摘要：This paper investigates the possibility to obtain better GDP forecasts in the early stages of Great Recession. Here, predictive performance refers to exclusively out-of-sample forecasts. Based on exploratory data analysis and general-to-specific modelling, this paper proposes a univariate predictive threshold model for the small open economy that outperforms its linear counterparts and correctly determines the course of events. This model does not explain any causal links; however, based on a set of economic arguments, it sets forward an idea regarding how a forecaster can act when principal determinant factors, responsible for a sudden, yet lasting change, are unknown, unmeasurable or cannot be influenced by national policy makers. A major dissimilarity between usual threshold models and the model presented in this paper is that while variables act differently under different conditions in the former, in this model, due to economic reasons, errors act differently. Alternatively, this paper can be viewed as a comparative GDP prediction study.
关键词：univariate time series models; forecasting and prediction methods; autocorrelated errors; threshold models; GDP forecasting.
