The effect of attention on the Bitcoin price

(1)

1

The effect of attention on the Bitcoin price

Name: Morris Meijs Supervisor: dr. T. Yorulmazer Student number: 11041293 Track (within Economics and Business): Economics and Finance Date: 29-06-2018

(2)

2

Abstract

After the introduction, the Bitcoin started to gather attention. This was because of its innovative system (unregulated, anonymous and low transactions fees), but the attention spread due to the attraction as a speculative asset, the risks and the possible regulation. A regression has been performed to determine the effect of the attention to the Bitcoin on the Bitcoin price. The effect of the price driver on the Bitcoin price is compared for the period of extreme increasing prices in 2017 to the period before 2017. The study shows an significant effect of the attention to the Bitcoin on the Bitcoin price in both periods. Additionally, the effect of the attention to the Bitcoin is higher in the booming period in 2017 relative to the period before 2017.

(3)

3

1. Introduction

4 2. The Bitcoin System

6 2.1 Transactions

7 2.2 Mining

8 2.2 Characteristics

9 3. Risks and Regulation

9 3.1 Bitcoin risks

9 3.2 Demand for regulation

10 4. Price drivers

12 4.1 Attention

12 4.2 Trading volume

12 4.3 Transaction activity

13 4.4 Macroeconomic factors

13 5. Empirical analyses

14 5.1 Methodology

14 5.2 Hypothesis

16 5.3 Data analysis

17 6. Results

20 6.1 Regression results

20 6.2 Analyses of the most important Bitcoin price fluctuations 22

7. Discussion

23 8. Conclusion

25 References

27

(4)

4

1. Introduction

The Bitcoin was introduced in 2008 as a peer-to-peer electronic cash system. A system in which an online transaction could be arranged by two parties, without needing a trusted third party (financial institution) to process the payments (Nakamoto, 2008).

Digital currency is not a new concept, because on several websites and online games it already existed. The difference with the Bitcoin is that these currencies couldn’t be spend outside their own virtual world. The Bitcoin is the next step of these virtual currencies as an actual medium of exchange in the real economy (Rice, 2013).

The Bitcoin started with a paper about the decentralized cryptocurrency published by Satoshi Nakamoto. The identity of this group/person remains unsolved. In 2009 the Bitcoin network was initiated, and the issuance of Bitcoins started. The decentralized currency started to rise in popularity in the first years, but it was only used as a means of private exchange. This

changed in 2012 when the first companies started to accept the Bitcoin as payment method (http://historyofbitcoin.org).

The Bitcoin became in only a decade one of the most thriving innovations in the money market. It has grown into a worthy opponent to existing payment and monetary systems (Mai, Shan, Bai, Wang & Chiang, 2018). The number of transactions grew to the amount of 11.2 million in December 2017. Because of the high returns the Bitcoin became attractive in the market of speculation. The price appreciated from 35 dollars in March 2013 to an all-time high of 19290 Dollar in December of 2017. This is an appreciation of 55014%

(5)

5

Figure 1: Bitcoin price

The Bitcoin received a lot of attention in the recent year from both the media and investors. This research examines the effect of the attention to the Bitcoin on the price of the Bitcoin. It focusses on the difference of this effect between the period of the extreme boom in 2017 and the period before 2017. This period was extraordinary, because the Bitcoin price didn’t experience a price increase of this level before. The appeal of the Bitcoin has different reasons.

For some people the increase in attention might be a result of the unregulated structure, the lower transactions cost than other payment systems and the anonymity which the Bitcoin facilitates (Blau, 2017).

According to Cheah and Fry (2015) the Bitcoin attractiveness is more the result of being an interesting speculative object rather than money.

This is the first research that aims to compare the effect of attention on the price of the Bitcoin in the period of extreme thriving prices in 2017 with the period before 2017. By examining this effect, it will be determined if there is a difference in the effect of attention on an extraordinary period such as 2017.

To gain a better understanding of the effect of attention in these different periods the research addresses this research question.

What is the difference in the effect of attention to the Bitcoin on the price of the Bitcoin in the period before 2017 and in the period of extraordinary price increase in 2017?

0 5 0 0 0 1 0 0 0 0 1 5 0 0 0 2 0 0 0 0 B it c o in p ri c e 1/1/2013 1/1/2014 1/1/2015 1/1/2016 1/1/2017 1/1/2018 date

(6)

6

In the research an OLS regression is used to find the linear relationship between the attention to the Bitcoin and the price. The research finds evidence of a significant effect of the attention to the Bitcoin on the price in both periods. In the period of the boom the attention to the Bitcoin has a larger effect than during the period before 2017. The data is collected for the period of the 1st of march, 2013 till the 11th of December 2017.

The organization of this thesis will be as follows. In the next section the Bitcoin system will be explained. Subsequently, the risks and the possible regulation of the Bitcoin is going to be discussed. Next, there will be section about the price drivers of the Bitcoin. The empirical analysis consists of a methodology (research method), the hypothesises and a data analysis. Afterwards, the results will be analysed and discussed. And the final section is the conclusion, which will give an answer to the research question.

2. The Bitcoin System

The Bitcoin is known as a virtual currency, or cryptocurrency. Cryptocurrency uses a security system called cryptography. This makes sure that transactions are protected and ad it controls the supply of new currency units. The blockchain system consists of a public record of all previous transactions of the bitcoin divided into blocks. This system will guarantee that currency units can’t be replicated, that transfers of Bitcoin units (as data) will go smoothly and the system ensures verification of transactions of currency units between two persons. This solves the double-spending problem (Sovbetov, 2018).

