University of Amsterdam, Amsterdam Business School
Master in International Finance
Master Thesis
What Influences the Prices of Cryptocurrencies from 2013 – 2018?
Eliza Permata Kristiani
Student ID: 11132302
September 2018
ABSTRACT
This paper aims to analyze what influences the price of cryptocurrencies by selecting 5 types of digital assets from the highest market capitalization (turnover) including the correlation between Bitcoin as the market leader compare to other digital assets selection, as well as comparing the prices of forks (2 selections). Conventionally, this correlation result attributed to a causal relationship between prices. These prices will be treated as endogenous variables and the control variables selected will be the foreign exchange currency USD/EUR, gold price, financial stock index S&P 500, Nikkei 225 and MSE.
This paper consist 4 aspects; first, this paper will study about the price correlation and causation between crypto currencies including the Bitcoin prices as the market leader correlation with other crypto assets. The second part will be analyzing the fork price correlate with their original of digital assets. Then, we will be adding the exogenous / control variables that have impact on the prices or the crypto market.
After the correlation and causal relationship are conducted, we will do some analysis of the event study by selecting some events that have the significant impact or changes on the regulation, technical or other announcement for the cryptocurrency news event.
Table of Contents
1. Introduction ... 5
2. Literature Review ... 7
2.1 Blockchain Technology ... 7
2.2 Cryptocurrency Regulations... 7
2.3 Security Risk of Cryptocurrenices ... 8
2.4 Other Factors ... 9
3. Research Questions ... 11
3.1 Control Variables ... 11
3.2 Macro-Financial ... 11
3.3 Cryptocurrency market trend/ attractiveness ... 13
4 Methodology ... 14
4.2 Augmented Dickey-Fuller Test – Unit root testing ... 14
4.3 VAR Residual Serial Correlation – LM Test ... 14
4.4 Granger Causality Test – (Augmented Granger Causality) ... 15
5 Data Collection and Preliminary Analysis ... 16
5.2 Descriptive Statistics ... 18
5.3 Correlation ... 19
5.3.1 Correlation on cryptocurrencies selected ... 19
5.3.2 Correlation on Forks ... 19
5.3.3 Correlation bitcoin and other variables ... 20
5.3.4 Correlation on popularity ... 20
6 Empirical Results ... 21
6.2 Augmented Dickey-Fuller Test – Unit root testing ... 21
6.3 VAR Residual Serial Correlation – LM Test ... 22
6.4 Granger Causality Test – (Augmented Granger Causality).. ... 22
6.4.1 Causality between Cryptocurrencies ... 23
6.4.2 Causality between Forks and the original of digital assets ... 23
6.4.3 Causality between Bitcoin and Exogeneous variables ... 24
6.3.4 Causality in Cryptocurrency popularity ... 25
7 Event Study Analysis ... 26
8 Conclusions ... 32
9 References ... 34
List of Figures
Figure 1. Internal and External Factors of Cryptocurrency Price...9
Figure 2. Time Series Pattern of Selected Cryptocurrency Assets and Control Variables...17
Figure 3.Timeline of an Event Study………...26
Figure 4. Event Selections with impact on Bitcoin Price 1...27
Figure 5. Event Selections with impact on Bitcoin Price 2...29
List of Tables
Table 1. List of Cryptocurrency Exchange Hacks Event...9Table 2. Descriptive Statistics of Selected Cryptocurrency and Control Variables...18
Table 3. Crypto Market Correlations...19
Table 4. Fork Correlations 1...20
Table 5. Fork Correlations 2...20
Table 6. Bitcoin vs Other Variables Correlations...20
Table 7. Cryptocurrency Correlation on Popularity...21
Table 8. Abbreviations Description...21
Table 9. ADF Test Result...22
Table 10. VAR Residual Serial Correlation – LM Test Result...22
Table 11. Granger Causality Test Result of Top 5 Selections of Cryptocurrencies...23
Table 12. Granger Causality Test Result Between Forks and the Original of Digital Assets 1...23
Table 13. Granger Causality Test Result Between Forks and the Original of Digital Assets 2...24
Table 14. Granger Causality Test Result Between Forks and the Original of Digital Assets 3...24
Table 15. Granger Causality Test Result Between Bitcoin vs Exogeneous Variables...25
Table 16. Granger Causality Test Result Between Bitcoin and Popularity…………...25
Table 17.Statistical Result of Events Selected on the Impact of Bitcoin Price...30
Table 18. Statistical Result of Events Selected on the Impact of Bitcoin Price in Multiple Days...30
Table 19. Statistical result of ME Event on The Impact of Bitcoin price...31
1. Introduction
Blockchain technology and cryptocurrency have received substantial attention given its innovative features, simplicity, and transparency. Its rapidly increasing usage and immense public interest in the subject has raised profound economic and society issues. Bitcoin (BTC) first was introduced in 2008 by Satoshi Nakamoto in the paper “Bitcoin: A Peer-to-Peer Electronic Cash System”. It is uncertain until now who is Satoshi Nakamoto or who is the inventor(s) of Bitcoin. The paper explains and introduce about the use of Blockchain on how is the main concepts of the system work as a safe electronic transaction system. It is the first of all an electronic payment system that based on cryptography that is the reason to be call the cryptocurrency or e-money does not mean that it has to have all the attributes of money, or it should work as a replacement of traditional fiat currencies. One study argues that Bitcoin has no intrinsic value but behaves more like a speculative investment than a currency because its market capitalization is high compared to the economic transactions it facilitates. The author also concludes that Bitcoin volatility adversely affects its usefulness as a currency. (Yernmack, 2013).
An empirical study conducted by researchers Michal Polasik and Anna Piotrowska, Radoslaw Kotkowski, Tomasz Wisniewski and Geoffrey Lightfoot (2014) found a positive link between media attention and the value of Bitcoin. The research team included stock market Schultz 13 fluctuations, the number of transactions conducted in Bitcoin and media appearances that help to tell the story of how popular the currency is. Moreover, Bitcoin started to become a fixture in world financial news till recently, few of Economist have explained that bitcoin is one of the things that are going to shape the future of finance and payment and compared them to digital gold. Ever since the market price of bitcoin has increase rapidly and have served more than 28 million users worldwide. 1
Due to the nature of Bitcoin and other open source projects such as the computer operating system known as Linux, it is possible that alternative versions of Bitcoin can arise. An occurrence known as a “hard fork” is what occurs when this split happens. A hard fork as it shares much of the code of Bitcoin. (Schulz, 2016) Basically, when a Blockchain diverge into two potential path forwards either with regard to a network’s transaction history or a new rule in deciding what will be the valid transaction is a ‘Fork’. There are many different types of forks – hard fork and a soft fork. For hard fork, it is a software upgrade that introduces the new rule that isn’t compatible with the older software. As for soft fork, by contrast, is any change that’s backward compatible. 2
Furthermore, on January 2018, total all cryptocurrencies reach the market capitalization of 795.83 billion USD3 with 40% came from Bitcoin4. Then drop significantly till the latest on April 2018 to 256.5
1 https://blockchain.info/charts/my-wallet-n-users
billion USD. Hence, the purpose of this paper is to analyze what influence the cryptocurrencies prices by looking at the correlation and causality relationship with the selected crypto assets prices including the forks from the original assets. This paper will also take into consideration of the macro economic impacts and stock market indexes whether to see if there have connections between them. In the following, we provide a brief introduction of other 4 crypto assets chosen for the analysis including 2 forks.
