Home » Articles » Exploring the best means for estimating bitcoin price volatility – A research study
Click Here To Hide Tor

Exploring the best means for estimating bitcoin price volatility – A research study

It is inarguable that cryptocurrencies, namely bitcoin, have attracted considerable attention during the past few years. This can be attributed to the innovative features of their underlying blockchain technology, the anonymity of their transactions and their decentralized framework that omits the need for intermediaries or third parties. However, being a highly volatile investment asset, many individuals are reluctant to participate in the cryptocurrency financial system.

A recently published research study aims at estimating bitcoin’s price volatility via a GARCH based model using data from four popular cryptocurrency exchanges including Coinbase, Bitstamp, Kraken, and Itbit. Throughout this article, we will take a look at the methodology and results of this research study.


Methodology of data collection:

Data of this study was obtained via Bloomberg’s databank. To promote uniformity of time range among the four bitcoin markets’ data obtained via the Bloomberg databank, authors of the paper focused on the period between December 11th, 2013 and August 23rd, 2017, which involved 1,376 observations. Returns were calculated via obtaining the logarithm to the base of ten of all observations.

To identify stationarity, the unit root tests as well as the ARCH test were checked. As a result, the ARCH and unit root tests yielded smooth results, so that there was no need for a correction to calculate ARCH family models.

Results of the study:

To identify Bitstamp’s volatility, the ARCH model was implemented via the smallest possible Akaike criterion (3.4896) using the ARCH/GARCH model family (TGARCH, GARCH, ARCH, and EGARCH). To provide robustness of the estimation results, three diagnostic tests were used. The deployed autocorrelation test, the ARCH model, with a probability larger than 5%, did not yield any autocorrelation problems. The ARCH LM test, which is a heteroscedasticity test, implemented onto the ARCH model’s residuals yielded a Chi-square probability larger than 5% and confirmed absence of any heteroscedasticity problems. On the other hand, the Jarque-Bera diagnostic test, using a probability value < 5% denoted that the collected data was not distributed normally. Nevertheless, using data that is not normally distributed cannot be considered a major problem by most researchers, so the used model can be considered acceptable.

For Coinbase, the ARCH model was also the best model using the smallest possible Akaike criterion (3.5299) among the ARCH/GARCH model family (TGARCH, GARCH, ARCH, and EGARCH). The same results obtained for Bitstamp’s data was also obtained from all used three diagnostic tests on Coinbase’s data.

For Itbit, the most appropriate model was the EGARCH model using the smallest possible Akaike criterion (3.4543) among all the other ARCH/GARCH family members. Similar to Bitstamp and Coinbase, all three diagnostic tests applied to Itbit’s data yielded the same results.

For Kraken’s data, the most appropriate model was the GARCH model with the smallest possible Akaike criterion (3.6620). As a result of the application of the autocorrelation test, the GARCH model, via a probability larger than 5%, did not yield any autocorrelation problems. The ARCH LM test model, which was applied to the GARCH model’s residuals, yielded a p-value lower than 5%, yet the difference was a very small number (0.0436), so we can consider the heteroscedasticity associated with the residuals negligible. The Jarque-Bera diagnostic test, via a probability value lower than 5%, denoted that the obtained data was not distributed normally.

Results of this study, which proved the appropriateness of implementing the ARCH/GARCH model family, are compliant with the results of previous studies that also proposed applying the ARCH/GARCH model family in order to predict bitcoin price’s volatility.


Having witnessed an enormous price surge between 2015 and 2017, bitcoin’s markets emerged to become an important asset that most investors seek to buy into. In 2016, the price unexpectedly jumped up to $4,300, gathering the attention of individual as well as professional investors. This increased attention towards the surging bitcoin price, necessitated conducting scientific research studies that explain the volatility and price trends in order to guide investors towards the best decisions. The relatively scarce number of research studies conducted until now do not offer investors the help they really need.

The research study we reviewed aims at filling the gap of current literature and developing means for estimating bitcoin’s price volatility in order to help investors formulate wise decision processes. Four bitcoin exchange platforms were studied: Coinbase, Bitstamp, Kraken, and Itbit. Conclusively, the study proves that the most appropriate model to predict the price volatility on Coinbase and Bitstamp is the ARCH model. For Itbit, the best model was found to be the EGARCH model, and for Kraken, the GARCH model was the best fitting model. This research advises investors to be thoroughly careful while investing in bitcoin due to the high speculative nature of this market and its unpredictable ups and downs.

Final thoughts:

We bump into news about bitcoin all over the internet every day. Such news is namely about bitcoin price hitting historical highs or governments, such as China, criminalizing bitcoin transactions. The interaction of effects of positive and negative news results in extreme chaos in the bitcoin market. This study has confirmed the high levels of volatility of bitcoin price, showing that bitcoin investors are exposed to considerable risks. As per the theory of finance, high risk levels can open the door to high profits, as well as big losses, so one should be very careful when investing in cryptocurrencies in general.

Leave a Reply

Your email address will not be published. Required fields are marked *


Captcha: *