Investigation of Swedish krona exchange rate volatility by APARCH-Support Vector Regression

This paper investigates daily exchange rate volatility behaviors with a focus on a small open economy’s currency, the Swedish krona (SEK), against four currencies: the U.S. dollar, Euro, the Pound Sterling (GBP), and the Norwegian krone (NOK) over the whole period from Jan. 2010 to March 2023, whereas the whole period is divided into different sub-sample periods based on the economic events. In the framework of APARCH models, we find that volatility behavior of the Swedish krona (SEK) exchange rates varies across different currency pairs (SEK being included in all cases) and sub-sample periods. Precisely, a negative asymmetric return-volatility relationship was found for the case of the SEK/EUR exchange rate, while an inverted asymmetric relationship was detected in the case of SEK/NOK exchange rate. Significant asymmetric effects of volatility in the SEK/USD and SEK/GBP exchange rates were not observed for either the whole period or the three sub-sample periods. As the return of exchange rate are all non-normally distributed, we then use a distribution-free support vector machine-based regression, called support vector regression (SVR), to estimate and forecast volatility in the framework of the chosen APARCH model for each krona exchange rate. The result shows that the SVR-APARCH based volatility forecasting performs better than the forecasting based on APARCH model estimated by maximum likelihood estimation (MLE).