Litecoin (LTC) - Based on the same peer to peer protocol used by Bitcoin, Litecoin was created by
Charles Lee which is often called as Bitcoin rival as it features improvements over Bitcoin protocol with faster confirmation time for transactions. (Litecoin Project 2017).
Ethereum (ETH) - Ethereum is a cryptographic smart contract that stores information, processed inputs
and output that is only accessible if conditions are met. This contract contains code compromising and executed by the Ethereum network. Vitalik Buterin – a cryptocurrency researcher and programmer, proposed it in late 2013. Then, the system went live on 30th July 2015.
Ripple (XRP) –The Ripple network is a Blockchain network that incorporates a payment system, and a
currency system known as XRP that is not based on proof-of-work. A unique property of Ripple is that XRP is not compulsory for transactions on the network, although it is encouraged as a bridge currency for more competitive cross border payments (Ripple 2017). The Ripple protocol is currently used by companies such as UBS, Santander, and Standard Chartered, and increasingly being used by the financial services industry as technology in settlements.
Stellar (XLM) – Stellar is cryptocurrency used by the Stellar payment network. The founder of the
Mt.Gox Bitcoin exchange – Jed Mccaled, launched it. It was originally based on the Ripple protocol and it does not provide a desktop client and all transactions are performed thought its web wallet.
Ethereum Classics (ETC) – it is a forked from Ethereum, which is an open-source, decentralized
cryptocurrency platform.
2. Literature Review
2.1 Blockchain Technology
A Blockchain is essentially a distributed database of records, or public ledger of all transactions or digital events that have been executed and shared among participating parties. Each transaction in the public ledger is verified by consensus of a majority of the participants in the system. Once entered, information can never be erased. The Blockchain contains a certain and verifiable record of every single transaction ever made. (Berkeley 2017)
There are many potential benefits of the Blockchain technology that are more than just economic – they extend into political, humanitarian, social and anti scientific domains and the technology capacity of the Blockchain is already being harnessed by specific groups to address real world problem. In addition to economic and political benefits, the coordination, record keeping and irrevocability of transactions using Blockchain technology are features that could be as fundamental for forward progress in society. Every asset could become a smart property by encoding every asset to the Blockchain with a unique identifier that can be tracked, controlled and exchanged on the Blockchain. (Swan 2015). Furthermore, Bitcoin and the other cryptocurrencies are the digital currencies that use Blockchain technology that is highly controversial but the underlying Blockchain technology has found many functionality in both financial and non-financial world.
2.2 Cryptocurrency Regulations
With the arising of popularity of crypto market, regulators around the world need to arrange for instance how to treat this in tax system or what type of regulations to use. For both Financial Crimes Enforcement Network (FinCen) and Internal Revenue Service (IRS), they have different views on the digital currency. They stated that Bitcoin or cryptocurrency in general could be seen as a virtual currency but not as a real currency that are absence of a global framework. Based on fact the government issued is one of the most common pitfalls of investing cryptocurrency. Most of the government gave warning about the difference the risk of unregulated of the many of the organization that facilitate the transactions as well as the added risk from the high volatility of the price behavior.
Furthermore, there were some particular highlight events where the regulations can have impact on the crypto market and prices. For instance, the announcement in Japan who considered cryptocurrency would be legal resulting the price of Bitcoin up by 2% in just 24 hours or globally more than 100% for the next two months.5 Another announcement from China that they would shut down crypto exchange and ban ICO (Initial coin Offerings) plunge the price almost 30% in 24 hours. An ICO is a method of
raising money, like an initial public offering, where a company offers early investors some of the new cryptocurrency it is looking to launch in exchange for their financial backing (usually done via bitcoin, Ethereum or in some cases fiat currency). Investors buying the new digital money are buying what are referred to as tokens in a new cryptocurrency project. The new cryptocurrency can then be sold or bought on cryptocurrency exchanges if the offering is successful. (Karry 2018) Nevertheless, Despite of the decentralize ledger of crypto currency, there are many decisions and regulated issues that need to be evolve as they also have impacts on their values.
2.3 Security Risk of Cryptocurrencies
One of the main threats to cryptocurrency ability is to preserve its value to the holders is the issue of cyber security. All virtual currency must be held in computer accounts or virtual wallets, which often has a major problem or common target for hackers. Due to the fact that cryptocurrencies are digital bearer assets that once transaction has been done, it cannot easily be reversed unless the recipient agrees to do it. With the market growth recent years, it has made cryptocurrency exchange as a more target for criminals or cyber-attack since they handles and store large amounts of cryptocurrencies.
In the Table 16, it shows the last 5 years of cryptocurrency exchange hack. Some of the biggest attack in crypto market is including the attack of Mt.Gox in 2014. Mt Gox was one of the largest exchanges at that time and they revealed 850,000 Bitcoin at value $ 470 million had been stolen. Then, the most recent attack of Coincheck in June 2018. Coincheck – Japanese cryptocurrency exchange had been hacked and confirmed to have digital coins to be stolen that worth more than $ 500 million. It reported that 500 million of one of the digital assets NEM that worth 58 billion yen had been taken from customer virtual wallet. This event had been outweighing the Mt Gox breach in 2014. Hence, the security flaws and hacks have been a major concern for investors of cryptocurrency market which including hack or thefts of the exchange or their virtual wallet. When the breach of Coincheck happened, Bitcoin price dropped 7% against the dollars. Ethereum and Ripple dropped 5% and 12% respectively.7
6 https://coiniq.com/cryptocurrency-exchange-hacks/
Table 1. List of Cryptocurrency Exchange Hacks Event
Regardless of how security risk drives the cryptocurrency market and prices, this will be not taking further action for empirical analysis testing in this paper due to the limited data and sources.
2.4 Other Factors
Other than the two factors that have been mentioned in previous section – Regulation and security risk. According to Poyser (2017), he mentioned some other factors that drive cryptocurrencies prices that divided into internal and external factors that include supply & demand, crypto market (attractiveness, trend and speculations), macro-financial. See below the summary of the factors.
