Machine Learning for Recession Prediction and Dynamic Asset Allocation

Abstract

We introduce a novel application of support vector machines (SVM), an important machine learning algorithm, to determine the beginning and end of recessions in real time. Nowcasting, forecasting a condition in the present time because the full information will not be available until later, is key for recessions, which are only determined months after the fact. We show that SVM has excellent predictive performance for this task, capturing all six recessions from 1973 to 2018 and providing the signal with minimal delay. We take advantage of the timeliness of SVM signals to test dynamic asset allocation between stocks and bonds. A dynamic risk budgeting approach using SVM outputs appears superior to an equal-risk contribution portfolio, improving the average returns by 85 bps per annum without increased tail risk.

Keywords: Asset Allocation, Forecasting, Macroeconomics, Recessions, Machine Learning, Business Cycle, SVM

JEL Classification: C14, C53, E32, E37