structural disturbances?
If the AR(1) process, for example, cost-push shock: cp(t) = Rho * cp(t-1) + epsilon(t)
It means that cost-push shock of this year depends on last year, is this AR(1) process reasonable, since it has a 1-year-lag effect rather than 1-quarter-lag effect as current literature on estimated DSGE model with quarterly data

Please see How to determine priors?
The AR question is tricky. If your true data generating process in quarterly data is an AR1, you would get a VARMA process at annual frequency. That being said, using an AR1 is still standard.

Hi Johannes, about adjusting the frequency of a model, do you know good sources where we can read more? Thank you!!

What exactly are you looking for? I cover some of this in my Guide to Observation Equations. You may also find the Appendix to our 2014 AER comment useful.