Figure 1. Internal and External Factors of Cryptocurrency Price
As an internal factor – supply & demand, cryptocurrency have controlled supply of coins settle by block height and block reward values, this is intrinsically related to the mining process. From this situation we can imply two things, firstly, crypto’s supply is exogenously determined and secondly, it is deflationary constructed. The problem was discussed by (Yermack 2013; Böhme et al. 2015; Garcia et al. 2014), and
the consensus is that it represents a serious drawback in its way to becoming a real currency according to the economic principles. Given that supply is deterministic, only the demand side that can affect crypto’s price (Ciaian, Rajcaniova & D. Kancs 2016; Kristoufek 2015; Baek & Elbeck 2014).
Moreover, the use of cryptocurrency in real transactions is tightly connected to fundamental aspects of its value. There are two possibly contradictory effects between the usage of the crypto and their price that might be cause by its speculative aspect. One effect stems from a standard expectation that the more often the coin has been used, the higher the demand and thus increase the price. (Kristaufek, 2015)
.
Besides the supply and demand, some authors have been studying the role of crypto as a safe heaven or hedge instruments. It examined the interconnection of commodity market including gold appropriateness as a weak safe-haven in the short run and as a hedge in the long run. (Bouoiyour & Selmi, 2016). Also the same stated on study from Kristoufek (2015) found one period that showed the correspondence amount the Financial Stress Index and Bitcoin’s price. Never the less, due to the limitation of the period and our focus on this paper analysis, the supply and demand will be not be investigate into deeper analysis.Another driver of the price of cryptocurrency is its attractiveness and the popularity. When there is more interest in the crypto market, there will be more demand as well thus increasing in prices. One of the biggest trigger influences on the trading or crypto market could be social media such twitter.
According to an experimental result from author Kim YB, Kim JG, Kim W, Im JH, Kim TH, Kang SJ, et al. (2016), the types of comment that most significantly influenced fluctuations in the price and the number of transaction of each cryptocurrency were identified. They also stated positive user comments significantly affected price fluctuations of Bitcoin or vice versa for negative comment.
With those saying, we will take into consideration of these factors as our control variables, which we will discuss more in the next section. We will examine short and long run factors that influence cryptocurrencies from 2013-2018.
3. Research Questions
Q1: Are there any correlation and causality in the price movement between Bitcoin and other 5 selections of cryptocurrencies?
Q2: How forks influence the original digital assets price rise and fall?
Q3: Does Bitcoin as the market leader prices have any correlation and causality relationship with Macro-financial such stock markets (NIKKEI 225, S&P 500, MSE), exchange rate (USD/EUR), and commodity (GOLD)?
Q4. How cryptocurrency market popularity correlates with the price?
3.1 Control Variables
3.2 Macro-Financial
As it previously stated from author Poyser (2017), other than the supply and demand variables, Marco – financial factors are also have impact on the price of Bitcoin or cryptocurrency. In a recent analysis from Bouri, Gupta, Tiwari, & Roubaud (2017) they have been studying whether Bitcoin can be a hedge or safe haven asset under market uncertainty scenarios, the authors found that Bitcoin acts as a hedge since it reacted positively to great ups and down financial movements, especially in the short run. Peter Fortune from Finance system who is the director of research at Boston FED stated that there is a comparison between stock market and Bitcoin.
Hence, for this section we will choose some example of the financial stock index from country that have been talking about implementing cryptocurrency in their countries such Japan, U.S.A, and Malta. Also some other economic fact of commodity – Gold and currency USD against Eur. Due to the fact that some author are try to comparing Gold with cryptocurrency assets or crypto it is as a digital currency. By choosing those we could see the result that can affect or lead to causation of these economies and cryptocurrency market. However, since there is a very limited academic article about this topic, we will use or able to find articles from the economical website.
3.2.1 NIKKEI225
In Asia, crypto currencies used to account for most of the volume in China or Chinese Yuen. Then the regulation started to clamp down on the digital currency exchanges in China and Japan Yen seems took over as the biggest trading pair as regulators adopting digital friendly rules. Moreover, the Nikkei 225
index is an industrial average that takes into account 225 corporations in Japan on Tokyo Stock Exchange. As this being said, this index could be an indication of economic performance in Japan and several other Asian countries for this paper analysis.
3.2.2 S&P500
Daratrek – an economic and financial consultancy company have analyze the relationship of Bitcoin and S&P 500 since January 2016 from period of 10 days, 30 days and 90 days. The result on the analysis is that there is a strong connection between two variables especially in the shorter period, while the longer the period the lower the correlation % ratios. The CEO of Federal Reserve Bank (FED) stated that the Fed is exploring the idea of its own digital currency and would be premature to estimate when the Fed would come up with its own offering. Cox (2017)
3.2.3 MSE
On July 2018. The Maltese Parliament has officially passed 3 bills into law, establishing the first framework for Blockchan and cryptocurrency that makes Malta as the first country in the world to provide an official set of regulation for Blockchain and cryptocurrency8. This news will represent and opportunity for industry player both ICOs and the crypto exchanges. One of the latest news Binance – the world seconds largest crypto exchange that has seen $ 923.3 million9 trades over 24 hours is making plans to create a Blockchain based bank with tokenized ownerships in Malta.
3.2.4 GOLD
Author Dyhrberg (2015) situated the hedging capability of Bitcoin somewhere between gold and the US dollar, then another author (Popper,2015) mentioned when some investors lose trust to mainstream currencies or to the entire economy, they might look for other alternative which is one of the reasons why Bitcoin or cryptocurrency has sometimes been called digital gold. Some economist have compared Bitcoin to gold as they have some familiarities that gold has some intrinsic value but it most likely does not justify its current market value while Bitcoin have some intrinsic value if its users are rational (Coindesk 2015). Also, several independent operators and companies mine both assets and neither of them has nationality or is controlled by a government (Dhyrberg 2015).
8 https://www.forbes.com/sites/rachelwolfson/2018/07/05/maltese-parliament-passes-laws-that-set-regulatory-framework-for-blockchain-cryptocurrency-and-dlt/#53148a5a49ed
3.2.5 USD/EUR
The rise of cryptocurrency could have a significant impact on monetary systems of countries as they are privately issued currencies, this not regulated by central banks. As of Bitcoin is raising the news and going into a global trend, government are curious about what will the future of finance world that it creates some plans about to adapt cryptocurrencies in their economies (Chen, 2017) the user and volume trading of cryptocurrencies getting stronger in many countries such as Japan, USA, South Korea, and other European or South America. The same as gold, one author Whelan (2013) has argued that Bitcoin or cryptocurrency is similar to the dollar. They both have no or limited intrinsic value and are used primarily as a medium exchange. The main difference is that a government regulates the dollar while the cryptocurrency is private money introduced by private sector.
