A carregar...
Publicação
Sentiment and volatility in cryptocurrency markets : an empirical investigation using GARCH-X, EVT, and regime-switching models
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 4.95 MB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
Volatility modeling plays a key role in understanding and managing financial risk, particularly in high-frequency and sentiment-driven markets such as cryptocurrency. However, traditional models often struggle to capture extreme fluctuations caused by sudden shifts in investor behavior. This study investigates whether public sentiment data obtained from platforms like LunarCrush and Google Trends can improve the forecasting of volatility and tail risk in crypto assets. To verify this, we apply a set of advanced time-series models to hourly price data for four major cryptocurrencies (BTC, ETH, DOGE, LINK) for the period 2020 to 2025. The modeling framework integrates multiple layers, including GARCH and EGARCH variants with external sentiment regressors, regime-switching volatility via Markov models, and tail modeling via Generalized Pareto Distribution. Model performance is assessed both in terms of volatility forecast accuracy and risk coverage metrics such as Value-at-Risk (VaR) and Expected Shortfall (ES). This article pays particular attention to changes in distribution behavior during panic and performs Monte Carlo simulations to assess forward looking tail risk. The results show that for both ETH and DOGE, the sentiment-enhanced GARCH model outperforms the standard model, especially during periods of heightened sentiment volatility. The regime-switching model shows that negative sentiment significantly increases the probability of entering a high-risk state, while the tail model suggests that once in such a state, the distribution of returns becomes quite heavy. DOGE exhibits the most severe tail risk amplification, while BTC and ETH show a more stable, but still significant, variation. In sum, this work provides evidence that sentiment signals not only predict short-term volatility, but also effectively capture the structural changes that drive extreme downside risk in cryptocurrency markets.
Descrição
Trabalho Final de Mestrado, Mathematical Finance, ISEG, 2025.