**2. Literature Review**

There has been a proliferating bulk of high-quality academic studies that investigate the factors determining performance of digital currencies. Ammous (2018) looks into whether Bitcoin, Ethereum, Litecoin, Ripple or Steem can fulfil the functions of money. He argues that only Bitcoin can be used as a store of value because of its commitment to increase its supply only by predetermined amounts. All cryptocurrencies are found to be inappropriate as units of account due to their large fluctuations. Moreover, theoretically, they can serve as a medium of exchange, but it is unlikely that they will become popular for this function. Bitcoin is found to be the most promising to function as money.

Brandvold et al. (2015) investigate how Bitcoin exchanges contribute to price discovery from April 2013 to February 2014. The unobserved components discovery model they employ leads to results supporting that MtGox was the main determinant of the price discovery process until its end. Furthermore, the Bitcoin's information share was found to be high, reflecting the increasing Chinese interest in Bitcoin. Overall, the information share is found to have been dynamic and evolve over time. Baur et al. (2018) investigate whether Bitcoin should be best considered as a medium of exchange or an asset for speculation by adopting data from July 2010 to June 2015. They argue that it is not a safe haven due to its weak correlation with conventional assets, such as stocks, bonds and commodities in normal times but also during crises. Findings indicate that Bitcoin is mostly used for speculative trading rather than as a medium of exchange and a new form of currency. By another perspective, Hayes (2017) investigates the determinants for value formation of cryptocurrencies by employing cross-sectional analysis for sixty-six of the most used digital currencies. He reveals that the level of competition in the network of producers, the rate of unit production as well as how di fficult is the algorithm for mining are determinants of cryptocurrency value.

Bouri et al. (2017) examine how Bitcoin has a ffected global uncertainty during the period from March 2011 to October 2016, by employing theWorld VIX as a measure of uncertainty. The wavelet-based quantile-in-quantile methodology employed provides evidence that Bitcoin acts as a hedger against global uncertainty. This is valid especially in short investment horizons for both lower and upper quantiles of Bitcoin returns and uncertainty. Balcilar et al. (2017) utilize a Generalized Autoregressive Conditional Heteroskedasticity (GARCH) methodology and a non-parametric causality-in-quantiles test for examining the causal nexus between trading volume and Bitcoin returns and volatility. Volume is found to be able to predict returns but not volatility at any point of the conditional distribution. However, the former is not true in Bitcoin bear and bull markets. It is also found that non-linearities are important for explaining behavior in tails regarding the causal nexus of Bitcoin returns with trading volume. Bitcoin trading based on the volume-return linkage can prove profitable. Furthermore, Pieters and Vivanco (2017) investigate whether Bitcoin prices follow the Rule of One Price across markets by evaluating eleven Bitcoin exchanges from June 2014 to July 2015. Findings indicate that failure of this rule is connected with markets where no ID is needed to fund an account, as they exhibit larger price deviations. Thereby, know-your-customer regulations can lead to a sizable effect on Bitcoin market. Further exploration of determinants of digital currency prices includes Gandal et al. (2018). They investigate the e ffect of suspicious trading activity on the Mt. Bitcoin currency exchange, in which almost 600,000 BTC were fraudulently acquired. They argue that this activity led to the spike of BTC price in late 2013, as trading volume increased significantly on those trading days. They support that a lack of regulations in markets lead to vulnerability regarding manipulation by speculators today.

Feng et al. (2018) propose a novel indicator for estimating the informed trades ahead of events related to cryptocurrencies and provide evidence that informed trading takes place in the Bitcoin market before such events. The preference of such traders is revealed for taking their positions two days before large positive events and one day before large negative events. This triggers large profits from trading. Corbet et al. (2018) employ a generalized variance decomposition methodology to measure the direction and intensity of spillovers across Bitcoin, Ripple, Litecoin and other selected markets. There is evidence that cryptocurrency markets are interconnected with each other and present similar patterns of connectedness with other important asset categories. Furthermore, Panagiotidis et al. (2018) examine the determinants of Bitcoin returns from June 2010 to June 2017 by adopting a Least Absolute Shrinkage and Selection Operator (LASSO) approach. Evidence uncovers that uncertainty negatively influences returns whereas currency exchange rates as well as interest rates, gold and oil have positive effects. The predicted signs are found for information demand and mixed impacts are traced from stock markets. Empirical outcomes reveal that search intensity and gold returns are the most influential. Amid literature on price determinants is Kim (2017) who examines the transaction cost of Bitcoin from April 2014 to April 2015. He argues that the bid-ask spreads of Bitcoin exchanges are lower than those of the retail foreign exchange market. This cost advantage equals 5% and is due to the simpler infrastructure and the e fficiency that characterizes the Bitcoin market.

