Kripto Para Birimlerinin Volatilite Yapılarının Karşılaştırmalı AnaliziFatih Kazova, Ayça Büyükyılmaz Ercan
2008 küresel finans krizi ile birlikte ortaya çıkan kripto para birimi, geleneksel para sisteminin yerini almak üzere geliştirilen alternatif bir değişim aracı olmuştur. Kripto para birimleri hızlı ve güvenli işlem yapabilmesi, aracı kurumları ortadan kaldırması ve düşük maliyetli olmasından dolayı giderek popüler hale gelmiştir. Ancak kripto para piyasasındaki sert dalgalanmalardan ötürü oluşan yüksek risk-getiri oranı nedeniyle literatürde kripto paraların getirilerinin yanında risklerinin de dikkate alınmasının önemi vurgulanmaktadır. Bu çalışmada pozitif ve negatif şokların kripto para birimlerinin getiri oranlarının volatilitesi üzerindeki etkilerinin araştırılması amaçlanmıştır. Bu doğrultuda piyasa değeri yüksek olan BTC, ETH, XRP, ADA, LTC, BCH, XLM, LINK, TRX ve DOGE kripto para birimleri seçilerek getiri serileri oluşturulmuştur. Bu getiri serilerinin volatiliteleri simetrik ve asimetrik koşullu değişen varyans modelleri kullanılarak analiz edilmiştir. Veri seti dönemi her bir kripto para birimi için değişmekle birlikte en geniş veri seti 01.01.2017-16.01.2021 dönemini kapsamaktadır. Elde edilen bulgular BTC, ADA, LINK getiri serilerinde meydana gelen negatif şokların pozitif şoklara göre volatilite üzerinde daha çok etkisi olduğunu göstermiştir. Diğer taraftan ETH, XRP, LTC, BCH, XLM, TRX, DOGE getiri serilerinde ise pozitif şokların volatilite üzerinde daha büyük bir etkiye sahip olduğu sonucuna ulaşılmıştır. Sonuç olarak bu çalışmada kullanılan veriler için getiri serilerinin volatilitelerinin modellenmesinde asimetrik koşullu değişen varyans modellerinin, simetrik koşullu değişen varyans modellerine göre daha anlamlı sonuçlar verdiğini göstermiştir.
Comparative Analysis of the Volatility Structure of CryptocurrenciesFatih Kazova, Ayça Büyükyılmaz Ercan
Cryptocurrency emerged as an alternative medium of exchange developed after the 2008 global financial crisis to replace the traditional money system. Cryptocurrencies have become increasingly popular because of their fast and secure transactions, elimination of intermediaries, and low cost. However, due to the high risk–return ratio arising from sharp fluctuations in the cryptomoney market, studies have emphasized that the risks of cryptocurrencies should be considered in addition to their returns. This study investigates the effects of positive and negative shocks on the volatility of the rates of return on cryptocurrencies. In this direction, a return series was created by choosing BTC, ETH, XRP, ADA, LTC, BCH, XLM, LINK, TRX, and DOGE cryptocurrencies with high market values. The volatility of these return series was analyzed using symmetric and asymmetric conditional heteroskedasticity models. Although the data set period varies for each cryptocurrency, the largest dataset covers the period from January 1, 2017, to January 16, 2021. The findings show that negative shocks in BTC, ADA, and LINK return series have more effect on volatility than positive shocks. Alternatively, it was concluded that positive shocks have a greater effect on volatility in ETH, XRP, LTC, BCH, XLM, TRX, DOGE return series. Therefore, for the data used in this study, it has been shown that the asymmetric conditional heteroskedasticity models give more meaningful results than the symmetric conditional heteroskedasticity models in modeling the volatility of the return series.
