Value at risk for Konsesjonskraft IKS

Abstract

ARMA models and uni- and multivariate ARMA-GARCH models are estimated based on log returns from time series of electricity price data from 2005 to 2019. The electricity price data are Nord Pool day-ahead spot prices and 1-3 years Nasdaq OMX forward contracts. The ARMA and ARMA-GARCH models were evaluated based on moving window, one-day walk forward calculations of AIC and BIC, and forecast testing towards an exponential weighted moving average benchmark model following a bootstrap procedure. Based on historical performance, models are chosen and used for out-of-sample forecasting as a basis for calculating value at risk for the required capital reserve for Konsesjonskraft IKS. The analyses did not provide any usable forecast because of not converging ARMA forecasts and divergent ARMA-GARCH models, thus, further model screening and selection is required. However, the study has contributed to creating a script for model screening including historical calculation of AIC and BIC and for forecasting evaluations to compare different model types, and the methodology is well prepared for further development including other explanatory factors. A univariate GARCH model approach is suggested for future development.