Milan Szabo
The paper proposes a novel application of Bayesian quantile regression to forecast a full distribution of macroeconomic variables that can be linked to, for example, an official projection of the variable published by a central bank, or a forecast from a survey of professional forecasters. The approach is employed to estimate the popular Growth-at-Risk, which maps current financial and economic conditions to the distribution of future GDP growth, focusing mainly on downside risks. The results show that the linkage improves distribution forecasting and, thanks to the additional information obtained from the linkage, reduces overfitting and makes Growth-at-Risk models more operational for countries with short time series. Additional improvements in consistency around the official projection enhance the credibility of the results when communicated by the central bank. The method can also be used to derive asymmetric fan charts around the official projection not only for real GDP growth as examined in the paper, but also for unemployment or inflation.
JEL codes: C53, E27, E32, E44
Keywords: Downside risk, fan charts, growth-at-risk, quantile regression
Issued: November 2020
Download: CNB WP No. 3/2020 (pdf, 1.9 MB)