(7)

7

2.1 Transactions

Figure 2: transactions (source: Nakamoto, 2008)

A wallet is used to store physical currency. For the Bitcoin there is also a program called a wallet. In the wallet there aren’t actual Bitcoins, but the wallet stores a Bitcoin address. The address contains the history of all transactions of a user. This address is a row of 34

components (numbers and letters) and is also called someone’s ‘public key’. Next to the public key there also exists a related private key of 64 components (numbers and letters). Although the two keys are linked, it is impossible to identify an owner’s private key from the related public key (Böhme, Christin, Edelman, and Moore, 2015).

It isn’t a problem that everyone is able to observe the public keys, because to complete a transaction it has to be confirmed with the corresponding private key. In order to do this a signature will be created by putting the private key and transaction specifics into someone’s personal bitcoin software in their computer. So, to validate a transaction it has to be confirmed with a digital signature, which is made of the private key. To verify transactions the other users only need the public key. Since transaction are permanent this will lower the demand of a third trusted party (Böhme et al., 2015).

A hash functions works as follows. It will transform any kind of data or text into a specified-length output (for blockchain a 64-character row) (Yang, Chen, Zhang, Yu & Zhang, 2017). The hash is extremely specific. If only one character in the data or text is changed the whole hash code will be different. In the blockchain this is convenient, because people will know if someone altered a transaction.

After the validation of a transaction, it will be included in the public ledger of all transactions and classified in a block. In every block a hash of the prior block is included. The hash

(8)

8

ensures that every block relates to the previous one. This creates a chain of blocks, for that reason the term blockchain (Sovbetov, 2018).

2.2 Mining

Providing virtual currency is conducted differently as providing physical currency. Physical currency depends on a third party (government). The government controls the process of printing money.

The mining process makes the Bitcoin self-sustaining. Miners are people who let the bitcoin distribution network use the processing power of their computers to generate data in new blocks. These blocks of data are a ledger of the entire history of transactions in Bitcoins. The blocks are coded with a hash. In order to mine new bitcoins your system has to generate a new hash code that will be accepted by the system. The miners actually guess the number, because the output (hash code) of the hash function is unpredictable. The guessed number and the block’s data will be implemented with the hash function to get the right combination. This is a complex mathematical task which requires a large amount of processing power. The miner whose system generates the new hash will announce this to the network. The miner will provide a proof-of-work that the new block is generated, which also includes a history of the transactions that have taken place since the last newly generated block and a reference of the previous block. The other miners will stop trying to determine the hash code of the specific block and will move on to the next one. The victorious miner will get Bitcoins or transaction fees in return. Ten minutes is the average processing time of a block (Böhme et al., 2015).

In the earlier period of the bitcoin there were less miners. They mined bitcoins at a high speed. All kinds of extra processors were used to accelerate the calculating power. The generating power grew, and the number of miners remained low. This resulted in faster generated hash codes, hence Bitcoin units.

This thriving grow couldn’t last, because the blockchain system hinders this acceleration. The generation of each new block gets more challenging. This diminishes the generation of new Bitcoins. So, in the future it gets more difficult to mine Bitcoins for the processing system (Böhme et al., 2015).

(9)

9

2.3 Characteristics

Characteristics of the Bitcoin are: - anonymity

- no centralized authority - no transaction fees.

The system provides anonymity, because an exchange in the peer-to-peer network is between two persons, who make transactions with each other only by exchanging hashes. This hash doesn’t provide the other person with any kind of personal information (Stegăroiu, 2018). Since the Bitcoin isn’t acknowledged as a currency by any nation, the Bitcoin isn’t subjected to regulations. Besides, there isn’t a trusted third party to process the payments, because of the public ledger, which makes this unnecessary (Stegăroiu, 2018).

No transactions fees are not completely the case. As mentioned earlier miners receive some sort of compensation. In general, the transactions fees of the Bitcoin are relatively low compared to other payment systems(Böhme et al., 2015).

3. Risks and regulation

3.1 Bitcoin risks

According to previous research there are several risks involved with the Bitcoin. These can be divided into long-term risks and direct risks.

Long-term risks

• The commonly use of the Bitcoin can result in a decreasing demand of traditional currencies issued by the central banks. The central banks could not only lose control over the supply of money, but also over monitoring the market. They would be less able to prevent a distortion in the market and their function to be a provider as lender of last resort would be disrupted (Sotiropoulou and Guégan, 2017).

(10)

10

• Additionally, the commonly adoption of the Bitcoin as a currency could be threatening for the financial system. For example, if a component of the system of the Bitcoin fails (Sotiropoulou and Guégan, 2017).

Direct risks

• The Bitcoin system can be used to hide the location of the destination the purchase of the payment is going to or is coming from. The user’s anonymity and the peer-to-peer network is an ideal system to obfuscate transaction.

• There is also a rise in frauds due to the higher privacy and the lack of regulatory oversight, which the use of the Bitcoin facilitates.

• Further will there be a security issue regarding the unvalidated transactions. These transactions will not be executed. Transaction are irreversible and errors in

transactions will be permanent as well. The decentralized aspect of the Bitcoin system puts the risk of an error in a transaction on the users (Sotiropoulou and Guégan, 2017).

