| Summary: | We compare a Global VAR model of actual and expected outputs with alternative models to assess the role of cross-country interdependencies and confidence in forecasting. Forecast performance is judged on point and density forecasts of growth, on probability forecasts of the occurrence of national and global recessionary events and, through a novel ‘fair bet’ exercise, on decision-making using probability forecasts. We find multi-country data and survey data are needed to fully capture the influence of global interactions and expectations in forecasts. We argue that output predictions should avoid simple point forecasts and focus on densities and events relevant to decision-makers.
|
|---|