Modelling and forecasting risk dependence and portfolio VaR for cryptocurrencies

Area of Science:

• Quantitative Finance
• Computational Economics
• Financial Econometrics

Background:

• Cryptocurrencies exhibit complex interdependencies and significant volatility.
• Accurate risk assessment is crucial for cryptocurrency portfolio management.
• Traditional models may struggle to capture dynamic correlations during market stress.

Purpose of the Study:

• To investigate the co-dependence and portfolio value-at-risk of major cryptocurrencies.
• To evaluate the performance of the generalized autoregressive score (GAS) model in capturing dynamic market behavior.
• To compare the GAS model against the dynamic conditional correlation (DCC) GARCH model for forecasting and risk management.

Main Methods:

• Utilized time series data for Bitcoin, Ethereum, Litecoin, and Ripple from January 2016 to December 2021.
• Applied the generalized autoregressive score (GAS) model to analyze cryptocurrency price series.
• Conducted out-of-sample probabilistic forecasts and backtests comparing GAS with DCC-GARCH.

Main Results:

• Found strong evidence of dynamic co-dependence among the analyzed cryptocurrencies.
• The GAS model demonstrated superior ability to handle volatility and correlation changes, particularly during turbulent market periods.
• GAS model significantly outperformed the DCC-GARCH model in probabilistic forecasting and risk measure insights.

Conclusions:

• The GAS model offers a robust framework for understanding and managing cryptocurrency portfolio risk.
• Dynamic co-dependence structures are a key feature of the cryptocurrency market.
• The findings provide valuable insights for investors and risk managers in the digital asset space.