• Moreover, Bitcoins can be used to evade taxes. Using traditional currency to buy Bitcoins and transferring them abroad anonymously to circumvent capital controls. The evasions already happened. There is significant evidence of the evasion of capital controls. From the Chinese Renminbi to the US Dollar and via the Bitcoin people conducted capital flight (Ju, Lu and Tu, 2016).

3.2 Demand for regulation

To avoid the risks described above there is a demand for the regulation of the Bitcoin. According to Sotiropoulou and Guégan (2017) there are three types of regulation of the Bitcoin. The regulation can apply to:

1. The Bitcoin system.

The regulation of the whole system means that all operations have to follow the

pre-determined rules. These rules apply to everyone involved. Normally the regulations are being overseen by a centralized authority. This is a problem with the Bitcoin, because it doesn’t have such an institution. In the Bitcoin system, the replacement of this institution are the minors and developers. These people are anonymous, and their identity can’t be discovered.

(11)

11

This is the reason why no regulated action can follow regarding these members. A possible regulatory action on the system would be to create an information program to collect and to inform all the dysfunctions of the system.

2. The uses of the Bitcoin.

The Bitcoin is being used for illegal activities. The anonymous part of the Bitcoin makes it attractive for illegal purposes such as money laundering and terrorist financing. Regulations can also be applied to prevent these illegal uses.

With traditional currencies the banks are the institutions that oversee the payment systems. They function as an important part of ensuring that regulations regarding money laundering and terrorist financing are preserved.

An institution like a bank is absent in the Bitcoin system. The identification of channels that could refer to money laundering and terrorist financing in Bitcoin payment system has to be overseen by another organisation. The Financial Action Task Force (FATF) are checking the entry into and exit from the Bitcoin domain. The Bitcoin exchanges are the most convenient target in this process.

3. Members of the Bitcoin system.

The users of the Bitcoin experience risk. The risk of an error in the transactions. Transactions are permanent and errors in transaction are irreversible as well. The decentralized aspect of the Bitcoin system makes it impossible to reverse a transaction with an error. Bitcoin users should have a full understanding of the Bitcoin and the including risks. Regulations will not be able to protect the users, but it can provide information about the consequences of these risks. It can make the users fully aware.

Besides looking at the users, the service providers can be more affected by regulations. The regulations will make the licencing regimes for service providers of the Bitcoin more comprehensive. The regimes focused to decrease the concerns about the user’s protection.

(12)

12

4. Price drivers

The Bitcoin price is influenced by several factors. These drivers are already examined in previous literature. An overview of significant papers will be given to illustrate the effect of these different drivers on the Bitcoin price.

There have been several researches on the effect of the attention to the Bitcoin on the price of the Bitcoin. This research aims to examine if there is a difference in this effect in the period of extreme price increase in 2017.

4.1 Attention

Ciaian, Rajcaniova and Kancs (2015) examined the Bitcoin price formation. He used the number of daily Wikipedia views as a proxy for the attractiveness of the Bitcoin. His results show that the effect of the attractiveness of the Bitcoin on the price is stronger in the period when the Bitcoin was little known compared to the later years when the Bitcoin became a bigger player in the financial market. In the long-run the results reflected that there was no effect on the price.

Kristoufek (2013) studies the relationship between the search queries Google Trends and Wikipedia views and the Bitcoin price. The study suggests a significant relationship, but the impact of increasing interest displays an irregularity between the price being positioned above or below the trend. If the price is situated above the trend, increasing interest results in rising prices and if the price is situated below the trend, increasing interest results in diminishing prices.

Shan et al. (2018) researches the connection between social media and the value of the Bitcoin. The study shows that the social media is a substantial predictor of the Bitcoin value. Positive forum posts have an effect of increasing Bitcoin values in the future. Moreover, Forum posts have a stronger effect on the Bitcoin values in the future compared to tweets.

4.2 Trading volume

Bacilar, Bouri, Gupta and Roubaud (2017) examined the relationship between the traded volume and the Bitcoin return and the volatility. The analyses reveal that the traded volume can’t predict the returns. Furthermore, they concluded that there was no causal relationship between trading volume and volatility.

(13)

13

Further, Sovbetov (2018) finds a strong significant effect of the traded volume on the Bitcoin price in the long-run. The effect of the traded volume on the Bitcoin is also significant in the short-run, but the impact is smaller. This indicates that the reaction to fluctuations in the trading volume is more severe in the long-run compared to the short-run.

Glaser, Zimmermann, Haferkorn, Weber and Siering (2014) also used the trading volume for their research, but for a different purpose. They examined if the user’s interest to the Bitcoin was because of its attraction as an investment vehicle or as a currency. The results indicated that generally uninformed people picture the Bitcoin as speculative vehicle rather than a medium of exchange.

4.3 Transaction activity

Koutmos (2018) examined the possible relationship between the transaction activity and Bitcoin returns. He used the Bitcoin transactions and the unique addresses as a proxy of the transaction activity. The results showed that there was a strong relationship between the transaction activity and the price movements, but the returns were more explanatory for the transaction activity than vice versa.

Additionally, Ciaian et al. (2015) also found a strong effect of the size of the Bitcoin economy (transaction volume and number of unique addresses) on the price.

4.4 Macroeconomic factors

According to Ciaian et al. (2015) the effect of the global macro-financial development on the Bitcoin price is only significant in the short-run. The effect is displayed by the Dow Jones index, oil price and exchange rate. The results show an insignificance of this effect in the long-run.

