This paper presents a practical method for assessing the uncertainty of long-term economic projections. Economic variables play a central role in the Congressional Budget Office’s analysis of federal spending and revenues, and the uncertainty of economic projections is a key driver of the uncertainty about the agency’s budget projections. The presented method quantifies the uncertainty of economic variables by using simulations from a multivariate statistical model in which variables are formulated as sums of unobserved stationary and nonstationary components. Experiments on artificial data demonstrate that the method performs fairly well compared with alternative methods in terms of long-term predictive accuracy and coverage.
