Document Type : Article-Based Dissertations
Authors
1 Ph.D. Candidate, Department of Economics, Faculty of Management and Economics, Islamic Azad University, Science and Research Branch, Tehran
2 Prof. Department of Economics, Faculty of Management and Economics, Islamic Azad University, Science and Research Branch, Tehran, Iran.
Abstract
This study aims to compare and evaluate two modeling approaches in econometrics in terms of forecasting accuracy, using a time series of Bitcoin prices for the period 2015 to 2023 AD. Bitcoin, as the most popular and well-known cryptocurrency, has become an important asset for many investors in financial markets and even ordinary people in recent years, many studies have been conducted to discover the price structure and changes of this asset over time. The high volatility of this cryptocurrency has created many complexities in these studies. Mixed data sampling (MIDAS) is a new approach for modeling data with different time frequencies in recent years and the elimination of important limitations in econometric modeling, it is increasingly being studied and investigated. In this study, the forecasting accuracy of ARIMA-GARCH and MIDAS models is compared. The models used in this research are based on intraday data of Bitcoin prices (with high frequency-5 minutes) on the variables of daily and weekly returns of Bitcoin. First, a variant of the ARIMA-GARCH model with different types of GARCH models and distributions and then the MIDAS model with different weighting functions are estimated. Finally, by dividing the sampling interval into two parts, both approaches are used to forecast the return of the Bitcoin time series and compared with the actual information. The results show that modeling with the mixed data sampling approach provides better results in terms of forecasting the return of both time series of daily and weekly returns of Bitcoin.
Keywords
)