CMStatistics 2022: Start Registration
View Submission - CFE
Title: Predictive performance of Bayesian VEC-SV-GARCH models before and during the Covid-19 pandemic Authors:  Lukasz Kwiatkowski - Krakow University of Economics (Poland) [presenting]
Justyna Wroblewska - Krakow University of Economics (Poland)
Anna Pajor - Krakow University of Economics (Poland)
Abstract: The main aim is to check whether taking into account long-term relations in heteroscedastic VAR models affects their predictive performance over the period of the pandemic. Additionally, we check whether updating the posterior upon the arrival of new observations affects the predictive performance of the models before and during the pandemic. In the empirical analysis, the so-called small model of monetary policy is considered separately for five economies: the United States, the United Kingdom, the Euro Area, Poland and Hungary. The heteroscedasticity is captured by means of hybrid specifications combining stochastic volatility and GARCH (SV-GARCH), or some of the special cases thereof. Estimation and prediction are performed within the Bayesian approach, with a focus on the evaluation and comparison of the models' predictive performance by means of predictive likelihood. The results indicate that allowing for conditional heteroskedasticity enhances the VEC models' predictive performance both before and during the pandemic. Also, in most cases, not updating the posterior decreases the predictive ability of models before as well as during the pandemic. For most developed economies, including long-term relations does not improve forecasts for long horizons over the pandemic. However, in most cases incorporating long-term relations improves forecasts for shorter horizons over the pandemic.