Machine Learning Modeling on Mixed-frequency Data for Financial Growth at Risk

Abstract

Determination of macroeconomic policies in real-time requires assessing the correct information regarding current economic conditions. This statement spurred researchers to develop methods involving high-frequency data for risk analysis. This paper extends the quarterly growth-at-risk (GaR) approach by involving a machine-learning approach based on the Mixed-Frequency Data Sampling Quantile Regression Neural Network (MIDAS-QRNN) model. This paper shows that the MIDAS-QRNN model has the best prediction accuracy and can show good PDB nowcasting. The monthly financial GaR can detect unusual economic growth movements during the COVID-19 pandemic.