Contrary, Sovbetov (2018) finds a small effect of the S&P 500 on the Bitcoin price in the long-run and a negative effect on the price in the short-run.

Furthermore, Bouoiyour and Selmi (2017) used the exchange trade ratio, velocity of the Bitcoin and the gold price as a proxy for the macroeconomic and financial determents. Their study reveals that the gold price and the velocity of the Bitcoin are important determents of the price when market is experiencing a drop.

(14)

14

5. Empirical analysis

5.1 Methodology

To determine whether the attention to the Bitcoin has a different influence on the price of the Bitcoin during a period of extreme increasing prices and during a period of less volatility a data set is collected for a term from 1st of March 2013 till the 11th of December 2017. This term is divided into two periods. A regression will be performed on the data of both periods to examine the difference of the effect of the attention to the Bitcoin on the price of the Bitcoin during a period of extreme price increase and during a period of less rising price.

To examine the relationship of the attention to the Bitcoin and the price, the following regression will be performed.

In this research an ordinary least square regression (OLS) will be used. This will estimate the linear relationship of the independent variables with the depended variable.

Assumptions for error-term OLS: - Normality

- Linearity

- Model specification

- Homoskedasticity (var (ɛt |X) = var(ɛt) = σ2 , t = 1, 2, . . . , n) - Errors in variables

Regression:

PBitcoin = β0 + β1*(GT)t + β2*(TradV)t + β3*(TransV)t + β4*SP500t + ɛt

PBitcoin is the price of the Bitcoin in US dollars. The average price of the largest exchanges in

the Bitcoin market are used to determine the price

.

For the regression the logarithm of the price is taken, and weekly data is used. The Bitcoin price is retrieved from data.Bitcoinity.org.

GTt is the Google Trends data of the Bitcoin. It reflects the amount of Bitcoin searches in Google-search engine.

(15)

15

(TradeV)t is the amount of the weekly traded volume of the Bitcoin. SP500t is the value of the S&P 500.

In the regression the Google Trends data captures the attention to the Bitcoin. It reflects the number of times the word ‘Bitcoin’ is searched for in Google in a certain week. To measure the interest in the Bitcoin this can be an excellent proxy with a good explanatory power. The Google Trends variable can be positively or negatively correlated to the price, because increasing interest will push the Bitcoin price higher if the price was situated above the trend, but further down if the Bitcoin price was situated below the trend (Kristoufek, 2013).

There are two kinds of users of the Bitcoin, which have different appeals to the Bitcoin. The attention of the users who picture the Bitcoin as a speculative asset and the attention of the users who see the Bitcoin as a currency (Glaseret et al, 2014). The Google Trends variable will capture the attention of both users.

The variable is standardized (ztrend), because the data in the research are in different scales. Standardizing will transform the data to comparable scales. Standardized output has two main advantages. The results of a standardized output can be used by researchers for meta-analyses and it will be possible to illustrate the relative effect on the dependent variable (Hunter & Hamilton, 2002). The Google Trends data is collected from Trends.Google.nl.

𝑧𝑡𝑟𝑒𝑛𝑑 = (𝐺𝑜𝑜𝑔𝑙𝑒𝑇𝑟𝑒𝑛𝑑)𝑡 − 𝐺𝑜𝑜𝑔𝑙𝑒𝑇𝑟𝑒𝑛𝑑̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ 𝜎(𝐺𝑜𝑜𝑔𝑙𝑒𝑇𝑟𝑒𝑛𝑑)

In this research three control variables are used. They are in the regression to control the effects of the Traded volume, the Transaction volume and the US market on the Bitcoin price.

The (tradev)t in this research represents the weekly traded volume of the Bitcoin (this

excludes transaction activities). The traded volume of the largest exchanges was added up the get the total amount. According to Balcilar et al (2017) volume influences the Bitcoin price, but this isn’t the case in bearish and bullish markets. This variable is also standardized. The weekly traded volume is retrieved from data.Bitcoinity.org.

(16)

16 𝑧𝑡𝑟𝑎𝑑𝑣 = (𝑇𝑟𝑎𝑑𝑣)𝑡 − 𝑇𝑟𝑎𝑑𝑣̅̅̅̅̅̅̅̅

𝜎(𝑇𝑟𝑎𝑑𝑣)

Transaction volume is the weekly number of Bitcoins that is used to purchase goods and services (this excludes the trading activities). According to Koutmos (2018) there is a strong relationship between the transaction activity and the Bitcoin price. The variable is also standardized. The weekly transaction volume is retrieved from data.Bitcoinity.org.

𝑧𝑡𝑟𝑎𝑛𝑠𝑣 = (𝑇𝑟𝑎𝑛𝑠𝑣)𝑡 − 𝑇𝑟𝑎𝑛𝑠𝑣̅̅̅̅̅̅̅̅̅̅ 𝜎(𝑇𝑟𝑎𝑛𝑠𝑣)

S&P 500 represents the influence of the developments in the American stock market on the Bitcoin price. The S&P 500 is a value-weighted index, which represents the market value of the 500 largest companies in the US (Kappou, Brooks & Ward, 2010). For the regression the logarithm of the S&P 500 value is taken. The value of the S&P 500 is retrieved from

DataStream (2018).

ɛt is the error-term of the regression. This will illustrate the effect of other factors that are not included in the model.

