| Abstract [eng] |
A comprehensive analysis from the financial risk perspective is performed for a portfolio consisting of three Baltic and five Nordic indices for the period spanning from 2016-12-20 to 2024-09-30 by estimating Value-at-Risk (VaR) and Expected Shortfall (ES). A combination of various backtesting methods, both traditional and comparative, is employed in order to assess the adequacy and compare the estimated VaR and ES models at several confidence levels and using four different calibration windows corresponding to one, two, three, and four years. It is found that the hierarchical Clayton copula with uniform margins produced by fitting the normal inverse Gaussian distribution to standardised residuals from GARCH-filtered logarithmic returns outperforms the benchmark estimates given by the historical simulation, multivariate normal inverse Gaussian distribution, and the hierarchical Frank copula approaches. The age-weighted historical simulation produces comparable results for a longer calibration window. Moreover, simple forecast combination strategies are explored, and an incremental VaR analysis is done for each of the eight indices in the portfolio by using the hierarchical Clayton copula approach. |
