This paper investigates a range of first-order implications of Bitcoin within economics and finance through a holistic approach and deduces a testable hypothesis concerning the ability of Twitter sentiment to predict Bitcoin price movements. In combination, a technological, theoretical and empirical review of Bitcoin ensure the necessary breadth and depth to appropriately advance the knowledge frontier of a highly relevant but less researched topic. Among the economic aspects investigated, a perspective on what type of money Bitcoin is and how it performs as money is included followed by an analysis of how Bitcoin impacts financial intermediation, money supply and money creation. This paper finds that Bitcoin can be characterized as synthetic commodity money while it does not perform well as money. Bitcoin makes trust in a centralized third party redundant in facilitating payments and can fundamentally alter the dynamics of money supply and creation. However, Bitcoin has currently no impact of economic significance. Among the financial aspects investigated, an analysis of how Bitcoin relates to traditional asset classes and the implications of considering Bitcoin as part of a new asset class was conducted. Subsequently, the shortcomings of conventional finance theory in explaining Bitcoin returns were analyzed followed by an assessment of the applicability of behavioral finance in relation to Bitcoin. This paper finds that Bitcoin is conceptually a currency but is mainly perceived as an investment vehicle. Furthermore, it was found that conventional finance theory is inadequate in explaining Bitcoin returns and that behavioral finance theory holds explanatory power concerning the ability of sentiment to influence prices. Based on the conclusions from this first part the paper, a testable hypothesis was deduced on the predictive ability of Twitter sentiment on Bitcoin price movements. This paper conducts a textual sentiment analysis on more than 1.1 million tweets using three different lexica to construct a sentiment variable as input in three different models. By using both an autoregressive distributed lag model, a logistic version of the same model and an artificial neural network, this paper confirms the hypothesis of Twitter sentiment’s predictive ability when considering a 6-hour time interval, which is reflected in the profitability of the associated trading strategy.
|Educations||MSc in Applied Economics and Finance, (Graduate Programme) Final Thesis|
|Number of pages||156|