A Single-Equation Approach to Model-Based Forecasting
Model-based forecasting approaches are essential elements in the process of effective decision making. One major benefit of using these forecasting approaches is that the model can serve as a baseline for adjusting economic guidance for risk assessment, and a model's forecasting performance can be monitored over time to compare expected versus actual outcomes. In addition, a model that is not performing as expected can be revised and rebuilt to improve guidance to decision makers. Decision makers utilize models to generate both long-term and short-term forecasts. The focus of this chapter, however, is on short-term forecasting.1 We discuss different methods to build forecasting models and describe which method may be appropriate for a forecaster depending on the forecasting objective and on the available data set.
This chapter focuses on single-equation, univariate, forecasting. A univariate model contains one dependent variable and one or more predictors (independent variables). Within a univariate forecasting framework, we discuss two approaches: unconditional and conditional forecasting. With the unconditional forecasting model, we do not need out-of-sample values of predictors to generate an out-of-sample forecast. Conditional forecasting, in contrast, implies that the forecasted values of the dependent variable are conditioned on the predictors' values. Another major difference between unconditional and conditional ...