3.3 Cryptocurrency market trend/ attractiveness
As it was previously stated, it is difficult to define Bitcoin or cryptocurrency since it has several capabilities, nonetheless, an attractiveness proxy predominantly addresses the payment method and investing asset, this study will include such variables. In most cases authors have rely on Google search trends and Wikipedia articles views (Kristoufek 2015; Glaser et al. 2014), Twitter sentiment analysis (Kaminski 2014; Georgoula et al. 2015) and online communities reactions (Ciaian, Rajcaniova & D. Kancs 2016; Dwyer 2015; Kim et al. 2016). Among all the variables in the studies, the attraction has the most relevant variance explanation power.
From the research of Barber and Odean (2013), there are two main factors of the extension to financial market to investing; delayed reaction to get the crucian information and overstated attention to specific news or information that could lead to overreaction and impulsive decision. Social judgement is intrinsic to cryptocurrency market since the valuation of any currency is contingent to the extension of the group that founds it valuable (Poyser, 2018)
4 Methodology
4.2 Augmented Dickey-Fuller Test – Unit root testing
When we want to analyze the relationship between the crypto currencies market and the control variables, it is important to detect series stationary status. The non-stationary will have an effect on the process and will result a bogus regression result. Therefore, we will perform the Augmented Dickey – Fuller test (1979) in the first step for testing unit roots.
For the null hypothesis, it implies that there has unit roots and series are non-stationary. Then, the alternative hypothesis is no unit roots and series are stationary, if it’s equal to zero then we cannot reject the null hypothesis and conclude the series in non-stationary and unit root exist which implies to do the next testing.
4.3 VAR Residual Serial Correlation – LM Test
In order to Asses the validity of the specification of an econometric model, it is useful to have a variety of diagnostic statistics, which provide evidence on the existence and possibly the type of
misspecification involved. The Lagrange Multiple (LM) test is to examine the fit of the model under the null for evidence of departures in the direction of interest.
Moreover, these LM test can be thought of as ways of examining the residual of a model for specific types of non-randomness. (Engel, 1982). Therefore, for checking whether the residuals are
autocorrelation or not, we will use the LM test.
The null hypothesis is no autocorrelation up to specified lag and the alternative hypothesis is there is autocorrelation existed up to specified lag. If the p-values of t-statistic are higher than 5% of the critical p-value, we don’t reject the null hypothesis. Therefore, there is no autocorrelation between variables.
4.4 Granger Causality Test – (Augmented Granger Causality)
Granger does not meant one variable at the reason for another variable rather by doing this test if will shows or prove that change in one variable is followed by change in another variable. Therefore, this test is basically to understand how one variable is determined to another variables. The application of Granger causality assumes that the analyzed signals are covariance stationary. (Hesse et all, 2003)
So in this case, we first need to test the ADF test, then we need to establish VAR model by using level data depending on how correlagram (trend or intercept or no trend) of the data. After that, we will run the appropriate maximum lag length for the variables in the VAR model. For this, we will follow the information criteria of LR, AIS, SIC and HQ to choose how many lags we should use. At the end we will run the test for Granger Causality. The EVIEWS runs bivariate regression as below:
Where the Yt and Xt are the stationary time series, εt and Ut are residuals and the constant term is α0
and β0. So for instance, in our analysis Yt is the Bitcoin price, the Xt will be the logs of another variable
that we try to test the causality.
Other than that, we will create a dummy variable as in time frame of 3 years, 1 year, and 3 month. Due to the launched data of the forks that are very recently a year ago, for the Forks analysis the time frame will be 1 year and 3 months.
5 Data Collection and Preliminary Analysis
For our analysis we use Augmented-Dicker Fuller, Autocorrelation and Granger Causal relationship, the daily closing prices for the following: Financial stock index – Nikkei Index (Nikkei 225) which is a stock index for Tokyo Stock Exchange and the currency in Japanese (JPY), S&P500 (GSPC) which is a stock index for the United States and the currency is in US Dollar (USD), and Malta Stock Exchange (MSE) and the currency is in EUR, for Commodity we choose the Gold price then currency exchange of USD/EUR both are in USD.
The time analysis period is 5 years from 2013-2018. However, due to the date launch of some digital assets, we used the data from the date of their existence and availability. For all crypto currencies assets we took the data from coinmarketcap.com as for the rest we took the data from Yahoo Finance.
Moreover, since the price of crypto currencies assets are in USD, we change different local currency with the exchange rate to USD for the empirical procedure. Also, in order to reduce the potential effects of heteroscedasticity, we convert data of the prices into logarithm form. All the empirical procedure is applying by using EVIEWS.
Nevertheless, for the trend or popularity topic, we will use the Google Trends that is a feature of Google search engine that illustrates how frequently a fixed search term was looked for Bitcoin and cryptocurrency. By using this, it will allowed you to compare up to five topics at one time to view their popularity along with the selection of which countries or worldwide who developing the popularity over time. This part analysis will just use for a general analysis for the market trend on how cryptocurrency news have been increasing or decreasing for the last 5 years due to the fact of the limitation of data provided from Google Trend.
Figures 2 below of the time series of the data for the 5 selections of crypto currencies including the forks (2 selection). We include as well the pattern from the exogenous variables in our analysis such stock index (NIKKEI 225, S&P500, and MSE), commodity (Gold), and monetary (USD/EUR currency rate). Sample period are from July 2016-2018. We selected from 2016 in order to take compare from all the period when all selected forks have already been launched in the market.
Figure 2. Time Series Pattern of Selected Cryptocurrency Assets and Control Variables
As shown in figure above, we could see how the daily prices of the crypto market compare to our selection of control variables are extremely high volatility. Although the graph above doesn’t take into consideration from the year 2009 – the first introduction of Bitcoin, the price have shown the volatility. Then, the series shows how market is quite steady from a year 2016. There might be some ups and down but till the beginning of the year of 2017 you could see how the pattern increases suddenly. Visually, the line shows a turning point from beginning of 2018 and it keeps getting down, as it will back as the original value in 2016. This is due to the fact that cryptocurrency market starts becoming better known in society, including one of the most trending topics in recent financial issues and economics. When the increasing number of news appears in society, it will give more attraction to
people. The more good news and speculative, the price level increasing to a market for taking profit, the individual start selling their assets or for instance the more bad news, the less confidence of the investor and triggered to wide selling the investment – panic selling. With this events happening, these could cause a market correction.
