Abstract

This study presents an in-depth empirical analysis of Bitcoin’s daily price dynamics and volatility from 2014 to 2024, leveraging advanced time series modelling techniques. Initial descriptive statistics reveal extreme price volatility, with daily closing prices ranging from $178 to over $106,000 and exhibiting positive skewness and platykurtic behavior, indicating frequent extreme upward returns and a flatterthan-normal distribution. Stationarity tests confirm the non-stationary nature of raw prices, necessitating log transformation and differencing, after which log returns are shown to be stationary and suitable for further modeling. Volatility modeling using various GARCH and ARCH specifications identifies the GARCH(3,3) model as the best fit based on multiple statistical criteria (AIC, BIC, MSE), capturing the persistence and clustering of Bitcoin’s volatility. The AR(2)- GARCH(1,1) model further elucidates significant negative autocorrelation in log returns and strong volatility persistence, with model diagnostics confirming its robustness and absence of remaining ARCH effects. Forecasts from this model indicate that while expected returns oscillate near zero, forecasted volatility steadily increases over the horizon, reflecting growing uncertainty in Bitcoin price movements. These findings provide valuable insights into Bitcoin’s complex price behavior and volatility dynamics, informing traders, investors, and policymakers on risk management and forecasting in highly volatile cryptocurrency markets.

Keywords: Bitcoin volatility, GARCH models, Cryptocurrency forecasting, Real-time data integration, Financial risk management, Python data analysis, Market volatility modeling, Econometric modeling, Digital asset analytics