5.2 Hypothesis

This paper concerns the effect of the attention to the Bitcoin on the price of the Bitcoin. The effect of this factor is measured by the Google Trends variable. The impact of the attention on the two different periods results in three hypotheses.

Hypothesis 1: The attention to the Bitcoin will influence the price of the Bitcoin in the period

before 2017 (Kristoufek, 2013). Period 1: H0: β1 = 0

(17)

17

Hypothesis 2: The attention to the Bitcoin will influence the price of the Bitcoin during the

booming period (Kristoufek, 2013). Period 2: H0: β1 = 0

H1: β1 ≠ 0

Hypothesis 3: The relationship between the attention to the Bitcoin and the price of the

Bitcoin is larger in the booming period compared to the period before 2017. H0: β1 (period 1) = β1 (period 2)

H1: β1 (period 1) < β1 (period 2)

5.3 Data analysis

The first period is before 2017 (1/3/2013-31/12/2016) and the second period is during the extreme boom of 2017 (1/1/2017-11/12/2017). The week of the 11th of December 2018 the Bitcoin price was at the all-time high. For the research weekly data is used, so in total there are 250 data points (first period: 200, second period: 50).

Before running the regression, the Bitcoin price and the variable for the attention to the Bitcoin (Google Trends) could be compared for both periods.

Figure 3: The Bitcoin price vs. the Google Trends data (period before 2017)

0 5 1 0 1 5 G o o g le t re n d s 0 2 0 0 4 0 0 6 0 0 8 0 0 1 0 0 0 B it c o in p ri c e 1/1/2013 1/1/2014 1/1/2015 1/1/2016 1/1/2017 Week

(18)

18

Figure 4: The Bitcoin price vs. the Google Trends data (period during 2017)

The Google Trends data represents the value of the attention of the Bitcoin in the given week relative to the highest point in the diagram for that period. A value of 100 is the peak

popularity for that term. A value of 50 means that the term is half as popular. A score of 0 means that there is insufficient data available for this term.

In the period before 2017 the Google Trends data moves less smoothly than in the period of 2017. This is because the attention to the Bitcoin in this period was in multiple terms relatively constant. In the graph it is observable that in these periods of relatively constant attention the Bitcoin price is also relatively stable.

In these illustrations the Google Trends data and the Bitcoin price move together during the booming period and before 2017 they don’t always move in line witch each other. This will push the prediction regarding hypothesis 3 to a rejection of the null-hypothesis. A regression will be run to determine the relationship between the two variables.

0 2 0 4 0 6 0 8 0 1 0 0 G o o g le t re n d s 0 5 0 0 0 1 0 0 0 0 1 5 0 0 0 2 0 0 0 0 B it c o in p ri c e 1/1/2017 4/1/2017 7/1/2017 10/1/2017 1/1/2018 Week

(19)

19

Summaries of the variables:

(1) (2) (3) (4) (5)

VARIABLES N Mean sd min max

Marketprice 200 5.823 0.662 3.827 6.862 SP500 200 6.766 0.0764 6.566 6.919 ztrend 200 3.19e-09 1.000 -0.948 5.088 ztradv 200 -9.22e-10 1.000 -0.648 4.395 ztransv 200 1.62e-09 1.000 -1.207 2.324 Table 1: period of 1/3/2013-31/12/2016 (1) (2) (3) (4) (5)

VARIABLES N mean sd min max

Marketprice 50 7.840 0.789 6.734 9.759 SP500 50 6.949 0.0256 6.907 7.019 ztrend 50 -2.53e-09 1.000 -0.677 4.124 ztradv 50 8.94e-10 1.000 -0.262 6.373 ztransv 50 -5.03e-09 1.000 -1.876 2.961 Table 2: period of 1/1/2017-11/12/2017 (1) (2) (3) (4) (5)

VARIABLES N mean sd min max

Marketprice 250 6.226 1.061 3.827 9.759 SP500 250 6.802 0.101 6.566 7.019 ztrend 250 -5.04e-09 1.000 -0.466 8.691 ztradv 250 6.26e-10 1.000 -0.575 5.601 ztransv 250 2.14e-09 1.000 -1.276 2.683 Table 3: period of 1/3/2013-11/12/2017

(20)

20

6. Results

Regression results

The regression analyses consist of three regressions. The first is of the period before 2017, the second is of the period of the boom in 2017 and the third is of the two periods combined.

Bitcoin price Bitcoin price Bitcoin price Robust VARIABLES 1/3/2013-31/12/2016 1/1/2017-11/12/2017 (Boom) 1/3/2013-11/12/2017 (both periods combined) 1/3/2013-11/12/2017

Ztrend (google trends) 0.331*** 0.452*** 0.373*** 0.373***

(0.0276) (0.0942) (0.0368) (0.0860) Ztradv (trad. volume) 0.0530 -0.0205 -0.0244 -0.0244

(0.0447) (0.0514) (0.0368) (0.0217) Ztransv (tran.volume) 0.0827* -0.204*** 0.161*** 0.161*** (0.0497) (0.0638) (0.0585) (0.0521) S&P500 5.274*** 15.73*** 5.792*** 5.792*** (0.432) (3.213) (0.503) (0.479) Constant -29.86*** -101.5*** -33.18*** -33.18*** (2.926) (22.33) (3.421) (3.256) Observations 200 50 250 250 R-squared 0.672 0.835 0.802 0.802

Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 Table 4: regression analyses

