Recession Risk Prediction with Machine Learning and Big Panel Data

The machine learning models have been considered a good choice for forecast recession, especially with multiple variables. In this paper, we compare the forecast ability of the machine learning models and traditional methods in the recent 5 years and focus on the recession forecast in two situations, the 2008 financial crisis for the U.S. and the 2011 Italy sovereign crisis. We find that some machine learning models perform well in the forecast of the trend of the GDP growth rate. For the recession forecast, the best models are different for different situations maybe because of the different reasons causing the recessions. In the recent 5 years, the recession-related to COVID-19 reflects some special and unprecedented. We may need to research the impact of it in some specific fields.