Risk measurement of cryptocurrencies using value at risk and expected shortfall

Sammanfattning: Cryptocurrencies are highly volatile and risky assets, therefore, it is of vital importance to find an appropriate model for risk measurement. This thesis compares three parametric and three non-parametric estimation methods to estimate the value at risk and the expected shortfall of five cryptocurrencies, namely Bitcoin (BTC), Ethereum (ETH), Binance coin (BNB), Ripple coin (XRP), and Cardano (ADA). We estimate the value at risk and expected shortfall using these methods at the confidence level of 95% and 99%. We then perform five backtesting procedures and use these test results to compare the performance of these estimation methods. Consequently, we can conclude that the volatility-weighted historical simulation (VWHS) method using the exponential weighted moving average (EWMA) model and GARCH-type models to rescale cryptocurrency loss for VaR and ES estimation perform the best in most cases. The basic historical simulation (BHS) method and the peak over threshold (POT) method also show positive performance in several cases. Meanwhile, the age-weighted historical simulation (AWHS) has a poor performance in almost all cases.