EABCN Training School: Large Dimensional VARs and Bayesian methods

General Description
 
We are pleased to announce the latest EABCN Training School; a three-day course entitled “Large Dimensional VARs and Bayesian Methods” taught by Todd Clark (Federal Reserve Bank of Cleveland).

It is primarily aimed at participants in the Euro Area Business Cycle Network, but applications will also be considered from doctoral students, post-doctoral researchers and economists working in central banks and government institutions outside of the network, as well as commercial organisations (fees are applicable for non-network non-academic organisations).

The course will teach Bayesian methods for the use of large vector autoregressions (VARs) in macroeconomic forecasting and structural analysis.  The morning lectures will focus on models and estimation methods.  The practice sessions will focus on hands-on illustrations with Matlab code.

Note that the course will assume basic familiarity with Bayesian methods and vector autoregressions. From that starting point, the course will focus on methods that have been developed for large models, along with their applications.

Tentative course outline
 
There will be about a 15-minute break between the lecture sessions on each day.
 
DAY 1
Lecture Session: 9:00 to 12:00 CET
Practical Session: 14:30 to 16:30 CET
 
Day 1 will begin with a brief overview of the essentials of Bayesian VARs (priors and algorithms) and then delve into key methods for estimating large BVARs, first in the homoscedastic case and then in the case of time-varying variances of innovations.
 
Session 1:  Large homoscedastic BVARs
Session 2:  Large BVARs with stochastic volatility
Practice Session:  Hands-on illustrations of the estimation of large BVARs
 

DAY 2
Lecture Session: 9:00 to 12:00 CET
Practical Session: 14:30 to 16:30 CET
 
Day 2 will focus on the key uses of large BVARs, in out-of-sample forecasting (both unconditional and conditional) and structural inference.
 
Session 1:  Use of large BVARs in out-of-sample forecasting
Session 2:  Use of large BVARs in structural inference
Practice Session:  Hands-on illustrations of out-of-sample forecasting and structural inference
 

DAY 3
Lecture Session: 9:00 to 12:00 CET
Practical Session: 14:30 to 16:30 CET
 
Day 3 will cover some of the important options for improving the large BVAR specifications and methods covered in Days 1 and 2, including computation, prior specifications, sensitivity to variable ordering that can arise with time-varying volatility, and accommodating temporary volatility spikes like those observed with COVID.
 
Session 1:  Options for improving computation and priors
Session 2:  Extensions of large BVARs
Practice Session:  Hands-on illustrations of options for speeding up computation, alternative priors, and order-invariant estimation

Administrative Information