Forecasting Labour Market Outcomes Using Bayesian Hierarchical Panel VARs

Bayesian methods are being increasingly used by international organizations such as the UN, ECB, and The Fed System, in the context of forecasting. In this seminar, Dr Tomasz Wozniak from the University of Melbourne will present progress and extensions of the baseline Bayesian hierarchical dynamic panel data model he developed in 2024 in collaboration with Research colleagues. The main features of the existing framework will be presented, while key extensions such as the use of groupings for parameter estimation and the advantages of the Bayesian approach for the treatment of missing observations will also be discussed. The accompanying R package bpvars, built with a view to allow for seamless solution of the model for ILO practitioners, will also be introduced and explained. The presentation will also include example forecasts for selected countries.

Dr. Tomasz Wozniak is an econometrician specialized in methods for empirical macroeconomic analysis. He works on new statistical models to describe and measure the economy and the effects of government policy decisions. Specifically, his research includes the application of Granger causality principles to analyse the volatility of financial asset returns and macroeconomic aggregates with nonlinear dynamics as well the investigation of novel empirical methods to compare the impact of alternative monetary policies represented as structural time series models.

He works as a Senior Lecturer at the Department of Economics of the University of Melbourne, and is co-founder of the Bayesian Analysis and Modelling Research Group and co-organizer of the annual Melbourne Bayesian Econometrics Workshop. Among others, he is the author of R packages bsvars, bsvarSIGNs and bpvars for structural and predictive analyses using Bayesian Structural Vector Autoregressions