On Business-Cycle Forecasting

An accurate forecasting for business cycles is important for economic agents who make decisions based on what they expect about the future state of the economy. Unfortunately, the performance of economic forecasting from both economists and industrial practitioners has been limited to date. The unsatisfactory track record is largely due to a restrictive definition of targeted variable, inflexible modeling specifications, and limited data utilization. This project aims to improve the performance for recession forecasting. It uses a less restrictive targeted variable and a more flexible and inclusive modeling framework to address the issues simultaneously. The proposed targeted variable can better capture a future recession and help economic agents in making relevant decisions. The unified modeling framework applies more flexible model specifications, extracts rich information from many economic and financial variables, and incorporates data with mixed frequencies. The project investigates whether the proposed model can outperform the key existing models in predicting recessions by exhibiting lower forecast errors and generating better early signals for the turning points of business cycles. These findings have significant implication for households, entrepreneurs, market practitioners, and policy makers who can use the proposed model to make a better assessment on the economic outlook, and therefore make timely decisions and policies accordingly.