The term blockchain was first used in an article published by Satoshi Nakamoto (2008). Blockchain can be defined as a system that keeps all transactions of the users of a system in the network through verification. The database of this system consists of controllable and reliable transactions composed of blocks. The concept of block can be expressed as the storage of data in the blockchain system. The emergence of the blockchain technology led to the creation of cryptocurrencies, which became increasingly popular because of their fast and secure transactions, elimination of middlemen, and low costs. Inflation rates increased, especially during the COVID-19 pandemic, with several countries increasing money printing. Therefore, the use of cryptocurrencies continues to increase because with limited total supply, cryptocurrencies such as Bitcoin are not affected by inflation. Although cryptocurrencies are not affected by inflation, they are highly volatile owing to several factors.
Financial time series can be analyzed using various methods. From a historical perspective, financial time series analyses have primarily used linear models. Over time, nonlinear models have been developed, considering that linear models are insufficient to explain the data. Various sub-models of nonlinear models have also been introduced to the literature in the process. Among these models, conditional heteroskedasticity models are frequently used. When it was first proposed, no successful results could be obtained because conditional heteroskedasticity models did not consider asymmetry. Therefore, conditional heteroskedasticity models that included asymmetry were developed. Better results are obtained by considering asymmetry in financial time series, because it is known that the effect of negative and positive shocks in financial time series returns on volatility is not always the same. This factor is considered with asymmetric models.
Several studies have used conditional heteroskedasticity models for analyzing volatility structures. In this study, cryptocurrencies, which are considered novel and significant part of the world money markets, are analyzed with conditional heteroskedasticity models. In this study, symmetric conditional heteroskedasticity models, such as Autoregressive Conditional Heteroskedasticity (ARCH), Generalized Autoregressive Conditional Heteroskedasticity (GARCH), Autoregressive Conditional Heteroskedasticity in Mean (ARCH-M), Generalized Autoregressive Conditional Heteroskedasticity in Mean (GARCH-M), and Integrated Generalized Autoregressive Conditional Heteroskedasticity(IGARCH), and asymmetric models, such as The Exponential Generalized Autoregressive Conditional Heteroskedasticity (EGARCH), The Threshold Generalized Autoregressive Conditional Heteroskedasticity (TGARCH), The Asymmetric Power Autoregressive Conditional Heteroskedasticity (APARCH), and Asymmetric Component Generalized Autoregressive Conditional Heteroskedasticity (ACGARCH), are discussed. In this study, unit root test, Autoregressive Moving Average (ARMA) models and Autoregressive Conditional Heteroskedasticity Lagrange Multiplier (ARCH LM) test were applied, respectively. Then, using symmetric and asymmetric conditional heteroskedasticity models, the volatility of cryptocurrencies such as BTC, ETH, XRP, ADA, LTC, BCH, LINK, XLM, TRX, and DOGE were estimated. Depending on the estimations, the models that yielded significant results were evaluated according to the Akaike Information Criterion (AIC)and Schwarz Information Criterion (SIC) information criteria and Log Likelihood (LL) values.
In traditional linear regression models, it is assumed that the variance in the estimated model will be constant over time. However, it is seen that the error term variance can change over time due to use of variables such as stocks, inflation rate, cryptocurrencies, and exchange rates in financial time series. This is called the heteroskedasticity. To be able to analyze with conditional heteroskedasticity models, some steps should be followed. The first step is to test whether there is an autocorrelation problem in the data to be analyzed and to use ARMA(p,q) models if necessary. In the second step, the ARCH LM test is applied to the residuals by developing the mean equation to test the ARCH effect in the series. Finally, in the third step, if there is an ARCH effect in the residuals because of the ARCH LM test, the estimation is made using appropriate conditional heteroskedasticity models. After evaluating the estimation results of the established model, it should be revised if necessary. Conditional heteroskedasticity models are divided into symmetric and asymmetric. While symmetric models assume that positive and negative shocks on volatility have the same effect, asymmetric models are based on the assumption that this effect is different.
Among the models used in the study, better results were obtained with EGARCH(1,1), TGARCH(1,1), and ACGARCH(1,1) models. Thus, positive shocks have more effect on volatility in the ETH, XRP, LTC, BCH, XLM, TRX, and DOGE return series, while negative shocks have more effect on BTC, ADA, and LINK return series.