In addition to that, the fluctuation of other variables such the financial index, gold and exchange rate between US dollar and euro compare to crypto market has not reach that peak of the decline or incline that show in the graph. It is clearly that crypto market is more volatile compare to the other variables choose. This volatility is a threat for crypto market to be accepted as global currency due to their inability to preserve stable value over period. However, cryptocurrency is also a relatively new currency that the value of the assets or its price formation is not yet well understood. (Kristoufek 2013)
5.2 Descriptive Statistics
Table 2. Descriptive Statistics of Selected Cryptocurrency and Control Variables
From the descriptive statistics table above, it shows the summary statistics of the return series for all variables. For each price index we calculate return as the first difference of the logarithm of closing price. The mean result describes the average value in the series while the standard deviation measures the dispersion or spread of the series. The maximum and minimum statistics measures upper and lower bounds of the variables under study during our chosen time span. Bitcoin and the rest of crypto assets have the highest daily mean and volatility. Other than that, compare to other financial index and currency exchange, the crypto market have all positive skewness compared to the rest that have lower to 0 or it is a negative skewness.
5.3 Correlation
In this part, the main purpose it to understand the correlations between the variables. By having this background, we could use the correlation result and interpret it in an intuitive way to add to the analysis on the subject. Then, the next section will establish the causal relationship.
5.3.1 Correlation on cryptocurrencies selected
By comparing the correlation matrix below between other 5 selections of cryptocurrency, first we could see the comparison for 3 period which it clearly see that all of them shows a positive correlation. Although some of them show a correlation but we could see some are lower and higher than the other. For example: XRP with other cryptocurrency that still show a positive number but they are very low correlated compare to Bitcoin and Litecoin.
Table 3. Crypto Market Correlations
Then, if we look at the shorter period – period 1 year or 3 months, the result are showing a high correlated between each cryptocurrency that close to 1 apart from XLM and XRP result that is still far from 1 but showing a positive correlation.
5.3.2 Correlation on Forks
According to the table below, for the correlation on forks in highly correlated they are the shorter the period including between Bitcoin and Bitcoin Cash or Ethereum and Ethereum Classic. The same goes if we try to compare the Bitcoin as the market leader of the cryptocurrency the result still show the positive correlation between each variables for 1 year or 3 months period. Kharpal (2017). The price of bitcoin took a hit after the cryptocurrency underwent another split with the newly created bitcoin gold seeing its value plunge over 60 percent within 10 days.
Table 4. Forks Correlations 1 Table 5. Fork Correlations 2
5.3.3 Correlation bitcoin and other variables
Another correlation check is to include the other variables that stated in previous section that could be the price influence of the cryptocurrency including the stock exchange, gold, and currency exchange. The result appears to be for 3-year period that Bitcoin are negatively correlated with GSPC, Gold and USD/EUR while compare to Nikkei and MSE there is a positive correlation. However, for the shorter period 1 year and 3 months, the results show a positive correlation for GSPC and currency exchange USD/EUR. Bouri, Molnar, & Hagfors (2016) conducted the analysis in daily and weekly data for the diversification of hedging between Bitcoin against other assets, the overall result shows Bitcoin can serve as an effective diversifier for most of the cases.
Table 6. Bitcoin vs Other Variables Correlations
5.3.4 Correlation on popularity
Gervis et al (2001), stated potential investor’s decisions may be affected by an increase or decrease of attention in the news. Lee (2014) mentioned some evidence for Bitcoin that by the alteration of positive and negative news generated high price cycles. As a result, this implies the driven investment behavior can affect the crypto market whether it negative or positive. In order to analyze more about this, we also calculate the correlation between the Bitcoin price and the news search. Below shows that it seems both positively correlated. From the figure below, we could see a strong correlation between the Bitcoin price and the trend no matter what is the horizon period – 5 year, 1 year or 3 months.
Table 7. Cryptocurrency Correlation on Popularity
6 Empirical Results
The table below is the description of the abbreviations that we use for the statistics testing analysis.
Table 8. Abbreviations Description
6.2 Augmented Dickey-Fuller Test – Unit root testing
According to Augmented Dickey-Fuller test (1979), in order to test the unit roots, we know that the null hypothesis exists unit roots in series and the alternative hypothesis is no units’ roots in series. Kristoufek (2013) study analysis about Bitcoin meets Google trends and Wikipedia, to cover various combinations of relationships; he initially studies standard transformation of the original series. He used the ADF test to form an ideal pair of for stationary vs unit root testing.
Therefore, we perform all ADF tests at level difference. As we can see from the results, all the p-values are smaller than the critical value – 5%. In this case, we can reject the null hypothesis, which this data indicated the series are stationary.
Table 9. ADF Test Results
6.3 VAR Residual Serial Correlation – LM Test
Besides checking the stationary of the data, we also perform the Residual serial correlation – LM test in order to check the autocorrelation based on the specified lag of variables. According to Dyhrberg (2015), Bitcoin like other financial assets is sensitive to certain shocks and may have positive time trend and shows clear non-stationary by using LM test.
So for the null hypothesis, the result is no autocorrelation up to specified lags while for the alternative hypothesis is autocorrelation up to the specified lags for variables. For the lag order selection criteria, we will run it in Eviews and determined based on likelihood ratio to choose the lag specification for the model. Hence, if you see the table below, all the p-values are higher than the critical value (5%) which we cannot reject the null hypothesis thus no autocorrelation.
Table 10. VAR Residual Serial Correlation – LM Test Result
6.4 Granger Causality Test – (Augmented Granger Causality)
According to Kristoufek (2013) research, by using Granger causality was tested for several variables, as the dependent variable that might also have en effect on the independent variable. With this test, he found such two-way relationship with both Wikipedia and Google trend search queries. Below we will see how the causal relationships for our analysis are.
6.4.1 Causality between Cryptocurrencies
In the case of below, the empirical results finds there is a granger causality between Bitcoin and other cryptocurrency selected for the last 3 years, 1 year and 3 months but for the rest of the results it shows the p-value is higher than the critical value (5%) resulting we can not reject the null hypothesis thus no causality.
Table 11. Granger Causality Test Result of Top 5 Selections of Cryptocurrencies.
6.4.2 Causality between Forks and the original of digital assets
For the causality between Forks and the original of digital assets, first we will look at the comparison between Bitcoin and Bitcoin Cash. The p-value here are smaller than critical value (5) so the null hypothesis need to be rejected and to conclude that there are causality between two of them including in 1 year period or 3 months period.
Table 12. Granger Causality Test Result between Forks and the Original of Digital Assets 1
As for the second comparison, we could see that the p-value for Ethereum granger to Ethereum Classic is below the critical value. Therefore, we need to reject and there is causality for 1-year period. However, for the rest the p-values show as higher than 5% and we cannot reject the hypothesis thus no causality between variables.
