A New Approach to Forecasting the Probability of Recessions after the COVID‐19 Pandemic*

Abstract

Standard recession forecasting based on economic indicators has become unsettled due to COVID‐19 pandemic's limited but influential data. This paper proposes a new non‐parametric approach to computing predictive probabilities of future recessions that is robust to influential observations and other data irregularities. The method simulates forecasts using past data histories embedded into a symbolic space. Then, the forecasts are converted into probability statements, which are weighted by the forecast probabilities of their respective symbols. Using GDP data from G7, our proposal outperforms other parametric approaches in classifying future national business cycle phases, especially including data from 2020 in the sample.