Further Topics in the Analysis of Non-Stationary Time Series
In this chapter three further topics are considered in some detail: estimation of models with I(2) variables; forecasting; and structural models with short-run behaviour driven by expectations. Though mathematically the notions of order of integration and cointegration are exact, in practice they are valid to the best approximation or resolution that the data may permit. To define an order of integration as a specific integer quantity is to assume that the series is approximated by a single well-defined time series process across the sample. Time series data for developed economies have exhibited many features, from behaviour that might be viewed as purely stationary through to series that require first or second differencing to render them stationary. Some nominal series in first differences may require further differencing, which suggests that the original nominal series are of order I(2) or higher when further differencing is required. In this chapter, discussion is limited to processes up until I(2).
KeywordsUnit Root Forecast Error Forecast Performance Forecast Horizon Forecast Error Variance
Unable to display preview. Download preview PDF.