Publicação
Forecasting Bitcoin returns volatility using GARCH methods
| Resumo: | This study aims to investigate the dynamics of Bitcoin’s price and volatility. The analysis begins by examining Bitcoin’s daily returns, identifying a stationary time series with pronounced volatility clusters. These characteristics, combined with (a)symmetry and (non)uniform dispersion, suggest the suitability of ARCH/GARCH models for statistical analysis. To determine the most appropriate model, a range of GARCH, EGARCH, and GARCH models with exogenous variable models are evaluated. The assessment includes a careful examination of AIC and BIC values and the interpretation of the coefficients of the model parameters. The statistical significance of coefficients confirms the impact of past squared returns and conditional variances on current volatility. The study culminates in a detailed analysis of Value at Risk (VaR) forecasting, with the EGARCH (1,1) model with a Student’s-t distribution emerging as the most effective in capturing Bitcoin returns’ VaR, based on the number of exceedances identified at 99% and 95% confidence levels. The research underscores the importance of choosing a model that aligns with the user’s risk profile and investment goals. However, it also acknowledges some limitations, such as the incapacity of using the exogenous variable in VaR forecasting and the potential for more advanced computational methods in future investigations. |
|---|---|
| Autores principais: | Loureiro, Clara Verdade |
| Assunto: | Bitcoin Volatilidade -- Volatility ARCH/GARCH models Modelos ARCH/GARCH |
| Ano: | 2023 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | ISCTE |
| Idioma: | inglês |
| Origem: | Repositório ISCTE |
| Resumo: | This study aims to investigate the dynamics of Bitcoin’s price and volatility. The analysis begins by examining Bitcoin’s daily returns, identifying a stationary time series with pronounced volatility clusters. These characteristics, combined with (a)symmetry and (non)uniform dispersion, suggest the suitability of ARCH/GARCH models for statistical analysis. To determine the most appropriate model, a range of GARCH, EGARCH, and GARCH models with exogenous variable models are evaluated. The assessment includes a careful examination of AIC and BIC values and the interpretation of the coefficients of the model parameters. The statistical significance of coefficients confirms the impact of past squared returns and conditional variances on current volatility. The study culminates in a detailed analysis of Value at Risk (VaR) forecasting, with the EGARCH (1,1) model with a Student’s-t distribution emerging as the most effective in capturing Bitcoin returns’ VaR, based on the number of exceedances identified at 99% and 95% confidence levels. The research underscores the importance of choosing a model that aligns with the user’s risk profile and investment goals. However, it also acknowledges some limitations, such as the incapacity of using the exogenous variable in VaR forecasting and the potential for more advanced computational methods in future investigations. |
|---|