Working Paper Revision

Real-Time Forecasting and Scenario Analysis using a Large Mixed-Frequency Bayesian VAR

Abstract: We use a mixed-frequency vector autoregression to obtain intraquarter point and density forecasts as new, high frequency information becomes available. This model, delineated in Ghysels (2016), is specified at the lowest sampling frequency; high frequency observations are treated as different economic series occurring at the low frequency. As this type of data stacking results in a high-dimensional system, we rely on Bayesian shrinkage to mitigate parameter proliferation. We obtain high-frequency updates to forecasts by treating new data releases as conditioning information. The same framework is used for scenario analysis to obtain forecasts conditional on a hypothetical future path of the variables in the system. We show that the methodology results in competitive point and density forecasts and illustrate the usefulness of the methodology by providing forecasts of real GDP growth given hypothetical paths of a central bank policy rate.

Keywords: Stacked vector autoregression; Mixed-frequency estimation; Vector autoregression; Bayesian methods; Forecasting; Nowcasting; conditional forecasts;

JEL Classification: C22; C52; C53;

https://doi.org/10.20955/wp.2015.030

Status: Published in International Journal of Central Banking

Bibliographic Information

Provider: Federal Reserve Bank of St. Louis

Part of Series: Working Papers

Publication Date: 2020-04-10

Number: 2015-030

Note: Publisher URL: https://www.ijcb.org/journal/ijcb21q5a8.pdf

Related Works

• Working Paper Revision: Real-Time Forecasting and Scenario Analysis using a Large Mixed-Frequency Bayesian VAR