We follow the idea of exploiting cross-sectional information to improve recession probability forecasts by aggregating indicator-specific turning point predictions to obtain economy-wide recession probabilities. This stands in contrast to most of the relevant literature, which relies on an aggregated economic indicator to identify business cycle turning points. Using smooth transition regressions we compare the forecast performance of both approaches to business cycle dating in a comprehensive real-time forecasting exercise for recessions in the US. Moreover, we propose a novel smooth transition modelling framework which makes use of the interrelation between business and growth cycles to forecast recession probabilities. Our real-time out-of-sample forecast evaluation reveals that (i) using cross-sectional information is benficial to predicting recession probabilities, (ii) aggregating indicator-specific turning point forecasts clearly outperforms turning point predictions based on a single indicator and (iii) the proposed smooth transition framework is able to provide informative recession probability forecasts for up to three months in the US.
Business cycles forecasting recessions STAR models turning points