In another vein, Adhami et al. (2018) analyze 253 Initial Coin O fferings (ICOs) of cryptocurrencies from August 2014 to August 2017 in order to reveal the determinants of their success. They document that when the code source is available, a token presale is organized and tokens allow contributors to access a specific service or to share profits, then success is higher. Bouri et al. (2018) employ the Generalized Supremum Augmented Dickey Fuller (GSADF) test in order to study co-explosivity in cryptocurrency markets. They detect multiple periods of explosivity, especially during the 2017 bullish market and this is more intense concerning Bitcoin. Explosivity in a digital currency market is found to a ffect explosivity in markets of other virtual coins. Moreover, Bouri et al. (2019b) reveal that the market of cryptocurrencies presents time-varying herding behavior. High comovement is detected in the cross-sectional returns' dispersion across digital currency markets. This implies that mimicking of others' decisions takes place by investors in digital currency markets.

Schilling and Uhlig (2019) present a model of an endowment economy with the US dollar and Bitcoin constituting two competing but intrinsically worthless currencies. They support that fluctuations in prices are not harmful for the medium-of-exchange function of Bitcoin. Moreover, they support that if all economic units were impatient, no speculation opportunities would occur in the Bitcoin market. Giudici and Abu-Hashish (2019) adopt a correlation network Vector Autoregressive (VAR) process and provide evidence that Bitcoin prices from di fferent exchanges are highly interrelated and that prices from larger exchanges drive prices in smaller ones. Furthermore, they support that the inclusion of Bitcoin in portfolios results in diversification benefits. Ji et al. (2019) investigate the spillovers among six major virtual currencies by building positive and negative returns-connectedness and volatility-connectedness networks. They argue that leading digital currencies are interconnected and that Litecoin exerts a significant impact on Bitcoin and other digital coins of primary importance. Asymmetries in spillovers are also detected and are larger in negative-return spillovers than in positive-returns impacts. Kyriazis et al. (2019) examine whether high-capitalization cryptocurrencies are a ffected by Bitcoin, Ethereum and Ripple during bearish times by employing a number of Generalized Autoregressive Conditional Heteroskedasticty (GARCH) specifications. Evidence reveals that the majority of digital coins investigated are complementary with the three drivers of the market and that no hedging abilities exist.

Bouri et al. (2019a) examine the persistence in the level and volatility of Bitcoin prices by taking into consideration the impact of structural breaks. They argue for the existence of long-memory in volatility of Bitcoin and identify structural changes in the dynamics of this leading digital currency. Moreover, Bouri et al. (2019c) investigate the ability of trading volume of seven major digital currencies on predicting their returns and volatility. They provide evidence that trading volume triggers extreme negative and positive e ffects of all currencies under scrutiny whereas only influences the volatility of Litecoin, Nem and Dash. Ferreira and Pereira (2019) examine the contagion e ffects in markets

of virtual currencies by employing detrended cross-correlation analysis coe fficients. They provide evidence that the contagion e ffect between Bitcoin and other digital coins has been more intense and that the digital currency market has been more integrated during bearish times. Hyun et al. (2019) examine dependency among Bitcoin, Ethereum, Litecoin, Ripple and Stellar using a copula directional dependence (CDD) approach. They provide evidence that dependency from Bitcoin to Litecoin is the highest one and that dependency from Ethereum to the other four currencies is higher than the other way around. By their own perspective, Dastgir et al. (2019) investigate the causality between the Google Trends search queries and returns of Bitcoin from January 2013 to December 2017, by applying a Copula-based Granger causality in Distribution (GCCD) test. They document that a bi-directional causal nexus between Bitcoin attention and Bitcoin returns exists mainly in the left tail (bad performance) and the right tail (superior performance) of the distribution, but not in the central distributions. In a somewhat similar vein, Shen et al. (2019) use the number of tweets from Twitter as a measure of attention and examine their influence on Bitcoin. They argue that tweets constitute a significant determinant of the next day's trading volume and realized volatility of Bitcoin.