Table 13. Granger Causality Test Result between Forks and the Original of Digital Assets 2
Last but not least in this part is to compare Bitcoin as the market leader with the other two of the forks. The result shows that for 1-year period, there is causality between Bitcoin cash and Bitcoin since the value is below 5% and for the rest of variables it seems that there are no causality because the p-value that is significantly higher than the critical p-value. So we cannot reject the hypothesis that is no causality. However, for the shorter period – 3 months, the testing result appear to be different. There is no causality for Bitcoin Cash and Bitcoin, nor Bitcoin and Ethereum Classic and BitcoinCash with another fork Ethereum Classic. This is by the fact that the result of p-value is higher than the critical value. As for Bitcoin to Bitcoin cash, Ethereum Classic to Bitcoin and Etherum Classic to Bitcoin Cash
Table 14. Granger Causality Test Result between Forks and the Original of Digital Assets 3
6.4.3 Causality between Bitcoin and Exogenous variables
According to the requirements, we find that no Granger causality between Bitcoin price and stock index or gold or currency exchange for last 3 years, 1 year and for last 3 months. The reason is p-values are higher than 0.05, we cannot reject the null hypothesis. We find no Granger Causality relationship between Bitcoin and other exogenous variables.
Table 15. Granger Causality Test Result between Bitcoin vs Exogenous Variables
6.3.4 Causality in Cryptocurrency popularity
By looking at the table below, for the last 5 years, 1 year and 3 month shows that there are causality between Bitcoin Price and the popularity of the Bitcoin itself. The probability of the testing has shown a result with p-value is smaller than the critical value. Therefore, we reject the null hypothesis.
Table 16. Granger Causality Test Result between Bitcoin and Popularity.
Regardless, Kristoufek (2013) argues that the hording effect dominates the market. According to him, the market forces of Bitcoin or other cryptocurrency supply and demand allowing for setting a fair price are missing, rather its price is driven by the investors faith in the future growth and is dominated by short-term investors, trend chasers, noise traders and speculators. This argument is also in line with the sceptics on BitCoin (e.g. Velde 2013; Hanley 2014; Yermack 2014) who argue that hoarding is one of the key weaknesses of BitCoin as a currency (alongside the security problem) compared to standard currencies.
7 Event Study Analysis
MacKinlay (1997) argues in his study that the abnormal return of a security around a specific event can be interpreted as the event’s impact on the stock price. Based on the historical returns, it will then get the expected return and the normal return. An actual return that significantly differs from the normal returns is an event-driven impact proof. Therefore, in order to have a deeper analysis in this paper for the event impact on the crypto market, it is absolute necessity to have a separate event study analysis. The conducted event studies in this paper will focus on the price movement of Bitcoin as it one of the leader of the cryptocurrency market. Below is the formula that MacKinlay use for the event studies. In order for an event to have a significant impact, the t-value of its abnormal return during the event – considering a 95% confidentiality, the event window has to be greater than 1.96.
Regardless, the t-value is define as below:
The ARit is represent the abnormal return, Rit is the actual return and E (Rit | Xt) is the normal return from the time period t. while Xt is the conditioning information for the market return, we will use the return of MSCI World Index for a comparison. In order to avoid the overlap data price of the MSCI World Index price that held every 5 days a week instead of 7 days as crypto market, we will exclude the weekend (Saturday and Sunday) price of Bitcoin in our analysis.
Furthermore, since the AR it only represent to one event on specific time, these abnormal returns must be aggregated. Then, this aggregation over time will conclude across event and time about the general indication on the event-driven impact that inform us the total impact of the event on the Bitcoin price or crypto market. The timing sequence will be represent as below.
Based on MacKinley theory, the event window is the period that the abnormal returns calculated with the comparison of the normal returns while the actual event is in the middle of the event window. The previous study were mostly use 31 trading days on the event window and 15 days of before and after the event. As from what we have discussed in the previous section analysis, the crypto currency market most of the time reacts in the short term to specifics events. Hence, we will take for the event window 2 days for pre and post event. Due to the fact, it seems it is difficult to analyze or capture the specific events that impact the price because of the overlapping event that occurred as well as the short-term reactions on the market if we select more days in the analysis. Regardless, we will also take one example of the macroeconomics situation and take 21 days of window event, since this event impact usually will take not as a short term.
For choosing the selected events, these were chosen based on how significantly the prices or news impact on the crypto market. Other than the event that has significant impact, we also taking a wider perspective and focus on other possibilities that effect in the society or market. For instance; the risk security or technical events which related to issue of exchange markets who got hacked, the regulations and macroeconomic/political events, or the news event such announcement or launch that got more market or people attention. See below the chart and more detail of the event selection. We in this case will examine how the relation behaves in different frequencies and time.
Figure 4. Event Selections with impact on Bitcoin Price 1
Event 1: 20th November 2013. The deputy governor of the PBOC (People’s Bank of China) and director of the State Administration of Foreign Exchange – Yi Gang commented that it would be
near future. But he added that he will adopt the perspective in the long term and people are free to participate in Bitcoin market.
Event 2: 5th December 2013. After the surging of Bitcoin popularity in China after the announcement in November 2013, China restricted its banks from using Bitcoin as currency as concerns of the finance stability and money laundry.
Event 3: 7th February 2014. Mt Gox foreign exchange currency has been hacked. As it stated in previous section, as one of the largest exchanges at that time the value of Bitcoin stole was 850,000 Bitcoin at value $ 470 million. This certainly had an impact on the market.
Event 4: 11th November 2014. Microsoft made informal announcement that they will accept Bitcoin as a payment to buy apps, games, and videos from online stores with the crypto-currency. The news raises the Bitcoin price at that moment overnight (Sparkes, M, 2014)
Event 5: 24th June 2015. The New York State Department of Financial Services (NYDFS) had officially adopted the BitLicense, which means for all business who run the digital business currency should submit the details and financial and legal histories in order to obtain the license.
Event 6: 17th September 2015. The US Commodity Future Trading Commission (CFTC) for the first time has ruled about Bitcoin and other digital currencies are commodities subjects.
Event 7: 4th March 2016. The Japanese government announce if they approved the use of Bitcoin as Monday on par with the currency of Japanese Yen itself. The cabinets also take into consideration the rising of virtual currencies or the potential of digitally transferred which would help the banking sector expend their information and technology businesses.
Figure 5. Event Selections with impact on Bitcoin Price 2
Event 8: 8th November 2016. The US President Election – the Trump effect. Bovaird (2016). During the controversial Donald Trump wont the 8th November presidential election, the price of Bitcoin had experienced swings in the days leading up to the US vote which had gone as high as $744.28 on 3rd November then to $680 mark later that day. Then, it rose more than 4% on the 9th November.
