New approaches to forecasting economic and financial time series

Grunddaten zu diesem Projekt

Beschreibung

We establish and analyze two new approaches to forecasting economic and financial time series. (i) In many real-world situations, the forecaster has available multiple (say M) individual forecasts of the same target variable. We establish a procedure for (potentially) reducing the forecast errors of the M individual forecast models, by interrelating the M forecast-error processes within a vector autoregressive framework. (ii) Many financial time series (like electricity prices) can be shown to include multifractal structures. We model the data-generating process (DGP) of the variable by incorporating (theoretical) multifractal elements, and analyze the forecasting capacities of these new DGPs.