FORECASTING NORD POOL ELECTRICITY MARKET VOLATILITY: TEST OF SYMMETRIC AND ASYMMETRIC GARCH-TYPE MODELS
Jokinen, Teemu (2010)
Jokinen, Teemu
2010
Kuvaus
Opinnäytetyö kokotekstinä PDF-muodossa.
Tiivistelmä
The purpose of this thesis is to compare the predictive power of three different GARCH-type volatility forecasting models on recently deregulated Nordic electricity market. The models included are GARCH, TARCH and EGARCH. The electricity spot price data from Nord Pool covers the period from October 29th 2000 to September 30th 2007. Three hypotheses were formed based on the findings in earlier studies and the characteristics of the models.
It is first hypothesized that the more complex model should generate most accurate forecasts. Second it is hypothesized that asymmetric volatility model results more accurate forecasts than the symmetric model. Third it is hypothesized that the volatility fore- casting capability is linked to forecasting horizon length and is decreasing over time.
Forecasts were constructed for 1-, 3-, and 5-day periods and the forecasting performance of different models is evaluated with well known error statistics: the root mean square error (RMSE), mean absolute percentage error (MAPE), the adjusted mean absolute percentage error (AMAPE), logarithmic error (LE) and heteroskedasticity adjusted mean square error (HMSE). The results suggest rejection of all three hypothesis and poor forecasting performance for GARCH-type models in electricity markets.
It is first hypothesized that the more complex model should generate most accurate forecasts. Second it is hypothesized that asymmetric volatility model results more accurate forecasts than the symmetric model. Third it is hypothesized that the volatility fore- casting capability is linked to forecasting horizon length and is decreasing over time.
Forecasts were constructed for 1-, 3-, and 5-day periods and the forecasting performance of different models is evaluated with well known error statistics: the root mean square error (RMSE), mean absolute percentage error (MAPE), the adjusted mean absolute percentage error (AMAPE), logarithmic error (LE) and heteroskedasticity adjusted mean square error (HMSE). The results suggest rejection of all three hypothesis and poor forecasting performance for GARCH-type models in electricity markets.