Event 9: 3rd August 2017. The announcement of Bitcoin Cash launched. The market impact on the launch was the Bitcoin price raised $2,925.03 on that day. 10
Event 10: 15th September 2017. China is shutting down all of Beijing Bitcoin and cryptocurrency exchange. Once the regional of Chinese authorities have ordered Beijing to do an immediate closure of the cryptocurrency exchange, the Bitcoin price went down for 5%. 11
Event 11: 31st October 2017. CME announce to launch Bitcoin Futures. Ozsoy (2017). The cryptocurrency market raised as much as 3.7% after the announcement the CME group to introduce Bitcoin Futures by the end of the year.
Event 12: 11 December 2017. CBOE (Chicago Board Options Exchange) Bitcoin futures are
launched. After the announcement, Bitcoin price was up to 10% that day – trading at $16,200. It was expected to even higher in January 2018. (Treanor, 2017)
Event 13: 31st January 2018. Coincheck Exchange that based in Japan confirmed coins that have been stolen worth around $530 Million. Bitcoin price initially dropped 7% against the dollar, while Ethereum and Ripple were down for 5% and 12% respectively. 12
Event 14: 4th July 2018. Malta establish first regulatory framework of Blockchain, Cryptocurrency and Distributed General Ledger.
By aggregating the specific events above using the average abnormal return on one single day, the table below shows the result of our analysis:
Table 17. Statistical Result of Events Selected on the Impact of Bitcoin Price
Based on the result, it is clearly indicate the immediate reaction once the news was announced with significant abnormal return toward it. For the good news event, we could see on the pre and post event the t-statistic result are higher that 1.96 which shows the Bitcoin user behavior that might confidence over the news and buying more of the assets. As for the bad news event, the few days before the event the result shows the negative result but not a significant abnormal return toward it but then during and after the announcement, the Bitcoin behavior indicate the direct impact on the price of Bitcoin.
Moreover, we take also consideration to analyze the outcome when we tried to take the sum of abnormal return (AR) in multiple days. From Table 17, you could see we divide the multiple days in the time period criteria. The outcome shows us only when the multiple days of -2 days to after 2 days of the event it has a significant impact on the price for the good news. The same goes for the bad news event, 2 days before and after the events, the result indicate the significant impact on the price of Bitcoin.
Table 18. Statistical Result of Events Selected on the Impact of Bitcoin Price in Multiple Days
Popper (2015). Bitcoin is an alternative to mainstream currencies and it’s often even considered as part of an alternative economy. If some investors lost trust to the entire economy, they might resort to Bitcoin. According to Luther & Salter (2017), following the Cyprus bailout announcement and in particular, depositors are more likely to turn to invest in Bitcoin. Therefore, in this case we would like to analyze one selection event that relate to the macroeconomic – Event 8.
Table 19. Statistical result of ME Event on The Impact of Bitcoin price
The figure in table 18 above shows the result that on the day of the announcement was not directly impact the price only the day after. Then, the on and off pattern of the return after the day of the announcement might be because of some investors that try to buy and take advantage while the price is in the dip and try to profit taking. There is also a possibility there are some investor still waiting for more sources and evidence. As for the day before the announcement, no conclusion can be drawn.
Last but not least, for the macroeconomic event analysis, we also take the consideration to analyze the result on multiple days. The table below clearly indicate which multiple days of time period in our calculation, and the result shows none of them have the significant impact on Bitcoin price.
Table 20. Statistical result of ME Event on The Impact of Bitcoin price in Multiple Days
8 Conclusions
This paper analyzed what are the influences factors of cryptocurrency for the last 5 years. At first our analysis is to see the correlation between each variable of the crypto market and also the selected of exogenous variables. The correlation results between each other of the crypto assets shows a highly correlated in the short term but it decreasing by the increasing of the period. As for the forks, it seems that for 1 year or short term of 3 months, the result appears to be strongly correlated. However, the long-term period in this case was taken as 1 year because of the recent launch of the fork. Perhaps in the future a new run analysis needs to be conducted.
For another correlation result with the exogenous variables, the result indicates that no correlation on the long term while for the period of 3 months is weakly correlated. Another factor with the trend, highly positive correlation had shown. This might be obvious since the more result in the trend; it might increase the demand.
Moreover, after the correlation test were conducted, we then checking the causal relationship between them. At first, we were checking whether the data is stationary as well as the auto correlation. Then, we continue our analysis using the Granger Causality test. According to our result, there is a causality relationship between each digital asset no matter the period zone is long or short term. When it comes to the original and fork asset, we found that there is some causality between the variable; for instance Bitcoin with Bitcoin cash then Ethereum with Ethereum Classics. We also test the causality between Bitcoin with the exogenous variables selected. The results clearly indicate that there is no causality between them for the long term or short-term period. Although the first analysis of the correlation shows a positive result in short term period. The last causal test that was conducted is with the trend and it is indeed obviously shows there is a causal relationship between them.
Furthermore, in this paper we also analyzed the impact of events to the Bitcoin price in event study analysis. On a single day analysis, the result revealed a direct impact for the good news on the day of announcement and the prior 2 days before or after the event. While for the bad news, it impact only after the announcement. When we analyze the event study on multiple days, the outcome shows us the impact on good news is significant for 2 days before and after the event and for the bad news 2 days or 1 days before or after the event it also impact the price of Bitcoin or crypto market in general. For
Macroeconomic event, based on the result on single days we cannot draw the conclusion whether it has the significant impact or not since the result shows the on and off pattern of the t-value. However, on multiple days we could conclude according to the result that there is no significant impact on the price.
After all the testing and analysis were conducted, we conclude that the cryptocurrency or blockchain might be the new phenomenon that is not yet establish by the whole society the limited information and missing of valuation on each assets or to understand what are the actual values of each crypto are still unclear which create the speculative investment in the market.
Besides, when it seems that there is an uncertainty in macroeconomic situation such politics, people or investors in general tend to search for another alternative to hedge their assets. They might find the cryptocurrency as another alternative to leverage their portfolio or as an asset class that might bring a valuable hedging mechanism.
Apart from that, the trend and popularity through the Google search or social media of cryptocurrency had a positive or negative impact on the prices. We could align this result to the behavioral finance theory of the fear of missing out (FOMO) so for instance when new people are talking or buying the cryptocurrency investment; they tend to look more information online with not really a professional investor management experience or background. Without having a financial or investing background, these people are easily impacted by news, social media, word of mouth or any rumors and behavioral biases that could increase or decrease the prices significantly. When it comes to speculation in investment that is more likely driven by individual investors this is called noise traders. Based on behavioral finance research (Kumar & Lee, 2006), noise traders are behaving according to less rational factors such fear or greed that tend to make impulsive decision. Taking also consideration how decentralized or opens the crypto exchange that everyone can register in one of his or her exchange and start doing the trading and increasing the purchasing price.