Table 5: correlation table (1/3/2013-31/12/2016) Table 6: correlation table (1/1/2017-11/12/2017)

The coefficients of the control variables transaction volume and of the S&P 500 are

significant in both periods. The coefficient of the variable for the traded volume of the Bitcoin is in both periods insignificant. This is partly comparable to the research of Balcilar et al (2017), because in their research the traded volume was also insignificant in bullish and bearish markets, which would be correct for the booming period. A possible reason for the

SP500 0.8660 0.7739 -0.2992 0.3729 1.0000 ztransv 0.2949 0.6353 0.0523 1.0000 ztradv -0.2484 -0.0975 1.0000 ztrend 0.8071 1.0000 Marketprice 1.0000 Market~e ztrend ztradv ztransv SP500

SP500 0.6309 -0.1563 0.3770 0.5548 1.0000 ztransv 0.4781 -0.0967 0.7895 1.0000 ztradv 0.4012 -0.0150 1.0000 ztrend 0.3914 1.0000 Marketprice 1.0000 Market~e ztrend ztradv ztransv SP500

(21)

21

insignificance in first period can be that the variable has a strong correlation with the variable of the transaction volume (0.7895).

Furthermore, the coefficient of the transaction volume in the first period is positive and in the period of the boom negative. The effect during a booming period of the transaction volume on the Bitcoin price is contrary to the effect on the period before 2017.

The coefficient of the attention to the Bitcoin (ztrend) is in both periods significant. In the booming period is the coefficient higher than in the period before 2017 (0.452>0.331). This signifies that the relationship between the attention and the Bitcoin price is larger in the period of extreme price increase. This is in line with earlier research. According to Kristoufek (2013) the attention to the Bitcoin would push the price to a higher level, if the price was situated above the trend.

In all three regressions is the R-squared is relatively high. The variance of the Bitcoin price is well explained by the independent variables. In the regression of the second period is the change in the R-squared noticeable (0.672 vs. 0.835). This would mean that in the booming period the model explains the Bitcoin price to a higher extend.

Overall will this lead to a rejection of the null-hypothesis of hypothesis 1, because the

attention to the Bitcoin has a significant effect on the price in the first period. Further will the null-hypothesis of the second hypothesis also be rejected, because the significant effect of the attention variable on the bitcoin price in the second period and the null-hypothesis of the third hypothesis will also be rejected due to the larger effect of the attention variable during the booming period.

(22)

22

6.2 Analyses of the most significant Bitcoin price fluctuations

Figure 5: Bitcoin price (source: Blockchain.info)

In 2010 the attention to the Bitcoin increased, because of an article of a new published source code. This caused an increase in demand at the exchanges, but through the limited supply the price increased. The price started below 14 cents at the beginning of 2010, the price rose and eventually settled at 29 cents (Wallace, 2011).

Due to an article in April of 2011 in Forbes magazine about cryptocurrency the price

increased massively to 8.89 dollars. The increase in the price didn’t stop, because an article in June 2011 about the use of Bitcoins in illegal markets resulted in a rise of the price to 27 dollars in just a week (Wallace, 2011).

In 2013, Cyprus agreed to a bailout, because of the banking crisis. Instead of using traditional currency (it was argued that this was less secure), Cyprus chose to use the cryptocurrency Bitcoin. This caused a massive increase of attention to the Bitcoin (Luther and Salter, 2017). The world started to adapt to the virtual currency and Bitcoin payments were accepted by organisations worldwide (Wingfield, 2013).

In 2013, the popularity of the Bitcoin experienced an enormous rise. Trade was conducted freely and the exchanges in China experience an accelerating growth. In the fall of 2013 the price increased above 1000 dollars. However, in December of 2013 China declared that the Bitcoin wasn’t a currency and not long after that the Chinese government prohibited the

(23)

23

institutions that worked as an intermediary between the businesses and the exchanges. This resulted an extreme price drop and the price decreased below 500 dollars (“A dream dispelled,” 2014).

In early 2014, The biggest exchange at time, the MT. Gox, experienced attacks from hackers. This resulted in a crash of the exchange and a fall in the Bitcoin price accompanied with a decline in the trust in the currency (Gandal, Hamrick, Moore and Oberman, 2018).

After 2014 the trust in the Bitcoin slowly came back and in 2015 the Bitcoin price to rise again. In 2016, The attention to Bitcoin grew and pushed the price up. In December of 2017, the Bitcoin price reached an all-time high. The Bitcoin bubble was set to burst and would collapse due to regulations of economic authorities (Li, Tao, Lobonţ and Su, 2018) .

In all the discussed price fluctuations the attention played an important role. This signifies the effect of the attention on the Bitcoin price in all periods.

7. Discussion

In future research also other extraordinary periods described in the analyses of extraordinary price fluctuations can be examined. Research about particular price drivers of the Bitcoin price or a comparison between different periods will help to explain the behaviour of the Bitcoin price more.

Additionally, several tests are conducted to examine the model used in this research. This is necessary for a critical approach to the used regression. The OLS assumptions of the model are tested. This is tested for both periods separately and for the two periods combined. The Breusch-Pagan/Cook-Weisberg test is conducted to check for heteroskedasticity (table A1, A2, A3), the skewness/kurtosis test for normality (table B1, B2, B3), the Ramsey RESET test to check for omitted variables (table C1, C2, C3) and the Breusch-Godfrey test to check for autocorrelation (table D1, D2, D3). This tables can be found in the Appendix.

