Point and density forecasting of macroeconomic and financial uncertainties of the USA

Abstract

We forecast macroeconomic and financial uncertainties of the USA over the period of 1960:Q3 to 2018:Q4, based on a large dataset of 303 predictors using a wide array of constant-parameter and time-varying models. We find that uncertainty is indeed forecastable, but while accurate point forecasts can be achieved without incorporating time variation in the parameters of the small-scale models for macroeconomic uncertainty and large-scale models for financial uncertainty, this is indeed a requirement, along with a large dataset, for producing precise density forecasts under both types of uncertainty.