Model-Based Forecasting

Abstract

Forecasting is a time-oriented activity, and forecasting is therefore an important area of application of dynamic econometric models. In this chapter, the theory of model-based forecasting is presented, and its relevance for applied forecasting is illustrated. The field of forecasting is large, stretching far beyond the scope of this chapter in a non-specialized textbook. The choice of topics and emphasis has been made with applied forecasting in mind, in particular the theory that any would-be macroeconomic forecaster is well advised to have as a background. One message is that several forecasting methods have a common ground in the VAR. Hence, methods that are often presented as alternative, nevertheless, have shared strengths and weaknesses, and that they may be better described as complementary forecasting methods. A second main theme in this chapter is the large impact that structural changes in the macroeconomy have on the properties of forecast errors. Since the empirical evidence, in the form of the forecast records of resourceful forecasters, shows that forecasts failures are common, we need a theory that explains why this is so. This seems to be a requisite not only for working theoretically toward improved forecast methods, but also for improving the robustness and use value of the forecasting methods that are available today.