MODEL UNCERTAINTY AND FORECAST ACCURACY

In time-series analysis, a model is rarely pre-specified but rather is typically formulated in an iterative, interactive way using the given time-series data. Unfortunately the properties of the fitted model,an d the forecasts from it, are generally calculated as if the model were known in the first Place: This is theoretically incorrect, as least s quares theory, for example, does not apply when the same data are used to formulates-and fit a model. Ignoring prior model selection leads t o biases, not only in estimates of model parameters but also in the su bsequent construction of prediction intervals. The latter are typicall y too narrow, partly because they do not allow for model uncertainty. Empirical results also suggest that more complicated models tend to gi ve a better fit but poorer ex-ante forecasts. The reasons behind these phenomena are reviewed. When comparing different forecasting models, the ETC is preferred to the AIC for identifying a model on the basis d f within-sample fit, but out-of-sample forecasting accuracy provides;t he;real test. Alternative approaches to forecasting, which avoid condi tioning on a single model; include Bayesian model averaging and using a forecasting method which is not model-based,but which is designed to be adaptable and robust.