A1732
Title: An ACD-POT MIDAS model for forecasting extreme returns in oil and metal markets during different economic conditions
Authors: Katarzyna Bien-Barkowska - Poznan University of Economics and Business (Poland) [presenting]
Agata Kliber - Poznan University of Economics and Business (Poland)
Abstract: The dynamics of extreme losses in oil and metal markets are investigated with the autoregressive conditional duration peaks over threshold (ACD-POT) MIDAS models. The models are tailored to the stylized facts about the dynamics of extreme events in financial markets; and hence, they can capture both the clustering of extreme-event days, and the autocorrelation in the sizes of extreme negative returns (i.e., threshold exceedances). Unlike in the existing versions of dynamic POT models, in the current approach, the time intervals between the extreme events can be either a continuous or a discrete variable. This latter assumption accounts for the fact, that the time series of financial prices are usually publicly available at daily intervals. Additionally, the standard ACD specification is enriched for the threshold exceedance times with the MIDAS component to capture the effect of the time-varying macro environment on the daily extreme loss event probability (ELEP) and the expected size of extreme losses. Forecasts of the value at risk (VaR) and expected shortfall (ES) are derived at different coverage probabilities and check the forecasting performance of the model for selected commodity assets.
