Pandemic Priors BVAR

Pandemic Priors: a simple, easy, and flexible way of estimating Bayesian VARs taking into consideration the pandemic period, as a Minnesota prior with time dummies

Update (Feb 2024): 

• New fully revised version of the paper, including forecast-based optimal priors and out-of-sample comparison of alternative methods.  Link here!

• Added Pandemic Priors extensions of the Giannone, Lenza, and Primiceri (2015) optimal priors (here) and the Chan (2022) asymmetric conjugate priors (here). 

Update (Feb 2023): Inclusion of a test for the optimal level of shrinkage for the pandemic period and a test for suitability of the Pandemic Priors

Update (Nov 2022): Added the flexibility of how much signal to take from pandemic period

Paper: Pandemic Priors

Abstract:

The onset of the COVID-19 pandemic and the Great Lockdown caused macroeconomic variables to display complex patterns that hardly follow any historical behavior. In the context of Bayesian vector autoregressions, an off-the-shelf exercise demonstrates how a very low number of extreme pandemic observations distort the estimated persistence of the variables, affecting forecasts and giving a myopic view of the economic effects after a structural shock.  I propose an easy and straightforward solution to deal with these extreme episodes by extending the Minnesota Prior with time dummies, which can be adapted to any conventional or state-of-the-art structural identification procedure. The method optimally defines the level at which the pandemic observations are downplayed, nesting the boundary cases of an uninformative prior that soaks all the variance and a traditional Minnesota Prior. The Pandemic Priors succeed in recovering historical relationships, properly identifying structural shocks, and providing more accurate forecasts than alternative methods.