Risk and Uncertainty in Cryptocurrency Markets

Abstract

This dissertation consists of three kinds of research. Each one has its purpose and aim to achieve. The first research tries to discover the most effective approach for forecasting the volatility of cryptocurrency returns utilising high-frequency data that can predict the volatility of dominant and less notable cryptocurrencies. The GARCH, IGARCH, EGARCH, GJR-GARCH, HAR, and LRE models were investigated, and univariate and comprehensive regression were used. Regarding univariate regression results, the HAR model beat the other models when forecasting one day ahead, while the EGARCH model outperformed the other models when forecasting seven and thirty days ahead. In addition, the HAR + EGARCH duo beat the other model couples when forecasting one, seven, and thirty days. Aside from the primary study, the out-of-sample analysis

yielded conflicting results. These results will benefit investors, portfolio managers, and other financial professionals. The second study seeks to investigate the relationship between cryptocurrency returns and uncertainty indices along with assessing the impact of the Covid-19 pandemic period on both indices and cryptocurrency returns, determining which index has the most significant influence on cryptocurrency market results, and determining which indices pair has the most significant influence on cryptocurrency market returns. Ten cryptocurrency returns, as well as eight uncertainty indices, were investigated. The Quantile Regression, Multivariate-Quantile Regression, and Granger Causality tests were used. According to the Quantile Regression results, the Cryptocurrency Policy Uncertainty index and the Cryptocurrency Price Uncertainty index considerably impact cryptocurrency returns. On the other hand, the other indices have no influence on cryptocurrency returns. The Multivariate-Quantile Regression findings demonstrated that when the cryptocurrency market experiences a bull wave, the UCRY Policy Index + Central Bank Digital Currency Attention Index combination strongly impacts cryptocurrency returns. Nonetheless, when the cryptocurrency market has a bull run, the UCRY Policy Index and the Cryptocurrency Environmental Attention (ICEA) index combination considerably impact cryptocurrency gains. During the crisis, most of the overall sample findings were verified. These insights will benefit investors, portfolio managers, and policymakers.

The third research strives to find the best model for forecasting the covariance matrix of cryptocurrency returns. To achieve this purpose, five models were thoroughly examined: BEKK, Diagonal BEKK, DCC, Asymmetric DCC, and LRE are all examples of BEKK. To assess prediction accuracy and capacity, three essential criteria were used: Euclidean distance (LE), Frobenius distance (LF), and the multivariate quasi-likelihood loss function (LQ). The LRE model outperformed the other models, predicting daily and weekly frequencies more accurately. Furthermore, the Mean Squared Error (MSE) and Mean Absolute Error (MAE) loss functions were used for validation. Except for LQ, the findings were in line with the forecasting criteria. These findings have significant implications for investors and portfolio managers aiming to enhance their risk management techniques. By utilizing the knowledge provided, they may be able to make better-informed decisions to lower portfolio risk.