Time-Varying VAR Model for Uncertainty Shocks

Hello everyone. I hope you are doing well.

I am sorry as I am going to ask this question which is not necessarily related to Dynare or DSGE modeling. I am trying to study the impacts of uncertainty shocks using the TVP-VAR model with stochastic volatility. For that purpose, I borrow codes from Professor Mumtaz’s website. I have attached the code. My main confusion with the use of the model to study uncertainty shocks comes from my inability to understand the following point:
a. Do we need to use a proxy of uncertainty like VXO or EPUI in order to study uncertainty shocks? As far as I understand, time-varying models themselves calculate volatility time-series based on the data we provide. For instance, Professor Mumtaz calculates volatility from the model in this paper: Dynamic Impacts of Uncertainty Shocks in the US Economy.
However, I also found some papers like this one: Time Varying Uncertainty Impacts on Chinese Economy where the authors use uncertainty proxies to study uncertainty impacts on the Chinese economy.

I will be more than happy if you could help me overcome this confusion on the basic question.

Thank you very much for your time and help.

TVPVARR.zip (54.8 KB)

There are essentially two approaches out there in the literature:

  1. You can can run a VAR with an uncertainty proxy. In that case, uncertainty shocks are identified as shocks to the level of the proxy. The VAR typically assumes homoskedasticity in this case or at least no shocks to structural error terms.
  2. You can run a VAR without an uncertainty proxy, but identify uncertainty shocks as shocks to the variance of the structural error terms. In that case, the model features heteroskedasticity.

Thank you, professor, for clearing my confusion.