Resumo

The current market scenario features technological innovations, faster information dissemination, and new products and services. The financial market has transformed significantly due to the internet, which has enabled low-cost interactions and information sharing (Suryono et al., 2020). A notable innovation in the last decade is cryptocurrencies, a decentralized financial product blending characteristics of currency, commodities, and stocks (Charfeddine et al., 2020). Bitcoin, created by Nakamoto (2008) and launched in 2009, initiated this new market, leading to the development of altcoins.

The aim of this study is to measure the risk of the 10 largest cryptocurrencies using Value-at-Risk (VaR) and Expected Shortfall (ES). Volatility forecasting was conducted using GARCH models. This study stands out for three reasons: it utilizes 63 GARCH model variations to identify the best for VaR and ES forecasts; it includes external variables, the Fear and Greed Index (FGI) and Market Value to Realized Value (MVRV), to enhance volatility modeling; and it analyzes 10 cryptocurrencies.

Risk is the possibility that the actual return on an investment will differ from the expected return. Risk management is crucial for investment decisions, helping investors maximize returns adjusted to their risk preferences. Value-at-Risk (VaR) is a traditional measure for assessing asset risk. However, Acerbi and Tasche (2002) argue that VaR, limited to identifying the maximum potential loss, is less suitable for risk analysis. They propose Expected Shortfall (ES), , which estimates the average loss in worst-case scenarios, providing a more comprehensive view of extreme risk.

For this study, data was collected on the dollar quotations of 10 cryptocurrencies from 14/05/2019 to 14/05/2023. The following cryptocurrencies were therefore selected for this study: Bitcoin (BTC), Ethereum (ETH), Ripple (XRP), Dogcoin (DOGO), Binance (BNB), Cardano (ADA), Tronix (TRX), Bitcoin Cash (BCH), Chainlink (LINK) and Litecoin (LTC). In addition, data was collected on two series of external variables, the FGI and the MVRV, both obtained via BGeometrics. In order to calculate VaR and ES, it was decided to use variations of GARCH model with and without the two external variables.

Overall, the calculated VaRs had a "Success" rate of 96.65% in estimating risk, implying a 3.35% failure margin. For ES, these values are 98.27% and 1.73%, respectively, indicating a lower margin of error for worse outcomes compared to VaR. FGI and MVRV are useful for predicting volatility in the four largest cryptocurrencies but less so for others. ES_Ext was superior to ES only for BTC, and for BNB, ADA, and BCH, it showed a loss of efficiency. Therefore, ES, while better than VaR, does not benefit from using FGI and MVRV in ARMA-GARCH models, except for BTC.

Overall, both metrics achieved a success rate exceeding 95%, demonstrating the suitability of the proposed methodology for risk management. ES proved to be a more accurate alternative than VaR. However, the inclusion of the external variable have bought little benefits for the models, especially for the ES estimation. Another key finding was the lack of a clear pattern in model parameter selection, particularly in GARCH family variations and error distributions.

Acerbi, C., & Tasche, D. (2002). On the coherence of expected shortfall. Journal of Banking & Finance, 26(7), 1487–1503 Nakamoto, S. (2008). Bitcoin: A Peer-to-Peer Electronic Cash System Suryono, R. R., Budi, I., & Purwandari, B. (2020). Challenges and Trends of Financial Technology (Fintech): A Systematic Literature Review. Information, 11(12), Artigo 12. Charfeddine, L., Benlagha, N., & Maouchi, Y. (2020). Investigating the dynamic relationship between cryptocurrencies and conventional assets: Implications for financial investors. Economic Modelling, 85, 198–217.