Forecasting National Recessions of the United States with State-level Climate Risks: Evidence from Model Averaging in Markov-switching Models

This paper utilizes Bayesian (static) model averaging (BMA) and dynamic model averaging (DMA) incorporated into Markov-switching (MS) models to forecast business cycle turning points of the United States (US) with state-level climate risks data, proxied by temperature changes and its (realized) volatility. We find that forecasts obtained from the DMA combination scheme provide timely updates of the US business cycles based on the information content of the metrics of state-level climate risks, particularly volatility of temperature, relative to the corresponding small-scale MS benchmarks that use national-level values of climate change-related predictors.