Forecasting US GDP growth rates in a rich environment of macroeconomic data

Forecasting GDP growth rates is a formidable challenge, compounded by the inherent volatility, the complexity of the economic landscape, and the presence of a multitude of economic indicators at varying data frequencies. This study employs the MIDAS-LASSO model, which represents a penalized approach designed for mixed-frequency data, to forecast US GDP growth rates, while considering a vast array of macroeconomic indicators, including the Macroeconomic Attention Index (MAI) of Fisher et al. (2022). The empirical analysis demonstrates that both macroeconomic indicators and MAI exhibit considerable power for forecasting US GDP growth rates. The MIDAS-LASSO model outperforms its competitors in terms of forecasting efficacy, particularly in scenarios involving a plethora of predictors. Further analysis scrutinizes the model's efficacy across business cycles and during significant economic downturns, and the pathways through which macroeconomic risks influence US GDP growth rates. These insights offer valuable contributions to the field of economic forecasting and present novel avenues for policymakers and analysts.