Forecasting Cryptocurrency Index CRIX

Since the CRIX index reflects the cryptocurrency market, investors and traders need to forecast this index in order to understand how the cryptocurrency market will develop in the future. In this paper, the effect of the VIX stock market volatility index on the CRIX cryptocurrency index forecast was studied using time series models. Daily values of the CRIX and VIX indices for the period from July 31, 2014, to February 28, 2019, were used. The data were divided into training and test samples in the ratio of 80:20. On the training sample, stationary time series models were evaluated (ARIMA, ARIMAX, ARIMA-GARCH, and ARIMAX-GARCH). The selection of lags in the models was determined based on the Akaike information criterion. After evaluating the models, a forecast was made for different time periods: 1, 2, 7, 30 days ahead. The accuracy of the prediction obtained was estimated using MAPE, RMSE, and Theil's U. The results were tested for stability using the expanding window forecast. The main result of this work is improving the accuracy of the cryptocurrency index forecast after inclusion the stock market volatility index with a single lag in the models when forecasting for 2 days and 7 days in advance.