(24)

24 Period 1:

• The null-hypothesis for homoskedasticity will not be rejected (table A1). In a

homoscedastic regression the variance of the errors is constant (Osbourne & Waters, 2002).

• The null-hypothesis for normality will not be rejected (table B1). In a non-normal regression, the distribution is highly skewed and kurtotic (Osbourne & Waters, 2002). • The null-hypothesis for no omitted variables will be rejected (table C1). This means

that there are omitted variables that influence the independent variables in the regression. A way to deal with this is to take the uncorrelated part of these omitted variables into the regression. This will improve the explanatory power of this regression (Baccarini, 2010).

• The null hypothesis for multicollinearity will not be rejected (table D1). The VIF-values are below 10, which indicates there is no multicollinearity.

Period 2:

• The null-hypothesis for homoskedasticity will not be rejected (table A2). • The null-hypothesis for normality will not be rejected (table B2).

• The null-hypothesis for no omitted variables will be rejected (table C2).

• The null hypothesis for multicollinearity will not be rejected (table D2). The VIF-values are below 10, which indicates that there is no multicollinearity.

Both periods combined:

• The null-hypothesis for homoskedasticity will be rejected (table A3).

Heteroscedasticity can lead to a distortion in the results. This will affect the chance of a Type I error (Osbourne & Waters, 2002). A way to solve this is to change the

variables. In the regression analyses the adjustment can be found in the robust column. In the regression for the combined periods heteroskedasticity proof standard errors are used to solve the problem.

• The null-hypothesis for normality will not be rejected (table B3). • The null-hypothesis for no omitted variables will be rejected (table C3).

• The null hypothesis for multicollinearity will not be rejected (table D3). The VIF-values are below 10, which indicates there is no multicollinearity.

(25)

25

8. Conclusion

After the introduction of the Bitcoin, it became one of the most thriving innovations in the financial market of the last decade. The unregulated structure, the lower transactions cost than other payment systems and the anonymity attracted a lot of attention from both the media and the investors. Further became the Bitcoin a popular speculative vehicle. Next to the attention to the impressive blockchain-system, it gathered also attention because of its specific risks and the possible demand for regulation.

In 2017, the price of the Bitcoin experienced an extreme price increase. The attention to the Bitcoin increased as well.

An OLS regression has been conducted to examine the effect of the attention to the Bitcoin on the Bitcoin price. This research compared two periods to find the possible difference of the effect of the attention in a period of extreme price increase compared to the period before with less price increase.

Previous research reveals that the Bitcoin price is effected by the attention to the Bitcoin. However, it depended on whether the price was situated above or below the trend. Previous researches didn’t examined the effect in the period of extreme increasing prices in 2017. This thesis has focused on the difference of a price increase of the Bitcoin in a period of extreme price increases of the Bitcoin compared to a period with much less price increases.

The regression consisted of three control variables (trading volume, transaction volume and the S&P 500) and was used to test three hypothesises. It was used to test if the attention to the Bitcoin influenced the Bitcoin price in the period before 2017, in the period in 2017 and if the effect was larger in the booming period compared to the period before 2017.

This research shows that the attention to the Bitcoin has a significant strong effect on the Bitcoin price in both periods. Additionally, the effect was larger in the period of extreme price increase relative to the period before 2017.

Furthermore, this research analysed the most important earlier price fluctuations of the Bitcoin. Before 2017 the most important fluctuations in the price of the Bitcoin were all

(26)

26

accompanied with an increase in the attention to the Bitcoin, which emphasises the effect of the attention to the Bitcoin on the Bitcoin price in all periods.

Testing the model resulted in omitted variable bias in both periods. Because of these omitted variables, the explanatory factor is not optimal. But this can be solved in future research by including the part of the omitted variables that is not correlated with the independent variables into the regression (Baccarini, 2010).

(27)

27

References

A dream dispelled; Bitcoin in China. (2014). The Economist, 411(8882), N/a.

Balcilar, Bouri, Gupta, & Roubaud. (2017). Can volume predict Bitcoin returns and volatility? A quantiles-based approach. Economic Modelling, 64, 74-81.

Beccarini, A. (2010). Eliminating the omitted variable bias by a regime-switching approach. Journal Of Applied Statistics, 37(1), 57-75.

Data.bitcoinity.org (beta version). (June 2018). Retrieved from http://data.bitcoinity.org/

Blau, B. (2017). Price dynamics and speculative trading in Bitcoin. Research in

International Business and Finance, 41, 493-499.

Böhme, R., Christin, N., Edelman, B., & Moore, T. (2015). Bitcoin: Economics, Technology, and Governance †. Journal of Economic Perspectives, 29(2), 213-238.

Bouoiyour, J., & Selmi, R. (2017). The Bitcoin price formation: Beyond the fundamental sources.

Carina-Elena Stegăroiu. (2018). THE ADVANTAGES AND DISADVANTAGES OF BITCOIN PAYMENTS IN THE NEW ECONOMY. Analele Universităţii Constantin

Brâncuşi Din Târgu Jiu : Seria Economie, 1(1), 67-72.

Cheah, & Fry. (2015). Speculative bubbles in Bitcoin markets? An empirical investigation into the fundamental value of Bitcoin. Economics Letters, 130, 32-36.

Ciaian, P., Rajcaniova, M., & Kancs, D. (2015). The economics of Bitcoin price formation. Applied Economics, 1-17.

