Econometric Analysis of Cryptocurrency Markets: A Time Series Approach

Abstract

Cryptocurrency markets have grown at unprecedented rates in the early 2020s due to increased demand and become further integrated into the broader financial system through new practices, regulatory developments, and institutional adoption. To help develop risk assessment tools and optimize trading strategies, this dissertation studies the stability, efficiency, and predictability of cryptocurrency markets using time series methods. The stability of cryptocurrency markets is analyzed in the context of stablecoins to better understand whether this class of cryptocurrencies lives up to its name and to mitigate the risks associated with the current practices. Daily stablecoin prices are modelled as a univariate double autoregressive process (DAR) with time-varying parameters and a set of simple plug-in measures of stability is introduced. The local analysis reveals significant temporal and cross-coin variation in the stability of large-cap stablecoins. To evaluate the efficiency of cryptocurrency markets, a statistical test for the no-arbitrage hypothesis is developed based on a time series model of fundamental and idiosyncratic risk factors. The test is conducted both globally and locally for major cryptocurrency markets via a method of simulated moments. The empirical application shows that the fragmented nature of the spot markets presents significant arbitrage opportunities for cryptocurrency traders, especially when coupled with higher price volatility. Motivated by their similarity with stablecoin prices, the out-of-sample predictability of funding rates of Bitcoin perpetual futures is evaluated using the one-step-ahead point forecasts generated from a set of DAR models. The results show that the discrepancy between the spot and futures prices of Bitcoin leads to predictable patterns in historical funding rates, with the degree of predictability varying over time and across exchange platforms.