Our Composite Forecast Model is a model of models, optimally finding weights between different qualitative and quantitative forecasts to generate a single composite forecast.
Assigned weights are time-varying and depend on the remaining time until the official data release. For example, high-frequency nowcast models tend to perform better for short-term forecasts, while qualitative forecasts tend to be better predictors for long-term forecasts; our model combines these forecasts using a locally linear random forest method to optimize the strengths of each forecast.
This model is updated daily. New releases will be made available between 16:00 and 20:00 ET.