Never the less, as it also stated in previous section, Bitcoin and cryptocurrency are relatively new topic for academic research that causes the limitation in this study. We believe that this paper analysis could be improve if we could empowering the adaptation of the cryptocurrency from the legalization or regulation of cryptocurrency as a payment tool or the security risk that impact the market.
9 References
Nakamoto, S. (2008). Bitcoin: A Peer-to-Peer Electronic Cash System. Unpublished manuscript. Retrieved at http://pdos.csail.mit.edu/6.824/papers/bitcoin.pdf.
Schut, M. (2017). Bitcoin analysis from an investor’s perspective: insight into market relations and diversification possibilities. Erasmus School of Economics.
Akben-Selcuk, E. (2016). Granger Causality Between Stock Prices and Trading Volume: Evidence from Turkey. Proceedings of the 22nd International Academic Conference.
Newey, W., and West, K. (1987). A Simple, Positive Semi-definite, Heteroskedasticity and Autocorrelation Consistent Covariance Matrix. Econometrica, 55(3), 703-08.
Brill, A., and Keene, L. (2014). Cryptocurrencies: The Next Generation of Terrorist Financing? Defence Against Terrorism Review, 6(1), 7-30.
Brière, M., Oosterlinck, K., & Szafarz, A. (2015). Virtual currency, tangible return: Portfolio diversification with bitcoin. Journal of Asset Management, 16(6), 365-373.
Russo , C, Migliozzi, B, Sam, C, (2017) This is where people are buying Bitcoin all over the world Retrieved at https://www.bloomberg.com/graphics/2017-bitcoin-volume/
Dyhrberg, A. H. (2016). Hedging capabilities of bitcoin. Is it the virtual Gold? Finance Research Letters, 16, 139-144.
Granger, C. W. (1969). Investigating Causal Relations by Econometric Models and Crossspectral Methods. Econometrica. 37(3),424–438
Quantstart (2016). Johansen Test for Cointegrating Time Series Analysis in R. Retrieved June 20, 2016. from https://www.quantstart.com/articles/Johansen-Test-for-Cointegrating-Time-Series-Analysis-in-R
Michal Polasik and Anna Piotrowska, Radoslaw Kotkowski, Tomasz Wisniewski and Geoffrey Lightfoot,(2014). Price fluctuations and the use of Bitcoin: an empirical enquiry.
Yermack, D., (2013). Is Bitcoin A Real Currency? An Economic Appraisal (No. w19747). National Bureau of Economic Research.
Kim YB, Kim JG, Kim W, Im JH, Kim TH, Kang SJ, et al. (2016) Predicting Fluctuations in Cryptocurrency Transactions Based on User Comments and Replies. PLoS ONE 11(8): e0161197. doi:10.1371/journal.pone.0161197
European Central Bank, (2012). Virtual currency schemes. Preprint. https://www.ecb.
europa.eu/pub/pdf/other/virtualcurrencyschemes201210en.pdf.
Wallace, B. (2014). The rise and fall of Bitcoin. Wired Magazine. Retrieved from http://www.wired.com/2011/11/mf_bitcoin/
Castor, A.(2017). A Short guide to Bitcoin fork (Retrieved from https://www.coindesk.com/short-guide-bitcoin-forks-explained/
Adam S. Hayes.(2016).Cryptocurrency value formation: an empirical study leading to a cost of production model on valuing bitcoin.
Novak, M.(2018). The implication of Blockchain for income inequality. file:///C:/Users/11132302/Downloads/SSRN-id3140440.pdf
Crosby, M, Nachiappan, Pattanayak, P, Verma, S, Kalyanaraman, V. (2017). Blockchain technology:
beyond Bitcoin Retrieved from
https://j2-capital.com/wp-content/uploads/2017/11/AIR-2016-Blockchain.pdf
Swan, M. (2015). Blockchain: Blue print for a new economy. O’reilly.
Sovbetov, Y. (2018). Factors influencing cryptocurrency prices: Evidence from Bitcoin, Ethereum,
Dash, Litecoin, and Monero. Retrieved from
https://mpra.ub.uni-muenchen.de/85036/1/MPRA_paper_85036.pdf
Cameron, A, Thrin, K. (2017). Four factors driving the price of Bitcoin. Retrieved from http://theconversation.com/four-factors-driving-the-price-of-bitcoin-87244
Kristoufek, L. (2015). What are the main drivers of the Bitcoin Price? Evidence from Wavelet Coherence Analysis. PLoS ONE 10(4): e0123923. doi:10.1371/journal.pone.0123923. Retrieved from http://journals.plos.org/plosone/article/file?id=10.1371/journal.pone.0123923&type=printable
MacKinlay, C. (1997). Event Studies in Economics and Finance. Retrived from https://pdfs.semanticscholar.org/aac6/83a678a12a3dcd73389aac7289868847ea73.pdf
Jakub, Bartos. (2015). Does Bitcoin follow the hypothesis of efficient market?. International Journal of Economic Sciences, Vol. IV(2), pp. 10-23., 10.20472/ES.2015.4.2.002
http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.902.3664&rep=rep1&type=pdf
Kharpal, Arjun. (2017). Bitcoin gold price plunges: what you need to know about hard fork. Retrieved from https://www.cnbc.com/2017/10/25/bitcoin-gold-price-plunges-what-is-hard-fork.html
Bovaird, Charles. (2016). From Brexit to Bitfinex: What shaped Bitcoin Price in 2016. Retrived from https://www.coindesk.com/brexit-bitfinex-shaped-bitcoins-price-2016/
10 Appendix
APPENDIX 1: Unit Root Testing
Bitcoin
Ethereum
Litecoin
Ripple
Bitcoin Cash
Ethereum Classic
Nikkei 225
S&P 500
Gold
USD/EUR
BTCP
APPENDIX 2: Determination of the optimal lag length
BTC ETH LTC XRP XLM
FORK: ETH ETC
APPENDIX 3: GRANGER CAUSALITY
BTC ETH LTC XRP XLM 3 year 1 year 3 monthsFORK: BTC BCH
3 month
FORK: ETH ETC
1 year
FORK: BTC BCH ETC 1 year 3 month
CRYPTO MARKET (BTC ETH LTC XRP XLM) CONTROL VARIABLE (NIKKEI, GSPC, MSE, GLD, USD.EUR)
3 year
1 year
3 month
1 year