Datastream. (2018) Thomson Reuters Datastream. [Online]. Available at: Subscription Service (Accessed: June 2018)

(28)

28

Data.bitcoinity.org (beta version). (June 2018). Retrieved from http://data.bitcoinity.org/

Gandal, Hamrick, Moore, & Oberman. (2018). Price manipulation in the Bitcoin ecosystem. Journal of Monetary Economics, 95, 86-96.

Glaser, Zimmermann, Haferkorn, Weber, & Siering. (2014). Bitcoin - Asset or Currency? Revealing Users' Hidden Intentions. ECIS 2014 (Tel Aviv).

Hunter, J., & Hamilton, M. (2002). The Advantages of Using Standardized Scores in Causal Analysis. Human Communication Research, 28(4), 552-561

Ju, L., Lu, T., & Tu, Z. (2016). Capital Flight and Bitcoin Regulation. International

Review of Finance, 16(3), 445-455

Kappou, Brooks, & Ward. (2010). The S&P500 index effect reconsidered: Evidence from overnight and intraday stock price performance and volume. Journal of Banking and

Finance, 34(1), 116-126.

Koutmos, D. (2018). Bitcoin returns and transaction activity. Economics Letters,167, 81-85.

Kristoufek, L. (2013). Bitcoin meets Google Trends and Wikipedia: Quantifying the relationship between phenomena of the Internet era. Scientific Reports, 3, Scientific Reports, 2013 Dec 4, Vol.3.

Li, Z. Z., Tao, R. R., Lobonţ, O. W., & Su, C. (2018). Does Bitcoin bubble burst? Quality and Quantity, 1-15.

Luther, & Salter. (2017). Bitcoin and the bailout. Quarterly Review of Economics and

(29)

29

Mai, F., Shan, Z., Bai, Q., Wang, X., & Chiang, R. (2018). How Does Social Media Impact Bitcoin Value? A Test of the Silent Majority Hypothesis. Journal of Management

Information Systems, 35(1), 19-52.

Nakamoto, S. (2008). Bitcoin: A peer-to-peer electronic cash system. Consulted, 1–9. Retrieved from https://Bitcoin.org/Bitcoin.pdf

Osbourne, Jason W., & Waters, Elaine. (2002). Four Assumptions of Multiple Regression That Researchers Should Always Test. Practical Assessment, Research &

Evaluation, 8(2), 1-5.

Rice, D. (2013). The Past and Future of Bitcoins in Worldwide Commerce. Business

Law Today, 1-4.

Sotiropoulou, A., & Guégan, D. (2017). Bitcoin and the challenges for financial regulation. Capital Markets Law Journal, 12(4), 466-479.

Urquhart, A. (2016). The inefficiency of Bitcoin. Economics Letters, 148, 80-82.

Wallace, Benjamin. (2011). The Rise And Fall Of Bitcoin. Wired, 19(12), N/a.

Wingfield, N. (2013, October 31). Bitcoin Pursues the Mainstream. New York Times

(1923-Current File), p. B1.

Yang, Yijun, Chen, Fei, Zhang, Xiaomei, Yu, Jianping, & Zhang, Peng. (2017). Research on the Hash Function Structures and its Application. Wireless Personal

Communications, 94(4), 2969-2985.

Yhlas Sovbetov. (2018). Factors Influencing Cryptocurrency Prices: Evidence from Bitcoin, Ethereum, Dash, Litcoin, and Monero. Journal of Economics and Financial

(30)

30

Appendix

1/3/2013-31/12/2016:

Table A1: test for heteroskedasticity

Table B1: test for normality

Table C1: test to check for omitted variables

Table D1: test for multicollinearity

1/1/2017-11/12/2017:

Prob > chi2 = 0.6794 chi2(1) = 0.17

Variables: fitted values of Marketprice Ho: Constant variance

Breusch-Pagan / Cook-Weisberg test for heteroskedasticity

r 200 0.2818 0.1842 2.95 0.2283 Variable Obs Pr(Skewness) Pr(Kurtosis) adj chi2(2) Prob>chi2 joint Skewness/Kurtosis tests for Normality

Prob > F = 0.0000 F(3, 192) = 14.66 Ho: model has no omitted variables

Ramsey RESET test using powers of the fitted values of Marketprice

Mean VIF 2.46 ztradv 1.16 0.861583 ztransv 1.79 0.558611 SP500 2.98 0.335817 ztrend 3.90 0.256267 Variable VIF 1/VIF

Prob > chi2 = 0.2185 chi2(1) = 1.51

(31)

31 Table B2: test for normality

Table D2: test for multicollinearity

1/3/2013-11/12/2017:

Table B3: test for normality

Ramsey RESET test using powers of the fitted values of Marketprice

Mean VIF 2.15 ztrend 1.03 0.968056 SP500 1.48 0.673516 ztradv 2.71 0.368543 ztransv 3.36 0.298009 Variable VIF 1/VIF

Prob > chi2 = 0.0035 chi2(1) = 8.51

Breusch-Pagan / Cook-Weisberg test for heteroskedasticity

(32)

32 Table D3: test for multicollinearity

Mean VIF 2.39 ztrend 1.49 0.672715 ztradv 1.49 0.670965 SP500 2.83 0.353152 ztransv 3.76 0.266018 Variable VIF 1/VIF

The effect of attention on the Bitcoin price