Modelling and forecasting price volatility in crypto-currency markets: An application to bitcoin

Modeling and forecasting financial market volatility has been an important research topic over the last three decades. The reason for the intense focus is that volatility is a prevalent feature of financial markets and the forecasting of volatility has numerous applications in economics and finance, i.e., risk management, portfolio management, option pricing, capital asset pricing, and even monetary policy. As a result, there are numerous articles and working papers that study the forecasting performance of various volatility models for a multitude of assets. This paper contributes to this literature by analyzing the volatility forecasts in a new and exciting asset class; cryptocurrencies.

The aim of this paper is to evaluate the forecasting performance of various GARCH models using data for Bitcoin (BTC/U.S. dollar exchange rate) over the period April 1, 2013 to October 24, 2017.

Modeling and forecasting time-varying financial market volatility are important for investors who are interested in the forecast of the variance of a series over the holding period for calculating measures of risk, pricing derivatives, and hedging. The long-run forecast of the conditional variance would be unimportant for these investors who hold the asset for a certain period only. In a seminal paper, Engle (1982) shows how to model the conditional variance of a time series. Bollerslev (1986) generalizes Engle’s work by allowing the conditional variance to be an ARMA process. The literature continued to grow by extending these works to the case of vector processes. Early articles on multivariate extensions are Engle, Granger and Kraft (1986), Diebold and Nerlove (1989), and Bollerslev, Engle and Wooldridge (1988). See Poon and Granger (2003) for an extensive survey. Bollerslev (2008) provides a comprehensive list of different volatility models discussed in the literature.