Forecasting bubbles with mixed causal-noncausal autoregressive models

This paper investigates one-step ahead density forecasts of mixed causal-noncausal models. It analyses and compares two data-driven approaches. The paper focuses on explosive episodes and therefore on predicting turning points of bubbles. Guidance in using these approximation methods are presented with the suggestion of using both of the approaches as they jointly carry more information. The analysis is illustrated with an application on Nickel prices