Прогнозирование последовательных крахов на финансовых рынках (на примере динамики индекса Доу-Джонса)
Financial markets tend to demonstrate extreme events in prices dynamics, among those are jumps leading to drastic prices’ changes and even regimes switching as well as for some instruments and for the markets overall.
Effective forecasting of price dynamics and actions planning by the market participants doesn’t necessary demand exact future price trajectories extrapolation, but estimation of duration of time periods, during which the prices won’t fall down more than a given mark could be enough.
In the paper an approach to forecasting durations between consequent market crashes is suggested. A significant autocorrelation was found in durations between crashes of DJIA during more than last 80 years. These autocorrelation allowed estimating a series of autoregressive conditional duration models to forecast time periods before next index crash. The models demonstrate significant forecasting power, especially in situations when consequent crashes appear rather frequently.