Investor Sentiment and Bitcoin Bubble Dynamics: An empirical analysis of the 2016–2018 period

This thesis examines the relationship between investor sentiment and Bitcoin market behaviour over the period 2016–2018, which includes the rapid increase in Bitcoin prices during 2017 and the subsequent market downturn in 2018. Drawing on the existing literature, this period is considered a relevant case for studying speculative bubble dynamics in cryptocurrency markets. The aim of the analysis is to investigate whether sentiment-related indicators, together with Bitcoin-specific market variables, help explain changes in Bitcoin returns and, in particular, its volatility. The dataset combines daily Bitcoin data with endogenous variables such as trading volume, momentum, drawdown, rolling volatility, and GARCH conditional volatility. It also includes exogenous variables, namely Google Trends, the VIX, gold returns, S&P 500 returns, and the US Dollar Index. These variables are used as proxies for investor attention, uncertainty, risk appetite, and wider financial-market conditions. The empirical analysis is conducted in MATLAB. It includes descriptive and distributional analysis of Bitcoin returns, autocorrelation analysis, stationarity and ARCH-effect tests, and estimation of a GARCH(1,1) model. Regression models are then used to assess which endogenous and exogenous variables are statistically significant in explaining next-day Bitcoin conditional volatility. A next-day return model is also estimated as a robustness check, while multicollinearity is examined through variance inflation factors. Finally, the main econometric findings will be represented through an agent-based model developed in NetLogo. The purpose of this model is to visually illustrate how sentiment-driven investment decisions may contribute to Bitcoin price movements, volatility, and bubble-like behaviour.