In finance, obtaining a precise estimation of the risk measures of the investment portfolio is crucial for achieving good performance. This thesis compares three methods for estimating the value at risk (VaR) and expected shortfall (ES), the most important risk measures in portfolio analysis. Two of the methods capture cross-sectional and serial dependence, while the third one uses the first four moments of the portfolio assets and ignores serial and cross-sectional dependencies. The first model uses an ARMA-GARCH vine copula model utilizing the software developed by Sommer (2022). In the first step it models the serial dependence with ARMA-GARCH models for each asset, separately. In a second step R-vine copulas are used to quantify cross-sectional dependence. The second method uses S-vine copulas proposed by Nagler et al. (2022) for modelling both dependencies at the same time based on translation invariance. For the last method Fleishman’s transformation introduced in Fleishman (1978) is used for estimating the risk measures. For all the methods, the estimation of risk measures is performed using Monte Carlo on the portfolio’s log returns. To conduct a time varying analysis, it is applied within a rolling window. The research includes several evaluation measures and backtests to compare the methods. A case study is conducted representing the BVK portfolio, composed of private equity, equity, fixed income, real estate and hedge fund indices. Two markets with 15 and 7 assets each, and three portfolios (equal weighted, market capitalization, and BVK) are considered in the study over a time period of 2005 to 2022. This comprehensive analysis helps us discover the strengths and weaknesses of each method.      «

In finance, obtaining a precise estimation of the risk measures of the investment portfolio is crucial for achieving good performance. This thesis compares three methods for estimating the value at risk (VaR) and expected shortfall (ES), the most important risk measures in portfolio analysis. Two of the methods capture cross-sectional and serial dependence, while the third one uses the first four moments of the portfolio assets and ignores serial and cross-sectional dependencies. The first...     »